Identification and validation of an explainable prediction model of acute kidney injury with prognostic implications in critically ill children: a prospective multicenter cohort study
Open accessThe DOI matches its Crossref record; 54 of the 54 references it cites were found in a registry or an index.
Abstract
Background: Acute kidney injury (AKI) is a common and serious organ dysfunction in critically ill children. Early identification and prediction of AKI are of great significance. However, current AKI criteria are insufficiently sensitive and specific, and AKI heterogeneity limits the clinical value of AKI biomarkers. This study aimed to establish and validate an explainable prediction model based on the machine learning (ML) approach for AKI, and assess its prognostic implications in children admitted to the pediatric intensive care unit (PICU). Methods: This multicenter prospective study in China was conducted on critically ill children for the derivation and validation of the prediction model. The derivation cohort, consisting of 957 children admitted to four independent PICUs from September 2020 to January 2021, was separated for training and internal validation, and an external data set of 866 children admitted from February 2021 to February 2022 was employed for external validation. AKI was defined based on serum creatinine and urine output using the Kidney Disease: Improving Global Outcome (KDIGO) criteria. With 33 medical characteristics easily obtained or evaluated during the first 24 h after PICU admission, 11 ML algorithms were used to construct prediction models. Several evaluation indexes, including the area under the receiver-operating-characteristic curve (AUC), were used to compare the predictive performance. The SHapley Additive exPlanation method was used to rank the feature importance and explain the final model. A probability threshold for the final model was identified for AKI prediction and subgrouping. Clinical outcomes were evaluated in various subgroups determined by a combination of the final model and KDIGO criteria. Findings: The random forest (RF) model performed best in discriminative ability among the 11 ML models. After reducing features according to feature importance rank, an explainable final RF model was established with 8 features. The final model could accurately predict AKI in both internal (AUC = 0.929) and external (AUC = 0.910) validations, and has been translated into a convenient tool to facilitate its utility in clinical settings. Critically ill children with a probability exceeding or equal to the threshold in the final model had a higher risk of death and multiple organ dysfunctions, regardless of whether they met the KDIGO criteria for AKI. Interpretation: Our explainable ML model was not only successfully developed to accurately predict AKI but was also highly relevant to adverse outcomes in individual children at an early stage of PICU admission, and it mitigated the concern of the "black-box" issue with an undirect interpretation of the ML technique. Funding: The National Natural Science Foundation of China, Jiangsu Province Science and Technology Support Program, Key talent of women's and children's health of Jiangsu Province, and Postgraduate Research & Practice Innovation Program of Jiangsu Province.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
Breiman, L. (2001) · Machine Learning · doi:10.1023/a:1010933404324 · Cited by 133,538 · cited 1×
“36 39 In this study, we employed the RF algo-rithm to develop a fi nal model with 8 features.”
MethodA Unified Approach to Interpreting Model Predictions
Lundberg, S., & Lee, S.-I. (2017) · arXiv · doi:10.48550/arxiv.1705.07874 · Cited by 23,707
DeLong, E. R., DeLong, D. M., & Clarke-Pearson, D. L. (1988) · Biometrics · doi:10.2307/2531595 · Cited by 22,649 · cited 1×
“The nonparametric method of Delong 32 was used to compare the difference between AUCs using MedCalc Version 19.6 (https://www.medcalc.org), and the features of the chosen ML model were gradually reduced until the AUC was dramatically decreased.”
