Abstract
Introduction
Patient-reported outcome (PRO) data are variably collected before and after total hip/knee arthroplasty (THA/TKA). We assessed the generalizability of incentivized, prospectively collected PRO data for THA/TKA patient-reported outcome performance measure (PRO-PM) development.
Methods
The Centers for Medicare & Medicaid Services (CMS) received PRO data voluntarily submitted by hospitals in a bundled payment model for THA/TKA procedures. Participating hospitals who collected and successfully submitted these data received an increase in their overall quality score, possibly resulting in a positive impact on model reconciliation payments. PRO data were collected from Medicare Fee-For-Service beneficiaries >= 65 years undergoing elective primary THA/TKA procedures from July 1 to August 31, 2016 at hospitals participating in the model. Pre-operative PRO and risk variable data were collected 0 – 90 days prior to surgery, while post-operative PRO data were collected 270 – 365 days following elective THA/TKA. PRO pre-op and post-op data were matched to Medicare claims data for determination of clinically eligible procedures and clinical comorbidities. We compared the characteristics of patients submitting PRO data to other elective primary THA/TKA recipients in the US.
Results
Four patient characteristics were associated with HOOS Jr. mean change scores (sex, narcotic use in past 90 days, other joint pain, and back pain) and four with KOOS Jr. mean change scores (sex, Hispanic ethnicity, other joint pain, and back pain). The frequency of simultaneous bilateral procedures, dementia, trauma, and dialysis were statistically significantly lower in patients submitting PRO data compared to other US Medicare beneficiaries undergoing elective primary THA/TKA, but no difference was greater than 1.5% absolute percentage points between groups.
Conclusions
Offering financial incentives in alternative payment models to encourage PRO data collection and submission can produce generalizable data for PRO measure development. The possibility of non-respondent biases will need to be specifically considered in measure development.
