Chief Data Officer / India
Three decisions that cost something to get right
- Annual revenue
- ₹1,480 Cr
- Year ended March 2026
- Beds across eight hospitals
- 2,050
- Kerala and Karnataka
- Patients in the group index
- 52 lakh
- After merge review; September 2026
- Colleagues in the data office
- 92
- Coding, records, engineering, analytics and governance
Decisions & consequences
01Delaying a group patient index so that uncertain matches were reviewed, not scored
In 2022 we were replacing separate registration and records systems at seven hospitals with one hospital information system, which needed a single patient index. Our registrations held 46 lakh patient identifiers, and the vendor's matching software flagged 7.9 lakh probable duplicates. Most were straightforward. About 61,000 pairs were not: the same name and date of birth with a different house name or phone number, or a shared family phone with a similar name.
The vendor proposed merging every pair above a set score. The chief operating officer supported this, because go-live was timed for the opening of our Mysuru hospital in early 2023. I asked medical-records staff to check a sample of 2,000 pairs from the band the vendor intended to merge automatically. Sixty-two of them, 3.1 per cent, were different people. A wrong merge could place one patient's allergy, blood group or medication history in another person's chart.
I recommended manual review of the whole uncertain band. The chief executive agreed after the medical director supported the clinical concern. The price was a fourteen-week delay to the unified record at three hospitals and about ₹2.4 crore in extended vendor support and temporary reviewers. Mysuru opened on the old system and migrated later, which meant duplicate work for its registration and billing teams. The operations team's frustration was justified.
Reviewers merged 38,400 pairs and kept 22,600 apart. In the first year after go-live, eleven wrong merges were reported, all from the band above the threshold, and each was reversed under a documented unmerge procedure. Duplicate registrations fell from 41 to 9 per thousand new visits once we standardised how desks record house names and mobile numbers.
The review queue is now a permanent function with six staff. What I would do differently is run the sample audit before the go-live date was agreed with the vendor. Had that evidence existed earlier, the Mysuru opening could have been planned around it rather than disrupted by it.
Keeping enrolment in national health accounts a patient's choice, at the cost of an incentive
In 2024 we connected our hospital information system to the Ayushman Bharat Digital Mission, so that patients could link their records to a health account number and share them with other providers by consent. A government incentive scheme paid facilities according to the volume of eligible digital record transactions. Our finance team projected ₹1.9 crore in the first year if most new registrations included creating or linking an account.
Operations proposed making account creation a default step at the registration desk, with desk staff measured on linkage rates. I disagreed. Participation in the mission is voluntary, and a step presented at a busy desk as part of registration would not feel voluntary to many patients, particularly elderly people or anyone arriving in an emergency. The chief financial officer argued that the incentive would fund part of the integration cost and that patients could withdraw later.
The chief executive accepted my alternative: a separate, explained step after registration, a printed notice in Malayalam, Kannada and English, and a desk script that made refusal easy. Staff were given no linkage quotas. We gave up most of the projected income and added an average of forty seconds to each registration that included linking, which mattered at the morning outpatient peak.
In the first year 23 per cent of new registrations were linked, against the 70 per cent planned, and we received ₹52 lakh of incentive. We recorded no complaint alleging that a patient had been enrolled without understanding it. Doctors retrieved 14,200 records from other providers with patient consent, mostly for referred cardiac and kidney patients, which avoided some repeat tests.
What changed was the board's understanding that consent design has a price. I would do one thing differently. I allowed the ₹1.9 crore projection to appear in a board paper before we had piloted the process at one hospital. That figure became the benchmark against which a sound decision looked like a shortfall, and I spent two committee meetings explaining a number I should never have let travel alone.
Reducing insurance deductions by correcting the clinical record rather than the bill
By 2025, cashless insurance and government-scheme patients accounted for about 41 per cent of our revenue. Across ₹608 crore of insured billing, insurers and third-party administrators deducted 6.8 per cent, about ₹41 crore a year, through partial approvals and disallowed items. The revenue-cycle head proposed thirty additional coders and a claims-editing tool that would recode bills before submission.
I asked for an analysis first. My team reviewed 18,000 claims with deductions from the previous year. Forty-four per cent of the deducted value traced to documentation: intensive-care days without a recorded reason, implants without matching invoices and labels, and stays extended without a request to the insurer for an enhanced approval. Only a small share came from coding errors. I was also concerned that a tool designed to maximise a bill could drift into describing more than the record supported.
The decision was contested. The revenue-cycle head wanted a quicker result, and several senior consultants objected to new mandatory fields in an already long discharge process. The medical director negotiated a compromise: four structured fields for intensive-care justification and implants, and an alert to the insurance desk whenever a stay passed its approved length. We postponed the tool and hired eight coders rather than thirty. We also stopped appealing deductions our own audit judged correct, writing off ₹3.2 crore of low-merit appeals instead of pursuing them.
Over the following twelve months, deductions on comparable admissions fell to 4.9 per cent of billed value, about ₹11.6 crore a year. Average settlement time fell from 34 to 27 days. Part of the improvement reflected changes in insurer practice, and I said so in the audit committee paper rather than claim all of it.
Deductions are now reported by department and by cause, and department heads review them monthly. If I did this again, I would bring nursing leadership in from the start. Much of the intensive-care justification already existed in nursing notes, and we spent two months designing physician fields before discovering it.