Discrimination as a Social Determinant of Stress and Health among NYC Taxi Drivers

Abstract

Discrimination is associated with poorer mental and physical health outcomes. Taxi drivers have a higher risk of exposure to discrimination and higher rates of chronic conditions. A cross-sectional needs assessment was conducted with a multilingual group of 535 male taxi drivers in New York City. Drivers reporting higher discrimination were more likely to have higher perceived stress and were more likely to have anxiety/depression and chronic pain, adjusting for confounders. Workplace-based interventions designed to help drivers cope with discrimination, stress, and chronic health conditions, interventions to educate the taxi riding public, and greater attention to these issues from administrative agencies are warranted.

INTRODUCTION

Taxi drivers, a large, sedentary, often immigrant workforce which faces multiple health issues (Gany et al., 2013; Gany et al., 2015; Gany et al., 2016a), are likely to encounter ethnic/racial discrimination (Abraham et al., 2008; Schwer et al., 2010). Ethnic/racial discrimination is a noxious and chronic stressor, with widespread deleterious effects, including mental and physical health disparities (Paradies et al., 2015; Pascoe and Richman, 2009; Williams and Mohammed, 2009; Williams et al., 1997; Schmitt et al., 2014). Meta-analyses have found that discrimination is negatively associated with psychological well-being constructs, such as self-esteem and positive affect, and positively associated with depression, anxiety, and stress including among immigrant populations (Schmitt et al., 2014; Paradies et al., 2015; Pascoe and Richman, 2009).

There are differences in the relationship between discrimination and health based on one’s ethnic/racial group, with stronger negative health effects for Asian Americans compared to other groups (Paradies et al., 2015). Perceived discrimination is linked to higher depression, anxiety, and anger among Asian Indians (Nadimpalli et al., 2016) and is predictive of a major depressive episode among Latino and Asian American immigrants (Singh et al., 2016). Although the correlations are not as strong as with mental health outcomes, research suggests that the experience of discrimination is also negatively related to physical health outcomes (Paradies et al., 2015; Schunck et al., 2015). Discrimination is associated with overall poorer self-reported health status among Asian Indians (Nadimpalli et al., 2016). With regard to specific health conditions, discrimination has been linked to a number of cardiovascular risks and outcomes, such as hypertension among African American and Latino groups (Brondolo et al., 2008; Sims et al., 2012), C-reactive protein (a marker of inflammation) in older African Americans (Lewis et al., 2011), abdominal fat in African American adults (Hickson et al., 2012), and diabetes in a multi-ethnic group (Whitaker et al., 2017). In one study, discrimination was also a significant predictor of chronic pain among older African American veterans (Burgess et al., 2009). Research has also suggested that age may be a moderator of the effect of discrimination (Chae et al., 2012).

Although the current study is cross-sectional and, thus, we are unable to examine mediating mechanisms, it is important to consider how discrimination might impact health outcomes. Discrimination is a chronic stressor. As such, from a theoretical standpoint, repeated experiences of discrimination may lead to heightened allostatic load (McEwen, 1998; McEwen, 1999), which initiates the dysregulation of biological processes and leads to chronic health issues (Cohen et al., 1995). Increased stress may influence behavioral responses, specifically health-related behaviors, which then impacts physiological responses such as allostatic load (McEwen, 1998). Mediating health behaviors have been linked to poorer health outcomes. For example, studies have suggested sleep as a possible mediator of the relationship between discrimination and mental and physical health outcomes (Tomfohr et al., 2016). In a recent systematic review, Slopen, Lewis, and Williams (2016) found that discrimination was linked to poorer sleep outcomes among various populations. Other work has linked experiences of discrimination with substance use (Gibbons et al., 2010), smoking and dietary fat (Sims et al., 2016), and poorer diet (Brodish et al., 2011). Early experimental work also indicates that discrimination may have immediate physiological effects (Neblett and Roberts, 2013). As suggested by the allostatic load model, individual characteristics may certainly influence these relationships: age, gender, ethnicity/race, ethnic/racial identity, to name a few. In addition to risky behaviors, Chen and Yang (2014) have also suggested that lower neighborhood social capital serves as a pathway for discrimination to influence poorer health. Indeed, these conceptual frameworks would suggest that the mostly immigrant taxi driving population, would particularly be at risk due to their social and occupational status.

