Loveland S (Bedford COE), Rosen AK
Hospital quality is commonly assessed using risk-adjusted mortality rates. However, better quality of care has also been associated with reduced readmission rates. We examined whether potential patient safety events [Patient Safety Indicators (PSIs)] were associated with readmission rates. Since the PSIs have been shown to be related to longer inpatient stays, higher mortality, and costs, we assumed that readmission rates would be higher for hospitalizations with PSI events than for hospitalizations without these events, indicating poorer quality care.
We combined FY03 through FY07 inpatient files, and obtained dates of death from the Vital Status file. We ran the AHRQ PSI software, obtaining observed PSI rates (PSIs related to death were excluded). Logistic regression models estimated the likelihood of readmission (readmission within 30 days of discharge), controlling for “any PSI event” (any of 13 PSIs) and comorbidities. For hospitalizations with readmissions, we examined type of readmission (emergent or elective) and whether the readmission occurred in the same hospital.
There were 1,916,486 acute-care hospitalizations: of those, 81.7% were single admissions without death or readmission, 2.7% ended in death, and 15.6% resulted in readmissions within 30 days. Comorbidity rates were generally higher in hospitalizations with readmissions (e.g., diabetic rates were 22.4% in hospitalizations without readmission vs. 24.6% in those with readmission). Observed rates of any PSI were 13.9 per 1,000 for discharges without readmissions/death, 18.2 for readmissions, and 72.9 for deaths. The probability of readmission was higher for hospitalizations with PSIs than those without PSIs; OR = 1.24 (95%CI 1.21-1.28). Among hospitalizations with readmissions, 86% were readmitted to same hospital, 43% were readmitted with trauma DRGs or during off-hours, and 52% were readmitted with the same Major Diagnostic Category.
Safety events were associated with higher readmission rates. Compared to hospitalizations without readmissions, those with readmissions had higher disease burden and were likely to be readmitted emergently. Readmission was diagnostically linked to the index hospitalization in half the readmissions.
Utilization of more than one outcome is important for accurate quality assessment. Inclusion of readmission as a quality indicator may flag additional opportunities for quality improvement missed by mortality.