Artificial Intelligence in healthcare in India

NAMS envisions a sustainable, high quality, affordable and patient focused care healthcare system in India, possible when steered by a special Cell comprising MoHFW, NAMS, NITI Aayog and MeitY.

An extensive review of up-to-date published literature and consensus statements and guidelines was undertaken by a Task Force of the National Academy of Medical Sciences (NAMS) specifically focusing on the application of AI in healthcare.

The Task Force has recommended that consensus be achieved amongst the various stakeholders in the health of the people of India, toward a national vision i.e., an open attitude towards judiciously using AI for healthcare.

NAMS envisions a healthcare system in India, both in the public sector and the private healthcare system, in which a collaborative, multidisciplinary approach will ensure a digital technology-adoptive population and an AI-enabled healthcare system, through standardised, evolving, evidence-based guidelines, to deliver sustainable, high quality, affordable and patient focused care.

The key issues identified by the Task Force include privacy issues, accountability, transparency, and explainability, training and standardization issues.

Recommendations have been made to bridge critical gaps/deficiencies as identified, including capacity building. Presently there is no formal system for upgradation and verification of skill sets of healthcare professionals in applying AI in healthcare. The measures required to be taken call for stakeholders to take effective action to create a future for India, where the health professionals are knowledgeable about the risk factors, beneficial outcomes, and overall risk-benefit ratio of AI in healthcare, and individuals feel empowered to talk with their healthcare providers about AI-enabled healthcare whenever appropriate.

There needs to be concerted efforts by policy makers and medical professional bodies, to focus attention on the policy gaps, and India specific recommendations on awareness and training for application of AI in healthcare.

Thus, there is a need for budgetary allocation of funds and policy initiatives, for assigning research priorities, and a need for convergence of medical specialties, to better serve the Indian population.

A community-based strategy is recommended to create awareness periodically. This is recommended to be steered by a special Committee or Cell that may be established by Ministry of Health and Family Welfare (MoHFW), with the cooperation of the National Academy of Medical Sciences (NAMS), in coordination with the NITI Aayog and Ministry of Electronics and Information Technology (MeitY).

Current situation in the country

• There is a shortage of qualified healthcare professionals, services, and infrastructure.
• The accessibility to healthcare is non-uniform, with a glaring disparity between rural and urban India. This problem is further accentuated by lack of consistent quality in healthcare across the country.
• With private spending making up most healthcare costs and the majority of these being out-of-pocket spending, which is one of the highest in the world, affordability continues to be an issue.
• The majority of patients only visit a hospital or doctor when their disease has progressed to an advanced stage, increasing the cost of treatment and decreasing the likelihood of a full recovery. This reactive approach to essential healthcare is largely the result of lack of awareness, access to services, and behavioral factors.
• The Covid-19 outbreak exposed the broken state of the current healthcare system, which works inefficiently by healthcare professionals, payers, and patients operating in silos.
• Only a small number of hospitals in India have either an HIS or an EMR that has digitalized patient records. Most records are still on paper. Even though India has an electronic health record (EHR) policy in place, sharing data between different hospital chains is still a work in progress because different chains have varied definitions of what “digitizing” information means.
• Due to its slow adoption rate, lack of enabling data ecosystems, inadequate availability of experienced workers and skilled personnel, high initial cost, unclear regulations, unattractive intellectual property regimes and deficiencies in its large-scale and extensive dissemination throughout the healthcare industry, there is still a sizable gap in the full utilization of artificial intelligence.
• Cost reduction will provide a significant challenge for hospitals as India transitions to a value-based healthcare ecosystem. Effective care management will be required, which means hospitals must actively manage their patients to enhance patient health outcomes. This will be crucial for both their reputation and the money and incentives they receive from insurance companies.

The main challenges are shortlisted as lack of data in ready-to-use form; inadequate infrastructure and connectivity , particularly in rural areas; lack of regulation and governance about how these technologies will be used, who will be responsible for deployment, and how data privacy and security will be protected; lack of skills due to the shortage of skilled personnel to build and deploy AI solutions; adequate and timely funding and investment that can limit or halt the development and deployment; and ethical and societal concerns about implications of AI, particularly in terms of the impact on jobs, privacy and equality.

Recommendations made to bridge the critical gaps
The measures required to be taken for stakeholders to take effective action to create a future for India, where.

