Human-centred artificial intelligence for mobile health sensing: challenges and opportunities
Open accessAbstract
Advances in wearable sensing and mobile computing have enabled the collection of health and well-being data outside of traditional laboratory and hospital settings, paving the way for a new era of mobile health. Meanwhile, artificial intelligence (AI) has made significant strides in various domains, demonstrating its potential to revolutionize healthcare. Devices can now diagnose diseases, predict heart irregularities and unlock the full potential of human cognition. However, the application of machine learning (ML) to mobile health sensing poses unique challenges due to noisy sensor measurements, high-dimensional data, sparse and irregular time series, heterogeneity in data, privacy concerns and resource constraints. Despite the recognition of the value of mobile sensing, leveraging these datasets has lagged behind other areas of ML. Furthermore, obtaining quality annotations and ground truth for such data is often expensive or impractical. While recent large-scale longitudinal studies have shown promise in leveraging wearable sensor data for health monitoring and prediction, they also introduce new challenges for data modelling. This paper explores the challenges and opportunities of human-centred AI for mobile health, focusing on key sensing modalities such as audio, location and activity tracking. We discuss the limitations of current approaches and propose potential solutions.
AI snapshot
Written from the paper’s full text.
TL;DR
This review examines human-centred AI for mobile health sensing across audio, mobility, and motion modalities, outlining applications and challenges. It argues that noisy, high-dimensional, sparse, heterogeneous, privacy-sensitive, and resource-constrained mobile data require interpretable, adaptive, fair, and rigorously validated AI.
Key findings
- Audio, location, and motion sensing can support mental and physical health applications; examples include COVID-19 detection from crowdsourced audio with ROC-AUC 0.71 in 2,478 participants, lung-health classification F1 scores of 0.912 (healthy versus unhealthy), 0.870 (obstructive versus non-obstructive), and 0.813 (COPD versus asthma), and wearable-based VO2max prediction with r = 0.82 (95% CI: 0.80–0.83) using Fenland Study N = 11,059, longitudinal cohort N = 2,675, and UK Biobank validation N = 181.
- Mobility data can model disease spread and aid diagnosis; travel restrictions in Wuhan reduced internationally exported cases by 70–80% until mid-February 2020, decreased mobility correlated with decreased COVID-19 case growth rates (Pearson correlation coefficients above 0.7), and driving features identified preclinical Alzheimer’s disease with F1 = 0.82 using data from 75 without and 64 with preclinical AD.
- Motion and wearable data enable fitness and mental-health monitoring; self-supervised embeddings achieved AUC 0.68 for cardio-fitness prediction and AUC 0.80 for physical activity energy expenditure, while large-scale studies link activity patterns to health outcomes.
- Cross-cutting challenges include noisy, high-dimensional, sparse, irregular, and heterogeneous data; costly labels; interpretability; personalization; fairness; resource constraints; limited validation; and privacy/security, with one study of 20,000 Google Play apps finding 45% relied on unencrypted communication and 23% sent personal data on unsecured traffic.
Methods
- Narrative review/position paper, not a systematic review or primary study; it focuses on three modalities—audio, location/mobility, and motion/activity tracking—with brief coverage of biosignals and RF sensing.
- Synthesizes prior literature and example datasets (e.g., table 1: Coswara, COVID-19 Sounds, RECOLA, E-DAIC, ADReSS, GLOBEM, MESA, FitRec) and discusses data pre-processing, feature extraction, modelling, and applications.
- No new data collection, experimental design, or statistical analysis is reported; conclusions are thematic.
Limitations
- Stated: current AI limitations include lack of interpretability, need for adaptation/personalization under distribution shift, fairness/robustness concerns, and difficulty determining causation.
- Stated: mobile/wearable constraints include resource limitations for on-device training/deployment, noisy or missing sensor data, irregular sampling, and label-quality issues; validation is limited by small datasets and lack of deployment standardization; privacy/security risks remain.
- Evident: as a narrative review, it does not report a systematic search strategy, inclusion criteria, or quantitative synthesis, so coverage and examples are selective.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Claims the paper makes
0 of 202 cited sources checked · 0 supported · 0 flagged
- Premisecitation
Deploying audio-based applications in real-world settings, especially for audio-based COVID-19 detection, remains challenging.
“While audio-based applications have seen significant advancements, deploying them in real-world settings, especially for audio-based COVID-19 detection, remains challenging [86].”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [86] Coppock H, Jones L, Kiskin I, Schuller B. 2021 COVID-19 detection from audio: seven grains of salt. Lancet Digit. Health 3, e537- e538. (10.1016/S2589-7500(21)00141-2) 34303644 PMC8294804 source
Not checked in the free run.
- Premisecitation
Speech contains rich information that can infer human mental and physical health.
“Speech, as one of the dominant sounds, contains rich information that can infer human mental health and physical health [34,39].”
Quote in the paperNot a substantial claim0 of 2 sources checked- cites [34] Tripathi S, Tripathi S, Beigi H. 2018 Multi-modal emotion recognition on iemocap dataset using deep learning. (http://arxiv.org/abs/1804.05788)
Not checked in the free run.
- cites [39] Ringeval F et al. 2019 AVEC 2019 workshop and challenge: state-of-mind, detecting depression with AI, and cross-cultural affect recognition. In Proc. 9th Int. Conf. on Audio/Visual Emotion Challenge and Workshop, pp. 3–12. source
Not checked in the free run.
- Premisenone given
Audio, location, and activity tracking typically involve passive sensing, minimizing the need for active user participation.
“Moreover, they typically involve passive sensing, minimizing the need for active user participation as opposed to camera-based vision techniques.”
Quote in the paperSubstantial claim - Premisecitation
Audio sensing has been applied to mental health conditions including depression, bipolar, schizophrenia, autism, stress, and mood.
“Depression, as one of the leading mental disorders, has been studied using audio sounds [76]. Other mental disorders have also been investigated using audio sensing, including bipolar [66], schizophrenia [77], autism [78]. Stress and mood which are highly correlated to mental health have also been studied using audio sounds [67,76].”
Quote in the paperNot a substantial claim0 of 3 sources checked- cites [66] Osmani V. 2015 Smartphones in mental health: detecting depressive and manic episodes. IEEE Pervasive Comput. 14, 10- 13. (10.1109/MPRV.2015.54) source
Not checked in the free run.
- cites [76] Ben-Zeev D, Scherer EA, Wang R, Xie H, Campbell AT. 2015 Next-generation psychiatric assessment: using smartphone sensors to monitor behavior and mental health. Psychiatr. Rehabil. J. 38, 218. (10.1037/prj0000130) 25844912 PMC4564327 source
Not checked in the free run.
- cites [77] Wang R et al. 2016 CrossCheck: toward passive sensing and detection of mental health changes in people with schizophrenia. In Proc. 2016 ACM Int. Joint Conf. on Pervasive and Ubiquitous Computing, pp. 886–897. source
Not checked in the free run.
- Premisecitation
Smartphone microphone can quantify flow and volume of air inspired and expired by lungs during breathing, reflecting degree of airflow obstruction in COPD and asthma.
“Some studies used the microphone of smartphones to quantify the flow and volume of air inspired and expired by the lungs during breathing, which reflects the degree of airflow obstruction in clinically associated pulmonary disorders such as COPD and asthma [82,83].”
Quote in the paperNot a substantial claim0 of 2 sources checked- cites [82] Chen NC, Wang KC, Chu HH. 2012 Listen-to-nose: a low-cost system to record nasal symptoms in daily life. In Proc. 2012 ACM Conf. on Ubiquitous Computing, pp. 590–591. source
Not checked in the free run.
- cites [83] Larson EC, Goel M, Redfield M, Boriello G, Rosenfeld M, Patel SN. 2013 Tracking lung function on any phone. In Proc. 3rd ACM Symp. on Computing for Development, pp. 1–2. source
Not checked in the free run.
- Premisecitation
Recent large-scale longitudinal studies have shown promise in leveraging wearable sensor data for health monitoring and prediction.
“While recent large-scale longitudinal studies have shown promise in leveraging wearable sensor data for health monitoring and prediction, they also introduce new challenges for data modelling.”
- Also stated in: § Introduction · plain
“recently there have been several large longitudinal studies like the UK Biobank [17], the Apple Study [18], the Fenland Study [19], Utsureko [20] and EmotionSense [21] with tens of thousands of participants using wearable sensors.”
Quote in the paperSubstantial claim0 of 3 sources checked- cites [17] Doherty A et al. 2017 Large scale population assessment of physical activity using wrist worn accelerometers: the UK Biobank study. PLoS ONE 12, e0169649. (10.1371/journal.pone.0169649) 28146576 PMC5287488 source
Not checked in the free run.
- cites [18] Perez MV et al. 2019 Large-scale assessment of a smartwatch to identify atrial fibrillation. N. Engl. J. Med. 381, 1909- 1917. (10.1056/NEJMoa1901183) 31722151 PMC8112605 source
Not checked in the free run.
- cites [22] Radin JM, Wineinger NE, Topol EJ, Steinhubl SR. 2020 Harnessing wearable device data to improve state-level real-time surveillance of influenza-like illness in the USA: a population-based study. Lancet Digit. Health 2, e85- e93. (10.1016/S2589-7500(19)30222-5) 33334565 PMC8048388 source
Not checked in the free run.
- Also stated in: § Introduction · plain
- Premisecitation
Changes in affective state and cognitive effects associated with depression alter mechanisms involved in speech production.
“It was proven that changes in affective state and cognitive effects associated with depression have altered the mechanisms involved in speech production [57,58].”
Quote in the paperSubstantial claim0 of 2 sources checked- cites [57] Levelt WJ, Roelofs A, Meyer AS. 1999 A theory of lexical access in speech production. Behav. Brain Sci. 22, 61- 69. (10.1017/S0140525X99451775) 11301520 source
Not checked in the free run.
- cites [58] Davidson RJ, Pizzagalli D, Nitschke JB, Putnam K. 2002 Depression: perspectives from affective neuroscience. Annu. Rev. Psychol. 53, 545- 574. (10.1146/annurev.psych.53.100901.135148) 11752496 source
Not checked in the free run.
- Premisecitation
Eating behaviours are analysed using audio collected from smartwatches and earables.
“Eating behaviours are also analysed using audio collected from smartwatches and earables [47,51,85].”
Quote in the paperNot a substantial claim0 of 3 sources checked- cites [47] Amft O, Stäger M, Lukowicz P, Tröster G. 2005 Analysis of chewing sounds for dietary monitoring. In UbiComp 2005: ubiquitous computing (eds M Beigl, S Intille, J Rekimoto, H Tokuda), pp. 56- 72. Berlin, Germany: Springer. (10.1007/11551201_4) source
Not checked in the free run.
- cites [51] Lotfi R, Tzanetakis G, Eskicioglu R, Irani P. 2020 A comparison between audio and IMU data to detect chewing events based on an earable device. In Proc. 11th Augmented Human Int. Conf., pp. 1–8. source
Not checked in the free run.
- cites [85] Kalantarian H, Sarrafzadeh M. 2015 Audio-based detection and evaluation of eating behavior using the smartwatch platform. Comput. Biol. Med. 65, 1- 9. (10.1016/j.compbiomed.2015.07.013) 26241487 source
Not checked in the free run.
- Interpretationreasoning
Audio can be used for both mental and physical health detection and monitoring.
“It can be used for both mental and physical health detection and monitoring.”
Quote in the paperSubstantial claim - Premisecitation
Large longitudinal studies introduce new challenges for data modelling.
“they also introduce new challenges for data modelling.”
- Also stated in: § Introduction · plain
“However, these large studies introduced new challenges for data modelling.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [23] Zhang X, Zhou Y, Ma Y, Chen BC, Zhang L, Agarwal D. 2016 Glmix: generalized linear mixed models for large-scale response prediction. In Proc. 22nd ACM SIGKDD Int. Conf. on Knowledge Discovery and Data Mining, pp. 363–372. New York, NY: ACM. source
Not checked in the free run.
- Also stated in: § Introduction · plain
- Interpretationreasoning
Statistical analysis and proof-of-concept results demonstrate the potential of audio sounds collected through mobile devices for health monitoring.
“These statistical analysis or proof-of-concept results demonstrate the potential of audio sounds collected through different mobile devices for health monitoring.”
Quote in the paperSubstantial claim - Premisecitation
Acoustic properties of speech in terms of formants and power distribution are best discriminators to distinguish between depression and suicidality individuals and healthy individuals.
“Acoustic properties of speech in terms of formants and power distribution were found to be best discriminators to distinguish between depression and suicidality individuals and healthy individuals [59].”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [59] France DJ, Shiavi RG, Silverman S, Silverman M, Wilkes M. 2000 Acoustical properties of speech as indicators of depression and suicidal risk. IEEE Trans. Biomed. Eng. 47, 829- 837. (10.1109/10.846676) 10916253 source
Not checked in the free run.
- Premisecitation
Obtaining quality annotations and ground truth for mobile health data is often expensive or impractical.
“Furthermore, obtaining quality annotations and ground truth for such data is often expensive or impractical.”
- Also stated in: § Introduction · plain
“Further, obtaining quality annotations and ground truth is costly or even impossible at this granularity.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [16] Brenner PS, DeLamater JD. 2014 Social desirability bias in self-reports of physical activity: is an exercise identity the culprit? Soc. Indic. Res. 117, 489- 504. (10.1007/s11205-013-0359-y) source
Not checked in the free run.
- Also stated in: § Introduction · plain
- Premisenone given
Audio contains rich information for health.
“Audio, as one of the most natural human communication signals, contains rich information.”
- Also stated in: § Applications · plain
“Audio sounds contain rich health-related information and different types of audio sounds additionally reveal complementary information.”
Quote in the paperSubstantial claim - Also stated in: § Applications · plain
- Premisecitation
Breathing patterns extracted from audio can monitor sleep apnoea.
“Breathing patterns are extracted from audio signal to monitor sleep apnoea [81]”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [81] Alqassim S, Ganesh M, Khoja S, Zaidi M, Aloul F, Sagahyroon A. 2012 Sleep apnea monitoring using mobile phones. In 2012 IEEE 14th Int. Conf. on e-Health Networking, Applications and Services (Healthcom), pp. 443–446. IEEE. source
Not checked in the free run.
- Premisecitation
Significant correlations exist between conversation duration and/or frequency computed from audio data and PHQ-9 depression scores, flourishing scale, and perceived stress scale.
“Significant correlations are obtained between conversation duration and/or frequency computed from audio data and PHQ-9 depression scores, flourishing scale and perceived stress scale (PSS) [63].”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [63] Wang R, Chen F, Chen Z, Li T, Harari G, Tignor S, Zhou X, Ben-Zeev D, Campbell AT. 2014 StudentLife: assessing mental health, academic performance and behavioral trends of college students using smartphones. In Proc. 2014 ACM Int. Joint Conf. on Pervasive and Ubiquitous Computing, pp. 3–14. source
Not checked in the free run.
- Premisecitation
The application of machine learning to mobile health sensing poses unique challenges due to noisy sensor measurements, high-dimensional data, sparse and irregular time series, heterogeneity in data, privacy concerns, and resource constraints.
“However, the application of machine learning (ML) to mobile health sensing poses unique challenges due to noisy sensor measurements, high-dimensional data, sparse and irregular time series, heterogeneity in data, privacy concerns and resource constraints.”
Quote in the paperSubstantial claim0 of 3 sources checked- cites [13] Ang WT, Khosla PK, Riviere CN. 2007 Nonlinear regression model of a low-g MEMS accelerometer. IEEE Sensors J. 7, 81- 88. (10.1109/JSEN.2006.886995) source
Not checked in the free run.
- cites [14] Lane ND, Miluzzo E, Lu H, Peebles D, Choudhury T, Campbell AT. 2010 A survey of mobile phone sensing. IEEE Commun. Mag. 48, 140- 150. (10.1109/MCOM.2010.5560598) source
Not checked in the free run.
- cites [15] Radu V, Tong C, Bhattacharya S, Lane ND, Mascolo C, Marina MK, Kawsar F. 2018 Multimodal deep learning for activity and context recognition. Proc. ACM Interact. Mob. Wearable Ubiquitous Technol. 1, 1- 27. (10.1145/3161174) source
Not checked in the free run.
- Premisecitation
It is possible to predict hospitalization using mobility behaviour.
“It is possible to predict hospitalization using mobility behaviour [119].”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [119] Zhang Y, Xu F, Li T, Kostakos V, Hui P, Li Y. 2021 Passive health monitoring using large scale mobility data. Proc. ACM Interact. Mob. Wearable Ubiquitous Technol. 5, 1- 23. (10.1145/3448078) source
Not checked in the free run.
- Premisecitation
Significant correlation exists between mobility trace characteristics and the depressive moods of individuals.
