Is There a VA Advantage - First Stage, Balance and Reduced Form

Figure 1: First Stage, Balance, and Reduced Form .25 .3 VA .35 .4 .45 A: First Stage coeff. = .878 (.035) N = 401,319, J = 1,217 .05 .1 .15 Leave−out ambulance Pr(VA) .2 .25 (Predicted) 28−day mortality .095 .1 .105 B: Balance and Reduced Form .09 red. form coeff. = −0.040 (.016) balance coeff. = −.002 (.003) N = 401,319, J = 1,217 .05 .1 .15 Leave−out ambulance Pr(VA) Mortality .2 .25 Predicted Mortality Note: Panel A shows a binned scatterplot of arrival at a VA hospital on the y-axis against the ambulance leave-out propensity to arrive at a VA hospital on the x-axis. The figure is a graphical representation of the first-stage regression in Equation (3). Panel B shows binned scatterplots of 28-day mortality and predicted 28-day mortality on the y-axis against the ambulance leave-out propensity to arrive at a VA hospital on the xaxis. Mortality bin means are shown in solid circles, while predicted mortality bin means are shown in hollow circles. The figure represents the reduced-form regression in Equation (4) and the corresponding balance regression replacing mortality with predicted mortality. The sample includes 401, 319 ambulance rides and 1, 217 combinations of ambulance company identifiers and Dartmouth Atlas Hospital Referral Regions (HRRs). The sample selection is given in Appendix Table A.1. Baseline controls are detailed in Appendix Table A.2 and include patient zip code dummies, ALS/BLS dummies, source of the ambulance ride, time categories, and patient prior utilization. −.1 −.08 Effect of VA −.06 −.04 −.02 0 Figure 2: OLS and IV Specifications baseline controls 1 2 3 hold−out controls 4 5 6 7 Specification OLS 8 9 10 11 12 IV Note: This figure shows OLS and IV estimates of the effect of the VA on 28-day mortality, represented in Equation (2) as 𝛽, with progressive sets of controls. Numbered incremental controls correspond to categories or subcategories of variables that are presented in order in Appendix Tables A.2 and A.3. Estimates are shown along solid lines, while 95% confidence intervals are shown in dashed lines. All specifications use the baseline sample, given in Appendix Table A.1. Figure 3: Complier Potential Outcomes .6 .7 .8 .9 1 A: Survival 7 28 98 189 Days VA 280 364 280 364 Non−VA 0 .01 .02 .03 B: Mortality Hazard Rates 7 28 98 189 Days VA Non−VA Note: This figure shows potential outcomes for ambulance compliers who arrive at a VA hospital and those who arrive at a non-VA hospital. Panel A shows survival outcomes as a function of days from the ambulance ride. “Days” indicate one-week intervals from the ambulance ride. Denote 𝑆𝑖 (𝑡; 𝑑) ∈ {0, 1} as an indicator for whether patient 𝑖 survives up to time 𝑡 after the ambulance ride, depending on whether the patient arrives at the VA (𝑑 = 1) or a non-VA hospital (𝑑 = 0). Observed survival is 𝑆𝑖 (𝑡) = 𝐷 𝑖 𝑆𝑖 (𝑡; 1) + (1 − 𝐷 𝑖 ) 𝑆𝑖 (𝑡; 0). We estimate complier VA survival, or 𝐸 [ 𝑆𝑖 (𝑡; 1)| 𝑖 ∈ C], by an IV regression with a dependent variable of 𝑆𝑖 (𝑡) 𝐷 𝑖 , the endogenous VA treatment 𝐷 𝑖 , and the same first-stage and reduced-form design matrix implied by Equations (3) and (4). We estimate complier non-VA survival, or 𝐸 [ 𝑆𝑖 (𝑡; 0)| 𝑖 ∈ C], by a similar IV regression with a dependent variable of 𝑆𝑖 (𝑡) (𝐷 𝑖 − 1). All regressions use a sample of ambulance rides with no prior ride in the last year and the same