Estimating oncologist variability in prescribing systemic cancer therapies to patients in the last 30 days of life.

INTRODUCTION

Clinical guidelines and quality improvement initiatives have identified reducing the use of end-of-life cancer therapies as an opportunity to improve care. We examined the extent to which oncologists differed in prescribing systemic therapies in the last 30 days of life.

METHODS

Using Surveillance, Epidemiology, and End Results-Medicare data, we identified patients who died of cancer from 2012 to 2017 (N = 17,609), their treating oncologists (N = 960), and the corresponding physician practice (N = 388). We used multilevel models to estimate oncologists' rates of providing cancer therapy for patients in their last 30 days of life, adjusted for patient characteristics and practice variation.

RESULTS

Patients' median age at the time of death was 74 years (interquartile range, 69-79); patients had lung (62%), colorectal (17%), breast (13%), and prostate (8%) cancers. We observed substantial variation across oncologists in their adjusted rate of treating patients in the last 30 days of life: oncologists in the 95th percentile exhibited a 45% adjusted rate of treatment, versus 17% among the 5th percentile. A patient treated by an oncologist with a high end-of-life prescribing behavior (top quartile), compared to an oncologist with a low prescribing behavior (bottom quartile), had more than four times greater odds of receiving end-of-life cancer therapy (OR, 4.42; 95% CI, 4.00-4.89).

CONCLUSIONS

Oncologists show substantial variation in end-of-life prescribing behavior. Future research should examine why some oncologists more often continue systemic therapy at the end of life than others, the consequences of this for patient and care outcomes, and whether interventions shaping oncologist decision-making can reduce overuse of end-of-life cancer therapies.