Applying a logistic regression-clustering joint model to analyze the causes of prolonged pre-analytic turnaround time for urine culture testing in hospital wards

TL;DR

This paper introduces a model combining logistic regression and clustering to identify and fix delays in urine culture testing workflows in hospitals.

Contribution

A novel joint model is proposed to analyze and improve pre-analytic turnaround time in urine culture testing.

Findings

Specimens collected during specific hours had significantly higher delay odds.

Targeted interventions reduced pre-TAT overtime rates and contamination rates significantly.

The joint model successfully identified workflow bottlenecks and improved specimen quality.

Abstract

In this study, we developed and validated a logistic regression-clustering joint model to: (1) quantify multistage workflow bottlenecks (collection/transport/reception) in urine culture pre-TAT prolongation (>115 min); and (2) assess the efficacy of targeted interventions derived from model-derived insights.

Using complete workflow data obtained from 1,343 urine culture specimens (January 2024–March 2024) collected at a tertiary hospital, we integrated binary logistic regression analysis with K-means clustering to quantify delay patterns. The analyzed variables included collection time, ward type, personnel roles, and patient demographics. Post-intervention data (May 2024–July 2024, *n* = 1,456) was also analyzed to assess the impact.

Analysis of the critical risk factors revealed that specimens collected between 04:00–05:59/10:00–11:59 had 142.92-fold higher delay odds (95% CI: 58.81–347.37). Those collected on SICU/ICU wards showed 9.98-fold higher risk (95% CI: 5.05–19.72) than general wards. Regarding intervention efficacy, pre-TAT overtime rates decreased by 58.6% (13.48% → 7.55%, P < 0.01). Contamination rate decreased by 59.8% (5.67% → 2.28%, P < 0.01). The median pre-TAT decreased by 15.9% (44 → 37 min, P < 0.01).

The joint model effectively identified workflow bottlenecks. Targeted interventions (dynamic transport scheduling, standardized training, and IoT alert systems) significantly optimized pre-TAT and specimen quality, providing a framework for improving clinical laboratory processes.

Figures