Predictors of readmission in nonagenarians: analysis of the American College of Surgeons National Surgical Quality Improvement Project dataset.

BACKGROUND

Increased longevity has led to more nonagenarians undergoing elective surgery. Development of predictive models for hospital readmission may identify patients who benefit from preoperative optimization and postoperative transition of care intervention. Our goal was to identify significant predictors of 30-d readmission in nonagenarians undergoing elective surgery.

METHODS

Nonagenarians undergoing elective surgery from January 2011 to December 2012 were identified using the American College of Surgeons National Surgical Quality Improvement Project participant use data files. This population was randomly divided into a 70% derivation cohort for model development and 30% validation cohort. Using multivariate step-down regression, predictive models were developed for 30-d readmission.

RESULTS

Of 7092 nonagenarians undergoing elective surgery, 798 (11.3%) were readmitted within 30 d. Factors significant in univariate analysis were used to develop predictive models for 30-d readmissions. Diabetes (odds ratio [OR]: 1.51, 95% confidence interval [CI]: 1.24-1.84), dialysis dependence (OR: 2.97, CI: 1.77-4.99), functional status (OR: 1.52, CI: 1.29-1.79), American Society of Anesthesiologists class II or higher (American Society of Anesthesiologist physical status classification system; OR: 1.80, CI: 1.42-2.28), operative time (OR: 1.05, CI: 1.02-1.08), myocardial infarction (OR: 5.17, CI: 3.38-7.90), organ space surgical site infection (OR: 8.63, CI: 4.04-18.4), wound disruption (OR: 14.3, CI: 4.80-42.9), pneumonia (OR: 8.59, CI: 6.17-12.0), urinary tract infection (OR: 3.88, CI: 3.02-4.99), stroke (OR: 6.37, CI: 3.47-11.7), deep venous thrombosis (OR: 5.96, CI: 3.70-9.60), pulmonary embolism (OR: 20.3, CI: 9.7-42.5), and sepsis (OR: 13.1, CI: 8.57-20.1), septic shock (OR: 43.8, CI: 18.2-105.0), were included in the final model. This model had a c-statistic of 0.73, indicating a fair association of predicted probabilities with observed outcomes. However, when applied to the validation cohort, the c-statistic dropped to 0.69, and six variables lost significance.

CONCLUSIONS

A reliable predictive model for readmission in nonagenarians undergoing elective surgery remains elusive. Investigation into other determinants of surgical outcomes, including social factors and access to skilled home care, might improve model predictability, identify areas for intervention to prevent readmission, and improve quality of care.