The Fifth Lever: the patient digital twin as a hospital's strategic asset

She is a composite, assembled from my years of running hospitals, but every hospital CEO in India will recognise her. A woman in her fifties calls the helpline of a large private hospital about chest discomfort. She browses the hospital's website that evening. Two days later she sits in the cardiology OPD, gives a blood sample, and gets a CT angiogram. Her son files an insurance pre-authorisation. A month on, she is admitted for two nights, undergoes an angioplasty, pays a bill, and writes one unhappy email about the discharge delay.

Count the systems that recorded her: the call-centre software, the website analytics, the OPD module, the laboratory information system, the radiology information system, the insurance desk, the inpatient module, the billing engine, the feedback tool. Nine systems did their jobs, and most did them well. Yet when she next arrives (at two in the morning, by ambulance, unable to speak for herself), the scene at the registration desk is one every emergency department in India knows: the family has left the papers at home, nobody remembers her registration number, and because it is an emergency the person at the desk does the only sensible thing – creates a new record. The hospital has met her nine times and does not know her.

The same game

India's large hospital chains are playing one game, and their own investor presentations say so. Max Healthcare's May 2026 deck lays the strategy out on a single slide: attract best-in-class clinical talent, optimise payor mix and capacity utilisation, expand beds through brownfields and greenfields, pursue mergers and acquisitions. Apollo told investors in August 2026 that it will add 5,798 beds by FY31 at a projected cost of about ₹11,152 crore. Medanta reported onboarding 110 doctors in a single quarter (Q3 FY26), most of them in cardiac sciences, critical care and cancer. Rainbow (the children's hospital chain) is doubling from 2,500 beds toward 5,000 at a planned ₹2,200 crore. Fortis compressed its plan into one sentence on its August 2026 earnings call: occupancy around 70 per cent, "with levers to improve by another 2-3 percentage points through new units and better clinical talent."

Four levers, then: capital into beds, doctors onto rosters, marketing into catchments, cost out of operations. And one scoreboard. The same Max deck charts Max against Medanta, Apollo and Fortis on exactly four measures – ARPOB (average revenue per occupied bed), occupancy, return on capital, and EBITDA per bed. The industry plays one game and publishes its own league table.

I pulled all four levers myself (at Max, Manipal, and what is now Gleneagles), and they built modern Indian private healthcare – hundreds of thousands of beds that did not exist twenty-five years ago, and clinical programmes that now compete globally. The trouble is that every lever is copyable. Capital builds a rival tower across the road. A cheque book moves a surgical team in one hiring season. Marketing playbooks travel with the agency, and cost programmes converge on the same consultants' benchmarks. When every player pulls the same four levers and reports the same four metrics, the returns get competed away. The game still works. It just no longer differentiates.

At least one insider has said this aloud. On Narayana Health's February 2026 earnings call, Vice Chairman Viren Shetty declined to report occupancy: "This is a number we are moving away from. We are not in the hotel business." Narayana, he said, had abandoned bed-led goals and now grows through "a combination of hospitals, clinics, and insurers", aiming at a patient's "total overall health spend" across the year rather than the next admission. Its specific bets (an insurance arm, an in-house software platform called Athma) will take years to judge. The decision behind them is already clear: a major chain has concluded that the four-lever game is fully priced, and has gone looking for the fifth.

The fifth lever

The asset missing from every deck is the one that compounds: a longitudinal, consented record of the patient herself. Beds, doctors, campaigns and cost programmes all buy episodes – an admission, a procedure, a package. A unified patient record buys something different. It grows more valuable with every interaction, it cannot be purchased ready-made at any price, and a competitor cannot poach it the way they poach a cardiac surgeon. Ten years of consented patient history is the only asset in this industry with a ten-year moat.

The four levers compete for episodes. The fifth competes for lifetimes. A hospital that knows a patient across a decade (her allergies, her stents, her family, her insurance, her unanswered complaint) holds a relationship that no billboard can dislodge.

And the window is closing, because others are building the front door to that relationship right now. PB Health (incubated by PolicyBazaar's parent) raised 218 million dollars in May 2025 and opened its first hospitals in Noida and Gurugram by mid-2026, explicitly promising an "integrated healthcare ecosystem" with digital health records at its centre. Tata 1mg (revenue ₹2,936 crore in FY26) is assembling a "Health Insights Hub" that consolidates a user's prescriptions, diagnostics and wearable data in one place. Amazon launched home diagnostics in India in June 2025 and now delivers medicines to every serviceable pin code. Of the three, only PB Health is building hospitals at all – Tata 1mg and Amazon own neither beds nor scanners. All of them are competing for the patient's memory. A hospital that declines to hold its patients' longitudinal story will find that story held by a payer or a platform – and will serve, from behind someone else's front door, as a supplier of procedures.

