Mining Electronic Health Records for Drugs Associated With 28-day Mortality in COVID-19: Pharmacopoeia-wide Association Study (PharmWAS)
Ivan Lerner, MD Arnaud Serret-Larmande, MD Bastien Rance, PhD Nicolas Garcelon, PhD; Anita Burgun, PhD Laurent Chouchana, PharmD Antoine Neuraz
JMIR Medical Informatics, doi:10.2196/35190
Background: Patients hospitalized for a given condition may be receiving other treatments for other contemporary conditions or comorbidities. The use of such observational clinical data for pharmacological hypothesis generation is appealing in the context of an emerging disease but particularly challenging due to the presence of drug indication bias. Objective: With this study, our main objective was the development and validation of a fully data-driven pipeline that would address this challenge. Our secondary objective was to generate pharmacological hypotheses in patients with COVID-19 and demonstrate the clinical relevance of the pipeline.
Methods: We developed a pharmacopeia-wide association study (PharmWAS) pipeline inspired from the PheWAS methodology, which systematically screens for associations between the whole pharmacopeia and a clinical phenotype. First, a fully data-driven procedure based on adaptive least absolute shrinkage and selection operator (LASSO) determined drug-specific adjustment sets. Second, we computed several measures of association, including robust methods based on propensity scores (PSs) to control indication bias. Finally, we applied the Benjamini and Hochberg procedure of the false discovery rate (FDR). We applied this method in a multicenter retrospective cohort study using electronic medical records from 16 university hospitals of the Greater Paris area. We included all adult patients between 18 and 95 years old hospitalized in conventional wards for COVID-19 between February 1, 2020, and June 15, 2021. We investigated the association between drug prescription within 48 hours from admission and 28-day mortality. We validated our data-driven pipeline against a knowledge-based pipeline on 3 treatments of reference, for which experts agreed on the expected association with mortality. We then demonstrated its clinical relevance by screening all drugs prescribed in more than 100 patients to generate pharmacological hypotheses. Results: A total of 5783 patients were included in the analysis. The median age at admission was 69.2 (IQR 56.7-81.1) years, and 3390 (58.62%) of the patients were male. The performance of our automated pipeline was comparable or better for controlling bias than the knowledge-based adjustment set for 3 reference drugs: dexamethasone, phloroglucinol, and paracetamol. After
Authors' Contributions IL contributed to conceptualization, data curation, formal analysis, methodology, software, and writing (original draft). ASL contributed to methodology and writing (reviewing and editing). LC contributed to conceptualization, formal analysis, and writing (reviewing and editing). BR and NG contributed to validation, and writing (reviewing and editing). AB contributed to project administration, resources, supervision, validation, and writing (reviewing and editing). AN contributed to conceptualization, data curation, formal analysis, software, resources, supervision, validation, writing (original draft), and writing (reviewing and editing).
Conflicts of Interest None declared.
Multimedia Appendix 1 Supplementary materials.
References
Austin, Grootendorst, Anderson, A comparison of the ability of different propensity score models to balance measured variables between treated and untreated subjects: a Monte Carlo study, Stat Med,
