SISA

"SISA" (Severity Index for Suspected Arbovirus) is a prediction model developed to aid in the clinical management of suspected arboviral infections, specifically in resource-limited settings where decisions regarding patient hospitalization are challenging. Developed alongside "SISAL" (Severity Index for Suspected Arbovirus with Laboratory), which incorporates laboratory data, SISA utilizes demographic and symptom data from patients to predict the necessity of hospitalization due to arboviral infections such as dengue, chikungunya, and Zika.

Development and Validation:

- Data Source: The models were developed using data from a surveillance study in Machala, Ecuador, a region with a high burden of arboviral infections. The data encompassed subjects who presented at sentinel medical centers with suspected arboviral infection from November 2013 to September 2017.

- Machine Learning Algorithms: Seven machine learning algorithms were compared to select the most effective one for each model. The selection was based on the ability to predict from a test dataset, specifically the area under the receiver operating characteristic curve (AUC).
- Performance:
- The generalized boosting model was selected as the best prediction algorithm for SISA, achieving an AUC of 0.91.
- The elastic net was identified as the most effective algorithm for SISAL, with an AUC of 0.94.

Features and Functionality:

- SISA: Uses patient intake data, specifically demographic and symptom information, to predict the likelihood of hospitalization due to suspected arboviral infection. This model is beneficial in settings where laboratory data may not be readily available.

- Accuracy: SISA and SISAL demonstrated high accuracy in predicting arbovirus hospitalization within the dataset used for their development. However, a sensitivity analysis indicated that SISA and SISAL are not directly comparable due to differences in their datasets and the inclusion of laboratory data in SISAL.