I recently read a solid research article about healthcare bias called Examination of Stigmatizing Language in the Electronic Health Record published in JAMA Network Open.
The article concluded that “findings suggest that stigmatizing language appears in patients’ EHR admission notes, varies by medical condition, and is more often used to describe non-Hispanic Black than non-Hispanic White patients. Therefore, efforts to understand and minimize the use of stigmatizing language might improve patients’ care and their trust in their clinicians.”
I agree with this conclusion, and I think this adds valuable information to discussions around bias within healthcare.
Yet one thing that concerns me about this study, and nearly every study that uses large EHR data sets to gain “insight,” (in academia or Corporate America) is the decision not to get the patients’ informed consent.
Consider this study which uses artificial intelligence (via natural language processing) to analyze the chart notes of 48,651 admission notes. Here’s how they justify doing this study without getting informed consent:
“The institutional review board (IRB) at Princeton University ceded review of this study to the IRB at Mass General Brigham, which approved it. Informed consent was waived because patient data were deidentified.” (highlight mine.)
I am concerned about the morality of large institutions (whether it’s within academia or Corporate America) waiving informed consent based on deidentifying patients’ data.
I understand the practical reasons they decided to do this study without consent.
First, it would be difficult and expensive to get informed consent from 48,000 patients.
Second, if they did get consent on a set of 48,000 patients, it would bias the results, making the study results suspect.
I also understand the typical ethical arguments supporting EHR studies without informed consent. It goes something like this: If we deidentify the patients, then analyzing their charts won’t do them any specific harm, and it potentially will help many future patients.
This is the Utilitarian Ethical argument: the greatest good for the greatest number.
But I don’t think Utilitarianism is the last word about the rightness of deidentification.
For argument’s sake, and using the philosophical Principle of Charity, let’s just assume that the process of deidentification is perfect and the results of this study improve the medical care for millions of people in the future.
Even then, it’s unlikely this specific study, which was done 3 years after the patients were seen, will directly help that patient for what they chose to be seen for when they went into their office visit.
Simply put, patients—without their consent—were treated as a means to an end rather than an end in themselves.
In a deontological ethical system, this is unethical. The academic institutions (and, in many cases, corporate institutions) are deeply invested in the ability to mine patients’ health data. Most of their EHR research programs would shut down without it. So, how can their ethics committees, which are the ones deciding to waive the patients’ rights to informed consent, possibly be unbiased?
Also, there is the question of harm. These types of studies are academia and Corporate America’s bread and butter, and as they require reducing huge numbers of patients to data sets—they are necessarily dehumanizing. I have a hard time seeing how dehumanizing patients—people—by the thousands ( or in cases of corporate healthcare, the millions) doesn’t cause some harm.