{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,10,6]],"date-time":"2026-10-06T18:41:38Z","timestamp":1791312098132,"version":"4.3.1"},"reference-count":79,"publisher":"F1000 Research Ltd","license":[{"start":{"date-parts":[[2026,2,20]],"date-time":"2026-02-20T00:00:00Z","timestamp":1771545600000},"content-version":"tdm","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"funder":[{"name":"VERSATILE EMERGING DISEASES CONSORTIUM","award":["874735"],"award-info":[{"award-number":["874735"]}]}],"content-domain":{"domain":["wellcomeopenresearch.org"],"crossmark-restriction":false},"short-container-title":["Wellcome Open Res"],"abstract":"<ns5:p>\n                    Antimicrobial resistance in gram-negative pathogens is a growing global health threat, particularly in low- and middle-income countries (LMICs). However, many LMICs lack the laboratory capacity needed for routine surveillance. Affordable, scalable methods are therefore required to guide the prioritisation of surveillance efforts. Although socioeconomic, governance, environmental, and health-system covariates are known to correlate with clinical AMR prevalence, these contextual covariates have not been fully utilised to estimate resistance in settings with limited data. We quantified these associations using a stacked ensemble modelling framework to generate national AMR prevalence estimates for nine priority gram-negative pathogen\u2013antibiotic combinations from 2005 to 2021. National AMR data were obtained from the\n                    <ns5:italic>Vivli AMR Register<\/ns5:italic>\n                    and\n                    <ns5:italic>ResistanceMap<\/ns5:italic>\n                    and linked with 178 covariates describing socioeconomic conditions, governance indicators, environmental exposures, water and sanitation, and health-system performance. Child models included generalized additive models, gradient boosting machines, random forests, support vector machines, cubist models, and xgboost regressors, whose estimations were combined using spatiotemporal gaussian process regression as a meta-learner. Cross-validated performance was robust for most pathogen\u2013antibiotic combinations, with\n                    <ns5:italic>\n                      R\n                      <ns5:sup>2<\/ns5:sup>\n                    <\/ns5:italic>\n                    values ranging from\n                    <ns5:italic>0.65 to 0.93.<\/ns5:italic>\n                    Our modelled estimates enable the identification of priority countries for AMR surveillance by jointly considering estimated prevalence, temporal trajectories, and model performance across pathogens. We identified priority countries across multiple pathogen\u2013antibiotic combinations, with estimated prevalence concentrated in regions such as Central Asia, the Middle East and North Africa, Latin America, and sub-Saharan Africa. Many countries within these regions remain unenrolled in the World Health Organization\u2019s Global Antimicrobial Resistance and Use Surveillance System or contribute only sparse data, underscoring persistent gaps in surveillance where resistance is estimated to be high. By integrating AMR data with relevant contextual covariates, our framework fills temporal and geographical gaps. These estimates support evidence-based prioritisation of surveillance strengthening in resource-constrained settings.\n                  <\/ns5:p>","DOI":"10.12688\/wellcomeopenres.25696.1","type":"journal-article","created":{"date-parts":[[2026,2,20]],"date-time":"2026-02-20T08:39:08Z","timestamp":1771576748000},"page":"134","update-policy":"https:\/\/doi.org\/10.12688\/wellcomeopenres.crossmark-policy","source":"Crossref","is-referenced-by-count":1,"title":["Prioritization\u00a0of antimicrobial resistance surveillance of gram-negative pathogens\u00a0-\u00a0a stacked modelling\u00a0approach"],"prefix":"10.12688","volume":"11","author":[{"ORCID":"https:\/\/orcid.org\/0000-0003-1687-7336","authenticated-orcid":false,"given":"Sneha","family":"P Kotian","sequence":"first","affiliation":[{"name":"Department of Global Health, Amsterdam Institute for Global Health and Development \u2013 AIGHD Amsterdam North Holland 1105 BP Amsterdam UMC Locatie AMC"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Rik","family":"Oldenkamp","sequence":"additional","affiliation":[{"name":"Amsterdam Institute for Life & Environment (A-LIFE) Section: Environment & Health Amsterdam North Holland 1081HZ VU University Amsterdam Institute for Environmental Studies"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Constance","family":"Schultsz","sequence":"additional","affiliation":[{"name":"Department of Global Health and Department of Medical Microbiology, Amsterdam Institute for Global Health and Development \u2013 AIGHD Amsterdam North Holland 1105BP Amsterdam UMC Location AMC Department of Medical Microbiology"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"2560","published-online":{"date-parts":[[2026,2,20]]},"reference":[{"key":"ref1","doi-asserted-by":"crossref","first-page":"629-655","DOI":"10.1016\/S0140-6736(21)02724-0","article-title":"Global burden of bacterial antimicrobial resistance in 2019: a systematic analysis.","volume":"399","author":"C Murray","year":"2022","journal-title":"Lancet."},{"key":"ref2","doi-asserted-by":"publisher","first-page":"1429-1433","DOI":"10.1093\/cid\/ciad758","article-title":"The real crisis in antimicrobial resistance: failure to anticipate and respond.","volume":"78","author":"R Bonomo","year":"2024","journal-title":"Clin. Infect. Dis."