Patient Identification in Hospitals: Duplicate Records, Impersonation and What Biometrics Changes
What is the first thing a hospital does with a patient? Before any diagnosis, someone at a registration desk has to decide which record belongs to this person. In most hospitals this decision is made by typing a name and a date of birth into a search field. When many patients share a few surnames, dates of birth are recorded approximately and a health card can be lent within a family, this search is wrong often enough to matter for both safety and money.
Three ways the record goes wrong
- The wrong record. Two patients with the same name, and the desk opens the one that comes first. The clinician then sees allergies, results and prescriptions of another person.
- A duplicate record. The desk does not find the patient, misspells the name or the patient gives a different date of birth, and a new record is created. The history is split, tests are repeated and the next clinician sees only half of it.
- A mixed record. A patient is treated under someone else's card. Two people's diagnoses and blood types end up in one record, and the insurer pays for the treatment of a person who is not covered.
The first two are mistakes, the third is often deliberate. All three come from the same root: the record is found by attributes that a person can share, forget or borrow.
How common it is
Even in health systems with good documents the numbers are high. In a study published in BMJ Quality & Safety, Allison McCoy and colleagues analyzed records of five US health organizations. Between 16% and 41% of records shared an exact first and last name with at least one other patient. Adding the date of birth reduced this to between 0.16% and 15%, still far from zero. The most common name at one site was shared by 2,552 records. The authors concluded that the rate of duplicate records may also be high and that better record management is needed to minimise the risk of harm.
A 2018 Black Book survey of 1,392 health technology managers found that respondents reported about 18% duplicate records in their organizations before they introduced patient matching tools, and respondents attributed a third of denied claims to inaccurate patient identification or information. Consider now a hospital where many patients have the same few surnames, where a date of birth is often recorded as the 1st of January of an estimated year, and where a national ID number is not always available. The share of collisions is likely higher.
The safety side has been known for a long time. In 2007 the WHO Collaborating Centre for Patient Safety Solutions, run by The Joint Commission, included patient identification in its nine patient safety solutions, noting that failures to identify patients correctly lead to medication, transfusion and testing errors, wrong-person procedures and even babies discharged to the wrong families.
When identification is also a fraud problem: India
India's Ayushman Bharat PM-JAY covers about 55 crore people with up to Rs 5 lakh per family per year for hospital treatment. Beneficiaries are verified with Aadhaar e-KYC when the card is issued and authenticated again when they receive treatment. By March 2025, according to the government's reply to the Rajya Sabha, states had rejected 3.56 lakh claims worth Rs 643 crore for fraud. By May 2026, according to a reply to the Lok Sabha, 2,359 hospitals had been de-empanelled and 1,200 suspended. Among the patterns the scheme's systems flag are ghost billing and beneficiary impersonation.
The Indian case also shows the other side. Critics point to India's food ration system, where Aadhaar authentication failures cut off genuine beneficiaries, and warn that the same can happen to patients. This is a real risk of any biometric gate in healthcare, and it is the reason to plan for exceptions from the start: a second modality, an emergency override with a later review and devices that do not depend on a stable connection.
What changes with biometrics at the registration desk
When the patient is identified by iris or face, the search by name disappears. The registration desk scans the patient and the system returns the existing record, or confirms that this person has no record yet and creates exactly one. This way:
- duplicates stop growing, and the old ones surface when the same person is found under two records;
- the wrong record does not open for a namesake;
- a borrowed card does not work, because the card no longer identifies the patient;
- the same check can repeat before a transfusion, a medication round or a sample collection, where the cost of an error is highest;
- a claim carries proof that the patient was in the facility, which we described for insurers in the article on Ghana and Kenya.
We recommend iris as the primary modality in hospitals. It is contactless, which matters for infection control, works for the elderly and for manual workers whose fingerprints are worn, and separates people well enough to search a regional or national patient register. Face complements it for patients who cannot keep their eyes open for the scanner.
What it does not solve
Biometrics identifies the person. It does not correct the record that is already wrong: merging existing duplicates still needs clinical review. It cannot be used for every patient, so emergency admissions need an override procedure. And it adds a new kind of sensitive data to the hospital. Templates should be stored apart from medical records, linked through a pseudonymous identifier, encrypted and covered by the hospital's data protection policy. Enrollment of children needs particular care and clear consent rules.
Where to start
The cheapest first step is to measure the problem. Enroll returning patients at the registration desks of one facility for a few weeks, and count how many of them already have more than one record. This number is a stronger argument for management than any presentation, and it gives the baseline for a wider rollout.
Sources: McCoy et al., Matching identifiers in electronic health records: implications for duplicate records and patient safety, BMJ Quality & Safety; Healthcare Finance News, Black Book patient matching survey, 2018; WHO Collaborating Centre for Patient Safety Solutions, nine patient safety solutions; PTI, 3.56 lakh claims worth Rs 643 cr rejected for frauds under Ayushman Bharat scheme, 2025; Outlook Money, 2,359 hospitals de-empanelled, 1,200 suspended (Lok Sabha reply), 2026; MediaNama, Ayushman Bharat fraud and Aadhaar authentication, 2025.