Page 110 - The value of total hip and knee arthroplasties for patients

Chapter 6
Discussion
The primary findings of the analyses were 1) that patient expectations for the outcome ofTHA andTKA consistently are part of a prediction model that predicts the outcomes pain and function 1 year post-operative. 2) that the more general CEQ expectancy subscale explains slightly more variance in function in TKA and function and pain inTHA as compared to the HSS total knee or total hip arthroplasty expectation surveys.
Comparisons with the literature
This study is, to the best of our knowledge, the first to assess the predictive value of expectations within a prediction model in which multiple other clinical and sociodemographic variables were entered; we however do want to discuss our results in the light of previous findings regarding patients’ outcome expectations for TKA and THA. Several studies have been published that examine the association between pre-operative expectations and outcomes of TKA and THA.These studies analyze their data from an etiological perspective, the aim those studies is to determine whether a particular independent variable really affects the dependent variable, and to estimate the magnitude of that effect [42;43] .Thus, in such studies patients’ expectations are the determinant of interest while other variables are regarded as confounders of the relationship between patients’ expectations and outcomes. For these studies contradictive results are found; some studies show a positive associations which suggest that higher expectations are related to better outcomes, others find no association or even negative associations[44].This variability in results of studies may be caused by the type of expectations examined, the measurement approach used, the outcome assessed, the timing of the outcome assessment or the use of univariate versus multivariate statistical methods[14].These studies however do not answer the question as to whether patient’s expectations can be used in clinical practice to predict the clinical course of the disorder. To answer this question one has to examine whether the predictive value increases by including the expectation variables in the regression analysis. Our study does answer that question by examining expectations within prediction models which “seek to get optimal predictions based on a linear combination of whatever variables are available” [43]. In our study we chose to include candidate variables in the multivariate models that mimic clinical routine, i.e. are easily accessible for professionals because they are already par t of regular anamnesis and routine outcome measurement. Our
108