Background:
Up to 25% of hospitalized adults experience early readmission. Pre‐discharge identification of patients likely to be readmitted would help target discharge‐planning to reduce readmissions, exposures to hospital‐associated risks, and costs. Pra scores of outpatients >65 years of age is a valid predictor of a hospital admission within 4 years. This study used the Pra score in hospitalized adults to estimate prediction of readmission within 1 month of an index hospitalization.
Method:
A cohort of willing internal medicine service adults in a Midwestern tertiary medical center whose life expectancy was > 6 months consented to study participation in June‐July of 2005. Proxies were used for cognitively or verbally impaired patients. Before discharge trained assistants interviewed subjects, abstracted the medical record, and derived Pra scores (using an equation incorporating age, gender, number of admissions and doctor/clinic visits in the last year, subjective health status, caretaker availability, and history of diabetes or coronary artery disease (CAD). Early readmission was defined as an admission for any reason, except for elective surgery, within 31 days of discharge from the index hospitalization. The 75th percentile was used to classify patients as likely (≥ .48) or unlikely (< .48) to be readmitted. After Day 31 the records (from the index hospital and 2 local hospitals) were queried for subjects' readmissions between discharge (day 0) and day 31. The analysis used percentages (with frequencies) for nominal variables and medians (with range) for skewed interval variables. Predictive values were reported with 95% confidence intervals (CI).
Summary of Results:
Of 156 consecutive patients, 35 were discharged before screened, while 60% (72/121) were eligible and consented. The sample was 56% (40/72) female; median age = 53 (18 ‐ 97), admissions = 1 (1 ‐14), and doctor/clinic visits = 10 (0 ‐100). Caretakers were available for 90% (67/72), 33% (24/72) had diabetes, and 18% (13/72) had CAD. Subjective health status was rated as poor/fair 60%, good/very good 37%, and excellent 3%. Readmissions occurred with 11% of all subjects. Predictive values included: sensitivity 75% (6/8), (CI): 0.45 ‐1.05; specificity 69% (44/64), CI: 0.57 ‐ 0.80; positive predictive value 23% (6/26), CI: 0.07 ‐ 0.39; negative predictive value 96% (44/46), CI: 0.88 ‐1.04; and odds ratio 6.143, CI: 1.28 ‐ 28.75.
Statement of Conclusions:
Multi‐institutional replication using larger samples is needed to confirm Pra's predictive ability. Within this heterogeneous sample of medical inpatient adults, readmission of high Pra versus low Pra patients was 6 times more likely. Pra's promising predictive ability may add valuable discharge planning information.
Author Disclosure Block:
N.L. Novotny, None.