Idea: Disentangle “taste” for familiarity from legitimate improvement in patient outcomes due to strength of PCP/specialist relationship.
How: Estimate a risk-adjustment model for orthopedic surgery outcomes, based on some existing literature, that includes an effect of relationships.
See evaluation reports on CJR: https://innovation.cms.gov/innovation-models/cjr
(near bottom of page)
Detailed lists of risk-adjustment variables:
First Annual Report, Exhibit 31, page 68
Second Annual Report appendix, p. E-26 (Exhibit E-16 for claims-based outcomes)
Third Annual Report appendix, Section E.III.D.2, p. E-28 (Exhibit E-15 for claims-based outcomes)
Fourth Annual Report appendix, p. C-28 (Exhibit C-9 for claims-based outcomes)
Einav et al. on the CJR mentions using "[f]ully saturated specifications that additionally control for all available patient characteristics still leave at least half of the variation unexplained." https://doi.org/10.1093/qje/qjab035
From Atul Gupta:
I think a better risk adjustment model is what CMS is using in the new version of BPCI, which they call “BPCI Advanced.” The risk adjustment for patient risk here seems more standard. This document provides a nice guide to implement it. see section 4.1.1. and Table 2. https://innovation.cms.gov/files/x/bpciadvanced-targetprice-my3.pdf
Idea: Disentangle “taste” for familiarity from legitimate improvement in patient outcomes due to strength of PCP/specialist relationship.
How: Estimate a risk-adjustment model for orthopedic surgery outcomes, based on some existing literature, that includes an effect of relationships.
See evaluation reports on CJR: https://innovation.cms.gov/innovation-models/cjr (near bottom of page) Detailed lists of risk-adjustment variables: First Annual Report, Exhibit 31, page 68 Second Annual Report appendix, p. E-26 (Exhibit E-16 for claims-based outcomes) Third Annual Report appendix, Section E.III.D.2, p. E-28 (Exhibit E-15 for claims-based outcomes) Fourth Annual Report appendix, p. C-28 (Exhibit C-9 for claims-based outcomes)
Einav et al. on the CJR mentions using "[f]ully saturated specifications that additionally control for all available patient characteristics still leave at least half of the variation unexplained." https://doi.org/10.1093/qje/qjab035
From Atul Gupta: I think a better risk adjustment model is what CMS is using in the new version of BPCI, which they call “BPCI Advanced.” The risk adjustment for patient risk here seems more standard. This document provides a nice guide to implement it. see section 4.1.1. and Table 2. https://innovation.cms.gov/files/x/bpciadvanced-targetprice-my3.pdf