},{"key":"ref3","doi-asserted-by":"publisher","first-page":"e1003787","DOI":"10.1371\/journal.pmed.1003787","article-title":"Gram-negative neonatal sepsis in low- and lower-middle-income countries and WHO empirical antibiotic recommendations: a systematic review and meta-analysis.","volume":"18","author":"S Wen","year":"2021","journal-title":"PLoS Med."},{"key":"ref4","doi-asserted-by":"publisher","first-page":"100940","DOI":"10.1016\/j.lanmic.2024.07.003","article-title":"WHO publishes updated list of bacterial priority pathogens.","volume":"5","author":"T Jesudason","year":"2024","journal-title":"Lancet Microbe."},{"key":"ref5","doi-asserted-by":"publisher","first-page":"2426-2438","DOI":"10.1016\/S0140-6736(24)00876-6","article-title":"The scope of the antimicrobial resistance challenge.","volume":"403","author":"I Okeke","year":"2024","journal-title":"Lancet."},{"key":"ref6","doi-asserted-by":"crossref","first-page":"1199-1226","DOI":"10.1016\/S0140-6736(24)01867-1","article-title":"Global burden of bacterial antimicrobial resistance 1990\u20132021: a systematic analysis with forecasts to 2050.","volume":"404","author":"M Naghavi","year":"2024","journal-title":"Lancet."},{"key":"ref7","doi-asserted-by":"publisher","DOI":"10.3390\/antibiotics10070834","article-title":"Determinants of antimicrobial resistance among European countries: beyond human and animal antimicrobial consumption.","volume":"10","author":"A Silva","year":"2021","journal-title":"Antibiotics (Basel)."},{"key":"ref8","doi-asserted-by":"publisher","DOI":"10.1073\/pnas.2013515118","article-title":"Filling the gaps in the global prevalence map of clinical antimicrobial resistance.","volume":"118","author":"R Oldenkamp","year":"2021","journal-title":"Proc. Natl. Acad. Sci. USA."},{"key":"ref9","doi-asserted-by":"publisher","first-page":"e893-e904","DOI":"10.1016\/S2542-5196(21)00280-1","article-title":"Global antibiotic consumption and usage in humans, 2000\u20132018: a spatial modelling study.","volume":"5","author":"A Browne","year":"2021","journal-title":"Lancet Planet Health."},{"key":"ref10","doi-asserted-by":"publisher","first-page":"e888-e899","DOI":"10.1016\/S2542-5196(23)00213-9","article-title":"Association of health, nutrition, and socioeconomic variables with global antimicrobial resistance: a modelling study.","volume":"7","author":"P Njage","year":"2023","journal-title":"Lancet Planet Health."},{"key":"ref11","doi-asserted-by":"crossref","first-page":"1808","DOI":"10.3390\/healthcare11121808","article-title":"Ensemble learning for disease prediction: a review.","volume":"11","author":"P Mahajan","year":"2023","journal-title":"Healthcare."},{"key":"ref12","doi-asserted-by":"publisher","first-page":"1323-1335","DOI":"10.1016\/S0140-6736(16)32381-9","article-title":"Future life expectancy in 35 industrialised countries: projections with a Bayesian model ensemble.","volume":"389","author":"V Kontis","year":"2017","journal-title":"Lancet."},{"key":"ref13","doi-asserted-by":"publisher","first-page":"265","DOI":"10.1186\/s12879-024-09138-x","article-title":"Predicting infectious diarrhea incidence using a stacking-based ensemble model.","volume":"24","author":"P Wang","year":"2024","journal-title":"BMC Infect. Dis."},{"key":"ref14","doi-asserted-by":"crossref","first-page":"e201-e216","DOI":"10.1016\/S2214-109X(23)00539-9","article-title":"Burden of bacterial antimicrobial resistance in the WHO African region in 2019.","volume":"12","author":"B Sartorius","year":"2024","journal-title":"Lancet Glob. Health."},{"key":"ref15","article-title":"AMR Register: antimicrobial resistance surveillance platform."},{"key":"ref16","article-title":"ResistanceMap: antibiotic resistance data.","year":"2023"},{"key":"ref17","article-title":"Antimicrobial Testing Leadership and Surveillance (ATLAS)."},{"key":"ref18","article-title":"World Development Indicators.","year":"2023"},{"key":"ref19","article-title":"Global Health Observatory: antibacterial consumption.","year":"2024"},{"key":"ref20","doi-asserted-by":"publisher","first-page":"269","DOI":"10.1186\/s12874-024-02392-2","article-title":"A MissForest-based imputation approach with recursive feature elimination in medical applications.","volume":"24","author":"Y Hu","year":"2024","journal-title":"BMC Med. Res. Methodol."},{"key":"ref21","doi-asserted-by":"publisher","first-page":"112-118","DOI":"10.1093\/bioinformatics\/btr597","article-title":"MissForest\u2014non-parametric missing value imputation for mixed-type data.","volume":"28","author":"D Stekhoven","year":"2012","journal-title":"Bioinformatics."},{"key":"ref22","doi-asserted-by":"publisher","first-page":"363-377","DOI":"10.1002\/sam.11348","article-title":"Random forest missing data algorithms.","volume":"10","author":"F Tang","year":"2017","journal-title":"Stat. Anal. Data Min."},{"key":"ref23","doi-asserted-by":"publisher","first-page":"122","DOI":"10.1186\/s13756-022-01160-5","article-title":"Utilising cumulative antibiogram data to enhance antibiotic stewardship capacity in Ghana.","volume":"11","author":"M Dakorah","year":"2022","journal-title":"Antimicrob. Resist. Infect. Control."},{"key":"ref24","article-title":"mgcv: Mixed GAM computation vehicle with automatic smoothness estimation.","author":"S Wood","journal-title":"R package."},{"key":"ref25","doi-asserted-by":"publisher","first-page":"2727-2739","DOI":"10.1007\/s10489-020-01997-6","article-title":"Gradient boosting machine approach for modelling COVID-19 transmission.","volume":"51","author":"L Shrivastav","year":"2021","journal-title":"Appl. Intell."