Unveiling Inequities

Decoding bias and disparities in diagnostic coding practices.

How discrepancies in ICD-10 coding illuminate and perpetuate systemic racism.

Report authored by Violet’s Clinical Research and Data Science teams.

Preface

At Violet, we have a strong belief that the systems we operate within are not built equitably for all people. This is particularly evident in health care, shown in the many health disparities that affect BIPOC and LGBTQIA+ communities, women, veterans, and more.

To begin combatting a broken system, we created a standard measure of provider cultural competence called Violet Benchmarks—a measure that allows for a shared, objective understanding of quality of care, particularly for BIPOC and LGBTQIA+ patients.

Gaurang Choksi

Gaurang Choksi

Founder & CEO at Violet

About

This report emerged from our standard analyses and due diligence review as we refine our benchmarking framework. Upon examining hundreds of thousands of medical claims, we identified and were struck by the stark disparities apparent in diagnostic and coding practices.

Through this series of investigations, we delve deeper into how these coding practices contribute to the escalating health disparities faced by BIPOC communities.

Analysis.

This report was developed through rigorous analysis of claims data, where community models were created to approximate demographic data for provider patient panels through geocoding. Led by senior data scientist Annie McKinnon and head of clinical and research Kay Nikiforova, the team aimed to validate community demographics models and uncovered disparities in diagnosis and coding percentages, particularly impacting BIPOC patients. These health disparity findings served as a tertiary validation of the models and sheds light on biased coding practices within the health care system.

Annie McKinnon

Annie McKinnon

Senior Data Scientist

Kay Nikiforova

Kay Nikiforova

Head of Clinical and Research

Chapter I

Investigating diagnostic coding discrepancies to uncover health inequities for BIPOC patients.

Chapter One Cover
Chapter I

In our analysis of prevalent diagnosis codes in the Black population, we uncovered some patterns which point to coding biases with ICD-10 code use.

While the prominence of diagnostic codes for known health disparities such as sickle cell anemia for Black patients isn't unexpected, the disproportionate presence of codes like Z7142 (Counseling for family member of alcoholic), Z7152 (Counseling for family member of drug abuser), and R461 (Bizarre personal appearance) was not expected, and points to the wider issue of systemic racism.

These findings strongly imply biases entrenched within both our medical system and broader societal structures.

United States Patient Population

16.1%

Black

8.6%

AAPI

(Asian American & Pacific Islander)

61.2%

White

Examining health disparities.

Despite making up a significantly smaller number of the overall patient population, we found common instances where Black and AAPI patients had double or nearly triple the percentage of certain diagnoses, compared to white patients.

2-3x

ICD-10 and CPT codes overrepresented in Black patients.

Slide or swipe right for more.

O759

Complication of labor and delivery.

% of national population based on US census

Percentage of make up of ICD-10 code: Complication of labor and delivery, unspecified

16.1%
61.2%
Black

26.9%

White

39.8%

O2690

Pregnancy-related conditions.

% of national population based on US census

Percentage of make up of ICD-10 code: Pregnancy related conditions, unspecified, unspecified trimester

16.1%
61.2%
Black

24.9%

White

49.3%

I256

Silent myocardial ischemia.

% of national population based on US census

Percentage of make up of ICD-10 code: Silent myocardial ischemia

16.1%
61.2%
Black

25.9%

White

50.1%

D5720

Sickle-cell/Hb-C disease.

% of national population based on US census

Percentage of make up of ICD-10 code: Sickle-cell/Hb-C disease without crisis.

16.1%
61.2%
Black

26.6%

White

43.9%

This obvious disproportionality drives home the stark statistics already seen in health disparities such as the Black maternal health crisis, sickle cell disease and related conditions, and racial disparities surrounding coronary artery disease.

In each of these cases, it is not a person’s race that is a determinant of their health in any biological sense, it is racism—structural, institutional, and interpersonal—that is the true cause of poor outcomes.

Coding discrepancies that indicate bias toward Black patients.

After looking at the health disparities present in the data, it also became apparent that there were stigmatizing codes attributed to Black and AAPI patients at significantly higher rates than white patients.

ICD-10 and CPT coding discrepancies for Black patients.

Slide or swipe right for more.

R461

Bizarre personal appearance.

% of national population based on US census

Percentage of make up of ICD-10 code: Bizarre personal appearance.

16.1%
61.2%
Black

31.7%

White

30.6%

Z743

Need for continuous supervision.

% of national population based on US census

Percentage of make up of ICD-10 code: Need for continuous supervision.

16.1%
61.2%
Black

28.1%

White

46.9%

Z550

Illiteracy and low-level literacy.

