Table 3. Health Care Use and Spending During First Year of CJR (2016)a.
| Mean (SD) Value in Control MSAs (n = 121) | Mean Difference (95% CI) Between Bundled Payment and Control MSAs | P Valueb | |
|---|---|---|---|
| Discharge destination, % | |||
| Institutional postacute carec | 33.7 (11.2) | −2.9 (−4.95 to −0.90) | .005 |
| Home health agencyd | 32.2 (19.4) | 1.3 (−1.73 to 4.34) | .40 |
| Home (without home health agency)d | 32.2 (23.3) | 2.6 (−0.79 to 5.90) | .14 |
| Otherd | 1.9 (3.0) | −0.40 (−0.84 to 0.05) | .08 |
| No. of days in institutional postacute care | 8.2 (2.7) | −0.48 (−1.06 to 0.11) | .12 |
| Medicare episode spending in institutional postacute care, $ | 3871 (1394) | −307 (−587 to −27) | .04 |
| Total Medicare episode spending, $e | 22 872 (3619) | −453 (−909 to 3) | .06 |
| Medicare spending net of reconciliation payment, $e,f | 22 872 (3619) | 234 (−214 to 683) | .31 |
Abbreviations: CJR, Comprehensive Care for Joint Replacement program; MSA, metropolitan statistical area.
Change with bundled payment was estimated with the use of 2-stage least-squares instrumental variable regression, with random assignment to CJR used as an instrument for inclusion in CJR. All regressions include indicators for strata and control for 2 lags of the dependent variable. All outcomes were measured during the episode. All measures are based on 2016 Medicare claims data unless otherwise noted; episodes starting between April 1, 2016, and September 15, 2016, are included; lags of the dependent variable are based on the corresponding periods in 2013 and 2014. N = 196 MSAs included overall. See Table 2 for variable definitions.
P values for differences are based on heteroskedasticity-robust standard errors.
Primary outcome.
The preanalysis plan specified that discharges to home health agencies would be combined with discharges to institutional postacute care rather than being analyzed separately, and analysis of the remaining categories was not specified. The changes are nonsubstantive and designed for ease of exposition.
Because control MSAs do not receive reconciliation payments, control MSA total Medicare episode spending and Medicare spending net of reconciliation payment are identical.
Medicare spending net of reconciliation payment equals total Medicare episode spending (previous row) plus any reconciliation payments made by Medicare. Data on reconciliation payment were downloaded from https://innovation.cms.gov/Files/x/cjr-py1reconpym.xlsx. Because reconciliation payment did not exist prior to 2016, analyses instead controlled for lags of total Medicare episode spending.