“GPS traces also produce similar conclusions [117]—significant correlation exists between mobility trace characteristics and the depressive moods of individuals.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [117] Canzian L, Musolesi M. 2015 Trajectories of depression: unobtrusive monitoring of depressive states by means of smartphone mobility traces analysis. In Proc. 2015 ACM Int. Joint Conf. on Pervasive and Ubiquitous Computing, pp. 1293–1304. source
Not checked in the free run.
- Premisereasoning
Features from mobility data are not standardized.
“Unlike many other sensing modalities (e.g. audio), features from mobility data are not standardized.”
Quote in the paperNot a substantial claim - Premisecitation
A patient's mobility patterns differ when accompanied by carers.
“A patient’s mobility patterns differ when accompanied by carers.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [115] Ghosh A, Puthusseryppady V, Chan D, Mascolo C, Hornberger M. 2022 Machine learning detects altered spatial navigation features in outdoor behaviour of Alzheimer’s disease patients. Sci. Rep. 12, 3160. (10.1038/s41598-022-06899-w) 35210486 PMC8873255 source
Not checked in the free run.
- Premisecitation
In one study, destinations for dementia patients (with 95% accuracy) are more predictable than the control population (with 92% accuracy).
“It was found that the destinations for the dementia patients (with 95% accuracy) are more predictable than the control population (with 92% accuracy).”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [113] Bayat S, Mihailidis A. 2021 Outdoor life in dementia: how predictable are people with dementia in their mobility? Alzheimer’s Dement. (Amst). 13, e12187. (10.1002/dad2.12187) 34027017 PMC8118112 source
Not checked in the free run.
- Premisecitation
In a city region the distribution of POI categories is a good predictor of the distribution of progression of chronic diseases.
“The main insight here is that in a city region the distribution of POI categories is a good predictor of the distribution of progression of chronic diseases.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [118] Wang Y, Zhou X, Mascolo C, Noulas A, Xie X, Liu Q. 2018 Predicting the spatio-temporal evolution of chronic diseases in population with human mobility data. In Proc. 27th Int. Joint Conf. on Artificial Intelligence, pp. 3578–3584. (10.24963/ijcai.2018/497) source
Not checked in the free run.
- Interpretationreasoning
Behavioural changes in mobility are useful in predicting or diagnosing a health problem passively.
“Such behavioural changes are useful in predicting or diagnosing a health problem passively.”
Quote in the paperNot a substantial claim - Premisecitation
Schwabe et al. proposed a mobility marked Hawkes model combining a Hawkes process to model transmission dynamics and a regularized Poisson regression to incorporate mobility covariates extracted from telecommunication data.
“For example, Schwabe et al. [100] proposed mobility marked Hawkes model combining Hawkes process to model the transmission dynamics of the disease and a regularized Poisson regression to incorporate mobility covariates extracted from telecommunication data.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [100] Schwabe A, Persson J, Feuerriegel S. 2021 Predicting COVID-19 spread from large-scale mobility data. (http://arxiv.org/abs/2106.00356) source
Not checked in the free run.
- Premisecitation
The most important driving based features were average jerk, number of night trips and radius of gyration.
“The most important driving based features were average jerk, number of night trips and radius of gyration.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [114] Bayat S, Babulal GM, Schindler SE, Fagan AM, Morris JC, Mihailidis A, Roe CM. 2021 GPS driving: a digital biomarker for preclinical Alzheimer disease. Alzheimer’s Res. Ther. 13, 1- 9. (10.1186/s13195-021-00852-1) 34127064 PMC8204509 source
Not checked in the free run.
- Premisecitation
Using disease simulation on aggregated inter-province movement counts in China as derived from Baidu location based services, travel restrictions in Wuhan introduced on 23 January 2020 reduced the number of cases exported internationally by 70-80% until mid February 2020.
“Using disease simulation on aggregated inter-province movement counts in China as derived from Baidu location based services [101] has shown that the travel restrictions in Wuhan introduced on 23 January 2020 had reduced the number of cases exported internationally by 70–80% until mid February 2020.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [101] Chinazzi M et al. 2020 The effect of travel restrictions on the spread of the 2019 novel coronavirus (COVID-19) outbreak. Science 368, 395- 400. (10.1126/science.aba9757) 32144116 PMC7164386 source
Not checked in the free run.
- Premisecitation
Alzheimer's patients have significantly different outdoor mobility behaviour than control population even in early stage: patients explore spatially less compared to controls.
“Further it has been shown that Alzheimer’s patients have significantly different outdoor mobility behaviour than control population even in early stage [115]: patients explore spatially less compared to controls.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [115] Ghosh A, Puthusseryppady V, Chan D, Mascolo C, Hornberger M. 2022 Machine learning detects altered spatial navigation features in outdoor behaviour of Alzheimer’s disease patients. Sci. Rep. 12, 3160. (10.1038/s41598-022-06899-w) 35210486 PMC8873255 source
Not checked in the free run.
- Premisecitation
Strategies to reopen an economy from lockdowns have been studied using check-in data, Wi-Fi association logs and GPS traces.
“Strategies to reopen an economy from lockdowns has been studied [106,107] using check-in data, Wi-Fi association logs and GPS traces.”
Quote in the paperNot a substantial claim0 of 2 sources checked- cites [106] Chang S, Pierson E, Koh PW, Gerardin J, Redbird B, Grusky D, Leskovec J. 2021 Mobility network models of COVID-19 explain inequities and inform reopening. Nature 589, 82- 87. (10.1038/s41586-020-2923-3) 33171481 source
Not checked in the free run.
- cites [107] Wang H, Ghosh A, Ding J, Sarkar R, Gao J. 2021 Heterogeneous interventions reduce the spread of COVID-19 in simulations on real mobility data. Sci. Rep. 11, 7809. (10.1038/s41598-021-87034-z) 33833298 PMC8034422 source
Not checked in the free run.
- Premisecitation
Individual mobility behaviour changes in fever, influenza or mental stress situations.
“Individual mobility behaviour changes in fever, influenza or mental stress situations [116].”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [116] Madan A, Cebrian M, Lazer D, Pentland A. 2010 Social sensing for epidemiological behavior change. In Proc. 12th ACM Int. Conf. on Ubiquitous Computing, pp. 291–300. source
Not checked in the free run.
- Premisecitation
Analysing CDR data for the most affected counties in the USA found that decreased mobility strongly correlates (Pearson correlation coefficients above 0.7) with decreased COVID-19 case growth rates.
“CDR data are also useful to draw similar conclusions—analysing the data for the most affected counties in the USA [102] found that the decreased mobility strongly correlates (Pearson correlation coefficients above 0.7) with decreased COVID-19 case growth rates.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [102] Badr HS, Du H, Marshall M, Dong E, Squire MM, Gardner LM. 2020 Association between mobility patterns and COVID-19 transmission in the USA: a mathematical modelling study. Lancet Infect. Dis. 20, 1247- 1254. (10.1016/S1473-3099(20)30553-3) 32621869 PMC7329287 source
Not checked in the free run.
- Premisecitation
Features extracted from driving trajectories are found to have potential to become digital biomarkers for Alzheimer's disease (AD).
“Features extracted from driving trajectories are found to have potential to become digital biomarkers for Alzheimer’s disease (AD) [114].”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [114] Bayat S, Babulal GM, Schindler SE, Fagan AM, Morris JC, Mihailidis A, Roe CM. 2021 GPS driving: a digital biomarker for preclinical Alzheimer disease. Alzheimer’s Res. Ther. 13, 1- 9. (10.1186/s13195-021-00852-1) 34127064 PMC8204509 source
Not checked in the free run.
- Interpretationcitation
Incorporating mobility in a disease spread model increases its predictive power.
“While traditional disease spread models like SEIR (susceptible exposed infected recovered) assume contacts between random pairs of individuals, incorporating mobility in the model increases its predictive power.”
Quote in the paperSubstantial claim- Rests on an assumption: Infectious diseases such as COVID-19 and H1N1 are spread primarily by human contacts.
- Premisecitation
Less mobile people with less visit diversity, and who visit fewer sports facilities are more likely to visit hospitals.
“Interesting findings include that less mobile people with less visit diversity, and who visit fewer sports facilities are more likely to visit hospitals; fast food store significantly affects health of younger people.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [119] Zhang Y, Xu F, Li T, Kostakos V, Hui P, Li Y. 2021 Passive health monitoring using large scale mobility data. Proc. ACM Interact. Mob. Wearable Ubiquitous Technol. 5, 1- 23. (10.1145/3448078) source
Not checked in the free run.
- Premisecitation
The mobility marked Hawkes model outperforms other models in forecasting infections 5-21 days in the future in Switzerland.
“The model outperforms the other models in forecasting the infections in the future for 5–21 days in Switzerland.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [100] Schwabe A, Persson J, Feuerriegel S. 2021 Predicting COVID-19 spread from large-scale mobility data. (http://arxiv.org/abs/2106.00356) source
Not checked in the free run.
- Premisecitation
Behavioural changes are reflected in total communication, interactions with respect to time of day, diversity and entropy of face-to-face interactions and movement.
“Behavioural changes are reflected in total communication, interactions with respect to time of day (e.g. late night, early morning), diversity and entropy of face-to-face interactions and movement.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [116] Madan A, Cebrian M, Lazer D, Pentland A. 2010 Social sensing for epidemiological behavior change. In Proc. 12th ACM Int. Conf. on Ubiquitous Computing, pp. 291–300. source
Not checked in the free run.
- Premisecitation
Navigation lifestyle analysis is helpful in studying longitudinal disease progression.
“Apart from diagnostic inference at snapshots, navigation lifestyle analysis is helpful in studying longitudinal disease progression [118].”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [118] Wang Y, Zhou X, Mascolo C, Noulas A, Xie X, Liu Q. 2018 Predicting the spatio-temporal evolution of chronic diseases in population with human mobility data. In Proc. 27th Int. Joint Conf. on Artificial Intelligence, pp. 3578–3584. (10.24963/ijcai.2018/497) source
Not checked in the free run.
- Premisecitation
Several studies across different countries have shown lockdowns' effectiveness.
“Several studies across different countries have shown their effectiveness [101,102].”
Quote in the paperNot a substantial claim0 of 2 sources checked- cites [101] Chinazzi M et al. 2020 The effect of travel restrictions on the spread of the 2019 novel coronavirus (COVID-19) outbreak. Science 368, 395- 400. (10.1126/science.aba9757) 32144116 PMC7164386 source
Not checked in the free run.
- cites [102] Badr HS, Du H, Marshall M, Dong E, Squire MM, Gardner LM. 2020 Association between mobility patterns and COVID-19 transmission in the USA: a mathematical modelling study. Lancet Infect. Dis. 20, 1247- 1254. (10.1016/S1473-3099(20)30553-3) 32621869 PMC7329287 source
Not checked in the free run.
- Premisecitation
The health status of an individual is predicted with a Bayesian network classifier, without having actual health measurements from the subjects.
“Finally the health status of an individual is predicted with a Bayesian network classifier, without having actual health measurements from the subjects.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [116] Madan A, Cebrian M, Lazer D, Pentland A. 2010 Social sensing for epidemiological behavior change. In Proc. 12th ACM Int. Conf. on Ubiquitous Computing, pp. 291–300. source
Not checked in the free run.
- Premisecitation
Using data from 75 without and 64 people with preclinical AD, a random forest classifier identified preclinical AD patients with F1 score 0.82.
“Using the data from 75 without and 64 people with preclinical AD, the authors built a random forest classifier for identifying preclinical AD patients with F1 score as 0.82.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [114] Bayat S, Babulal GM, Schindler SE, Fagan AM, Morris JC, Mihailidis A, Roe CM. 2021 GPS driving: a digital biomarker for preclinical Alzheimer disease. Alzheimer’s Res. Ther. 13, 1- 9. (10.1186/s13195-021-00852-1) 34127064 PMC8204509 source
Not checked in the free run.
- Premisecitation
Fast food store significantly affects health of younger people.
“fast food store significantly affects health of younger people.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [119] Zhang Y, Xu F, Li T, Kostakos V, Hui P, Li Y. 2021 Passive health monitoring using large scale mobility data. Proc. ACM Interact. Mob. Wearable Ubiquitous Technol. 5, 1- 23. (10.1145/3448078) source
Not checked in the free run.
- Premisecitation
FitRec, an LSTM-based approach to modelling HR and activity data for personalized fitness recommendations, was able to learn activity-specific contextual and personalized dynamics of individual user HR profiles during exercise segments.
“For example, FitRec, an LSTM-based approach to modelling HR and activity data for personalized fitness recommendations, was able to learn activity-specific contextual and personalized dynamics of individual user HR profiles during exercise segments [56].”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [56] Ni J, Muhlstein L, McAuley J. 2019 Modeling heart rate and activity data for personalized fitness recommendation. In The World Wide Web Conf., pp. 1343–1353. source
Not checked in the free run.
- Premisecitation
We recently developed a self-supervised model which exploits the multi-modal data of modern wearables to learn meaningful representations which generalize to several outcomes with transfer learning.
“Further, we recently developed a self-supervised model which exploits the multi-modal data of modern wearables to learn meaningful representations which generalize to several outcomes with transfer learning [144].”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [144] Spathis D, Perez-Pozuelo I, Brage S, Wareham NJ, Mascolo C. 2021 Self-supervised transfer learning of physiological representations from free-living wearable data. In Proc. Conf. on Health, Inference, and Learning, CHIL ’21, pp. 69–78. New York, NY: ACM. source
Not checked in the free run.
- Premisecitation
The validity of non-exercise estimation of VO2max is still much lower than what can be achieved with dynamic exercise testing.
“However, the validity of such estimation is still much lower than what can be achieved with dynamic exercise testing [139].”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [139] Gonzales TI, Westgate K, Hollidge S, Lindsay T, Jeon J, Brage S. 2020 Estimating maximal oxygen consumption from heart rate response to submaximal ramped treadmill test. medRxiv. (10.1101/2020.02.18.20024489) source
Not checked in the free run.
- Premisecitation
The model detects fitness change over time (e.g. after 7 years).
“It also detects fitness change over time (e.g. after 7 years).”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [136] Spathis D, Perez-Pozuelo I, Gonzales TI, Wu Y, Brage S, Wareham N, Mascolo C. 2022 Longitudinal cardio-respiratory fitness prediction through wearables in free-living environments. npj Digit. Med. 5, 176. (10.1038/s41746-022-00719-1) 36460766 PMC9718831 source
Not checked in the free run.
- Premisecitation
Long travel times to health facilities strongly correlate with increased mobility in geographically isolated areas.
“They found that long travel times to health facilities strongly correlate with increased mobility in geographically isolated areas.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [121] Wesolowski A, O’Meara WP, Tatem AJ, Ndege S, Eagle N, Buckee CO. 2015 Quantifying the impact of accessibility on preventive healthcare in sub-Saharan Africa using mobile phone data. Epidemiology 26, 223. (10.1097/EDE.0000000000000239) 25643101 PMC4323566 source
Not checked in the free run.
- Interpretationreasoning
Leveraging movement and cardiovascular response concurrently likely produces better representations than either signal taken in isolation.
“thus, leveraging these two signals concurrently likely produces better representations than either signal taken in isolation.”
Quote in the paperSubstantial claim - Interpretationreasoning
A correlation analysis without considering age will lead to a negative correlation between park visits and health status.
“Thus a correlation analysis without considering age will lead to a negative correlation between park visits and health status.”
Quote in the paperNot a substantial claim- Rests on an assumption: Older people visit parks more often but have worse health status than younger people.
- Interpretationreasoning
Extracting meaning from wearable data can be challenging, since sensors measure low-level signals (e.g. acceleration) as opposed to the more high-level events that are usually of interest (e.g. arrhythmia, infection or obesity onset).
“However, extracting meaning from these data can be challenging, since sensors measure low-level signals (e.g. acceleration) as opposed to the more high-level events that are usually of interest (e.g. arrhythmia, infection or obesity onset).”
Quote in the paperSubstantial claim - Premisecitation
Non-exercise models are usually traditional regression models and incorporate variables like sex, age, body mass index (BMI), RHR and self-reported physical activity to infer VO2max.
“These are usually traditional regression models and incorporate variables like sex, age, body mass index (BMI), RHR and self-reported physical activity to infer VO2max [138].”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [138] Cao ZB, Miyatake N, Higuchi M, Miyachi M, Ishikawa-Takata K, Tabata I. 2010 Predicting VO2max with an objectively measured physical activity in Japanese women. Med. Sci. Sports Exerc. 42, 179- 186. (10.1249/MSS.0b013e3181af238d) 20010115 source
Not checked in the free run.
- Interpretationreasoning
FitRec requires prior segmentation of activities, which can be a constraint when applying these techniques in free-living, unconstrained conditions.
“This approach is helpful but requires prior segmentation of activities, which can be a constraint when applying these techniques in free-living, unconstrained conditions.”
Quote in the paperSubstantial claim - Premisecitation
Causal relation between mobility and health status has also been analysed.