Over 90% of drivers in New York City are immigrant males, primarily from South Asia, the West Indies, and Africa (TLC, 2016), and, hence, also potentially particularly susceptible to discrimination and its health consequences. Within New York City, there are different types of vehicles driven: medallion yellow taxis (primarily in Manhattan), green cabs (primarily in Northern Manhattan and outer boroughs), and liveries. Yellow taxis are the most expensive vehicles to drive per hour. Drivers typically work 10–12 hour shifts (Gany et al., 2013), either during the day, night, or in varied shifts. In addition to demographic variables such as years in the United States, English proficiency, education, income group, and health insurance status, these occupational factors may influence stress, and mental and physical health.

Taxi drivers are at a higher risk for chronic diseases and conditions, including cardiovascular disease (Chen et al., 2005; Elshatarat and Burgel, 2016; Kurosaka et al., 2000), cancer (Hansen et al., 1998), diabetes (Lim and Chia, 2015), fatigue (Lim and Chia, 2015), compromised immune function (Nakano et al., 1998), and pain (Chen et al., 2004). Given that they may also have lower rates of health insurance than the general population (Gany et al., 2016b), taxi drivers in metropolitan cities, such as San Francisco, Los Angeles, and New York City, are particularly vulnerable populations with regard to health status (Apantaku-Onayemi et al., 2012; Blasi and Leavitt, 2006; Burgel et al., 2012).

A few studies have described taxi drivers’ exposure to discrimination. Foreign-born taxi drivers in Las Vegas report significantly higher rates of assault compared to native-born drivers (Schwer et al., 2010). In Toronto, researchers have found that taxi drivers face a lack of safety from customers and a lack of respect and protection from police, due to their immigrant status and lack of English proficiency (Abraham et al., 2008). In a report on the taxi industry in Los Angeles, researchers reported that over a third of the drivers they interviewed recalled an incident of racially targeted harassment, including slurs and hostile comments (Blasi and Leavitt, 2006).

Although some prior work has documented the discrimination experienced by taxi drivers, researchers have not examined how discrimination, as a chronic psychosocial stressor, is associated with poorer driver mental and physical health outcomes. This study examines the impact of discrimination, and the potential moderating effect of region of origin and age, on perceived stress and health risk, specifically for self-reported history of anxiety/depression, Type 2 diabetes, hypertension, hypercholesterolemia, and chronic pain among taxi drivers. These health outcomes were chosen based on prior literature indicating significant associations with discrimination: mental health (Nadimpalli et al., 2016; Pardies et al., 2015), hypertension (Brondolo et al., 2008; Sims et al., 2008), and chronic pain (Burgess et al., 2009). As discrimination has also been linked with overall poorer health in minorities (Nadimpalli et al., 2016) and taxi drivers are particularly at risk for diabetes (Lim and Chia, 2015) and cardiovascular disease (Chen et al., 2005; Elshatarat and Burgel, 2016; Kurosaka et al., 2000), we also chose to assess self-reported Type 2 diabetes and hypercholesterolemia.

MATERIALS AND METHODS

This cross-sectional study utilized a needs assessment questionnaire developed by the Taxi Network, a community-based participatory research group of taxi drivers, community-based organizations, research staff, and research faculty led by the Immigrant Health and Cancer Disparities Center at Memorial Sloan Kettering Cancer Center and the South Asian Council for Social Services.

Measurement

The needs assessment consisted of questions on demographics, workplace and financial profiles, driving attitudes, perceived discrimination, health care access, health behaviors, health conditions, and health treatment status. Only measures relevant to the current analyses are discussed.

Health conditions.

Health conditions were assessed by self-report only. Participants were asked if they were ever diagnosed by a doctor with Type 2 diabetes, hypertension, hypercholesterolemia, anxiety/depression, chronic pain (including back, joint, knees, and/or foot pain, and headaches) with a binary ‒ yes/no ‒ response.

