We describe physical symptoms and signs, laboratory data and psychosocial assessments for 84 children (age 1-17) hospitalized for the first time with suspected or definite rheumatic disease. At 16 months (range 7-28) followup change of disease severity was assessed based on hospital records. The patients were categorized as unchanged/worse (n = 36) or improved (n = 48). A linear discriminant model applying a combined set of 6 medical and 2 psychosocial variables characterized correctly 68 of 84 patients (81%). The variables were age at onset, disease severity, months of disease duration, presence of rheumatic disease in the family, erythrocyte sedimentation rate (ESR), thrombocytes, total behavior problem score on the Child Behavior Checklist and maternal distress. By discriminant function analysis of biopsychosocial variables more homogeneous groups may be obtained. Study of such groups may improve prediction of the prognosis.