Predictors of Inter-Hospital Transfer from a Secondary-Level Hospital in Trincomalee District, Sri Lanka

Abstract

Background: Inter-hospital transfers from secondary-level hospitals are often driven by patient severity and limited access to specialized care. Identifying key predictors in Trincomalee District, Sri Lanka, is essential to improve referral efficiency and patient outcomes. Aims: To identify demographic, clinical, and institutional factors associated with inter-hospital transfer from a secondary-level public hospital in Sri Lanka and to develop a practical predictive model for administrative decision-making. Study Design: Retrospective observational study with supplementary qualitative inquiry. Place and Duration of Study: Base Hospital Kanthale, Trincomalee District, Sri Lanka, using records of patients admitted from January 2025 to April 2025. Methodology: Secondary data were extracted for 357 adult patients aged 16 years and above using a structured checklist. Patients with pre-arranged referral plans and those requiring immediate resuscitation and intubation were excluded. Variables covered demographic factors, clinical status, and institutional service availability. Semi-structured interviews with the medical administrator and clinicians involved in acute care were used to contextualize transfer decisions. Descriptive statistics, chi-square testing, analysis of variance, and multivariable logistic regression were used. Predictive performance was assessed using accuracy, precision, recall, and F1 score. Results: Of 357 patients, 79 (22.0%) were transferred to a higher level of care. Mean age was 49.8 years, and 53.3% were male. Pulse rate differed significantly between transferred and non-transferred groups (P = .023). Significant categorical associations were found for level of consciousness (P = .003), specialist-related service availability (P < .001), CT/MRI availability (P < .001), and day of admission (P < .001). In the adjusted model, semi-conscious state (OR 0.26, P = .044), specialist-related service constraint (OR 1032.45, approximate 95% CI 126.08–8459.79; P < .001), CT/MRI-related constraint (OR 10.69, P < .001), and admission on public holidays (OR 50.98, P < .001) were significant predictors. Model accuracy was 91.3%, while precision, recall, and F1 score were each 80.3%. Conclusion: Inter-hospital transfer was driven mainly by institutional constraints rather than demographic characteristics alone. Limited specialist coverage, restricted access to advanced imaging, and service pressure during public holidays were major determinants. The predictive model may support better resource allocation and more rational transfer decisions.