“Causal relation between mobility and health status has also been analysed [120].”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [120] Zhang Y, Xu F, Xia T, Li Y. 2021 Quantifying the causal effect of individual mobility on health status in urban space. Proc. ACM Interact. Mob. Wearable Ubiquitous Technol. 5, 1- 30. (10.1145/3494990) source
Not checked in the free run.
- Premisecitation
Mobile phone-derived measures of mobility can predict which regions are lacking preventive care.
“Furthermore, they have shown that the mobile phone-derived measures of mobility can predict which regions are lacking preventive care.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [121] Wesolowski A, O’Meara WP, Tatem AJ, Ndege S, Eagle N, Buckee CO. 2015 Quantifying the impact of accessibility on preventive healthcare in sub-Saharan Africa using mobile phone data. Epidemiology 26, 223. (10.1097/EDE.0000000000000239) 25643101 PMC4323566 source
Not checked in the free run.
- Premisecitation
A recent large-scale analysis using Fenland Study (N = 11 059), its longitudinal cohort (N = 2675), and UK Biobank Validation Study (N = 181) showed that combination of wearables and other biomarkers as inputs to neural networks yields a strong correlation to ground truth in a holdout sample (r = 0.82, 95% CI: 0.80-0.83).
“A recent large-scale analysis used data from Fenland Study (N = 11 059), along with its longitudinal cohort (N = 2675), and a third external cohort using the UK Biobank Validation Study (N = 181) who underwent maximal VO2max testing, to show that the combination of wearables and other biomarkers as inputs to neural networks yields a strong correlation to ground truth in a holdout sample (r = 0.82,”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [136] Spathis D, Perez-Pozuelo I, Gonzales TI, Wu Y, Brage S, Wareham N, Mascolo C. 2022 Longitudinal cardio-respiratory fitness prediction through wearables in free-living environments. npj Digit. Med. 5, 176. (10.1038/s41746-022-00719-1) 36460766 PMC9718831 source
Not checked in the free run.
- Premisecitation
Physical activity is characterized by both movement and the associated cardiovascular response to movement.
“In particular, physical activity is characterized by both movement and the associated cardiovascular response to movement (e.g. heart rate increases after exercise and the dynamics of this increase are dictated by fitness levels [132]); thus, leveraging these two signals concurrently likely produces better representations than either signal taken in isolation.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [132] Jones AM, Carter H. 2000 The effect of endurance training on parameters of aerobic fitness. Sports Med. 29, 373- 386. (10.2165/00007256-200029060-00001) 10870864 source
Not checked in the free run.
- Interpretationcitation
These results demonstrate the value of wearables for fitness estimation that today can be measured only with laboratory tests.
“These results demonstrate the value of wearables for fitness estimation that today can be measured only with laboratory tests.”
Quote in the paperSubstantial claim - Premisecitation
Large population-scale studies showed that HR responses can predict VO2max.
“while large population-scale studies showed that these HR responses can predict VO2max [136].”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [136] Spathis D, Perez-Pozuelo I, Gonzales TI, Wu Y, Brage S, Wareham N, Mascolo C. 2022 Longitudinal cardio-respiratory fitness prediction through wearables in free-living environments. npj Digit. Med. 5, 176. (10.1038/s41746-022-00719-1) 36460766 PMC9718831 source
Not checked in the free run.
- Premisecitation
A study used cell tower location data from 14 million subscribers tracked for a year in rural Kenya to find the correlation between the mobility of the people, access to healthcare and uptake of two preventive healthcare interventions: childhood immunizations and antenatal care.
“A study [121] used cell tower location data from 14 million subscribers tracked for a year in rural Kenya to find the correlation between the mobility of the people, access to healthcare and uptake of two preventive healthcare interventions: childhood immunizations and antenatal care.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [121] Wesolowski A, O’Meara WP, Tatem AJ, Ndege S, Eagle N, Buckee CO. 2015 Quantifying the impact of accessibility on preventive healthcare in sub-Saharan Africa using mobile phone data. Epidemiology 26, 223. (10.1097/EDE.0000000000000239) 25643101 PMC4323566 source
Not checked in the free run.
- Premisecitation
The causal framework is extended to health prediction models selecting causal features resulting in a robust model.
“The framework is extended to health prediction models selecting causal features resulting in a robust model.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [120] Zhang Y, Xu F, Xia T, Li Y. 2021 Quantifying the causal effect of individual mobility on health status in urban space. Proc. ACM Interact. Mob. Wearable Ubiquitous Technol. 5, 1- 30. (10.1145/3494990) source
Not checked in the free run.
- Premisecitation
Heart rate responses to exercise have been shown to be strongly predictive of CVD, coronary heart disease (CHD) and all-cause mortality.
“For instance, heart rate (HR) responses to exercise have been shown to be strongly predictive of CVD, coronary heart disease (CHD) and all-cause mortality [133].”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [133] Savonen KP, Lakka TA, Laukkanen JA, Halonen PM, Rauramaa TH, Salonen JT, Rauramaa R. 2006 Heart rate response during exercise test and cardiovascular mortality in middle-aged men. Eur. Heart J. 27, 582- 588. (10.1093/eurheartj/ehi708) 16399774 source
Not checked in the free run.
- Interpretationreasoning
In many scenarios the effect of access to resources on health can be better understood and managed by studying the locations of the facilities and their users.
“In many such scenarios the effect can be better understood and managed by studying the locations of the facilities and their users.”
Quote in the paperSubstantial claim- Rests on an assumption: Access to resources heavily affects health outcome be it the facilities for health being in the neighbourhood or critical machines in a hospital.
- Premisecitation
In healthy individuals, HR responses to activity are defined by an increase in HR that is concurrent to the increasing intensity of the activity.
“In healthy individuals, HR responses to activity are defined by an increase in HR that is concurrent to the increasing intensity of the activity [134].”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [134] Ellestad MH, Wan M. 1975 Predictive implications of stress testing. Follow-up of 2700 subjects after maximum treadmill stress testing. Circulation 51, 363- 369. (10.1161/01.CIR.51.2.363) 1112017 source
Not checked in the free run.
- Premisecitation
Indoor positioning is used to protect visually impaired, elderly people from collisions.
“Further indoor positioning is used to protect visually impaired, elderly people from collisions [126].”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [126] Xiao F, Miao Q, Xie X, Sun L, Wang R. 2018 Indoor anti-collision alarm system based on wearable Internet of Things for smart healthcare. IEEE Commun. Mag. 56, 53- 59. (10.1109/MCOM.2018.1700706) source
Not checked in the free run.
- Premisecitation
The dynamics of heart rate increase after exercise are dictated by fitness levels.
“the dynamics of this increase are dictated by fitness levels [132]”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [132] Jones AM, Carter H. 2000 The effect of endurance training on parameters of aerobic fitness. Sports Med. 29, 373- 386. (10.2165/00007256-200029060-00001) 10870864 source
Not checked in the free run.
- Premisecitation
Self-reported sadness was found to be an indicator of depression.
“Self-reported sadness was found to be an indicator of depression [151], while self-reported happiness is linked to longevity [152], personality traits [153] and reduced mortality risk [154].”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [151] Cheng H, Furnham A. 2003 Personality, self-esteem, and demographic predictions of happiness and depression. Pers. Individ. Differ. 34, 921- 942. (10.1016/S0191-8869(02)00078-8) source
Not checked in the free run.
- Premisecitation
Other more complex sensing modalities such as EEG and EOG are less employed in consumer wearable and mobile devices, rather in only clinical devices that limit their employment.
“Apart from these two types of sensors, there are also other more complex sensing modalities such as electroencephalography (EEG) and electrooculography (EOG) [171], e.g. for sleep monitoring, but they are less employed in consumer wearable and mobile devices, rather in only clinical devices that limit their employment.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [171] Guillodo E et al. 2020 Clinical applications of mobile health wearable-based sleep monitoring: systematic review. JMIR mHealth uHealth 8, e10733. (10.2196/10733) 32234707 PMC7160700 source
Not checked in the free run.
- Premisecitation
Clinical evidence shows that CRF is potentially a stronger predictor of mortality than well-established risk factors like hypertension, type 2 diabetes, high cholesterol or smoking.
“Clinical evidence shows that CRF is not only potentially a stronger predictor of mortality than well-established risk factors like hypertension, type 2 diabetes, high cholesterol or smoking, but that using CRF to complement these traditional risk factors significantly improves the precision of risk prediction models for adverse CVD health outcomes [148].”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [148] Ross R et al. 2016 Importance of assessing cardiorespiratory fitness in clinical practice: a case for fitness as a clinical vital sign: a scientific statement from the American Heart Association. Circulation 134, e653- e699. (10.1161/CIR.0000000000000461) 27881567 source
Not checked in the free run.
- Premisecitation
The apps lack in transparent quantification of individual advantage in obeying the app's advice.
“Further, the apps lack in transparent quantification of individual advantage in obeying the app’s advice [179].”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [179] Kaptchuk G, Goldstein DG, Hargittai E, Hofman J, Redmiles EM. 2020 How good is good enough for COVID19 apps? The influence of benefits, accuracy, and privacy on willingness to adopt. (http://arxiv.org/abs/2005.04343)
Not checked in the free run.
- Premisecitation
Most solutions to model interpretability have been implemented post hoc (e.g., attention mechanism) focusing on explanations in certain domains such as vision.
“Despite previous attempts to address the challenges surrounding model interpretability, most solutions have been implemented post hoc (such as the attention mechanism) focusing on explanations in certain domains, such as vision [184].”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [184] Samek W, Montavon G, Lapuschkin S, Anders CJ, Müller KR. 2021 Explaining deep neural networks and beyond: a review of methods and applications. Proc. IEEE 109, 247- 278. (10.1109/JPROC.2021.3060483) source
Not checked in the free run.
- Premisecitation
Mood and general mental wellbeing have been associated with several clinical outcomes.
“Mood and general mental wellbeing have been associated with several clinical outcomes.”
Quote in the paperSubstantial claim - Premisecitation
The majority of related literature has applied some kind of supervised learning, such as logistic regression or SVMs, which cannot capture nonlinear combinations of features.
“Regarding ML models, the majority of related literature has applied some kind of supervised learning, such as logistic regression or SVMs, which cannot capture nonlinear combinations of features.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [158] Thieme A, Belgrave D, Doherty G. 2020 Machine learning in mental health: a systematic review of the HCI literature to support the development of effective and implementable ML systems. ACM Trans. Comput.-Hum. Interact. 27, 1- 53. (10.1145/3398069) source
Not checked in the free run.
- Interpretationcitation
Uncertainty-aware modelling approaches enable prediction of uncertainty associated with final outcome predictions, providing confidence levels for users to trust or distrust systems.
“Another potential solution for enhancing interpretability is adopting uncertainty-aware modelling approaches. These enable the prediction of uncertainty associated with the final outcome predictions, thereby providing confidence levels for users to either trust or distrust the systems.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [185] Xia T, Han J, Qendro L, Dang T, Mascolo C. 2021 Uncertainty-aware COVID-19 detection from imbalanced sound data. (http://arxiv.org/abs/2104.02005) 38319779 source
Not checked in the free run.
- Premisecitation
Very recent work has demonstrated that by prompting LLMs with wearable time-series data and applying few-shot tuning, they are capable of making meaningful inferences on numerous health tasks for both clinical and wellness contexts.
“Very recent work has demonstrated that by prompting LLMs with wearable time-series data and applying few-shot tuning, they are capable of making meaningful inferences on numerous health tasks for both clinical and wellness contexts [147].”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [147] Liu X, McDuff D, Kovacs G, Galatzer-Levy I, Sunshine J, Zhan J, Poh MZ, Liao S, Di Achille P, Patel S. 2023 Large language models are few-shot health learners. (http://arxiv.org/abs/2305.15525)
Not checked in the free run.
- Interpretationreasoning
Timely and accurate mood reporting is important to anticipate clinical outcomes.
“Independently of the scale used, timely and accurate mood reporting is important to anticipate clinical outcomes.”
Quote in the paperSubstantial claim- Rests on an assumption: People without economic privilege have a limited capacity to self-isolate at home, which undermines the effectiveness of these approaches.
- Premisecitation
CRF is an important modifiable marker of cardiovascular health embodied by a strong inverse relationship with the incidence of CVD, type 2 diabetes, cancer, mortality and other adverse health outcomes.
“A concept that is particularly central to physical activity is that of CRF, an important modifiable marker of cardiovascular health embodied by a strong inverse relationship with the incidence of CVD, type 2 diabetes, cancer, mortality and other adverse health outcomes [131].”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [131] Mandsager K, Harb S, Cremer P, Phelan D, Nissen SE, Jaber W. 2018 Association of cardiorespiratory fitness with long-term mortality among adults undergoing exercise treadmill testing. JAMA Netw. Open 1, e183605. (10.1001/jamanetworkopen.2018.3605) 30646252 PMC6324439 source
Not checked in the free run.
- Premisecitation
Using CRF to complement traditional risk factors significantly improves the precision of risk prediction models for adverse CVD health outcomes.
“using CRF to complement these traditional risk factors significantly improves the precision of risk prediction models for adverse CVD health outcomes [148].”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [148] Ross R et al. 2016 Importance of assessing cardiorespiratory fitness in clinical practice: a case for fitness as a clinical vital sign: a scientific statement from the American Heart Association. Circulation 134, e653- e699. (10.1161/CIR.0000000000000461) 27881567 source
Not checked in the free run.
- Interpretationreasoning
Deploying large models on resource-limited devices such as smartphones and wearable devices presents a mounting challenge, especially regarding memory capacity and battery life.
“This task-agnostic shift, however, presents a mounting challenge in deploying these systems on resource-limited devices such as smartphones and wearable devices, especially when considering aspects like memory capacity and battery life.”
Quote in the paperSubstantial claim - Premisecitation
Federated learning has shown promise to preserve privacy while building useful models out of the data.
“More recently ML on decentralized data in terms of federated learning has shown promise to preserve privacy while building useful models out of the data [206,207].”
Quote in the paperNot a substantial claim0 of 2 sources checked- cites [206] Rieke N et al. 2020 The future of digital health with federated learning. npj Dig. Med. 3, 119. (10.1038/s41746-020-00323-1) PMC7490367 33015372 source
Not checked in the free run.
- cites [207] Xia T, Han J, Ghosh A, Mascolo C. 2023 Cross-device federated learning for mobile health diagnostics: a first study on COVID-19 detection. In ICASSP 2023-2023 IEEE Int. Conf. on Acoustics, Speech and Signal Processing (ICASSP), pp. 1–5. IEEE. source
Not checked in the free run.
- Premisecitation
Gender could be relevant as a sensitive attribute in depression screening due to reported gender differences in the disorder's signals.
“For example, in a depression screening application, gender could be relevant as a sensitive attribute due to reported gender differences in the disorder’s signals [192].”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [192] Parker G, Brotchie H. 2010 Gender differences in depression. Int. Rev. Psychiatry 22, 429- 436. (10.3109/09540261.2010.492391) 21047157 source
Not checked in the free run.
- Premisecitation
Despite recent efforts within the mobile and ubiquitous computing community, fairness and robustness are still underexplored with regard to mobile health sensing.
“Despite recent efforts within the mobile and ubiquitous computing community [196], fairness and robustness are still underexplored with regard to mobile health sensing.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [196] Yfantidou S, Constantinides M, Spathis D, Vakali A, Quercia D, Kawsar F. 2023 Beyond accuracy: a critical review of fairness in machine learning for mobile and wearable computing. (http://arxiv.org/abs/2303.15585)
Not checked in the free run.
- Interpretationreasoning
Fine-tuning or retraining models on edge devices is a challenge.
“However, when it requires fine-tuning or retraining the models on edge devices, it becomes a challenge.”
Quote in the paperSubstantial claim - Interpretationreasoning
Personalized systems can adapt and cater to heterogeneity among users, further enhancing effectiveness of disease monitoring and healthcare interventions.
“Such systems can adapt and cater to the heterogeneity among users, further enhancing the effectiveness of disease monitoring and healthcare interventions.”
Quote in the paperSubstantial claim - Premisecitation
Despite progress and even with advanced transformers, hurdles related to latency, energy consumption and compute resource management persist.
“Despite progress and even with advanced transformers [200], hurdles related to latency, energy consumption and compute resource management persist.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [200] Li Z, Wallace E, Shen S, Lin K, Keutzer K, Klein D, Gonzalez J. 2020 Train big, then compress: rethinking model size for efficient training and inference of transformers. In Proc. 37th Int. Conf. on Machine Learning, pp. 5958–5968. PMLR, vol. 119.
Not checked in the free run.
- Premisecitation
Existing methodologies like continual learning or transfer learning may present potential avenues for adaptation and personalization such as for electronic health record, but their exploration and application in mobile health remain relatively underdeveloped.