},{"key":"ref27","doi-asserted-by":"crossref","first-page":"16","DOI":"10.1186\/1472-6947-10-16","article-title":"Application of support vector machine modelling for disease prediction.","volume":"10","author":"W Yu","year":"2010","journal-title":"BMC Med. Inform. Decis. Mak."},{"key":"ref28","first-page":"2129-2154","article-title":"Human resources for health and universal health coverage.","volume":"399","author":"A Haakenstad","year":"2022","journal-title":"Lancet."},{"key":"ref29","doi-asserted-by":"crossref","first-page":"3-36","DOI":"10.1111\/j.1467-9868.2010.00749.x","article-title":"Fast stable REML and marginal likelihood estimation of semiparametric GLMs.","volume":"73","author":"S Wood","year":"2011","journal-title":"J. R. Stat. Soc. Ser. B."},{"key":"ref30","article-title":"gbm: Generalized boosted regression models.","author":"G Ridgeway","journal-title":"R package."},{"key":"ref31","doi-asserted-by":"crossref","first-page":"5-32","DOI":"10.1023\/A:1010933404324","article-title":"Random forests.","volume":"45","author":"L Breiman","year":"2001","journal-title":"Mach. Learn."},{"key":"ref32","article-title":"Cubist: Rule- and instance-based regression modelling.","author":"M Kuhn","journal-title":"R package."},{"key":"ref34","doi-asserted-by":"publisher","DOI":"10.18637\/jss.v028.i05","article-title":"Building predictive models in R using the caret package.","volume":"28","author":"M Kuhn","year":"2008","journal-title":"J. Stat. Softw."},{"key":"ref35","doi-asserted-by":"publisher","DOI":"10.18637\/jss.v063.i19","article-title":"Bayesian spatial modelling with R-INLA.","volume":"63","author":"F Lindgren","year":"2015","journal-title":"J. Stat. Softw."},{"key":"ref36","first-page":"e955-e970","article-title":"AMR burden in the WHO Eastern Mediterranean Region.","volume":"10","author":"T Mestrovic","year":"2025","journal-title":"Lancet Public Health."},{"key":"ref37","first-page":"e897-e913","article-title":"AMR burden in the WHO European region in 2019.","volume":"7","author":"T Mestrovic","year":"2022","journal-title":"Lancet Public Health."},{"key":"ref38","first-page":"100561","article-title":"Burden of antimicrobial resistance in the Americas in 2019.","volume":"25","year":"2023","journal-title":"Lancet Reg Health Am."},{"key":"ref39","first-page":"e406-e418","article-title":"Subnational prevalence of antimicrobial-resistant Salmonella Typhi and Paratyphi A.","volume":"12","author":"A Browne","year":"2024","journal-title":"Lancet Glob. Health."},{"key":"ref40","doi-asserted-by":"crossref","first-page":"382","DOI":"10.3390\/antibiotics10040382","article-title":"Antibiotic resistance in Acinetobacter baumannii in Kazakhstan.","volume":"10","author":"A Lavrinenko","year":"2021","journal-title":"Antibiotics."},{"key":"ref41","first-page":"875","article-title":"baumannii in Bulgarian ICUs.","volume":"11","author":"T Strateva","year":"2023","journal-title":"Microorganisms."},{"key":"ref42","doi-asserted-by":"crossref","first-page":"421","DOI":"10.3390\/microorganisms13020421","article-title":"Carbapenem-resistant Gram-negative bacteria in Moldova.","volume":"13","author":"V Nirca","year":"2025","journal-title":"Microorganisms."},{"key":"ref43","first-page":"1200","article-title":"AMR trends of A. baumannii in Slovakia.","volume":"12","author":"Y Jalali","year":"2023","journal-title":"Antibiotics."},{"key":"ref44","doi-asserted-by":"publisher","first-page":"e0251508","DOI":"10.1371\/journal.pone.0251508","article-title":"Molecular characterisation of carbapenem-resistant A. baumannii in Egypt.","volume":"16","author":"R Hassan","year":"2021","journal-title":"PLoS One."},{"key":"ref45","doi-asserted-by":"publisher","first-page":"612-621","DOI":"10.1016\/j.ijid.2022.07.018","article-title":"AMR patterns in febrile illness pathogens.","volume":"122","author":"T Roberts","year":"2022","journal-title":"Int. J. Infect. Dis."},{"key":"ref46","doi-asserted-by":"publisher","first-page":"4177-4191","DOI":"10.2147\/IDR.S259562","article-title":"Carbapenemase-producing Gram-negative bacteria in Tunisia.","volume":"13","author":"O Dziri","year":"2020","journal-title":"Infect. Drug Resist."},{"key":"ref47","doi-asserted-by":"publisher","first-page":"254-260","DOI":"10.4103\/sjmms.sjmms_855_20","article-title":"Antibiotic resistance in A. baumannii.","volume":"9","author":"M Sannathimmappa","year":"2021","journal-title":"Saudi J. Med. Med. Sci."},{"key":"ref48","doi-asserted-by":"publisher","first-page":"776-777","DOI":"10.1016\/j.ajic.2011.09.019","article-title":"Antibiotic resistance in A. baumannii from Syria.","volume":"40","author":"A Hamzeh","year":"2012","journal-title":"Am. J. Infect. Control."},{"key":"ref49","doi-asserted-by":"publisher","first-page":"746","DOI":"10.1186\/s12879-019-4295-6","article-title":"AMR trends in Harare, Zimbabwe.","volume":"19","author":"M Mhondoro","year":"2019","journal-title":"BMC Infect. Dis."},{"key":"ref50","doi-asserted-by":"publisher","first-page":"1308","DOI":"10.1186\/s40064-016-2986-7","article-title":"aeruginosa and A. baumannii in Uganda.","volume":"5","author":"D Kateete","year":"2016","journal-title":"Springerplus."},{"key":"ref51","doi-asserted-by":"publisher","first-page":"e0287762","DOI":"10.1371\/journal.pone.0287762","article-title":"Carbapenem resistance in sub-Saharan Africa.","volume":"18","author":"M Arowolo","year":"2023","journal-title":"PLoS One."},{"key":"ref52","doi-asserted-by":"publisher","first-page":"492-507","DOI":"10.1093\/jac\/dkz464","article-title":"Bloodstream infections due to 3GCR Enterobacteriaceae in sub-Saharan Africa.","volume":"75","author":"R Lester","year":"2020","journal-title":"J. Antimicrob. Chemother."