Working Paper: CEPR ID: DP13157
Authors: Ity Shurtz; Alon Eizenberg; Adi Alkalay; Amnon Lahad
Abstract: Primary care is a notable example of a service industry where capacity-constrained suppliers face fluctuating demand levels. Unable to adjust prices, such providers may degrade service quality when faced with high demand levels. Little is known, however, on the nature of such adjustments in the primary care context. We study how physicians trade off one key input --- their time with patients --- with other inputs, such as prescriptions, lab tests and referrals. Employing detailed administrative data from eleven clinics of a large Israeli HMO, we use the absence of colleagues as a source of exogenous variation in physician workload. We find no evidence that physicians' workload affects the intensity with which they prescribe painkillers, or refer patients to the Emergency Room, and very little evidence for an effect on the prescription of antibiotics. We do find, however, that physician time and the use of diagnostic inputs are complements: a one minute decrease in the (daily) average visit length causes a 9 percent decrease in referrals to specialists, and a 3.8 percent decrease in referrals to lab tests. Following recent literature, we complement the traditional use of an exclusion restriction within a linear model by estimating non-parametric bounds on Average Treatment Effects using alternative assumptions. Such alternative estimators rule out the possibility that physician time and the use of diagnostic tools are substitutes in an economically-meaningful fashion, while still leaving a broad scope for the possibility that those are complements. Taken together, the results indicate that the shadow cost of physician capacity is not reflected in poor treatment decisions, but may instead be manifested in the underprovision of tests and referrals to specialists --- fundamental aspects of long-term preventive care.
Keywords: physician workload; capacity constraints; partial identification
JEL Codes: No JEL codes provided
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| physician workload (I11) | intensity of treatment (I12) |
| decrease in average visit length (C41) | increase in antibiotic prescriptions (H51) |
| decrease in average visit length (C41) | decrease in referrals to specialists (I11) |
| decrease in average visit length (C41) | decrease in lab test referrals (I11) |
| physician time and use of diagnostic inputs (I11) | underprovision of diagnostic tests and referrals (I11) |
| physician workload (I11) | limits on scope of issues addressed during visits (I10) |
| physician workload (I11) | adverse effect on long-term preventive care (I11) |
| physician workload (I11) | heterogeneous effects on older vs younger patients (J14) |