Background Delays to rheumatologic care represent addressable barriers to improved outcomes for inflammatory arthritis (IA)1. A self-administered tool has been developed for early IA detection in pre-rheumatology care settings. Psychometric properties and discriminant validity of this tool in rheumatology triage have been previously studied2,3. A scoring algorithm enhanced discriminant validity of unweighted tool in rheumatology triage3.
Objectives To determine if predictive performance of unweighted early IA detection tool differs from primary care physician (PCP)’s diagnosis in primary care. Secondary objective: To contrast relative performance of tool and PCP diagnosis between primary care and rheumatology triage settings.
Methods Prospective study of 139 patients with 6-52 wks musculoskeletal complaints in Canadian primary care setting was conducted. Participants completed tool and received rheumatology consultation regardless of PCP’s recommendation. Predictive performance of participants’ positive responses to 6 of 11 tool items and IA diagnosis by PCP were compared using rheumatologist’s blinded diagnosis as reference standard. Relative performance of the tool and PCP diagnosis in primary care was compared to rheumatology triage setting where 143 referred patients were studied.
Results 19 patients were diagnosed by rheumatologist to have IA. Compared to non-IA, cases had more swollen joints (p<0.0001), greater duration of morning stiffness (p<0.0001) and higher erythrocyte sedimentation rate (p=0.002). Similarly, cases were more frequently rheumatoid factor positive (p<0.0001), treated with NSAIDs by PCP (p<0.0001), had constitutional symptoms of IA diagnosis (p<0.01) and had positive metacarpophalangeal and metatarsophalangeal squeeze tests (p<0.0001 respectively). Predictive performance of tool included a significant odds ratio (OR) (95% CI) of 5.07 (1.59-16.20), sensitivity (Se) of 0.79 (0.61-0.97) and specificity (Sp) of 0.58 (0.49-0.66). In contrast, predictive performance of PCP diagnosis included non-significant OR of 2.94 (0.91-9.49), Se of 0.26 (0.07-0.46) and Sp of 0.89 (0.84-0.95). Among 56 participants missing PCP diagnosis (40.29%), OR for the tool was 6.91 (0.75-63.53). Median (IQR) number of positive rheumatology-derived cases was 7 (6-8), compared to 5 (4-6) for PCP-determined IA cases (p=0.02) and 5 (3-7) for rheumatologist-diagnosed non-IA (p=0.0014). In the triage setting (PCP referrals to rheumatology), the unweighted tool had OR of 7.03 (2.76-17.88) compared to 28.25 (8.93-89.32) for PCP diagnosis. Weighting the tool, OR was 25.28 (9.07-70.41). Among 34 referrals without PCP diagnosis (23.78%), OR was 49.5 (4.88-501.75) for unweighted tool and 69.67 (6.44-754.16) for weighted tool.
Conclusions The unweighted, self-administered early IA detection tool has discriminant validity both in primary care and rheumatology triage settings. In primary care, the tool generally outperforms PCP diagnosis. In the rheumatology triage setting, the tool may be especially advantageous among referrals without a PCP diagnosis. In primary care, significant predictive performance of the tool may be further enhanced with a scoring algorithm.
Mottonen T, et al. Arthritis Rheum 2002;46(4):894-8.
Bell MJ, et al. BMC Musculoskelet Disord 2010;11:50.
Tavares R, et al. Arthritis Care Res 2012 (submitted).
Disclosure of Interest None Declared