Predictive variables for hemodialysis and death in Japanese spotted fever, and the association between distance from rivers and incidence

Abstract

The number of patients with Japanese spotted fever (JSF) and its case fatality rate have been increasing in Japan and other East Asian countries. Better clinical and laboratory biomarkers are needed to avoid misdiagnosing JSF and to predict severe cases. In addition to determining these predictors, we aimed to examine the association between the incidence of JSF and the distance from rivers, in Hiroshima Prefecture, one of the most JSF prevalent areas in Japan. Patients diagnosed with JSF from 2009 to 2017 in two hospitals in Onomichi City in Hiroshima Prefecture were studied, and their clinical characteristics and laboratory data were collected retrospectively from medical charts. A random forest was used to identify predictors of severe JSF leading to hemodialysis or death. A multivariable negative binomial regression model was utilized to analyze the association between the cumulative incidence in each postal code area and the distance from the residential postal code area to the closest river. Out of 82 patients with JSF (mean age at diagnosis, 74.1 ± 10.6 years; 34 (41.5 %) men), 6 cases were regarded as severe (among them 5 hemodialysis patients and 3 deaths). Twenty-eight (34.1 %) patients were misdiagnosed at least once at the initial hospital visit. Laboratory examination showed 34.5 % had atypical lymphocytes, 73.8 % had no eosinophils, 75.6 % had an elevated aspartate aminotransferase (AST) level, and 69.5 % had hyponatremia. Among cases without urine leucocytes, 63.3 % had proteinuria and 63.3 % had hematuria. Low serum total protein was the strongest predictor of severe JSF, followed by high blood urea nitrogen (BUN) and low albumin. Geospatial analysis showed a significant negative association between the cumulative incidence of JSF cases and the distance from rivers in an adjusted model: the cumulative incidence decreased by 0.51 times (95 % CI: 0.30 to 0.86) for every kilometer of distance from the residential postal code area to the closest river. Some laboratory data may be useful in averting misdiagnosis of JSF and in predicting severe cases. Additional studies should be done in order to clarify the mechanism and association of the incidence of JSF with the distance from the nearest river.