Feasibility of identifying prognostic factors for total joint arthroplasty in electronic primary health care records

Abstract

The ability to identify osteoarthritis (OA) patients at high risk of progressing to joint arthroplasty surgery could enable earlier, targeted nonsurgical management in primary care. The ability to do so using only information routinely recorded within the primary care electronic healthcare records (EHR) would have several advantages but the feasibility of this is unclear. The aim of this thesis was to identify factors associated with future primary total knee or hip arthroplasty (TKA/THA) that were available within the EHR. Initially, a systematic review identified 35 published articles reporting 42 factors potentially associated with primary TKA or THA in patients with osteoarthritis. A series of searches was then undertaken which obtained codelists based on Read morbidity and process of care codes and prescription medications based on the British National Formulary (BNF) subchapters for 13 of these factors. These codelists and a search of the consultation free text were then used in case-control studies of 874 patients receiving primary THA/TKA between 2005-2011 and 4,370 age-sex-practice-matched controls in the Consultations in Primary Care Archive (CiPCA) database. These analyses were used to determine which factors (i) met the minimum prevalence among cases and controls to warrant further analysis (3% in prior 5 years); (ii) were associated with the outcome of primary TKA/THA (unadjusted p<0.05). To identify other potential factors, an additional ‘hypothesis free’ analysis was conducted examining the associations between outcome and all third- level Read codes and BNF subchapters (unadjusted odds ratio <0.75 or >1.33). In total 92 and 79 factors met the minimum prevalence and were associated with TKA and THA respectively. After adjusting for OA, 106 and 83 risk factors were associated with TKA and THA respectively. The studies in this thesis have identified ‘building blocks’ for a future multivariable risk prediction algorithm based within the primary care EHR.