MethodInfluentialNew Equations to Estimate GFR in Children with CKD
Schwartz, G. J., Mun[Combining Tilde]oz, A., Schneider, M. F., Mak, R. H., Kaskel, F., Warady, B. A., & Furth, S. L. (2009) · Journal of the American Society of Nephrology · doi:10.1681/asn.2008030287 · Cited by 3,777
Bellomo, R., Kellum, J. A., & Ronco, C. (2012) · The Lancet · doi:10.1016/s0140-6736(11)61454-2 · Cited by 2,598
Ronco, C., Bellomo, R., & Kellum, J. A. (2019) · The Lancet · doi:10.1016/s0140-6736(19)32563-2 · Cited by 1,910
PRISM III: An updated Pediatric Risk of Mortality score
Pollack, M. M., Patel, K. M., & Ruttimann, U. E. (1996) · Critical Care Medicine · doi:10.1097/00003246-199605000-00004 · Cited by 1,759
16., O. B. O. T. A. D. Q. I. W., Chawla, L. S., Bellomo, R., Bihorac, A., Goldstein, S. L., Siew, E. D., Bagshaw, S. M., Bittleman, D., Cruz, D., Endre, Z., et al. (2017) · Nature Reviews Nephrology · doi:10.1038/nrneph.2017.2 · Cited by 1,537
Epidemiology of Acute Kidney Injury in Critically Ill Children and Young Adults
Kaddourah, A., Basu, R. K., Bagshaw, S. M., & Goldstein, S. L. (2017) · New England Journal of Medicine · doi:10.1056/nejmoa1611391 · Cited by 1,075 · cited 1×
“1,2 Notably, AKI is associated with an increase in the length of ICU stays and ICU mortality.”
BackgroundKDOQI US Commentary on the 2012 KDIGO Clinical Practice Guideline for Acute Kidney Injury
Palevsky, P. M., Liu, K. D., Brophy, P. D., Chawla, L. S., Parikh, C. R., Thakar, C. V., Tolwani, A. J., Waikar, S. S., & Weisbord, S. D. (2013) · American Journal of Kidney Diseases · doi:10.1053/j.ajkd.2013.02.349 · Cited by 935 · cited 3×
“2,3 Instead of detecting AKI before it occurs, SCr only begins to rise after kidney injury has already set in, which may limit the timing of intervention for AKI.”
BackgroundSepsis associated acute kidney injury
Poston, J. T., & Koyner, J. L. (2019) · BMJ · doi:10.1136/bmj.k4891 · Cited by 818 · cited 1×
“To date, numerous studies have concentrated on AKI prediction, 33,34 but a single analyte seems to be dif fi cult to put into clinical practice due to the heterogeneous underlying pathophysiology of AKI.”
BackgroundOstermann, M., Zarbock, A., Goldstein, S., Kashani, K., Macedo, E., Murugan, R., Bell, M., Forni, L., Guzzi, L., Joannidis, M., et al. (2020) · JAMA Network Open · doi:10.1001/jamanetworkopen.2020.19209 · Cited by 677 · cited 1×
“54 In the present study, we used the RF model in a special and innovative way to de fi ne clinical subgroups, and the probability of 30.8%, as the optimal cutoff value that maximized sensitivity and speci fi city, was used as the threshold for subgrouping.”
MethodInfluentialHow Machine Learning Will Transform Biomedicine
Goecks, J., Jalili, V., Heiser, L. M., & Gray, J. W. (2020) · Cell · doi:10.1016/j.cell.2020.03.022 · Cited by 582 · cited 1×
“35 Among 11 ML models, the RF model had the best AUC value with a good net bene fi t and a high threshold probability in feature reduction.”
MethodObesity in the critically ill: a narrative review
Schetz, M., De Jong, A., Deane, A. M., Druml, W., Hemelaar, P., Pelosi, P., Pickkers, P., Reintam-Blaser, A., Roberts, J., Sakr, Y., et al. (2019) · Intensive Care Medicine · doi:10.1007/s00134-019-05594-1 · Cited by 563 · cited 1×
“44,48,51 Therefore, these clinical variables could contribute to the fi nal model, and the combination of them may be superior to a single marker in predicting AKI.”
