Working Paper: NBER ID: w24054
Authors: David H. Howard; Jason Hockenberry; Guy David
Abstract: Advocates for “personalized medicine” tests claim they can reduce health care spending by identifying patients unlikely to benefit from costly treatments. But most tests are imperfect, and so physicians have considerable discretion in how they use the results. We show that when physicians face incentives to provide a treatment, the introduction of an imperfect prognostic test will increase treatment rates. We study the interaction of incentives and information in physicians’ choice between conventional radiotherapy and intensity modulated radiation therapy (IMRT) for Medicare patients with breast cancer. IMRT is far more costly. Patients with left-side tumors are more likely to benefit from IMRT, though it is unnecessary for the vast majority of patients. IMRT use is 18 percentage points higher in freestanding clinics, where physician-owners share in the lucrative fees generated by IMRT, than in hospital-based clinics. Patients with left-side tumors are more likely to receive IMRT in both types of clinics. However, IMRT use in patients with right-side tumors (the low benefit group) treated in freestanding clinics is actually higher than use in patients with left-side tumors (high benefit group) treated in hospital-based clinics. Prognostic information affects use but does nothing to counter incentives to overuse IMRT.
Keywords: personalized medicine; physician incentives; intensity-modulated radiation therapy; breast cancer
JEL Codes: I11; I18
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| introduction of an imperfect prognostic test (C52) | increase in treatment rates (C22) |
| type of clinic (freestanding vs. hospital-based) (I11) | likelihood of receiving IMRT (C67) |
| tumor laterality (left-side vs. right-side tumors) (L15) | treatment decisions (D87) |
| physician incentives (J33) | treatment rates for patients with left-side tumors (C34) |
| financial incentives (M52) | treatment thresholds (C22) |
| introduction of personalized medicine tests (C91) | increased treatment rates (C22) |
| observed differences in IMRT receipt (I14) | not solely attributable to observable patient characteristics (D91) |