Harnessing EHR data for health research
The DOI matches its Crossref record; 94 of the 94 references it cites were found in a registry or an index.
Abstract
With the increasing availability of rich, longitudinal, real-world clinical data recorded in electronic health records (EHRs) for millions of patients, there is a growing interest in leveraging these records to improve the understanding of human health and disease and translate these insights into clinical applications. However, there is also a need to consider the limitations of these data due to various biases and to understand the impact of missing information. Recognizing and addressing these limitations can inform the design and interpretation of EHR-based informatics studies that avoid confusing or incorrect conclusions, particularly when applied to population or precision medicine. Here we discuss key considerations in the design, implementation and interpretation of EHR-based informatics studies, drawing from examples in the literature across hypothesis generation, hypothesis testing and machine learning applications. We outline the growing opportunities for EHR-based informatics studies, including association studies and predictive modeling, enabled by evolving AI capabilities-while addressing limitations and potential pitfalls to avoid.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
In the citing paper
“For example, changes in physical activity and movement patterns, driving behavior, sleep patterns, everyday speech, eye tracking, facial expressions, technology use, writing, texting, and keyboard typing dynamics have all been shown to differentiate those with and without neurodegenerative disease.111–119”
From Review of Artificial Intelligence for Clinical Use in Alzheimer's Disease and Related Dementias (2026) · cites this paper as [119]
Citations
Cites
Sudlow, C., Gallacher, J., Allen, N., Beral, V., Burton, P., Danesh, J., Downey, P., Elliott, P., Green, J., Landray, M., et al. (2015) · PLOS Medicine · doi:10.1371/journal.pmed.1001779 · Cited by 12,728
Austin, P. C. (2011) · Multivariate Behavioral Research · doi:10.1080/00273171.2011.568786 · Cited by 12,242 · cited 1×
“Many approaches exist for comparator matching 36–38 , with each approach having its benefits and limitations. of these limitations is essential when designing methods for EHR data analysis and interpretation of results, particularly if the insights will be utilized to inform treatment and care that…”
Dissecting racial bias in an algorithm used to manage the health of populations
Obermeyer, Z., Powers, B., Vogeli, C., & Mullainathan, S. (2019) · Science · doi:10.1126/science.aax2342 · Cited by 7,137 · cited 1×
“For example, race correction in interpreting glomerular filtration rate or deciding on a Cesarean section may lead to increased rates of procedures within a demographic group, which may be picked up by association studies or predictive models 30,54,65 .”
Matching Methods for Causal Inference: A Review and a Look Forward
Stuart, E. A. (2010) · Statistical Science · doi:10.1214/09-sts313 · Cited by 5,478
MatchIt : Nonparametric Preprocessing for Parametric Causal Inference
Ho, D. E., Imai, K., King, G., & Stuart, E. A. (2011) · Journal of Statistical Software · doi:10.18637/jss.v042.i08 · Cited by 4,385 · cited 2×
“Current and emerging advancements in the EHR informatics field include causal inference methods 38,72 , state transition models 73 , and transformer models 74,75 to account for temporal relationships in the records.”
Large language models in medicine
Thirunavukarasu, A. J., Ting, D. S. J., Elangovan, K., Gutierrez, L., Tan, T. F., & Ting, D. S. W. (2023) · Nature Medicine · doi:10.1038/s41591-023-02448-8 · Cited by 3,944 · cited 1×
“With transformer models and large language models (LLMs), applications can be extended for multiple tasks in one model—including summarization, chatting, future outcome prediction, patient risk alerts and differential diagnoses 58 .”
MIMIC-IV, a freely accessible electronic health record dataset
Johnson, A. E. W., Bulgarelli, L., Shen, L., Gayles, A., Shammout, A., Horng, S., Pollard, T. J., Hao, S., Moody, B., Gow, B., et al. (2023) · Scientific Data · doi:10.1038/s41597-022-01899-x · Cited by 3,401 · cited 2×
“These datasets are now available as anonymized ‘controlled access’ datasets at many institutions and healthcare systems within the United States and worldwide 6,7 , and some—such as MIMIC 8 , AllofUs 9 and the UK Biobank 10 —are publicly accessible for research.”
Using Big Data to Emulate a Target Trial When a Randomized Trial Is Not Available: Table 1.
Hernán, M. A., & Robins, J. M. (2016) · American Journal of Epidemiology · doi:10.1093/aje/kwv254 · Cited by 3,115
The “All of Us” Research Program
The All of Us Research Program Investigators (2019) · New England Journal of Medicine · doi:10.1056/nejmsr1809937 · Cited by 2,440
Hidden in Plain Sight — Reconsidering the Use of Race Correction in Clinical Algorithms
Vyas, D. A., Eisenstein, L. G., & Jones, D. S. (2020) · New England Journal of Medicine · doi:10.1056/nejmms2004740 · Cited by 1,731 · cited 1×
“For example, race correction in interpreting glomerular filtration rate or deciding on a Cesarean section may lead to increased rates of procedures within a demographic group, which may be picked up by association studies or predictive models 30,54,65 .”
