A cross-sectional survey of pond cleaning participants was conducted from October 6 to 8, 1999, in the subdistrict of Khumuang, Buriram Province, Thailand. The survey was conducted with a convenience sample of 315 persons from the total number who participated in the pond cleaning (n=500). All 315 persons were part of an agricultural community located in the Khumuang subdistrict, approximately 400 km from Bangkok and 100 km from Thai-Cambodian border.

Twelve teams, each consisting of three or four health-care providers (doctor or nurse, health officer, and health volunteer), visited this village from October 6 to 8, 1999. All 315 persons from the village involved in the cleaning of Nong Tad were asked to participate in this study. Two hundred twenty-eight (72%) workers consented to be interviewed and were given a standardized questionnaire. Questions included demographics and risk factor information, such as the working site in the pond, lunch eating site, clothing worn while working, other probable exposure to leptospirosis, and presence of wounds during their participation in the cleaning activity. If a participant answered yes to having skin wounds during the pond cleaning activity, his or her body was examined for lesions or scars. and the interviewer subsequently noted the affected body parts. Participants were asked if they had been ill since participating in the pond cleaning activity. Illness was defined as having symptoms meeting the WHO criteria for leptospirosis (20). Clinical information was collected through review of medical records for those participants who reported being ill and who were seen at the Khumuang Hospital. Serum samples were obtained 1 week after the interview from 104 (45.6%) of the 228 participants; the other 124 persons refused to undergo phlebotomy. Therefore, the interviewer did not know the seropositivity status of the participants at the time of the interview. The timing of the serum collections was within 1 month after exposure to the pond. Infected and noninfected pond cleaning participants were evaluated for risk factors for infection in a nested case-control study. All specimens were tested in the Thailand Ministry of Health Laboratories by using the Lepto-Dipstick Test (Organon, Dublin, Ireland), a commercial test kit with sensitivity and specificity exceeding 80% (21). This study was determined to be a public health response that did not require IRB review.

A case was defined as the presence of immunoglobulin (Ig) M anti-leptospiral antibodies by the Lepto-Dipstick Test in a person from the Khumuang subdistrict who participated in the survey. Persons with a positive IgM antibody response were considered to have incident cases because the serum was collected and tested 1 month after exposure and IgM antibodies last an average of 3–6 months (22–24). Asymptomatic infection was defined as a positive IgM response in a person who did not report having fever, myalgias, headache, or other evidence of leptospirosis.

Descriptive statistics and subsequent multivariable analysis were derived through the use of SAS software release 8.1 (SAS Institute, Inc., Cary, NC). All variables were dichotomized except for age, which was treated as a continuous variable.

Demographic, environmental, and behavioral exposure variables were compared for infected and noninfected persons by univariate analyses. All two-way interactions between variables were tested. A multivariable model was created by inclusion of all exposures significant by univariate analysis as well as age and sex to control for confounding, and a backward elimination procedure was performed to identify exposure variables most strongly associated with seropositivity for leptospirosis infection. Confounding in the absence of interaction was assessed by comparing odds ratios (OR) of the exposure variables in the gold standard model controlling for the covariates, age and sex, with the odds ratios of the exposure variables in the reduced models without age and sex, respectively. If a difference of >10% between the OR was detected, confounding was present and the covariate was retained in the model (25). The variance-covariance matrix allowed for the calculation of 95% confidence intervals (CI) for the OR involving the estimated coefficient of any significant interaction term. A SAS macro was used to calculate the conditional indices and variance decomposition proportions, allowing for the assessment of multicolinearity for two or more variables.

Of the 43 persons with IgM anti-Leptospira antibodies, only 17 (39.5%) reported having illness that met the WHO criteria for leptospirosis; the remaining 26 (60.5%) had asymptomatic infection. Clinical information was available for 13 (76.5%) of the 17 infected persons whose illness met the WHO criteria for leptospirosis. Four people did not seek treatment at Khumuang Hosptial for their illness; their clinical information was not available. All 13 persons who sought treatment at Khumuang Hospital had a fever, 10 above 39°C. Other predominant clinical presentations included chills 84.6% (11/13), headache 76.9% (10/13), myalgia 84.6% (11/13), and calf pain 76.9% (10/13).

Univariate associations between exposures and leptospiral infection among the 104 persons sampled are presented in Table 1. The median age of infected persons (35, range 15–65) was not significantly different from the median age of noninfected persons (40, range 15–63). Although the infection rate in women (32.7%) was lower than that of men (62.8%), the difference was not statistically significant (p=0.091).

In univariate analysis, the pond sites where people worked and places where lunch was eaten were not associated with infection. Infection was not significantly associated with the presence of hand wounds (p=0.092), leg wounds (p=0.069), or six or more leg wounds (p=0.064). Yet having more than two wounds anywhere on the body was significantly associated with infection (p=0.004). Additionally, wearing trousers or long skirts was significantly protective against infection (p=0.038). Trousers or long skirts were worn by 48 (98%) of 49 women and 39 (70.9%) of 55 men (p=0.0002), and footwear was worn by 30 (61.2%) of 49 women and 12 (21.8%) of 55 men (p<0.0001).

All variables associated with or protective against leptospirosis infection with a p value <0.1 by univariate analysis were included in the multivariable logistic regression model. No interaction was detected between any of the exposures. Age was retained in the model throughout the backwards elimination procedure to control for confounding. Sex did not confound any variables that remained in the model after backwards elimination and was removed from the final model. Although any type of footwear was protective in univariate, it was not independent in multivariable analysis. Multivariable analysis by using a backwards elimination procedure (p<0.05) while controlling for age indicated that having a total of more than two wounds anywhere on the body while working in the pond remained independently associated with infection (OR_adjusted=3.97, 95% CI 1.56 to 10.2), while wearing trousers or long skirts was protective (OR_adjusted=0.23, 95% CI 0.067 to 0.701) (Table 2). Multicolinearity was not detected among any of the variables in the multivariate model.