Towards a general practice electronic health record definition of clinical OA
The National Institute for Health and Care Excellence recommends a symptom-based osteoarthritis definition. Clinicians might be expected to record patients meeting this definition as having osteoarthritis; however, a record of peripheral joint pain is sometimes used instead. Patients with a recorded formal osteoarthritis diagnosis have been shown to be more likely to have care consistent with guidelines. Due to the potential impact of recorded diagnosis on clinical management and database research, our aim was to estimate the proportion of people aged ≥45 years recorded with joint pain who have osteoarthritis.
Pseudo-anonymised MOSAICS trial consultation data were reviewed for patients recorded in a 12-month period with a joint pain Read code for the hand, hip, knee or foot but not osteoarthritis (N05 codes). Two academic GPs with an osteoarthritis interest assessed consultation narrative to determine if patients could be reclassified to ‘osteoarthritis’, ‘not osteoarthritis’, or ‘joint pain unclassified’. Disagreement triggered a third independent narrative review. A majority decision was binding; where none was reached, the consultation was allocated to ‘joint pain unclassified’. Where multiple joint pain consultations were recorded, patients were reallocated hierarchically to ‘osteoarthritis’, ‘not osteoarthritis’, or ‘joint pain unclassified’. By combining the reclassified consultations with N05 Read-coded osteoarthritis consultations, revised estimates of age-standardised consultation prevalence for clinical osteoarthritis were produced. Clinical factors significantly associated with ‘osteoarthritis’ were determined through a multilevel multinomial logistic regression analysis.
A decision was reached on 3091 (96%) of 3209 joint pain coded consultations (895 after arbitration) in 2058 patients. 515 (25%) patients were classed as ‘osteoarthritis’, 718 (35%) as ‘not osteoarthritis’, and 825 (40%) as ‘joint pain unclassified’ (inter-rater agreement for multiple raters κ = 0.49 [‘moderate’]). The age-standardised annual consultation rate for formally-coded osteoarthritis (N05) was 323 per 10,000 people aged ≥45 years (95% CI: 304,342). Addition of the reclassified joint pain consultations increased this to between 474 (451,498) and 720 (691,749) per 10,000, based on the extreme assumption that either no or all patients with ‘joint pain unclassified’ had osteoarthritis. The ‘best estimate’, in which patients with ‘joint pain unclassified’ were allocated 25:35 to ‘osteoarthritis’ and ‘not osteoarthritis’, was 576 (550,602) per 10,000. The factor most associated with higher likelihood of being reclassified to ‘osteoarthritis’ was overweight/obesity. Consultations for joint pain at the ankle/foot, wrist/hand and unspecified sites were less likely to be reclassified to ‘osteoarthritis’.
Use of Read codes representing a formal osteoarthritis diagnosis underestimates the prevalence of people aged ≥45 who have osteoarthritis in primary care; true consultation rates are likely to be 1.5—2.2 times greater, with implications for clinical practice and research. Factors associated with under-diagnosis may be usable in algorithms to improve case definitions for electronic health record database research.