“Although existing methodologies like continual learning or transfer learning may present potential avenues for adaptation and personalization such as for electronic health record [190], their exploration and application in the context of mobile health remain relatively underdeveloped.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [190] Armstrong J, Clifton DA. 2022 Continual learning of longitudinal health records. In 2022 IEEE-EMBS Int. Conf. on Biomedical and Health Informatics (BHI) pp. 1–6. IEEE. source
Not checked in the free run.
- Premisecitation
Efforts have been made on both software and hardware to optimize deep learning inference, from model pruning and compression to parallel network computations.
“Some efforts have been made on both the software and hardware to optimize deep learning inference, from model pruning and compression techniques that reduce model size and complexity to parallel network computations [198].”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [198] Shuvo MMH, Islam SK, Cheng J, Morshed BI. 2022 Efficient acceleration of deep learning inference on resource-constrained edge devices: a review. Proc. IEEE 11, 42- 91. (10.1109/JPROC.2022.3226481)
Not checked in the free run.
- Interpretationreasoning
This limits applications in various contexts, e.g., personalizing the model with new user data or making continual adaptation with new data.
“This limits its applications in various contexts, e.g. personalizing the model with more new user data input, or making continual adaptation with new data.”
Quote in the paperSubstantial claim
- Classified as probably not a substantial claim. 79
- May combine two claims. 61
Probably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationCOVID-19 has been studied using voice, breathing, and cough sounds.
“COVID-19 has been studied using different audio sounds, including voice, breathing and cough sounds [10,29,79,80].”
Quote in the paperNot a substantial claim0 of 4 sources checked- cites [10] Han J et al. 2021 Sounds of COVID-19: exploring realistic performance of audio-based digital testing. npj Digit. Med. 5, 16. (10.1038/s41746-021-00553-x) PMC8799654 35091662 source
Not checked in the free run.
- cites [29] Coppock H, Gaskell A, Tzirakis P, Baird A, Jones L, Schuller B. 2021 End-to-end convolutional neural network enables COVID-19 detection from breath and cough audio: a pilot study. BMJ Innov. 7, 356- 362. (10.1136/bmjinnov-2021-000668) 34192022 source
Not checked in the free run.
- cites [79] Brown C, Chauhan J, Grammenos A, Han J, Hasthanasombat A, Spathis D, Xia T, Cicuta P, Mascolo C. 2020 Exploring automatic diagnosis of COVID-19 from crowdsourced respiratory sound data. (http://arxiv.org/abs/2006.05919) source
Not checked in the free run.
- cites [80] Pahar M, Klopper M, Warren R, Niesler T. 2021 COVID-19 cough classification using machine learning and global smartphone recordings. Comput. Biol. Med. 135, 104572. (10.1016/j.compbiomed.2021.104572) 34182331 PMC8213969 source
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationArtificial intelligence has made significant strides across domains, demonstrating potential to revolutionize healthcare.
“Meanwhile, artificial intelligence (AI) has made significant strides in various domains, demonstrating its potential to revolutionize healthcare.”
- Also stated in: § Introduction · boosted
“At the same time, artificial intelligence (AI) has taken a huge step forward across domains ranging from object recognition in images [5], to winning against the best players in the games of Go, Atari and chess [6], or outperforming experts in breast cancer screening [7].”
Quote in the paperNot a substantial claim0 of 3 sources checked- cites [5] He K, Zhang X, Ren S, Sun J. 2016 Deep residual learning for image recognition. In Proc. IEEE Conf. on Computer Vision and Pattern Recognition, pp. 770–778. source
Not checked in the free run.
- cites [6] Schrittwieser J et al. 2020 Mastering atari, go, chess and shogi by planning with a learned model. Nature 588, 604- 609. (10.1038/s41586-020-03051-4) 33361790 source
Not checked in the free run.
- cites [7] McKinney SM et al. 2020 International evaluation of an AI system for breast cancer screening. Nature 577, 89- 94. (10.1038/s41586-019-1799-6) 31894144 source
Not checked in the free run.
- Also stated in: § Introduction · boosted
May combine two claims.
PremisecitationA pre-trained CNN MobileNet detects respiratory disease with F1 scores of 0.912 for healthy versus unhealthy, 0.870 for obstructive versus non-obstructive, and 0.813 for COPD versus asthma.
“A pre-trained CNN based network of MobileNet is employed to detect respiratory disease, and achieves an F1 score of 0.912 on healthy versus unhealthy, 0.870 on obstructive versus non-obstructive and 0.813 on COPD versus asthma classification [74].”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [74] Xu X, Nemati E, Vatanparvar K, Nathan V, Ahmed T, Rahman MM, McCaffrey D, Kuang J, Gao JA. 2021 Listen2Cough: leveraging end-to-end deep learning cough detection model to enhance lung health assessment using passively sensed audio. Proc. ACM Interact. Mob. Wearable Ubiquitous Technol. 5, 1- 22. (10.1145/3448124) source
Not checked in the free run.
May combine two claims.
PremisecitationVoice biomarkers such as jitter, shimmer, and spectral contrast show consistent trends for recovery from acute respiratory illness and can distinguish recovering participants from healthy users.
“Voice biomarkers have also been analysed for recovery from acute respiratory illness [61]. The findings suggest that jitter, shimmer and spectral contrast within certain frequency bands show consistent trends for recovering participants, and can be informative to distinguish them from healthy users as well.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [61] Tracey B et al. 2021 Voice biomarkers of recovery from acute respiratory illness. IEEE J. Biomed. Health Inform. 26, 2787- 2795. (10.1109/JBHI.2021.3137050) 34932491 source
Not checked in the free run.
Probably not a substantial claim: may not pass S4 (based). May combine two claims.
InterpretationreasoningAutomatic detection and monitoring of mental health could provide prompt diagnosis and reduce the burden.
“Automatic detection and monitoring of mental health could provide prompt diagnosis and reduce the burden.”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S4 (based).
Premisenone givenAudio, location, and activity tracking are popular within both research and commercial applications.
“We particularly focus on these areas considering their popularity within both research and commercial applications.”
Quote in the paperNot a substantial claimMay combine two claims.
PremisecitationA CNN-based network validated using crowdsourced voice, cough, and breathing sounds yields 0.71 ROC-AUC for COVID-19 detection, suggesting potential of crowdsourced audio for COVID-19 detection.
“A convolutional neural network (CNN) based network structure is validated using a combination of crowdsourced voice, cough and breathing sounds [10,49]. The system is developed and validated using a total of 2478 participants and yields 0.71 in terms of ROC-AUC, suggesting the potential of crowdsourced audio for COVID-19 detection.”
Quote in the paperSubstantial claim0 of 2 sources checked- cites [10] Han J et al. 2021 Sounds of COVID-19: exploring realistic performance of audio-based digital testing. npj Digit. Med. 5, 16. (10.1038/s41746-021-00553-x) PMC8799654 35091662 source
Not checked in the free run.
- cites [49] Xia T et al. 2021 COVID-19 sounds: a large-scale audio dataset for digital respiratory screening. In 35th Conf. on Neural Information Processing Systems Datasets and Benchmarks Track (Round 2).
Not checked in the free run.
- Rests on an assumption: Deep learning techniques generally show superior performance over traditional ML techniques with adequate and high-quality data.
May combine two claims.
InterpretationreasoningThe advantages of audio, location, and activity tracking have led to extensive research in these modalities, suggesting their potential for future large-scale population screening and monitoring.
“These advantages have led to extensive research in these three modalities [24–26], suggesting their potential for future large-scale population screening and monitoring.”
Quote in the paperSubstantial claim0 of 3 sources checked- cites [24] Asadzadeh A, Kalankesh LR. 2021 A scope of mobile health solutions in COVID-19 pandemics. Inform. Med. Unlocked 23, 100558. (10.1016/j.imu.2021.100558) 33842688 PMC8019236 source
Not checked in the free run.
- cites [25] Bo F, Yerebakan M, Dai Y, Wang W, Li J, Hu B, Gao S. 2022 IMU-based monitoring for assistive diagnosis and management of IoHT: a review. Healthcare 10, 1210. 35885736 PMC9318359 source
Not checked in the free run.
- cites [26] Lai S, Farnham A, Ruktanonchai NW, Tatem AJ. 2019 Measuring mobility, disease connectivity and individual risk: a review of using mobile phone data and mHealth for travel medicine. J. Travel Med. 26, taz019. (10.1093/jtm/taz019) 30869148 PMC6904325 source
Not checked in the free run.
Probably not a substantial claim: may not pass S4 (based). May combine two claims.
Premisenone givenAudio sounds contain rich health-related information and different types of audio sounds reveal complementary information.
“Audio sounds contain rich health-related information and different types of audio sounds additionally reveal complementary information.”
Quote in the paperNot a substantial claimClassified as probably not a substantial claim.
Premisenone givenAudio, location, and activity tracking rely on sensors commonly found in the majority of mobile and wearable devices.
“Additionally, these three modalities rely on sensors commonly found in the majority of mobile and wearable devices, unlike biosignal collection devices such as ECG chest straps or EEG headsets.”
Quote in the paperSubstantial claimMay combine two claims.
PremisecitationSignificant acoustic emissions occur at similar joint angles for repeated exercises, suggesting reliability of audio sounds for knee activities.
“The significant acoustic emissions were found to occur at similar joint angles for repeated exercises, suggesting the reliability of the audio sounds for knee activities.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [65] Teague CN, Heller JA, Nevius BN, Carek AM, Mabrouk S, Garcia-Vicente F, Inan OT, Etemadi M. 2020 A wearable, multimodal sensing system to monitor knee joint health. IEEE Sensors J. 20, 10 323- 10 334. (10.1109/JSEN.2020.2994552) source
Not checked in the free run.
May combine two claims.
PremisecitationHeart rate and breathing rate are estimated from in-ear microphones, demonstrating potential of earables for health monitoring.
“Heart rate and breathing rate are also estimated from in-ear microphones, and demonstrated the potential of earables for health monitoring [44,84].”
Quote in the paperNot a substantial claim0 of 2 sources checked- cites [44] Chen G, Imtiaz SA, Aguilar-Pelaez E, Rodriguez-Villegas E. 2015 Algorithm for heart rate extraction in a novel wearable acoustic sensor. Healthcare Technol. Lett. 2, 28- 33. (10.1049/htl.2014.0095) PMC4613720 26609401 source
Not checked in the free run.
- cites [84] Martin A, Voix J. 2017 In-ear audio wearable: measurement of heart and breathing rates for health and safety monitoring. IEEE Trans. Biomed. Eng. 65, 1256- 1263. (10.1109/TBME.2017.2720463) 28678697 source
Not checked in the free run.
May combine two claims.
InterpretationcitationThe general conclusion of the clinical relevance of audio sounds to health and the distinguishable information between healthy and non-healthy conditions holds.
“Slightly different conclusions might be made for different studies due to the data variability and data size, but the general conclusion of the clinical relevance of audio sounds to health and the distinguishable information between still holds.”
Quote in the paperSubstantial claim- Rests on an assumption: Lung disease alters the sound generation mechanism.
May combine two claims.
PremisecitationBreathing and cough sounds contain clinically relevant information and are highly related to pulmonary health.
“Breathing and cough are two types of sounds that contain clinically relevant information and highly related to pulmonary health [10,43].”
Quote in the paperNot a substantial claim0 of 2 sources checked- cites [10] Han J et al. 2021 Sounds of COVID-19: exploring realistic performance of audio-based digital testing. npj Digit. Med. 5, 16. (10.1038/s41746-021-00553-x) PMC8799654 35091662 source
Not checked in the free run.
- cites [43] Amrulloh Y, Abeyratne U, Swarnkar V, Triasih R. 2015 Cough sound analysis for pneumonia and asthma classification in pediatric population. In 2015 6th Int. Conf. on Intelligent Systems, Modelling and Simulation, pp. 127–131. IEEE. 24110049 source
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationBreathing rate via smartphones for sleep monitoring detects sleep events including snore, cough, turn over, and get up.
“another study explored the potential of breathing rate by leveraging smartphones for sleep monitoring, which detects sleep events including snore, cough, turn over and get up [30].”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [30] Ren Y, Wang C, Yang J, Chen Y. 2015 Fine-grained sleep monitoring: hearing your breathing with smartphones. In 2015 IEEE Conf. on Computer Communications (INFOCOM), pp. 1194–1202. IEEE. source
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing).
Premisenone givenPhysical health studies using audio were mainly carried out for respiratory disease and sleep health.
“Physical health studies were mainly carried out for respiratory disease and sleep health.”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S4 (based).
Premisenone givenLeveraging mobile sensing datasets has lagged behind other areas of machine learning.
“Despite the recognition of the value of mobile sensing, leveraging these datasets has lagged behind other areas of ML.”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationBone conducted sounds collected via wearables have been analysed for health applications.
“In addition to the normal airborne sound, bone conducted sounds collected via wearables have also been analysed for health application, which is generally collected via earables or sensors placed near neck.”
Quote in the paperNot a substantial claim0 of 3 sources checked- cites [44] Chen G, Imtiaz SA, Aguilar-Pelaez E, Rodriguez-Villegas E. 2015 Algorithm for heart rate extraction in a novel wearable acoustic sensor. Healthcare Technol. Lett. 2, 28- 33. (10.1049/htl.2014.0095) PMC4613720 26609401 source
Not checked in the free run.
- cites [45] Kusche R, Klimach P, Malhotra A, Kaufmann S, Ryschka M. 2015 An in-ear pulse wave velocity measurement system using heart sounds as time reference. Curr. Dir. Biomed. Eng. 1, 366- 370. (10.1515/cdbme-2015-0090) source
Not checked in the free run.
- cites [47] Amft O, Stäger M, Lukowicz P, Tröster G. 2005 Analysis of chewing sounds for dietary monitoring. In UbiComp 2005: ubiquitous computing (eds M Beigl, S Intille, J Rekimoto, H Tokuda), pp. 56- 72. Berlin, Germany: Springer. (10.1007/11551201_4) source
Not checked in the free run.
Classified as probably not a substantial claim.
Premisenone givenAdvances in wearable sensing and mobile computing have enabled the collection of health and well-being data outside traditional laboratory and hospital settings.
“Advances in wearable sensing and mobile computing have enabled the collection of health and well-being data outside of traditional laboratory and hospital settings, paving the way for a new era of mobile health.”
- Also stated in: § Introduction · plain
“recent advances in wearable sensing and mobile computing, along with their wide adoption, have created new pathways for the collection of health and well-being data outside of laboratory and hospital settings, in a longitudinal fashion.”
Quote in the paperSubstantial claim- Also stated in: § Introduction · plain
Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.
Premisenone givenCheck-in data record visits, have low temporal granularity and less scaling power to a large population, but can capture longitudinal lifestyle with limited data size.
“Check-in data record when a person logs visits to venues, for example posts in social media, such as Twitter, Foursquare, Facebook, etc. Check-ins have low temporal granularity and less scaling power to a large population. However, they can capture longitudinal lifestyle with limited data size.”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationContact tracing in the context of disease spreading has long been studied.
“Also contact tracing in the context of disease spreading has long been studied [110].”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [110] Cebrian M. 2021 The past, present and future of digital contact tracing. Nat. Electron. 4, 2- 4. (10.1038/s41928-020-00535-z) source
Not checked in the free run.
Probably not a substantial claim: may not pass S4 (based).
Premisenone givenSeveral health problems correlate with mobility behaviour changes.
“Several health problems correlate with mobility behaviour changes.”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing), S4 (based).
Premisenone givenCollocation information is often more important than the location itself, for example, in the study of infectious disease spread.
“Collocation information is often more important than the location itself, for example, in the study of infectious disease spread.”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing), S4 (based).
Premisenone givenLocalization sensors are ubiquitous in mobile phones and wearable devices.
“Localization sensors are ubiquitous in mobile phones and wearable devices.”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.
Premisenone givenGPS consumes a lot of battery power and its continuous precise recordings are often privacy invasive.
“Moreover, it consumes a lot of battery power and its continuous precise recordings are often privacy invasive.”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
Premisenone givenSimple features characterizing individual trips or the spatial spread of movements are commonly used in the literature.
“Simple features characterizing individual trips or the spatial spread of movements are commonly used in the literature (table 2).”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationMobility has been shown to play a major role in spreading other diseases like malaria-people travelling to different cities carry the parasite and infect the mosquitoes there and in turn the population.
“Apart from the above infectious diseases, mobility has been shown to play a major role in spreading other diseases like malaria [111]—people travelling to different cities carry the parasite and infect the mosquitoes there and in turn the population.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [111] Wesolowski A, Eagle N, Tatem AJ, Smith DL, Noor AM, Snow RW, Buckee CO. 2012 Quantifying the impact of human mobility on malaria. Science 338, 267- 270. (10.1126/science.1223467) 23066082 PMC3675794 source
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationTracking locations and analysing mobility has contributed to many day-to-day applications such as traffic management, routing, recommender systems, and cellular network load balancing.