% of national population based on US census

Percentage of make up of ICD-10 code: Illiteracy and low-level literacy.

16.1%
61.2%
Black

27.5%

White

47.9%

F8082

Social pragmatic communication disorder.

% of national population based on US census

Percentage of make up of ICD-10 code: Social pragmatic communication disorder.

16.1%
61.2%
Black

22.9%

White

53.4%

Looking closer

Why do BIPOC patients have these codes attributed to them at disproportionate rates?

Our hypothesis behind the use of codes like “bizarre personal appearance” and “need for continuous supervision” is due to doctors overpathologizing Black patients, or in other cases, attributing a negative bias they wouldn’t otherwise apply to white patients.

Literacy, language, and communication related coding discrepancies in the Black patient population can be also attributed to the bias of misattributing language differences as language disorders—particularly in Black children.

The numbers

Examining nationwide data, researchers found that each year, 14 percent of states overrepresent the number of Black children with speech and language impairments.

United States Map
14%

Coding discrepancies that indicate dismissive attitudes toward AAPI patients and Eastern medicine.

We found that the frequency of codes for “encounter for childbirth instruction” and “worries” significantly exceeds the AAPI patient population size, indicating bias and misconceptions around patient health literacy and symptom presentation.

Eastern medicine.

ICD-10 and CPT coding discrepancies for AAPI patients.

Slide or swipe right for more.

Z322

Encounter for childbirth instruction.

% of national population based on US census

Percentage of make up of ICD-10 code: Encounter for childbirth instruction.

8.6%
61.2%
AAPI

22.1%

White

46.9%

Z2809

Immunization not carried out because of other contraindication.

% of national population based on US census

Percentage of make up of ICD-10 code: Immunization not carried out because of other contraindication.

8.6%
61.2%
AAPI

13.8%

White

53.4%

A bias against non-allopathic and Eastern medicine.

For AAPI patients, the code for “encounter for childbirth instruction” was nearly triple the population size—but why is this code so high for AAPI patients in particular?

Our hypothesis is that clinicians are not trusting AAPI patients to do correct childbirth processes either because of the prevalence of non-allopathic medicine practices in AAPI cultures, or because clinicians are assuming these patients don’t know the “right” childbirth practices in terms of Western medicine.

For immunization contraindication, our hypothesis is that it’s patients using non-allopathic medicine, and the result of clinicians not having knowledge of or knowing how to work with traditional medicine when combined with Western allopathic medicine.

Stereotypes and health-related stigmas against AAPI patients.

The coding for “worries” was more than double the population size of AAPI patients. Since these encounters occurred in a primary care setting, we suspect that it’s related to the misconception that AAPI patients disproportionately somaticize their emotional symptoms. Some doctors believe that if a patient presents with symptoms that don’t seem to have a cause, it’s a result of worry. Alternatively, there’s a commonly held belief that AAPI patients won’t go to mental health providers and instead will go to a primary care provider for care.

R4582

Worries.

% of national population based on US census

Percentage of make up of ICD-10 code: Worries.

8.6%
61.2%
AAPI

18.4%

White

52.5%

Chapter II

Kinder Coding: A look into how white patients get more provider time and preferential ICD-10 codes.

Chapter One Cover
Chapter II

Now lets jump into the flip side of coding biases — into something we’ve nicknamed kinder coding — the ways in which white patients are given the benefit of the doubt in clinical encounters.

What is kinder coding?

Research indicates that, on average, doctors allocate more time to white patients. Our findings further reveal that clinicians not only spend more time but are also more inclined to provide additional context and consideration to the injuries and ailments of white patients.

70
Minutes

Average physician facetime per year with white patients.

52.4
Minutes

Average physician facetime per year with Black patients.

53
Minutes

Average physician facetime per year with Hispanic patients.

Reference: Gaffney, Adam et al. “Trends and Disparities in the Distribution of Outpatient Physicians' Annual Face Time with Patients, 1979-2018.” Journal of general internal medicine vol. 38,2 (2023): 434-441. doi:10.1007/s11606-022-07688-x

Looking at the data.

In the following chart you’ll see evidence that clinicians are taking time enough in their sessions to get a high level of specific information from white patients, while getting considerably less details from patients of other races.

ICD-10 and CPT coding "irregularities / inconsistencies / exceptions / phenomenon" for white patients.

