AI is trending, but is it risky?
AI applications can create an annual savings of as much as $150 billion for the US healthcare economy by 2026
Human brain was considered a robust machine, until it developed machines as brains. The development of artificial intelligence (AI), an amalgamation of various computer-based algorithms and software such as machine learning (ML), natural language processing (NLP), etc. are increasingly gaining ground, facilitating better management and attempting to improve treatment results. The descriptive use of AI in healthcare quantifies events that have already occurred and uses this data to gain further insights. The predictive use of AI, though contested, is said to improve efficiencies and reduce costs – it can help in primary screening at rural healthcare centers that are usually short staffed. The third is the prescriptive use of AI that suggests possible treatment based on predictive AI.
According to Accenture, key clinical health AI applications can create an annual savings of as much as $150 billion for the US healthcare economy by 2026. The AI health market in the US is expected to reach $6.6 billion by 2021— a compound annual growth rate of 40 per cent. In India too, several healthcare providers have collaborated with firms offering AI-based services to improve both access and outcome.
Promoting evidence-based practice in healthcare
The most remarkable feature of AI-based healthcare services is that it established evidence-based practice on a firm ground. In a country like India where quacks are a reality and access to healthcare is a major sticking point of governance, artificial intelligence is a messiah bridging the gap. Diagnostic laboratories are increasingly adopting artificial intelligence, combining them with efficient human findings, to get faster results that have low probability of error. Thermal analytics are used for detecting early-stage breast cancer while some detect tuberculosis from chest x-rays and acute infections from ultrasound images. Mental health, mainly depression, is considered a silent killer that does not distinguish between a celebrity and a commoner, and seeking help is stigmatised. It has found a friend in artificial intelligence – chatbots, an AI-enabled system, has been developed to provide support and suggest practitioners to consult.
The most beneficial use of AI in healthcare is probably through telemedicine that uses electronic communications and software to provide clinical services to patients remotely. Typically, it is used for follow-up visits, management of chronic conditions, medication management, specialist consultation and other clinical services. However, in remote and rural areas with little or sporadic access to healthcare, telemedicine is virtually godsend. The government of Telangana has collaborated with Microsoft for its ‘the Rashtriya Bal Swasthya Karyakram’ wherein the use of Microsoft Intelligent Network for Eyecare (MINE), a cloud-based analytics platform to reduce avoidable blindness in children.
Pharmaceutical companies have also benefitted a lot riding the AI bandwagon. Pharma companies have been using AI in manufacturing, extraction, processing, purification and packaging of components used for preparing medicines. When Ebola virus was wreaking havoc, an AI-powered programme was used to scan existing medicines that could be redesigned to fight the disease. The programme found two potential medicines in one day, whereas the traditional analysis could have continued for months or years. Beside drug discovery, pharma companies are also using AI for better competitor differentiation, optimal resource allocation for maximum market share, ability to maximise growth, customising sales and marketing and better customer engagement, and pharmaceutical supply chain management.
Other prominent use of AI in healthcare is managing data and databases that of hospitals, diagnostic centres and insurance providers. Using ‘Machine Learning’ that relies on neural networks (a computer system modeled on the human brain), companies are leveraging multilevel probabilistic analysis that allows computers to simulate and expand the way human mind processes data. Managing large amount of data, sifting through them for accurate information and at times suggesting possible treatment have contributed in reducing the cost of operations and improving service to care-seekers. Machine Learning helps in automating claims management, weeding fraudulent claims and improves customer experience for medical insurance companies while a combination of big data and AI can help them analyse lifestyle habits of customers and customise offerings.
Depending on a mechanical brain may have implications
Even the best of AI-based applications in healthcare have proven that human beings are irreplaceable, at least as of now. IBM’s ‘Watson for Oncology’ aid doctors in the diagnosis and treatment of 7 types of cancer by analysing data, research evidence and improve the quality of the report. However, physicians across globe have criticised as “a ‘mechanical turk’ – a human-driven engine masquerading as an artificial intelligence.” Instead, it is suspected, it convenes a small panel of cancer experts, who formulate recommendations for specific patient profiles. Some physicians have also complained bias of ‘Watson for Oncology’ towards the US patients and standard of care, something that is very unlikely of any AI-based system whose greatest strength lies in objectivity.
Given that artificial intelligence has now reached a stage where it can suggest medication in order to assist a doctor, it is dubious that the results will meet the intended outcome – imagine a situation wherein a doctor, discussing a critical case with a patient, being breathed down his or her neck. Overdependence on AI may also cause a conflict over possible medication or treatment of a patient and norms must be established to deal with confidentiality in the doctor-patient - AI relationship, informed consent (for different procedures and use/access of personal data - health related and beyond), and standards for AI driven medical research. While doctor-patient confidentiality has been codified in the Medical Council of India (MCI) Code of Ethics Regulations, 2002, setting professional standards for medical practitioners in India, any such code in using AI has not yet been set.
In view of security breaches, security data access by AI is also a major concern among patients. In 2016, a Mumbai-based diagnostic laboratory database was hacked, leaking medical records (including HIV reports) of over 35,000 patients. We need to examine if the existing laws and regulations are enough to deal with the flip sides of using artificial intelligence in healthcare.
In India, the doctor-patient relationship is special – the doctor is often held in high regard and patients bestow him or her with complete trust. A significant part of the population in India lack literacy and live in poverty outside of the ambit of formally recognised medical systems; the doctor is the god on earth. As medical practitioners, the well-being of our patients is as important as easing our burden, if not more.
Dr Dharminder Nagar is MD with Paras Healthcare.