Perceived Stress Scale (PSS-10).

Perceived stress, measured by the Perceived Stress Scale-10 (Cohen et al., 1983), included 10 items with a 5-item response scale of 0 “Never” to 4 “Very often.” The PSS-10 is a widely used assessment which is appropriate for community-based samples and which captures subjective evaluations of personal stress. A total score is obtained by reversing positive items and then summing all items; a higher score indicates higher stress. Sample items included “In the last month, how often have you been upset because of something that happened unexpectedly?” and, “In the last month, how often have you felt difficulties were piling up so high that you could not overcome them?” In the current study, Cronbach’s alpha was .76.

Everyday Discrimination Scale (EDS).

Perceived discrimination was measured using five items from the Everyday Discrimination Scale (Williams et al., 1997): “People have acted as if they think you are not smart,” “People have acted as if they think you were dishonest,” “People have acted as if they are better than you,” “You have been called names or insulted,” and “You have been threatened or harassed.” The EDS is commonly utilized to assess subjective experiences of discrimination. Items were rated on a 4-point scale from 0 = “Never” to 3 = “4 or more times.” Items were averaged, with a higher score indicating increased discrimination. Internal consistency in this sample (α = .78) was similar to that in other studies (α = .77) using the shortened version of the EDS (Sternthal et al., 2011).

Survey Administration

In-person interviewer-administered surveys were conducted from March 2013 to November 2016, during all shifts and seasons, at taxi garage bases, taxi stands, airport holding lots, gas stations, and other community-serving institutions, such as restaurants, worship centers, and community based organizations, in New York City. All drivers congregating in these areas were approached by a research assistant, the study was explained, and the driver was invited to participate in the study. For study inclusion, participants were required to be between 19 and 85 years of age, male, with a minimum of three months experience as a New York City taxi driver. Given that over 95% of the driving population is male (2016), females were excluded from this study. The survey was administered in English and in Arabic, Bengali, Hindi, Chinese, Punjabi, Spanish, and Urdu. Written consent was obtained from participants. Participants were given a small cash incentive to participate.

This study was approved by the Institutional Review Board at Memorial Sloan Kettering Cancer Center.

Data Analysis

For all initial multivariable regression models, we controlled for a baseline set of confounders chosen a priori: age, region of origin, years in the United States, English proficiency, education, income group, health insurance status, type of for-hire-vehicle driven, and work shift driven. Linear regressions were conducted to assess the predictive effect of everyday discrimination on perceived stress, and a series of logistic regressions were conducted to examine whether discrimination predicted the diagnoses of one of the five individual health conditions. We also considered the potential moderating effects of age and region of origin on the relationships between discrimination and our outcomes of interest using interaction terms of discrimination with age and region of origin respectively. Bonferroni corrections were applied to account for multiple testing across our five models (type I error =.05/7 = 0.007). All statistical modeling was conducted using R statistical software version 3.3.1.

Missing Data

All descriptive statistics were calculated via an available case analysis, removing any missing data from the frequencies and percentages reported. For the multivariable regression models, estimates were computed both using the complete cases as well as based on four multiply imputed datasets under the missing at random assumption. Multiple imputations were conducted using the mi package in R (Su et al., 2011). Multiple imputed estimates were derived using Rubin’s combining rules for the imputed datasets (Rubin, 2004).

