Identifying Increased Risk of Readmission and In-hospital Mortality Using Hospital Administrative Data
We read with interest the paper written by Moore et al1 suggesting that they were the first group planning a comorbidity index able to evaluate risk of readmission. We would like to report that in 2010, Crane et al2 demonstrated that by the use of an electronic medical record it could be possible to identify risk of readmissions. Authors evaluated retrospectively >12,000 adults aged 60 years and above, abstracting electronic medical records and administrative databases. They developed a score called Elder Risk Assessment (ERA) index, including age, marital status, length of hospital stay, and history of diabetes, coronary artery disease, congestive heart failure, stroke, chronic obstructive pulmonary disease, neoplasia, and dementia. The range of the score varied from −7 to 32, and patients with a score ≥16 had the highest risk of visits, emergency room visits/hospital admissions and hospital stay. Their final model had an area under the curve of 0.678 similar to results obtained by Moore and colleagues, their c-statistic was 0.634 (95% confidence interval, 0.633–0.634) for the readmissions index. In contrast Moore et al,1 used 29 comorbidity measures. In a recent study3 we tested 613 readmitted patients of 13,237 admissions. Age, sex, length of hospital stay, and deaths were retrospectively analyzed. Readmissions with diagnosis coincident with the index hospitalization were classified as avoidable, whereas those with a different diagnosis were defined as nonavoidable. When ERA score was ≥16, it was able to identify high-risk patients for readmission. Patients with nonavoidable readmissions were older, more frequently female, diabetic, and had higher ERA score than subjects with avoidable readmission. Multivariate logistic regression analysis showed that nonavoidable readmissions were independently associated with female sex, and age, whereas only age and ERA score were independently associated with death at the end of follow-up.