Leptospirosis in children is often under-diagnosed, especially in those who do not present with the severe icteric form of disease. In a semi-rural region of north central Thailand, we found that leptospirosis accounted for at least 6% of all acute undifferentiated febrile illnesses, and 19% of the non-dengue illnesses, among children presenting to hospital during the rainy season. As expected, the prevalence of leptospirosis was much lower in the metropolitan environment of Bangkok. We used the PanBio IgM ELISA to screen for leptospirosis in well-timed acute and convalescent blood samples, and performed MAT on any equivocal or positive samples. It is unlikely that a large number of leptospirosis cases were missed, since the sensitivity of the PanBio IgM ELISA as used in this study has been reported as high (76–90%) 
. Some characteristics of the children we identified with leptospirosis were similar to what has been previously reported. There was a male predominance, cases peaked during times of flooding (i.e. late rainy season), and illness was generally less severe than what is typically reported in adults 
. Our cohort of children with leptospirosis had self-limited and even milder disease than what has been reported from studies of pediatric leptospirosis in Brazil, Reunion Island, and India 
. One reason may be differences in the predominant circulating Leptospira
serovars. Serovars that have been reported to be associated with severe illness in children are Icterhemorrhagiae, Copenhageni, Canicola, and Sejroe 
. In our cohort of Thai children with leptospirosis, we noted predominant seroreactivity to serovars Autumnalis and Andamana (serogroups Autumnalis and Andamana, respectively). However, others have reported different serovar associations with disease severity 
or none at all 
. Another reason for the mild disease seen here could be the slightly younger age distribution of the children in our study. An age-dependent association with the severity and case-fatality rate of leptospirosis has been observed across many regions of the world 
. Host factors that may contribute to this association could include higher organism loads with increasing age, or age-dependent changes in innate and adaptive immune responses to leptospiral infection. Finally, 39% of the children received antibiotics, but not because leptospirosis was suspected. We cannot determine if early antibiotic therapy played a role in ameliorating the severity of some pediatric leptospiral disease.
In our cohort of 18 children with confirmed leptospirosis, none were diagnosed correctly by the time of hospital discharge. The most common specific alternative diagnoses were dengue and scrub typhus. Leptospirosis is often indistinguishable from dengue at the critical early stages of illness, and the two are confused commonly 
. A lack of affordable and accurate diagnostic tests in many settings also contributes to the diagnostic confusion. Predictive models using more readily available clinical and laboratory characteristics would provide useful information to practitioners. We found that presenting symptoms within the first 3 days of illness were not helpful in distinguishing children with leptospirosis from dengue. A lower petechial count on the standardized tourniquet test was associated with leptospirosis compared to dengue in our study and among adult patients in Bangladesh 
. However, the predictive value of this single test was poor, and the association did not remain significant in a multivariate predictive model. Conjunctival abnormalities were seen in 2/18 (11%) of the children with leptospirosis in this study. Conjunctival inflammation or hemorrhage may have been confused with conjunctival suffusion (with or without hemorrhage), a condition classically described in leptospirosis 
. The children with leptospirosis also tended to have a lower degree of hepatomegaly than the age-and sex-matched dengue patients. Similar observations were reported among children with leptospirosis and dengue in Mumbai, India 
. Unfortunately, none of the aforementioned physical signs had sufficient prevalence or discriminatory capability to be useful early in illness.
We found that the most striking difference on presentation between children with leptospirosis and dengue existed in their WBC count and differential. The most significant single laboratory value independently associated with leptospirosis compared to dengue was the absolute neutrophil count (ANC). Neutrophilia is often reported in leptospirosis 
. However, the ANC has not been previously reported as a useful indicator of leptospiral infection in children 
, or distinguishing leptospirosis from dengue in any age group 
. The predominance of mature neutrophils and paucity of band forms on the WBC differential might also be an additional useful clue to the diagnosis of leptospirosis. We found that the combination of 3 laboratory values early in illness-ANC, albumin, and AST-provided the best ability to distinguish between leptospirosis and dengue among children with acute undifferentiated febrile illnesses presenting to the hospital. This predictive model will need to be tested and validated in a prospective fashion in order to determine its potential clinical utility.
Our study was limited to symptomatic children presenting to a hospital for evaluation. Its conclusions cannot be extrapolated to the spectrum of disease that does not present to the hospital and may have different clinical and laboratory manifestations. Some potentially useful clinical variables may also not have emerged in our analysis due to the relatively small number of children with leptospirosis. With increased recognition of children with leptospirosis in Kamphaeng Phet, Thailand 
, future comparative studies will have greater statistical and discriminatory power. In the tropics, leptospirosis can be a significant cause of “dengue-like” febrile illness among children presenting to the hospital during the rainy season. Increased awareness of pediatric leptospirosis, and an enhanced ability to discriminate between leptospirosis and dengue early in illness, will help guide the appropriate use of healthcare resources in often resource-limited settings.