RESULTS

Descriptive Statistics

Data from 535 drivers were included in the analyses (Table 1). Drivers were between 21 and 80 years old (M= 44.09, SD = 11.33). The majority of surveys were conducted in English (n = 286); the rest were conducted in Chinese (n = 91), Bengali (n = 47), Arabic (n = 25), Hindi (n = 7), Punjabi (n = 15), Urdu (n = 42), and Spanish (n = 22). Just a quarter of the drivers indicated that they spoke English very well (23.7%). Drivers were predominately from South Asian countries (43.7%), followed by African or Afro-Caribbean countries (22.6%), East Asian countries (18.1%), and other countries (15.5%). Of these other countries, 19 drivers (4%) were born in the United States. Drivers predominantly drove medallion yellow cabs (52.3%), followed by livery cabs (28.4%), and green cabs (19.3%). Average gross monthly income was reported based on three groups: almost half reported making more than $2256 per month on average (44.7%) while approximately a third of the sample reported making less than $1806 per month (31.4%); the rest made between $1806 and $2256 (24%). Most drivers worked during the day shift (57.7%). The majority of drivers indicated that they had health insurance (74%). Most drivers reported visiting a doctor for a routine check-up in the last year (76.1%). Other drivers indicated visiting a doctor within the past two years (9.9%), past five years (5.8%), five or more years (3.9%), and never (3.2%). The remaining drivers either did not know or had missing data (1.2%). Most drivers reported at least one health condition (67%): 39.2% hypercholesterolemia, 31.2% reported chronic pain, 20.7% reported hypertension, 18.4% of drivers reported having diabetes, and 7.8% reported anxiety/depression.

Table 1.

Study Descriptives

Variables	Level
n		535
Age (mean (sd))		44.2 (11.4)
Years in the US (mean (sd))		16.9 (9.8)
Birth Region	Other	83 (15.5)
	South Asian	234 (43.7)
	African/Afro-Caribbean	121 (22.6)
	East Asian	97 (18.1)
Income Group	<$1806	144 (31.4)
	$1806 – $2256	110 (24.0)
	More than $2256	205 (44.7)
Shift	Day	306 (57.7)
	Night	138 (26.0)
	Varies	86 (16.2)
Type of Vehicle	Medallion Yellow Cab	280 (52.3)
	Livery Cab	152 (28.4)
	Green Cab	103 (19.3)
Education	< High School	78 (14.7)
	12th Grade/HS Graduate	166 (31.3)
	College Graduate	151 (28.4)
	Post college/Graduate school	42 (7.9)
	Some College	94 (17.7)
English Proficiency	Not at All	16 (3.4)
	Not Well	121 (26.0)
	Well	218 (46.9)
	Very Well	110 (23.7)
Health Insurance		393 (74.0)
Everyday Discrimination (mean (sd))		0.89 (0.75)
Perceived Stress (mean (sd))		12.5 (7.0)
Diabetes		92 (18.4)
Hypertension		102 (20.7)
Hypercholesterolemia		192 (39.2)
Anxiety/Depression		38 (7.8)
Chronic Pain		158 (31.2)

Differences Based on Demographic and Occupational Factors

Discrimination.

Reported discrimination varied by demographic and occupational factors. Those from African/Afro-Caribbean nations reported higher rates of discrimination (M = 1.08) than other groups, with drivers from East Asian countries reporting the least discrimination (M = .54). Those driving during the night shift also reported more discrimination (M = 1.12) than those driving during the day or varied shifts. Yellow cab drivers (M = 1.01) reported more discrimination than green (M= .88) or livery drivers (M= .67). Those reporting that they spoke English only well reported higher discrimination (M = .96) than those who spoke English very well (M= .86) or not at all/not well (M= .69). Age and number of years in the United States were not related to rates of discrimination.

Stress.

Drivers from East Asian countries (M= 9.9) reported lower rates of stress than drivers from South Asian (M= 12.7), African/Afro-Caribbean (M= 12.8), and Other regions (M = 14.2, p<.01). Higher income drivers (M= 11.7) reported lower stress than other drivers. Night shift drivers (M= 13.8) experienced more stress than drivers during other shifts. Medallion yellow cab drivers reported more stress (M= 13.6) than green cab or livery drivers (p<.01). Drivers reporting that they spoke English well reported more stress (M= 13.2) than drivers who spoke English very well (M= 12.7) or not at all/not well (M= 11.3). Age was negatively correlated with stress (r = −.18, p<.01), but number of years in the United States was not related. Unless noted, these differences were significant at p<.05.

Health Outcomes.

