Regression Analysis of the Main Environmental Risk Factors for Mountainous Subtype of Zoonotic VL in China

Mountain-type zoonotic visceral leishmaniasis (MT-ZVL) has re-emerged in China in recent years, posing a growing public health concern. Canines are the primary reservoir hosts, yet the environmental determinants of canine infection remain insufficiently understood. This study aimed to identify key environmental factors associated with MT-ZVL infection.

A total of 1484 canine blood samples were collected from 89 villages across five endemic provinces in China. After excluding cases with missing data, 1374 cases were ultimately included in the analysis. Candidate predictors included canine hair length, age, sex, weight, NDVI, altitude, and temperature. We developed a model using ridge-penalized logistic regression, tuned hyperparameters, and validated internally through 30 iterations of 10-fold stratified cross-validation. Model discriminatory power was evaluated using the area under the receiver operating characteristic (AUC) curve and its 95% confidence interval. Additionally, the optimal cutoff value corresponding to the maximum Youden's index was determined, and sensitivity, specificity, positive predictive value, negative predictive value, and accuracy were calculated. Variable importance was ranked based on the absolute values of the standardized coefficients.

The final model, built using ridge regression (optimal λ = 0.0215), included seven predictors: canine hair length, age, sex, weight, NDVI, altitude, and temperature. Based on 30 repetitions of 10-fold stratified cross-validation, the model demonstrated moderate discriminatory power, with an AUC of 0.68 (95% CI 0.62-0.74). The optimal threshold, determined by maximizing the Youden index, was 0.024. At this threshold, sensitivity reached 97.8%, specificity was 35.9%, the negative predictive value was as high as 99.8%, the positive predictive value was 4.9%, and the overall accuracy was 37.9%. Among the variables, NDVI had the highest relative importance (OR = 14.51), followed by canine hair length. This model has an extremely low risk of missed diagnoses and is suitable for screening to rule out the disease; however, due to its high false-positive rate, it should not be used alone for definitive diagnosis.

Environmental factors, particularly vegetation coverage, play a critical role in shaping the risk of canine infection in MT-ZVL endemic areas. These findings provide important evidence for improving surveillance strategies and implementing targeted control measures.