Press Ganey Administration of Hospital Consumer Assessment of Healthcare Providers and Systems Survey Result in a Biased Responder Sample for Hip and Knee Arthroplasties.

BACKGROUND: Press Ganey administration of Hospital Consumer Assessment of Healthcare Providers and Systems Survey after discharge is sometimes used to monitor hospital performance and adjust reimbursements. HYPOTHESIS: significant differences exist between responders and nonresponders. We assessed baseline characteristic differences between responders and nonresponders; sampling bias; responding predictability in total joint arthroplasty patients; and differences in sampling and response rates among different hospitals.
METHODS: A prospective database of arthroplasty patients from January 1, 2016 to September 30, 2016 was used to compare responders' and nonresponders' baseline characteristics at 4 hospitals. A univariate analysis between groups was performed. A multiple logistic regression model was used to assess whether Press Ganey sampling was predictable. We identified receiving and responding predictors.
RESULTS: We captured 96.6% (3255 of 3369) of hip and knee arthroplasties. Hospital Consumer Assessment of Healthcare Providers and Systems Survey sampling rate was 60% and response rate was 36% (1157). Responders were more likely Caucasians, nonsmokers, discharged home, have shorter hospital stays, have higher baseline joint pain and physical composite scores, and have better mental health composite scores. Concordance indices suggest reasonable-to-very-strong model predictability for those sampled (range 0.56-0.91) and those responding (range 0.61-0.78). Completion predictors were Caucasian race (P < .0001), younger ages (P < .0001), discharged home (P < .0001), negative smoking status (P = .02), quit smoking (P = .0026), higher baseline mental health composite scores (P = .0096), and diagnoses of femoroacetabular impingement (P = .0056), osteoarthritis (P = .0111), or prosthesis failure (P = .0036).
CONCLUSION: Responders/nonresponders were significantly different in several characteristics. It can be predicted who will likely be sampled and who will complete. Responders were not representative of arthroplasty population. Research is needed for more representative sampling methods. LEVEL OF EVIDENCE: Level III.