I, Prabal De, have no financial or non-financial competing interest with respect to this manuscript.
I investigate the association of perceived discrimination based both on race and other attributes such as age, gender, and insurance status on self-reported health access and health outcomes in a diverse and densely populated metropolitan area.
Restricted data from the 2016 round of the New York City Community Health Survey was used to create prevalence estimates for both racial and non-racial discrimination. Logistic regression models were used to estimate the association of these discrimination measures with health access and health outcome variables.
Among residents who perceived discrimination receiving health care during the previous year, 15% reported the reason behind such discrimination to race, while the rest chose other reasons. Among the non-race based categories, 34% reported the reason behind such discrimination to be insurance status, followed by other reasons (26.83%) and income (11.76%). Non-racial discrimination was significantly associated with the adjusted odds of not receiving care when needed (AOR = 6.96; CI: [5.00 9.70]), and seeking informal care (AOR = 2.24; CI: [1.13 4.48] respectively, after adjusting for insurance status, age, gender, marital status, race/ethnicity, nativity, and poverty. It was also associated with higher adjusted odds of reporting poor health (AOR = 2.49; CI: [1.65 3.75]) and being diagnosed with hypertension (AOR = 1.75; CI: [1.21 2.52]), and diabetes (AOR = 1.84; CI: [1.22 2.77]) respectively.
Perceived discrimination in health care exists in multiple forms. Non-racial discrimination was strongly associated with worse health access and outcomes, and such experiences may contribute to health disparities between different socioeconomic groups.
Disparities in both mental and physical health outcomes between various population groups have long been a concern among health researchers and policymakers [
However, patients in a medical setting can potentially perceive some institutional or interpersonal behaviors as discriminatory owing to several personal attributes, not just race/ethnicity. Subsequently, providers, researchers, and policymakers have come to understand that
Using a representative dataset from one of the most populous and diverse cities in the US, this research investigates whether individuals report experiencing perceived discrimination (henceforth, discrimination) while seeking health care not only due to their race/ethnicity, but also because of their other attributes such as age, gender, type of insurance, and immigration status, the latter group being termed collectively as
There are several barriers to producing empirical evidence on different discrimination types. First, self-reported discrimination data have either not been routinely collected or not made readily available. Not many surveys include discrimination questions, and even if they do, such items are restricted to race/ethnicity-related questions, such as the Behavioral Risk Factor Surveillance System (BRFSS) surveys (which has now stopped including the question). Specific discrimination modules in some surveys, such as the ones used in the current study, are not publicly available. Addressing this question is critical, because if an individual does not perceive discrimination for their race or ethnicity, but for their other attributes such as age, insurance status, or income, the impact of the latter on healthcare might still be significantly negative, and a narrower definition of discrimination based solely on race/ethnicity would confound the nature of such relationships.
The findings show that both racial and non-racial discrimination are associated negatively with health access and health outcomes. In particular, in several cases, the magnitude of association for non-racial discrimination is both larger in magnitude and statistically significant. New York has previously been described as an enormous 'city-region' [
All data are anonymized. The restricted variable was obtained as part of a Data Use Agreement with the NYC Dept. of Hygiene and Mental Health. Ethical approval was provided by the City University of New York (CUNY) Institutional Review Board (IRB, #2018–0473).
This study utilized individual-level data from the 2016 New York City Department of Health and Mental Hygiene (DOHMH) Community Health Survey (CHS), which is an annual, stratified random digit-dialed phone survey of approximately 10,000 adult New York City residents accessed on September 21, 2019 [
The question assessing discrimination in the 2016 CHS asked respondents, "Thinking of your experiences trying to get health care treatment in the past 12 months, have you felt you were hassled, made to feel inferior, or discriminated against for any reason?" Available response categories were 'Yes,' 'No,' 'Did not seek health care treatment in the past 12 months,' and 'Don't know.' Additionally, that year, the survey asked a series of follow up questions to respondents who answered 'yes.' They were asked, "What was the reason or reasons you felt discriminated against while trying to get health care treatment in the past 12 months:" and presented with the options Race/ethnicity or skin color, Age, Language, Disability, Bodyweight, insurance status or type, Income level, Religion, Sexual orientation, Gender, Gender identity, Immigration status, and Other reason. Since the primary purpose of this study is to underline the existence and potential role of non-racial discrimination in health care, a categorical variable is created to assess discrimination: no discrimination (base category) vs. racial discrimination vs. non-racial discrimination. Specifically, individuals answering 'No' to the above question is assigned No Discrimination (= 1); individuals citing the reason to be race/ethnicity or skin color is assigned Race-based discrimination (= 2), and individuals citing any other reason listed above (except race) for discrimination is assigned to Non-racial discrimination (= 3); individuals replying ‘Don't Know’ are treated as missing. Therefore, for the respondents in both categories, only the race-based category is considered to make the estimates for non-race based measures more conservative. There are only twenty individuals who responded 'yes' to both race-based and at least one of the non-race based discrimination questions. As a robustness check, I performed the same analysis on the sample that excluded them. The results are very similar, as presented in Table A1 in
Outcome variables include both health access and health status, including both physical and mental health status, as the discrimination can lead to differentials in access to health care, which may lead to disparities in actual health status [
Other variables used in the multivariable analysis include demographic, insurance, education, and income. Five race/ethnicity categories are available in the data—non-Hispanic White, non-Hispanic Black, Hispanic, Asian and Pacific Islanders, and Others. The income-to-poverty ratio in each respondent's household is classified as an income of less than 200% of the federal poverty level and between 200% and 599% of the federal poverty level, and greater than 600% of the federal poverty level. Education is dichotomized as a college graduate and above vs. non-graduate. Likewise, individuals are categorized as having been employed or not.