Swipe to view full table.
Code
Code description
White patients
61.2%
Black patients
16.1%
American Indian patients
1.2%
AAPI patients
8.6%
Native Hawaiian patients
0.2%
Other race patients
11.4%
Z5689
Other problems related to employment
81.0%
8.7%
1.1%
4.0%
0.1%
4.2%
V8655XA
Driver of 3- or 4- wheeled all-terrain vehicle (ATV) injured in nontraffic accident, initial encounter
80.9%
8.1%
1.4%
2.9%
0.2%
4.5%
Y93C9
Activity, other involving computer technology and electronic devices
80.7%
8.8%
1.8%
2.3%
0.3%
6.1%
Y92007
Garden or yard of unspecified non-institutional (private) residence as the place of occurrence of the external cause
79.7%
8.4%
1.3%
3.4%
0.1%
5.2%
Y9365
Activity, lacrosse and field hockey
79.4%
7.9%
1.0%
4.5%
0.1%
4.6%
Y93K1
Activity, walking an animal
79.3%
8.4%
1.2%
3.9%
0.2%
4.7%
V8656XA
Driver of dirt bike or motor/cross bike injured in nontraffic accident, initial encounter
78.9%
8.3%
1.2%
3.7%
0.1%
5.3%
Y92008
Other place in unspecified non-institutional (private) residence as the place of occurrence of the external cause
78.3%
9.1%
1.3%
4.2%
0.1%
6.1%
W293XXA
Contact with powered garden and outdoor hand tools and machinery, initial encounter
78.1%
9.3%
1.2%
3.9%
0.2%
4.9%
Y93H2
Activity, gardening and landscaping
78.0%
9.0%
1.2%
4.1%
0.2%
4.9%

What’s in a code?

Most of the top codes that white patients are disproportionately coded with ICD-10 W and Y codes. W codes provide information about the circumstances of the injury or ailment (e.g. falls, bites, explosions, etc.) The code even gets specific enough to differentiate between Fell from chair, initial encounter (W07.XXXA) and Fall from chair, subsequent (W07.XXXD). Y codes also reference external causes, such as locations for the injury or ailment, and also have a high level of specificity for them. Y codes include details such as Activities involving caregiving (Y93.F) and Nursing home as the place of occurrence of the external cause (Y92.12).

What is so “kind” about this way of coding?

When a clinician uses codes to specify that an injury was caused due to “Activity, food preparation and clean up” in “Unspecified place in single-family (private) house” it increases the possibility that the provider takes the patient’s story at face value and puts the blame towards external circumstances.

Juxtapose that with fault-bearing codes such as “Inappropriate diet and eating habits” (Z724) and “Acculturation difficulty” (Z603), which BIPOC patients disproportionately make up, and the implications these coding discrepancies may have for care are quite alarming.

More context

Injury due to:

Y93.G1 Activity, food preparation and clean up
Y92.019 Unspecified place in single-family (private) house
VS

Assigning fault

Injury due to:

Z724 Inappropriate diet and eating habits
Z603 Acculturation difficulty

The importance of documenting initial patient encounters in an aware and sensitive manner.

18%

Only 18 percent of text in inpatient progress notes were originally manually input, with the majority being imported from prior documentation.

2.5x

A recent article in Health Affairs reviewed over 40k notes from EHRs and found that Black patients had over 2.5 times the odds of having at least one negative descriptor in the history and physical notes compared to white patients.

Another recent journal article found that Black patients were being coded for higher rates of noncompliance with Z-codes.

History taking and clinical notes are vital sources of information often referenced by other care providers. Negative descriptors documented in the admission history and physical are prone to being replicated in subsequent notes, thereby perpetuating and potentially magnifying biases. This underscores the importance for providers documenting the initial patient encounter to do so with awareness and sensitivity.

Chapter III

Investigating coding practices for social determinants of health.

Chapter One Cover
Chapter IIi

Social determinants of health (SDoH) relate to how individuals are born, live, age, learn, work, worship, and play, which affects their functioning, quality of life, health, and health risks. The CMS Framework for Health Equity includes collecting, analyzing, and reporting social determinants of health for patients. There has also been an increasing push for increased coding of social determinants of health for patients, including payments for SDoH risk assessments through CMS in 2024.

Social determinants of health coding can offer a valuable insight into the components of a patient’s life that may be affecting them and their health. If providers take time to document these factors in a patient’s medical record, there is a better opportunity to address their SDoH needs, including case management or offering resources for housing, food, employment, and more. Social determinants of health disparities disproportionately affect BIPOC patients through less access to infrastructural support and health resources. In our research, we aimed to understand if our data reflected these SDoH disparities.

Social determinants of health-related issues disproportionately affect Black patients.

Our findings detailed in the graph below show that Black patients disproportionately made up common Z codes for SDoH needs, compared to the average percentage of population based on the US Census race categories.

Z codes overrepresented in Black patients.

