The prevalence of tropical fevers, caused by various pathogens transmitted through vectors, contributes significantly to morbidity and mortality in tropical regions. Challenges exist in diagnosis due to overlapping clinical symptoms, false positive results of disease-specific tests, reliability of the tests, mixed infection. Additionally, these available tests are time-consuming and expensive. Consequently, there is growing interest among researchers to develop certain diagnostics which are simple, quick, easy to use and cheaper. In this regard, artificial intelligence emerges as a pivotal domain, facilitating advancements in disease prediction, mutation detection, pre-emption of next-generation viral diseases, and new drug development etc. Thus, our objective is to devise and assess a technology-integrated tool tailored for the diagnosis and differentiation of tropical fevers within tertiary care hospital settings.

In the present study, phase 1 involves the identification of tropical fevers and associated clinical parameters through a retrospective audit and qualitative interviews with physicians. The case definition includes criteria for acute febrile illness with overlapping symptoms such as fever, rashes, headache, cough, nausea, vomiting, jaundice, abdominal pain, diarrhoea, thrombocytopenia, encephalopathy, respiratory distress, and renal failure etc. Inclusion and exclusion criteria are established to select cases from medical records, excluding other infections such as pneumonia, sepsis, urinary tract infection have similar symptoms with confirmed diagnosis and immunocompromised patients. Consequently, a qualitative interview with physicians will be conducted to understand diagnostic challenges and the importance of clinical and demographical variables. Based on both of these stages of phase 1, the disease and the parameters required for the development of the diagnostic tool will be finalized.

This will be followed by phase 2 which mainly focuses on the construction of the model through retrospective study. Inclusion criteria specify patients aged 18-60 with confirmed diagnoses of tropical fevers, while exclusion criteria exclude children, mixed infections, and missing data. Data from 2019 to 2023 will be collected retrospectively and used in the development of the model. Python language will be used for model construction, employing machine learning techniques like SVM and logistic regression. Data preprocessing, feature selection, and optimization techniques will be applied, and performance metrics will evaluate model efficacy.

Once the model is developed, it will be implemented in the hospital setting in two stages. First part will involve feasibility study for assessing model performance in two hospital settings and calculating sensitivity and specificity to determine sample size for final evaluation. The second stage involves the single-centric observational study for the implementation of the model. Total of 400 patients will be enrolled after Inform consent and baseline demographical, and biochemical parameters will be collected from patient data sheet .The collected data will be used in predicting diagnoses through the developed model and predictions will be compared with specific confirmatory laboratory test results. Model performance will be assessed using various metrics such as Sensitivity, Specificity, Accuracy, Precision, F-measure, Kappa value, Hamming loss, and Area under receiver operating characteristic (AUROC), and overall model performance will be notified to the physicians.