doi:10.1002/sim.2580Austin, Optimal caliper widths for propensity-score matching when estimating differences in means and differences in proportions in observational studies, Pharm Stat,
doi:10.1002/pst.433Bejan, Cahill, Staso, Choi, Peterson et al., DrugWAS: drug-wide association studies for COVID-19 drug repurposing, Clin Pharmacol Ther,
doi:10.1002/cpt.2376Boutron, Chaimani, Devane, Meerpohl, Tovey et al., Interventions for preventing and treating COVID-19: protocol for a living mapping of research and a living systematic review, Cochrane Database Syst Rev,
doi:10.1002/14651858.CD013769Brookhart, Schneeweiss, Rothman, Glynn, Avorn et al., Variable selection for propensity score models, Am J Epidemiol,
doi:10.1093/aje/kwj149Castelnuovo, Bonaccio, Costanzo, Gialluisi, Antinori et al., Common cardiovascular risk factors and in-hospital mortality in 3,894 patients with COVID-19: survival analysis and machine learning-based findings from the multicentre Italian CORIST Study, Nutr Metab Cardiovasc Dis,
doi:10.1016/j.numecd.2020.07.031Choi, Carroll, Beck, Mosley, Roden et al., Evaluating statistical approaches to leverage large clinical datasets for uncovering therapeutic and adverse medication effects, Bioinformatics,
doi:10.1093/bioinformatics/bty306Chouchana, Beeker, Garcelon, Rance, Paris et al., Correction to: association of antihypertensive agents with the risk of in-hospital death in patients with Covid-19, Cardiovasc Drugs Ther,
doi:10.1101/2020.11.23.20237362Chouchana, Beeker, Garcelon, Rance, Paris et al., Universities/Inserm COVID-19 research collaboration‚ AP-HP Covid CDR Initiative‚"Entrepôt de Données de Santé" AP-HP Consortium, Cardiovasc Drugs Ther,
doi:10.1007/s10557-021-07155-5Cohen, Davidson, Gallus, Lassen, Prins et al., Efficacy and safety of fondaparinux for the prevention of venous thromboembolism in older acute medical patients: randomised placebo controlled trial, BMJ,
doi:10.1136/bmj.38733.466748.7cCore, R: A Language and Environment for Statistical Computing
Daniel, Serre, Orlova, Bréant, Paris et al., Initializing a hospital-wide data quality program. The AP-HP experience, Comput Methods Programs Biomed,
doi:10.1016/j.cmpb.2018.10.016Denny, Bastarache, Ritchie, Carroll, Zink et al., Systematic comparison of phenome-wide association study of electronic medical record data and genome-wide association study data, Nat Biotechnol,
doi:10.3410/f.718185859.793488460Denny, Ritchie, Basford, Pulley, Bastarache et al., PheWAS: demonstrating the feasibility of a phenome-wide scan to discover gene-disease associations, Bioinformatics,
doi:10.1093/bioinformatics/btq126Duarte, Pelorosso, Nicolosi, Salgado, Vetulli et al., Telmisartan for treatment of Covid-19 patients: an open multicenter randomized clinical trial, EClinicalMedicine,
doi:10.1016/j.eclinm.2021.100962Ekström, Bornefalk-Hermansson, Abernethy, Currow, Safety of benzodiazepines and opioids in very severe respiratory disease: national prospective study, BMJ,
doi:10.1136/bmj.g445Friedman, Hastie, Tibshirani, Regularization paths for generalized linear models via coordinate descent, J Stat Softw,
doi:10.18637/jss.v033.i01Gue, Tennyson, Gao, Ren, Kanji et al., Development of a novel risk score to predict mortality in patients admitted to hospital with COVID-19, Sci Rep,
doi:10.1038/s41598-020-78505-wGérard, Romani, Fresse, Viard, Parassol et al., French Network of Pharmacovigilance Centers. "Off-label" use of hydroxychloroquine, azithromycin, lopinavir-ritonavir and chloroquine in COVID-19: a survey of cardiac adverse drug reactions by the French Network of Pharmacovigilance Centers, Therapie,
doi:10.1016/j.therap.2020.05.002Ho, Imai, King, Stuart, MatchIt: nonparametric preprocessing for parametric causal inference, J Stat Softw,
