Personal clinical history predicts antibiotic resistance in urinary tract infections

Abstract

The prevalence of antibiotic resistance in urinary tract infections (UTIs) often renders the prescribed antimicrobial treatment ineffective, highlighting the need for personalized prediction of resistance at time of care. Here, crossing a 10-year longitudinal dataset of over 700,000 community-acquired UTIs with over 6,000,000 personally-linked records of antibiotic purchases, we show that the resistance profile of infections can be predicted based on patient-specific demographics and clinical history. Age, gender, and retirement home residence had strong, yet differential and even non-monotonic, associations with resistance to different antibiotics. Resistance profiles were also associated with the patient's records of past urine samples and antibiotic usage, with these associations persisting for months and even longer than a year. Drug usage selected specifically for its own cognate resistance, which led indirectly, through genetic linkage, also to resistance to other, even mechanistically unrelated, drugs. Applying machine learning models, these association patterns allowed good personalized predictions of resistance, which could inform and better optimize empirical prescription of antibiotics.The resistance of bacterial pathogens to commonly used antibiotics is a growing public health concern, threatening the efficacy of life-saving antibiotic drugs 1,2 . Antibiotic use and misuse can benefit resistant strains, exacerbating the problem over time [3][4][5] . At the single patient level, the efficacy of antimicrobial treatment is critically dependent on correctly matching antibiotic choice to the specific susceptibilities of the pathogen 6-8 . Ideally, correct prescription should be based on direct measurement of the antibiotic susceptibilities of the infecting pathogen. In practice, though, to save time and resources, drugs are often prescribed empirically in absence of culture susceptibility measurements, risking incorrect and ineffective treatment. difficult to discern the direction of causality 17,35 . Finally, the time extent of these positive and negative associations of resistance with prior antibiotic usage and prior resistant samples is not well resolved, and it is also unclear whether and how these associations vary across resistances to different antibiotics.Here, we present a systematic big-data analysis of a large population of UTI patients to unravel predictive features of antibiotic resistance. We analyze a patient-level longitudinal dataset of community and retirement-home acquired UTI cultures collected by Maccabi Healthcare Services (MHS), Israel's second largest Health Maintenance Organizations, serving a diverse population of~2 million patients. We first analyze correlations between demographic factors and antibiotic resistance. Then, comparing resistance data of multiple infections from the same patient, we unravel long-term "memory" of resistance over time. We also combine these culture records with patient-linked records of antibiotic use to quantify the extent and time of direct and indi...

Recent Advances in the Race to Design a Rapid Diagnostic Test for Antimicrobial Resistance

“…Thus, a faster time-to-readout can be achieved or even a higher noise level can be allowed in an assay, solely by including a specialized computer algorithm into the assay. Even more futuristic may be the ability to predict resistance of pathogens simply by patient history . As of now, machine learning for assisting AST platforms is only bottlenecked by the programming knowledge of those designing platforms, as well as the limited clinical data sets collected in some cases.…”

Machine learning based prediction of antibiotic sensitivity in patients with critical illness

“…In the space of bacterial infection, retrospective analyses of etiology and descriptive studies have been reported in regular intervals 31–33 . However, collective and methodical applications of predictive modeling is yet to be extensively implemented in the field of empiric treatment via antimicrobial sensitivity analysis barring a few interesting attempts 34,35 . Also the data used in the existing studies were not publicly available.…”