BackgroundAcute kidney injury in the critically ill: an updated review on pathophysiology and management
Pickkers, P., Darmon, M., Hoste, E., Joannidis, M., Legrand, M., Ostermann, M., Prowle, J. R., Schneider, A., & Schetz, M. (2021) · Intensive Care Medicine · doi:10.1007/s00134-021-06454-7 · Cited by 549
Opening the black box: interpretable machine learning for geneticists
Azodi, C. B., Tang, J., & Shiu, S.-H. (2020) · Trends in Genetics · doi:10.1016/j.tig.2020.03.005 · Cited by 473
Prediction of the development of acute kidney injury following cardiac surgery by machine learning
Tseng, P.-Y., Chen, Y.-T., Wang, C.-H., Chiu, K.-M., Peng, Y.-S., Hsu, S.-P., Chen, K.-L., Yang, C.-Y., & Lee, O. K.-S. (2020) · Critical Care · doi:10.1186/s13054-020-03179-9 · Cited by 442 · cited 2×
“20,21 Regardless, scarce research has focused on AKI in critically ill pediatric patients, with the SHAP method being used to explain prediction models.”
BackgroundMethodMachine learning for the prediction of acute kidney injury in patients with sepsis
Yue, S., Li, S., Huang, X., Liu, J., Hou, X., Zhao, Y., Niu, D., Wang, Y., Tan, W., & Wu, J. (2022) · Journal of Translational Medicine · doi:10.1186/s12967-022-03364-0 · Cited by 342 · cited 1×
“20,21,52 As far as we know, the cause and course of AKI exhibit extensive heterogeneity, and the immature kidney function of children differs from that of adults.”
BackgroundThe development of a machine learning inpatient acute kidney injury prediction model.
Koyner, J. L., Carey, K. A., Edelson, D. P., & Churpek, M. M. (2018) · Critical Care Medicine · doi:10.1097/ccm.0000000000003123 · Cited by 323
Basu, R. K., Zappitelli, M., Brunner, L., Wang, Y., Wong, H. R., Chawla, L. S., Wheeler, D. S., & Goldstein, S. L. (2014) · Kidney International · doi:10.1038/ki.2013.349 · Cited by 297 · cited 9×
“Supplementary Table S6 lists several probability cutoff values that had a Youden index exceeding 0.8 in the training cohort, as well as their sensitivity, speci-fi city, PPV, and NPV for predicting AKI, severe AKI, and persistent AKI in 1823 critically ill children.”
MethodInfluentialHilmi, I. A., Damian, D., Al-Khafaji, A., Planinsic, R., Boucek, C., Sakai, T., Chang, C.-C. H., & Kellum, J. A. (2015) · British Journal of Anaesthesia · doi:10.1093/bja/aeu556 · Cited by 284 · cited 2×
“44,48,51 Therefore, these clinical variables could contribute to the fi nal model, and the combination of them may be superior to a single marker in predicting AKI.”
BackgroundManagement of patients at risk of acute kidney injury
Vanmassenhove, J., Kielstein, J., Jörres, A., & Biesen, W. V. (2017) · The Lancet · doi:10.1016/s0140-6736(17)31329-6 · Cited by 234 · cited 10×
“Supplementary Table S6 lists several probability cutoff values that had a Youden index exceeding 0.8 in the training cohort, as well as their sensitivity, speci-fi city, PPV, and NPV for predicting AKI, severe AKI, and persistent AKI in 1823 critically ill children.”
BackgroundInfluentialMachine learning model for early prediction of acute kidney injury (AKI) in pediatric critical care
Dong, J., Feng, T., Thapa-Chhetry, B., Cho, B. G., Shum, T., Inwald, D. P., Newth, C. J. L., & Vaidya, V. U. (2021) · Critical Care · doi:10.1186/s13054-021-03724-0 · Cited by 233 · cited 2×
“Although a few retrospective studies were conducted to identify AKI prediction models in hospitalized children, 14 – 16,53 these models performed less accurately than our fi nal model 14, The ML technique has been described as a “ black-box ” with little explanation about how predictions are derived.”
