Can the application of machine learning to electronic health records guide antibiotic prescribing decisions for suspected urinary tract infection in the Emergency Department?

Urinary tract infections (UTIs) are a major cause of emergency hospital admissions, but it remains challenging to diagnose them reliably. Application of machine learning (ML) to routine patient data could support clinical decision-making. We developed a ML model predicting bacteriuria in the ED and evaluated its performance in key patient groups to determine scope for its future use to improve UTI diagnosis and thus guide antibiotic prescribing decisions in clinical practice. We used retrospective electronic health records from a large UK hospital (2011-2019). Non-pregnant adults who attended the ED and had a urine sample cultured were eligible for inclusion. The primary outcome was predominant bacterial growth ≥104 cfu/mL in urine. Predictors included demography, medical history, ED diagnoses, blood tests, and urine flow cytometry. Linear and tree-based models were trained via repeated cross-validation, re-calibrated, and validated on data from 2018/19. Changes in performance were investigated by age, sex, ethnicity, and suspected ED diagnosis, and compared to clinical judgement. Among 12,680 included samples, 4,677 (36.9%) showed bacterial growth. Relying primarily on flow cytometry parameters, our best model achieved an area under the ROC curve (AUC) of 0.813 (95% CI 0.792-0.834) in the test data, and achieved both higher sensitivity and specificity compared to proxies of clinician's judgement. Performance remained stable for white and non-white patients but was lower during a period of laboratory procedure change in 2015, in patients ≥65 years (AUC 0.783, 95% CI 0.752-0.815), and in men (AUC 0.758, 95% CI 0.717-0.798). Performance was also slightly reduced in patients with recorded suspicion of UTI (AUC 0.797, 95% CI 0.765-0.828). Our results suggest scope for use of ML to inform antibiotic prescribing decisions by improving diagnosis of suspected UTI in the ED, but performance varied with patient characteristics. Clinical utility of predictive models for UTI is therefore likely to differ for important patient subgroups including women <65 years, women ≥65 years, and men. Tailored models and decision thresholds may be required that account for differences in achievable performance, background incidence, and risks of infectious complications in these groups.

Impact of antibiotic use on patient-level risk of death in 36 million hospital admissions in England.

OBJECTIVES: Initiatives to curb hospital antibiotic use might be associated with harm from under-treatment. We examined the extent to which variation in hospital antibiotic prescribing is associated with mortality risk in acute/general medicine inpatients. METHODS: This ecological analysis examined Hospital Episode Statistics from 36,124,372 acute/general medicine admissions (≥16y) to 135 acute hospitals in England, 01/April/2010-31/March/2017. Random-effects meta-regression was used to investigate whether heterogeneity in adjusted 30-day mortality was associated with hospital-level antibiotic use, measured in defined-daily-doses (DDD)/1,000 bed-days. Models also considered DDDs/1,000 admissions and DDDs for narrow-spectrum/broad-spectrum antibiotics, parenteral/oral, and local interpretations of World Health Organization Access, Watch, and Reserve antibiotics. RESULTS: Hospital-level antibiotic DDDs/1,000 bed-days varied 15-fold with comparable variation in broad-spectrum, parenteral, and Reserve antibiotic use. After extensive adjusting for hospital case-mix, the probability of 30-day mortality changed -0.010% (95% CI: -0.064,+0.044) for each increase of 500 hospital-level antibiotic DDDs/1,000 bed-days. Analyses of other metrics of antibiotic use showed no consistent association with mortality risk. CONCLUSIONS: We found no evidence that wide variation in hospital antibiotic use is associated with adjusted mortality risk in acute/general medicine inpatients. Using low-prescribing hospitals as benchmarks could help drive safe and substantial reductions in antibiotic consumption of up-to one-third in this population.

Development of risk prediction models to predict urine culture growth for adults with suspected urinary tract infection in the emergency department: protocol for an electronic health record study from a single UK university hospital.

BACKGROUND: Urinary tract infection (UTI) is a leading cause of hospital admissions and is diagnosed based on urinary symptoms and microbiological cultures. Due to lags in the availability of culture results of up to 72 h, and the limitations of routine diagnostics, many patients with suspected UTI are started on antibiotic treatment unnecessarily. Predictive models based on routinely collected clinical information may help clinicians to rule out a diagnosis of bacterial UTI in low-risk patients shortly after hospital admission, providing additional evidence to guide antibiotic treatment decisions. METHODS: Using electronic hospital records from Queen Elizabeth Hospital Birmingham (QEHB) collected between 2011 and 2017, we aim to develop a series of models that estimate the probability of bacterial UTI at presentation in the emergency department (ED) among individuals with suspected UTI syndromes. Predictions will be made during ED attendance and at different time points after hospital admission to assess whether predictive performance may be improved over time as more information becomes available about patient status. All models will be externally validated for expected future performance using QEHB data from 2018/2019. DISCUSSION: Risk prediction models using electronic health records offer a new approach to improve antibiotic prescribing decisions, integrating clinical and demographic data with test results to stratify patients according to their probability of bacterial infection. Used in conjunction with expert opinion, they may help clinicians to identify patients that benefit the most from early antibiotic cessation.