All statistical analyses use survey weights provided in the NYC CHS data to control for complex survey design. First, the (weighted) prevalence estimates of various categories of self-reported discrimination are calculated to describe the overall prevalence of such a phenomenon. Similarly, weighted averages and prevalence estimates for all the relevant variables and categories are also calculated. Next, to assess the association between discrimination and health access and outcomes, two sets of multivariable logistic regression models are estimated with both racial and non-racial discrimination as predictors of interest (no discrimination being the reference group). In the first model, the primary outcomes of interest correspond to health care access, where the dependent variables are
Approximately 6% of New York City adults reported experiencing hassles, being made to feel inferior, or being discriminated against for
Among individuals who experienced discrimination in the past 12 months, 30.62% reported insurance status to be the underlying reason for such discrimination—the highest among all the categories. The next highest category is 'other,' presumably due to the fact that all reasons were not mentioned in the questionnaire (
(A) Trying to get health care treatment in the past 12 months, have you felt you were hassled, made to feel inferior, or discriminated against for any reason? (B) If experienced discrimination, what was the reason?.
Variable | Percent |
---|---|
White Non-Hispanic | 35% |
Black Non-Hispanic | 22 |
Hispanic | 27 |
Asian/PI Non-Hispanic | 13 |
Other Non-Hispanic | 02 |
Private | 47 |
Medicare | 16 |
Medicaid | 24 |
Others | 03 |
Uninsured | 11 |
18-24yrs | 13 |
25–44 yrs | 40 |
45–64 yrs | 32 |
65+ yrs | 15 |
<100% FPL | 26 |
100 - <200% FPL | 22 |
200 - <400% FPL | 17 |
400 - <600% FPL | 16 |
>600% FPL | 19 |
46 | |
42 | |
60 | |
34 | |
52 | |
35 |
Rows report weighted prevalence rates (expressed in %). FPL: Federal Poverty Level.
SOURCE. Author's analysis of New York City Community Health Survey data for 2016.
(1) | (2) | |
---|---|---|
1 if Did not get needed care | 1 if Medical advice from informal source | |
Discrimination based on race/ethnicity | 6.97 |
2.15 |
[4.15–11.70] | [0.85–5.43] | |
Discrimination based on other categories | 6.96 |
2.24 |
[5.00–9.70] | [1.13–4.48] | |
Medicare | 1.66 |
0.74 |
[1.11–2.47] | [0.38–1.43] | |
Medicaid | 1.85 |
1.19 |
[1.37–2.51] | [0.67–2.12] | |
Others | 1.77 | 1.72 |
[0.86–3.62] | [0.78–3.80] | |
Uninsured | 2.06 |
3.95 |
[1.43–2.96] | [2.29–6.80] | |
1 if Born in US | 0.97 | 0.68 |
[0.73–1.28] | [0.44–1.04] | |
1 if Male | 1.14 | 1.75 |
[0.93–1.40] | [1.24–2.48] | |
1 if Married | 0.76 |
0.95 |
[0.61–0.96] | [0.65–1.38] | |
1 if college graduate | 1.12 | 1.23 |
[0.87–1.43] | [0.80–1.89] | |
employed | 0.96 | 1.05 |
[0.74–1.23] | [0.67–1.65] | |
1 if Non-English at home | 0.76 | 1.33 |
[0.53–1.09] | [0.82–2.18] | |
Black Non-Hispanic | 0.94 | 0.87 |
[0.69–1.29] | [0.49–1.55] | |
Hispanic | 1.06 | 0.96 |
[0.77–1.45] | [0.58–1.60] | |
Asian/PI Non-Hispanic | 0.84 | 0.52 |
[0.55–1.27] | [0.24–1.12] | |
Others | 1.32 | 1.13 |
[0.68–2.56] | [0.49–2.64] | |
25–44 yrs | 1.44 |
0.88 |
[0.99–2.08] | [0.51–1.53] | |
45–64 yrs | 1.18 | 0.49 |
[0.81–1.71] | [0.28–0.87] | |
65+ yrs | 0.71 | 0.62 |
[0.44–1.14] | [0.32–1.20] | |
100 - <200% FPL | 0.99 | 0.79 |
[0.74–1.32] | [0.51–1.23] | |
200 - <400% FPL | 1.14 | 0.90 |
[0.81–1.59] | [0.51–1.56] | |
400 - <600% FPL | 0.93 | 0.90 |
[0.63–1.37] | [0.50–1.60] | |
>600% FPL | 0.61 |
0.94 |
[0.39–0.97] | [0.46–1.95] | |
Observations | 9,390 | 9,388 |
SOURCE. Author's analysis of New York City Community Health Survey data for 2016.