MethodResultFlechet, M., Güiza, F., Schetz, M., Wouters, P., Vanhorebeek, I., Derese, I., Gunst, J., Spriet, I., Casaer, M., Van den Berghe, G., et al. (2017) · Intensive Care Medicine · doi:10.1007/s00134-017-4678-3 · Cited by 169
Basu, R. K., Kaddourah, A., Goldstein, S. L., Akcan‐Arikan, A., Arnold, M. A., Cruz, C., Goldsworthy, M., Jaimon, N., Alexander, S. I., Festa, M. S., et al. (2017) · The Lancet Child & Adolescent Health · doi:10.1016/s2352-4642(17)30181-5 · Cited by 165 · cited 1×
“In addition, the renal angina index (RAI), as a predictive indicator for AKI, 24,25 was determined using the data obtained during the fi rst 24 h of PICU admission, and the predictive value of the RAI was subsequently compared to that of the prediction model.”
MethodAcute Kidney Injury in Children With Type 1 Diabetes Hospitalized for Diabetic Ketoacidosis
Hursh, B. E., Ronsley, R., Islam, N., Mammen, C., & Panagiotopoulos, C. (2017) · JAMA Pediatrics · doi:10.1001/jamapediatrics.2017.0020 · Cited by 162 · cited 1×
“26,27 Second, we built the AKI prediction model in heterogeneous PICU cohorts, regardless of the causes of AKI.”
MethodLi, Y., Sperrin, M., Ashcroft, D. M., & van Staa, T. P. (2020) · BMJ · doi:10.1136/bmj.m3919 · Cited by 135 · cited 1×
“40 The SHAP method was employed to assist feature selection.”
MethodSelf-citationBerkelmans, G. F. N., Read, S. H., Gudbjörnsdottir, S., Wild, S. H., Franzen, S., van der Graaf, Y., Eliasson, B., Visseren, F. L. J., Paynter, N. P., & Dorresteijn, J. A. N. (2022) · Journal of Clinical Epidemiology · doi:10.1016/j.jclinepi.2022.01.011 · Cited by 125
Internal and External Validation of a Machine Learning Risk Score for Acute Kidney Injury
Churpek, M. M., Carey, K. A., Edelson, D. P., Singh, T., Astor, B. C., Gilbert, E. R., Winslow, C., Shah, N., Afshar, M., & Koyner, J. L. (2020) · JAMA Network Open · doi:10.1001/jamanetworkopen.2020.12892 · Cited by 119
Palmieri, T. L., Lavrentieva, A., & Greenhalgh, D. (2009) · Intensive Care Medicine · doi:10.1007/s00134-009-1638-6 · Cited by 116
Sutherland, S. M., Chawla, L. S., Kane-Gill, S. L., Hsu, R. K., Kramer, A. A., Goldstein, S. L., Kellum, J. A., Ronco, C., Bagshaw, S. M., & on behalf of the 15 ADQI Consensus Group (2016) · Canadian Journal of Kidney Health and Disease · doi:10.1186/s40697-016-0099-4 · Cited by 115 · cited 1×
“18 To overcome these issues of the “ black-box ” , the SHapley Additive exPla-nation (SHAP) method was utilized to explain the ML models and visualize individual variable predictions.”
MethodDevelopment of a Multicenter Ward–Based AKI Prediction Model
Koyner, J. L., Adhikari, R., Edelson, D. P., & Churpek, M. M. (2016) · Clinical Journal of the American Society of Nephrology · doi:10.2215/cjn.00280116 · Cited by 113
Zhang, Y., Yang, D. H., Liu, Z., Chen, C., Ge, M., Li, X., Luo, T., Wu, Z., Shi, C., Wang, B., et al. (2021) · Journal of Translational Medicine · doi:10.1186/s12967-021-02990-4 · Cited by 105 · cited 2×
“20,21 Regardless, scarce research has focused on AKI in critically ill pediatric patients, with the SHAP method being used to explain prediction models.”