Hernan, M. A. (2002) · American Journal of Epidemiology · doi:10.1093/aje/155.2.176 · Cited by 1,445
PheWAS: demonstrating the feasibility of a phenome-wide scan to discover gene–disease associations
Denny, J. C., Ritchie, M. D., Basford, M. A., Pulley, J. M., Bastarache, L., Brown-Gentry, K., Wang, D., Masys, D. R., Roden, D. M., & Crawford, D. C. (2010) · Bioinformatics · doi:10.1093/bioinformatics/btq126 · Cited by 1,219 · cited 3×
“With availability of genetic data, genetic association with phenotypes can also be explored to elucidate biological hypotheses (for example, phenome-wide association studies 33,43,44 ).”
The imperative for regulatory oversight of large language models (or generative AI) in healthcare
Meskó, B., & Topol, E. J. (2023) · npj Digital Medicine · doi:10.1038/s41746-023-00873-0 · Cited by 1,047 · cited 1×
“Nevertheless, we are at the cusp of unlocking the potential of EHRs and AI to facilitate advances in health and healthcare in the coming decades 59,60 .”
Balance diagnostics after propensity score matching
written on behalf of AME Big-Data Clinical Trial Collaborative Group, Zhang, Z., Kim, H. J., Lonjon, G., & Zhu, Y. (2019) · Annals of Translational Medicine · doi:10.21037/atm.2018.12.10 · Cited by 1,031 · cited 1×
“…Propensity score matching is one approach, whose benefits include the reduction of baseline differences between comparison groups in potential confounding variables (for example, age and other demographic features, comorbid conditions) that could influence the studied outcome of interest 39 .”
Structured information extraction from scientific text with large language models
Dagdelen, J., Dunn, A., Lee, S., Walker, N., Rosen, A. S., Ceder, G., Persson, K. A., & Jain, A. (2024) · Nature Communications · doi:10.1038/s41467-024-45563-x · Cited by 771 · cited 1×
“Although extracting data from these notes has historically been extremely difficult, LLMs have shown to be extremely good at these tasks 66–68 , which will considerably enhance the overall availability of data for studies going forward.”
Digital twins for health: a scoping review
Katsoulakis, E., Wang, Q., Wu, H., Shahriyari, L., Fletcher, R., Liu, J., Achenie, L., Liu, H., Jackson, P., Xiao, Y., et al. (2024) · npj Digital Medicine · doi:10.1038/s41746-024-01073-0 · Cited by 760 · cited 1×
“Generative AI can learn from the underlying data to create simulated patient phenotypes and trajectories, which can be utilized in applications such as clinical summarization and digital twins 56,57 .”
Electronic health records to facilitate clinical research
Cowie, M. R., Blomster, J. I., Curtis, L. H., Duclaux, S., Ford, I., Fritz, F., Goldman, S., Janmohamed, S., Kreuzer, J., Leenay, M., et al. (2017) · Clinical Research in Cardiology · doi:10.1007/s00392-016-1025-6 · Cited by 651
BEHRT: Transformer for Electronic Health Records
Li, Y., Rao, S., Solares, J. R. A., Hassaine, A., Ramakrishnan, R., Canoy, D., Zhu, Y., Rahimi, K., & Salimi-Khorshidi, G. (2020) · Scientific Reports · doi:10.1038/s41598-020-62922-y · Cited by 595
Hersh, W. R., Weiner, M. G., Embi, P. J., Logan, J. R., Payne, P. R. O., Bernstam, E. V., Lehmann, H. P., Hripcsak, G., Hartzog, T. H., Cimino, J. J., et al. (2013) · Medical Care · doi:10.1097/mlr.0b013e31829b1dbd · Cited by 567
Identification of type 2 diabetes subgroups through topological analysis of patient similarity
Li, L., Cheng, W.-Y., Glicksberg, B. S., Gottesman, O., Tamler, R., Chen, R., Bottinger, E. P., & Dudley, J. T. (2015) · Science Translational Medicine · doi:10.1126/scitranslmed.aaa9364 · Cited by 549 · cited 2×
“…or dimensionality reduction approaches (for example, body mass index as one dimen-sion instead of height and weight as two dimensions), can also aid in visualizing clusters of pathways or differences in patient populations due to the observed clinical profile 21,22,40 (for example, body mass index”
Validation of a common data model for active safety surveillance research
Overhage, J. M., Ryan, P. B., Reich, C. G., Hartzema, A. G., & Stang, P. E. (2012) · Journal of the American Medical Informatics Association · doi:10.1136/amiajnl-2011-000376 · Cited by 548 · cited 1×
“After data collection, extraction takes the form of converting raw data into standardized representations of data, for example, through a Clinical Data Warehouse or the Observational Medical Outcomes Partnership Common Data Model 12 .”
Real-world data: A brief review of the methods, applications, challenges and opportunities
Liu, F., & Panagiotakos, D. (2022) · BMC Medical Research Methodology · doi:10.1186/s12874-022-01768-6 · Cited by 496 · cited 3×
“With longitudinal health data collected on millions of patients, and billions of dollars invested over a decade of this ‘big data’ collection and data storage effort, there is rising interest in leveraging EHRs for research and clinical applications 4,5 .”
Digital health for aging populations
Chen, C., Ding, S., & Wang, J. (2023) · Nature Medicine · doi:10.1038/s41591-023-02391-8 · Cited by 491 · cited 1×
“Augmentation of EHR datasets with molecular datasets, such as genomics data or imaging datasets, as well as data from digital devices and technologies (such as wearable sensors and electronic surveillance systems) can add biological support for cohort selection approaches and phenotyping 33,34 .”