“Tracking locations and analysing mobility has contributed to many day-to-day applications, for example, traffic management and forecasting [87,88], routing individuals [89] and delivery packages [90], recommender systems [91], dynamic load balancing systems for cellular networks [92], to name a few.”
Quote in the paperNot a substantial claim0 of 4 sources checked- cites [87] Wang X, Ma Y, Wang Y, Jin W, Wang X, Tang J, Jia C, Yu J. 2020 Traffic flow prediction via spatial temporal graph neural network. In Proc. Web Conf. 2020, pp. 1082–1092. source
Not checked in the free run.
- cites [88] Liang Y, Ouyang K, Sun J, Wang Y, Zhang J, Zheng Y, Rosenblum D, Zimmermann R. 2021 Fine-grained urban flow prediction. In Proc. Web Conf. 2021, pp. 1833–1845. source
Not checked in the free run.
- cites [89] Ben Ticha H, Absi N, Feillet D, Quilliot A. 2018 Vehicle routing problems with road-network information: state of the art. Networks 72, 393- 406. (10.1002/net.21808) source
Not checked in the free run.
- cites [91] Noulas A, Scellato S, Lathia N, Mascolo C. 2012 A random walk around the city: new venue recommendation in location-based social networks. In 2012 Int. Conf. on Privacy, Security, Risk and Trust and 2012 Int. Conf. on Social Computing, pp. 144–153. IEEE. source
Not checked in the free run.
Probably not a substantial claim: may not pass S2 (checkable). May combine two claims.
Premisenone givenComplex spatio-temporal mobility data are hard to describe in a succinct unique way and different applications need to take varied perspectives.
“This is primarily because complex spatio-temporal mobility data are hard to describe in a succinct unique way and different applications need to take varied perspectives.”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.
Premisenone givenGPS localization is highly accurate within a few metres but requires clear line of sight to at least three satellites; accuracy reduces significantly in regions with obstruction.
“While GPS localization is highly accurate (within a few metres), it requires clear line of sight to at least three satellites all the time; its accuracy reduces significantly in regions with obstruction.”
Quote in the paperNot a substantial claimClassified as probably not a substantial claim. May combine two claims.
PremisecitationTo reduce disease spread and retain social activities, reopening strategies need to consider the mobility of the people, for example, protect the most mobile people or restricting large gatherings.
“The primary conclusions are that to reduce the disease spread and retain social activities, reopening strategies need to consider the mobility of the people, for example, protect the most mobile people or restricting large gatherings.”
Quote in the paperSubstantial claim0 of 2 sources checked- cites [106] Chang S, Pierson E, Koh PW, Gerardin J, Redbird B, Grusky D, Leskovec J. 2021 Mobility network models of COVID-19 explain inequities and inform reopening. Nature 589, 82- 87. (10.1038/s41586-020-2923-3) 33171481 source
Not checked in the free run.
- cites [107] Wang H, Ghosh A, Ding J, Sarkar R, Gao J. 2021 Heterogeneous interventions reduce the spread of COVID-19 in simulations on real mobility data. Sci. Rep. 11, 7809. (10.1038/s41598-021-87034-z) 33833298 PMC8034422 source
Not checked in the free run.
May combine two claims.
PremisecitationSpatial navigation patterns are affected by dementia and are considered as one of the earliest symptoms.
“Spatial navigation patterns are affected by dementia and are considered as one of the earliest symptoms [112].”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [112] Coughlan G, Laczó J, Hort J, Minihane AM, Hornberger M. 2018 Spatial navigation deficits—overlooked cognitive marker for preclinical Alzheimer disease? Nat. Rev. Neurol. 14, 496- 506. (10.1038/s41582-018-0031-x) 29980763 source
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).
Premisenone givenWi-Fi access points offer similar localization service with increased spatial and temporal resolution.
“Wi-Fi access points also can offer similar localization service with increased spatial and temporal resolution.”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationMobility restrictions had been used and studied in the past, for example to stop the spread of Ebola in Sierra Leone and H1N1 influenza.
“While in the above we reviewed COVID-19 related studies for their timely relevance, mobility restrictions had been used and studied in the past, for example to stop the spread of Ebola in Sierra Leone [108], H1N1 influenza [109], etc.”
Quote in the paperNot a substantial claim0 of 2 sources checked- cites [108] Peak CM et al. 2018 Population mobility reductions associated with travel restrictions during the Ebola epidemic in Sierra Leone: use of mobile phone data. Int. J. Epidemiol. 47, 1562- 1570. (10.1093/ije/dyy095) 29947788 PMC6208277 source
Not checked in the free run.
- cites [109] Bajardi P, Poletto C, Ramasco JJ, Tizzoni M, Colizza V, Vespignani A. 2011 Human mobility networks, travel restrictions, and the global spread of 2009 H1N1 pandemic. PLoS ONE 6, e16591. (10.1371/journal.pone.0016591) 21304943 PMC3031602 source
Not checked in the free run.
May combine two claims.
PremisecitationHealthwalks framework generates combinations of features using a deterministic automaton and its effectiveness is shown using an eight class disease classification.
“Healthwalks [94] builds such a framework that takes user location traces, associated points of interest and user social demographics as input and automatically generates combinations of features using a deterministic automaton (DFA). The effectiveness of the generated features is shown using an eight class disease classification.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [94] Lin Z, Lyu S, Cao H, Xu F, Wei Y, Samet H, Li Y. 2020 Healthwalks: sensing fine-grained individual health condition via mobility data. Proc. ACM Interact. Mob. Wearable Ubiquitous Technol. 4, 1- 26. (10.1145/3432229)35846237 source
Not checked in the free run.
- Rests on an assumption: Health applications typically benefit from interpretable features to build explainable models.
Probably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationThe effect of mobility for transmitting infectious diseases has long been studied, for example in Spanish flu.
“The effect of mobility for transmitting such diseases has long been studied, for example in Spanish flu [99].”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [99] Coronavirus: how they tried to curb Spanish flu pandemic in 1918. See https://www.bbc.co.uk/news/in-pictures-52564371 (accessed 21 January 2023).
Not checked in the free run.
May combine two claims.
PremisereasoningCDR is suited for outdoor localization; Wi-Fi localization is suited for indoor localization.
“While CDR is suited for outdoor localization for its wide adaptation, Wi-Fi localization is suited for indoor localization for its higher resolution and availability in indoor spaces.”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing), S4 (based).
InterpretationreasoningThis method of diagnosis is inherently less costly to deploy at large scale than traditional survey and clinical test based methods simply due to pervasive presence of mobile phones and wearable devices in the population.
“This method of diagnosis is inherently less costly to deploy at large scale than traditional survey and clinical test based methods simply due to pervasive presence of mobile phones and wearable devices in the population.”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationRecently researchers have been using tools from algebraic topology and geometry to map trajectories to points in Euclidean space to make them amenable to ML tools.
“While the above features are useful in many applications, recently researchers have been using tools from algebraic topology [95] and geometry [96] to map trajectories to points in Euclidean space to make them amenable to ML tools.”
Quote in the paperNot a substantial claim0 of 2 sources checked- cites [95] Ghosh A, Rozemberczki B, Ramamoorthy S, Sarkar R. 2018 Topological signatures for fast mobility analysis. In Proc. 26th ACM SIGSPATIAL Int. Conf. on Advances in Geographic Information Systems, pp. 159–168. source
Not checked in the free run.
- cites [96] Phillips JM, Pourmahmood-Aghababa H. 2020 Orientation-preserving vectorized distance between curves. (http://arxiv.org/abs/2007.15924)
Not checked in the free run.
Classified as probably not a substantial claim.
PremisecitationThe study exploits the causal relationship between chronic diseases and daily habits captured by check-ins.
“The study exploits the causal relationship between chronic diseases and daily habits captured by check-ins.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [118] Wang Y, Zhou X, Mascolo C, Noulas A, Xie X, Liu Q. 2018 Predicting the spatio-temporal evolution of chronic diseases in population with human mobility data. In Proc. 27th Int. Joint Conf. on Artificial Intelligence, pp. 3578–3584. (10.24963/ijcai.2018/497) source
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationProgress is being made for using trajectories in ML algorithms.
“Further, progress is being made for using trajectories in the ML algorithms [97,98].”
Quote in the paperNot a substantial claim0 of 2 sources checked- cites [97] Takeuchi K, Imaizumi M, Kanda S, Tabei Y, Fujii K, Yoda K, Ishihata M, Maekawa T. 2021 Fréchet kernel for trajectory data analysis. In Proc. 29th Int. Conf. on Advances in Geographic Information Systems, pp. 221–224. source
Not checked in the free run.
- cites [98] Buchin K, Driemel A, Gudmundsson J, Horton M, Kostitsyna I, Löffler M, Struijs M. 2019 Approximating (k, ℓ)-center clustering for curves. In Proc. 30th Annual ACM-SIAM Symp. on Discrete Algorithms, pp. 2922–2938. Philadelphia, PA: SIAM. source
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.
PremisecitationMobility restrictions are effective in flattening the curve but immediately affected the economy and societal functions.
“Although mobility restrictions are effective in flattening the curve they immediately affected the economy and societal functions [103].”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [103] Bonaccorsi G et al. 2020 Economic and social consequences of human mobility restrictions under COVID-19. Proc. Natl Acad. Sci. USA 117, 15 530- 15 535. (10.1073/pnas.2007658117) PMC7355033 32554604 source
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.
PremisereasoningCDR records the location of the base station, has coarse temporal resolution, accuracy of the order of hundreds of metres, and is less privacy invasive.
“Call data record (CDR) records the location of the base station (the cell tower a phone is connected to). It has coarse temporal resolution as the tracking generally takes place when a person uses call, messaging or data service. As a cell tower usually serves a large area its accuracy is of the order of hundreds of metres. Its coarse spatial and temporal resolution makes it less privacy invasive.”
Quote in the paperNot a substantial claimMay combine two claims.
PremisecitationWearable devices, such as activity trackers and smartwatches, increasingly provide opportunities for non-intrusive objective monitoring of biological signals such as HR and movement during free-living, potentially enabling more precise prediction of VO2max without the need to conduct a specific exercise test.
“Wearable devices, such as activity trackers and smartwatches, increasingly provide opportunities for non-intrusive objective monitoring of biological signals such as HR and movement during free-living, potentially enabling more precise prediction of VO2max without the need to conduct a specific exercise test [140].”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [140] Plasqui G, Westerterp KR. 2005 Accelerometry and heart rate as a measure of physical fitness: proof of concept. Med. Sci. Sports Exerc. 37, 872- 876. (10.1249/01.MSS.0000161805.61893.C0) 15870644 source
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationIn the context of health signals, self-supervised tasks can range from predicting the arrow of time, to contrastive learning and masking.
“In the context of health signals, these tasks can range from predicting the arrow of time, to contrastive learning and masking [145,146].”
Quote in the paperNot a substantial claim0 of 2 sources checked- cites [145] Spathis D, Perez-Pozuelo I, Marques-Fernandez L, Mascolo C. 2022 Breaking away from labels: the promise of self-supervised machine learning in intelligent health. Patterns 3, 100410. (10.1016/j.patter.2021.100410) 35199063 PMC8848012 source
Not checked in the free run.
- cites [146] Yuan H, Chan S, Creagh AP, Tong C, Clifton DA, Doherty A. 2022 Self-supervised learning for human activity recognition using 700,000 person-days of wearable data. (http://arxiv.org/abs/2206.02909) PMC11015005 38609437
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationStudies of over 1 million participants showed that power users demonstrated the greatest weight loss.
“Weight loss was the subject of other studies of over 1 million participants [129], which showed that power users demonstrated the greatest weight loss.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [129] Serrano KJ, Coa KI, Yu M, Wolff-Hughes DL, Atienza AA. 2017 Characterizing user engagement with health app data: a data mining approach. Transl. Behav. Med. 7, 277- 285. (10.1007/s13142-017-0508-y) 28616846 PMC5526821 source
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationRecent advances in deep learning architectures for sequential modelling based upon wearable and mobile sensing have been used for health predictions and recommendations.
“Recent advances in deep learning architectures for sequential modelling based upon wearable and mobile sensing have been used for health predictions and recommendations [141].”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [141] Ballinger B et al. 2018 Deepheart: semi-supervised sequence learning for cardiovascular risk prediction. (http://arxiv.org/abs/1802.02511) source
Not checked in the free run.
Probably not a substantial claim: may not pass S2 (checkable).
InterpretationreasoningObtaining these outcomes in large populations can be valuable for downstream health-related inferences that would normally be costly and burdensome.
“Obtaining these outcomes in large populations can be valuable for downstream health-related inferences that would normally be costly and burdensome.”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.
PremisecitationAn analysis of exercise patterns in a global social network of 1.1 million runners demonstrated that exercise is 'contagious', whose effect depends on gender and relative levels of activity.
“Another analysis of exercise patterns in a global social network of 1.1 million runners demonstrated that exercise is ‘contagious’, whose effect depends on gender and relative levels of activity [127].”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [127] Aral S, Nicolaides C. 2017 Exercise contagion in a global social network. Nat. Commun. 8, 14753. (10.1038/ncomms14753) 28418379 PMC5399289 source
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationRecent work using self-supervised learning has shown promise in the same data modalities such as ECG data.
“Recent work using self-supervised learning has shown promise in the same data modalities such as ECG data [142,143].”
Quote in the paperNot a substantial claim0 of 2 sources checked- cites [142] Sarkar P, Etemad A. 2019 Self-supervised learning for ECG-based emotion recognition. (http://arxiv.org/abs/1910.07497) source
Not checked in the free run.
- cites [143] Kiyasseh D, Zhu T, Clifton DA. 2020 Clocs: contrastive learning of cardiac signals. (http://arxiv.org/abs/2005.13249)
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.
InterpretationcitationThe model's latent space can also be used for fitness-aware patient subtyping paving the way to scalable interventions and personalized trial recruitment.
“Notably, the model’s latent space can also be used for fitness-aware patient subtyping paving the way to scalable interventions and personalized trial recruitment.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [136] Spathis D, Perez-Pozuelo I, Gonzales TI, Wu Y, Brage S, Wareham N, Mascolo C. 2022 Longitudinal cardio-respiratory fitness prediction through wearables in free-living environments. npj Digit. Med. 5, 176. (10.1038/s41746-022-00719-1) 36460766 PMC9718831 source
Not checked in the free run.
May combine two claims.
PremisecitationThe model maps activity data to future HR responses and can be used as a feature extractor for wearable data.
“The model maps activity data to future HR responses and can be used as a feature extractor for wearable data.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [144] Spathis D, Perez-Pozuelo I, Brage S, Wareham NJ, Mascolo C. 2021 Self-supervised transfer learning of physiological representations from free-living wearable data. In Proc. Conf. on Health, Inference, and Learning, CHIL ’21, pp. 69–78. New York, NY: ACM. source
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.
PremisecitationTen million users of a weight monitoring app were used to show that people are more likely to lose weight when they had more friends of the opposite sex.
“Ten million users of a weight monitoring app were used to show that people are more likely to lose weight when they had more friends of the opposite sex [128].”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [128] Wang Z, Derr T, Yin D, Tang J. 2017 Understanding and predicting weight loss with mobile social networking data. In Proc. 2017 ACM on Conf. on Information and Knowledge Management, pp. 1269–1278. source
Not checked in the free run.
May combine two claims.
PremisecitationBy extracting RSSI and phase profiles as fingerprints, the system estimates distance to obstacles without observations of eyes and thus guides users to move safely.
“By extracting the RSSI and phase profiles as fingerprints it estimates the distance to obstacles without observations of eyes and thus guiding the users to move in the domain safely.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [126] Xiao F, Miao Q, Xie X, Sun L, Wang R. 2018 Indoor anti-collision alarm system based on wearable Internet of Things for smart healthcare. IEEE Commun. Mag. 56, 53- 59. (10.1109/MCOM.2018.1700706) source
Not checked in the free run.
May combine two claims.
PremisecitationPoor physical access to health facilities causes reduced uptake of preventive health services and is likely to affect low-income households the most.
“Poor physical access to health facilities causes reduced uptake of preventive health services and is likely to affect low-income households the most.”
Quote in the paperNot a substantial claim- Rests on an assumption: Poor physical access to health facilities causes reduced uptake of preventive health services and is likely to affect low-income households the most.
Probably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationAn app was developed for elderly people to locate an accessible and adequate healthcare centre that can attend to their needs.
“An app was developed [122] for elderly people to locate an accessible and adequate healthcare centre that can attend to their needs.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [122] Huh JH, Kim TJ. 2019 A location-based mobile health care facility search system for senior citizens. J. Supercomput. 75, 1831- 1848. (10.1007/s11227-018-2342-5) source
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).
InterpretationreasoningSince wearables are used in daily, unconstrained environments, activities such as drinking coffee or alcohol, as well as stress, may confound simple heuristics.