doi:10.18637/jss.v042.i08Kohane, Aronow, Avillach, Beaulieu-Jones, Bellazzi et al., What every reader should know about studies using electronic health record data but may be afraid to ask, J Med Internet Res,
doi:10.2196/22219Lerner, Pharmwas, GitLab
Li, Raghunathan, Db, Large-sample significance levels from multiply imputed data using moment-based statistics and an F reference distribution, J Am Stat Assoc,
doi:10.2307/2290525Neuraz, Chouchana, Malamut, Beller, Roche et al., Phenome-wide association studies on a quantitative trait: application to TPMT enzyme activity and thiopurine therapy in pharmacogenomics, PLoS Comput Biol,
doi:10.1371/journal.pcbi.1003405Neuraz, Lerner, Digan, Paris, Tsopra et al., Natural language processing for rapid response to emergent diseases: case study of calcium channel blockers and hypertension in the COVID-19 pandemic, J Med Internet Res,
doi:10.2196/20773Nopp, Moik, Jilma, Pabinger, Ay, Risk of venous thromboembolism in patients with COVID-19: a systematic review and meta-analysis, Res Pract Thromb Haemost,
doi:10.1002/rth2.12439Patel, Ioannidis, Assessment of vibration of effects due to model specification can demonstrate the instability of observational associations, J Clin Epidemiol,
doi:10.1016/j.jclinepi.2015.05.029Patel, Ji, Sundquist, Ioannidis, Sundquist, Systematic assessment of pharmaceutical prescriptions in association with cancer risk: a method to conduct a population-wide medication-wide longitudinal study, Sci Rep,
doi:10.1038/srep31308Ryan, Madigan, Stang, Schuemie, Hripcsak, Medication-wide association studies, CPT Pharmacomet Syst Pharmacol,
doi:10.1038/psp.2013.52Samama, Cohen, Darmon, Desjardins, Eldor et al., A comparison of enoxaparin with placebo for the prevention of venous thromboembolism in acutely ill medical patients, N Engl J Med,
doi:10.1056/nejm199909093411103Stuart, Lee, Leacy, Prognostic score-based balance measures can be a useful diagnostic for propensity score methods in comparative effectiveness research, J Clin Epidemiol,
doi:10.1016/j.jclinepi.2013.01.013Stürmer, Rothman, Avorn, Glynn, Treatment effects in the presence of unmeasured confounding: dealing with observations in the tails of the propensity score distribution; a simulation study, Am J Epidemiol,
doi:10.1093/aje/kwq198Stürmer, Webster-Clark, Lund, Ellis, Lunt, Propensity score weighting and trimming strategies for reducing variance and bias of treatment effect estimates: a simulation study, Am J Epidemiol,
doi:10.1093/aje/kwab041Van Buuren, Flexible Imputation of Missing Data, Second Edition
Walker, Patrick, Lauer, Hornbrook, Marin et al., A tool for assessing the feasibility of comparative effectiveness research, CER,
doi:10.2147/cer.s40357Wang, Cao, Zhang, Yang, Liu et al., Remdesivir and chloroquine effectively inhibit the recently emerged novel coronavirus (2019-nCoV) in vitro, Cell Res,
doi:10.1038/s41422-020-0282-0Williamson, Walker, Bhaskaran, Bacon, Bates et al., Factors associated with COVID-19-related death using OpenSAFELY, Nature,
doi:10.1038/s41586-020-2521-4Witte, Didelez, Covariate selection strategies for causal inference: classification and comparison, Biom J,
doi:10.1002/bimj.201700294Wong, Mackenna, Morton, Schultze, Walker et al., Use of non-steroidal anti-inflammatory drugs and risk of death from COVID-19: an OpenSAFELY cohort analysis based on two cohorts, Ann Rheum Dis,
doi:10.1136/annrheumdis-2020-219517Zhou, Tong, Li, Thomas, PSweight: An R Package for Propensity Score Weighting Analysis
{ 'indexed': {'date-parts': [[2022, 4, 12]], 'date-time': '2022-04-12T18:41:20Z', 'timestamp': 1649788880603},
'reference-count': 53,