What a patient digital twin is

A patient digital twin is a living digital counterpart of one person: everything the hospital knows about her, held in one place, updated with every interaction, and able to answer questions about her future. This is the fifth lever in its mature form – the patient understood, and forecastable, ahead of her own disease. The Lancet Digital Health gave medicine its working definition in 2025: a "constantly updating virtual copy" that permits "analysis, simulation, and prediction", built so that the virtual patient "evolves alongside" the real one. The idea has already reached the operating theatre. In April 2026, a Johns Hopkins team reported in the New England Journal of Medicine that it had guided ventricular-tachycardia ablations by operating first on each patient's virtual heart (reconstructed from cardiac MRI) – over a mean follow-up of 405 days, eight of the ten patients stayed free of the arrhythmia without drugs. And Eric Topol, the American cardiologist who has chronicled medicine's digital turn, reaches further still: with deep data on enough people, "you could identify your digital twins early in life, and because those twins have already moved ahead of you in the world, you'd get a sense of what might lie in store" (January 2026). The research field is compounding to match – PubMed indexed three papers mentioning digital twins in 2018 and 912 in 2025.

Engineering, which coined the term, also polices it. The classic ladder (Kritzinger and colleagues, 2018) runs from a digital model (no automatic data flow), through a digital shadow (data flows in automatically, one way), to the twin proper (a living model, dynamically updated, that predicts and informs decisions – the US National Academies' 2024 definition). Almost nothing in medicine clears the top rung yet: a 2025 review in npj Digital Medicine found that 18 of 149 published healthcare "digital twins" (about 12 per cent) met the formal criteria, and Topol says the same thing in plainer words – "we still don't have any good digital twin resources yet, not even theoretically, but we could do this, and it would be phenomenal and useful."

For a hospital, the ambition comes in two stages. Stage one is the shadow: one central, consent-first, longitudinal record into which every interaction writes itself automatically – the phone call, the OPD visit, the lab result, the scan, the admission, the dialysis session, the reimbursement claim, the complaint. The nine systems feed one asset, instead of one viewer stitching nine screens together at display time. Stage two is the twin proper: a model of the patient you can question and project ("which of our diabetics is drifting toward the ICU?", "which post-angioplasty patients have gone silent?"). Stage two is arriving in fragments – a randomised trial of a metabolic "whole-body digital twin" run in Indian hospitals reported meaningfully larger blood-pressure reductions than standard care (JACC: Advances, 2024, with the caveat that the vendor's scientists led the study) – but stage two is worthless without stage one. And stage one is precisely where Indian hospitals are weakest.

A 2025 paper in the Bulletin of the World Health Organization, describing why India's National Cancer Grid had to commission its own interoperable EMRs, lists the defects of existing Indian hospital systems plainly, ending with "a lack of interoperability." An 800-bed private teaching hospital studied the same year still ran entirely on paper. Fortis's own chief information officer described the 2024–25 milestone as digitising OPD records in fifteen hospitals and inpatient records in one. The shiniest digital assets confirm the pattern: Apollo 24|7 serves 46 million registered users, yet in the eighteen pages of Apollo's Q3 FY26 earnings transcript the platform is measured entirely in users and GMV – the words "patient record" never appear. These are channels and commerce layered onto the old module stack. A front door is not a memory.

Two exceptions stand out. Narayana has built an in-house platform it claims holds over ten million patient records and has launched a "unified 360-degree patient view" – the most architecturally serious attempt among the chains, though it spans only Narayana's own ecosystem. And the government has built the identity rails: 93.95 crore ABHA numbers and 105 crore linked health records as of July 2026. Divide those two official figures and the state of the nation appears – roughly 1.1 records per person, most of them from government programme feeds. India has identity at astonishing scale and longitudinal depth of almost nothing. The rails are laid. Nobody has yet run the train.

Consent is the spine of the whole design. A twin is built with the patient: every record enters because she agreed, she can see what she agreed to, and she can walk away. This separates a patient digital twin from a surveillance file – and, from May 2027, Indian law will demand most of it anyway.