baseline controls as described in Figure 1. Panel B presents implied weekly mortality hazard rates, as given by Equation (5). Figure 4: Observed Risk-Adjusted Outcomes .6 .7 .8 .9 1 A: Survival 7 28 98 189 Days VA 280 364 280 364 Non−VA 0 .01 .02 .03 B: Mortality Hazard Rates 7 28 98 189 Days VA Non−VA Note: This figure shows observed risk-adjusted outcomes for patients who arrive at a VA hospital and those who arrive at a non-VA hospital. Panel A shows survival outcomes as a function of days from the ambulance ride. “Days” indicate one-week intervals from the ambulance ride. Denote 𝑆𝑖 (𝑡; 𝑑) ∈ {0, 1} as an indicator for whether patient 𝑖 survives up to time 𝑡 after the ambulance ride, depending on whether the patient arrives at the VA (𝑑 = 1) or a non-VA hospital (𝑑 = 0). Observed survival is 𝑆𝑖 (𝑡) = 𝐷 𝑖 𝑆𝑖 (𝑡; 1) + (1 − 𝐷 𝑖 ) 𝑆𝑖 (𝑡; 0). We estimate VA survival, or 𝐸 [ 𝑆𝑖 (𝑡)| 𝐷 𝑖 = 1], by an OLS regression with a dependent variable of 𝑆𝑖 (𝑡) 𝐷 𝑖 and the same design matrix implied by Equation (2); we estimate non-VA survival, or 𝐸 [ 𝑆𝑖 (𝑡)| 𝐷 𝑖 = 0], by a similar OLS regression with a dependent variable of 𝑆𝑖 (𝑡) (𝐷 𝑖 − 1). All regressions use a sample of ambulance rides with no prior ride in the last year and the same baseline controls as described in Figure 1. Panel B presents implied weekly mortality hazard rates, as given by Equation (6). Figure 5: Complier Spending 0 10,000 20,000 30,000 40,000 A: Cumulative Spending 7 28 98 189 Days VA 280 364 Non−VA 0 1,000 2,000 3,000 B: Conditional Spending Rate 7 28 98 189 Days VA 280 364 Non−VA Note: This figure shows potential spending outcomes for ambulance compliers who arrive at a VA hospital and those who arrive at a non-VA hospital. Denote Spending𝑖 (𝑡; 𝑑) as the potential cumulative spending function for patient 𝑖 up to time 𝑡 after the ambulance ride, depending on whether the patient arrives at the VA (𝑑 = 1) or a non-VA hospital (𝑑 = 0). If a veteran 𝑖 dies at 𝑡, Spending𝑖 (𝑡; 𝑑) will be constant for all 𝑡 ≥ 𝑡. Panel A shows cumulative spending per patient as a function of days from the ambulance ride.  We estimate cumulative spending for compliers who arrive at a VA hospital, or 𝐸 Spending𝑖 (𝑡; 1) 𝑖 ∈ C , by an IV regression with a dependent variable of Spending𝑖 (𝑡) × 𝐷 𝑖 , the endogenous VA treatment 𝐷 𝑖 , and the same first-stage and reducedform  design matrix implied  by Equations (3) and (4). We estimate complier non-VA cumulative spending, or 𝐸 Spending𝑖 (𝑡; 0) 𝑖 ∈ C , by a similar IV regression with a dependent variable of Spending𝑖 (𝑡) (𝐷 𝑖 − 1). All regressions use a sample of ambulance rides with no prior ride in the last year and the same baseline controls as described in Figure 1. Panel B presents implied weekly spending rates for compliers, conditional on survival. Figure 6: Modal Hospital Effect and Health IT Adoption −.05 Effect of modal hospital 0 .05 A: Modal Hospital Effect by Year 2000 2005 2010 2015 Year 0 .2 Any technology adoption .4 .6 .8 1 B: Health IT Adoption by Year 2011 2012 2013 2014 Year Note: Panel A of this figure shows the IV estimate of the modal non-VA hospital effect on 28-day mortality by calendar year. The first-stage and reduced-form equations are given in Equations (10) and (11). The overall sample is the same alternative sample designed to study choice among non-VA hospitals for patients with only non-VA utilization in the prior year. Results for