Antimicrobial Resistance in Tuberculosis: Leveraging AI to Combat Adverse Outcomes

Authored by: Aparna Chaudhary

Antimicrobial resistance (AMR) is one of the most critical global health challenges, threatening decades of progress in treating infectious diseases. Among the diseases impacted by AMR, tuberculosis (TB) stands out due to its global prevalence and severe consequences, particularly in high-burden countries like India. Despite significant progress in TB control, the rise of multidrug-resistant TB (MDR-TB) and extensively drug-resistant TB (XDR-TB) has created serious obstacles to  treatment and disease control.

Non-adherence to TB treatment is a crucial factor contributing to the development of drug resistance. Ensuring that people with TB (PwTB) complete their treatment regimens is essential for controlling the disease and preventing AMR. Addressing the factors leading to treatment non-adherence, such as loss to follow-up (LFU), requires innovative solutions. 

Understanding Antimicrobial Resistance in Tuberculosis

TB is caused by Mycobacterium tuberculosis, a bacterium that primarily affects the lungs but can affect other organs. TB treatment involves a combination of antibiotics administered over 6-9 months for drug-sensitive TB. In cases of drug-resistant TB, the treatment duration can extend to 18-20 months. Longevity treatment is essential to eliminate bacteria that can remain dormant in the body. However, the lengthy duration makes it challenging for many patients to adhere to the regimen, increasing the risk of incomplete treatment and drug resistance.

When patients do not complete their treatment, surviving bacteria may become resistant to one or more antibiotics given under the treatment regimen. MDR-TB arises when bacteria become resistant to at least isoniazid and rifampicin, the two most potent anti-TB drugs. XDR-TB occurs when bacteria become resistant to a broader array of drugs, making treatment options more limited and less effective. Additionally, these individuals continue transmitting the infection, and the development of drug resistance in the community through droplet and aerosol dispersal of drug-resistant TB from people with active infection when they cough, speak, or sneeze without covering their mouth and nose.

TB patients who remain non-adherent to their treatment, i.e., stop or interrupt the treatment for two or more consecutive months, are characterized as "Loss to follow-up".

In India, non-adherence to TB treatment is driven by factors like drug side effects, socioeconomic challenges, and gaps in healthcare infrastructure. According to the India TB Report 2023, 2.6 % (54,127) of new and relapse TB cases were lost to follow-up (LFU). Moreover, 10 -12% of these LFUs report back with drug resistance(Rahayu SR et al, 2023). Thus, poor adherence becomes one of the critical drivers of drug-resistant TB, making it a public health priority to address these challenges.

The Role of Predictive AI in Reducing AMR in TB

Advancements in artificial intelligence offer powerful tools to drive innovation in TB care, enabling proactive approaches to improve treatment adherence and combat drug resistance. Wadhwani AI’s predictive AI solution, developed in collaboration with the Central TB Division, the Ministry of Health and Family Welfare, and USAID, is a promising breakthrough designed to reduce adverse outcomes like LFU and ensure treatment adherence. This can play a pivotal role in preventing the development of AMR among TB patients in India.

The AI model generates a risk score for each patient based on several parameters recorded at treatment initiation in India's TB surveillance portal, Ni-kshay. Patients are then classified as "high-risk" or "low-risk" based on their likelihood of experiencing adverse outcomes. The solution currently targets the top 35% of patients with high-risk scores, correctly identifying 69% of patients who experience adverse outcomes.

By identifying high-risk patients early, the AI solution enables healthcare workers to intervene proactively with tailored support, such as counseling, reminders, or assistance in overcoming logistical barriers. This  approach helps improve treatment adherence and ensures that patients complete their treatment  reducing the chances of AMR development. The solution can  also help optimize resource allocation for patients who need more support. Eventually, this will aid in strengthening patient-centric care by addressing the unique needs of individuals  at higher risk. 

Conclusion

The rise of antimicrobial resistance to TB poses a serious public health threat, particularly in high-burden countries like India where it threatens to reverse years of progress in TB control. Ensuring treatment adherence is one of the most effective ways to prevent the development of drug-resistant TB. 

The predictive AI solution developed by Wadhwani AI  addresses this challenge by identifying patients at risk of LFU or non-adherence and enabling timely interventions. This prevents the development and spread of drug-resistant TB strains, thus safeguarding public health.

By reducing loss to follow-up and ensuring treatment adherence, AI can play a pivotal role in combating the development of AMR in TB, which will ultimately benefit patients and the broader community.