Why doctors remain Cautious about AI in Healthcare
By Arunima Rajan
Amidst the growing enthusiasm for AI's potential to revolutionize Indian healthcare, a sector plagued by resource shortages, many doctors express concerns. They question the practicality of AI tools like ChatGPT in complex medical scenarios, highlighting issues such as personal data misuse, the lack of physical patient interaction, and the high implementation costs. These reservations underscore the challenges in harmonizing AI advancements with the realities of healthcare.
“It is indeed overly optimistic at this nascent stage of its evolution to assume that these innovations alone will improve India healthcare at the public health level. The ChatGPT version of AI was released in Nov 2022, just over a year ago. There has not been enough time for scientific validation and trial time to declare it as transformative,” says Dr Sunil Chandy, Chief Medical Officer of ITC India.
But to what extent can artificial intelligence systems surpass the performance of medical professionals in India?
The former Director of CMC Vellore points out that Artificial Intelligence has immense potential to impact the delivery of healthcare if it is positioned cautiously in the spectrum of care delivery.
“AI can serve the first and last segments very well as they are logistic and predictive in nature . The middle segment is less predictive and nuanced with individual variations which must be left to the clinicians judgement. With all its prowess, AI cannot surpass human capacity to deal with the multifaceted nature of illnesses which often requires sensitivity, concern, empathy and beyond-the-mile involvement,” adds Chandy.
Dr Tanvi Shah, assistant director of Apex Group of Hospitals seconds his views. “ChatGPT acts as a virtual health coach and by analysing patient data it suggests further scope of action. Every patient is unique and the clinical process is complex. So, ChatGPT can help with the analysis of the vast data, highlight the red flags, summarise the history, help get to the diagnosis faster and probably diagnose basic illness but cannot replace human intelligence, empathy and expertise for diagnosis and care as of now. Human body is dynamic and unique. Symptoms could be common for many diseases. So, as of now using it for diagnosis will only increase anxiety among patients,” she adds.
AI’s Impact on Public Health
Dr Chandy highlights that artificial intelligence is influencing healthcare sector in India in three major areas:
In public health – by dredging of large volume data to create predictive trends, predict imminent pandemics and population based control of many non-communicable diseases
In personal healthcare – by providing more accuracy in clinical decision-making, imaging and diagnostic accuracy, in pathology, and cancer care
In health-policy making to plan, administer and strategize health priorities on local, regional and national levels
However amidst these advancements, the question arises: Can artificial intelligence reinstate empathy and human connection in doctor-patient relationships? Dr Chandy firmly believes that the answer is no. He points out that empathy and compassion are real entities where physicality is essential. Virtual reality and algorithmic logic cannot masquerade as vehicles for empathy and compassion as the latter are unmeasurable and sensory in nature. AI can facilitate an encounter but cannot replace the human doctor where these soft qualities need to be harnessed.
Patient Perspective: Human Touch in Therapy
Lakshmi, a homemaker from Mumbai seconds his views. “I initially turned to a mental health AI chatbot, lured by its affordability with a 6000 Rs annual membership, a stark contrast to the 2500 Rs per session fee of a professional therapist. However, over time, I realized that what I saved in money, I lost in depth. The conversations with the AI felt hollow, missing the soulful connection and understanding that only a human professional could offer. This realization led me to invest the extra money in consulting a real therapist, understanding that some aspects of healing simply can't be digitized,” she explains.
Dr Satish Kumar PV, Head, Radiology, Kauvery Hospital, Bangalore provides a contrasting insight.
“AI tools quickly and accurately create detailed narrative radiology reports of a patient can greatly ease the workload of busy radiologists, not merely identify the presence or absence of an abnormality, complex diagnostic information, detailed descriptions, and nuanced findings. In short, they mirror how human radiologist describe what they see in a scan,” he explains.
He explains that AI can assist in automating various operational tasks within radiology, including evaluating appropriateness of imaging, managing patient scheduling, selecting examination protocols, and enhancing workflow of radiologists’ reporting. Another significant area benefitting from AI is post-scanning image reconstruction. AI is widely used for lung cancer screening, breast cancer applications, chest X-ray, and even recent applications in ultrasounds for maternal care where two deep-learning neural network models are used to predict gestational age and fetal malpresentation (non-cephalic vs cephalic) from ultrasound videos captured using a “blind sweep” method. It can be applied in low-resource settings. Triage tools, a popular choice for AI algorithm developers, help radiologists in prioritizing which images, to review first, particularly in time-sensitive fields including stroke, intracranial hemorrhage, pulmonary embolism, and spine fractures. Additionally, AI plays a vital role in opportunistic screening during whole body scans offering predictive insights into future health risks, such as heart attacks or strokes, by estimating the risks.
Dr Kumar admits that limitations of AI include lack of human judgement since AI algorithms are meant to analyze medical images based on patterns and data they have been educated on. “This leads to inaccurate diagnosis when it comes to reaching the ultimate diagnosis and choosing treatment plan. Other limitations include obstacles in the form of regulation on laws, resistance and acceptance of AI and cost involved with it.”
Global Digital Health Application Types
Source: World Economic Forum Report
The question of whether AI will eventually replace radiologist is a common concern among medical professionals. Dr Kumar, however, believes that AI is more likely to complement the role of radiologist than entirely replace them. He anticipates that AI will revolutionize the field of radiology in next two decades as technology evolves and become more accessible.