'publisher': 'JMIR Publications Inc.',
'issue': '3',
'content-domain': {'domain': [], 'crossmark-restriction': False},
'abstract': '<jats:sec>\n'
' <jats:title>Background</jats:title>\n'
' <jats:p>Patients hospitalized for a given condition may be receiving other '
'treatments for other contemporary conditions or comorbidities. The use of such observational '
'clinical data for pharmacological hypothesis generation is appealing in the context of an '
'emerging disease but particularly challenging due to the presence of drug indication '
'bias.</jats:p>\n'
' </jats:sec>\n'
' <jats:sec>\n'
' <jats:title>Objective</jats:title>\n'
' <jats:p>With this study, our main objective was the development and validation of '
'a fully data-driven pipeline that would address this challenge. Our secondary objective was '
'to generate pharmacological hypotheses in patients with COVID-19 and demonstrate the clinical '
'relevance of the pipeline.</jats:p>\n'
' </jats:sec>\n'
' <jats:sec>\n'
' <jats:title>Methods</jats:title>\n'
' <jats:p>We developed a pharmacopeia-wide association study (PharmWAS) pipeline '
'inspired from the PheWAS methodology, which systematically screens for associations between '
'the whole pharmacopeia and a clinical phenotype. First, a fully data-driven procedure based '
'on adaptive least absolute shrinkage and selection operator (LASSO) determined drug-specific '
'adjustment sets. Second, we computed several measures of association, including robust '
'methods based on propensity scores (PSs) to control indication bias. Finally, we applied the '
'Benjamini and Hochberg procedure of the false discovery rate (FDR). We applied this method in '
'a multicenter retrospective cohort study using electronic medical records from 16 university '
'hospitals of the Greater Paris area. We included all adult patients between 18 and 95 years '
'old hospitalized in conventional wards for COVID-19 between February 1, 2020, and June 15, '
'2021. We investigated the association between drug prescription within 48 hours from '
'admission and 28-day mortality. We validated our data-driven pipeline against a '
'knowledge-based pipeline on 3 treatments of reference, for which experts agreed on the '
'expected association with mortality. We then demonstrated its clinical relevance by screening '
'all drugs prescribed in more than 100 patients to generate pharmacological '
'hypotheses.</jats:p>\n'
' </jats:sec>\n'
' <jats:sec>\n'
' <jats:title>Results</jats:title>\n'
' <jats:p>A total of 5783 patients were included in the analysis. The median age at '
'admission was 69.2 (IQR 56.7-81.1) years, and 3390 (58.62%) of the patients were male. The '
'performance of our automated pipeline was comparable or better for controlling bias than the '
'knowledge-based adjustment set for 3 reference drugs: dexamethasone, phloroglucinol, and '
'paracetamol. After correction for multiple testing, 4 drugs were associated with increased '
'in-hospital mortality. Among these, diazepam and tramadol were the only ones not discarded by '
'automated diagnostics, with adjusted odds ratios of 2.51 (95% CI 1.52-4.16, Q=.01) and 1.94 '
'(95% CI 1.32-2.85, Q=.02), respectively.</jats:p>\n'
' </jats:sec>\n'
' <jats:sec>\n'
' <jats:title>Conclusions</jats:title>\n'
' <jats:p>Our innovative approach proved useful in generating pharmacological '
'hypotheses in an outbreak setting, without requiring a priori knowledge of the disease. Our '
'systematic analysis of early prescribed treatments from patients hospitalized for COVID-19 '
'showed that diazepam and tramadol are associated with increased 28-day mortality. Whether '
'these drugs could worsen COVID-19 needs to be further assessed.</jats:p>\n'
