With more than 70 hospitals across India and a patient list in the millions, Apollo Hospitals Group has a monumental task on its hands. But within that, the organization has an ethos beyond “primum non nocere” - the core healthcare principle that translates to “first, do no harm” - it wants to use technology to further treatment for its patients.

“We are more advanced than any other healthcare organization in India, for sure,” an Apollo Health Axis’ tech leader, who requested anonymity, tells DCD.

As far back as 2013, Apollo was boasting in its Annual Reports that it was storing patient records in a “private cloud” and that they were accessible over the Internet.

“For 15 years or so, we ran from in-house data centers,” he says. “But we figured out, for the pace of growth we were doing, we needed a much more advanced or easy way of scaling, and that’s where we went into the cloud and saw we could scale with minimal to zero downtime.”

The migration process itself only took around a year and a half to complete, though there were some understandable hesitations before migrating to the cloud - concerns about how data would be managed and stored, whether the organization had the expertise not only to migrate but to then function in the new environment. “Performance was also a concern raised,” he adds.

Besides that, the migration process ran relatively smoothly.

Apollo has, for the most part, opted to use Microsoft Azure - leveraging customized versions of Microsoft products - with a small use of Google Cloud Platform, though he notes this isn’t for “enterprise” applications.

While cloud computing is increasingly being adopted by the healthcare sector, it is still common for hospitals to have on-site data centers.

For Apollo, the migration was unavoidable. As the organization grows - so does the patient list, and each patient comes with a trove of data that doesn’t really go out of date.

“Healthcare data is huge,” he says. “Every day, we have some [hundreds of thousands of] patients in India within all of our hospitals, and each of these patients comes with several sets of data.

“There are the physical documents we get from outside, which have to be uploaded into the system. The document management system is a whole set of work we are doing in parallel, and then there is the work within the Apollo entity itself. Patients go through investigation, consultation, imaging, through non-clinical operation services, and then there is the pharmacy piece of it. We want each of the touchpoints the patient has with us to be recorded so that we are compliant with regulations.”

The regulatory complication means that, as he puts it, they can’t “push” the data. For local compliance reasons, the absolute minimum time period health data has to be stored for is seven years, but Apollo pushes this even further.

“In our environment, we do so much analysis that is collected from the beginning of the patient journey,” he explains.

“Say they were born in 2020 and joined Apollo then, we would keep the data for their entire life, so there is no concept that we would want to push it. We wanted to accept the reality that yes, this data is going to grow day by day, and we will figure it out.”

The organization has separated its hospitals into nine different regions, and each region has its own cloud instance. This is for a variety of reasons, one being that there are regulations, even at a state level, that mean all “transactions” for a patient have to occur in that specific locale.

It also makes it easier to dig through the data if it is somewhat siloed according to location. Beyond this, he says the separation acts as a solid backup plan.

“Even if one center (cloud instance) goes down, there are still eight other centers that are working, and it ensures a high availability,” he says.

One of the company’s current tasks is building a data warehouse for all nine regions in a centralized place as well, which is currently somewhere in the region of 10 to 12 terabytes of data, and growing daily.

The importance of this is more or less self-explanatory. “At the end of the day, all data has to be pushed to a central database where everyone can access everything, so if a patient visits a different region, they can still access it.”

Risky business

Healthcare data is, by its very definition, personal. As companies become more IT literate, there is an increasing understanding that the cloud - just because it isn’t in your physical building - isn’t inherently less secure.

Regardless of where data is stored, however, the cybersecurity question must be considered and resolved.

“When it comes to security, it’s what you define that matters, not where you define it,” he argues. “If you are able to set up your encryptions and security boundary right in terms of cybersecurity.

“We needed a much more advanced or easy way of scaling, and that’s where we went into the cloud.

“It's the same equivalent of if we had a set of vulnerabilities in a data center, as in the cloud. It’s the same - you just need to know how you define your network protocols, transfer protocols, and security standards, etc.”

By putting data into the cloud, you are in some ways putting the lifeblood of your organization into another's hands. While cloud providers have various SLAs and uptime commitments, outages are an unavoidable fact of operating.

