Background: The simplified HOSPITAL score is an easy-to-use prediction model that accurately identifies patients at high-risk of 30-day unplanned readmission before hospital discharge. The predictors include the last available hemoglobin and sodium levels at discharge. Because an earlier stratification risk of readmission would allow more preparation time for transitional care interventions, we aimed to assess whether the score would perform similarly by using hemoglobin and sodium levels at time of admission instead of discharge.
Methods: We prospectively screened all consecutive adult patients discharged home from the department of general internal medicine at four hospitals. We excluded patients who refused to give consent, who died during hospitalization, or who stayed hospitalized for less than 24 hours. The primary outcome was the composite of first unplanned readmission or death within 30 days after discharge of index admission. We compared the performance of the simplified score with lab at discharge and with lab at admission according to their discriminatory power (Area under the Receiver Operating Characteristic Curve, AUROC) and the Net Reclassification Improvement (NRI). The diagnostic characteristics were calculated using the cut-points that maximize Youden’s index.
Results: Among the 923 patients included, 126 (13.5%) had a 30-day unplanned readmission or death. The median hemoglobin levels at admission and discharge were 13.2 and 12.4 g/dL, and the median sodium levels were 137 and 139 mmol/L, respectively. Both score versions categorized 316 patients (34%) as likely to be readmitted. The two versions of the simplified HOSPITAL score showed both a very good accuracy (Brier score 0.11, 95% confidence interval 0.10-0.13) and model fit (Hosmer-Lemeshow goodness-of-fit test p=0.38, and p=0.52, respectively). Their AUROC were 0.64 (95%CI 0.59-0.69), and 0.65 (95%CI 0.60-0.70), respectively, without statistical difference (p=0.51). The optimal cut-point according to Youden’s index was 4 score points. Compared with the model at discharge, the model with lab at admission showed improvement in discrimination based on the continuous NRI (29%; 95% CI 11-47; P=0.003) mainly driven by an improvement in prediction of non-cases that compensated the worse prediction for cases: 75% (596/797) non-cases had a lower risk if lab at admission was used, while 60% (50/126) cases had a lower risk if lab at admission was used.
Conclusions: The simplified HOSPITAL score using lab values at time of admission performs at least as good as with lab values at discharge in order to identify high-risk patient for 30-day unplanned readmission or death. The simplified HOSPITAL score with lab at admission offers a readmission risk stratification early during the hospital stay, and therefore allows a timely preparation of transition care interventions to the patients who may benefit the most.