“However, since wearables are used in daily, unconstrained environments, activities such as drinking coffee or alcohol, as well as stress, may confound simple heuristics.”
Quote in the paperNot a substantial claimMay combine two claims.
PremisecitationMobile and wearable sensors allow for continuous and ubiquitous monitoring of an individual's physical activity profiles, which, when combined with cardio-respiratory information, provide valuable insights into that individual's health and fitness status.
“Mobile and wearable sensors allow for continuous and ubiquitous monitoring of an individual’s physical activity profiles, which, when combined with cardio-respiratory information, provide valuable insights into that individual’s health and fitness status [131].”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [131] Mandsager K, Harb S, Cremer P, Phelan D, Nissen SE, Jaber W. 2018 Association of cardiorespiratory fitness with long-term mortality among adults undergoing exercise treadmill testing. JAMA Netw. Open 1, e183605. (10.1001/jamanetworkopen.2018.3605) 30646252 PMC6324439 source
Not checked in the free run.
May combine two claims.
PremisecitationThe learned embeddings showed strong performance at inferring physiologically meaningful variables ranging from anthropometrics to fitness, outperforming unimodal autoencoders and common biomarkers.
“The value captured by the learned embeddings was showcased through strong performance at inferring physiologically meaningful variables ranging from anthropometrics to fitness, outperforming unimodal autoencoders and common biomarkers.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [144] Spathis D, Perez-Pozuelo I, Brage S, Wareham NJ, Mascolo C. 2021 Self-supervised transfer learning of physiological representations from free-living wearable data. In Proc. Conf. on Health, Inference, and Learning, CHIL ’21, pp. 69–78. New York, NY: ACM. source
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.
PremisecitationIndoor positioning is being used in hospitals to track medical devices and staffs for better resource management, enhancing patient experience, and repair and maintenance management.
“Indoor positioning is being used in hospitals to track medical devices and staffs for better resource management, enhancing patient experience, and repair and maintenance management [124,125].”
Quote in the paperNot a substantial claim0 of 2 sources checked- cites [124] Indoor positioning use cases for healthcare institutions. See https://www.esri.com/arcgis-blog/products/arcgis-indoors/health/indoor-positioning-use-cases-for-healthcare/ (accessed 26 January 2023).
Not checked in the free run.
- cites [125] Hospital RTLS. See https://indoo.rs/industries/healthcare/ (accessed 26 January 2023).
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationGiven rapid growth in adoption of Internet-enabled wearable devices, sensor time-series comprise a considerable amount of user-generated data.
“Given the rapid growth in adoption of Internet-enabled wearable devices, sensor time-series comprise a considerable amount of user-generated data [137].”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [137] Blalock DW, Guttag JV. 2016 Extract: strong examples from weakly-labeled sensor data. In 2016 IEEE 16th Int. Conf. Data Mining (ICDM), pp. 799–804. IEEE. source
Not checked in the free run.
Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based).
Interpretationnone givenThe emergence of large foundation AI models including large language models (LLMs) holds a promise to revolutionize this area.
“The emergence of large foundation AI models including large language models (LLMs) holds a promise to revolutionize this area.”
Quote in the paperNot a substantial claimMay combine two claims.
PremisecitationLarge scale studies of physical activity and well-being leveraging mobile devices' built-in accelerometers have shown promise as global physical activity surveillance tools, demonstrating inequalities across different countries after analysing data from over 700 000 people.
“Large scale studies of physical activity and well-being leveraging mobile devices’ built-in accelerometers have shown promise as global physical activity surveillance tools, demonstrating, for example, inequalities across different countries after analysing data from over 700 000 people [2].”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [2] Althoff T, Sosic R, Hicks JL, King AC, Delp SL, Leskovec J. 2017 Large-scale physical activity data reveal worldwide activity inequality. Nature 547, 336- 339. (10.1038/nature23018) 28693034 PMC5774986 source
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationFor 50 000 people, lower overall activity but more frequent transitions between active and inactive periods was associated with similar CVD to higher overall activity but with fewer transitions.
“The relationship between physical activity and cardiovascular disease (CVD) was studied for 50 000 people in [130], finding that lower overall activity but more frequent transitions between active and inactive periods was associated with similar CVD to higher overall activity but with fewer transitions.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [130] McConnell MV et al. 2017 Feasibility of obtaining measures of lifestyle from a smartphone app: the MyHeart Counts Cardiovascular Health Study. JAMA Cardiol. 2, 67- 76. (10.1001/jamacardio.2016.4395) 27973671 source
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.
PremisecitationGreater home walkability was associated with increased BMI, as was greater school access to green space.
“Greater home walkability was associated with increased BMI, as was greater school access to green space.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [123] Burgoine T, Jones AP, Brouwer RJN, Neelon SEB. 2015 Associations between BMI and home, school and route environmental exposures estimated using GPS and GIS: do we see evidence of selective daily mobility bias in children? Int. J. Health Geogr. 14, 8. (10.1186/1476-072X-14-8) 25656299 PMC4429367 source
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationGreater school walkability was associated with significantly lower mean BMI.
“Burgoine et al. [123] tracked school children for a week with GPS devices and found that greater school walkability was associated with significantly lower mean BMI.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [123] Burgoine T, Jones AP, Brouwer RJN, Neelon SEB. 2015 Associations between BMI and home, school and route environmental exposures estimated using GPS and GIS: do we see evidence of selective daily mobility bias in children? Int. J. Health Geogr. 14, 8. (10.1186/1476-072X-14-8) 25656299 PMC4429367 source
Not checked in the free run.
May combine two claims.
PremisecitationThe embeddings achieved an AUC of 0.68 for cardio-fitness prediction and an AUC of 0.80 for physical activity energy expenditure.
“For instance, the embeddings achieved an AUC of 0.68 for cardio-fitness prediction and an AUC of 0.80 for physical activity energy expenditure.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [144] Spathis D, Perez-Pozuelo I, Brage S, Wareham NJ, Mascolo C. 2021 Self-supervised transfer learning of physiological representations from free-living wearable data. In Proc. Conf. on Health, Inference, and Learning, CHIL ’21, pp. 69–78. New York, NY: ACM. source
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationThe Apple Watch has introduced cardio-respiratory fitness (CRF) features.
“More recently, the Apple Watch has introduced cardio-respiratory fitness (CRF) features [135] while large population-scale studies showed that these HR responses can predict VO2max [136].”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [135] Apple. 2021 Using Apple Watch to estimate cardio fitness with VO2 max. See https://www.apple.com/healthcare/docs/site/Using_Apple_Watch_to_Estimate_Cardio_Fitness_with_VO2_max.pdf.
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationMost wearable devices, particularly those that are wrist-worn, incorporate accelerometry sensors, which are very affordable tools to objectively study physical activity patterns.
“Most wearable devices, particularly those that are wrist-worn, incorporate accelerometry sensors, which are very affordable tools to objectively study physical activity patterns [17].”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [17] Doherty A et al. 2017 Large scale population assessment of physical activity using wrist worn accelerometers: the UK Biobank study. PLoS ONE 12, e0169649. (10.1371/journal.pone.0169649) 28146576 PMC5287488 source
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationPPG sensors have been widely employed on smartwatches, such as the Apple Watch, to monitor cardiovascular health.
“Due to their non-invasive and easy-to-carry nature, PPG sensors have been widely employed on smartwatches, such as the Apple Watch, to monitor cardiovascular health [169].”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [169] Falter M, Budts W, Goetschalckx K, Cornelissen V, Buys R. 2019 Accuracy of Apple Watch measurements for heart rate and energy expenditure in patients with cardiovascular disease: cross-sectional study. JMIR mHealth uHealth 7, e11889. (10.2196/11889) 30888332 PMC6444219 source
Not checked in the free run.
May combine two claims.
PremisecitationDespite some promising studies which attempt to infer VO2max from data collected during free-living conditions, these mostly stem from small-scale cohorts with less than 50 participants and use contextual data from treadmill activity, which again limits their application in real-world contexts.
“Despite some promising studies which attempt to infer VO2max from data collected during free-living conditions, these mostly stem from small-scale cohorts with less than 50 participants and use contextual data from treadmill activity, which again limits their application in real-world contexts [150].”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [150] Altini M, Casale P, Penders J, Amft O. 2016 Cardiorespiratory fitness estimation in free-living using wearable sensors. Artif. Intell. Med. 68, 37- 46. (10.1016/j.artmed.2016.02.002) 26948954 source
Not checked in the free run.
May combine two claims.
PremisecitationDue to false negatives in detecting face-to-face contacts, digital contact tracing may not be as effective as its manual counterpart-a study estimated that digital tracing alone can reduce the R number by 44% whereas manual tracing of all contacts can reduce R number by 61%.
“However, due to false negatives in detecting face-to-face contacts, digital contact tracing may not be as effective as its manual counterpart—a study [178] estimated that digital tracing alone can reduce the R number (average number of people infected by a single infectious person) by 44% whereas manual tracing of all contacts can reduce R number by 61%.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [178] Kucharski AJ et al. 2020 Effectiveness of isolation, testing, contact tracing, and physical distancing on reducing transmission of SARS-CoV-2 in different settings: a mathematical modelling study. Lancet Infect. Dis. 20, 1151- 1160. (10.1016/S1473-3099(20)30457-6) 32559451 PMC7511527 source
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing).
PremisereasoningSince neutral mood might be uninformative and make the predictions harder, authors often omit the middle-scoring 40-60% of reports.
“Since neutral mood might be uninformative and make the predictions harder, authors often omit the middle-scoring 40–60% of reports.”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationMany mobile apps have been developed using Bluetooth and Wi-Fi technology to understand outbreak epidemiology, individual screening and contact tracing.
“Many mobile apps have been developed using such technology to understand outbreak epidemiology, individual screening and contact tracing [172–174].”
Quote in the paperNot a substantial claim0 of 3 sources checked- cites [172] Pandit JA, Radin JM, Quer G, Topol EJ. 2022 Smartphone apps in the COVID-19 pandemic. Nat. Biotechnol. 40, 1013- 1022. (10.1038/s41587-022-01350-x) 35726090 PMC13501915 source
Not checked in the free run.
- cites [173] Toussaert S. 2021 Upping uptake of COVID contact tracing apps. Nat. Hum. Behav. 5, 183- 184. (10.1038/s41562-021-01048-1) 33479504 source
Not checked in the free run.
- cites [174] Salathé M. 2023 COVID-19 digital contact tracing worked—heed the lessons for future pandemics. Nature 619, 31- 33. (10.1038/d41586-023-02130-6) 37400650 source
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationSome similar apps in different countries have also been employed for COVID-19 tracing.
“Some similar apps in different countries have also been employed for COVID-19 tracing [176,177].”
Quote in the paperNot a substantial claim0 of 2 sources checked- cites [176] Abbas HSM, Xu X, Sun C. 2021 Role of COVIDsafe app and control measures in Australia in combating COVID-19 pandemic. Transform. Govern. People Process Policy 15, 708- 719. (10.1108/TG-01-2021-0004)
Not checked in the free run.
- cites [177] Lalmuanawma S, Hussain J, Chhakchhuak L. 2020 Applications of machine learning and artificial intelligence for Covid-19 (SARS-CoV-2) pandemic: a review. Chaos Solit. Fractals 139, 110059. (10.1016/j.chaos.2020.110059) PMC7315944 32834612 source
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.
PremisecitationRecent work explored fine-grained mood prediction through regression and clustering, as well as novel multi-dimensional formulations of the Affect Grid through multi-task learning.
“Instead, recent work explored fine-grained mood prediction through regression and clustering [163], as well as novel multi-dimensional formulations of the Affect Grid through multi-task learning [164].”
Quote in the paperNot a substantial claim0 of 2 sources checked- cites [163] Spathis D, Servia-Rodriguez S, Farrahi K, Mascolo C, Rentfrow J. 2019 Passive mobile sensing and psychological traits for large scale mood prediction. In Proc. 13th EAI Int. Conf. on Pervasive Computing Technologies for Healthcare, pp. 272–281. source
Not checked in the free run.
- cites [164] Spathis D, Servia-Rodriguez S, Farrahi K, Mascolo C, Rentfrow J. 2019 Sequence multi-task learning to forecast mental wellbeing from sparse self-reported data. In Proc. 25th ACM SIGKDD Int. Conf. on Knowledge Discovery & Data Mining, pp. 2886–2894. source
Not checked in the free run.
May combine two claims.
InterpretationreasoningGiven logistical constraints and inherent risk of maximal exercise testing, scalability of fitness assessment in large populations has been limited, meaning relatively little is known about population levels of fitness, or their possible changes over time.
“Given these logistical constraints and the inherent risk of maximal exercise testing, scalability of fitness assessment in large populations has been limited, meaning relatively little is known about population levels of fitness, or their possible changes over time.”
Quote in the paperSubstantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationRF signal provides non-contact sleep monitoring solutions.
“It provides non-contact sleep monitoring solutions.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [183] Yue S, Yang Y, Wang H, Rahul H, Katabi D. 2020 BodyCompass: monitoring sleep posture with wireless signals. Proc. ACM Interact. Mob. Wearable Ubiquitous Technol. 4, 1- 25. (10.1145/3397311)35846237 source
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationSome recent works employ recurrent neural networks, feed-forward layers, multi-task learning and autoencoders to fill in missing sensor data, or to learn better representations.
“Some recent works employ recurrent neural networks [20], feed-forward layers [159], multi-task learning [160] and autoencoders to fill in missing sensor data [161], or to learn better representations [162].”
Quote in the paperNot a substantial claim0 of 3 sources checked- cites [20] Suhara Y, Xu Y, Pentland A. 2017 Deepmood: forecasting depressed mood based on self-reported histories via recurrent neural networks. In Proc. 26th Int. Conf. on World Wide Web, pp. 715–724. source
Not checked in the free run.
- cites [159] Mikelsons G, Smith M, Mehrotra A, Musolesi M. 2017 Towards deep learning models for psychological state prediction using smartphone data: challenges and opportunities. In Workshop on Machine Learning for Health (ML4H) at 31st Conf. on Neural Information Processing Systems, Long Beach, CA, USA, 4–9 December 2017.
Not checked in the free run.
- cites [160] Taylor SA, Jaques N, Nosakhare E, Sano A, Picard R. 2017 Personalized multitask learning for predicting tomorrow’s mood, stress, and health. IEEE Trans. Affect. Comput. 11, 200- 213. (10.1109/TAFFC.2017.2784832) 32489521 PMC7266106 source
Not checked in the free run.
May combine two claims.
PremisecitationPPG sensors may have limitations, such as being bulky to use in daily life (e.g. ECG sensors in a chest strap), inaccurate measurement under motions (e.g. PPG sensors), sensitivity to environmental conditions, high power consumption, etc.
“However, they may still have their own limitations, such as being bulky to use in daily life (e.g. ECG sensors in a chest strap), inaccurate measurement under motions (e.g. PPG sensors), sensitivity to environmental conditions, high power consumption, etc. [170].”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [170] Ismail S, Akram U, Siddiqi I. 2021 Heart rate tracking in photoplethysmography signals affected by motion artifacts: a review. EURASIP J. Adv. Signal Process. 2021, 1- 27. (10.1186/s13634-020-00714-2) source
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.
PremisecitationOthers focused on detecting depression by tracking medication, sleep patterns and actions, location or keypress acceleration.
“Others focused on detecting depression by tracking medication, sleep patterns and actions [20], location [11] or keypress acceleration [165].”
Quote in the paperNot a substantial claim0 of 3 sources checked- cites [20] Suhara Y, Xu Y, Pentland A. 2017 Deepmood: forecasting depressed mood based on self-reported histories via recurrent neural networks. In Proc. 26th Int. Conf. on World Wide Web, pp. 715–724. source
Not checked in the free run.
- cites [11] Palmius N, Tsanas A, Saunders KE, Bilderbeck AC, Geddes JR, Goodwin GM, De Vos M. 2016 Detecting bipolar depression from geographic location data. IEEE Trans. Biomed. Eng. 64, 1761- 1771. (10.1109/TBME.2016.2611862) 28113247 PMC5947818 source
Not checked in the free run.
- cites [165] Cao B, Zheng L, Zhang C, Yu PS, Piscitello A, Zulueta J, Ajilore O, Ryan K, Leow AD. 2017 DeepMood: modeling mobile phone typing dynamics for mood detection. In Proc. 23rd ACM SIGKDD Int. Conf. on Knowledge Discovery and Data Mining, pp. 747–755. New York, NY: ACM. source
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationThe two most widely adopted sensor data types are ECG and photoplethysmography (PPG).
“The two most widely adopted sensor data types are ECG and photoplethysmography (PPG) [166–168].”
Quote in the paperNot a substantial claim0 of 3 sources checked- cites [166] Sahoo PK, Thakkar HK, Lee MY. 2017 A cardiac early warning system with multi channel SCG and ECG monitoring for mobile health. Sensors 17, 711. (10.3390/s17040711) 28353681 PMC5421671 source
Not checked in the free run.
