Predischarge blood count and nutritional risk marker patterns associated with 30-day unplanned readmissions for hematological toxicity after radiotherapy for nasopharyngeal carcinoma: a multicenter cohort study

BackgroundEarly unplanned readmissions after radiotherapy-based treatment may reflect residual toxicity and gaps in supportive care coordination. We examined associations between predischarge hematological and albumin marker patterns and 30-day hematological toxicity-related unplanned readmission in nasopharyngeal carcinoma (NPC).MethodsThis multicenter retrospective cohort study included 607 patients discharged after hospitalization involving radiotherapy-based treatment at three tertiary hospitals in southwestern China between December 2018 and March 2026. The primary complete-case sample comprised 588 patients with predischarge absolute neutrophil count (ANC), platelet count (PLT), and albumin measurements. Core markers were ANC <1.5 × 10⁹/L, PLT <200 × 10⁹/L, and albumin <35 g/L. We used association rule mining and Firth penalized logistic regression. Chemotherapy and supportive care variables were examined in sensitivity analyses, and all-cause 30-day unplanned readmission was a supportive secondary outcome.ResultsBased on the chronologically first readmission within 30 days, 77 of 607 patients (12.7%) were classified as having an unplanned readmission; 58 of the 77 (75.3%) were classified as hematological toxicity-related. In the primary sample, hematological readmission rates were 4.9%, 13.1%, and 17.5% for zero, one, and two or more core markers, respectively. Adjusted odds ratios (ORs) were 2.86 [95% confidence interval (CI) 1.50–5.46] and 4.33 (95% CI 1.82–10.30) for one and two or more markers versus zero markers. Corresponding ORs were 2.93 and 4.55 after chemotherapy adjustment and 2.59 and 3.47 after supportive care adjustment. The all-cause analysis showed a similar gradient (OR 2.65, 95% CI 1.52–4.61; OR 3.18, 95% CI 1.43–7.05).ConclusionRoutine predischarge measures were associated with early readmission risk, but the findings do not establish causality or a clinical decision rule. Prospective external validation is required before clinical use.