Privacy Concerns
The potential for the misuse of personal data, particularly by insurance companies and employers for crucial decision-making regarding job opportunities or insurance coverage stands out as the most significant risk associated with AI. “It is already playing out in most healthcare ecosystems. Data privacy is a concern and the anecdotal instances of data-misuse serves as a red alert for AI based informatics. Cyber security efforts are ongoing, but the potential for data misuse remains huge and expensive to administer,” adds Dr Sunil Chandy.
Dr Harish Pillai, CEO, Metro Pacific Health, notes that he is less anxious now regarding the potential for misuse of personal data due to the passing of the Digital Personal Data Protection( DPDP) Act 2023 by our Parliament. “They have specified the rights and obligation of both individuals as well as Data fiduciaries who collect, store and process data. This is a progressive legislation that took a lot of deliberations and public feedback from all stakeholders,” he adds.
Source: https://services.google.com/fh/files/blogs/india_economy_report_2023.pdf
AI has the potential to make healthcare services more affordable and accessible in even the remotest areas efficiently. However, the affordability of AI solutions for healthcare providers in India remains a significant concern.
Dr Pillai points out that investment in a technology architecture framework by healthcare entities in both public and private sectors is inevitable. “Without foundation systems such as data and process standardizations ; infrastructure setups inclusive of networks, servers, storage , security ; enterprise systems ; core applications and client interface solutions ; it will be similar to driving an automobile blind folded . The initial Capex and the subsequent opex relative to the vision of affordability in service rendering should always be kept in mind. Robust technology setups provide an analytical ability that can craft operational efficiency strategies resulting in affordable services that do not compromise margin contributions and hence sustainability. So yes AI Solutions even in the context of emerging economies can be affordable if deployed effectively,” he adds.
TOP-DOWN AND PRESCRIPTIVE?
The hesitancy within the Indian medical community towards the adoption of emerging technologies is also a notable concern.
“Yes there is hesitancy, for the following reasons
Thus far, the development of health-tech has been top-down and prescriptive without the involvement of doctors at the ground level. It needs to be participatory.
The language of technology is innately alien to doctors in its very nature. The twain between the two has not yet met. There is diffidence to embrace the new tech-language coming into healthcare.
There is a perception, and true to a certain extent, that health-tech is driven by its business potential and less for patient welfare. The rapid expansion of the stakeholder base in healthcare is largely driven by its market potential.
Physicians are trained to accept scientific validation and evidence before mass usage of any new drug or device. The plethora of commercial tech-applications without adequate clinical trials prevents them from quick acceptance,” adds Dr Chandy.
Gaurav Parchani is the CTO & Co-Founder of Dozee, an AI-based contactless Remote Patient Monitoring (RPM) & Early Warning System (EWS). He adds that technology is a force multiplier in healthcare, enabling providers to scale their capabilities and maintain quality care. “Without technology, skilled healthcare providers can offer good care on an individual level, but the limitations of the current system become apparent and worsen over time. Technology, however, addresses these challenges by facilitating scalability. It acts as a force multiplier when integrated with existing processes, enhancing the ability to deliver quality healthcare on a large scale, differentiating between serving a thousand patients and catering to a million,” he explains.
He adds that it's crucial to recognize that while technology serves as a catalyst for healthcare transformation, it is not a one-size-fits-all solution. The integration of technology, algorithms, and machines requires a thoughtful and collaborative strategy due to the inherent complexities in healthcare delivery. Machine learning algorithms, for instance, can analyse extensive health data, offering insights for personalized treatment plans, predictive analytics, and efficient resource allocation. To ensure successful implementation, collaboration among healthcare professionals, policymakers, technology developers, and the community is essential. Prioritizing ethical considerations, data privacy, and security measures builds trust in these technological advancements. In conclusion, while technology, algorithms, and machines are pivotal in improving healthcare, they are not standalone solutions. A holistic, collaborative approach that combines technological innovation with human expertise is necessary for sustainable and impactful results in the Indian healthcare system.
Parchani suggests that to address and mitigate the risk of personal data misuse, several measures can be implemented:
Robust Data Governance: Establishing and adhering to robust data governance practices, including clear policies on data collection, storage, and sharing, along with implementing encryption and secure storage protocols to safeguard sensitive information.
Transparency and Explainability : Ensuring AI algorithms are transparent and explainable, allowing individuals to understand how their data is being used. Clear communication builds trust and empowers individuals to make informed decisions about sharing their information.
Data Minimization and Purpose Limitation: Collecting only necessary data for specific purposes and limiting its use to those defined purposes to prevent unwarranted intrusions into individuals' privacy.
User Consent and Control: Providing individuals with control over their data and obtaining explicit consent for its use. Users should have the ability to opt in or out of data sharing, ensuring awareness and comfort with how their information will be utilized.
Regular Audits and Compliance Checks: Conducting regular audits and compliance checks to ensure data practices align with regulatory frameworks and ethical standards. This proactive approach identifies and rectifies potential issues before they can lead to misuse.
He also emphasises that while the potential for personal data misuse is a significant concern, it's crucial to balance it by highlighting the positive impact AI can have on healthcare, personalized services, and societal progress when ethically developed and responsibly deployed.