Understanding the Dual Nature of Physician Preferences (Figures)

Figure 1: Histogram of Elasticities When Prices Increase Significantly Notes: Data from CMS Medicare 20% claims, 2003‐2010. This figure shows the elasticities (calculated simply as the annual percent change in quantity divided by the annual percent change in price) for HCPCS with annual price increases ranging from 45% to 55%. It is evident that quantity increases for about half of the HCPCS, while quantity falls for the other half. The right tail of elasticities is truncated at 1.5, excluding four outliers. Figure 2: Estimated Elasticities Plot (a): RVU + GAF Price Elasticities Plot (b): GAF Price Elasticities Notes: In Plot (a), data comes from the CMS Medicare 20% claims. Each pair of dots (own‐ and cross‐price elasticities) is estimated from a separate regression. In Plot (b), data comes from the CMS Medicare 5% claims. Each dot comes from a separate regression. For both plots, only the HCPCS with statistically significant own and cross price elasticity measures at the 5% level. The graph is truncated to show elasticities less than 5 or greater than ‐5. Table 1: Summary Statistics, by Sign of Own‐Price HCPCS Elasticity RVU+GAF, 2003-2010 PatientDriven (1) PhysicianDriven (2) GAF, 1993-2005 PatientDriven (3) PhysicianDriven (4) Panel A: Patient Socioeconomic Status Gross income ($1000s) Employer coverage Private coverage 31.85 0.40 0.14 33.04** 0.43** 0.15 27.84 0.38 0.116 28.73* 0.40*** 0.111 Less than high school Some college College grad More than college 0.29 0.17 0.34 0.17 0.27** 0.18** 0.36** 0.18** 0.328 0.160 0.316 0.156 0.307*** 0.167 0.323 0.157 Panel B: Patient Cost-Sharing OOP ($) Coinsurance ($) Deductible ($) 56.76 48.94 1.81 38.87*** 33.37*** 1.54* 19.79 17.67 2.11 14.42*** 12.48** 1.94 Any OOP (%) Fraction OOP 0.87 0.21 0.82*** 0.20*** 0.71 0.20 0.70 0.21 Panel C: Profitability Allowed - Submitted -667.55 -508.84** -125.66 -96.56* No. of Obs. (MCBS) No. of Obs. (CCF) 317 291 244 467 160 185 516 230 Notes: Data for Panels A is from MCBS 2003‐2010. Data from Panels B and C are from CCF 2003‐2010 or 1993‐2005. Columns (1) and (2), or Columns (3) and (4) are statistically different at the * 10% level, ** 5% level, *** 1% level. Sample universe limited to HCPCS with an identifiable substitute. In Panel A, the insurance coverage and education variables are measures of the fraction of patients with each characteristic. In Panel B, Any OOP is the fraction of patients who had OOP>0. Fraction OOP is the average fraction of total payments attributed to out of pocket costs. In Panel C, Allowed‐Submitted is the difference between the actual CMS payment and the amount providers bill CMS. Table 2: Probability of Being Physician‐Driven, by Changes in “Profitability” Changes in “Profitability” 1(Physician-driven) Above Median (1) Below Median (2) 0.740 0.406*** Notes: Column (1) shows the probability that the own‐price elasticity, calculated using changes in annual profitability that greater than the median, is positive. Column (2) shows the probability that the elasticity, calculated using changes in profitability less than the median, is positive. Above‐ and below‐ median are identified according to the data for each HCCPS. The means in columns (1) and (2) are statistically different at the *** 1% level. Profitability is measured using the “allowed‐submitted” measure. Table 3: Differential Effect of a Physician Reimbursement Reform Dependent Var: Log(Q), ∈ 2.9, 3.3 All (1) DualEligible (2) Not DualEligible (3) GAF x 1(Year 1997) x 1(Physician-Driven) 1.727*** (0.0478) 1.162*** (0.0250) -0.234*** (0.0265) -1.053*** (0.0258) 0.222*** (0.0275) -0.103*** (0.0343) 0.176*** (0.0357) 1.828*** (0.0779) 0.640*** (0.0428) -0.378*** (0.0415) -0.423*** (0.0443) 0.369*** (0.0432) -0.309*** (0.0563) 0.459*** (0.0584) 2.646*** (0.0492) 0.745*** (0.0280) -0.232*** (0.0280) -0.610*** (0.0291) 0.209*** (0.0293) -0.254*** (0.0382) 0.353*** (0.0399) R-squared No. of Obs. (CCF) 0.823 1,573,830 0.760 576,455 0.837 1,240,860 GAF 1(Year 1997) 1(Physician-Driven) GAF x 1(Year 1997) GAF x 1(Physician-Driven) 1(Year 1997) x 1(Phy. Driven) Notes: CCF data at the HCPCS‐county‐year level from 1993‐2002. All regressions control for year, county, and HCPCS fixed effects. Bootstrapped standard errors in parenthesis, * 10% level, ** 5% level, *** 1% level. Column (1) examines all Medicare beneficiaries. Column (2) considers the dual‐eligibles, and Column (3) focuses on the non‐dual eligibles. In each column, the sample is limited to HCPCS and counties with observations in each year. Table 4: Effect of Cost‐Sharing, Profitability, and Income on Elasticity Magnitudes Dependent Variable: Log(Q), Income (1) Log(Price) Log(Income) Log(Price) x Log(Income) 0.0970*** (0.0285) -0.113*** (0.0292) 0.0221*** (0.00659) Patient Cost Sharing (2) Price-Cost Margin (3) 0.210*** (0.0169) 0.135*** (0.0197) Log(Price) x Fraction OOP Covered Charges Log(Price) x Covered Charges Observations R-squared 9,696 0.982 9,696 0.982 All (4) 0.969*** (0.259) -0.118** (0.0484) 0.0717** (0.0316) -0.124*** (0.0293) 0.0246*** (0000661) 0.577*** (0.191) -0.264*** (0.0532) 1.1296*** (0.262) -0.132*** (0.0486) 9,696 0.982 9,696 0.982 0.370** (0.184) -0.215*** (0.0518) Fraction OOP 9.57 Notes: Robust standard errors in parenthesis, * 10% level, ** 5% level, *** 1% level. All regressions also include the log(substitute price index), CCI, and HCPCS and year fixed effects. Sample is limited to HCPCS with identifiable substitutes and patient income data from MCBS. Figure 5: Effects of Counterfactual Price Changes on Quantity 6% Percent Change in Total Quantity 4% 2% 0% 2.96% Scenario 1: Uniform Price Change Scenario 2: Selected Price Change ‐2% ‐4% ‐2.96% ‐6% ‐8% ‐10% ‐12% ‐14% ‐9.20% 20% Price Decrease 20% Price Increase ‐12.16% Notes: Scenario 1 shows the percent change in total quantity when 2010 prices uniformly decrease by 20% (dark red) or uniformly increase by 20% (light red). Scenario 2 shows the percent change in total quantity when 2010 prices decrease by 20% for only the physician‐driven HCPCS (dark red) or increase by 20% for only the patient‐driven HCPCS (light red).