East Asian (OR =0.36, 95% CI [0.15, 0.89]) and African/Afro-Caribbean (OR=0.35, 95% CI [0.18, 0.67]) drivers reported significantly lower rates of chronic pain in comparison to other drivers. Night shift drivers suffered from higher rates of anxiety (OR= 2.18 [1.01, 4.7]). South Asian drivers were three times as likely (OR=3.18, 95% CI [1.33, 7.56]) to have diabetes compared to African/Afro-Caribbean and Other drivers. East Asian drivers were five times less likely (OR=0.18, 95% CI [0.04, 0.76]) to have diabetes compared to drivers from African/Afro-Caribbean and Other regions.

Multivariable Regression: Effect of Discrimination on Stress, Anxiety/Depression and Chronic Pain

Multivariable regression models found discrimination to be significantly associated with stress, anxiety/depression, and chronic pain (Table 2). A one unit increase in the everyday discrimination scale was associated with a 2.45 (95% CI [1.66, 3.24]) increase in perceived stress. Those with more discrimination were 88% more likely to report anxiety/depression (OR=1.88, 95% CI [1.23, 2.87]) and 52% more likely to report chronic pain (OR=1.52, 95% CI [1.15, 2.02]). Discrimination was not significantly associated with Type 2 diabetes, hypertension, or hypercholesterolemia. These results are based on combined analyses from four multiple imputed datasets adjusting for relevant confounders described in the methods. Age and region of origin did not moderate the effect of discrimination.

Table 2.

Effect of Everyday Discrimination on Stress and Health Outcomes from Multivariable Regression Models

Outcome	Multiple Imputation (N=535,m=4)			Complete Case Analysis				Model
	Effect of Discrimination	95% CI	p-value	Effect of Discrimination	95% CI	p-value	N (%)
Perceived Stress	2.45	(1.66, 3.24)	<0.0001	2.62	(2.47, 2.77)	<0.0001	371 (69%)	Linear
Chronic Pain	1.52	(1.15, 2.02)	0.0034	1.86	(1.29, 2.67)	<0.0001	360 (67%)	Logistic
Anxiety/Depression	1.88	(1.23, 2.87)	0.0038	1.35	(0.75, 2.44)	0.3138	354 (66%)	Logistic
Diabetes	1.13	(0.79, 1.62)	0.4904	1.17	(0.75, 1.84)	0.4921	359 (67%)	Logistic
Hypercholesterolemia	1.21	(0.89, 1.63)	0.2224	1.14	(0.80, 1.63)	0.4757	352 (66%)	Logistic
Hypertension	1.31	(0.93, 1.85)	0.1248	1.29	(0.86, 1.95)	0.2234	354 (66%)	Logistic

We display the regression coefficient for the linear model, the relative risk ratio (RRR) for the Poisson model, and the odds ratio (OR) for the logistic regression models. Multiple Imputed Estimates were derived using Rubin’s combining rules for the m=4 imputed datasets..

DISCUSSION

Taxi drivers are a vulnerable population. In this research, we describe how the social burden of chronic discrimination is associated with health outcomes in this population. Our findings support prior research showing links between experiences of discrimination and mental and physical health outcomes (Paradies et al., 2015; Pascoe and Richman, 2009; Williams and Mohammed, 2009; Williams et al., 1997; Schmitt et al., 2014), and extends this work by documenting these relationships within an occupational group routinely exposed to discrimination. Within our large representative sample of New York City taxi and for-hire vehicle drivers, we found that discrimination significantly predicted higher levels of perceived stress, and a greater likelihood of anxiety/depression and chronic pain, when controlling for sociodemographic and workplace factors. A largely immigrant group, drivers may be particularly vulnerable to discrimination from multiple sources, including passengers and police (Abraham et al., 2008). Age and region of origin did not moderate these relationships, such that the relationship between discrimination and outcomes did not differ based on region of origin. However, there were notable differences in levels of discrimination, stress, chronic pain, and diabetes by demographic and occupational factors.

