Cost-Effectiveness of AI-Based Chatbot for Tuberculosis Screening in Primary Health Care Settings in Indonesia

Abstract

Background: Tuberculosis (TB) still poses serious public health problems in Indonesia, with many gaps in early detection and case finding by primary care. Digital health innovations, such as AI (artificial intelligence)–based chatbots, provide a scalable solution to improve symptom screening and triage. This study aimed to assess the cost-effectiveness of an AI-based chatbot for TB screening versus standard symptom-based screening in primary health care settings in Indonesia. Methods: A decision analytic model was created to evaluate AI-based chatbot screening as compared with conventional screening from the health system perspective over a one-year time horizon. Model inputs were obtained from secondary data, including the Indonesian TB system (SITB), published literature, and data from pilot implementation in 50 primary health centers. Effectiveness was defined as the number of additional TB cases detected and DALYs averted. Costs comprised screening, diagnostic confirmation, and treatment initiation. ICERs were calculated. Sensitivity analyses were performed both deterministically and probabilistically. Results: Among a cohort of 120,000 individuals screened, the AI-based chatbot strategy detected 4,320 presumptive TB cases and confirmed 1,080 TB cases, versus the standard screening, which referred only 3,150 presumptive cases that yielded only 840 confirmed cases. This added up to a further 240 TB cases diagnosed (+28.6%). The intervention prevented an estimated 385 DALYs, relative to 290 DALYs under standard screening practices, resulting in an incremental gain of 95 DALYs averted. Total costs for the chatbot strategy were USD 412,560 versus USD 398,200 for standard screening. This yielded an ICER of USD 151 per DALY averted, below Indonesia’s willingness-to-pay threshold (USD 1,200/DALY). Conclusion: In Indonesia’s primary health care system, AI-based chatbot screening is a highly cost-effective intervention to improve TB case detection. Scaling up this approach can increase efficiency, speed case finding, and strengthen progress toward national TB elimination targets.