Tuesday, October 20, 2015: 1:30 PM
Grand Ballroom B (Hyatt Regency St. Louis at the Arch)

Sze-chuan Suen, MS1, Margaret L. Brandeau, PhD1 and Jeremy D. Goldhaber-Fiebert, PhD2, (1)Department of Management Science and Engineering, Stanford University, Stanford, CA, (2)Stanford Health Policy, Centers for Health Policy and Primary Care and Outcomes Research, Department of Medicine, Stanford University, Stanford, CA

Purpose: Effectively treating tuberculosis (TB) requires administering drugs to which the infection is not resistant. Though costly, drug sensitivity testing (DST) of patients receiving first-line treatment can triage those with multi-drug-resistant (MDR) TB to appropriate but expensive treatment alternatives. In India, patients receive DST if they have not responded to four months of treatment, as measured by the imperfect but inexpensive sputum smear (SS) test. We seek to determine the optimal time to administer DST and the patterns of SS results that should prompt DST. If DST is administered too soon, many patients without MDR TB will be unnecessarily tested. If administered too late, patients with MDR TB may continue to transmit disease and experience declining health.

Method: We use a partially observed Markov decision process (POMDP) to determine the optimal timing and frequency of SS test information collection and DST testing in India. We calculate parameters such as patient response to treatment, patient dynamics while on treatment (the possibility of default or death), and discounted lifetime costs and health benefits using clinical studies and our previously published TB microsimulation model. We solve the POMDP using value iteration on a constrained feasible belief set.

Result: Current policy appears suboptimal given India's relatively high national estimates of MDR TB prevalence and transmission. For these estimates, DST should be administered to all patients at the outset of treatment. We project that this could save $7800 per TB patient in discounted net monetary benefits after accounting for averted downstream transmissions. However, in settings where the risk of transmission or MDR prevalence is much lower than the national average, a patient's SS result sequence can change the optimal DST timing, and individually tailored testing policies would be optimal.  See Figure 1: national averages for India lie in the white region but districts with low MDR prevalence or transmission have different optimal policies that vary by patient SS outcomes. 

Conclusion:  India should revise the drug sensitivity testing protocol in their first-line national TB treatment program to provide DST during the first month of treatment in areas of average or high MDR TB prevalence and transmission, and may wish to consider individually tailored DST regimens in low transmission, low MDR prevalence areas to reduce financial costs.