A number of studies indicate a consistent negative relationship between discrimination and perceived stress, and our results echo those findings (Paradies et al., 2015). In our study, drivers who reported that they had encountered discrimination more frequently also indicated that they experienced higher levels of general stress. The chronic and interpersonal nature of discrimination may lead to a heightened stress response (McEwen, 1998; McEwen, 1999). Research on stress proliferation suggests that exposure to one stressor may also lead to other secondary stressors (Pearlin et al., 2005). As such, those experiencing discrimination may hypothetically view future social interactions in more negative ways, leading to additional stress (Brondolo et al., 2008). Prior work also suggests that among minority populations, discrimination negatively influences markers of stress, such as ambulatory blood pressure even when controlling for personality factors (Brondolo et al., 2008), and telomere length (Chae et al., 2014; Chae et al., 2016). Experimental work on discrimination further indicates that exposure to discrimination may affect autonomic responses (Neblett and Roberts, 2013). Research has suggested that discrimination influences outcomes through stress-mediated pathways (Perry et al., 2013).

Our findings suggest that discrimination significantly impacts drivers psychologically and physically. We found that higher discrimination predicted a higher likelihood of reporting anxiety/depression and chronic pain. This supports a large body of research linking discrimination and health outcomes (Paradies et al., 2015; Williams and Mohammed, 2009; Pascoe and Richman, 2009; Schmitt et al., 2014). Research consistently indicates that discrimination appears to have robust relationships with various internalizing symptoms and specifically with anxiety and depressive symptoms across different immigrant groups (Nadimpalli et al., 2016; Singh et al., 2016). Mental health may be of particular concern in immigrant groups (Karasz et al., 2016). Although few drivers in our sample reported anxiety or depression, discrimination had a predictive effect. Given the likely stigma surrounding mental health issues in this sample, future research should consider culturally sensitive mental health interventions. Despite weaker documented links between discrimination and physical health in prior research (Paradies et al., 2015), our results indicate a higher risk of reporting chronic pain based on experienced discrimination. Few studies have investigated the link between discrimination and chronic pain. Burgess and colleagues found that among older African American veterans, higher reports of lifetime discrimination (e.g., based on specified situations, such as at school, being hired, getting service in a store or restaurant, on the street or public setting, from the police or courts, among others) were related to the likelihood of experiencing moderate to severe pain over the past four weeks (Burgess et al., 2009). In a nationally representative sample of Asian Americans, everyday discrimination was also associated with chronic illnesses more generally and chronic pain specifically (Gee et al., 2007). Chronic pain may also be a significant area of concern among taxi drivers, given the long hours spent in seated positions and low levels of physical activity.

In past research, some researchers have found that discrimination was related to cardiovascular risks and outcomes, including diabetes, hypertension, and high cholesterol (Lewis et al., 2011; Hickson et al., 2012). Meta-analyses, though, have indicated that across studies, discrimination has little to no correlation with diabetes, hypertension, high cholesterol, and heart conditions/illnesses (Paradies et al., 2015; Pascoe and Richman, 2009). In this study, we did not find any relationship between discrimination and reports of any of these conditions.

Finally, the lack of moderating effects suggests that the discrimination-distress relationship does not significantly differ across age or region of origin in this specific occupational group. Discrimination appears to predict mental and physical health issues regardless of these demographic factors. Although we sampled from a wide age range, it is possible that the links between discrimination and health vary more in childhood and adolescence, and less so in adulthood. In a life course perspective of the impact of discrimination on health inequity, Gee, Walsemann, and Brondolo (2012) also suggest that discrimination may have larger influences on mental and physical health outcomes particularly during sensitive periods.

Meta-analyses have also indicated that ethnicity/race moderates the relationship between discrimination and health, with the association being strongest for Asian Americans in comparison to Latinos and African Americans (Paradies et al., 2015). Although ethnicity/race does not directly equate with region of origin, we expected that based on these findings, there would be a moderating effect, based on region of origin, which we did not find. However, we did find important differences in levels of discrimination and our outcomes of interest based on demographic and occupational factors. In prior work, African American/Black individuals have reported higher discrimination in comparison to other ethnic/racial minority groups (Kim et al., 2014), and our findings support this. East Asian drivers reported lower discrimination and lower stress than other groups. Researchers have linked skin color with discrimination (e.g., Klonoff & Landrine, 2000; Monk, 2015) and different ethnic/racial minorities experience varying cultural stereotypes. Further, there may be additional factors not measured in this study, such as ethnic/racial identity, dimensions of which may moderate the association between discrimination and health for certain groups (e.g., Lee, 2005).

