Health Care Demand Elasticities by Type of Service among the Privately Insured

The structure of health insurance coverage for different types of health care spending has implications for health plan choice, access to care, health care cost, biased selection, risk adjustment, and the extent of insurance protection. It is well understood that health care services differ in their responsiveness to insurance coverage, and yet few studies have examined this heterogeneity comprehensively.

We estimate own-price demand elasticities for distinct health care services, parsing total spending by types of service in three ways. First, we estimate elasticities by the broad categories of inpatient care, outpatient and prescription drugs. Second, using a combination of both service location (surgery ward, mental health facility, etc) and procedures (MRI, specialty visit, etc), we classify spending into 10 types of services, which enables us to distinguish primary care, mental health/substance abuse, prevention, surgery and such. Finally we consider demand responsiveness for 27 types of services which allows even more granularity.

Our analysis uses Truven’s MarketScan private claims database between 2007 and 2013 to estimate two-part models of spending on each type of service as a function of the cost share for that service. For the first part, we model the decision to seek care for each type of service while for the second part we model conditional spending on a type of service given at least some spending. Patient health is controlled for by individual fixed effects and risk scores, while endogeneity of cost share (the consumer’s price) is controlled for by IV techniques.

Our sample includes unbalanced panel data on more than 30 million individuals, insured through contracts with 87 large employers, each of which offers multiple types of health plans (PPO, HMO, etc). Our identification strategy relies on two different types of variation in cost sharing. First, employers vary annually in the set of health plans offered, which also vary in their copays, coinsurance, deductibles and stoplosses. This induces cost share variation over time, across consumers, and across health plans. We use employer offering of an insurance plan type, alone and interacted with the enrollees’ prior year risk scores, as instruments for plan choices. Second, in health plans with deductibles and stoplosses, as consumers spend past these thresholds in a given year, prices go down. This is in contrast with other plans with fixed copayments or coinsurance that do not decline during the year.  A simple time trend or monthly dummies is an appropriate instrument for this within-year cost share variation.

Preliminary results using five employers with 2.2 million person-years of spending show significant variation in elasticities across types of services. Provisional estimates find that annual outpatient spending is 2.5 times as responsive to cost sharing as the average of all health care services, while inpatient services are 35 percent less responsive than the average. Pharmacy spending is 25 percent more responsive than the average of all spending. At a finer level, services such as surgery and mental health/substance abuse are the most elastic, while lab tests and emergency rooms are the most inelastic.