• The health professionals are knowledgeable about the risk factors, beneficial outcomes, and overall risk-benefit ratio of AI in healthcare, and individuals feel empowered to talk with their healthcare providers about AI-enabled healthcare whenever appropriate.
• Evidence-based practices for the development, deployment, and use of AI / ML in healthcare are clearly understood and routinely applied by all medical professionals in all settings.
• New scientific evidence is constantly being uncovered to fill gaps in knowledge, and these findings are quickly and easily disseminated to the healthcare professional educators and put into practice by healthcare professionals.

Key issues/gaps identified in the current infrastructure.
Data is often referred to as the lifeblood of AI, and powerful algorithms generated by AI are steadily entering and transforming the decision-making processes in all areas of health care, public health, and medical research. Lessons learned from the pandemic and exponential growth of data in the Indian healthcare system warrant the use of AI for gathering evidence-based insights with agility and ease in real world medicine. Digitization in India took a rapid pace during the pandemic, replacing the conventional systems with advanced technology to connect people and provide them with real time public health information by the central, state and local governments. But the country faces significant gaps, delays and challenges in data management and governance at enterprise levels of the healthcare system. India’s massive wealth of data and insights is trapped in silos within the legacy core that must be freed to create a data-driven healthcare system. Government legislation to facilitate data collection and sharing will create the foundation for adopting and implementing AI technologies in the vast healthcare system.

Way forward
The NAMS Task Force after due deliberations on the need for current evidence about AI-enabled healthcare practices, has proposed the conduct of a rapid multi centric cross-sectional study on a pan India basis, to ascertain a representative view of the real world (non AI-enabled healthcare) practices. This is intended to be undertaken to develop indigenous digital health interventions that improve the health outcomes of the population and reduce the cost and time of healthcare delivery.

Short Term

• We need to identify the areas within healthcare where current strategies are not effective, and integration and complex analysis of novel, unstructured data are necessary to make accurate predictions.
• Automation of routine administrative tasks to enhance workflow and save costs.
• We need to identify the existing challenges and explore solutions for enhanced assessment and adoption of technology in the real world environment.
• Different stakeholders have adopted different interpre-tations interpretations of ‘digitizing’ records. Hence, there needs to be a guiding document that erases ambiguity from these terminologies.
• There is an urgent need for an integrated technology solution that is Interoperable, Scalable, and Affordable.
• Adoption of a cloud-based health management system can reduce the challenges arising from bureaucracy and red-tapeism, and operational costs.
• Investment should be made in the training of healthcare leadership and workforce. A comprehensive understanding of AI in national curricula for students studying medicine and public health is also necessary for integration of AI into healthcare systems.
• Entry of startups should be encouraged by facilitative infrastructural government initiatives.
• The principle of ART – Accountability, Responsibility and Transparency should be embedded in all AI applications. Focus should be on human-in-the-loop and human-centric designs that enable healthcare personnel to comprehend how a decision is made and how to apply this information to therapy.

Medium Term

• Data entry into Electronic Medical Record (EMR) systems can be automated by use of technologies like ambient AI, IoT and 5G.
• To promote coordination between academia, government, industry, NGOs, and patient advocacy groups, the government must also invest in and create public-private partnerships across the healthcare sector.
• Evidence generation, adoption and full integration of solutions need to be incentivized.
• Scaling up governance and regulatory frameworks will ensure adequate protection of privacy, equity, and openness.
• Critical investments should be made on sustainable data pipelines and infrastructure, design, and processes, as well as innovative business models.

Long Term

• Having an entire department dedicated towards clinical artificial intelligence (in each health facility or in each region) may be considered in future.
• As an option instead of working in silos, a multidisciplinary group involving all stakeholders such as administration, technology, and clinical staff, is recommended.
• Data stewardship is essential to build trust and long-term integration of AI into India’s healthcare system.
• Implement new AI tools alongside existing tools, so that practitioners can feel comfortable with the new tools and /workflows and experience their added value in practice firsthand while awaiting further research studies on the clinical evidence, implementation, and benefits of AI.
• Long-term success depends on a controlled approach that scales up AI in healthcare while ensuring meaningful human control and informed consent.