- cites [167] Wannenburg J, Malekian R, Hancke GP. 2018 Wireless capacitive-based ECG sensing for feature extraction and mobile health monitoring. IEEE Sensors J. 18, 6023- 6032. (10.1109/JSEN.2018.2844122) source
Not checked in the free run.
- cites [168] Lopez Perales CR, Van Spall HG, Maeda S, Jimenez A, Laţcu DG, Milman A, Kirakoya-Samadoulougou F, Mamas MA, Muser D, Casado Arroyo R. 2021 Mobile health applications for the detection of atrial fibrillation: a systematic review. Europace 23, 11- 28. (10.1093/europace/euaa139) 33043358 PMC7842109 source
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationRF signal has been extended for sleep posture monitoring, which transmits radio signals and observes their reflections from the surroundings.
“Recently, RF signal has also been extended for sleep posture monitoring [183], which transmits the radio signals and observes their reflections from the surroundings.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [183] Yue S, Yang Y, Wang H, Rahul H, Katabi D. 2020 BodyCompass: monitoring sleep posture with wireless signals. Proc. ACM Interact. Mob. Wearable Ubiquitous Technol. 4, 1- 25. (10.1145/3397311)35846237 source
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.
PremisecitationCRF is frequently used in sports as an indicator of endurance capacity, having strong predictive value for other sport-related performance traits.
“Beyond its implications in medicine, CRF is frequently used in sports as an indicator of endurance capacity, having strong predictive value for other sport-related performance traits [148].”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [148] Ross R et al. 2016 Importance of assessing cardiorespiratory fitness in clinical practice: a case for fitness as a clinical vital sign: a scientific statement from the American Heart Association. Circulation 134, e653- e699. (10.1161/CIR.0000000000000461) 27881567 source
Not checked in the free run.
May combine two claims.
PremisecitationAs Wi-Fi can travel through almost any indoor corner, human body movement or gesture changes can lead to small changes in signal propagation, and thus can be used to detect activities.
“As Wi-Fi can travel through almost any indoor corner, human body movement or gesture changes can lead to small changes in signal propagation, and thus can be used to detect activities.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [182] Liu J, Liu H, Chen Y, Wang Y, Wang C. 2019 Wireless sensing for human activity: a survey. IEEE Commun. Surv. Tutor. 22, 1629- 1645. (10.1109/COMST.2019.2934489) source
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.
PremisecitationStudentLife combined sensing and self-reports to assess the impact of student workload on stress, whereas Snapshot tracked their mood and sleep.
“For instance, StudentLife [63] combined sensing and self-reports to assess the impact of student workload on stress, whereas Snapshot [160] tracked their mood and sleep.”
Quote in the paperNot a substantial claim0 of 2 sources checked- cites [63] Wang R, Chen F, Chen Z, Li T, Harari G, Tignor S, Zhou X, Ben-Zeev D, Campbell AT. 2014 StudentLife: assessing mental health, academic performance and behavioral trends of college students using smartphones. In Proc. 2014 ACM Int. Joint Conf. on Pervasive and Ubiquitous Computing, pp. 3–14. source
Not checked in the free run.
- cites [160] Taylor SA, Jaques N, Nosakhare E, Sano A, Picard R. 2017 Personalized multitask learning for predicting tomorrow’s mood, stress, and health. IEEE Trans. Affect. Comput. 11, 200- 213. (10.1109/TAFFC.2017.2784832) 32489521 PMC7266106 source
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.
Interpretationnone givenPeople without economic privilege have a limited capacity to self-isolate at home, which undermines the effectiveness of these approaches.
“Moreover, people without economic privilege have a limited capacity to self-isolate at home, which undermines the effectiveness of these approaches.”
Quote in the paperNot a substantial claim- Rests on an assumption: Extracting meaning from wearable data can be challenging, since sensors measure low-level signals as opposed to the more high-level events that are usually of interest.
May combine two claims.
PremisecitationThis type of assessment requires trained staff and expensive laboratory settings with specialized equipment and often test criteria for maximal effort are not met.
“This type of assessment requires trained staff and expensive laboratory settings with specialized equipment and often test criteria for maximal effort are not met [149].”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [149] Swain DP et al. 2014 ACSM’s resource manual for guidelines for exercise testing and prescription. New York, NY: Wolters Kluwer Health/Lippincott Williams & Wilkins.
Not checked in the free run.
May combine two claims.
PremisecitationSelf-reported happiness is linked to longevity, personality traits and reduced mortality risk.
“while self-reported happiness is linked to longevity [152], personality traits [153] and reduced mortality risk [154].”
Quote in the paperNot a substantial claim0 of 3 sources checked- cites [152] Veenhoven R. 2008 Healthy happiness: effects of happiness on physical health and the consequences for preventive health care. J. Happiness Stud. 9, 449- 469. (10.1007/s10902-006-9042-1) source
Not checked in the free run.
- cites [153] Geukes K, Nestler S, Hutteman R, Küfner AC, Back MD. 2017 Trait personality and state variability: predicting individual differences in within-and cross-context fluctuations in affect, self-evaluations, and behavior in everyday life. J. Res. Personal. 69, 124- 138. (10.1016/j.jrp.2016.06.003) source
Not checked in the free run.
- cites [154] Aichele S, Rabbitt P, Ghisletta P. 2016 Think fast, feel fine, live long: a 29-year study of cognition, health, and survival in middle-aged and older adults. Psychol. Sci. 27, 518- 529. (10.1177/0956797615626906) 26917212 source
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationSingapore implemented a mobile phone application, TraceTogether, which pushes notifications to devices in close proximity as 2 m apart and sharing information.
“For example, Singapore implemented a mobile phone application, TraceTogether, which pushes notifications to devices in close proximity as 2 m apart and sharing information [175].”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [175] Whitelaw S, Mamas MA, Topol E, Van Spall HG. 2020 Applications of digital technology in COVID-19 pandemic planning and response. Lancet Digit. Health 2, e435- e440. (10.1016/S2589-7500(20)30142-4) 32835201 PMC7324092 source
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.
PremisecitationWi-Fi signal is one of the key modalities for human activity recognition and gesture recognition.
“Wi-Fi signal is one of the key modalities for human activity recognition and gesture recognition [182].”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [182] Liu J, Liu H, Chen Y, Wang Y, Wang C. 2019 Wireless sensing for human activity: a survey. IEEE Commun. Surv. Tutor. 22, 1629- 1645. (10.1109/COMST.2019.2934489) source
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.
PremisecitationUtsureko and EmotionSense, two independent smartphone applications for mood monitoring through self-reports, were used by more than 24 000 and 17 000 users, respectively.
“On a larger scale, Utsureko [20] and EmotionSense [21], two independent smartphone applications for mood monitoring through self-reports, were used by more than 24 000 and 17 000 users, respectively.”
Quote in the paperNot a substantial claim0 of 2 sources checked- cites [20] Suhara Y, Xu Y, Pentland A. 2017 Deepmood: forecasting depressed mood based on self-reported histories via recurrent neural networks. In Proc. 26th Int. Conf. on World Wide Web, pp. 715–724. source
Not checked in the free run.
- cites [21] Servia-Rodríguez S, Rachuri KK, Mascolo C, Rentfrow PJ, Lathia N, Sandstrom GM. 2017 Mobile sensing at the service of mental well-being: a large-scale longitudinal study. In Proc. 26th Int. Conf. on World Wide Web, pp. 103–112. International World Wide Web Conferences Steering Committee. source
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationBoth Bluetooth and Wi-Fi signals have been widely used for indoor localization particularly for older age groups, in assisting their living quality and also the emergency response.
“Both Bluetooth and Wi-Fi signals have been widely used for indoor localization particularly for older age groups, in assisting their living quality and also the emergency response [180,181].”
Quote in the paperNot a substantial claim0 of 2 sources checked- cites [180] Qureshi UM, Umair Z, Hancke GP. 2019 Indoor localization using wireless fidelity (WiFi) and bluetooth low energy (BLE) signals. In 2019 IEEE 28th Int. Symp. on Industrial Electronics (ISIE), pp. 2232–2237. IEEE. PMC6696175 31349727 source
Not checked in the free run.
- cites [181] Vuong N, Goh S, Chan S, Lau C. 2013 A mobile-health application to detect wandering patterns of elderly people in home environment. In 2013 35th Annual Int. Conf. of the IEEE Engineering in Medicine and Biology Society (EMBC), pp. 6748–6751. IEEE. 24111292 source
Not checked in the free run.
May combine two claims.
PremisecitationBinary prediction is quite common in the mood prediction literature, where mood is simplified to a binary state, so that extreme depression is binned in the same class as moderate unhappiness.
“Further, binary prediction is quite common in the mood prediction literature, where mood is simplified to a binary state [21], so that, for instance, extreme depression is binned in the same class as moderate unhappiness.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [21] Servia-Rodríguez S, Rachuri KK, Mascolo C, Rentfrow PJ, Lathia N, Sandstrom GM. 2017 Mobile sensing at the service of mental well-being: a large-scale longitudinal study. In Proc. 26th Int. Conf. on World Wide Web, pp. 103–112. International World Wide Web Conferences Steering Committee. source
Not checked in the free run.
May combine two claims.
PremisecitationSmartphones and wearable devices have enabled timely delivery of experience sampling, allowing a near real-time detection of clinical outcomes and relapses.
“To this end, smartphones and wearable devices have enabled timely delivery of experience sampling [157], allowing a near real-time detection of clinical outcomes and relapses.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [157] Csikszentmihalyi M, Larson R. 1987 Validity and reliability of the experience-sampling method. J. Nerv. Ment. Dis. 175, 526- 536. (10.1097/00005053-198709000-00004) 3655778 source
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing).
Interpretationnone givenOn-device computation is a popular way to mitigate privacy risks.
“Another popular way to mitigate privacy risks is to use on-device computation.”
Quote in the paperNot a substantial claimMay combine two claims.
PremisecitationOut of 20 000 apps from Google Play studied, 45% rely on unencrypted communication, and 23% of personal data (e.g., location information and passwords) are sent on unsecured traffic.
“For example, out of 20 000 apps from Google Play studied in [202] it was revealed that 45% of the apps rely on unencrypted communication, and as much as 23% of personal data (e.g. location information and passwords) are sent on unsecured traffic.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [202] Tangari G, Ikram M, Sentana IWB, Ijaz K, Kaafar MA, Berkovsky S. 2021 Analyzing security issues of android mobile health and medical applications. J. Am. Med. Inform. Assoc. 28, 2074- 2084. (10.1093/jamia/ocab131) 34338763 PMC8449631 source
Not checked in the free run.
May combine two claims.
PremisecitationDifferent individuals may exhibit varying symptoms or responses to the same treatment, making a case for improved personalization.
“Additionally, different individuals may exhibit varying symptoms or responses to the same treatment. This requires meticulous attention for systems designed around individual needs, making a case for improved personalization [189].”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [189] Kathan A et al. 2022 Personalised depression forecasting using mobile sensor data and ecological momentary assessment. Front. Digit. Health 4, 964582. (10.3389/fdgth.2022.964582) 36465087 PMC9715619 source
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).
Interpretationnone givenRobustness to distribution shifts can be dealt with using domain adaptation or personalization techniques tailored to individuals and learning from their unique preferences.
“Robustness to distribution shifts can also be dealt with using domain adaptation or personalization techniques that are tailored to individuals and learn from their unique preferences (see previous paragraph).”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationNew opportunities in interpretability include integrating domain knowledge from clinicians into modelling approaches.
“New opportunities are blossoming in the realm of interpretability, not only encompassing previously discussed areas but also highlighting the potential of integrating domain knowledge from clinicians into modelling approaches [188].”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [188] Ramanarayanan V, Lammert AC, Rowe HP, Quatieri TF, Green JR. 2022 Speech as a biomarker: opportunities, interpretability, and challenges. Perspect. ASHA Spec. Interest Groups 7, 276- 283. (10.1044/2021_PERSP-21-00174) source
Not checked in the free run.
Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing).
InterpretationreasoningThese characteristics may require additional mechanisms in AI algorithms to handle effectively.
“These characteristics may require additional mechanisms in AI algorithms to handle effectively.”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing). May combine two claims.
InterpretationreasoningApplying ideas from ML fairness could possibly solve some of these problems; however, sensing data and models have certain particularities.
“Applying ideas from ML fairness could possibly solve some of these problems; however, sensing data and models have certain particularities.”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S4 (based). May combine two claims.
InterpretationreasoningExisting studies have not been tailored for mobile health, but they share similar challenges.
“Although the existing studies have not been tailored for mobile health, they share similar challenges.”
Quote in the paperNot a substantial claimMay combine two claims.
PremisecitationDeep learning can identify spurious correlations but often fails to determine causation, which can be essential in a clinical context.
“Evidence also suggests that while deep learning can identify spurious correlations, it often fails to determine causation, which can be essential in a clinical context [187].”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [187] Obermeyer Z, Emanuel EJ. 2016 Predicting the future—big data, machine learning, and clinical medicine. N. Engl. J. Med. 375, 1216. (10.1056/NEJMp1606181) 27682033 PMC5070532 source
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationActive learning methods could enable selection of uncertain samples or users for further analysis such as additional review by domain experts (physicians).
“Furthermore, active learning methods could enable the selection of uncertain samples or users for further analysis such as an additional review by domain experts (physicians) [186].”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [186] Budd S, Robinson EC, Kainz B. 2021 A survey on active learning and human-in-the-loop deep learning for medical image analysis. Med. Image Anal. 71, 102062. (10.1016/j.media.2021.102062) 33901992 source
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationVisualization tools such as the What-If Tool and FairLens can help surface potential data outliers and help correct them before they creep into models.
“For example, visualization tools such as the What-If Tool [193] and FairLens [194] can help surface any potential data outliers and help correct them before they creep into models.”
Quote in the paperNot a substantial claim0 of 2 sources checked- cites [193] Wexler J, Pushkarna M, Bolukbasi T, Wattenberg M, Viégas F, Wilson J. 2019 The what-if tool: interactive probing of machine learning models. IEEE Trans. Vis. Comput. Graph. 26, 56- 65. (10.1109/TVCG.2019.2934619) 31442996
Not checked in the free run.
- cites [194] Panigutti C, Perotti A, Panisson A, Bajardi P, Pedreschi D. 2021 FairLens: auditing black-box clinical decision support systems. Inf. Process. Manag. 58, 102657. (10.1016/j.ipm.2021.102657) source
Not checked in the free run.
May combine two claims.
PremisecitationPeople of colour are consistently misclassified by health sensors such as oximeters as they were validated on predominantly white populations.
“At the same time, people of colour are consistently misclassified by health sensors such as oximeters as they were validated on predominantly white populations [191].”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [191] Sjoding MW, Dickson RP, Iwashyna TJ, Gay SE, Valley TS. 2020 Racial bias in pulse oximetry measurement. N. Engl. J. Med. 383, 2477- 2478. (10.1056/NEJMc2029240) 33326721 PMC7808260 source
Not checked in the free run.
Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing).
Premisenone givenSecurity and privacy in mobile health applications are of particular importance.
“While security and privacy in mobile health applications are of particular importance, we point the reader to other material that is available [208,209].”
Quote in the paperNot a substantial claim0 of 2 sources checked- cites [208] Kotz D, Gunter CA, Kumar S, Weiner JP. 2016 Privacy and security in mobile health: a research agenda. Computer 49, 22- 30. (10.1109/MC.2016.185) 28344359 PMC5362254 source
Not checked in the free run.
- cites [209] Dehling T, Gao F, Schneider S, Sunyaev A. 2015 Exploring the far side of mobile health: information security and privacy of mobile health apps on iOS and Android. JMIR mHealth uHealth 3, e3672. (10.2196/mhealth.3672) PMC4319144 25599627 source
Not checked in the free run.
Classified as probably not a substantial claim. May combine two claims.
InterpretationreasoningTransmitting local device data to a server is bandwidth-intensive and raises privacy issues; emphasis is on distributed training methods.
“Instead of transmitting local device data to a server (a process that is bandwidth-intensive and raises privacy issues), the emphasis is on distributed training methods.”
Quote in the paperSubstantial claimMay combine two claims.
Interpretationnone givenIt is relatively straightforward to distinguish a person's skin tone given a picture, but much harder to do so from oximetry measurements, necessitating collection of supplementary metadata such as demographics.
“In other words, while it is relatively straightforward to distinguish a person’s skin tone given a picture, it is much harder to do so from oximetry measurements, necessitating the collection of supplementary metadata, such as demographics.”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.
PremisecitationBoth privacy and security are popular topics of research and several approaches have been proposed including anonymity frameworks, differential privacy, and blockchain-based solutions.