BackgroundMethodSelf-citationKang, M.-W., Kim, J., Kim, D., Oh, K., Joo, K., Kim, Y., & Han, S. (2020) · Critical Care · doi:10.1186/s13054-020-2752-7 · Cited by 105
A New Criterion for Pediatric AKI Based on the Reference Change Value of Serum Creatinine
Xu, X. P., Nie, S., Zhang, A., Jianhua, M., Liu, H., Xia, H., Xu, H. M., Liu, Z., Feng, S., Zhou, W., et al. (2018) · Journal of the American Society of Nephrology · doi:10.1681/asn.2018010090 · Cited by 99
Self-citationMyers, S. R., Glaser, N. S., Trainor, J. L., Nigrovic, L. E., Garro, A. C., Tzimenatos, L. S., Quayle, K. S., Kwok, M. Y., Rewers, A., Stoner, M. J., et al. (2020) · JAMA Network Open · doi:10.1001/jamanetworkopen.2020.25481 · Cited by 98 · cited 1×
“26,27 Second, we built the AKI prediction model in heterogeneous PICU cohorts, regardless of the causes of AKI.”
MethodMatsa, R., Ashley, E., Sharma, V., Walden, A. P., & Keating, L. (2014) · Critical Care · doi:10.1186/cc13958 · Cited by 65 · cited 1×
“55,56 The fi nal model still performed well for predicting AKI that occurred on days 2 – 7 of the PICU stay.”
MethodA prediction and interpretation framework of acute kidney injury in critical care
Gong, K., Lee, H. K., Yu, K., Xie, X., & Li, J. (2020) · Journal of Biomedical Informatics · doi:10.1016/j.jbi.2020.103653 · Cited by 57
Prediction of acute kidney injury in ICU with gradient boosting decision tree algorithms
Gao, W., Wang, J., Zhou, L., Luo, Q., Lao, Y., Lyu, H., & Guo, S. (2021) · Computers in Biology and Medicine · doi:10.1016/j.compbiomed.2021.105097 · Cited by 56
A Time-Updated, Parsimonious Model to Predict AKI in Hospitalized Children
Sandokji, I. A., Yamamoto, Y., Biswas, A., Arora, T., Ugwuowo, U. C., Simonov, M., Saran, I., Martin, M. C., Testani, J. M., Mansour, S. G., et al. (2020) · Journal of the American Society of Nephrology · doi:10.1681/asn.2019070745 · Cited by 53
Sanchez-Pinto, L. N., & Khemani, R. G. (2016) · Pediatric Critical Care Medicine · doi:10.1097/pcc.0000000000000750 · Cited by 52 · cited 2×
“Although a few retrospective studies were conducted to identify AKI prediction models in hospitalized children, 14 – 16,53 these models performed less accurately than our fi nal model 14, The ML technique has been described as a “ black-box ” with little explanation about how predictions are derived.”
MethodResultValidation of a Machine Learning Model to Predict Childhood Lead Poisoning
Potash, E., Ghani, R., Walsh, J., Jorgensen, E., Lohff, C. J., Prachand, N., & Mansour, R. (2020) · JAMA Network Open · doi:10.1001/jamanetworkopen.2020.12734 · Cited by 52 · cited 1×
“36 39 In this study, we employed the RF algo-rithm to develop a fi nal model with 8 features.”
MethodShawwa, K., Ghosh, E., Lanius, S., Schwager, E., Eshelman, L. J., & KASHANI, K. B. (2020) · Clinical Kidney Journal · doi:10.1093/ckj/sfaa145 · Cited by 47 · cited 1×
“17 Several studies have ranked the feature importance of ML models for AKI prediction in critically ill adult patients, 7 – 12 but only two of those explained the model at the individual adult level.”
BackgroundWang, L., McGregor, T. L., Jones, D. P., Bridges, B. C., Fleming, G. M., Shirey-Rice, J. K., McLemore, M. F., Chen, L., Weitkamp, A. O., Byrne, D. W., et al. (2017) · Pediatric Research · doi:10.1038/pr.2017.116 · Cited by 47
Alfieri, F., Ancona, A., Tripepi, G. L., Rubeis, A., Arjoldi, N., Finazzi, S., Cauda, V., & Fagugli, R. M. (2023) · PLoS ONE · doi:10.1371/journal.pone.0287398 · Cited by 39
Kuai, Y., Li, M., Chen, J., Jiang, Z., Bai, Z., Huang, H., Wei, L., Liu, N., Li, X., Lü, G., et al. (2022) · Critical Care · doi:10.1186/s13054-022-04083-0 · Cited by 35 · cited 1×
“41,42 We used eGFR instead of SCr in the development of model due to the fact that eGFR gives more information than just SCr alone and re fl ects the physical conditions of speci fi c children of different ages.”
