Using analytical approaches to inform diagnostic strategies for febrile illnesses in the context of universal health coverage

The diagnosis of Acute febrile illness (AFIs) is an immense challenge in resource-limited setting due to their non-specific clinical presentation, and inadequate access to optimal clinical expertise and diagnostic tools. Except for malaria, there exist no reliable rapid diagnostic tests (RDTs) or clinical guidance to manage AFIs. The indiscriminate uses of commercially available suboptimal RDTs in clinical decisions have risked patients to substandard care and inappropriate use of antimicrobial drugs. The aim of this thesis is to understand and develop strategies in the utilization of currently available diagnostic tests to improve integrated fever management with focus on algorithm innovations in resource-limited contexts. Utilizing innovative data analytical approaches and mathematical simulations, and built upon three different case studies, this thesis attempts to understand the better application of available diagnostic tests against AFIs. The first case study compares the application of two-test algorithms of antigen-based rapid testing and reverse transcriptase polymerase chain reaction with their singular testing in hypothetical population samples of COVID-19 cases and evaluates the diagnostic outcomes of each testing approaches making the trade-offs with resource requirements and time to results. The second case study explores the role of clinical signs and symptoms and routine laboratory test results in classification of non-malarial acute febrile illness into bacterial and non-bacterial infections. The study utilizes application of statistical methods, machine learning algorithm and principal component analysis on a secondary dataset of 1915 AFI patients from a multinational fever study. The third case study, using a typhoid RDT in a simulated hypothetical population cohort, explores conditions and potential scenarios where currently available RDTs, although suboptimal, can improve health outcomes of AFI patients. In this thesis, I describe the results of these case studies along with the corresponding design and implementation of mathematical model frameworks and data analytical techniques. The results are further elaborated and interpreted for their generalization to other causes of AFIs and their implications for integrated fever management. In conclusion, this thesis uses a multifaceted approach to better understand the current complexities in the diagnosis of AFIs and provides important insights for better utilization of presently available tools in improved fever management. By emphasizing on the evidence-based approach that incorporates contextual priorities, epidemiology and health system characteristics, this study contributes to advancing universal health coverage by promoting effective AFI diagnosis and management tailored to population needs.