Conceptualizing, Contextualizing, and Operationalizing Race in Quantitative Health Sciences Research
Lett, E., Asabor, E., Beltrán, S., Cannon, A. M., & Arah, O. A. (2022) · The Annals of Family Medicine · doi:10.1370/afm.2792 · Cited by 398 · cited 1×
“These identities can differ between regions and countries and may represent the impact of an exposure as opposed to an inherent biological phenomenon 61 .”
Improving large language models for clinical named entity recognition via prompt engineering
Hu, Y., Chen, Q., Du, J., Peng, X., Keloth, V. K., Zuo, X., Zhou, Y., Li, Z., Jiang, X., Lu, Z., et al. (2024) · Journal of the American Medical Informatics Association · doi:10.1093/jamia/ocad259 · Cited by 314
A Review of Generalizability and Transportability
Degtiar, I., & Rose, S. (2023) · Annual Review of Statistics and Its Application · doi:10.1146/annurev-statistics-042522-103837 · Cited by 295 · cited 2×
“Since certain diseases may manifest concurrently and single datasets may have sample size limitations, some studies are expanding to include multiple cohorts or controls of various complexities 49,83–85 .”
A population-based phenome-wide association study of cardiac and aortic structure and function
Bai, W., Suzuki, H., Huang, J., Francis, C., Wang, S., Tarroni, G., Guitton, F., Aung, N., Fung, K., Petersen, S. E., et al. (2020) · Nature Medicine · doi:10.1038/s41591-020-1009-y · Cited by 273 · cited 1×
“Medication and procedure associations can be explored to understand patient populations and generate hypotheses that may impact these populations 41,42 .”
Sai, S., Gaur, A., Sai, R., Chamola, V., Guizani, M., & Rodrigues, J. J. P. C. (2024) · IEEE Access · doi:10.1109/access.2024.3367715 · Cited by 268 · cited 1×
“…of the specific tasks being performed by the model (for example, extraction, classification and early risk stratification), and to clearly define the baseline population, temporal relationships, input data and interpretation of the output (such as disease onset, mortality, readmission or cost) 55 .”
Ten common statistical mistakes to watch out for when writing or reviewing a manuscript
Makin, T. R., & Orban de Xivry, J.-J. (2019) · eLife · doi:10.7554/elife.48175 · Cited by 238 · cited 2×
“Presently, there is increased emphasis on control selection, including the utilization of matching methods, propensity score models or regression to account for multiple covariates or identifying controls from external databases 48,81–83 .”
Deep representation learning of electronic health records to unlock patient stratification at scale
Landi, I., Glicksberg, B. S., Lee, H.-C., Cherng, S., Landi, G., Danieletto, M., Dudley, J. T., Furlanello, C., & Miotto, R. (2020) · npj Digital Medicine · doi:10.1038/s41746-020-0301-z · Cited by 213 · cited 2×
“…or dimensionality reduction approaches (for example, body mass index as one dimen-sion instead of height and weight as two dimensions), can also aid in visualizing clusters of pathways or differences in patient populations due to the observed clinical profile 21,22,40 (for example, body mass index”
Gillum, R. F. (2013) · The American Journal of Medicine · doi:10.1016/j.amjmed.2013.03.024 · Cited by 177 · cited 1×
“Since the beginning of the practice of medicine, recordkeeping has been an important aspect of clinical care 1 .”
Opportunities and challenges in using real-world data for health care
Rudrapatna, V. A., & Butte, A. J. (2020) · Journal of Clinical Investigation · doi:10.1172/jci129197 · Cited by 174 · cited 2×
“EHRs provide an extensive, rich, longitudinal dataset with great opportunity for answering scientific questions, developing AI models, and advancing therapeutics in human health and disease 4,51,76,77 .”
A multiomics approach to heterogeneity in Alzheimer’s disease: focused review and roadmap
Badhwar, A., McFall, G. P., Sapkota, S., Black, S. E., Chertkow, H., Duchesne, S., Masellis, M., Li, L., Dixon, R. A., & Bellec, P. (2020) · Brain · doi:10.1093/brain/awz384 · Cited by 165
Phenome-Wide Association Studies
Bastarache, L., Denny, J. C., & Roden, D. M. (2022) · JAMA · doi:10.1001/jama.2021.20356 · Cited by 159 · cited 1×
“With availability of genetic data, genetic association with phenotypes can also be explored to elucidate biological hypotheses (for example, phenome-wide association studies 33,43,44 ).”
Matching With Multiple Control Groups With Adjustment for Group Differences
Stuart, E. A., & Rubin, D. B. (2007) · Journal of Educational and Behavioral Statistics · doi:10.3102/1076998607306078 · Cited by 150
A multiomics approach to heterogeneity in Alzheimer's disease: focused review and roadmap.
Badhwar, A., McFall, G., Sapkota, S., Black, S., Chertkow, H., Duchesne, S., Masellis, M., Li, L., Dixon, R., & Bellec, L. P. (2019) · Brain : a journal of neurology · doi:10.1101/19008615 · Cited by 146
Reflection on modern methods: selection bias—a review of recent developments
Infante‐Rivard, C., & Cusson, A. (2018) · International Journal of Epidemiology · doi:10.1093/ije/dyy138 · Cited by 146
Differentially private synthetic medical data generation using convolutional GANs
Torfi, A., Fox, E. A., & Reddy, C. K. (2022) · Information Sciences · doi:10.1016/j.ins.2021.12.018 · Cited by 145 · cited 1×
“By contrast, generative adversarial networks (GANs) have shown to be extraordinarily good at synthetically generating realistic covariates at the patient level 69,70 or even generating complete individual (synthetic) patients in ways that preserve cause-and-effect relationships 71 .”
