Research and data

Patient identification: A call to arms (or ears?)

patient identification

In 2009, the Obama administration passed the Health Information Technology for Economic and Clinical Health (HITECH) Act, forever changing how healthcare is documented, coordinated, and reported on in the United States. This act put the transition from paper-based records to electronic health records (EHRs) front and center. After its passage, the adoption rate of EHR systems took off, jumping from 3.2% in 2008 to 86% in 2017.

This transition came with promises of improved efficiency, improved quality of care, and interoperability of the patient record. Promises that by many standards have gone unmet. While frustrations differ across hospital systems and across users, a National Physician Poll conducted by Stanford Medicine in 2018 found that 99% of physicians agreed the most important function of an EHR is to maintain a high-quality record of patient data over time. This same poll found that one in four primary care physicians do not believe their EHRs meet this function.

Why do EHRs struggle to maintain a comprehensive patient record over time? It all roots back to patient identification—how records are stored and matched to the correct patient.

In EHR systems, patient records are stored in a searchable database, known as the Master Patient Index. Patient identification relies on patient-provided identifiers such as first name, last name, and date of birth. On admission, registration staff use these identifiers to search the Master Patient Index. Misspelled names, nicknames used, incorrectly entered dates of birth, or any combination, provide many opportunities for patient identification errors.

The most common error occurs when a person is unable to be found and linked to their existing patient record. When this happens, a duplicate record is created that contains incomplete and missing patient data. When Master Patient Indexes become saturated with duplicate patient records, quality of care, efficiency, and downstream reporting is negatively impacted. A recent AHIMA study reported that hospital systems have an average duplicate patient record rate of 20%.

What are the effects of patient identification errors?

Quality of Care

An incomplete patient record (with missing lab test results, vitals, medications, allergies, and provider notes) carries significant patient safety risks that run the gambit. Missed imaging results may lead to repeat imaging which carries associated costs, radiation exposure, and delays in treatment, while a missed allergy could lead to dangerous, life-threatening prescriptions.


Hospitals incur administrative costs for cleaning duplicate records from the Master Patient Index, while costs resulting from duplicate or unnecessary tests, labs, and medications land on the patient. An audit of Children’s Medical Center in Texas found that 22% of patient records stored in the hospital’s Master Patient Index were duplicates. In this case, the facility hired outside help to clean the database, removing or merging a total of 250,000 duplicate patient medical records. Cost estimates from the organization were an average of $96/record in administrative costs. In cases where duplicate records could be traced to duplicate tests, labs and/or medications, unnecessary care costs averaged $1,100 per patient.

Hospital Reporting

Metrics reported at the hospital level are often used to justify process, policy, and funding decisions. Because many of these metrics are measured using aggregate data directly from text fields in patient-level records, patient identification errors can cause over- and under-reporting that affect how hospital resources are allocated. The metric of patient readmission rate, for example, is often used to determine the quality of care. When a duplicate record is used on readmission, the patient is not counted toward the patient readmission rate.

Population Health And Aggregate Data

Aggregate data at the population level is used to inform funding decisions and resource allocation, to quantify vaccination coverage and burden of disease in populations, to inform and prioritize research focus areas, and to coordinate targeted health interventions. When we consider that this data is contaminated by information stored in duplicate records, the downstream effects of patient identification errors are made painfully clear.

Biometrics as a solution for more accurate patient identification?

Patient identification errors present an undeniable barrier to EHRs achieving their most important function: maintaining high-quality patient data over time. Recent proposals around how to overcome this barrier include the use of biometrics for identification. In contrast to other proposals like a universal patient identifier, biometrics cannot be lost, forgotten, or stolen, presenting a reliable option for patient identification.

A major issue raised around biometrics, however, concerns their reliability, especially in infant populations. While iris scanning, facial recognition, and fingerprint scanning have repeatedly proven unreliable in infants (5 years and under), ear biometric identification holds hope.

To date, identification using an image of the ear has proven reliable in adult populations. Little has been documented, however, about how ear growth affects patient identification in early life. Our group at BUSPH, in collaboration with our team in Zambia (Project SEARCH), is currently studying this. In the summer of 2020, we completed data collection from a longitudinal cohort study of 224 infants who were followed at monthly intervals from 6 days through 9 months. Data analysis from the growth study is underway (stay tuned!), and will provide important insight on the feasibility of using ear biometrics as a method for patient identification.

Lauren Etter is a Global Health Research Fellow at Boston University School of Public Health. She manages Project SEARCH (Scanning EARs for Child Health) under Dr. Christopher Gill. The Project SEARCH team is a collaboration of computer science, public health, and engineering expertise from Boston University and the University of Zambia (UNZA). Their current research is focused on developing a mobile application to use ear biometrics for patient identification in the Zambian healthcare setting. To read more about our work:

New call-to-action

You may also like

data analysis
In the news
Why Health Data Is Good Infrastructure

We need to strengthen and build robust health data utilities that can link and connect public health and clinical data systems. It makes sense to build these data utilities as public-private…

mental health and physical health are intrinsically tied together
Research and data
Mental health and physical health cannot be separated

The mind and body are deeply intertwined, and mental wellbeing is crucial for physical wellbeing. People with mental illness, such as depression, are at increased risk of chronic disease, such…

How machine learning helps identify people at high risk of suicide after discharge
Research and data
How Machine Learning Helps Predict Suicide After Discharge from Psychiatric Hospitalization

Suicide is a major public health problem, with approximately 800,000 suicide deaths globally each year. Thus, it’s crucial to accurately predict who will engage in suicidal behavior so that they…