},{"key":"ref53","article-title":"Ceftriaxone resistance trends in Sudan.","author":"Y Hamadalneel","year":"2024","journal-title":"Cureus."},{"key":"ref54","first-page":"77","article-title":"Antibiotic resistance in Ugandan tertiary hospitals.","volume":"9","author":"J Mayito","year":"2024","journal-title":"Trop. Med. Infect. Dis."},{"key":"ref55","doi-asserted-by":"publisher","first-page":"1381-1390","DOI":"10.2147\/IDR.S403360","article-title":"Carbapenem-resistant Pseudomonas and Acinetobacter in Ethiopia.","volume":"16","author":"A Abdeta","year":"2023","journal-title":"Infect. Drug Resist."},{"key":"ref56","first-page":"1-7","article-title":"Resistance to third-generation cephalosporins.","volume":"2018","author":"F Gashe","year":"2018","journal-title":"J. Pharm (Cairo)."},{"key":"ref57","first-page":"1-9","article-title":"Carbapenem-resistant A. baumannii in Peru. Emerg.","volume":"7","author":"S Levy-Blitchtein","year":"2018","journal-title":"Microbes Infect."},{"key":"ref58","doi-asserted-by":"publisher","first-page":"378","DOI":"10.1186\/s12879-024-09248-6","article-title":"Molecular epidemiology of carbapenem-resistant Gram-negative bacilli in Ecuador.","volume":"24","author":"C Soria-Segarra","year":"2024","journal-title":"BMC Infect. Dis."},{"key":"ref59","doi-asserted-by":"publisher","first-page":"e0274423","DOI":"10.1371\/journal.pone.0274423","article-title":"Community-acquired UTIs in Bangladesh.","volume":"17","author":"M Islam","year":"2022","journal-title":"PLoS One."},{"key":"ref60","doi-asserted-by":"publisher","first-page":"26-32","DOI":"10.3390\/idr14010004","article-title":"AMR in paediatric patients in Myanmar.","volume":"14","author":"T San","year":"2022","journal-title":"Infect. Dis. Rep."},{"key":"ref61","doi-asserted-by":"publisher","first-page":"e0311124","DOI":"10.1371\/journal.pone.0311124","article-title":"Global mapping of A. baumannii carbapenem resistance.","volume":"19","author":"M Beig","year":"2024","journal-title":"PLoS One."},{"key":"ref62","doi-asserted-by":"publisher","first-page":"e002418","DOI":"10.1136\/bmjgh-2020-002418","article-title":"AMR in Pacific Island countries.","volume":"5","author":"M Loftus","year":"2020","journal-title":"BMJ Glob. Health."},{"key":"ref63","article-title":"Socioeconomic status and risk of priority pathogens: a global evidence map.","author":"S Blackmon","year":"2024"},{"key":"ref64","doi-asserted-by":"publisher","first-page":"385-390","DOI":"10.1093\/jac\/dkg072","article-title":"Influence of population density on antibiotic resistance.","volume":"51","author":"N Bruinsma","year":"2003","journal-title":"J. Antimicrob. Chemother."},{"key":"ref65","doi-asserted-by":"publisher","first-page":"2439-2454","DOI":"10.1016\/S0140-6736(24)00862-6","article-title":"AMR avertible by existing interventions.","volume":"403","author":"J Lewnard","year":"2024","journal-title":"Lancet."},{"key":"ref66","doi-asserted-by":"crossref","first-page":"829","DOI":"10.3390\/microorganisms13040829","article-title":"Healthcare-associated infections due to carbapenem-resistant A. baumannii.","volume":"13","author":"A Russo","year":"2025","journal-title":"Microorganisms."},{"key":"ref67","doi-asserted-by":"publisher","first-page":"706-718","DOI":"10.1016\/S1473-3099(22)00796-4","article-title":"Global response to antimicrobial resistance, 2020\u201321.","volume":"23","author":"J Patel","year":"2023","journal-title":"Lancet Infect. Dis."},{"key":"ref68","doi-asserted-by":"publisher","first-page":"e0116746","DOI":"10.1371\/journal.pone.0116746","article-title":"Governance and corruption as drivers of antimicrobial resistance.","volume":"10","author":"P Collignon","year":"2015","journal-title":"PLoS One."},{"key":"ref69","doi-asserted-by":"publisher","first-page":"12","DOI":"10.1186\/s12992-023-00913-0","article-title":"Socioeconomic and governance indicators shaping AMR in Europe.","volume":"19","author":"A Maugeri","year":"2023","journal-title":"Glob. Health."},{"key":"ref70","doi-asserted-by":"publisher","first-page":"15","DOI":"10.1186\/s13756-021-01040-4","article-title":"Transparency and accountability of AMR national action plans in Africa.","volume":"11","author":"A Harant","year":"2022","journal-title":"Antimicrob. Resist. Infect. Control."},{"key":"ref71","doi-asserted-by":"publisher","first-page":"e398-e405","DOI":"10.1016\/S2542-5196(18)30186-4","article-title":"Anthropological and socioeconomic drivers of AMR.","volume":"2","author":"P Collignon","year":"2018","journal-title":"Lancet Planet Health."},{"key":"ref72","doi-asserted-by":"publisher","first-page":"e291-e303","DOI":"10.1016\/S2542-5196(23)00026-8","article-title":"Global AMR drivers at the human\u2013animal interface.","volume":"7","author":"K Allel","year":"2023","journal-title":"Lancet Planet Health."},{"key":"ref73","doi-asserted-by":"crossref","first-page":"48","DOI":"10.12688\/wellcomeopenres.18829.1","article-title":"Anaemia and susceptibility to invasive bacterial infections.","volume":"8","author":"K Abuga","year":"2023","journal-title":"Wellcome Open Res."},{"key":"ref74","doi-asserted-by":"publisher","first-page":"75-84","DOI":"10.1136\/jech-2020-216029","article-title":"Type 2 diabetes and antibiotic-resistant infections.","volume":"76","author":"R Carrillo-Larco","year":"2022","journal-title":"J. Epidemiol. Community Health."},{"key":"ref76","doi-asserted-by":"crossref","first-page":"16","DOI":"10.1038\/s41545-025-00447-5","article-title":"AMR enteric bacteria in water sources in Bangladesh.","volume":"8","author":"N Sharif","year":"2025","journal-title":"NPJ Clean Water."