Clinical presentation of leptospirosis can be varied and non specific, hence the recognition of red flags of complications is of paramount importance. The aim of our study was to determine such red flags and to formulate a risk predictor model. Of the many clinical and laboratory predictors of mortality which have been described, we assessed nine clinical and eight laboratory variables in our study. Furthermore many studies have been conducted to determine the predictors of mortality, but studies conducted to determine the predictors of complications were not found in Sri Lankan literature .
Out of the 17 variables assessed, only oliguria, jaundice and arrhythmias were predictors of leptospirosis complications. Although previous studies have shown that 70% of leptospirosis patients were having arrhythmias, in our study it was only 12.9%. In another study acute renal failure was found in 40% of patients which is a higher value compared to ours. In the same study it was shown that oliguria is a sensitive sign of severe leptospirosis. This finding is somewhat similar to the finding of ours. Why the clinical severity of leptospirosis varies so much is so far unexplained but difference in serovars could be a good reason. Although univariate analysis showed a significant difference for most of the laboratory variables, none of them were significant as predictors in multivariate analysis. Similarly, oliguria and jaundice showed a significant difference as predictors in multivariate analysis despite having no significant difference in univariate analysis. This finding could be due to the confounding effects of the variables. All three variables included in the predictor model were strong predictors and are included in surveillance case definition of leptospirosis in Sri Lanka as well as other countries.
Although jaundice was found to be a predictor of complications in our study, a study done in Turkey  has shown that majority of their patients had anicteric leptospirosis which could also lead to severe disease. Many previous studies have shown leucocytosis, thrombocytopenia and blood urea as early predictors . Interestingly, none of the laboratory measures were shown to have any value as early predictors of renal or cardiac complications in our study. This is a valuable finding as most patients present to rural or district hospitals which lack laboratory facilities, therefore predicting their complications and need for transfer could be decided using the three clinical criteria alone.
Although many previous studies have shown pulmonary involvement with haemoptysis as a major complication , it was not found in our study. Possible reason for this could be a different pathogenic serovar circulating in the area, but this could not be confirmed as we didn't analyze for serotypes.
Thrombocytopenia was a common finding in our patients. Despite this bleeding was uncommon and unlike in other studies it did not predict the development of myocarditis or acute renal failure. Similarly leucocytosis was also a common finding and it is a valuable investigation in differentiating leptospirosis from dengue fever which is deadlier than leptospirosis. It was also not proven to be effective as a predictor of complications.
As seen with other similar studies  acute renal failure was a major complication in our patients. Oliguria has shown to predict acute renal failure but not BU, K, and creatinine. It may be that serum biochemical markers lag behind the clinical presentation of acute renal failure in leptospirosis but the exact mechanism is unknown.
As oliguria, jaundice and arrhythmia point towards involvement of three distinct organ systems, it is possible that all three clinical features herald multi organ involvement in severe disease with different organs involved in different severity.
Many outcome prediction scores have been developed for the critically ill patients e.g. APACHE 11, but none of them have been validated for leptospirosis. According to our study, presence of the three clinical variables oliguria, jaundice and arrhythmias predict acute renal failure and myocarditis hence can be utilized to decide which patient should receive intensive/high dependency care.
Limitations of the study
Present study was carried out within the limited resources for diagnosis of leptospirosis as shown previously. Due to this lack of point of care diagnostic facilities, there were several limitations to our study. Although none of the laboratory confirmed cases died, three patients suspected of having leptospirosis died before serology could be performed, and were not included in the study, which could have introduced a selection bias in the sample. The cases included were based on a single sample MAT titre, which is not the standard test for confirmation. Further, it has been shown that development of antibody might take more than seven days in some patients. These could have missed out some of the patients with leptospirosis. Previous studies have shown that complications of leptospirosis depend on infecting serovars, which in this study we were unable to study due to resource limitations. Although all patients were studied as a single sample, complications and predictors could be varied for different serovars. Duration from the onset of the illness to the admission was varied which could have had some confounding effect on the investigations done on admission. This could be a reason why laboratory measures were poor in predicting the complications. Due to the lack of facilities we couldn't confirm myocarditis histologically and it was done with clinical, electrocardiographical and echocardiographical evidence. But histological confirmation would not change the management of a leptospirosis patient with myocarditis.