Three questions for the next board review

Every CIO who reads this far will say, reasonably, "we already have most of this." Sometimes they will be right. So a leadership team needs a test it can run in one meeting, in plain language, with no technical vocabulary. Three questions, in rising order of difficulty.

First, the timeline test. Pick one patient (you pick, in the meeting, not a name supplied in advance) and ask to see everything the institution knows about her in one time-ordered view: calls, web visits, consultations, results, scans, admissions, bills, complaints. Allow minutes, not a week. The surprise and the clock matter, because any organisation can compile one heroic dossier given a fortnight's warning. This test catches fragmentation, and most hospitals fail here.

Second, the cohort test. Ask one question that cuts across patients and across time. How many patients had an abnormal result this year and never returned? How many callers reporting chest pain were admitted anywhere in the network within thirty days? The answer should come in days, from the hospital's own systems, without a consulting engagement. This is the test that separates an asset from a view: a "patient 360" dashboard that stitches nine systems together at display time will pass the first test and fail this one, because you cannot compute across data that only meets on a screen. Everything downstream (safety monitoring, research, commercial insight) lives or dies on cohort questions.

Third, the mirror test. Can the patient see it? Her own longitudinal record, what she consented to, and the means to share it with an outside doctor of her choosing. A "twin" the patient has never seen and never authorised is a surveillance file wearing a better name. This is where even the genuine builders grow quiet.

Most hospitals fail the first test. Nearly all fail the second. The third is where the industry's real distance from patient-centricity shows. A board can run all three before lunch.

Why are Patient Digital Twins critical?

First, safety. Every clinician has met a version of the allergy recorded in the OPD module that the inpatient system never shows. The international evidence for what a unified record does to safety and outcomes is old and strong. Kaiser Permanente, running one longitudinal record across millions of members, lifted hypertension control in Northern California from 43.6 per cent to 80.4 per cent between 2001 and 2009 (JAMA, 2013) – an achievement a fragmented system cannot even attempt, because a fragmented system cannot count its hypertensives.

Second, patient convenience. The woman from the ambulance never again repeats her history, never carries films in a plastic bag, never re-does a test because the last result is in another building's system. Quantified Indian evidence on duplicate testing does not yet exist (a gap worth closing), but a Dutch study found 56 per cent of GP-ordered tests were repeated when patients crossed into hospital care – and any Indian patient will tell you our walls are higher.

Third, architectural freedom. With the twin at the centre, the nine modules become replaceable plug-ins around a permanent asset, instead of nine silos each holding hostages. The hospital can change its LIS vendor without losing ten years of results. Vendor lock-in weakens the day the data stops living inside the vendor's product.

Fourth, real-world evidence – the benefit most Indian hospital boards have never discussed. Consented, de-identified longitudinal clinical data has become a scientific instrument and a serious business. Roche paid 1.9 billion dollars in 2018 for Flatiron Health, whose asset was curated longitudinal records from oncology clinics. The US FDA used such data in 2019 to extend a breast-cancer drug to male patients, who are too few for conventional trials. Israel's Clalit (holding two decades of unified records) produced the world's first large real-world evaluation of the Pfizer covid vaccine (1.19 million matched people, NEJM, February 2021) weeks before any other system could – and Israel's government had quite literally traded early vaccine supply for its data's analytical output. No Indian hospital group operates anything comparable today. The most prominent attempt (Karkinos) passed through insolvency into Reliance's hands in December 2024 for ₹375 crore. Meanwhile India's National Cancer Grid alone sees roughly 750,000 new cancer patients a year. The country with some of the world's richest disease experience contributes almost none of the world's real-world evidence. For the first hospital group that builds the consent and the curation to change that – ethically, transparently, with the patient's agreement earning the patient a share of the benefit – the gap is an opening.

Fifth, commerce. A hospital that truly knows its patients can serve them across the year (the screening due, the follow-up lapsed, the parent ageing into cardiac risk) rather than waiting with the marketing department for the next episode to walk in. Narayana's phrase for the prize is the patient's "total overall health spend."