the overall IV estimates are shown in Appendix Figure A.10. Details of the sample selection are given in Appendix Table A.13. Estimates are shown in connected dots, while 95% confidence intervals are shown in dashed lines. Panel B of the figure shows the percent of rides going to hospitals after health IT adoption in our analytic sample. Health IT adoption is defined from a dataset from the Office of the National Coordinator of Health Information Technology (ONC). This dataset merges hospital attestation data from the Medicare EHR Incentive Program with certified EHR product information from ONC’s Certified Health IT Product List (CHPL), and we code the use of any certified product as health IT adoption. 2,862,557 8,828,997 Number of patients Number of ambulance rides 1,118,302 3,465,588 365,163 1,051,093 188,299 491,193 Sample characteristics Add zip × Add VA prior ambulance utilization 0.863 0.962 76.13 75.62 0.187 0.200 $20,874 $20,243 0.045 0.050 0.652 0.685 6.44 6.54 0.264 0.565 0.626 0.539 3.25 3.12 0.655 0.674 0.270 0.270 0.104 0.100 0.166 0.336 188,299 401,319 Add no ride in prior month 0.963 76.03 0.194 $20,905 0.051 0.705 6.14 0.529 0.482 2.16 0.684 0.269 0.097 0.330 Note: This table presents characteristics of observations remaining at each step of creating the baseline sample, detailed in Appendix Table A.1. Dually eligible 0.899 77.04 0.111 $21,724 0.255 0.610 6.53 0.136 0.695 2.77 0.696 0.272 0.115 0.044 Restrictions Male Age Black Income Rural zip code Residential source Comorbidity count Prior VA ED visit Prior Medicare ED visit Ambulance rides in prior year Advanced Life Support Weekend rate 28-day mortality Present at VA Add zip × hospital 0.883 76.89 0.163 $21,453 0.043 0.600 6.69 0.197 0.675 3.05 0.655 0.269 0.109 0.088 Table 1: Characteristics of Baseline Sample No No No No 0.878 (0.035) -0.045 (0.018) 0.097 401,319 -0.024 (0.001) 0.097 401,319 Yes No No No Yes Yes No No Yes Yes Yes No Dependent variable: 28-day mortality (2) (3) (4) A: OLS -0.023 -0.026 -0.022 (0.001) (0.001) (0.001) 0.097 0.097 0.097 401,319 401,319 401,319 B: IV 0.853 0.839 0.837 (0.034) (0.034) (0.034) -0.037 -0.036 -0.038 (0.018) (0.018) (0.017) 0.097 0.097 0.097 401,319 401,319 401,319 Yes Yes Yes Yes 0.860 (0.043) -0.049 (0.023) 0.097 401,319 -0.021 (0.001) 0.097 401,319 (5) Note: This table shows OLS and IV estimates of the effect of VA hospitals on 28-day mortality. Panel A gives OLS estimates, 𝛽ˆ𝑂𝐿𝑆 , for 𝛽 in Equation (2). Panel B gives IV estimates, 𝛽ˆ𝐼 𝑉 , as well as the first stage coefficient, 𝜋ˆ 1 in Equation (3), with respect to the leave-out probability of the assigned ambulance company to transport patients to the VA. Baseline controls in all specifications are described in Appendix Table A.2 and include patient zip code dummies, ALS/BLS dummies, source of the ambulance ride, time categories, and patient prior utilization. Patient background controls include demographics, socioeconomic status, combat history, eligibility for benefits, and counts of prior utilization. All additional controls are described in further detail in Appendix Table A.3. The estimation sample is described in Appendix Table A.1. Patient background controls Comorbidity controls Ambulance diagnosis controls Ambulance and co-rider controls Outcome mean Observations IV estimate First stage Outcome mean Observations VA hospital (1) Table 2: Effect of VA Hospitals on Mortality -0.090 (0.032) 0.589 401,319 -0.004 (0.003) 0.589 401,319 -0.468 (0.434) 4.380 401,319 0.514 (0.045) 4.380 401,319 Hospital days (2) ED revisits (3) A: OLS -0.036 (0.007) 0.318 401,319 B: IV 0.029 (0.044) 0.318 401,319 0.379 (0.174) 1.443 401,319 0.200 (0.017) 1.443 401,319 Outpatient visits (4) -2,548 (822) 12,173 401,319 932 (87) 12,173 401,319 Spending (5) Note: This table shows OLS and IV estimates of the effect of VA hospitals on various outcomes. Hospital days is defined as the number of inpatient days immediately following the ED visit; if the patient is not admitted from the visit, then this equals 0 for that visit. Outpatient visits is defined as the number of VA and non-VA outpatient visits within one month of the ride. ED revisits is defined as subsequent ED visits up to 14 days following the ride. Spending is defined as total spending over the 28 days following the ambulance ride. Panel A gives OLS estimates, 𝛽ˆ𝑂𝐿𝑆 , for 𝛽 in Equation (2). Panel B gives IV estimates, 𝛽ˆ𝐼 𝑉 , as well as the first stage coefficient, 𝜋ˆ 1 in Equation (3), with respect to the leave-out probability of the assigned ambulance company to transport patients to the VA. All regressions use baseline controls, which are described in Appendix Table A.2 and include patient zip code dummies, ALS/BLS dummies, source of the ambulance ride, time categories, and patient prior utilization. The estimation sample is described in Appendix Table A.1. Outcome mean Observations IV estimate Outcome mean Observations VA hospital Admission (1) Dependent variable Table 3: Effect of VA Hospitals on Other Outcomes Table 4: Complier Characteristics Male Overall 0.963 Age 76.0 Black 0.194 Income $20,905 Rural zip code 0.051 Residential source 0.705 Comorbidity count 6.143 Mental illness 0.427 Substance abuse 0.144 Prior VA ED visit 0.529 Prior Medicare ED visit 0.482 Ambulance rides in prior year 2.156 Advanced Life Support 0.684 Predicted VA user 0.847 Predicted mortality 0.097 Compliers 0.952 (0.006) 74.9 (0.433) 0.257 (0.028) $16,972 ($611) 0.091 (0.025) 0.647 (0.033) 5.447 (0.113) 0.444 (0.015) 0.163 (0.011) 0.712 (0.012) 0.336 (0.014) 2.178 (0.084) 0.600 (0.024) 0.939 (0.004) 0.095 (0.004) Ratio 0.99 [0.98 - 1.00] 0.99 [0.97 - 1.00] 1.33 [1.05 - 1.61] 0.81 [0.75 - 0.87] 1.78 [0.82 - 2.75] 0.92 [0.83 - 1.01] 0.89 [0.85 - 0.92] 1.04 [0.97 - 1.11] 1.13 [0.97 - 1.28] 1.35 [1.30 - 1.39] 0.70 [0.64 - 0.75] 1.01 [0.93 - 1.09] 0.88 [0.81 - 0.95] 1.11 [1.10 - 1.12] 0.98 [0.90 - 1.06] Note: This table presents average complier characteristics and the ratio between this average and the average among all veterans in the sample. Average complier characteristics and standard errors are calculated by performing two-stage least squares using the first stage Equation (3) and a reduced-form equation replacing the outcome variable in Equation (4) with 𝑋𝑖 𝐷 𝑖 , where 𝑋𝑖 is the characteristic corresponding to ride 𝑖. Regressions use baseline controls described in Appendix Table A.2; the regression sample is the baseline sample described in Appendix Table A.1. Standard errors for each average are presented in parentheses. The corresponding 95% confidence intervals for each ratio are presented in brackets. Table 5: Means of Hospital Characteristics VA Baseline sample Volume, Size, and Capabilities ED visits Admissions Average daily census Total staffed beds Teaching hospital Urban location Trama