Development and validation of a risk prediction model for acute care use among older advanced cancer patients on...

Patients with advanced cancer are at risk for unplanned emergency department (ED) visits and hospital stays. The purpose was to develop and validate a risk prediction model to identify patients at the highest risk for acute care use.

We identified advanced cancer patients aged 65 years and older treated on SWOG trials from 1999 to 2014 using data linked to Medicare claims. The primary outcome was acute care use (ED visits or hospital stays). A 60% random sample training set was used to identify candidate variables. An adverse risk model was built by summing adverse factors and creating high- vs low-risk groups by splitting at the median. This risk model was tested in the 40% validation set.

Among 1397 patients from 6 trials, 839 composed the training set. The proportion of patients with at least 1 hospital stay or ED visit was 67.5%. Adverse risk factors were performance status (0 vs ≥1), coronary artery disease, hypertension, and liver disease. Patients with at least 2 factors (high risk; 57.3%) vs 0/1 risk factor (low risk; 42.7%) were more likely to experience acute care (79.6% vs 51.1%, respectively), corresponding to a more than threefold increase in odds ratio (OR = 3.38, 95% confidence interval = 2.48 to 4.62). Results were similar in the test set, indicating successful validation. Among all patients, quartile-level proportions were 48.9% for zero risk factors vs 84.0% for 3/4 risk factors. The C statistic was 0.703.

A limited set of 4 variables predicted a threefold increased risk of acute care use in older patients with advanced cancer. Personalized interventions aimed at preventing acute care use could improve the quality of cancer care.