Improving Tuberculosis Treatment Adherence Support: The Case for Targeted Behavioral Interventions
研究结核病治疗依从性支持平台，发现个人赞助人外联可将次日依从性验证几率提高35%，并利用机器学习动态风险预测实现针对性干预，提升资源效率。
Problem definition: Lack of patient adherence to treatment protocols is a main barrier to reducing the global disease burden of tuberculosis (TB). We study the operational design of a treatment adherence support (TAS) platform that requires patients to verify their treatment adherence on a daily basis. Academic/practical relevance: Experimental results on the effectiveness of TAS programs have been mixed; and rigorous research is needed on how to structure these motivational programs, particularly in resource-limited settings. Our analysis establishes that patient engagement can be increased by personal sponsor outreach and that patient behavior data can be used to identify at-risk patients for targeted outreach. Methodology: We partner with a TB TAS provider and use data from a completed randomized controlled trial. We use administrative variation in the timing of peer sponsor outreach to evaluate the impact of personal messages on subsequent patient verification behavior. We then develop a rolling-horizon machine learning (ML) framework to generate dynamic risk predictions for patients enrolled on the platform. Results: We find that, on average, sponsor outreach to patients increases the odds ratio of next-day treatment adherence verification by 35%. Furthermore, patients’ prior verification behavior can be used to accurately predict short-term (treatment adherence verification) and long-term (successful treatment completion) outcomes. These results allow the provider to target and implement behavioral interventions to at-risk patients. Managerial implications: Our results indicate that, compared with a benchmark policy, the TAS platform could reach the same number of at-risk patients with 6%–40% less capacity, or reach 2%–20% more at-risk patients with the same capacity, by using various ML-based prioritization policies that leverage patient engagement data. Personal sponsor outreach to all patients is likely to be very costly, so targeted TAS may substantially improve the cost-effectiveness of TAS programs. History: This paper has been accepted for the Manufacturing & Service Operations Management Special Section on Responsible Research in Operations Management. Funding: The authors are grateful for financial research support from the MIT Sloan Health Systems Initiative. Supplemental Material: The e-companion is available at https://doi.org/10.1287/msom.2021.1046 .