center Advanced cardiac care Hospital Sample Non-VA Baseline National sample average 17,780 6,310 227 322 0.59 0.90 0.11 0.74 41,600 13,676 200 283 0.27 0.97 0.64 0.70 38,416 12,445 182 258 0.22 0.89 0.61 0.58 Staffing ED staff per 1,000 ED visits Nurses per 1,000 patient-days Physicians per 1,000 patient-days Hospitalists per 1,000 patient-days Intensivists per 1,000 patient-days 0.78 6.20 5.31 0.17 0.08 0.55 5.45 7.75 0.29 0.16 0.73 5.63 8.73 0.40 0.20 Spending and Relative Outcomes Relative spending Mortality rate Readmissions rate 1.13 7.68 12.33 1.01 12.27 18.08 1.00 12.23 18.14 Payment and Organization Capitated lives covered Network participant Hospital system HMO PPO ACO 8,087 0.51 0.75 0.19 0.21 0.04 11,399 0.46 0.62 0.20 0.19 0.09 Health IT Adoption Adoption by 2011 Adoption by 2012 Adoption by 2013 Adoption by 2014 0.25 0.61 0.87 0.93 0.20 0.56 0.83 0.87 Note: This table presents average characteristics of VA and non-VA hospitals. Non-VA hospital characteristics are further presented for the baseline sample and for the national average. The national average weights hospital characteristics by their yearly admissions in the American Hospital Association (AHA) Annual Survey. The average in the baseline sample weights hospital characteristics by rides in that sample, described in Appendix Table A.1. Hospital characteristics are described in further detail in Appendix A.5. IT adoption Yes N/A 0.106 338,313 0.745 (0.011) -0.015 (0.009) -0.005 (0.001) Yes N/A No IT adoption Full No Yes ACO adoption Yes N/A Dependent variable: 28-day mortality (2) (3) (4) A: OLS -0.006 -0.012 (0.001) (0.005) -0.006 (0.001) -0.006 (0.001) B: IV 0.689 0.506 (0.008) (0.026) -0.004 -0.011 (0.006) (0.034) -0.015 (0.006) -0.005 (0.005) 0.113 0.112 0.107 1,075,528 1,414,197 58,968 Yes N/A No ACO adoption 0.112 1,354,196 0.703 (0.007) -0.006 (0.005) -0.006 (0.001) (5) Full No Yes -0.015 (0.019) -0.006 (0.005) 0.112 1,413,573 -0.008 (0.003) -0.006 (0.001) (6) Note: This table shows OLS and IV estimates of the effect of presenting to a veteran’s modal hospital on 28-day mortality, depending on whether the modal hospital has adopted health IT or whether the modal hospital has joined an Accountable Care Organization (ACO). Columns 1 and 2 show results estimated in subsamples defined by whether the modal hospital has adopted health IT or not. Columns 5 and 6 show results estimated in subsamples defined by whether the modal hospital has joined an ACO or not. The first-stage and reduced-form equations for the IV estimation (Panel B) are given in Equations (10) and (11); while this table presents results with hospital fixed effects, results do not qualitatively depend on the inclusion of hospital fixed effects. Columns 3 and 6 present results estimated on the overall sample with interactions for adoption status; these specifications are described in detail in Appendix A.7. We include baseline controls defined in Appendix Table A.2. The overall sample is the same alternative sample designed to study choice among non-VA hospitals for patients with only non-VA utilization in the prior year. Details of the sample selection are given in Appendix Table A.13, Panel B. Sample Outcome mean Observations Fixed effects Hospital identities Hospital ever adopted × No adoption × Adoption Modal hospital First stage × No adoption × Adoption Modal hospital (1) Table 6: Modal Hospital Mechanisms