A90: Pediatric Henoch‐NSchöAdonlein Purpura: Can Length of Hospital Stay Be Predicted?

Background/Purpose: Henoch Schonlein purpura (HSP) is a common, usually benign, arthritis in childhood. However, at times, it follows a protracted course involving prolonged length of hospital stay (LOS). Objective: To determine clinical and laboratory characteristics that allow prediction of prolonged LOS in children with HSP. Methods: We retrospectively collected all consecutive cases of HSP that were admitted to the Dana‐Duek Children's hospital between Jan. 1, 2000 and Sep. 30, 2013. All charts were thoroughly reviewed for demographic characteristics (e.g. age, gender), LOS (in days by date), presenting and later symptoms (abdominal pain, vomiting, GI bleeding, palpable purpura, arthralgia or periarthritis, fever) and laboratory data on admission and discharge (CBC, CRP, ESR, serum creatinine, urinalysis). In univariate analysis we studied the variables that were statistically related to LOS. Variables that significantly predicted LOS in univariate analysis or that were clinically relevant (i.e. Proteinuria (defined as positive protein in urinalysis in 2 consecutive tests), systemic fever (defined as rectal temperature ≥38.0°C), gastrointestinal (GI) bleeding, arthralgia, CRP concentration, age (years) and abdominal pain as sole presenting symptom on admission) were analyzed in stepwise backward multiple regression analysis. Variables that remained significant in the final analysis were correlated with LOS, were each given specific weight (according to their contribution to the final R square) and were used to assemble an HSP severity score. Discriminate analysis was used to determine the score best discriminating patients with prolonged LOS (4 or more days) from the others. Results: Eighty nine children (aged 9 months–17 years) were retrieved for analysis. In univariate analysis, variables that were associated with prolonged LOS were: abdominal pain as initial sole presentation (P < 0.001), CRP (P = 0.001) and systemic fever (P = 0.018). Four variables remained significant in the final stepwise multiple regression analysis, and were used as parameters in assembling the HSP severity score. The HSP severity score consisted of sole abdominal pain at presentation (0 or 3 points), Fever (0 or 2 points), CRP ≥45 (0 or 1 point) and Age ≥6 (0 or 0.5 points). A score of ≥3 predicted a prolonged LOS (4 or more days) with a sensitivity of 76%, a specificity of 79% and a positive predictive value of 93%. Conclusion: Four selected variables available on the day of admission allow predicting the severity (in terms of LOS) of HSP. This score will be validated in an ongoing prospective study.