},{"key":"ref77","first-page":"12","article-title":"AMR surveillance systems in East Africa.","volume":"47","author":"M Matee","year":"2023","journal-title":"Bull. Natl. Res. Cent."},{"key":"ref78","doi-asserted-by":"publisher","first-page":"8","DOI":"10.3389\/fpubh.2020.00454","article-title":"AMR surveillance and stewardship in Tanzania.","volume":"8","author":"R Sangeda","year":"2020","journal-title":"Front Public Health."},{"key":"ref79","first-page":"80","article-title":"Wounds and AMR surveillance in Myanmar.","volume":"6","author":"W Sandar","year":"2021","journal-title":"Trop. Med. Infect. Dis."},{"key":"ref80","doi-asserted-by":"publisher","first-page":"429-433","DOI":"10.1016\/j.cmi.2022.05.006","article-title":"Unbiased AMR prevalence estimates through population-based surveillance.","volume":"29","author":"F Leth","year":"2023","journal-title":"Clin. Microbiol. Infect."},{"key":"ref81","doi-asserted-by":"publisher","DOI":"10.6084\/m9.figshare.31077844","article-title":"Supplementary appendix for prioritisation of antimicrobial resistance surveillance of gram-negative pathogens using a stacked modelling approach.","author":"S Kotian","year":"2025","journal-title":"Figshare."},{"key":"ref82","doi-asserted-by":"publisher","DOI":"10.5281\/zenodo.18269992","article-title":"Extended data for prioritisation of antimicrobial resistance surveillance of gram-negative pathogens using a stacked modelling approach [code].","author":"S Kotian","year":"2025","journal-title":"Zenodo."}],"updated-by":[{"DOI":"10.12688\/wellcomeopenres.25696.2","type":"new_version","label":"New version","source":"publisher","updated":{"date-parts":[[2026,6,25]],"date-time":"2026-06-25T00:00:00Z","timestamp":1782345600000}}],"container-title":["Wellcome Open Research"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/wellcomeopenresearch.org\/articles\/11-134\/v1\/xml","content-type":"application\/xml","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/wellcomeopenresearch.org\/articles\/11-134\/v1\/pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/wellcomeopenresearch.org\/articles\/11-134\/v1\/iparadigms","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,6,25]],"date-time":"2026-06-25T07:25:24Z","timestamp":1782372324000},"score":1,"resource":{"primary":{"URL":"https:\/\/wellcomeopenresearch.org\/articles\/11-134\/v1"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,2,20]]},"references-count":79,"URL":"https:\/\/doi.org\/10.12688\/wellcomeopenres.25696.1","relation":{"has-review":[{"id-type":"doi","id":"10.21956\/wellcomeopenres.28294.r149236","asserted-by":"subject"},{"id-type":"doi","id":"10.21956\/wellcomeopenres.28294.r152442","asserted-by":"subject"},{"id-type":"doi","id":"10.21956\/wellcomeopenres.28294.r149236","asserted-by":"object"},{"id-type":"doi","id":"10.21956\/wellcomeopenres.28294.r152442","asserted-by":"object"}],"is-version-of":[{"id-type":"doi","id":"10.12688\/wellcomeopenres.25696.2","asserted-by":"object"}]},"ISSN":["2398-502X"],"issn-type":[{"value":"2398-502X","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,2,20]]},"assertion":[{"value":"Approved, Approved with reservations","URL":"https:\/\/wellcomeopenresearch.org\/articles\/11-134\/v1#article-reports","order":0,"name":"referee-status","label":"Referee status","group":{"name":"current-referee-status","label":"Current Referee Status"}},{"value":"10.21956\/wellcomeopenres.28294.r149236, Chlo\u00e9 Smit, University of Technology Sydney, Broadway, New South Wales, 26 Mar 2026, version 1, 1 approved, 1 approved with reservations","URL":"https:\/\/wellcomeopenresearch.org\/articles\/11-134\/v1#referee-response-149236","order":0,"name":"referee-response-149236","label":"Referee Report","group":{"name":"article-reports","label":"Article Reports"}},{"value":"10.21956\/wellcomeopenres.28294.r152442, Gang Wang, Universit\u00e9 Catholique de Louvain, Brussels, 06 Apr 2026, version 1, 1 approved, 1 approved with reservations","URL":"https:\/\/wellcomeopenresearch.org\/articles\/11-134\/v1#referee-response-152442","order":1,"name":"referee-response-152442","label":"Referee Report","group":{"name":"article-reports","label":"Article Reports"}},{"value":"<b>SNEHA KOTIAN<\/b>; \n<i>Posted: 12 Jun 2026<\/i>; We thank Reviewer 2 for their careful and constructive assessment of our manuscript.&nbsp; We have revised the manuscript and added&nbsp;Supplementary Appendix 2&nbsp;to address these points, and we provide a point-by-point response below.&nbsp;  \n<b>Comment no 1:<\/b>&nbsp; A major concern is that&nbsp;the&nbsp;validation strategy does not fully match the intended application.&nbsp;The model is used to generate&nbsp;estimates&nbsp;for countries with no surveillance data, yet performance is assessed using stratified train-test splitting within the observed dataset, plus internal cross-validation and bootstrap resampling. This&nbsp;demonstrates&nbsp;internal predictive skill, but it is not equivalent to testing how well the framework&nbsp;extrapolates to&nbsp;completely unobserved countries or regions. For a surveillance-prioritization tool, this distinction matters&nbsp;greatly&nbsp;in&nbsp;practice.&nbsp;I therefore think the study design&nbsp;is only&nbsp;partly&nbsp;sufficient as presented. The authors should either perform a more stringent external-style validation, such as leave-one-country-out, leave-one-region-out, or temporally blocked validation, or explicitly temper their claims about model performance in data-absent settings.&nbsp; The authors should clarify that&nbsp;internal cross-validation performance does not fully&nbsp;establish&nbsp;transportability to countries with no surveillance&nbsp;data and&nbsp;ideally provide a more demanding validation scheme closer to the intended use case.