Diagnostic uncertainty and urinary tract infection in the emergency department: a cohort study from a UK hospital.

BACKGROUND: Suspected urinary tract infection (UTI) syndromes are a common reason for empirical antibiotics to be prescribed in the Emergency Department (ED), but differentiating UTI from other conditions with a similar presentation is challenging. We investigated how often an ED diagnosis of UTI is confirmed clinically/microbiologically, and described conditions which present as UTI syndromes. METHODS: Observational study using electronic health records from patients who attended the ED with suspected UTI and had a urine sample submitted for culture. We compared the ED diagnosis to diagnosis at discharge from hospital (ICD-10 codes), and estimated the proportion of cases with clinical/microbiological evidence of UTI. RESULTS: Two hundred eighty nine patients had an ED diagnosis of UTI syndrome comprising: lower UTI (191), pyelonephritis (56) and urosepsis (42). In patients admitted to hospital with an ED diagnosis of lower UTI, pyelonephritis or urosepsis, clinical/microbiological evidence of UTI was lacking in 61/103, 33/54 and 31/42 cases respectively. The ED diagnosis was concordant with the main reason for admission in less than 40% of patients with UTI syndromes, and antibiotics were stopped within 72 h in 37/161 patients. CONCLUSIONS: Clinical/microbiological evidence of UTI was lacking in 60-70% of patients, suggesting scope to revise empirical prescribing decisions for UTI syndromes in light of microbial culture and clinical progression.

Surveillance of infection severity: a registry study of laboratory diagnosed Clostridium difficile.

BACKGROUND: Changing clinical impact, as virulent clones replace less virulent ones, is a feature of many pathogenic bacterial species and can be difficult to detect. Consequently, innovative techniques monitoring infection severity are of potential clinical value. METHODS AND FINDINGS: We studied 5,551 toxin-positive and 20,098 persistently toxin-negative patients tested for Clostridium difficile infection between February 1998 and July 2009 in a group of hospitals based in Oxford, UK, and investigated 28-day mortality and biomarkers of inflammation (blood neutrophil count, urea, and creatinine concentrations) collected at diagnosis using iterative sequential regression (ISR), a novel joinpoint-based regression technique suitable for serial monitoring of continuous or dichotomous outcomes. Among C. difficile toxin-positive patients in the Oxford hospitals, mean neutrophil counts on diagnosis increased from 2003, peaked in 2006-2007, and then declined; 28-day mortality increased from early 2006, peaked in late 2006-2007, and then declined. Molecular typing confirmed these changes were likely due to the ingress of the globally distributed severe C. difficile strain, ST1. We assessed the generalizability of ISR-based severity monitoring in three ways. First, we assessed and found strong (p<0.0001) associations between isolation of the ST1 severe strain and higher neutrophil counts at diagnosis in two unrelated large multi-centre studies, suggesting the technique described might be useful elsewhere. Second, we assessed and found similar trends in a second group of hospitals in Birmingham, UK, from which 5,399 cases were analysed. Third, we used simulation to assess the performance of this surveillance system given the ingress of future severe strains under a variety of assumptions. ISR-based severity monitoring allowed the detection of the severity change years earlier than mortality monitoring. CONCLUSIONS: Automated electronic systems providing early warning of the changing severity of infectious conditions can be established using routinely collected laboratory hospital data. In the settings studied here these systems have higher performance than those monitoring mortality, at least in C. difficile infection. Such systems could have wider applicability for monitoring infections presenting in hospital.

Comparison of Acinetobacter baumannii isolates from the United Kingdom and the United States that were associated with repatriated casualties of the Iraq conflict.

Acinetobacter isolates associated with casualties from the Iraq conflict from the United States were compared with those from the United Kingdom by pulsed-field gel electrophoresis and integron analysis. Representatives of the main outbreak strain associated with casualties from both countries were indistinguishable in DNA profile. Two further outbreak strains were common to both sets of isolates.