' </jats:sec>',
'DOI': '10.2196/35190',
'type': 'journal-article',
'created': {'date-parts': [[2022, 3, 11]], 'date-time': '2022-03-11T18:02:31Z', 'timestamp': 1647021751000},
'page': 'e35190',
'source': 'Crossref',
'is-referenced-by-count': 1,
'title': 'Mining Electronic Health Records for Drugs Associated With 28-day Mortality in COVID-19: '
'Pharmacopoeia-wide Association Study (PharmWAS)',
'prefix': '10.2196',
'volume': '10',
'author': [ { 'ORCID': 'http://orcid.org/0000-0002-5466-1707',
'authenticated-orcid': False,
'given': 'Ivan',
'family': 'Lerner',
'sequence': 'first',
'affiliation': []},
{ 'ORCID': 'http://orcid.org/0000-0002-2642-2514',
'authenticated-orcid': False,
'given': 'Arnaud',
'family': 'Serret-Larmande',
'sequence': 'additional',
'affiliation': []},
{ 'ORCID': 'http://orcid.org/0000-0003-4417-1197',
'authenticated-orcid': False,
'given': 'Bastien',
'family': 'Rance',
'sequence': 'additional',
'affiliation': []},
{ 'ORCID': 'http://orcid.org/0000-0002-3326-2811',
'authenticated-orcid': False,
'given': 'Nicolas',
'family': 'Garcelon',
'sequence': 'additional',
'affiliation': []},
{ 'ORCID': 'http://orcid.org/0000-0001-6855-4366',
'authenticated-orcid': False,
'given': 'Anita',
'family': 'Burgun',
'sequence': 'additional',
'affiliation': []},
{ 'ORCID': 'http://orcid.org/0000-0002-9626-3571',
'authenticated-orcid': False,
'given': 'Laurent',
'family': 'Chouchana',
'sequence': 'additional',
'affiliation': []},
{ 'ORCID': 'http://orcid.org/0000-0001-7142-6728',
'authenticated-orcid': False,
'given': 'Antoine',
'family': 'Neuraz',
'sequence': 'additional',
'affiliation': []}],
'member': '1010',
'published-online': {'date-parts': [[2022, 3, 30]]},
'reference': [ { 'key': 'ref1',
'unstructured': 'COVID-19 Dashboard by the Center for Systems Science and Engineering '
'(CSSE) at Johns Hopkins University '
'(JHU)2021-08-27https://coronavirus.jhu.edu/map.html'},
{'key': 'ref2', 'doi-asserted-by': 'publisher', 'DOI': '10.1002/14651858.CD013769'},
{'key': 'ref3', 'doi-asserted-by': 'publisher', 'DOI': '10.1056/nejmoa2021436'},
{'key': 'ref4', 'doi-asserted-by': 'publisher', 'DOI': '10.1016/j.therap.2020.05.002'},
{'key': 'ref5', 'doi-asserted-by': 'publisher', 'DOI': '10.1136/annrheumdis-2020-219517'},
{'key': 'ref6', 'doi-asserted-by': 'publisher', 'DOI': '10.1007/s10557-021-07155-5'},
{'key': 'ref7', 'doi-asserted-by': 'publisher', 'DOI': '10.1016/j.ijantimicag.2020.106248'},
{'key': 'ref8', 'doi-asserted-by': 'publisher', 'DOI': '10.1016/j.antiviral.2020.104787'},
{'key': 'ref9', 'doi-asserted-by': 'publisher', 'DOI': '10.1038/s41422-020-0282-0'},
{'key': 'ref10', 'doi-asserted-by': 'publisher', 'DOI': '10.1016/j.mayocp.2020.05.036'},
{'key': 'ref11', 'doi-asserted-by': 'publisher', 'DOI': '10.1002/rth2.12439'},
{'key': 'ref12', 'doi-asserted-by': 'publisher', 'DOI': '10.1136/bmj.38733.466748.7c'},
{'key': 'ref13', 'doi-asserted-by': 'publisher', 'DOI': '10.1056/nejm199909093411103'},
{'key': 'ref14', 'doi-asserted-by': 'publisher', 'DOI': '10.1093/bioinformatics/btq126'},
{'key': 'ref15', 'doi-asserted-by': 'publisher', 'DOI': '10.3410/f.718185859.793488460'},
{'key': 'ref16', 'doi-asserted-by': 'publisher', 'DOI': '10.1371/journal.pcbi.1003405'},
{'key': 'ref17', 'doi-asserted-by': 'publisher', 'DOI': '10.1038/psp.2013.52'},
{'key': 'ref18', 'doi-asserted-by': 'publisher', 'DOI': '10.1038/srep31308'},
{'key': 'ref19', 'doi-asserted-by': 'publisher', 'DOI': '10.1093/bioinformatics/bty306'},
{'key': 'ref20', 'doi-asserted-by': 'publisher', 'DOI': '10.1002/cpt.2376'},
{'key': 'ref21', 'doi-asserted-by': 'publisher', 'DOI': '10.1002/sim.2580'},
{'key': 'ref22', 'doi-asserted-by': 'publisher', 'DOI': '10.1093/aje/kwj149'},