Oskotsky, T., Marić, I., Tang, A., Oskotsky, B., Wong, R. J., Aghaeepour, N., Sirota, M., & Stevenson, D. K. (2021) · JAMA Network Open · doi:10.1001/jamanetworkopen.2021.33090 · Cited by 135 · cited 1×
“This study design was used to investigate statin and antidepressant exposure and COVID-19 mortality 90,91 , as well as antidepressant use in pregnancy outcomes 92 .”
Self-citationElectronic health record: Wiring Europe’s healthcare
Kierkegaard, P. (2011) · Computer law & security review · doi:10.1016/j.clsr.2011.07.013 · Cited by 122 · cited 1×
“These datasets are now available as anonymized ‘controlled access’ datasets at many institutions and healthcare systems within the United States and worldwide 6,7 , and some—such as MIMIC 8 , AllofUs 9 and the UK Biobank 10 —are publicly accessible for research.”
Taubes, A., Nova, P., Zalocusky, K. A., Kosti, I., Bicak, M., Zilberter, M. Y., Hao, Y., Yoon, S. Y., Oskotsky, T., Pineda, S., et al. (2021) · Nature Aging · doi:10.1038/s43587-021-00122-7 · Cited by 118 · cited 1×
“…there is great potential for exciting applications in the future for deriving insight into biology based on EHR trends and molecular–phenotype associations, which will allow improved predictive modeling 46 , subtyping 40 , drug repurposing and therapeutic response investigations 76,78 .”
Self-citationAccuracy of an automated knowledge base for identifying drug adverse reactions
Voss, E. A., Boyce, R. D., Ryan, P. B., van der Lei, J., Rijnbeek, P. R., & Schuemie, M. J. (2016) · Journal of Biomedical Informatics · doi:10.1016/j.jbi.2016.12.005 · Cited by 116 · cited 1×
“EHRs provide an extensive, rich, longitudinal dataset with great opportunity for answering scientific questions, developing AI models, and advancing therapeutics in human health and disease 4,51,76,77 .”
Norgeot, B., Muenzen, K., Peterson, T. A., Fan, X., Glicksberg, B. S., Schenk, G., Rutenberg, E., Oskotsky, B., Sirota, M., Yazdany, J., et al. (2020) · npj Digital Medicine · doi:10.1038/s41746-020-0258-y · Cited by 114 · cited 1×
“Text information may require further automated de-identification, which can introduce biases from perceived lower relevance of concepts arising from removal of information (as opposed to lack of association) 16 .”
Self-citationA causal roadmap for generating high-quality real-world evidence
Dang, L. E., Gruber, S., Lee, H., Dahabreh, I. J., Stuart, E. A., Williamson, B. D., Wyss, R., Diaz, I. M., Ghosh, D., Kıcıman, E., et al. (2023) · Journal of Clinical and Translational Science · doi:10.1017/cts.2023.635 · Cited by 113 · cited 1×
“Causal inference approaches can be applied to EHR data to further elucidate causal relationships 52,71,86–88 .”
Carrigan, G. J., Whipple, S., Capra, W. B., Taylor, M. D., Brown, J. S., Lu, M. W., Arnieri, B., Copping, R., & Rothman, K. J. (2019) · Clinical Pharmacology & Therapeutics · doi:10.1002/cpt.1586 · Cited by 112 · cited 1×
“Presently, there is increased emphasis on control selection, including the utilization of matching methods, propensity score models or regression to account for multiple covariates or identifying controls from external databases 48,81–83 .”
EHR foundation models improve robustness in the presence of temporal distribution shift
Guo, L., Steinberg, E., Fleming, S. L., Posada, J., Lemmon, J., Pfohl, S. R., Shah, N. H., Fries, J., & Sung, L. (2023) · Scientific Reports · doi:10.1038/s41598-023-30820-8 · Cited by 107 · cited 1×
“Current and emerging advancements in the EHR informatics field include causal inference methods 38,72 , state transition models 73 , and transformer models 74,75 to account for temporal relationships in the records.”
Yazdany, J., Bansback, N., Clowse, M., Collier, D., Law, K., Liao, K. P., Michaud, K., Morgan, E. M., Oates, J. C., Orozco, C., et al. (2016) · Arthritis Care & Research · doi:10.1002/acr.23089 · Cited by 104 · cited 1×
“With availability of genetic data, genetic association with phenotypes can also be explored to elucidate biological hypotheses (for example, phenome-wide association studies 33,43,44 ).”
Krumholz, A. (1991) · JAMA · doi:10.1001/jama.265.5.622 · Cited by 95 · cited 1×
“For example, due to potential impacts on a patient’s mental capacity or driving rights when diagnosed with certain neurological disorders (such as epilepsy 63 ), the timing of a diagnosis may be delayed.”
Murali, L., Gopakumar, G., Viswanathan, D. M., & Nedungadi, P. (2023) · Journal of Biomedical Informatics · doi:10.1016/j.jbi.2023.104403 · Cited by 93 · cited 1×
“Current and emerging advancements in the EHR informatics field include causal inference methods 38,72 , state transition models 73 , and transformer models 74,75 to account for temporal relationships in the records.”