Occupational variables also exert influence. Medallion yellow taxi drivers and night shift drivers appeared to be most at risk, with yellow taxi drivers reporting higher levels of discrimination and stress. Yellow taxis are the iconic New York taxi symbol, yet the industry is changing, with yellow taxi drivers facing increasing financial strain and competition from app-based for hire vehicles and green taxis. Night shift drivers also reported higher levels of discrimination and stress, and reported higher rates of anxiety/depression. It is clear that these night shift drivers experience heightened disadvantages as a result of the timing of their work, perhaps due to a lack of safety during these hours. Other researchers have documented the dangers to which taxi drivers are exposed (Abraham et al., 2008; Schwer et al., 2010). Future work should consider how these risks might be mitigated.

Limitations

There are limitations to note in this study. As the study was cross-sectional, we were unable to appropriately assess mediating mechanisms. Given that there is no instantaneous link between discrimination and health, but rather many multi-factorial pathways that interact in a complex manner, larger, longitudinal designs and innovative experimental work are necessary to appropriately examine causal links (Williams and Mohammed, 2009). We also utilized self-reports of health, which may not fully capture the presence of conditions, especially given the rate of uptake of medical services. Additionally, we asked if drivers had been diagnosed with each health condition. Given the relatively high rates of being uninsured in this population (26%) in comparison to the uninsured rates of 10.5–12.5% in the New York Metro Area in 2013–2014 (Smith and Medalia, 2015), and that 24% of our sample had not visited a doctor in the past year, it is possible that drivers were unaware of their health conditions. Thus, health conditions may be underreported in this study. We did not utilize objective measures of health, including stress biomarkers and biometrics, which may provide valuable data to support the association between discrimination and health. Finally, with regard to our measurement of discrimination, we did not collect information on which particular identity (e.g. occupation, country of origin, race) elicited participants’ perception of discrimination. As a result, we cannot specifically determine whether perceived discrimination had to do with their ethnicity/race, immigrant status, and/or occupation.

Statistical limitations are related to overall model specification, which specifically include issues of unobserved confounding and assumptions underlying multiple imputations. In terms of missing data, unobserved health outcomes could affect probability of missingness (i.e. someone with anxiety might be less likely to admit being diagnosed and thus refuse to answer the question). This could bias our imputations by violating the missing at random assumption (Rubin, 1976). Further, given uneven sample sizes due to the convenience sample, our non-significant results may not necessarily indicate that there is no difference between groups but may instead reflect a methodological challenge. Despite these limitations, we believe our large sample size, strict Bonferroni p-value adjustments, and inclusion of many relevant confounders provide us protection from any extreme bias in generalizing our results. Strengths of the study include a large sample size, a representative sample with a wide distribution of ages, languages, regions of birth, and levels of education.

Conclusions and Implications

As a predominantly immigrant group which already grapples with challenging occupation-based issues that negatively impacting health, this study points to the importance of simultaneously considering the psychosocial determinants of health when developing health interventions for taxi drivers. Thus, researchers, health care professionals and providers should consider culturally sensitive interventions designed to help drivers manage discriminatory events, bolster coping strategies, and access mental health services. Recent work has suggested that coping styles and cultural beliefs could have buffering effects for the negative effects of experienced discrimination (Nadimpalli et al., 2016). Tailored educational interventions around these resources may be valuable for drivers. Further, local and state governmental agencies that regularly interface with drivers should consider policies that protect drivers from discrimination and harassment, including education initiatives targeting the taxi-riding public and agency representatives. Systematic changes to minimize perceived discrimination are necessary to address health disparities.

This study contributes to this important and burgeoning field of study on the influence of discrimination on the health and well-being of ethnic/racial minority populations. We found that higher reports of discrimination significantly predicted higher rates of perceived stress, anxiety/depression, and chronic pain among New York City taxi drivers. Our results indicate the toxicity of discrimination on the stress and health outcomes of this vulnerable group.