Predicting hospital readmissions in the oncology population.

177 Background: The 30-day readmission rate is established as an important indicator of quality of care. The LACE index is commonly used in the general medical setting to predict readmission but its ability to predict readmission with sensitivity and specificity in the oncology population has not yet been examined. At our cancer center, palliative care (PC) consultation is associated with an increased risk for readmission but it is not an element in the LACE index.

METHODS: We sought to characterize the operating characteristics of the LACE Index using receiver operating characteristics analyses to predict unplanned readmissions to our cancer center over a 6-week period beginning March 2016. Data was gathered from chart review to calculate a total LACE score for each unplanned admission. Logistic regression was used to examine the individual components of the LACE index and whether a PC consult improved the performance of the index.

RESULTS: A total of 329 patients with unplanned admissions were included. Fifty-nine (17.9%) were readmitted within 30 days of discharge. There was no difference between the median LACE scores of those readmitted compared to those who were not (Md = 10.0; p = .93). Receiver operating characteristic (ROC) curve analyses of LACE scores yielded an area under the curve estimate relative to 30-day readmissions of .45 indicative of poor overall accuracy. ROC analyses also showed that the previously established LACE cutoff score of 10 had sensitivity of .54 and specificity of .57 relative to readmissions. The positive predictive value was .81 and the negative predictive value was .18. In logistic regression analysis, only direct referral center/emergency department visits were an independent predictor of readmission, with a c-statistics of .64 for readmission. The inclusion of a PC consult did not improve the performance of the index.

CONCLUSIONS: The LACE Index performed poorly in predicting 30-day readmission in the oncology setting; the inclusion of whether a PC consult took place did not improve the index's utility. Further research is required to create a new tool or enhance existing indices to predict readmission in the oncology population.