Identification of predictive parameters of juvenile idiopathic arthritis among a cohort of patients with muskoloskeletal pain

Methods Clinical records of children referred to our Unit for musculoskeletal complaints between June 2011 and December 2013 were collected during the first evaluation. A database was built, considering: characteristics of pain (frequency, pattern, precipitating factors), joint swelling, morning stiffness, constitutional symptoms, laboratory test results (ESR, CRP, ANA, RF), family history. After the final diagnosis was made, each patient was categorized according to three subgroups: JIA, postinfective arthritis, non-inflammatory disorders and the three categories were compared for each one of the parameters recorded. Leave-on-out cross validation and logistic regression model were used to identify the parameters associated with the final diagnosis of JIA Results A total of 178 children were evaluated throughout this study. The final diagnosis were: JIA (36), infection-related arthritis (28), non-inflammatory disorders (114). The comparison between the three groups showed a pattern of signs and symptoms specific for each one of the categories, in particular JIA and non-inflammatory disorders: persistent joint swelling, the rest as precipitating factor, the presence of morning stiffness and a persistent pain were statistically associated with the diagnosis of JIA (all p values<0.0001), while absence of joint swelling, activity as a precipitating factor, recurrent pain on evenings/nights, resulted statistically associated with non-inflammatory disorders (all p values<0.001). Focusing on JIA diagnosis, we then analyzed the impact of each one of the identified parameters on the final diagnosis, obtaining the formula y = k + b1x1 + b2x2 + b3x3 + b4x4 (Y= probability of having JIA; k=15,735; x1= joint swelling’s pattern; x2= precipitating factors; x3=morning stiffness; x4=pain frequency). This formula gives the probability that a child with musculoskeletal complaints will receive a final diagnosis of JIA (sensitivity 90.9%, specificity 95.3%)