By Scott Weiner, MD, MPH, Brigham and Women’s Hospital
The word “stigma” has roots in both Latin and Greek, meaning a mark on the skin usually from a pointed instrument, tattoo or burning. In Shakespearean times, stigma became used more figurately, to indicate a mark of disgrace attached to a person on account of evil conduct (1). People who use drugs are frequently subjected to stigma, particularly in our society which equates substance use disorder as a conscious choice and moral failing. In more recent years, we have learned that this is simply not true, and reasons for drug use are complex and not the “fault” of the individual. Still, the persisting stigma of people who use drugs can be extremely detrimental towards these individuals, especially when the stigma comes from medical professionals.
Even just reading a vignette in which a patient is referred to as “a substance abuser” vs. “having a substance use disorder” can change impressions of medical professionals about the patient (2). For this reason, the National Institute on Drug Abuse (NIDA) created an initiative called “Words Matter”, with a list of terms to use and avoid when talking about addiction (3). Some of the words are not surprising, such as the heavily stigmatizing “junkie” or “abuser”. Other words are less obvious. For example, referring to a urine drug screen result as “dirty” may make the individual feel that the word is being applied to them and it is better to just use factual language like “the test was positive for oxycodone”. Other words like “alcoholic” and “addict” may be used by patients themselves in the context of recovery (e.g. Alcoholics Anonymous) but really have no place in the medical environment where person-first language is not only preferred, but vital to approaching care from an anti-stigmatizing perspective.
Our research team was curious to investigate how frequently stigmatizing language appeared in our medical records. We are the largest heath system in Massachusetts and have 11 hospitals including two large academic centers. Since other clinicians read notes and respond differently based upon presence of stigmatizing language, and patients can now see many of their medical records through the 21st Center Cures Act, we thought it would be important to highlight the presence of these words in our charts for the awareness of our clinical staff.
Looking for stigmatizing language is not simple because it depends on the context that the word is use in. A sentence stating, “the patient’s room is dirty” is not the same as saying “their urine is dirty”. Likewise, it is acceptable to refer to a disease as “alcoholic cirrhosis” or “alcoholic hepatitis” but not “the patient is an alcoholic”. Therefore, in our study recently published in the Journal of Addiction Medicine, we relied on a technique called “Natural Language Processing (NLP)” to answer the question about the incidence of stigmatizing language in our medical records (4). NLP evaluates each sentence and then places the language in context. NLP is a type of artificial intelligence.
As a first step, we identified patients seen in our system over a one-year period who had a diagnosis code indicating a substance use disorder (including both alcohol and/or drugs). We applied an NLP algorithm that we created to 10,000 clinical notes. We then reviewed about 200 notes that were determined to be stigmatizing and 200 that were not. We further refined our algorithm based on this and applied it to over 500,000 clinical notes.
The results were surprising to us. For over 30,000 patients with at least one diagnosis of a substance use disorder, 18% of notes had stigmatizing language and about 62% of patients had at least one note with stigma. By far, the most common word used was “abuse”. This is probably less stigmatizing than other words, especially since there are numerous federal agencies that still use the term (including NIDA itself). However, we also observed several instances of “abuser”, “addict”, “junkie”, “drunk” and other such words. Notes most likely to obtain these words were the history and physical notes that are usually written upon a patient’s admission to the hospital. We also noted that physician assistants were about twice as likely to use stigmatizing language than physicians or nurse practitioners, a finding we do not have a good explanation for but may indicate the need for education.
Finally, we were curious to see if there were patient characteristics that were most likely to be associated with having stigmatizing language. Sure enough, male patients (adjusted odds ratio (aOR) 1.17), middle aged patients (45-54 years, aOR 1, with all other groups <1) and those with Medicaid (aOR 1.41) or Medicare insurance (aOR 1.23) were more likely. We saw no difference between patients who were White or Black, but Asian patients were much less likely (aOR 0.45) to have stigmatizing language, perhaps reflecting preconceived notions about substance use between these racial groups. Patients of Hispanic ethnicity also had lower odds (aOR 0.88) compared with non-Hispanic patients.
The point of our paper was simply awareness. We do not believe that any use of stigmatizing language that we detected was malicious or purposeful, but simply these are words that are in the vernacular and clinicians may not be aware of how their use may negatively affect our patients. A next step for our team is to determine if use of these words are associated with actual clinical outcomes.