Black and Violet

Electronic Health Records Coded with Bias

If a patient is Black, health providers are more than twice as likely to put negative words in that patient’s health history. These are descriptors like hysterical, noncompliant, unpleasant, or uncooperative. Those word choices don’t suggest a good relationship with a patient. This conclusion comes from an analysis of records for 18,459 patients, published recently in Health Affairs. Their electronic health records (EHRs) seem to be coded with bias.

Likewise, a new study in JAMA Network Open also shows a negative skew in the language used for Black patients in their EHRs. But that’s not all. It also shows a bias for using stigmatizing language to describe patients living with type 2 diabetes.

Words really do matter, because they help us give form to our thoughts. They also reflect the thinking we bring into our relationships with other persons.

Labeling and Blaming

These studies cannot possibly tell us what lies in the hearts of providers who inject bias into health records for vulnerable populations. In each study, researchers have found correlations. For whatever reason, negative words find their way into descriptors of Black patients more often than White patients. If the diagnosis is type 2 diabetes, providers more often describe them as non-adherent or a failure. No doubt these biases are implicit. They likely don’t reflect an intent to diminish a patient.

Nonetheless, the potential for harm is very real. Dean Schillinger, Director of the Center for Vulnerable Populations at San Francisco General Hospital, describes it:

“In medicine, we tend to label people in derogatory ways when we don’t truly ‘see’ them – when we don’t know them or understand them. The process of labeling provides a convenient shortcut that leads some physicians to blame the patient for their illnesses.”

A Subtle Breach of Ethics

Health providers have ethical obligation to respect patient autonomy, to benefit patients, and do them no harm. In a subtle way, the simple act of labeling a patient as non-compliant violates all three of those obligations. Tammy Sinkfield-Morey reminds us that a provider can never be sure their plans for a patient’s care will turn out to be the right choice. But a good provider can see opportunity when a patient wants something different. That “non-compliant” label presumes a patient is at fault, when the truth might be that the care is not adequate for a patient’s needs.

The fact that patient records seem coded with bias does not mean we need to be pointing fingers at bad providers or bad patients. Rather, it signals a need to work harder on recognizing our implicit biases and moving beyond them.

Click here for the study in Health Affairs and here for the study in JAMA Network Open. For further perspective, click here and here.

Black and Violet, painting by Wassily Kandinsky / WikiArt

Subscribe by email to follow the accumulating evidence and observations that shape our view of health, obesity, and policy.


February 17, 2022