Exploring heterogeneity in hospital length of stay post emergency abdominal surgery in Ireland: A quantile regression approach

Abstract

Background Emergency Abdominal Surgery (EAS) refers to a range of complex intra-abdominal surgical procedures associated with high mortality risk and long length of hospital stay (LOS). LOS is often used as a proxy measure for hospital resource utilisation in hospital capacity management and planning. Our objective was to explore the heterogeneity in LOS among EAS patients admitted at publicly funded hospitals in Ireland. Methods We analysed national hospital inpatient data (2014 – 2022) for adults discharged following EAS. We used Quantile Regression (QR) methods to explore the heterogeneous effects along the LOS distribution between 10 th - 90 th percentiles. We compared QR with Ordinary Least Squares (OLS) estimates, and identified from which point in the LOS distribution heterogeneous effects were different from OLS estimates. Results From the National Healthcare Quality Reporting System (NHQRS) records for 15,408 EAS adult inpatient episodes were obtained for analysis. We observed significant (p < 0.001) heterogeneous effects across most quantiles of the LOS distribution. LOS was longer for patients with Charlson comorbidity indices of 4 or higher, American Society of Anaesthesiologists physical status scores of 2 and higher, admissions to critical care units, hospital readmissions within 30-days, discharges to nursing home and other hospital, and for patients treated in Model 4 hospitals. LOS was shorter for patients with a cancer diagnosis and patients who died during admission. Across these factors, statistically significant heterogeneous effects above OLS estimates were observed at the 70 th to the 90 th quantile. Conclusions The QR methods identified the presence of significant heterogeneity across the entire LOS distribution. Relative to OLS mean estimates, QR is a better method for identifying heterogeneous effects by exploring the entire LOS distribution. Our results highlight the importance of using appropriate methods for estimating skewed outcomes. This is important to provide valid and relevant empirical analysis to inform policy.