NOTES. Logistic Regression models are estimated using the svy suite of commands in Stata 15, using weights to control for the complex survey design. FPL: Federal Poverty Level. AOR: Adjusted Odds Ratio; 95% Confidence Intervals are in brackets.
*** p<0.01
** p<0.05
* p<0.1
The pattern is similar for seeking informal care, and column (2) reports the point estimates. In this case, individuals experiencing race-based discrimination are 2.15 times more likely to get medical advice from an informal source (AOR = 2.15; CI: [0.85 5.43]), though this particular estimate is not statistically significant. The next row in the same column shows that for individuals experiencing other types of discrimination, the magnitude of the AOR is similar, but the effect is significant at 5% level (AOR = 2.24; CI: [1.13 4.48]. Notably, these results are obtained after controlling for health insurance status, and lack of insurance is independently associated with higher probabilities of either lacking medical treatment or seeking that from informal sources.
The results in
(1) | (2) | (3) | (4) | |
---|---|---|---|---|
1 if General Health Poor | 1 if Depressed | 1 if has High Pressure | 1 if has diabetes | |
Discrimination based on race/ethnicity | 3.76 |
6.20 |
2.44 |
1.33 |
[2.15–6.59] | [3.44–11.16] | [1.32–4.51] | [0.59–3.00] | |
Discrimination based on other categories | 2.49 |
3.09 |
1.75 |
1.84 |
[1.65–3.75] | [2.00–4.77] | [1.21–2.52] | [1.22–2.77] | |
Medicare | 1.45 |
1.47 |
1.44 |
1.05 |
[1.11–1.89] | [0.95–2.28] | [1.13–1.83] | [0.80–1.38] | |
Medicaid | 1.16 | 1.47 |
1.01 | 0.83 |
[0.91–1.48] | [1.03–2.09] | [0.80–1.26] | [0.63–1.09] | |
Others | 1.12 | 0.88 | 0.84 | 0.55 |
[0.68–1.85] | [0.45–1.71] | [0.56–1.27] | [0.31–0.97] | |
Uninsured | 1.10 | 0.98 | 0.73 |
0.59 |
[0.78–1.55] | [0.60–1.62] | [0.52–1.01] | [0.37–0.94] | |
1 if Born in US | 1.06 | 1.24 | 1.22 |
0.89 |
[0.86–1.32] | [0.92–1.67] | [1.01–1.48] | [0.69–1.14] | |
1 if Male | 0.91 | 0.91 | 0.98 | 1.20 |
[0.77–1.07] | [0.72–1.15] | [0.84–1.14] | [0.99–1.45] | |
1 if Married | 0.94 | 0.60 |
0.87 |
1.02 |
[0.79–1.13] | [0.47–0.79] | [0.74–1.02] | [0.83–1.25] | |
1 if college graduate | 0.69 |
0.49 |
0.78 |
0.59 |
[0.56–0.85] | [0.37–0.65] | [0.65–0.94] | [0.47–0.74] | |
employed | 0.48 |
0.60 |
0.68 |
0.50 |
[0.40–0.58] | [0.46–0.78] | [0.57–0.81] | [0.40–0.62] | |
1 if Non-English at home | 1.49 |
0.85 | 1.11 | 0.86 |
[1.16–1.91] | [0.59–1.24] | [0.88–1.40] | [0.64–1.15] | |
Black Non-Hispanic | 0.99 | 0.54 |
2.05 |
1.82 |
[0.75–1.29] | [0.37–0.79] | [1.63–2.58] | [1.33–2.48] | |
Hispanic | 1.09 | 0.83 | 1.57 |
1.99 |
[0.85–1.41] | [0.57–1.20] | [1.23–2.01] | [1.48–2.69] | |
Asian/PI Non-Hispanic | 1.84 |
0.50 |
0.87 | 1.45 |
[1.37–2.47] | [0.28–0.86] | [0.65–1.16] | [0.97–2.15] | |
Others | 1.57 | 0.97 | 1.23 | 1.47 |
[0.89–2.77] | [0.48–1.97] | [0.71–2.13] | [0.82–2.62] | |
25–44 yrs | 2.81 |
1.40 | 2.65 |
4.02 |
[1.85–4.29] | [0.91–2.15] | [1.73–4.05] | [1.56–10.34] | |
45–64 yrs | 7.25 |
1.91 |
11.31 |
20.48 |
[4.84–10.86] | [1.26–2.91] | [7.49–17.08] | [8.16–51.43] | |
65+ yrs | 7.45 |
1.04 | 20.06 |
30.39 |
[4.88–11.37] | [0.61–1.75] | [13.08–30.76] | [11.88–77.73] | |
100 - <200% FPL | 0.83 |
0.64 |
0.90 | 0.81 |
[0.67–1.03] | [0.47–0.88] | [0.71–1.12] | [0.63–1.03] | |
200 - <400% FPL | 0.53 |
0.54 |
0.91 | 0.78 |
