Artificial Intelligence in Public Health Screening

Dr. Nobijith Roy

Artificial Intelligence in Public Health Screening

A Report, based on the Keynote Speech delivered
– By –
Dr. Nobhojit Roy, Team Lead, Health Systems Strengthening, State Resource Unit, CARE-India, Bihar
@ Artificial Intelligence in Radiology 2018 Symposium, November 10, 2018
Organized by Telerad Tech and Image Core Lab


India’s Technology March

In the recently concluded AIR Symposium 2018 in Bengaluru, Dr. Nobhojit Roy presented some rather interesting views on AI in his Keynote address. In his opening remarks, Dr Nobhojit Roy expressed immense pride in India’s Information Technology (IT) journey and especially the lead that it has taken in the space of Artificial Intelligence and machine learning.  In the same breath, he also felt that application of AI in public health screening is still conceptual and nascent, and remains to be tested on the ground.

Availability of Drugs in a Public Health System is a challenge

Availability of drugs at the point of care is important for addressing public health challenges such as tuberculosis (TB) in India. It is often experienced that a patient gets diagnosed with TB in a peripheral facility, but the facility does not have the required medicines. As a result of this, the patient is forced to travel to district hospitals or any other higher centers. Dr Nobhojit felt that availability of drugs and well-equipped diagnostic facility at the primary care level is a pre-requisite for the success of flagship programmes such as the Ayushman Bharat (the National Health Protection Scheme). There would be no point in having a diagnosis and not being able to link it to the essential drug list.

Expressing his concern, Dr. Roy said, that linking diagnostics to the drug list and making it available all under one roof seemed to be the real challenge. However, he was optimistic that availability of medicines was getting better on ground and that the essential medicines lists were getting linked with the health & wellness centers and at the Primary Health Center (PHC) level. He added that 7 diagnostic tests at the sub health center level were being promised while at the PHC the situation had improved to provide up to 90 diagnostic tests. Artificial intelligence would have a very vital role to play in Supply Chain Logistics for drug warehousing and supply, but needs to be based on a robust database. Current platforms of e-aushadhi are being used sparingly by the EAG states.

Manpower Training

In order to improve their skill sets, most of the Physicians go for overseas training at prestigious schools, return and perform successful surgical procedures here. Recovery during post-surgical care is usually handed over to the nurses who been through inadequate training courses in nursing. The surgical outcomes is dependent on the weakest link in the chain – the inadequately trained nurses and not on the highly trained physicians. India seems to lag behind here as we need to approach the systems-approach over the individual-centred approach.

Dr. Roy expressed his concern, stating that even at the PHC level, there are 69% vacancies to be filled in many states. The manpower at these centres are primarily a ANM nurse. Setting expectations, he said that the aim should be to push for an MBBS degree, as a bare minimum to improve consulting at that level. The patient should also receive a data warehouse of education material and patient care pathways which are being developed to aid in the proper treatment and diagnosis. Dr. Roy said that Accredited Social Health Activists (ASHA) are receiving a 5-day training on universal screening of Non-Communicable Diseases (NCDs) while Multi-Purpose Workers (MPWs) are being trained for 4-days on universal health screening. However, he felt that it would be a herculean task to train more than 1,20,000 of ASHA workers being present on ground.

Teleradiology – On ground realities

Post evaluation of all telemedicine projects around 14 states in India, Dr. Nobhojit Roy, was confident that Teleradiology was the best operational model. He expressed satisfaction that all image-based Teleradiology applications were doing well on ground. At the same time, however, he mentioned that it should be linked back to other systems. Not linking it back would mean creating a weak link in the overall function of the health system, using telemedicine.

Having enrolled in a Public Private Partnership (PPP) mode across 9 states – AP, Meghalaya, Rajasthan, Tripura, West Bengal, Assam, Odisha, Uttar Pradesh, and Uttarakhand, Dr. Roy stated that Teleradiology was doing reasonably well. He assured that the government was supportive of CT scans services at the district level hospitals providing full technical support across 24 states – 13 under the PPP module and the rest 11 under the in-house trust state model. Realizing that the out of pocket expenditure on diagnostics is high, financially, the government had provided US$ 172,937,239 (INR 1218 crores) to these 33 states for a free drug- free diagnostic initiative. The concurrent Biomedical Equipment maintenance program is important alongside, to reduce downtime and to be able to provide consistent diagnostic services to the population.

NCT PPP module – the future

Dr. Roy was optimistic that the government was taking steps to ensure that certain services are made available to the public under the Public Private Partnership (PPP) model. For eg; availability of mammography, biopsies at the diagnostic level is being envisaged as the rollout for the program; As a part of the aspirational plan, Cardiology- coronary angiography would be available at the District hospitals under Ayushman Bharat PMJAY scheme, and for Pulmonology- clinical examination during screening and PCG, bronchoscopy x-ray for diagnostics.

He lauded the ministry of health for having partnered with NICE (National Institute of Clinical Excellence), known for making the best guidelines in the world. Over the last three and a half years, jointly, they have come up with standard treatment guidelines for the district hospital. These guidelines will surely make all the health workers well aware of standard treatments and will help in overcoming ad-hoc treatments based on personal belief systems. The patient flow pathway algorithms are a part of it including patient education material. These may serve to build our algorithms for AI, if implemented well.

Adding to this. Dr. Roy mentioned that pathology pap smear was very amenable to AI. And this was a clear example of how AI can be looked at. The productivity of the pathologists can be enhanced using AI in this sphere. With a limited number of pathologists available to be able to look at all these slides that are likely to be generated by screening, could both time-consuming as well as tedious.  So, it would make sense to get at least the initial screening aided by AI.

Like-wise in the case of sputum microscopic, Dr. Roy hoped that where there are already various models in play, there needed to be capabilities of tying up teleradiology with tele-pathology. Dr. Roy believed that this was the space where it would pose a real challenge because according to him most image-based scans, even the retinal scans were proficient for mass screenings. Ethical issues with regards to data sharing will remain and will accompany the regulatory aspect of it. Real ethical and legal challenges notwithstanding, AI can potentially score in electronic health records, big data, epidemiology outbreak, and crowdsourcing. The aim should be augmented intelligence and it is multidisciplinary as the industry needs to get the doctors, data engineers, legal and social scientists, and the logisticians together. There is also the need for a strong curriculum to handle big data as the industry lacks in the same and thus it can’t be used effectively, Dr. Roy concluded.