The clock is set

Until now, everything here was a choice. From this point the calendar intervenes. India's Digital Personal Data Protection Act received its final Rules on 13 November 2025, and the substantive duties commence eighteen months later, in mid-May 2027. The window may yet shrink: in January 2026 the government proposed cutting those eighteen months to twelve, which would pull the deadline back to November 2026 for large data fiduciaries (the proposal sits ungazetted as of this writing). What the law demands is, almost line for line, stage one's plumbing. Consent must be specific, informed, and provable – and the burden of proving it sits on the hospital (Section 6(10)). Every patient gains the right to see her data, correct it, and have it erased, subject to the legally mandated retention of clinical records. Breaches must be reported to every affected patient without delay and to the regulator in detail within 72 hours. Penalties reach ₹250 crore per breach of the security obligations.

Two consequences follow for a hospital board. The first is financial: the consent-and-rights plumbing will be paid for either way, so the only real decision is what the money buys. Spent defensively (a consent patch bolted onto each of nine modules), it buys a compliance certificate. Spent architecturally, the same mandated rupees lay the consent spine of the twin, and the hospital exits its compliance programme holding an asset. The second consequence is practical: consider what happens when one patient exercises her erasure right against nine systems holding fragments of her under imperfectly matched identities. The marketing list, the analytics copies, the stale numbers – all must go, and a hospital that cannot find her everywhere cannot delete her anywhere. Fragmentation is about to change from an inconvenience into a liability. And the mirror test from the boardroom stops being idealism in May 2027 and becomes a statutory right with a regulator attached.

How to begin

A group can begin without a hero project. Seven moves, in order.

First, do not rip and replace. The nine modules stay – they are good at their workflows, and the CFO has already paid for them. The twin begins as a layer above them, not a five-year migration that outlives its sponsors.

Second, solve identity. One patient, one identity, across every system – the genuinely hard problem in India, where families share phone numbers and names take many spellings. The ABHA number now gives every hospital a national anchor to build on. This step is unglamorous, and everything else depends on it.

Third, build consent in from the first day. Every properly captured consent is an asset; every record captured without one is, from May 2027, a liability. Starting the twin and starting DPDP compliance are the same project – budget them as one.

Fourth, capture events first, then reach back for history. The twin grows by listening. Let each system publish its events (an appointment, a result, an admission, a call) into the central record as they happen, so the asset compounds from the first month. The archives can wait – and then they should not wait forever. Once identity and consent are solved for a returning patient, pull her old records into the twin as well: the decade of laboratory results and discharge summaries is exactly the longitudinal depth the asset exists to hold. Done patient by patient, at each return visit, the back-population needs no big-bang migration, and every old record arrives with fresh consent attached.

Fifth, prove value with one lighthouse. Choose something a patient feels and a CFO can measure within a year: discharge follow-up for cardiac patients, or a diabetes cohort that never repeats its history. One working demonstration defends the programme better than any strategy paper.

Sixth, give the asset an owner. Here, in my experience, is where such efforts live or die. As a CIO side-project the twin loses every quarterly battle to the ERP upgrade. It needs a named senior owner whose only job is the patient asset, and one metric the board reviews at every meeting – say, the share of active patients holding a complete, consented, longitudinal record. Boards get what they measure. I have sat through enough reviews on both sides of the table to promise that this single agenda line changes more than any platform selection.

Seventh, only then, the intelligence. Prediction, simulation, the twin you can question, the research partnerships – all real, all valuable, all later. The advantage begins with plumbing. AI is what the asset earns, not what it is.

None of these moves requires waiting for a vendor's roadmap or a ministry's circular. A group could take the first three decisions this quarter.

The chains have spent twenty-five years building towers, and India needed the towers. The next decade will be decided by something quieter: which institution builds a memory. The woman from the ambulance is still out there, in every catchment in the country – nine systems, nine fragments, one person. The next time she calls at two in the morning, the hospital that answers as if it knows her (allergy, angioplasty, unhappy email and all) is the hospital she will never leave. That is the fifth lever.

Dr. Ajay Bakshi is a healthcare leader who has led major hospital networks in India including Max, Manipal and Gleneagles. A neurosurgeon by training, he now advises hospital boards on strategy and works on AI in healthcare.

