Tamil Nadu becomes first Indian state to implement TB death prediction model

Then, health staff feeds these variables into TB SeWA, which then tells whether a particular patient is severely ill or not. Under TN-KET, all severely ill (very severely undernourished or having respiratory distress or poor physical status) have to be prioritised for admission, Dr Frederick said.

Until now, TB SeWA flagged patients as 'severely ill' based on these variables, helping health staff prioritise them for inpatient care, Dr Manoj Murhekar, Director of NIE said.

"The new feature goes a step further - calculating and displaying predicted probability of death for adults with TB," Dr Murhekar said.

This objective risk percentage aims to overcome any subjective inference regarding severity and guide frontline staff to act immediately and firmly for the hospital admission of severely ill adults with TB at the time of diagnosis, he said.

"How this feature addition is helpful is that the predicted probability of death varies widely between a 'severely ill' and 'not severely ill' patient. The predicted probability for a severely ill adult with TB death ranges from 10 per cent to as high as 50 per cent, depending on how many of the conditions are present.

"In contrast, for patients not flagged as 'severely ill', the predicted probability drops sharply to just 1-4 per cent," explained Dr Hemant Shewade, a senior Scientist at NIE.

Data of last three years show that about 10-15 per cent of adults with TB in Tamil Nadu are found to be severely ill at diagnosis, he told PTI.

"This clear risk estimate will help ensure that the sickest patients are admitted to hospitals without delay," Dr Shewade stated.

Elaborating further, Dr Shewade said that while the average time from diagnosis to admission of a severely ill patient under TN-KET is one day, around a quarter of severely ill patients still face delays of up to three to six days in the state.

"There are still some severely ill patients - like 25 per cent roughly who get admitted after some delay. This feature will advocate and motivate healthcare workers to take immediate decisions on their referral for admission," he said.

"Over time, it will also help us analyse whether the average time from diagnosis to admission for a severely ill TB patient has reduced or not, and eventually this will contribute towards further reducing TB deaths in the state. Two-thirds of TB deaths occur within two months of diagnosis," he said.

The predictive death model has been developed using data from nearly 56,000 TB patients diagnosed in public health facilities across Tamil Nadu between July 2022 and June 2023.

Dr Frederick said that for predicting TB deaths, the five triage variables used under TN-KET alone were as accurate as all baseline variables captured in India's national TB portal Ni-kshay.

Dr Shewade further added that all baseline variables in Ni-kshay are available only by around three weeks (too late to use for prediction) while the five-triage variable are captured within a day in Tamil Nadu.

All 2,800 public health facilities in Tamil Nadu -- from primary health centres to medical colleges -- currently use the TB SeWA application alongside a paper-based triage tool, said Dr Frederick. "Tamil Nadu is so far the only state in India to systematically record and use these five triage variables to guide patient management," she said.

According to a study by the NIE, after implementing TN-KET for around three years now, the losses in the care cascade have significantly reduced, and around two-thirds of districts have documented a reduction in TB death rates.

The ICMR-NIE scientists emphasized that the initiative sets an important example for other states, where TB deaths, especially the early fatalities, remain a persistent challenge despite free diagnosis and treatment.