&nbsp; \n<b>Answer no 1:&nbsp;<\/b>&nbsp; We agree that&nbsp;validation should be aligned as closely as possible with the intended application of the model.&nbsp;We also&nbsp;recognise&nbsp;that&nbsp;our original wording did not make the validation design sufficiently clear.&nbsp;As described in the manuscript, we&nbsp;used&nbsp;a stringent validation approach consistent with the reviewer\u2019s recommendation,&nbsp;using&nbsp;stratified&nbsp;country level&nbsp;holdout&nbsp;validation.&nbsp;&nbsp;Thus, all observations from a given country across all available years were assigned exclusively to either the training or the testing set, with stratification by WHO region and World Bank income group to preserve geographic and socioeconomic representation in both sets.&nbsp;The reported out-of-sample performance therefore reflects estimation for entirely unseen countries within the testing dataset.&nbsp; To further&nbsp;contextualise&nbsp;estimates&nbsp;in data-absent settings, we also calculated&nbsp;Mahalanobis&nbsp;distance as a supplementary measure of how far countries without surveillance data lay from the covariate space represented in the training data.&nbsp;(Supplementary Appendix&nbsp;2)&nbsp;&nbsp; We have revised the manuscript to make the country-level holdout explicit, and added the following text to the&nbsp;methods&nbsp;and discussion&nbsp;section&nbsp;(Lines: 114-120; 158-164;&nbsp;588-602)&nbsp; &nbsp; &nbsp;  \n<b>Comment no 2:&nbsp;<\/b>&nbsp; A second major issue is the harmonization of different AMR data sources.&nbsp;ResistanceMap&nbsp;provides aggregated country-level resistance estimates from invasive isolates obtained from blood and cerebrospinal fluid,&nbsp;whereas&nbsp;ATLAS&nbsp;contains&nbsp;isolate-level susceptibility data from multiple specimen types including blood, sputum, urine, abscess material, and wound swabs. The manuscript also reclassifies intermediate susceptibility as resistant to&nbsp;align&nbsp;the datasets. While this harmonization is understandable, it introduces non-trivial comparability concerns,&nbsp;especially&nbsp;pathogen\u2013drug combinations whose resistance prevalence may vary&nbsp;substantially by&nbsp;specimen type, clinical syndrome, and care setting. In my view, this issue is central rather than peripheral because the study\u2019s outputs are intended to inform surveillance investment decisions. The authors should provide sensitivity analyses using more comparable subsets where&nbsp;feasible&nbsp;or&nbsp;at&nbsp;least expand the discussion of the&nbsp;likely direction&nbsp;and&nbsp;magnitude&nbsp;of bias introduced by combining these data streams.&nbsp; The authors should justify more fully the&nbsp;harmonisation&nbsp;of&nbsp;ATLAS&nbsp;and&nbsp;ResistanceMap,&nbsp;including the handling of different specimen types and the reclassification of intermediate isolates as resistant, and add sensitivity analyses if&nbsp;feasible.&nbsp;\n<b> Answer&nbsp;no&nbsp;2:<\/b>&nbsp; We agree with the reviewer that&nbsp;harmonising&nbsp;ResistanceMap&nbsp;and&nbsp;ATLAS&nbsp;introduces potential comparability concerns.&nbsp;ResistanceMap&nbsp;provides aggregated country-level resistance estimates from invasive isolates obtained from blood and cerebrospinal fluid,&nbsp;whereas&nbsp;ATLAS&nbsp;includes isolate-level data from a broader range of specimen types, including blood, sputum, urine, abscess material, and wound swabs. We also agree that&nbsp;resistance prevalence may vary by specimen type, clinical syndrome, and care setting, and that&nbsp;this issue is relevant for interpreting outputs intended to guide surveillance&nbsp;prioritisation.&nbsp; To address this concern, we performed an&nbsp;additional&nbsp;sensitivity&nbsp;analysis&nbsp;(Supplementary Appendix 2)&nbsp;in which&nbsp;ATLAS&nbsp;was restricted to blood and cerebrospinal fluid specimens only, while&nbsp;ResistanceMap&nbsp;was kept unchanged. Intermediate susceptibility results were classified as resistant, consistent with the main analysis. The blood\/CSF-restricted&nbsp;ATLAS&nbsp;sensitivity estimates were consistent with the observed resistance estimates used in the main analysis, which combine&nbsp;ResistanceMap&nbsp;and all eligible&nbsp;ATLAS&nbsp;specimens. For&nbsp;carbapenem-resistant&nbsp;A. baumannii&nbsp;and&nbsp;third-generation&nbsp;cephalosporin-resistant&nbsp;E. coli,&nbsp;the Pearson correlation coefficients between the two sets of estimates were 0.992 and 0.968, respectively. These findings suggest that, for the evaluated pathogen\u2013antibiotic&nbsp;combinations and overlapping country-years, restricting&nbsp;ATLAS&nbsp;to invasive specimens did not change the observed resistance estimates&nbsp;significantly.&nbsp; However, a full blood\/CSF-restricted analysis was only partly&nbsp;feasible&nbsp;and would have&nbsp;substantially reduced&nbsp;the usable&nbsp;ATLAS&nbsp;data.(Supplementary Appendix 2)&nbsp;Across&nbsp;the pathogen\u2013antibiotic&nbsp;combinations included in our analysis, only 12.1% to 25.2% of&nbsp;ATLAS&nbsp;isolates were from blood or cerebrospinal fluid; the remaining isolates were from other or unknown specimen types. Restricting&nbsp;ATLAS&nbsp;to blood\/CSF would therefore have reduced the number of eligible country-year observations, particularly in already data-sparse settings. In addition,&nbsp;ResistanceMap&nbsp;is available only as an aggregated resistance measure and cannot be further stratified by specimen type, clinical syndrome, or care setting.