Predicting acute care use following the initiation of systemic therapy for solid tumors.

Abstract

6 Background: Emergency department (ED) visits and hospitalizations are undesirable and costly. We developed and validated the PROACCT (PRediction Of Acute Care use during Cancer Treatment) score to predict at least one acute care visit during the first 30 days (AC30) after initiating systemic therapy for cancer. Methods: Using administrative data, we identified patients in Ontario with the 18 most common solid tumors who started systemic treatment from July 1, 2014 to June 30, 2015. Southwestern and Northeastern Ontario formed the development and validation cohorts, respectively. We created a score to predict AC30 using multivariate logistic regression in the development cohort. Combinations of tumor sites and regimens were grouped into quintiles based on AC30. The score was assessed in the validation cohort. Results: AC30 occurred in 23% (4438/19359) of patients. Eleven factors predicted AC30 and formed the score: tumor site and regimen (2nd quintile: 2; 3rd-4th: 3; 5th: 4), recent ED visit (2), recent palliative radiation (1), rural residence (1) and Edmonton Symptom Assessment Scale anxiety (4+: 1), lack of appetite (4+: 1), pain (4+: 1), and wellbeing (4+: 1). Among the 204 tumor-regimen combinations, tumors with poor prognoses, such as pancreatic and lung, and platinum- and taxane-containing regimens, carried the highest risk for AC30. The score had moderate discrimination (c-statistic 0.65; P< 0.001) and strong calibration (Table) in the validation cohort. Conclusions: PROACCT identifies factors that predict AC30 in patients starting systemic treatment for solid tumors and could be incorporated into electronic health records to select patients for preventative interventions. [Table: see text]