Diagnostic Entropy as an Administrative Measure of Cross-System Diagnostic Dispersion in Noncommunicable Disease...

Existing comorbidity indices such as the Charlson Comorbidity Index (CCI) summarise comorbidity burden but do not capture how a patient's diagnoses are distributed across clinical speciality domains, a dimension hypothesised to relate to cross-system care coordination demands. This gap limits their utility for nursing workforce planning in settings with high multimorbidity prevalence.

This study aimed to provide preliminary evidence for the construct validity of diagnostic entropy as an administrative measure of cross-system diagnostic dispersion, by testing its incremental predictive value for prolonged hospitalisation, beyond diagnosis count and the CCI.

A retrospective methodological validation study used inpatient discharge records from a tertiary hospital in southwest China, 2023-2025 (n = 31,470). Diagnostic entropy was calculated using the Shannon entropy formula applied to ICD-10 chapter distributions across each patient's valid diagnoses. Four nested binary logistic regression models were fitted with LOS > 10 days as the primary outcome. Four sensitivity analyses examined robustness to extreme LOS values, coding inclusion rules, insurance-type stratification and repeat admissions from the same patient.

Diagnostic entropy had a mean of 1.746 (SD: 0.569); 72.3% of patients were in the high-entropy group (H ≥ 1.5). Entropy showed low-to-moderate correlation with the CCI (Spearman r = 0.304, p < 0.001) and moderate correlation with diagnosis count (r = 0.544, p < 0.001), suggesting that entropy captures a related but nonequivalent aspect of comorbidity-related complexity. After controlling for demographic characteristics, diagnosis count and CCI, entropy remained independently associated with prolonged hospitalisation (OR = 1.112, 95% CI: 1.072-1.153, p < 0.001), with an incremental Nagelkerke R² = 0.0014. This association was evident primarily among patients with seven or more diagnoses and was not statistically significant among patients with fewer diagnoses. High-entropy patients had longer stays, higher costs and higher rates of prolonged hospitalisation compared to low-entropy patients. Sensitivity analyses produced consistent results (OR range: 1.103-1.126).

These findings contribute discriminant evidence toward the construct validity of diagnostic entropy as an administrative measure of cross-system diagnostic dispersion. Whether this dispersion translates into greater nursing coordination burden remains a hypothesis requiring prospective validation.

Diagnostic entropy, calculated from routinely collected discharge diagnoses, may offer nurse managers a low-cost, scalable indicator of cross-system diagnostic dispersion among NCD inpatients. Wards with a higher proportion of high-entropy patients could potentially require greater coordination across speciality teams, though this application awaits prospective, multicentre validation linking entropy to nursing-sensitive outcomes.