“Both privacy and security are popular topics of research and several approaches have been proposed including anonymity frameworks [203] to more recently proposed differential privacy [204], and blockchain-based solutions [205].”
Quote in the paperNot a substantial claim0 of 3 sources checked- cites [203] Sweeney L. 2002 k-anonymity: a model for protecting privacy. Int. J. Uncertain Fuzziness Knowl.-Based Syst. 10, 557- 570. (10.1142/S0218488502001648) source
Not checked in the free run.
- cites [204] Liu W, Zhang Y, Yang H, Meng Q. 2023 A survey on differential privacy for medical data analysis. Ann. Data Sci. (10.1007/s40745-023-00475-3) PMC10257172 38625247 source
Not checked in the free run.
- cites [205] Rajasekaran AS, Maria A, Rajagopal M, Lorincz J. 2022 Blockchain enabled anonymous privacy-preserving authentication scheme for internet of health things. Sensors 23, 240. (10.3390/s23010240) 36616838 PMC9823844 source
Not checked in the free run.
May combine two claims.
PremisecitationApple's CRF feature is publicly validated on 221 participants while non-commercial methods use population-scale data from more than 11 000 participants.
“For example, Apple’s CRF feature is publicly validated on 221 participants [135] while non-commercial methods use population-scale data from more than 11 000 participants [136].”
Quote in the paperSubstantial claim0 of 2 sources checked- cites [135] Apple. 2021 Using Apple Watch to estimate cardio fitness with VO2 max. See https://www.apple.com/healthcare/docs/site/Using_Apple_Watch_to_Estimate_Cardio_Fitness_with_VO2_max.pdf.
Not checked in the free run.
- cites [136] Spathis D, Perez-Pozuelo I, Gonzales TI, Wu Y, Brage S, Wareham N, Mascolo C. 2022 Longitudinal cardio-respiratory fitness prediction through wearables in free-living environments. npj Digit. Med. 5, 176. (10.1038/s41746-022-00719-1) 36460766 PMC9718831 source
Not checked in the free run.
Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing). May combine two claims.
InterpretationreasoningData validation methods can facilitate fairness and robustness and should be applied across sensitive attributes.
“Similarly, data validation methods can facilitate fairness and robustness and, therefore, should also be applied across sensitive attributes.”
Quote in the paperNot a substantial claimMay combine two claims.
PremisecitationEnsemble learning for sound-based COVID-19 detection can infer uncertainty of model predictions, and false predictions often yield higher uncertainty, enabling suggestions to users with certainty higher than a threshold to repeat audio test or take clinical tests if digital diagnosis fails.
“One study [185] employed ensemble learning for sound-based COVID-19 detection to infer the uncertainty of the model predictions and showed that false predictions often yield higher uncertainty, enabling us to suggest to users with certainty higher than a threshold to repeat the audio test on their phones or to take clinical tests if digital diagnosis still fails.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [185] Xia T, Han J, Qendro L, Dang T, Mascolo C. 2021 Uncertainty-aware COVID-19 detection from imbalanced sound data. (http://arxiv.org/abs/2104.02005) 38319779 source
Not checked in the free run.
The paper relies on this without support.
Most audio health studies were carried out in laboratory or controlled environment, and only limited work has explored audio in mobile health context in free-living environment.
“Despite its wide application and great potential, most of them were carried out using data collected in the laboratory or a controlled environment. Only limited work has explored the potential of audio within the mobile health context in free-living environment.”
The paper relies on this without support.
Lung disease alters the sound generation mechanism.
“Lung disease is also expected to alter the sound generation mechanism.”
The paper relies on this without support.
Many studies using wearables such as smartwatches and earables involve limited numbers of participants.
“Many studies using wearables such as smartwatches and earables only involve limited number of participants, where conventional ML techniques are suitable to avoid overfitting.”
The paper relies on this without support.
Deep learning techniques generally show superior performance over traditional ML techniques with adequate and high-quality data.
“Deep learning techniques have been widely employed as they generally show superior performance over traditional ML techniques with adequate and high-quality data.”
The paper relies on this without support.
Mobility naturally captures the lifestyle of individuals and thus related health information.
“Mobility naturally captures the lifestyle of individuals and thus related health information; for example, the type of restaurant a person visits reveals food habit and hence correlated health problems.”
The paper relies on this without support.
Mobility has seen limited applications for mobile health primarily due to lack of large relevant datasets.
“In spite of its potential to unlock secrets of human health conditions, mobility has seen limited applications for mobile health primarily due to lack of large relevant datasets.”
The paper relies on this without support.
Health applications typically benefit from interpretable features to build explainable models.
“Health applications typically benefit from interpretable features to build explainable models.”
The paper relies on this without support.
Health applications are yet to take advantage from progress in using trajectories in ML algorithms.
“However, health applications are yet to take advantage from these progresses.”
The paper relies on this without support.
Infectious diseases such as COVID-19 and H1N1 are spread primarily by human contacts.
“Infectious diseases such as COVID-19 and H1N1 are spread primarily by human contacts.”
The authors state this as an assumption.
Traditional disease spread models like SEIR assume contacts between random pairs of individuals.
“While traditional disease spread models like SEIR (susceptible exposed infected recovered) assume contacts between random pairs of individuals, incorporating mobility in the model increases its predictive power.”
The paper relies on this without support.
Fine-grained mobility within a local geographical region in terms of venues visited and number of face-to-face meetups affects the growth rate and demographic patterns of infections in that region.
“According to the scale of movement, mobility has different effects on disease spreading, for example, fine-grained mobility within a local geographical region in terms of venues visited and number of face-to-face meetups affect the growth rate and demographic patterns of infections in that region.”
The paper relies on this without support.
Movement between countries or large geographical regions may spread a new disease or a variant to a different country or region.
“On the contrary, movement between countries or large geographical regions may spread a new disease or a variant to a different country or region.”
The paper relies on this without support.
People and venues have non-uniform effect in spreading a disease, e.g. large gatherings act as super-spreaders.
“People and venues have non-uniform effect in spreading a disease, e.g. large gatherings act as super-spreaders.”
The paper relies on this without support.
Older people visit parks more often but have worse health status than younger people.
“for example, older people visit parks more often but have worse health status than younger people.”
The paper relies on this without support.
Access to resources heavily affects health outcome be it the facilities for health being in the neighbourhood or critical machines in a hospital.
“Access to resources heavily affects health outcome be it the facilities for health being in the neighbourhood or critical machines in a hospital.”
The paper relies on this without support.
Poor physical access to health facilities causes reduced uptake of preventive health services and is likely to affect low-income households the most.
“Poor physical access to health facilities causes reduced uptake of preventive health services and is likely to affect low-income households the most.”
The paper relies on this without support.
Timely and accurate mood reporting is important to anticipate clinical outcomes.
“Independently of the scale used, timely and accurate mood reporting is important to anticipate clinical outcomes.”
The paper relies on this without support.
People without economic privilege have a limited capacity to self-isolate at home, which undermines the effectiveness of these approaches.
“Moreover, people without economic privilege have a limited capacity to self-isolate at home, which undermines the effectiveness of these approaches.”
The paper relies on this without support.
Extracting meaning from wearable data can be challenging, since sensors measure low-level signals as opposed to the more high-level events that are usually of interest.
“However, extracting meaning from these data can be challenging, since sensors measure low-level signals (e.g. acceleration) as opposed to the more high-level events that are usually of interest (e.g. arrhythmia, infection or obesity onset).”
The paper relies on this without support.
Since wearables are used in daily, unconstrained environments, activities such as drinking coffee or alcohol, as well as stress, may confound simple heuristics.
“However, since wearables are used in daily, unconstrained environments, activities such as drinking coffee or alcohol, as well as stress, may confound simple heuristics.”
The paper relies on this without support.
Lack of interpretability hinders wide-scale adoption of AI in healthcare and makes it challenging for users and medical practitioners to trust these systems.
“A significant hurdle that hinders the wide-scale adoption of AI systems in healthcare is the lack of interpretability, making it challenging for users and medical practitioners to trust these systems.”
The paper relies on this without support.
Sequential deep learning can generate fairly accurate predictions for motion sequences or audio features in longitudinal time series, but it remains difficult to interpret how the model makes judgments based on historical samples.
“while sequential deep learning can generate fairly accurate predictions when dealing with motion sequences or audio features in longitudinal time series, it remains difficult to interpret how the model makes such judgments based on historical samples.”
The paper relies on this without support.
There is an emerging trend away from solely favouring straightforward ML systems like linear regression or SVMs to ensure explainability or overlooking interpretability in favour of performance gains using opaque deep learning models.
“There is an emerging trend away from solely favouring straightforward ML systems like linear regression or SVMs to ensure explainability or overlooking interpretability in favour of performance gains using opaque deep learning models.”
The paper relies on this without support.
Data distribution shift occurs because data attributes can shift due to factors like ageing or disease progression.
“Owing to factors like ageing or disease progression (e.g. deterioration), the attributes of such data can shift, a phenomenon referred to as the data distribution shift.”
The paper relies on this without support.
Models trained on data gathered from a specific temporal cutoff may underperform in later data collection.
“Models trained on data gathered from a specific temporal cutoff may underperform in later data collection.”
The paper relies on this without support.
Skin lesion classification models underperform when assessing black patients compared to white patients.
“such as skin lesion classification models underperforming when assessing black patients, compared to white patients.”
The paper relies on this without support.
Sensing data typically deal with small-scale studies, proof-of-concept datasets often collected in the laboratory or in the wild.
“they typically deal with small-scale studies, proof-of-concept datasets often collected in the laboratory or in the wild.”
The paper relies on this without support.
Sensing data are mostly sequential in nature, with biases being harder to surface compared to tabular datasets.
“Contrary to the tabular format of such datasets, sensing data are mostly sequential in nature, with biases being harder to surface.”
The paper relies on this without support.
The literature is increasingly interested in approaches that mitigate biases.
“More technically, the literature is increasingly interested in approaches that mitigate biases.”
The paper relies on this without support.
Healthcare applications are increasingly relying on domain-specific models at the moment.
“These applications are increasingly relying on domain-specific models at the moment.”
The paper relies on this without support.
The focus is shifting towards efficient deep learning training on edge devices.
“Thus, the focus is shifting towards efficient deep learning training on edge devices.”
The paper relies on this without support.
Very few phones and wearable devices have medical-grade sensors and are costly.
“Very few phones and wearable devices have medical-grade sensors and are costly.”
The paper relies on this without support.
The majority of devices on the market have a cheap sensor which produces noisy data or is easily interfered with by motion artefacts.
“The majority of devices on the market have a cheap sensor which produces noisy data or is easily interfered with by motion artefacts.”
The paper relies on this without support.
Proper denoising techniques before passing data for modelling or specific model structures such as multi-task learning have been validated to denoise signals or eliminate motion artefacts.
“Proper denoising techniques before passing data for modelling or specific model structures such as multi-task learning have been validated to denoise the signals or eliminate motion artefacts.”
The paper relies on this without support.
Sensor data might be missing during motions, or data are sampled at irregular time intervals in a longitudinal way.
“Moreover, sensor data might be missing during motions, or data are sampled at irregular time intervals in a longitudinal way.”
The paper relies on this without support.
To make the study scalable, some labels are collected in a crowdsourced way, by self-reporting.
“To make the study scalable, some labels are collected in a crowdsourced way, by self-reporting.”
The paper relies on this without support.
Data collection is time-consuming and cost-intensive.
“Data collection is time-consuming and cost-intensive.”
The paper relies on this without support.
Most mobile health apps are validated only using a limited number of participants.
“Most mobile health apps are validated only using a limited number of participants.”
The paper relies on this without support.
New drugs or treatments undergo strict certification and regulatory screenings before they become available to the public.
“New drugs or treatments undergo strict certification and regulatory screenings before they become available to the public.”
The paper relies on this without support.
Most commonly, vendors conduct randomized controlled trials in order to rule out side effects and validate efficacy.
“Most commonly, the vendors conduct randomized controlled trials in order to rule out side effects and validate the efficacy thereof.”
The paper relies on this without support.
In mobile health, validation is still a thorny issue.
“In mobile health, validation is still a thorny issue.”
The paper relies on this without support.
Data remain siloed within organizations and companies and most of the time treated as a ‘moat’, or a competitive advantage of each company.
“Data remain siloed within organizations and companies and most of the time treated as a ‘moat’, or a competitive advantage of each company.”
The paper relies on this without support.
The few mobile health products that become available are labelled as ‘well-being’ apps so as to bypass the strict regulatory screening for medical devices.
“The few mobile health products that become available are labelled as ‘well-being’ apps so as to bypass the strict regulatory screening for medical devices.”
The paper relies on this without support.
Sensing signals from mobile phones often contain privacy-sensitive information.
“Last, sensing signals from mobile phones often contain privacy-sensitive information.”
The paper relies on this without support.
Continuously recorded audio from phones will reveal personal conversations, and mobility data can leak sensitive location visits and meetups.
“For example, continuously recorded audio from phones will reveal personal conversations, and mobility data can leak sensitive location visits and meetups.”
The paper relies on this without support.
Not every sensing modality poses the same threat; motion data are comparatively less privacy sensitive.
“However, not every sensing modality poses the same threat; for example, motion data are comparatively less privacy sensitive.”
The paper relies on this without support.
Mobility data expose sensitive personal information.
“On the other hand, mobility data expose sensitive personal information.”
The paper relies on this without support.
Security has always been a significant challenge in mobile health applications.
“Moreover, security has always been a significant challenge in mobile health applications.”
The paper relies on this without support.
Several studies have identified security challenges that affect the development of secure mobile health applications.
“Several studies have identified security challenges that affect the development of secure mobile health applications.”
The paper relies on this without support.
Security and privacy in mobile health applications are of particular importance.
“While security and privacy in mobile health applications are of particular importance, we point the reader to other material that is available [208,209].”
The citation lists haven’t been fetched yet
References 197 · Cited by 30
Research works printed but not linked: 6
- Xia T et al. 2021 COVID-19 sounds: a large-scale audio dataset for digital respiratory screening. In 35th Conf. on Neural Information Processing Systems Datasets and Benchmarks Track (Round 2).
- 10.1145/3494990 · DOI · Paper page
- Mikelsons G Smith M Mehrotra A Musolesi M. 2017 Towards deep learning models for psychological state prediction using smartphone data: challenges and opportunities. In Workshop on Machine Learning for Health (ML4H) at 31st Conf. on Neural Information Processing Systems Long Beach CA USA 4–9 December 2017.
- Agarwal A Dudík M Wu ZS. 2019 Fair regression: quantitative definitions and reduction-based algorithms. In Proc. 36th Int. Conf. on Machine Learning pp. 120–129. PMLR vol. 97.
- Yfantidou S Constantinides M Spathis D Vakali A Quercia D Kawsar F. 2023 Beyond accuracy: a critical review of fairness in machine learning for mobile and wearable computing. (http://arxiv.org/abs/2303.15585) · arXiv
- Li Z Wallace E Shen S Lin K Keutzer K Klein D Gonzalez J. 2020 Train big then compress: rethinking model size for efficient training and inference of transformers. In Proc. 37th Int. Conf. on Machine Learning pp. 5958–5968. PMLR vol. 119.
- Authors
- Dang, T., Spathis, D., Ghosh, A., & Mascolo, C.
- Affiliations
- University of Cambridge (GB)
- University of Cambridge, Cambridge, UK
- Corresponding author
- Dang, Ting
- Venue
- Royal Society Open Science
- Publisher
- The Royal Society
- Published
- 2023-11
- Volume
- 10
- Issue
- 11
- Pages
- 230806-230806
- ISSN
- 2054-5703 (ISSN-L)
- PMC
- PMC10646451
- OpenAlex
- W4388705935
- Semantic Scholar
- 10a23ad25e86cf86f3e58e91d75f3685c0e1f905
- Cited by
- 30
- References
- 197
- Citations by year
- 2026: 12 · 2025: 11 · 2024: 7
- FWCI
- 0.78
- Type
- journal article
- Language
- en
- Open access
- gold
- doi.org
- Locations
- Royal Society Open Science (publishedVersion, cc-by)
- PubMed Central (publishedVersion, cc-by)
- PubMed (publishedVersion)
- University of Birmingham Research Portal (University of Birmingham) (submittedVersion, cc-by)
- Apollo (University of Cambridge) (submittedVersion, cc-by)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
- Funders
- H2020 European Research Council (Project 833296)
- Wellcome Trust Project (213939)
- Topics
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Context-Aware Activity Recognition Systems
- Health Sciences › Health Professions › General Health Professions › Mobile Health and mHealth Applications
- Social Sciences › Transportation › Human Mobility and Location-Based Analysis
- Keywords
- mobile health, wearable sensing, wearable sensor data, health monitoring, health prediction, noisy sensor measurements, sparse time series, data heterogeneity, privacy, resource constraints, audio sensing, position sensing, activity monitoring
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-08.
Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-08.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