References

1. Max Healthcare Investor Presentation, May 26, 2026. Strategy slide (13), peer benchmarking on ARPOB/occupancy/ROCE/EBITDA per bed (11), bed-expansion plans (16).
2. Apollo Hospitals Q1 FY27 investor presentation coverage (Investing.com, August 2026): 5,798 beds through FY31, ~₹11,152 crore; Apollo 24|7 user counts.
3. Medanta (Global Health Ltd) Investor Presentation Q3 FY26 (BSE filing, February 2026): 110 doctors onboarded; bed and capex plans.
4. Rainbow Children's Medicare Q1 FY27 coverage (July 2026): 2,500-bed expansion, ~₹2,200 crore.
5. Fortis Healthcare Q1 FY27 earnings call coverage (August 2026): occupancy and clinical-talent quote.
6. Narayana Hrudayalaya Q3 FY26 earnings call transcript (BSE filing, February 2026): Viren Shetty on occupancy, bed-led goals, and total health spend.
7. PB Fintech's hospital venture raises $218 mn (Business Standard, May 2025); PB Health's entry (Business Today, July 2026).
8. Tata 1mg FY26 financials (Entrackr, 2026); Amazon Diagnostics launch (June 2025).
9. Kritzinger W, et al. Digital Twin in manufacturing: a categorical literature review and classification. IFAC-PapersOnLine 51(11), 2018. The model / shadow / twin taxonomy.
10. National Academies of Sciences, Engineering, and Medicine. Foundational Research Gaps and Future Directions for Digital Twins. National Academies Press, 2024.
11. Tudor BH, et al. A scoping review of human digital twins in healthcare applications and usage patterns. npj Digital Medicine, 2025. 18 of 149 studies (12.08%) met the NASEM criteria.
12. Shamanna P, et al. Digital Twin in Managing Hypertension Among People With Type 2 Diabetes: 1-Year Randomized Controlled Trial. JACC: Advances 3(9), 2024.
13. Pramesh CS, et al. National Cancer Grid initiative for electronic medical records, India. Bulletin of the World Health Organization, 2025.
14. Saleh H, et al. Doctors' perceptions on electronic medical records at a large private hospital in Ujjain, India. BMC Health Services Research, 2025.
15. EMR adoption: a strategic shift (CIO, Fortis) (Financial Express Futech, July 2025): OPD records digitised in 15 hospitals, IPD in one, 2024–25.
16. Apollo Hospitals Q3 FY26 earnings call transcript (February 2026). Checked in full on 14 August 2026: the platform is discussed in users and GMV; "patient record" does not appear.
17. Athma and Narayana Health (BW Healthcare World); AIRA launch (Business Standard, August 2025). Record counts are company statements.
18. ABDM Mission Steering Group figures, 10 July 2026 (Ministry of Health release): 93.95 crore ABHA numbers, 105 crore linked records; PIB backgrounder, 6 July 2026; 100-crore milestone, PIB, 22 May 2026.
19. Jaffe MG, et al. Improved blood pressure control associated with a large-scale hypertension program. JAMA 310(7), 2013. Kaiser Permanente Northern California, 43.6% to 80.4%.
20. Koning NR, et al. Duplicate laboratory testing after referral. BJGP Open, 2022. 56% of GP-ordered tests repeated in Dutch outpatient referral.
21. Roche to acquire Flatiron Health (Roche media release, February 2018, USD 1.9 billion); FDA approval summary, palbociclib for male breast cancer (Clinical Cancer Research, 2020).
22. Dagan N, et al. BNT162b2 mRNA Covid-19 vaccine in a nationwide mass vaccination setting. NEJM, February 2021 (Clalit, 1.19 million matched individuals); Israel–Pfizer Real-World Epidemiological Evidence Collaboration Agreement, January 2021.
23. Reliance acquires Karkinos for ₹375 crore (Business Standard, December 2024); National Cancer Grid (~750,000 new cancer patients annually).
24. Digital Personal Data Protection Rules, 2025 (Gazette of India, G.S.R. 846(E), notified 13 November 2025) and the DPDP Act, 2023: phase-in schedule (Rule 1), consent burden of proof (s.6(10)), patient rights (ss.11–12), breach notification (Rule 7), penalty schedule (s.33). MeitY's January 2026 timeline-compression proposal remains ungazetted as of this writing.
25. Sadée C, et al. Medical digital twins: enabling precision medicine and medical artificial intelligence. The Lancet Digital Health 7(7), 2025.
26. Chrispin J, Prakosa A, et al. (Trayanova N, senior author). Digital Twin-Guided Ablation for Ventricular Tachycardia. New England Journal of Medicine 394(13), April 2026. Ten patients; mean follow-up 405 days.
27. Reinventing Modern Medicine: interview with Eric Topol. EMJ Innovations 10(1), 2026. PubMed counts: search "digital twin", run 14 August 2026 (3 papers dated 2018; 912 dated 2025).