&nbsp; These results highlight a trade-off between comparability and coverage. Restricting&nbsp;ATLAS&nbsp;to invasive specimens improves comparability with&nbsp;ResistanceMap&nbsp;but&nbsp;substantially reduces&nbsp;surveillance coverage. Conversely,&nbsp;retaining&nbsp;the broader&nbsp;ATLAS&nbsp;dataset preserves country-year coverage, which is important for a study focused on surveillance gaps and&nbsp;prioritisation. Because&nbsp;ATLAS&nbsp;contributed fewer tested isolates than&nbsp;ResistanceMap,&nbsp;changes to the&nbsp;ATLAS&nbsp;specimen definition had limited influence on the combined observed resistance estimates in the sensitivity analysis&nbsp;&nbsp; We have added the&nbsp;detailed&nbsp;sensitivity analyses&nbsp;in&nbsp;(Supplementary Appendix 2)&nbsp;for&nbsp;your&nbsp;reference.&nbsp;&nbsp;  \n<b>Comment no 3:<\/b>&nbsp; Third, although the statistical framework is sophisticated&nbsp;and in many respects&nbsp;appropriate, the&nbsp;interpretation of the model outputs and covariate importance is&nbsp;somewhat&nbsp;stronger&nbsp;than the design strictly supports. The manuscript appropriately&nbsp;states&nbsp;that&nbsp;covariate importance scores do not imply causality, and the results section notes that&nbsp;governance, health-system financing, socioeconomic variables, WASH, agricultural, nutritional, and environmental factors rank highly. However, parts of the Discussion read as though these factors are being interpreted as mechanistic drivers rather than predictive correlates within a modelling framework. This is especially important because policy readers may overread such results as evidence for causal leverage points. I therefore rate the statistical analysis and its interpretation as&nbsp;partly&nbsp;appropriate: the modelling itself is broadly reasonable, but the interpretation should be more restrained and more clearly separated into predictive versus causal language.&nbsp; The authors should distinguish more clearly between predictive covariate importance and causal&nbsp;inference, and&nbsp;avoid implying&nbsp;mechanism&nbsp;where the modelling design only supports association.&nbsp; \n<b>Answer no 3:<\/b>&nbsp; We&nbsp;agree with the reviewer that&nbsp;correlation does not imply&nbsp;causality and&nbsp;realise&nbsp;we might not have made this distinction clear enough throughout the manuscript.&nbsp;We&nbsp;have revised the manuscript&nbsp;in response to this&nbsp;suggestion.&nbsp;(Lines: 432-435)&nbsp;We&nbsp;removed text that&nbsp;could have been interpreted as implying causal mechanisms.&nbsp;(Lines: 482-483)&nbsp;  \n<b>Comment no 4:<\/b>&nbsp; Fourth, the manuscript would&nbsp;benefit&nbsp;from a stronger&nbsp;uncertainty-aware&nbsp;prioritisation&nbsp;framework. The authors derive three&nbsp;prioritisation&nbsp;indicators:&nbsp;absolute&nbsp;estimated prevalence in 2021, short-term change from 2016 to 2021, and long-term change from 2005 to 2021. These are sensible indicators. However, if the real goal is to decide where surveillance should be expanded first, then uncertainty should play a more direct role in the ranking or&nbsp;categorisation&nbsp;of countries. A country with a high predicted&nbsp;prevalence&nbsp;but very wide uncertainty may&nbsp;warrant&nbsp;a different operational response from a country with&nbsp;similarly high predicted prevalence and tighter uncertainty.&nbsp;Since the meta-model is probabilistic and the manuscript notes that&nbsp;95% uncertainty intervals were generated, the paper would be&nbsp;considerably stronger&nbsp;if uncertainty were incorporated explicitly into the&nbsp;prioritisation&nbsp;logic rather than being treated&nbsp;mainly as&nbsp;a technical output.&nbsp; The authors should integrate predictive uncertainty more explicitly into the surveillance-prioritisation&nbsp;framework.&nbsp; \n<b>Answer no 4:<\/b>&nbsp; We agree that&nbsp;uncertainty should be incorporated more directly into the interpretation of surveillance priorities, rather than presented only as a technical&nbsp;model&nbsp;output. In response, we added a rank-stability analysis to assess how robust country&nbsp;prioritisation&nbsp;rankings were to predictive uncertainty.&nbsp;This analysis propagated uncertainty by repeatedly sampling plausible resistance values from the estimated uncertainty intervals and re-ranking countries across 1,000 simulations. Countries were then classified as having high, moderate, or low rank stability based on the probability of&nbsp;remaining&nbsp;in the same rank quintile as the point estimate.&nbsp;(Supplementary Appendix 2)&nbsp; The results showed that&nbsp;prioritisation&nbsp;rankings were&nbsp;generally robust&nbsp;to predictive uncertainty. For&nbsp;carbapenem-resistant&nbsp;A. baumannii, 55.2% of country-year estimates showed high rank stability, 41.3% moderate stability, and 3.4% low stability. For&nbsp;third-generation cephalosporin-resistant&nbsp;E. coli, 83.0% of country-year estimates were classified as high stability, 15.5% as moderate stability, and 1.5% as low stability. In 2021, the highest-ranked countries for both pathogen\u2013antibiotic combinations were&nbsp;generally classified&nbsp;as high stability.&nbsp;&nbsp;  \n<b>Comment no 5:<\/b>&nbsp; Fifth, the paper acknowledges important data limitations that, in my view, should have even greater weight in&nbsp;the conclusions. The authors note that&nbsp;passive laboratory surveillance may over-represent&nbsp;resistant isolates, that&nbsp;the surveillance data used to train the models&nbsp;predominantly originate&nbsp;from high-income countries, and that&nbsp;lower-middle-income countries&nbsp;remain&nbsp;underrepresented. These are highly relevant limitations for a manuscript aiming to guide surveillance priorities in LMICs.