Tang, A. S., Rankin, K. P., Cerono, G., Miramontes, S., Mills, H., Roger, J., Zeng, B., Nelson, C., Soman, K., Woldemariam, S., et al. (2024) · Nature Aging · doi:10.1038/s43587-024-00573-8 · Cited by 86 · cited 4×
“Excitingly, augmentation of clinical phenotypes with known disease—molecular associations from heterogeneous knowledge databases is emerging as a promising opportunity for generating and prioritizing biological hypotheses—as evidenced by recent efforts in multiple sclerosis and AD 45,46 .”
InfluentialSelf-citationMurugadoss, K., Rajasekharan, A., Malin, B., Agarwal, V., Bade, S., Anderson, J. R., Ross, J. L., Faubion, W. A., Halamka, J. D., Soundararajan, V., et al. (2021) · Patterns · doi:10.1016/j.patter.2021.100255 · Cited by 72 · cited 1×
“Furthermore, de-identification methods are applied during the preprocessing step to ensure proper privacy is maintained 8,13 .”
On mining latent treatment patterns from electronic medical records
Huang, Z., Dong, W., Bath, P. A., Ji, L., & Duan, H. (2014) · Data Mining and Knowledge Discovery · doi:10.1007/s10618-014-0381-y · Cited by 68 · cited 1×
“Understanding and characterizing patterns in the data can give insight into affected populations, practices for treatment prescription and current clinical behaviors 26–28 , as well as identifying potential differences in subgroups and disparities in current care 29–31 .”
Krumholz, A. (1991) · JAMA · doi:10.1001/jama.1991.03460050076024 · Cited by 65
A systematic review of automatic text summarization for biomedical literature and EHRs
Wang, M., Wang, M., Yu, F., Yang, Y., Walker, J., & Mostafa, J. (2021) · Journal of the American Medical Informatics Association · doi:10.1093/jamia/ocab143 · Cited by 63 · cited 1×
“Generative AI can learn from the underlying data to create simulated patient phenotypes and trajectories, which can be utilized in applications such as clinical summarization and digital twins 56,57 .”
Belonwu, S., Li, Y., Bunis, D. G., Rao, A. A., Solsberg, C. W., Tang, A., Fragiadakis, G. K., Dubal, D. B., Oskotsky, T., & Sirota, M. (2021) · Molecular Neurobiology · doi:10.1007/s12035-021-02591-8 · Cited by 62 · cited 1×
“…between two groups where one group is chosen as the reference group (black-identifying individuals versus white-identifying individuals)— similar to comparison of specific cell types in bioinformatics studies (for example, microglia versus all other cells rather than microglia versus neuron 62 ).”
Self-citationPreventing harm from non-conscious bias in medical generative AI
Hastings, J. (2024) · The Lancet Digital Health · doi:10.1016/s2589-7500(23)00246-7 · Cited by 60 · cited 1×
“Nevertheless, we are at the cusp of unlocking the potential of EHRs and AI to facilitate advances in health and healthcare in the coming decades 59,60 .”
Gervasi, S. S., Chen, I. Y., Smith-McLallen, A., Sontag, D., Obermeyer, Z., Vennera, M., & Chawla, R. (2022) · Health Affairs · doi:10.1377/hlthaff.2021.01287 · Cited by 60
Lotfata, A., Moosazadeh, M., Helbich, M., & Hoseini, B. (2023) · International Journal of Health Geographics · doi:10.1186/s12942-023-00343-6 · Cited by 60 · cited 1×
“In leveraging EHR data, one of the greatest opportunities includes improved understanding of the magnitude and implications of disease burdens, as well as patterns of healthcare service utilization and treat-ment within specific demographics or locations 17–20 .”
Nelson, C., Bove, R. M., Butte, A. J., & Baranzini, S. E. (2021) · Journal of the American Medical Informatics Association · doi:10.1093/jamia/ocab270 · Cited by 57 · cited 1×
“Excitingly, augmentation of clinical phenotypes with known disease—molecular associations from heterogeneous knowledge databases is emerging as a promising opportunity for generating and prioritizing biological hypotheses—as evidenced by recent efforts in multiple sclerosis and AD 45,46 .”
PheValuator: Development and evaluation of a phenotype algorithm evaluator
Swerdel, J. N., Hripcsak, G., & Ryan, P. B. (2019) · Journal of Biomedical Informatics · doi:10.1016/j.jbi.2019.103258 · Cited by 51 · cited 1×
“Since clinical datasets may contain potential mis-classifications or lack specificity, exploratory analysis of the data can aid in the development of phenotyping algorithms and cohort selection approaches that may best represent the biological characteristics of a desired population 32 .”
Xu, J., Wang, F. “., Xu, Z., Adekkanattu, P. M., Brandt, P., Jiang, G., Kiefer, R. C., Luo, Y., Mao, C., Pacheco, J. A., et al. (2020) · Learning Health Systems · doi:10.1002/lrh2.10246 · Cited by 50 · cited 1×
“Indeed, some prior association analyses and models for AD pick up the prescription of an anti-dementia drug before an official AD diagnosis 64 .”
A review of Automatic end-to-end De-Identification: Is High Accuracy the Only Metric?
Yogarajan, V., Pfahringer, B., & Mayo, M. (2020) · Applied Artificial Intelligence · doi:10.1080/08839514.2020.1718343 · Cited by 47 · cited 1×
“…However, the de-identification process itself may contribute to biases and inaccuracies in the data; examples include inappropriate or inad-vertent removal of patients, shifting of event dates and assigning all individuals over 89 years of age as being 90 years old 14,15 .”
