Background: Hospital readmissions are common, detrimental for patients and associated with significant costs for the healthcare system. The aim of this study was to evaluate Charlson, Charlson age, simplified HOSPITAL and LACE score in 30-day non-elective readmission prediction at a Brazilian tertiary care teaching public hospital in Southern Brazil.
Methods: Retrospective cohort study including all hospital discharges from the Internal Medicine Service at the study facility from September to November 2017. To analyze the simplified HOSPITAL score, the following variables were assessed: hemoglobin level at discharge < 12 g/L, cancer diagnosis, sodium level at discharge < 135 mmol/L, index type of admission: non-elective, hospital admissions during the previous 12 months and length of stay ≥ 5 days. To analyze the LACE score, length of stay, acute (emergent) admission, comorbidity (Charlson) and visits to emergency department during previous 6 months were evaluated. The area under the ROC curve (AUC) was calculated to evaluate the ability of the score in predicting readmissions at the same hospital and other two electronic medical records connected hospitals. Comparison of ROC curves was performed by DeLong test. A P < 0.05 was considered statistically significant.
Results: A total of 1,038 hospital discharges were included [median age 65.9 (IQR: 51.9-75.9) years, 52.1% female]. Cancer diagnosis was described in 201 (19.4%) index admissions [median Charlson score 2 (1-3) points]. Low hemoglobin and sodium level at discharge were observed in 627 (60.4%) and 89 (8.7%) of index hospitalizations, respectively. Most of the index admissions were non non-elective (1,026, 98.8%) and a length of stay ≥ 5 days was observed in 923 (88.9%) [median 10 (IQR: 6-17) days]. Rate of ≥ 2 admissions during the previous 12 months was 15.5% (≥ 2 visits to emergency department during the previous 6 months were 28.2%). Overall 30-day non-elective readmission rate was 14.5% (95%CI: 12.4-16.6%). AUC was 0.67 (95%CI: 0.63-0.72; P < 0.001) for Charlson age, 0.68 (95%CI: 0.63-0.73; P < 0.001) for simplified HOSPITAL, 0.69 (95%CI: 0.65-0.74; P < 0.001) for Charlson and 0.72 (95%CI: 0.67-0.76; P < 0.001) for LACE score. When AUCs were compared, the only statistical significant observed difference was between Charlson age and LACE score (P = 0.03).
Conclusions: The analyzed scores showed only poor (AUC 0.67-0.69) at best fair discriminatory power (AUC 0.72 for LACE score) to predict 30-day non-elective readmission. LACE score had a better AUC when compared to Charlson age (P = 0.03), with no other statistical significance observed. Local adjustments in the scores should be done to improve they performance. Identification of a high readmission risk group of patients before hospital discharge could allow prioritize efforts to make a better transition of care.