Objective: To investigate inpatient rehabilitation rates after private total knee replacements (TKRs) in Australia since 2009; to quantify the contributions of hospital-, surgeon-and patient-related factors to predicting inpatient rehabilitation. Design: Retrospective cohort study; multivariate linear regression analysis of linked, de-identified Medibank administrative claims data and hospital casemix protocol data, adjusted for patient-related characteristics. Setting, participants: 35 389 patients undergoing Medibank-funded TKRs in 170 private hospitals in Australia, 2009−2016. Main outcome measures: Hospital inpatient rehabilitation rate; relative contributions of patient-and provider-related characteristics to variation in inpatient rehabilitation rates. Results: The overall inpatient rehabilitation rate increased from 31% in 2009 to 45% in 2016, but varied between hospitals (range, 0−100%). The reduction in mean acute length of stay during this period explained about 15% of this increase, and about 30% was explained by patient-related factors; more than half of the increase was explained by neither reduced length of stay or patient-related factors. Patient-related characteristics explained little of the variation in rates between hospitals. Rates at 27% of hospitals lay above the 95% confidence limit for the mean inpatient rehabilitation rate in private hospitals (38%), both before and after adjusting for patient-related factors. Provider characteristics explained three times as much of the variation as patient characteristics (75% v 25%); hospital-related factors made the largest contribution to variation (47%). Conclusion: Inpatient rehabilitation after TKR has increased in private health care during the past 8 years. Substantial variation in inpatient rehabilitation rates is not explained by patient-related factors, suggesting that some inpatient rehabilitation is low value care.