[0.41–0.70] | [0.37–0.78] | [0.71–1.18] | [0.57–1.05] | |
400 - <600% FPL | 0.43 |
0.34 |
0.74 |
0.74 |
[0.31–0.58] | [0.21–0.55] | [0.57–0.97] | [0.53–1.03] | |
>600% FPL | 0.25 |
0.25 |
0.74 |
0.51 |
[0.17–0.37] | [0.15–0.43] | [0.54–1.01] | [0.33–0.79] | |
Observations | 9,358 | 8,869 | 9,412 | 9,416 |
Logistic Regression models are estimated using the svy suite of commands in Stata 15, using weights to control for the complex survey design. FPL: Federal Poverty Level. AOR: Adjusted Odds Ratio; 95% Confidence Intervals are in brackets.
*** p<0.01
** p<0.05
* p<0.1.
For diabetes, the association with race-based discrimination is not statistically significant, while the one with non-racial discrimination is significant at level 1%. Finally, to check if the above results are sensitive to the choice of models, the models for physical and mental health outcomes were re-estimated by adding three additional control variables—smoking, heavy drinking, and Body Mass Index (BMI). The revised models did not qualitatively alter the findings (for detailed results of the sensitivity analyses, please see Table A3 in
Important limitations of these findings include its cross-sectional and self-reported nature. There is a possibility of
In spite of making some progress, the successive goals of eliminating health disparities by the US government in its Healthy People reports have not been met [
This study empirically investigated the association between self-reported experiences of racial and non-racial discrimination and health access and outcome in one of the most diverse cities in the US. There are two major findings. The first is that in health care settings, the estimated prevalence rate of reporting perceived discrimination due to insurance status is higher than such rates due to race or ethnicity among adults in NYC. Second, though the overall pattern of association between the outcome variables and the two broad categories of discrimination was similar, the magnitudes and significance of estimates were more varied. In some cases, it was the non-racial discrimination that was significantly (and negatively) associated with health outcomes as the corresponding adjusted odds ratios were higher in magnitude and more significant.
There are two possible explanations for these findings. First, when individuals seek health care, they tend to become more vulnerable as patients. If they perceive that they are treated differently from other patients based on some personal attributes such as race, income, insurance, or immigration status, their health status may get affected either directly or indirectly through a lack of future care. Second, while providers may be more sensitive to treating patients of different racial backgrounds, given the prominence of race and ethnicity in discrimination studies and training, they may unconsciously show bias based on other group characteristics.
The results have important implications for addressing the health disparities in New York City and beyond. They underscore why efforts should be made to address
(DOCX)
(DOCX)
(DOCX)
I gratefully acknowledge the peer reviewer–Dr. M. Barton Laws, and Dr. Amber Seligman and Tamar Marder from the NYC Department of Health (DoH) for critical comments on an earlier version of this paper. Access to the restricted variable from the NYC DoH is also gratefully acknowledged.