Background Strategies for case detection in epidemiological surveys have explored the performance of questionnaires administered by mail, self-report or in face-to-face interview by health professionals, further confirmed by a rheumatologist. The cost of such strategies and the moderate validity of these methods has led to the preliminary development of a telephone questionnaire administered by lay people that would be useful for inflammatory rheumatic diseases.
Objectives To assess the accuracy of a questionnaire for epidemiological detection of rheumatoid arthritis (RA) and spondylarthropathies (SPA) suitable for telephone survey conducted by lay people.
Methods A questionnaire was created with a list of items derived from signs, symptoms, self-report diagnosis and currently accepted epidemiological criteria for classification of RA (ACR 1987) and SPA (ESSG 1991) by a group of rheumatologists and epidemiologists. RA, SPA and control patients were recruited in rheumatology outpatient clinics in 9 university hospitals in France. All patients were examined and had diagnosis confirmed by a certified rheumatologist. In each centre, lay people from self-help group or social league were trained by a unique team of professional poll staff to conduct a telephone interview in a standard manner using this questionnaire. Each group of RA and SPA was compared to the control group for sensitivity (Se) and specificity (Sp) of each item and overall concordance with clinical diagnosis.
Results A total of 230 RA, 175 SPA and 195 controls with mean age of 55.6, 46.4 and 55.0 years, and female/male ratio of 2.6, 0.7 and 2.2 respectively were recruited. They were interviewed on the phone by 3 to 10 people in each centre. In RA-control comparison, a set of 5 items, most belonging to the ACR criteria, were found the most informative in logistic regression analysis, with self-report diagnosis showing the highest performance (Se = 0.98, Sp = 0.86). In SPA-control comparison, a few items belonging to the set of ESSG criteria were the most informative, with self-report diagnosis showing the highest performance (Se = 0.85, Sp = 0.97). The overall concordance of each set with clinical diagnosis for RA and SPA was 98% and 93% respectively. When self-report diagnosis was not considered, queries on peripheral joint and spinal pain contributed to detection performance most significantly, with 90% and 78% concordance with clinical diagnosis of RA and SPA respectively. When modelling the selected items in a hypothetical population with 0.5% or 1% prevalence rate, the positive predictive value ranged from 4 to 26%, and the negative predictive value was 99%.
Conclusion A questionnaire using clinical diagnosis self-report and presentation of epidemiological criteria in common language by lay people has been developed. Given its performances in this sample at high risk for error of specificity, it is available for case detection in epidemiological survey in the general population, provided diagnosis is further confirmed by a certified rheumatologist as part of the case ascertainment strategy.