Development, Evaluation, and Multisite Deployment of a Machine Learning Decision Tree Algorithm To Optimize Urinalysis Parameters for Predicting Urine Culture Positivity.

PittUDT, a recursive partitioning decision tree algorithm for predicting urine culture (UC) positivity based on macroscopic and microscopic urinalysis (UA) parameters, was developed in support of a broader system-wide diagnostic stewardship initiative to increase appropriateness of UC testing. Reflex algorithm training utilized results from 19,511 paired UA and UC cases (26.8% UC positive); the average patient age was 57.4 years, and 70% of samples were from female patients. Receiver operating characteristic (ROC) analysis identified urine white blood cells (WBCs), leukocyte esterase, and bacteria as the best predictors of UC positivity, with areas under the ROC curve of 0.79, 0.78, and 0.77, respectively. Using the held-out test data set (9,773 cases; 26.3% UC positive), the PittUDT algorithm met the prespecified target of a negative predictive value above 90% and resulted in a 30 to 60% total negative proportion (true-negative plus false-negative predictions). These data show that a supervised rule-based machine learning algorithm trained on paired UA and UC data has adequate predictive ability for triaging urine specimens by identifying low-risk urine specimens, which are unlikely to grow pathogenic organisms, with a false-negative proportion under 5%. The decision tree approach also generates human-readable rules that can be easily implemented across multiple hospital sites and settings. Our work demonstrates how a data-driven approach can be used to optimize UA parameters for predicting UC positivity in a reflex protocol, with the intent of improving antimicrobial stewardship and UC utilization, a potential avenue for cost savings.

The impact of autocancellation of uncollected Clostridioides difficile specimens after 24 hours on reported healthcare-associated infections: A quality improvement intervention.

OBJECTIVE: To assess the impact of a 24-hour autocancellation of uncollected Clostridioides difficile samples in reducing reported healthcare-associated infections (HAIs). DESIGN: Quality-improvement, before-and-after implementation study. SETTING: The study was conducted in 17 hospitals in Pennsylvania. INTERVENTIONS: Clostridioides difficile tests that are not collected within 24 hours are automatically canceled ("autocancel") through the electronic health record. The intervention took place at 2 facilities (intervention period November 2021-July 2022) and subsequently at 15 additional facilities (April 2022-July 2022). Quality measures included percentage of orders canceled, C. difficile HAI rate, percent positivity of completed tests, and potential adverse outcomes of canceled or delayed testing. RESULTS: Of 6,101 orders, 1,090 (17.9%) were automatically canceled after not being collected for 24 hours during the intervention periods. The reported C. difficile HAI rates per 10,000 patient days did not significantly change. These rates were 8.07 in the 6-month preintervention period and 8.77 in the intervention period for facilities A and B combined (incidence rate ratio [IRR], 1.09; 95% CI, 0.88-1.34; P = .43), and were 5.23 HAIs per 10,000 patient days in the 6-month preintervention period and 5.33 in the intervention period for facilities C-Q combined (IRR, 1.02; 95% CI, 0.79-1.32; P = .87). From the preintervention to the intervention periods, the percent positivity rates of completed C. difficile tests increased by 1.1% for facilities A and B and by 1.4% for facilities C-Q. No adverse outcomes were observed. CONCLUSIONS: The 24-hour autocancellation of uncollected C. difficile orders reduced testing but did not result in reported HAI reduction.

Epidemiological profile of linezolid-resistant coagulase-negative staphylococci.

BACKGROUND: Surveillance studies have shown that <0.1% of coagulase-negative staphylococci are linezolid resistant; however, at our institution, 4% of such organisms were found to be resistant. We investigated the risk factors for and the epidemiological profile of linezolid-resistant coagulase-negative staphylococci. METHODS: Susceptibility testing and pulsed-field gel electrophoresis were performed to analyze the genetic relatedness of both linezolid-resistant and linezolid-susceptible isolates. Clinical data were retrieved from medical records, and a case-case-control study was performed to identify unique risk factors for linezolid resistance. RESULTS: Isolates recovered from 25 patients with linezolid-resistant coagulase-negative staphylococci were examined; all but 1 of the isolates were identified as Staphylococcus epidermidis, and all but 1 had a minimum inhibitory concentration of linezolid of >256 microg/mL. Pulsed-field gel electrophoresis showed that 21 (84%) of 25 linezolid-resistant isolates exhibited genetic relatedness, whereas linezolid-susceptible isolates were of diverse clones. Unique, independent predictors of linezolid resistance included receipt of linezolid in the 3 months preceding isolation of the coagulase-negative staphylococci (odds ratio, 20.6; 95% confidence interval, 5.8-73.0). CONCLUSION: Linezolid-resistant coagulase-negative staphylococci have emerged at our institution and are predominately of a single clone. We believe that the most likely scenario to explain this emergence is that person-to-person spread of linezolid-resistant coagulase-negative staphylococci led to establishment of skin colonization with the strain. Subsequent use of linezolid was followed by selection of the linezolid-resistant strain, which then became the dominant skin flora. The potential for a parallel scenario involving clonal dissemination followed by selection of linezolid-resistant methicillin-resistant Staphylococcus aureus is a real possibility.

SHV-type extended-spectrum beta-lactamase production is associated with Reduced cefepime susceptibility in Enterobacter cloacae.

Cefepime is a potentially useful antibiotic for treatment of infections with Enterobacter cloacae. However, in our institution the MIC(90) for E. cloacae bloodstream isolates is 16 microg/ml. PCR amplification of bla genes revealed that one-third (15/45) of E. cloacae bloodstream isolates produced SHV-type extended-spectrum beta-lactamases (ESBLs) in addition to hyperproduction of AmpC-type beta-lactamases. The majority (11/15) of ESBL producers also produced the TEM-1 beta-lactamase. The SHV types included SHV-2, -5, -7, -12, -14, and -30. All but two of the ESBL-producing E. cloacae isolates, but none of the non-ESBL-producing strains, had MICs of cefepime of >or=2 microg/ml. The MIC(90) for cefepime for ESBL-producing strains was 64 mug/ml, while for non-ESBL producers it was 0.5 microg/ml. Using current Clinical and Laboratory Standards Institute breakpoints for cefepime, two thirds (10/15) of ESBL-producing isolates would have been regarded as susceptible to cefepime. Phenotypic ESBL detection methods were generally unreliable with these E. cloacae isolates. Based on these results, pharmacokinetic, pharmacodynamic, and clinical reevaluation of cefepime breakpoints for E. cloacae may be prudent.