Performance of Pediatric Readmission Measures

BACKGROUND: Readmission rates are frequently used as a hospital quality metric; yet multiple measures exist to evaluate pediatric readmission rates. We sought to assess how four different measures of pediatric readmission compare with assessment of both preventable and unplanned readmission. METHODS: Clinicians on hospital medicine, cardiology, neonatology, and neurology teams reviewed medical records for 30-day readmissions using an abstraction tool with high interrater reliability for preventability assessment. Readmissions between July 2014 and June 2016 were classified separately as preventable or not preventable and planned or unplanned. We compared the classifications to four existing readmission metrics: all-cause readmission, unplanned readmission/time flag classification, the pediatric all-condition readmission, and potentially preventable readmission. We calculated sensitivity and specificity for all readmission metrics. RESULTS: Among 30-day readmissions considered, 1,643 were eligible for medical record review; 1,125 reviews were completed by the clinical teams (68%). On medical record review, the majority of readmissions were determined not preventable (85%). Only 15% were classified as unplanned and preventable. None of the four readmission measures had appropriate sensitivity or specificity for identifying preventable readmission. The unplanned readmission/time flag classification had the highest sensitivity (95%) and specificity (90%) in identifying unplanned readmissions. CONCLUSION: None of the existing pediatric readmission measures can reliably determine preventability. The unplanned readmission/time flag measure performed best in identifying unplanned readmissions.

Differences in 30-day preventable readmission rates after index cancer surgery in New York State hospitals.

6640 Background: 30-day readmission rates are currently being used as a measure of performance quality. Among surgical patients, readmissions may be reducible for certain complications such as deep venous thrombosis or wound infection. We report 30-day readmission rates for potentially preventable readmissions following surgical treatment of the most common malignancies in the US. Methods: The most common cancer hospitalizations were identified from the Healthcare Cost and Utilization Project. Previously reported ICD-9 codes of preventable readmissions from cancer surgery were used to assess 30-day readmissions in New York State in 2009. We measured comorbidity using CMS hierarchical condition categories. Hospital teaching status was based on the American Hospital Association designation. Random effect hierarchical logistic regression models were run to account for clustering within hospitals. Results: 21,945 index admissions for cancer surgery occurred in 2009 at 169 teaching and 73 non-teaching hospitals. The most common operations were for prostate, breast, colon, lung, and renal cancer. 51% of patients were male and 12% were black. The overall readmission rate was 9.3% with readmissions being higher in non-teaching hospitals (11.2%) vs. teaching hospitals (8.6%) (p<0.0001). There was a significant interaction between hospital teaching status and patient race. In teaching hospitals, there was no racial difference in readmission. However, in non-teaching hospitals, black patients were more likely to be readmitted (15.1% vs 10.9%; p=0.02). Multivariate models found that being male (OR=1.17; 95% CI: 1.04; 1.31; p=0.007), undergoing surgery at a non-teaching hospital (OR=1.16; 95% CI: 1.00; 1.35; p=0.048), black race (OR=1.47; 95% CI: 1.04; 2.08; p=0.029), and certain comorbidities increased a patient’s risk of 30-day readmission for a preventable cause. Conclusions: The 30-day preventable readmission rate after index hospitalizations for cancer surgery is higher in non-teaching hospitals, and this difference is more pronounced for black patients. Clinical protocols in teaching hospitals may play a role in this phenomenon. Efforts to address remediable causes of this disparity are warranted.