MethodSelf-citationZhang, X., Yue, P., Zhang, J., Yang, M., Chen, J., Zhang, B., Luo, W. H., Wang, M., Da, Z., Lin, Y., et al. (2022) · EClinicalMedicine · doi:10.1016/j.eclinm.2022.101431 · Cited by 35
Self-citationClinical features and outcomes of acute kidney injury among patients with acute hepatitis A
Choi, H. K., Song, Y. G., Han, S. H., Ku, N. S., Jeong, S. J., Baek, J. H., Kim, H., Kim, S. B., Kim, C. O., Kim, J. M., et al. (2011) · Journal of Clinical Virology · doi:10.1016/j.jcv.2011.07.013 · Cited by 27
Xu, Z., Zhang, Y., Zhang, C., Xiong, F., Zhang, J., & Xiong, J. (2022) · Aging and Disease · doi:10.14336/ad.2021.1125 · Cited by 22
Prediction of urinary dickkopf-3 for AKI, sepsis-associated AKI, and PICU mortality in children
Hu, J., Zhou, Y., Huang, H., Kuai, Y., Chen, J., Bai, Z., Li, X., & Li, Y. (2022) · Pediatric Research · doi:10.1038/s41390-022-02269-4 · Cited by 22 · cited 2×
“44,48,51 Therefore, these clinical variables could contribute to the fi nal model, and the combination of them may be superior to a single marker in predicting AKI.”
BackgroundSelf-citationJiang, X., Hu, Y., Guo, S., Du, C., & Cheng, X. (2022) · Scientific Reports · doi:10.1038/s41598-022-21428-5 · Cited by 21 · cited 1×
“10,11 The Acute Dialysis Quality Initiative (ADQI) group also advised that the prediction model should present information about clinical variables in fl uencing the risk of AKI.”
MethodIkemura, N., Spertus, J. A., Kimura, T., Katsumata, Y., Fujisawa, T., Ueda, I., Fukuda, K., Takatsuki, S., & Kohsaka, S. (2021) · Journal of the American Heart Association · doi:10.1161/jaha.120.019983 · Cited by 14
He, L., Liang, S., Liang, Y., Fang, M., Li, J., Deng, J. H., Fang, H., Li, Y., Jiang, X., & Chen, C. (2022) · Internal and Emergency Medicine · doi:10.1007/s11739-022-03187-3 · Cited by 14
Self-citationHypertension, Blood Pressure Variability, and Acute Kidney Injury in Hospitalized Children
Nugent, J. T., Ghazi, L., Yamamoto, Y., Bakhoum, C. Y., Wilson, F. P., & Greenberg, J. H. (2023) · Journal of the American Heart Association · doi:10.1161/jaha.122.029059 · Cited by 8 · cited 10×
“Supplementary Table S6 lists several probability cutoff values that had a Youden index exceeding 0.8 in the training cohort, as well as their sensitivity, speci-fi city, PPV, and NPV for predicting AKI, severe AKI, and persistent AKI in 1823 critically ill children.”
ResultInfluential
Cited by
Huang, L., Shi, Z., Tang, F., & Wang, H. (2026) · IEEE Transactions on Biomedical Engineering · doi:10.1109/tbme.2025.3590149 · Cited by 2
Metadata
- Authors
- Hu, J., Xu, J., Li, M., Jiang, Z., Mao, J., Feng, L., Miao, K., Li, H., Chen, J., Bai, Z., Li, X., Lu, G., & Li, Y.
- Venue
- eClinicalMedicine
- Publisher
- Elsevier BV
- Published
- 2024-02
- Volume
- 68
- Pages
- 102409
- ISSN
- 2589-5370 (ISSN-L)
- OpenAlex
- W4390599059
- Cited by
- 213
- References
- 54
- FWCI
- 52.78
- Type
- journal article
- Language
- en
- Open access
- gold
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.