AI Analysis Reveals Racial Bias in Medical Records: Black Patients More Likely to Face Credibility Doubts

Study Reveals Racial Bias in Medical Record Documentation

A groundbreaking study published in the open-access journal PLOS One has uncovered a concerning pattern of racial bias in how clinicians document patient credibility in electronic health records (EHRs). Led by Mary Catherine Beach of Johns Hopkins University, the research analyzed over 13 million clinical notes from a large health system in the mid-Atlantic United States 1

1

2

Source: News-Medical

AI-Powered Analysis of Medical Records

The study employed artificial intelligence (AI) tools to examine 13,065,081 EHR notes written between 2016 and 2023. These notes pertained to 1,537,587 patients and were authored by 12,027 clinicians. The AI was programmed to identify language that suggested clinicians doubted the sincerity or narrative competence of patients 1

1

2

Key Findings: Racial Disparities in Patient Credibility

While less than 1% (0.82%) of the medical notes contained language undermining patient credibility, a clear racial disparity emerged:

1. Non-Hispanic Black patients were more likely to have notes that undermined their credibility (adjusted odds ratio [aOR] 1.29, 95% CI 1.27-1.32) compared to White patients.
2. Notes about Black patients were more likely to question their sincerity (aOR 1.16; 95% CI 1.14-1.19) and competence (aOR 1.50; 95% CI 1.47-1.54).
3. Clinicians were less likely to use language supporting the credibility of Black patients (aOR 0.82; 95% CI 0.79-0.85) compared to White or Asian patients 1

   1

   2

   2

   .

Source: Medical Xpress

Implications for Healthcare Disparities

The researchers suggest that this pattern of documentation could contribute to ongoing racial disparities in healthcare. Dr. Beach and her colleagues emphasize that these findings likely represent "the tip of an iceberg" in terms of unconscious bias in medical practice 1

1

2

Related Stories

Researchers Call for Standardization of Race and Ethnicity Data in Medical AI to Address Bias

AI in Medicine: Study Reveals Socioeconomic Bias in Treatment Recommendations

AI Uncovers Hidden Bias in Amputation Rates Among Minority and Rural Patients

AI Medical Advice Skewed by Typos, Slang, and Gender Bias, MIT Study Reveals

Limitations and Future Directions

The study acknowledges several limitations, including its focus on a single health system and the inability to examine clinician characteristics such as race, age, or gender. Additionally, the AI models used, while highly accurate, may have misclassified some notes 1

1

2

Recommendations for Addressing Bias

To combat this issue, the researchers propose two key strategies:

1. Enhance medical training to increase awareness of unconscious biases among future clinicians.
2. Develop AI tools for writing medical notes that are programmed to avoid biased language 1

   1

   2

   2

   .

Dr. Beach stated, "For years, many patients - particularly Black patients - have felt their concerns were dismissed by health professionals. By isolating words and phrases suggesting that a patient may not be believed or taken seriously, we hope to raise awareness of this type of credibility bias with the ultimate goal of eliminating it" 1

1

2

This study highlights the potential of AI in uncovering and addressing systemic biases in healthcare, paving the way for more equitable and unbiased medical practices in the future.