To keep heart rates low, we discussed the possibility of an outage in hypothetical terms.

“With an outage, the major risk is obviously the patient care,” he says. “That’s the most important thing, and the unavailability of the information surrounding this, such as transactions, billing, and EMR (electronic medical records) data, is a major disruption. It could delay diagnoses, getting treatment, prevent prescriptions from getting processed, or lab results from arriving.”

Apollo, with its technology-forward strategy, has the added complication that the hospital beds are all remotely connected so that staff at its control center can monitor well-being 24/7.

“If that goes down, and the real-time monitoring crashes out, then there’s a chance of people getting into code blues, for example,” he adds. As is to be expected, we have various strategies to avoid just this - including disaster recovery zones.

Additionally, the group does regular “drills” - which the exec compares to fire drills - to practise dealing with such an issue. “Worst case, we have printed paper forms,” he adds, though noting “it's a huge task” with each patient having possibly up to 100 forms.

This exact “worst case” scenario hit Guy’s and St Thomas’ NHS Foundation Trust hospitals in the UK in 2022.

The two data centers supporting the hospitals overheated, leading to an outage and forcing staff to turn to paper and cancel or delay appointments.

The following year, it was estimated that the outage had cost the NHS £1.4 million ($1.7m).

AI-Patient confidentiality

Though these challenges around security are not insignificant, adopting the cloud has brought many benefits to the healthcare sector, not least that it opens the door for wider adoption of AI.

It is when discussing the application of AI that the exec gets really excited.

Apollo has been using AI for several years to assist with its imaging to give doctors an overview of what a scan is demonstrating - for example, measuring a tumour's comparative growth between scans.

Now, AI is even able to provide a “valuable explanation.”

“It not only gives the information, but it can explain what it's been taught. If it knows the depth and density of a tumor and its location, it can, using these scenarios, give an explanation of why that might be, and then the doctor can review that information.”

One of the main ways AI is used is for “big data analysis,” which enables the company to look at longitudinal data from a patient, for example, if they have a chronic disease, and see how it is progressing and what comorbidities are likely to develop or require treatment.

Today, Apollo is also using AI to recognize audio, and feed information directly into the EMR, so its doctors can focus on working with patients rather than writing copious notes.

“The auditor captures information as the doctor and patient speak and then summarizes the information that is most important and puts it in an instructional way for the doctor to review and edit, so 90 percent of the work is done, and there’s just the 10 percent job of editing,” he explains.

“The risk with this is that there are more than 18 official languages in India, but we are in the process of enabling those languages, so even if they speak Punjabi or Tamil, for example, the data is still picked up and then translated into English.”

Currently, the company has around 30 to 40 engineers who “take care of generating the clinical algorithms,” and some of them are already on the Microsoft marketplace and being offered to the outside world.

For example, in 2018, Apollo and Microsoft teamed up on the AI Network for Healthcare, developing an AI-focused network in cardiology. The two developed and deployed machine learning models to help predict patient risks for heart disease and assist doctors on treatment plans.

Later that year, they launched an API (application program interface) for cardiovascular risk scores that is purely for the Indian population, and more than twice as accurate as using data from other nationalities.

In January 2025, the two collaborated again, with Microsoft and Apollo committing to innovate and develop products together on topics such as disease progression, genomics, and multimodal models. That commitment also saw Microsoft agree to further assist Apollo with their data strategy, engineering platform, and integrating AI to create new intellectual property (IP) that could be used in markets outside of India.

Apollo has also worked with Google Cloud to develop a “Clinical Intelligence Engine” with Vertex AI and generative AI models. With the use of AI tools, of course, comes the necessity of training language models, which isn’t something that Apollo itself is engaging with - instead taking existing models, frequently from Microsoft, and “feeding” their data in.

It also isn’t a concept that the exec is willing to rule out, however, with him noting that perhaps Apollo’s data could be used to feed into a future more. He excitedly adds: “Who knows, maybe one day Apollo will have its own to run.”