Freitas, J. K. D., Johnson, K. W., Golden, E., Nadkarni, G. N., Dudley, J. T., Böttinger, E. P., Glicksberg, B. S., & Miotto, R. (2021) · Patterns · doi:10.1016/j.patter.2021.100337 · Cited by 47 · cited 1×
“…or dimensionality reduction approaches (for example, body mass index as one dimen-sion instead of height and weight as two dimensions), can also aid in visualizing clusters of pathways or differences in patient populations due to the observed clinical profile 21,22,40 (for example, body mass index”
Glicksberg, B. S., Li, L., Badgeley, M. A., Shameer, K., Kosoy, R., Beckmann, N. D., Pho, N., Hakenberg, J., Ma, M., Ayers, K. L., et al. (2016) · Bioinformatics · doi:10.1093/bioinformatics/btw282 · Cited by 46
Glicksberg, B. S., Oskotsky, B., Thangaraj, P. M., Giangreco, N., Badgeley, M. A., Johnson, K. W., Datta, D., Rudrapatna, V. A., Rappoport, N., Shervey, M. M., et al. (2019) · Bioinformatics · doi:10.1093/bioinformatics/btz409 · Cited by 45 · cited 2×
“EHR data offer an opportunity to systematically characterize and phenotype various aspects of health, including gaining a broad understanding of the prevalence of diseases, drugs and complex combinations of comorbid disorders or concomitant medications within a specific disease group 21–25 .”
Tang, A. S., Oskotsky, T., Havaldar, S., Mantyh, W. G., Bicak, M., Solsberg, C. W., Woldemariam, S., Zeng, B., Hu, Z., Oskotsky, B., et al. (2022) · Nature Communications · doi:10.1038/s41467-022-28273-0 · Cited by 45 · cited 3×
“From a cross-sectional view of disease, broad associations identified from deep clinical phenotyping can aid in exploratory research 23 .”
Self-citationZhu, R., Vora, B., Menon, S. S., Younis, I. R., Dwivedi, G., Meng, Z., Datta‐Mannan, A., Manchandani, P., Nayak, S., Tammara, B. K., et al. (2023) · Clinical Pharmacology & Therapeutics · doi:10.1002/cpt.2988 · Cited by 45 · cited 2×
“…there is great potential for exciting applications in the future for deriving insight into biology based on EHR trends and molecular–phenotype associations, which will allow improved predictive modeling 46 , subtyping 40 , drug repurposing and therapeutic response investigations 76,78 .”
HIPAA and the Leak of “Deidentified” EHR Data
Mandl, K. D., & Perakslis, E. D. (2021) · New England Journal of Medicine · doi:10.1056/nejmp2102616 · Cited by 44 · cited 1×
“…However, the de-identification process itself may contribute to biases and inaccuracies in the data; examples include inappropriate or inad-vertent removal of patients, shifting of event dates and assigning all individuals over 89 years of age as being 90 years old 14,15 .”
Sinha, P. K., Sunder, G., Bendale, P., Mantri, M., & Dande, A. (2012) · Wiley eBooks · doi:10.1002/9781118479612 · Cited by 42
Shi, J., Wang, D., Tesei, G., & Norgeot, B. (2022) · Frontiers in Artificial Intelligence · doi:10.3389/frai.2022.918813 · Cited by 37 · cited 2×
“By contrast, generative adversarial networks (GANs) have shown to be extraordinarily good at synthetically generating realistic covariates at the patient level 69,70 or even generating complete individual (synthetic) patients in ways that preserve cause-and-effect relationships 71 .”
Self-citationSu, C., Zhang, Y., Flory, J., Weiner, M. G., Kaushal, R., Schenck, E. J., & Wang, F. (2021) · npj Digital Medicine · doi:10.1038/s41746-021-00481-w · Cited by 35
Gold, R. B., Sheppler, C. R., Hessler, D., Bunce, A., Cottrell, E., Yosuf, N., Pisciotta, M., Gunn, R., Leo, M. C., & Gottlieb, L. M. (2021) · JMIR Research Protocols · doi:10.2196/31733 · Cited by 33 · cited 1×
“Prospective methods to evaluate algorithmic performance and biases should also be developed as part of the implementation process to allow iterative evaluation and improvement of algorithms or models, which could help ensure equitable performance across diverse cohorts 79 .”
Varga, A. N., Morel, A. E. G., Lokkerbol, J., van Dongen, J. M., van Tulder, M. W., & Bosmans, J. E. (2022) · Statistics in Medicine · doi:10.1002/sim.9628 · Cited by 32
Engels, E. A., Parsons, R., Besson, C., Morton, L. M., Enewold, L., Ricker, W., Yanik, E. L., Arem, H. E., Austin, A. A., & Pfeiffer, R. M. (2016) · Cancer Epidemiology Biomarkers & Prevention · doi:10.1158/1055-9965.epi-16-0212 · Cited by 28 · cited 1×
“Medication and procedure associations can be explored to understand patient populations and generate hypotheses that may impact these populations 41,42 .”
Yan, C., Zhang, X., Yang, Y., Kang, K., Were, M. C., Embí, P. J., Patel, M. B., Malin, B., Kho, A., & Chen, Y. (2023) · JAMA Network Open · doi:10.1001/jamanetworkopen.2023.36383 · Cited by 27 · cited 1×
“This study design was used to investigate statin and antidepressant exposure and COVID-19 mortality 90,91 , as well as antidepressant use in pregnancy outcomes 92 .”
