When and what to test for: A cost-effectiveness analysis of febrile illness test-and-treat strategies in the era of responsible antibiotic use

Background: Febrile illness caused by viral and bacterial diseases (e.g., dengue and leptospirosis) often have similar symptoms and are difficult to differentiate without diagnostic tests. If not treated appropriately, patients may experience serious complications. The question of what diagnostic tests to make available to providers in order to inform antibiotic therapy remains an open problem for health services facing limited resources. Methods and findings: We formulated the problem of minimizing the weighted average of antibiotic underuse and overuse to inform the optimal diagnostic test and antibiotic treatment options for given occurrence probabilities of several bacterial and viral infections. We modeled the weight of antibiotic overuse as a monetary penalty per unnecessarily administered course, which we varied in both the base case and sensitivity analysis. Detailed Markov cohort models of febrile illness progression were used to estimate the weight of antibiotic underuse. The model accounted for multiple infections simultaneously and incorporated test, treatment, and other direct and indirect costs, as well as the effect of delays in seeking care and test turnaround times. We used the Markov models to numerically estimate disability-adjusted life years (DALYs), pre-penalty costs, and likelihood of antibiotics overuse per patient for fifteen different strategies in two example settings in Thailand, one with a higher probability of bacterial infections (Northern Thailand, Scenario A) and one with a higher probability of viral infections (Bangkok, Scenario B). We found that empirical antibiotic treatment to all patients always incurs the lowest pre-penalty cost (Scenario A: $47.5/patient, $100.6/patient, $149.5/patient for patients seeking care on day one, day four, and day ten respectively; Scenario B: $94.1/patient, $108.7/patient, $122.1/patient on day one, day four, and day ten respectively), and the lowest DALYs, (Scenario A: 0.2 DALYs/patient, 0.9 DALYs/patient, 1.7 DALYs/patient on day one, day four, and day ten, respectively; Scenario B: 0.5 DALYs/patient, 0.7 DALYs/patient, 0.9 DALYs/patient on day one, day four, and day ten, respectively). However, such strategy resulted in the highest proportion of antibiotic overuse per patient (Scenario A: 38.1%, 19.3%, 7.5% on day one, day four, and day ten, respectively; Scenario B: 82.9%, 42.1%, 16.3% on day one, day four, and day ten, respectively). Consequently, empirical antibiotic treatment became suboptimal with antibiotic overuse penalties above $12,800/course, $18,400/course, $23,900/course for patients presenting on day one, day four, and day ten in Scenario A and above $1,100/course, $1,500/course, $1,600/course for patients presenting on day one, day four, and day ten in Scenario B. Conclusions: Empirical antibiotic treatment to all patients provided the best outcomes if antibiotic overuse was not the primary concern or if presenting with viral disease (such as dengue) was unlikely. Empirical antibiotic treatment to severe patients only was in most cases not beneficial. Otherwise, strategies involving diagnostic tests became optimal. In particular, our results indicated that single test strategies (bacterial RDT or viral PCR) were optimal in regions with a greater probability of presenting with viral infection. PCR-led strategies (e.g., parallel bacterial PCR, or multiplex PCR) are robust under parameter uncertainty (e.g., with uncertain disease occurrence probabilities).