This hospital network in India has deployed IBM’s Watson to help diagnose cancer. Will it work?

Will developing countries with lack of data and problems of affordability be able to benefit from AI in healthcare? And how?

1.5 million Indians suffered from cancer in 2016, projected to grow to 1.7 million in 2020. At the same time, the country faces an acute affordability and awareness issue- only 12.5% patients come in for treatment at early stages [1]. As a result, 736,000 Indians died from cancer in 2016, projected to go up to 880,000 in 2020 [1]. These problems are compounded by the severe supply gap- the country has only 1,000 oncologists (ratio of 1:1500). In contrast the US has a ratio of 1:100. [2]

While the country grapples with lack of access and affordability, quality is also a major challenge. Many oncologists are unable to diagnose cancer early on, struggling to keep up with the large volume of research, medical records and clinical trials. Further, doctors face an increasing battle to stay up to date about best practices in treatment and care management. [3]

In this context, the use of AI to help detect cancer can be a potential solution. AI could improve quality of diagnosis and reduce costs (while increasing access) by reducing the time per oncologist spends on diagnosis (so that it can increase the number of patients served by limited oncologists)- so that oncologists’ critical time is used for human-intensive tasks such as consultation and treatment. [4]

Large hospital networks in India such as Apollo, Manipal and Fortis, which together serve ~60% of India’s organized private market [5] have the opportunity and responsibility to adopt new innovative ways of delivering care to patients. In 2016, Manipal (~12.5% of the organized private market share [5]) announced the national launch of IBM Watson across 6 hospitals. [3] The idea is to use a cognitive computing platform, to provide information and insights to physicians to help them identify personalized, evidence-based cancer care options across India. Manipal oncologists can access IBM Watson for Oncology for patients with breast, colorectal and lung cancer that are registered with a Manipal facility pan-India.

12 oncologists have been trained to use this solution to help patients in better diagnosis and administer treatment accordingly. In the long term, Manipal hopes to gain 3 benefits for its oncologists: [6]

  • Ability to keep pace with the growing and changing body of relevant guidelines, trials, articles, and patient data.
  • Confidently derive key insights from the relevant information and medical advances that are applicable to a patient’s condition.
  • Develop a more individualized, patient-centric approach to oncology while helping to increase time for patient-physician interactions.

Manipal hospitals tested the technology before deploying it, in a double-blinded study. Watson’s evidence-based recommendations was concordant with 90% of Manipal’s recommendations on its most complex cases. [7]

One challenge is that Watson or AI in cancer detection is in early stages and hasn’t proven to be significantly better than human approach. Especially in the case of rare cancers or recommending personalized treatment, keeping up with the rapidly evolving cancer treatments has been tougher than imagined. Manipal found that Watson was similar to doctors’ recommendations in most of the cases, so it has stopped using it for every case and now uses it only for 30% of the complex cases- 9% of which are influenced by care recommended by Watson. [8]

This fails the purpose of deploying Watson in the first place. In the Indian context, where 87.5% patients are not able to access early care [1], success of Watson or any AI based system will be determined by its ability to increase access for a majority of the population by reducing costs. To be of use in the Indian context, Manipal should use Watson in all cases rather than the complex ones, so that oncologists’ limited time is freed up to consult a much higher proportion of patients, potentially in areas which don’t have a doctor. This will not only increase access, but also improve quality as Watson will have more data from Indian patients, which it is lacking now. In the long term, it also needs to serve the rural population (70% of India’s population) which is not served by the overwhelmingly urban oncologist base. One potential solution is to combine the use of telemedicine with IBM Watson to reach patients in remote areas.

But some pertinent questions remain unanswered. Who will pay for the cost? Unlike any other country in the world, an overwhelming majority of costs in India are being borne by patients (65% of healthcare costs are out of pocket [9]). Can the use of Watson increase the number of patients treated to an extent that it reduces the average cost? Even if it does, what about data integrity? EMRs are just being adopted in India. In such an environment is it sustainable to make diagnosis dependent on a system such as Watson which relies on and gets better with higher quality of data?

(799 words)


1 Press Trust of India, “Over 17 lakh new cancer cases in India by 2020: ICMR”, India Today, May 18, 2016,, accessed November 2018

2 Sushmi Dey, “India has 1.8 mn cancer patients but only one oncologist to terat every 2000”, Business Standard, May 24, 2014,, accessed November 2018

3 IBM, “Manipal Hospitals announces national launch of IBM Watson for oncology”,, accessed November 2018

4 Rashmi Mabiyan, “How artificial intelligence can help transform Indian healthcare”, May 23, 2018,, accessed November 2018

5 India Brand Equity Foundation, “Healthcare”, January 2017,, accessed November 2018

6 IBM, “Watson for Oncology”,, accessed November 2018

7 Presentation by Prof. Dr. S.P. Somashekhar, Chairman, Manipal Comprehensive Cancer Center, Manipal Hospitals, Bangalore, India at the San Antonio Breast Cancer Symposium, December 9th, 2016-, accessed November 2018

8 Daniela Hernandez and Ted Greenwald, “IBM has a Watson Dilemma”, The Wall Street Journal, August 11, 2018,, accessed November 2018

9 The World Bank, “Out-of-pocket expenditure (% of current health expenditure”, 2015,, accessed November 2018


Will 3D Printing Change the Future of Homebuilding and End Global Homelessness?


Open innovation in the Public Sector: The Argentinean case

Student comments on This hospital network in India has deployed IBM’s Watson to help diagnose cancer. Will it work?

  1. I agree with you that the question of “who pays” is important. I think it’s not insurmountable. Training datasets used for Watson and other western-based AIs lack information many countries in the world.The lack of global data is an opportunity to commercialize existing national data sets and earn money to subsidize access to health AI benefits for the populations in these countries, while improving the AI’s predictive power.

  2. When I first looked at the title of this post, I got really excited to think about using AI such as IBM Watson in a developing country like India. However, not only did you raise the point about unproven efficacy of Watson in diagnosing cancer compared to oncologists, but it also raises the question of what India truly needs and is ready for. To your point, EMRs are only just being used, which brings into question the type of data that an Indian population would have in comparison to a more developed healthcare infrastructure. Watson would most likely need Indian-specific data to train itself on since Indian populations are usually genetically susceptible to different diseases compared to white Western populations, meaning that it might not be ready to diagnosis in India.

    This also reminded me of something an old boss of mine once said when talking about a small nonprofit that wanted to purchase an overly expensive and complicated IT solution for their organization – “They want to buy a cruise ship, but they’re swimming in a pond. They only need a rowboat”. Similarly, despite the enormous power of Watson, I am not sure it would have the most impact in India. An article in BMJ Global Health ( explores the implications of using AI in developing countries; while they clearly outline the huge potential for clinicians to be able to do more for patients, it also mentions that while diagnostic algorithms are useful, they may be irrelevant because some countries may not be equipped to care for patients even if they are diagnosed. While this may not be the case in all parts of India, I think it is important to think about the lower hanging fruit that exists (diagnosing other more commonly treated conditions, improving access, etc) that could be applicable in developing countries.

    P.S. Spangler – please server better food on the weekends 🙂

Leave a comment