Tamil Nadu Launches Predictive Model for TB Mortality

Tamil Nadu has introduced a pioneering model to predict the probability of death among adults diagnosed with Tuberculosis (TB). This model is integrated into the state’s existing TB SeWA application. The objective is to reduce the time between diagnosis and hospital admission for severely ill TB patients. This initiative aims to lower the mortality rate further.

Background and Recent Implementation

The model was developed by the Indian Council of Medical Research’s National Institute of Epidemiology (ICMR-NIE). It was launched recently and incorporated into Tamil Nadu’s TB SeWA system, used since 2022. TB SeWA supports the Tamil Nadu
Kasanoi Erappila Thittam (TN-KET) initiative, which focuses on differentiated care for TB patients.

Triage Variables and Patient Assessment

Under TN-KET, healthcare workers assess every newly diagnosed adult TB patient using five key variables – body mass index (BMI), pedal oedema, respiratory rate, oxygen saturation, and ability to stand unaided. These indicators help identify patients who are very severely undernourished, in respiratory distress, or physically weak. The data is entered into TB SeWA, which flags patients as severely ill or not.

Predictive Model Features

The new model advances beyond flagging severity. It calculates and displays the predicted probability of death for each patient. This percentage offers an objective risk estimate, helping healthcare workers prioritise hospital admission urgently. Severely ill patients show a death probability ranging from 10% to 50%. Non-severely ill patients have a much lower risk, between 1% and 4%.

Impact on Healthcare Response

Data shows about 10-15% of TB patients in Tamil Nadu are severely ill at diagnosis. The model helps reduce delays in hospital admission, which currently average one day but can extend up to six days for some patients. Immediate admission of high-risk patients is crucial, as two-thirds of TB deaths occur within two months of diagnosis.

Data and Accuracy

The predictive model was built using data from nearly 56,000 TB patients diagnosed between July 2022 and June 2023. The five triage variables alone predict mortality as accurately as the full baseline data from India’s national TB portal, Ni-kshay. Unlike Ni-kshay data which is available only after three weeks, TN-KET variables are collected within a day, enabling timely intervention.

Statewide Usage and Significance

All 2,800 public health facilities in Tamil Nadu, from primary centres to medical colleges, use TB SeWA alongside a paper-based triage tool. Tamil Nadu is the only state in India to systematically record and apply these five variables for patient management. The initiative has led to reduced losses in the TB care cascade and documented declines in death rates in two-thirds of the districts.

Model’s Broader Implications

The success of this predictive model provides a replicable example for other states struggling with high TB mortality. Early identification and rapid admission of severely ill patients remain critical to reducing TB deaths nationwide despite free diagnosis and treatment services.

Questions for UPSC:

1. Critically discuss the role of predictive analytics in improving public health outcomes with reference to Tuberculosis management in India.
2. Examine the challenges and opportunities in implementing digital health tools in rural and urban healthcare settings in India.
3. Analyse the impact of early diagnosis and triage systems on disease mortality rates. How can these be applied to other infectious diseases?
4. Estimate the potential benefits and limitations of integrating artificial intelligence and machine learning in national disease control programmes.

Answer Hints:

1. Critically discuss the role of predictive analytics in improving public health outcomes with reference to Tuberculosis management in India.

1. Predictive analytics enables early identification of high-risk TB patients, allowing timely interventions.
2. Tamil Nadu’s TB SeWA model predicts death probability using five triage variables, improving prioritization for hospital admission.
3. Objective risk estimates reduce subjective bias in clinical decision-making and speed up care for severely ill patients.
4. Data-driven approaches help reduce mortality by shortening diagnosis-to-admission time, crucial since two-thirds of TB deaths occur within two months.
5. Analytics support resource allocation and targeted care, optimizing limited healthcare infrastructure.
6. Challenges include data quality, integration with existing systems, and ensuring frontline worker training and compliance.

2. Examine the challenges and opportunities in implementing digital health tools in rural and urban healthcare settings in India.

1. Opportunities – Improved data collection, real-time monitoring, and standardized triage as seen in Tamil Nadu’s TB SeWA across 2,800 facilities.
2. Digital tools enable faster decision-making and better patient management even in remote areas with mobile connectivity.
3. Challenges include infrastructure gaps, digital literacy among healthcare workers and patients, and inconsistent internet access in rural areas.
4. Integration with existing paper-based systems and workflows can be complex and require capacity-building.
5. Urban settings may have better infrastructure but face high patient loads and privacy concerns.
6. Successful adoption depends on government support, training, and continuous technical assistance.

3. Analyse the impact of early diagnosis and triage systems on disease mortality rates. How can these be applied to other infectious diseases?

1. Early diagnosis coupled with triage identifies severely ill patients quickly, enabling prompt treatment and reducing mortality.
2. TN-KET’s use of five simple clinical variables allows rapid risk stratification within a day, critical for TB outcomes.
3. Reducing delays from diagnosis to admission decreases early fatalities, especially in diseases with acute progression.
4. Similar triage systems can be adapted for diseases like COVID-19, dengue, and malaria to prioritize care and manage hospital resources.
5. Data-driven triage supports differentiated care models, improving efficiency and patient outcomes.
6. Challenges include disease-specific variable selection and ensuring triage tool accuracy and usability across diverse settings.

4. Estimate the potential benefits and limitations of integrating artificial intelligence and machine learning in national disease control programmes.

1. Benefits – AI/ML can analyze large datasets to develop predictive models improving early detection and personalized care.
2. Enables dynamic risk scoring and real-time decision support for frontline health workers, as demonstrated by TB mortality prediction in Tamil Nadu.
3. Facilitates efficient resource allocation and monitoring of disease trends at population level.
4. Limitations include data privacy concerns, algorithm biases, and the need for high-quality, representative data.
5. Infrastructure and digital literacy gaps may hinder equitable access and implementation in rural areas.
6. Continuous validation and local customization are necessary to maintain accuracy and relevance over time.