&nbsp;The model may still be useful as a screening or prioritisation aid, but the conclusions should&nbsp;emphasise&nbsp;more clearly that&nbsp;the estimates are not substitutes for representative population-based surveillance and should not be interpreted as precise prevalence measurements in countries without routine data.&nbsp;On this point, the conclusions are only&nbsp;partly&nbsp;supported by the results in their current wording.&nbsp; The authors should make the practical interpretation of the outputs more cautious,&nbsp;emphasising&nbsp;that&nbsp;these estimates are best used to guide surveillance expansion rather than replace representative surveillance.&nbsp; \n<b>Answer no 5:<\/b>&nbsp; We agree with the reviewer that&nbsp;these limitations should be given&nbsp;appropriate weight&nbsp;in the interpretation of the study findings.&nbsp;We have revised the conclusion. (Lines:&nbsp;613-622)&nbsp; At&nbsp;the same time, we would like to clarify that&nbsp;passive laboratory-based surveillance may over-represent&nbsp;resistant isolates, but this does not&nbsp;preclude&nbsp;its use for&nbsp;prioritising&nbsp;surveillance strengthening. Rather, it highlights the need to interpret the outputs in relation to the surveillance system they are intended to support.&nbsp;Our estimates are most appropriately viewed as tools to&nbsp;identify&nbsp;where laboratory surveillance capacity, coverage, and reporting could be expanded or strengthened, while&nbsp;recognising&nbsp;that&nbsp;population-based surveillance systems would provide more precise estimates of resistance prevalence.&nbsp;To further support interpretation in countries without&nbsp;observed&nbsp;AMR data, we now&nbsp;also&nbsp;refer to the&nbsp;Mahalanobis&nbsp;distance analysis. This analysis assesses how&nbsp;similar countries without surveillance data are to countries represented in the training dataset. Countries with larger&nbsp;Mahalanobis&nbsp;distances are less well represented within the observed covariate space,&nbsp;indicating&nbsp;that&nbsp;predictions for these settings involve greater extrapolation from the training data. These countries may therefore be&nbsp;important targets for future surveillance expansion, both to strengthen local evidence generation and to improve the empirical basis for future modelled estimates.&nbsp;&nbsp;  \n<b>Comment no 6:<\/b>&nbsp; On reproducibility, I also think&nbsp;\u201cPartly\u201d&nbsp;is the most&nbsp;accurate&nbsp;answer. The manuscript provides substantial methodological detail,&nbsp;identifies&nbsp;the software environment and packages, notes compliance with GATHER, and makes the analysis code available on&nbsp;Zenodo. Those are clear strengths. However, the core&nbsp;ATLAS&nbsp;dataset is not openly available because it is under third-party license, and&nbsp;ResistanceMap&nbsp;is available only upon request. This means the study is not fully reproducible from openly accessible source data, even though code and supplementary materials are provided. I therefore would not mark \u201cYes\u201d for full&nbsp;source-data&nbsp;availability or full reproducibility.&nbsp; The authors should clarify the degree of reproducibility that&nbsp;is realistically achievable given the licensed and request-only source datasets&nbsp; \n<b>Answer no 6:<\/b>&nbsp; We agree that&nbsp;the study cannot be reproduced from openly downloadable AMR source data alone&nbsp;as&nbsp;the underlying AMR datasets are governed by third-party access conditions. The&nbsp;ATLAS&nbsp;dataset is subject to&nbsp;licence&nbsp;restrictions and cannot be redistributed by the authors, and&nbsp;ResistanceMap&nbsp;data are available upon request from One Health Trust.&nbsp;However, we would respectfully clarify that&nbsp;this reflects the governance conditions of the source datasets rather than a limitation in the transparency of the analytical workflow. This approach is common for studies using licensed or request-access health surveillance datasets, where authors are not&nbsp;permitted&nbsp;to re-host the original data but can provide sufficient information for other researchers to request access through the&nbsp;appropriate mechanisms.&nbsp;To support reproducibility as fully as possible within these constraints, we&nbsp;have&nbsp;provided the&nbsp;relevant&nbsp;details&nbsp;and access routes for both AMR data sources. Researchers who obtain access to the&nbsp;ATLAS&nbsp;and&nbsp;ResistanceMap&nbsp;datasets through these mechanisms should therefore be able to reproduce the analytical workflow and model outputs. We have revised the&nbsp;data availability statement to make this distinction&nbsp;clear.&nbsp;(Lines&nbsp;639-640)&nbsp; We thank the reviewer again for these helpful comments. The revisions have strengthened the manuscript.","URL":"https:\/\/wellcomeopenresearch.org\/articles\/11-134\/v1#referee-comment-6606","order":2,"name":"referee-comment-6606","label":"Referee Comment","group":{"name":"article-reports","label":"Article Reports"}},{"value":"This work was supported by Wellcome [227486\/A\/23\/Z] , European Union\u2019s Horizon 2020 programme  VEO [874735]. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.","order":3,"name":"grant-information","label":"Grant Information"},{"value":"This is an open access article distributed under the terms of the Creative Commons Attribution Licence, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.","order":0,"name":"copyright-info","label":"Copyright"}]}}