Digital pathology

• Creation of five centers of excellence in Computational Pathology. This could be done along the lines of the Swiss Digital Pathology Infrastructure (SDPI) to“organically facilitate (a) multi-site clinicaltrials and (b) implementation of tools for increased diagnostic accuracy, quality and patient safety, while simultaneously unlocking the hidden value of DP images by (c) advancing the utilization of artificial intelligence (AI) towards improved patient-level diagnosis, prognosis and therapy response prediction.”
• Mandate progressive digitalization of physical slide collections and the use of digital tools in the analytical phase of diagnosis.
• Initiate India Specific Grand Challenges for Annotation and Diagnosis. Large country level data base collection for various cancers and non-cancer conditions must be created. The huge patient load in India can make this generation a fast process. Grand challenges and crowdsourcing with incentive-based approach to attract postgraduate Pathology students to participate in such gaming competitions should be one of the immediate activities that can be started under the aegis of NAMS.
• Central regulations and guidelines for maintenance of medical records and best practices in Pathology lab. This is the most important step for the future when we are looking forward to acquiring AI based solutions for health problems and when we are implementing AI over expected similar standards of pathology tissue handling. ICMR can spearhead this activity.
• Wide scale support for purchase of equipment (e.g., digital microscopes) and software must be initiated by the government quickly. An ecosystem for multi-institution collaboration to create a country repository of digitized slides (including WSI) must be initiated by the government at the earliest. Such a facility can be meaningfully used only with a proactive role of NAMS and the government by training and teaching data science basics and holding webinars demonstrating the importance and applications of AI in pathology in various medical colleges across India.
• Given the bias in datasets and the difficulty in creating datasets which are properly annotated and of the desired quality, emphasis has to be on the development of unsupervised, weakly supervised and transfer learning based models for effective use of AI in pathology.
• Health grid to be created which can allow access to datasets and (some) compute and storage resources.
• India specific regulatory and ethics recommendations for use of AI in clinical diagnosis must be defined.

Suggested policy activities and advocacy for policy makers
AI systems are poised to drastically alter the way businesses and governments operate on a global scale, with significant changes already under way. This technology has manifested itself in multiple forms including natural language processing, machine learning, and autonomous systems, but with the proper inputs can be leveraged to make predictions, recommendations, and even decisions.

As sectors such as healthcare invest significant resources into AI, it is critical to understand the current and proposed legal frameworks regulating this novel technology. Specifically for multinational enterprises operating globally or international collaboration on healthcare innovation, the task of ensuring that their AI technology complies with applicable regulations will be complicated by the differing standards that are emerging from around the world.

Recommendations for health/medical professionals
Given the promise of AI for impacting the practice of medicine, there are multiple perspectives on how it may impact medical education. Since health professionals in training or junior health professionals are likely to see significant adoption of AI during their careers, it is of interest to teach how AI may be used to augment clinical workflows. There is also interest in using AI as a tool for quantitative assessment of trainees, particularly in the context of medical procedures. Finally, given the likely role AI will play in many specialties of medicine, there is a question of the appropriate level of training a clinician should have with this technology. There is a role for educational research to guide how AI can be incorporated into medical education.

In the context of augmenting clinical workflows, AI has the potential to serve as an instructional and learning tool for triaging cases, particularly those involving image and video analysis. Specifically, AI-based tools may be used to confidently identify cases for which the diagnosis is straightforward or clear (e.g., of echocardiograms), so that the clinician may focus on the challenging cases. Such integration of AI technology into medical workflows is likely to occur across all specialties for which image and video analysis is central.

AI also has potential as a quantitative assessment tool, particularly in the context of training implementation of procedures. For example, in the context of training and evaluating surgeons’ videos may be taken of a procedure, and AI may be used to segment the video into a series of underlying steps or tasks. One may use AI to evaluate the quality of each step as performed by a trainee, relative to a skilled surgeon, and may identify areas for trainee improvement.

Given the impact that AI is likely to have in many aspects of medicine, there is a question of how much training a clinician should have in the underlying technology. It is unlikely that most clinicians will need expertise in the methodology of AI, but they likely should understand the fundamentals of how it works, to appreciate opportunities for its use, as well as to recognize its limitations to translate its meaning when interacting with other health professionals and with patients.

Based on a Task Force report by Directorate General of Health Services, Ministry of Health and Family Welfare. For complete report, https://www.medicalbuyer.co.in/ai-in-healthcare-in-india-nams/