Swipe to view full table.
Code
Code description
White patients
61.2%
Black patients
16.1%
American Indian patients
1.2%
AAPI patients
8.6%
Native Hawaiian patients
0.2%
Other race patients
11.4%
Z550
Illiteracy and low-level literacy
47.9%
27.5%
1.1%
11.1%
0.2%
12.2%
Z021
Encounter for pre-employment examination
41.9%
25.8%
1.3%
11.3%
0.2%
19.3%
Z558
Other problems related to education and literacy
49.7%
25.2%
1.2%
9.9%
0.2%
13.7%
Z594
Lack of adequate food
46.5%
25%
1.2%
10.7%
0.3%
16.3%
Z597
Insufficient social insurance and welfare support
48.9%
24.9%
1.2%
10.2%
0.2%
14.3%
Z5941
Food insecurity
47.6%
24.5%
1.2%
10.7%
0.4%
15.6%
Z599
Problem related to housing and economic circumstances, unspecified
48.7%
24.1%
1.3%
9.9%
0.2%
15.7%
Z5901
Sheltered homelessness
46.1%
24%
1.4%
9.7%
0.3%
18.4%
Z560
Unemployment, unspecified
47.7%
23%
1.3%
9.2%
0.2%
18.4%
Z742
Need for assistance at home and no other household member able to render care
48.5%
22.2%
1.3%
11.2%
0.2%
16.6%

Investigating biased coding for social determinants of health.

Let’s turn our attention back to biased coding practices. Our data showed that white patients were repeatedly coded for employment needs while Black patients were disproportionately given codes like “low levels of literacy” or “education problems.”

Z5689

Other problems related to employment

% of national population based on US census

Percentage of make up of ICD-10 code: Other problems related to employment.

16.1%
61.2%
Black

8.7%

White

81%

When filtering by for codes that white patients make up the largest percentage of, the most frequent code is “Other problems related to employment” (Z5689). Going back to our kinder coding idea, we can hypothesize that providers are practicing the same coding style for white patients regarding this specific SDoH category—by coding for problems relating to employment, providers are offering more context to the patient’s issues. This places the blame on issues that are external to the patient, essentially giving them the benefit of the doubt.

Why does bias occur?

We’re not saying that all of these clinicians are being openly racist when they walk into the exam room in these coding instances. Bias in health care often stems from unconscious attitudes and stereotypes that influence decision-making processes. Moreover, with the growing amount of administrative burden, clinicians may be relying on prior assumptions with patients as they have to rush through appointments.

Bias happens when clinicians...

  • Are unaware of how factors such as race, gender, socioeconomic status, and body language can trigger biased impressions of a patient.
  • Don’t take time to understand the type of communication a patient uses (e.g., non-native English speakers, AAVE, etc.) or acknowledge that medical settings are stressful for patients who have been repeatedly mistreated, and as a result, they may have difficulty communicating.
  • Unintentionally give certain patients more facetime by getting more details and putting additional effort into documenting the encounter.
  • Don’t have adequate time for care and resort to using prior beliefs and assumptions to speed up appointments.

In summary

This report sheds light on some of the more subtle and insidious issues that hinder equitable access to care.

Chapter I

Chapter I

Revealed diagnostic coding biases disproportionately affecting BIPOC patients, exposing hidden systemic inequities within the health care system.

Chapter II

Chapter II

Exposed the phenomenon of "kinder coding," where provider biases favor white patients with the benefit of the doubt.

Chapter III

Chapter III

Examined coding practices related to social determinants of health, highlighting the disproportionate burden on BIPOC communities.

Actionable steps

By implementing these strategies and prioritizing health equity, we can work towards building an equitable health care system, inclusive of every patient’s race, sexual orientation, gender, ability, language, citizenship, and more.

  • Health care providers and staff should undergo comprehensive cultural competence education and training to recognize and mitigate biases.
  • Continuously collect and analyze data on patient demographics, diagnostic coding, and health care outcomes to identify and address disparities in real-time.
  • Increase representation of BIPOC individuals in leadership positions within health care organizations.
  • Support policy initiatives aimed at addressing systemic racism and inequities within the health care system, including reforms to reimbursement practices and standardized coding guidelines.
  • Engage with BIPOC communities to understand their unique health care challenges and preferences, and collaborate on initiatives to improve access to care and health outcomes.

Violet is the industry’s first provider cultural competence analytics and training platform, democratizing health care by providing transparency into the inclusivity of a provider, care delivery organization, or network. Beyond establishing the infrastructure for inclusive care, we supply health care organizations with clinical training tailored to improve health outcomes for BIPOC, LGBTQIA+, neurodivergent communities, and beyond.

Visit Violet