Risk factors and prediction models for the presence of bacterial pathogens with clinically significant antimicrobial resistance phenotypes: a systematic review

Thu, 20 Aug 2026 15:17:46 +0000

Objective.

To systematically synthesize factors associated with the presence, colonization, infection, or acquisition of bacterial pathogens with clinically significant antimicrobial resistance phenotypes; to assess the variability of available quantitative estimates of these associations; and, separately, to characterize prediction models, clinical scores, and algorithms developed to estimate the probability of such outcomes.

Materials and Methods.

A systematic review was conducted of publications identified through a PubMed/MEDLINE search covering the period from January 1, 2000, to April 30, 2026, using terms related to risk factors and prediction, antimicrobial resistance, and clinical infectious syndromes. The review was prepared in accordance with Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 principles. Of 2,277 bibliographic records identified, 1,316 underwent detailed assessment and role classification after initial screening. Primary risk-factor studies and prediction-model studies were analyzed separately. For risk-factor studies, characteristics of the investigated factors, clinical events, resistance phenotypes, study populations, and quantitative measures of association were extracted and standardized. Prediction models were analyzed at the level of distinct model units, including predictors, model structure, discrimination, calibration, and validation. Risk of bias was assessed using Quality In Prognosis Studies (QUIPS) for risk-factor studies and PROBAST+AI for prediction models. Owing to substantial clinical and methodological heterogeneity, the primary synthesis was descriptive.

Results.

The primary synthesis included 382 publications: 317 risk-factor studies and 65 predictionmodel studies. A total of 1,071 risk-factor records were extracted and, after within-publication consolidation, yielded 1,001 analytical units. Quantitative estimates were available for 738 (68.9%) records, and at least one adjusted estimate was reported in 179 (56.5%) studies. The most frequently represented domains included prior antimicrobial therapy, hospitalization or other healthcare exposure, demographic characteristics, comorbidities, invasive devices and procedures, and prior microbiological history or previously documented antimicrobial resistance. According to QUIPS, the overall risk of bias was low in 3 (0.9%) studies, moderate in 291 (91.8%), and high in 23 (7.3%).

Conclusions.

The evidence base on risk factors for antimicrobial resistance is concentrated around a limited set of recurring clinical domains. However, quantitative estimates of their associations vary substantially according to the resistance phenotype, clinical event, study population, and healthcare setting. Accordingly, the identified factors should primarily be considered candidate variables for subsequent risk stratification rather than universally applicable quantitative effect estimates. Prediction models demonstrate the feasibility of multidimensional risk assessment, but their direct clinical implementation is limited by heterogeneity, limited reproducibility, infrequent reporting of calibration, and scarce external validation. Practical use of published risk factors and prediction models requires assessment of their transportability and mandatory evaluation in the local target population.

Babicheva A.I. on КМАХ

Risk factors and prediction models for the presence of bacterial pathogens with clinically significant antimicrobial resistance phenotypes: a systematic review

Objective.

Materials and Methods.

Results.

Conclusions.