{'key': 'ref23', 'doi-asserted-by': 'publisher', 'DOI': '10.1002/bimj.201700294'},
{'key': 'ref24', 'doi-asserted-by': 'publisher', 'DOI': '10.1093/aje/kwq198'},
{'key': 'ref25', 'doi-asserted-by': 'publisher', 'DOI': '10.1093/aje/kwab041'},
{'key': 'ref26', 'doi-asserted-by': 'publisher', 'DOI': '10.1097/00001648-199311000-00013'},
{'key': 'ref27', 'doi-asserted-by': 'publisher', 'DOI': '10.2147/cer.s40357'},
{'key': 'ref28', 'doi-asserted-by': 'publisher', 'DOI': '10.1016/j.cmpb.2018.10.016'},
{ 'key': 'ref29',
'unstructured': 'Anatomical Therapeutic Chemical (ATC) '
'Classification2021-05-02https://www.who.int/tools/atc-ddd-toolkit/atc-classification'},
{'key': 'ref30', 'doi-asserted-by': 'publisher', 'DOI': '10.1198/016214506000000735'},
{'key': 'ref31', 'doi-asserted-by': 'publisher', 'DOI': '10.1002/pst.433'},
{'key': 'ref32', 'doi-asserted-by': 'publisher', 'DOI': '10.1016/j.jclinepi.2013.01.013'},
{ 'key': 'ref33',
'doi-asserted-by': 'publisher',
'DOI': '10.1111/j.2517-6161.1995.tb02031.x'},
{'key': 'ref34', 'doi-asserted-by': 'publisher', 'DOI': '10.1016/j.jinf.2020.04.021'},
{'key': 'ref35', 'doi-asserted-by': 'publisher', 'DOI': '10.1038/s41586-020-2521-4'},
{'key': 'ref36', 'doi-asserted-by': 'publisher', 'DOI': '10.1038/s41598-020-78505-w'},
{'key': 'ref37', 'doi-asserted-by': 'publisher', 'DOI': '10.1016/j.numecd.2020.07.031'},
{ 'key': 'ref38',
'doi-asserted-by': 'crossref',
'author': 'van Buuren, S',
'year': '2018',
'journal-title': 'Flexible Imputation of Missing Data, Second Edition',
'DOI': '10.1201/9780429492259'},
{'key': 'ref39', 'doi-asserted-by': 'publisher', 'DOI': '10.2307/2290525'},
{ 'key': 'ref40',
'year': '2018',
'journal-title': 'R: A Language and Environment for Statistical Computing'},
{'key': 'ref41', 'doi-asserted-by': 'publisher', 'DOI': '10.18637/jss.v045.i03'},
{'key': 'ref42', 'doi-asserted-by': 'publisher', 'DOI': '10.18637/jss.v033.i01'},
{'key': 'ref43', 'doi-asserted-by': 'publisher', 'DOI': '10.18637/jss.v042.i08'},
{ 'key': 'ref44',
'unstructured': 'ZhouTTongGLiFThomasLPSweight: An R Package for Propensity Score '
'Weighting Analysis20202021-08-27http://arxiv.org/abs/2010.08893'},
{ 'key': 'ref45',
'unstructured': 'pharmwas2021-11-29https://gitlab.com/lerner.ivan/pharmwas'},
{'key': 'ref46', 'doi-asserted-by': 'publisher', 'DOI': '10.1136/bmj.g445'},
{'key': 'ref47', 'doi-asserted-by': 'publisher', 'DOI': '10.1016/s0749-0704(18)30042-3'},
{'key': 'ref48', 'doi-asserted-by': 'publisher', 'DOI': '10.1016/j.eclinm.2021.100962'},
{'key': 'ref49', 'doi-asserted-by': 'publisher', 'DOI': '10.1101/2020.11.23.20237362'},
{'key': 'ref50', 'doi-asserted-by': 'publisher', 'DOI': '10.2196/22219'},
{'key': 'ref51', 'doi-asserted-by': 'publisher', 'DOI': '10.1016/j.jclinepi.2015.05.029'},
{'key': 'ref52', 'doi-asserted-by': 'publisher', 'DOI': '10.2196/20773'},
{ 'key': 'ref53',
'unstructured': 'LernerIpharmwasGitLab2021-11-29https://gitlab.com/lerner.ivan/pharmwas'}],
'container-title': 'JMIR Medical Informatics',
'original-title': [],
'language': 'en',
'deposited': { 'date-parts': [[2022, 4, 12]],
'date-time': '2022-04-12T18:21:52Z',
'timestamp': 1649787712000},
'score': 1,
'resource': {'primary': {'URL': 'https://medinform.jmir.org/2022/3/e35190'}},
'subtitle': [],
'short-title': [],
'issued': {'date-parts': [[2022, 3, 30]]},
'references-count': 53,
'journal-issue': {'issue': '3', 'published-online': {'date-parts': [[2022]]}},
'URL': 'http://dx.doi.org/10.2196/35190',
'relation': {},
'ISSN': ['2291-9694'],
'subject': ['Health Information Management', 'Health Informatics'],
'container-title-short': 'JMIR Med Inform',
'published': {'date-parts': [[2022, 3, 30]]}}