Carrigan, G. J., Bradbury, B. D., Brookhart, M. A., Capra, W. B., Chia, V., Rothman, K. J., Sarsour, K., Taylor, M. D., & Brown, J. S. (2022) · Current Epidemiology Reports · doi:10.1007/s40471-022-00305-9 · Cited by 20 · cited 1×
“Since certain diseases may manifest concurrently and single datasets may have sample size limitations, some studies are expanding to include multiple cohorts or controls of various complexities 49,83–85 .”
Antidepressant use during pregnancy and the risk of preterm birth – a cohort study
Amit, G., Yanover, C., Bivas‐Benita, M., Girshovitz, I., Helbitz, T., Sudry, T., Bar, V., Tang, A., Le, B. L., Roger, J. M., et al. (2024) · npj Women s Health · doi:10.1038/s44294-024-00008-0 · Cited by 20
Self-citationWen, H., Chang, W.-P., Hsu, M., Ho, C.-H., & Chu, C. (2019) · JMIR Medical Informatics · doi:10.2196/12630 · Cited by 18 · cited 1×
“These datasets are now available as anonymized ‘controlled access’ datasets at many institutions and healthcare systems within the United States and worldwide 6,7 , and some—such as MIMIC 8 , AllofUs 9 and the UK Biobank 10 —are publicly accessible for research.”
Belthangady, C., Giampanis, S., Jankovic, I., Stedden, W., Alves, P., Chong, S., Knott, C., & Norgeot, B. (2022) · Nature Communications · doi:10.1038/s41467-022-33732-9 · Cited by 17 · cited 2×
“https://doi.org/10.1038/s41591-024-03074-8 For example, an approach that combines deep learning, causal inference and network meta-analysis applied to clinical data from a claims database enabled personalized treatment of type 2 diabetes 52 .”
InfluentialSelf-citationCostello, J. M., Steurer, M. A., Baer, R. J., Witte, J. S., & Jelliffe‐Pawlowski, L. L. (2022) · Paediatric and Perinatal Epidemiology · doi:10.1111/ppe.12820 · Cited by 14
Sperry, M. M., Oskotsky, T., Marić, I., Kaushal, S., Takeda, T., Horváth, V., Powers, R. K., Rodas, M. J., Furlong, B. A., Soong, M., et al. (2023) · PLoS Computational Biology · doi:10.1371/journal.pcbi.1011050 · Cited by 14 · cited 1×
“This study design was used to investigate statin and antidepressant exposure and COVID-19 mortality 90,91 , as well as antidepressant use in pregnancy outcomes 92 .”
Self-citationLou, S. S., Liu, H., Harford, D., Lu, C., & Kannampallil, T. (2023) · Journal of the American Medical Informatics Association · doi:10.1093/jamia/ocac239 · Cited by 13 · cited 1×
“Understanding and characterizing patterns in the data can give insight into affected populations, practices for treatment prescription and current clinical behaviors 26–28 , as well as identifying potential differences in subgroups and disparities in current care 29–31 .”
EHR Foundation Models Improve Robustness in the Presence of Temporal Distribution ShiftPreprint
Guo, L. L., Steinberg, E. H., Fleming, S. L., et al. (2022) · medRxiv · doi:10.1101/2022.04.15.22273900 · Cited by 12
Zaballa, O., Pérez, A., Gómez-Inhiesto, E., Acaiturri-Ayesta, T., & Lozano, J. A. (2020) · PLoS ONE · doi:10.1371/journal.pone.0244004 · Cited by 11
Woldemariam, S. R., Tang, A., Oskotsky, T., Yaffe, K. S., & Sirota, M. (2023) · Communications Medicine · doi:10.1038/s43856-023-00280-2 · Cited by 11 · cited 3×
“…or to identify controls of the same grouping (for example, black-identifying patients with cancer versus black-identifying healthy control patients) 35 , as opposed to a comparative analysis between two groups where one group is chosen as the reference group (black-identifying individuals versus…”
Self-citationSmith, M. A., Gigot, M., Harburn, A., Bednarz, L., Curtis, K., Mathew, J., & Farrar‐Edwards, D. (2023) · Journal of Clinical and Translational Science · doi:10.1017/cts.2022.521 · Cited by 10 · cited 1×
“Understanding and characterizing patterns in the data can give insight into affected populations, practices for treatment prescription and current clinical behaviors 26–28 , as well as identifying potential differences in subgroups and disparities in current care 29–31 .”
A Causal Roadmap for Generating High-Quality Real-World EvidencePreprint
Dang, L. E., Gruber, S., Lee, H., et al. (2023) · arXiv (Cornell University) · doi:10.48550/arxiv.2305.06850 · Cited by 10 · cited 1×
“Causal inference approaches can be applied to EHR data to further elucidate causal relationships 52,71,86–88 .”
Murugadoss, K., Rajasekharan, A., Malin, B., et al. (2020) · medRxiv · doi:10.1101/2020.12.22.20248270 · Cited by 8
Freitas, J. K. D., Johnson, K. W., Golden, E. A., et al. (2020) · medRxiv · doi:10.1101/2020.11.14.20231894 · Cited by 7
Sperry, M. M., Oskotsky, T., Marić, I., et al. (2022) · medRxiv · doi:10.1101/2022.04.12.22273802 · Cited by 5
Self-citationSteurer, M. A., Costello, J. M., Baer, R. J., Oltman, S. P., Feuer, S. K., Pacheco‐Werner, T., Rogers, E. E., Jankowska, M. M., Block, J., McCarthy, M., et al. (2020) · Paediatric and Perinatal Epidemiology · doi:10.1111/ppe.12652 · Cited by 3 · cited 1×
“In leveraging EHR data, one of the greatest opportunities includes improved understanding of the magnitude and implications of disease burdens, as well as patterns of healthcare service utilization and treat-ment within specific demographics or locations 17–20 .”
Gervasi, S. S., Chen, I. Y., Smith–McLallen, A., Sontag, D. A., Obermeyer, Z., Vennera, M., & Chawla, R. (2022) · DSpace@MIT (Massachusetts Institute of Technology) · Cited by 3
Karlin, L., Arnulf, B., Chevret, S., Robin, M., de Latour, R. P., Malphettes, M., Kabbara, N., Asli, B., Adès, L., Rocha, V., et al. (2008) · Blood · doi:10.1182/blood.v112.11.1133.1133 · Cited by 2
Roger, J. M., Xie, F., Costello, J. M., Tang, A., Liu, J., Oskotsky, T., Woldemariam, S. R., Kosti, I., Le, B., Snyder, M. P., et al. (2023) · Research Square · doi:10.21203/rs.3.rs-2631220/v2 · Cited by 2 · cited 1×
“Nevertheless, due to the convoluted nature of the data, conclusions from EHR studies are suggestive, and confidence in conclusions should be supported by similar conclusions across study sites, study designs and ‘-omics’ modalities 23,53 .”
Self-citation
Research works printed but not linked: 2
- Mullainathan, S. & Obermeyer, Z. Diagnosing physician error: a machine learning approach to low-value health care. Q. J. Econ. 137, 679–727 (2022). · DOI
- Yoon, J., Jordon, J. & van der Schaar, M. GAIN: missing data imputation using generative adversarial nets. Preprint at https://arxiv.org/abs/1806.02920v1 (2018). · arXiv
Cited by
Chicco, D., Sichenze, A., & Jurman, G. (2025) · BioData Mining · doi:10.1186/s13040-025-00465-6 · Cited by 277
Wang, K., Liu, F., Wu, W., Hu, C., Shen, X., Wang, M., Li, G., Zeng, F., Liu, L., Wong, I. N., et al. (2025) · Nature Medicine · doi:10.1038/s41591-025-04006-w · Cited by 45
“https://doi.org/10.1038/s41591-025-04006-w deep learning models have the capacity to extract such information, previous studies have largely focused on specific diseases within narrow age ranges 27–29 .”
InfluentialWu, J., He, K., Mao, R., Shang, X., & Cambria, E. (2025) · Information Fusion · doi:10.1016/j.inffus.2025.103283 · Cited by 40
Hu, Y., Yan, H. X., Liu, M., Gao, J., Xie, L., Zhang, C., Wei, L. S., Ding, Y., & Jiang, H. (2024) · BMC Medical Research Methodology · doi:10.1186/s12874-024-02422-z · Cited by 28
Teles, A. S., de Moura, I. R., Silva, F., Roberts, A., & Stahl, D. R. (2025) · Information Fusion · doi:10.1016/j.inffus.2025.102981 · Cited by 27
Metadata
- Authors
- Tang, A. S., Woldemariam, S. R., Miramontes, S., Norgeot, B., Oskotsky, T. T., & Sirota, M.
- Affiliations
- University of California, San Francisco (US)
- Corresponding author
- Sirota, Marina
- ORCID
- Venue
- Nature Medicine
- Published
- 2024-07
- Volume
- 30
- Issue
- 7
- Pages
- 1847-1855
- ISSN
- 1078-8956 (ISSN-L)
- 1546-170X
- OpenAlex
- W4400324874
- Cited by
- 87
- References
- 89
- Citations by year
- 2026: 41 · 2025: 41 · 2024: 5
- FWCI
- 2.99
- Type
- journal article
- Language
- en
- Open access
- closed
- Licence
- www.springernature.com/gp/researchers/text-and-data-mining (tdm)
- www.springernature.com/gp/researchers/text-and-data-mining (vor)
- Locations
- Nature Medicine (publishedVersion)
- PubMed (publishedVersion)
- Funders
- U.S. Department of Health & Human Services | NIH | Eunice Kennedy Shriver National Institute of Child Health and Human Development (P01HD106414)
- U.S. Department of Health & Human Services | NIH | National Institute of Arthritis and Musculoskeletal and Skin Diseases (P30 AR070155)
- U.S. Department of Health & Human Services | NIH | National Institute on Aging (R01AG060393)
- March of Dimes Foundation
- U.S. Department of Health & Human Services | National Institutes of Health (T32GM007618)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Machine Learning in Healthcare
- Health Sciences › Health Professions › Health Information Management › Electronic Health Records Systems
- Health Sciences › Health Professions › Health Information Management › Medical Coding and Health Information
- Keywords
- electronic health records, data bias, missing data, research design, hypothesis generation, hypothesis testing, machine learning, predictive modeling, association studies, precision medicine, population health, real-world clinical data
- MeSH terms
- Machine Learning, Humans, Medical Informatics, Biomedical Research, Precision Medicine, Electronic Health Records
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.
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
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.