Taxi Drivers and Modifiable Health Behaviors: Is Stress a Predictor?

Abstract

Background:

Taxi drivers, an immigrant male population, may exhibit poor health behaviors and increased health risks.

Objectives:

The current study examined stress and demographics as predictors of physical activity (PA), nutrition, sleep, and smoking, and the co-occurrence of these behaviors among taxi drivers.

Methods:

A cross-sectional needs assessment was conducted in New York City. The sample (n = 252) was comprised of male taxi drivers, 98% of whom were born outside of the U.S., with the majority from South Asian countries (61.5%), and 44.92 years old on average (SD = 11.21).

Results:

We found low rates of fruit/vegetable consumption and PA. Rates of stress, PA, and smoking varied by demographic factors. Stress positively predicted sleep disturbances and negatively predicted smoking. Aside from a relationship between sugar consumption and smoking, other health behaviors were not associated.

Conclusions:

While stress appears to impact some indicators of modifiable health behaviors, its lack of relationship with others points to more persistent health issues. Demographic differences found for PA and smoking also point to groups that may especially benefit from interventions. These findings suggest the need for targeted health interventions for taxi drivers in large metropolitan cities.

Background

Taxi and app-based drivers are a prominent and growing presence in major cities throughout the U.S. In New York City (NYC) alone, there are over 50,000 yellow taxi medallion drivers and over 100,000 app-based drivers (1). These drivers often work 10–12 hours a day, six days a week (2). Research across various cities has shown that these long hours, combined with the sedentary nature of their work, high stress, and low rates of insurance and healthcare access (3), contribute to poor health outcomes (4–6).

Workplace conditions can deleteriously impact health through their facilitation of negative health behavior patterns (2, 7–9). Due to the opportunity cost of time spent on any activity besides driving, many drivers strive to maximize their driving time and hesitate to take breaks for health-promoting activities, such as using the bathroom, eating lunch, and stretching (2, 10). South Asian taxi drivers in NYC report having little to no physical activity and eating unhealthy fast foods at work due to these time constraints and difficulty parking (2). Among taxi drivers in Chicago, only 4.6% of drivers reported eating the recommended five daily servings of fruits and vegetables and 40% of drivers reported never exercising (7). Taxi drivers also report suffering from sleep related issues, including driving fatigue (11) and sleep apnea (12), in part due to increased obesity rates.

In combination with, and as a result of, stressful working conditions, long hours, low pay, and lack of health care access, drivers’ health behaviors likely contribute to poorer overall health among drivers. Drivers report stress as the foremost barrier to good health and job satisfaction, often indicating that they would take another job as soon as it became available (2). Given the prominent occupational factors negatively influencing drivers’ health, protective health behaviors become especially important to foster in this population.

Results from several promising workplace-based health screenings and interventions have suggested that there are opportunities to improve taxi driver health behaviors (3, 13–15). However, because the taxi driver population has very limited time and resources, and because sustainability and disseminability of health behavior interventions is key (16), it is useful to investigate the most critical driver health behavior needs to design the most effective, efficacious interventions. Although notable work has underscored the importance of systemic problems that NYC drivers face, such as a lack of health insurance (17) and ethnic/racial discrimination (18), no research has comprehensively investigated the most pervasive health behaviors in NYC taxi drivers that likely contribute to their poor health status. To direct health behavior interventions, it is worthwhile to understand health behavior patterns in this population, specifically, whether stress significantly predicts health behaviors, the demographic groups in which poor health behaviors tend to persist, and whether health behaviors tend to co-occur. With this knowledge, interventionists can make informed decisions, such as whether to target underlying contributors to multiple health behaviors such as stress. This paper will examine whether taxi drivers’ health behaviors are predicted by stress and demographic factors, and whether these behaviors occur together or separately. These results will inform primary health behavior intervention targets.

Methods

This study examined data collected in a cross-sectional health needs assessment conducted by the Taxi Network, a community-based participatory research program, led by the Immigrant Health and Cancer Disparities Service, Memorial Sloan Kettering Cancer Center. This study was approved by the Institutional Review Board at Memorial Sloan Kettering Cancer Center.

Needs Assessment Questionnaire

A needs assessment questionnaire queried drivers on demographics, workplace and financial profiles, health care access, perceived stress, and health behaviors, including physical activity, nutrition, sleep, and smoking.

Perceived Stress

Perceived stress was measured using the Perceived Stress Scale-10 (19), which includes 10 items with a five-item response scale of 0 “Never” to 4 “Very often.” The PSS-10 is widely used and is appropriate for community-based samples. It captures subjective evaluations of personal stress. Sample questions include, “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?”. A total score is obtained by reverse coding positive items and then summing all items; a higher score indicates higher stress. Summed scores 13 and below are typically considered indicative of “low stress,” between 14 and 26, “moderate stress,” and 27 and above, “high stress.”

Health Behaviors

Physical activity

Participants were asked if, the number of days per week, and the number of minutes/day they had participated in, physical activities, and what type, in the past month. Moderate physical activities were defined as brisk walking, bicycling, gardening, and other similar activities. Vigorous physical activities were defined as running, aerobics, heavy yard work, soccer, cricket, and other similar activities.

Diet

Nutrition was assessed using the Dietary Screener Questionnaire (20), a focused dietary assessment (21) that queries how often in the last month the participant had consumed certain foods. Questions included “During the past month, how often did you eat fruit? Include fresh, frozen or canned fruit.” Scoring algorithms produce scores for the number of daily fruits and vegetables eaten (in cups), and the number of teaspoons of sugar consumed daily (including from sugar sweetened beverages) (20, 22). The DSQ has been used in large scale studies and shown to have good validity to actual consumption (23).

Sleep

Sleep quantity and disturbances were assessed using the sleep scale designed for the Medical Outcomes Study (MOS-SS), a 12-item scale that measures self-reported sleep quantity, quality, and disturbances over the past four weeks. Sample items include: “How often in the past four weeks did you feel that your sleep was not quiet (moving restlessly, feeling tense, speaking, etc., while sleeping)?” and “How often in the past four weeks did you awaken short of breath or with a headache?” Items were rated on a six-point scale (1 = All the time; 6 = None of the time. Sleep quantity was measured by one item (“On the average, how many hours did you sleep each night during the past 4 weeks?”) and sleep disturbances were operationalized as the subscale for a sleep problem index (SLP-II), a summary measure of sleep quality derived from scores on nine of the 12 items.

Smoking

Participants were asked if they currently smoked cigarettes. Response options were “No” and “Yes, currently” and “Yes, in the past, but not anymore.” For regression analyses, we coded this into a binary variable to account for current smoking (0 = No, never smoked or smoked in the past; 1 = Yes, currently smoke).

Participant Recruitment and Survey Administration

In-person interviewer-administered surveys were conducted from November 2014 to November 2016 in New York City, 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. All drivers congregating in these areas were invited to participate in the study. For study inclusion, participants were required to be between 19 and 85 years of age, male, and with a minimum of three months experience as a New York City taxi driver. For the current analyses, only male drivers with responses to all questions of interest were included. The survey was administered in English and in Arabic, Bengali, Hindi, Punjabi, Spanish, and Urdu.

Data

To determine the prevalence of our targeted health behaviors, we used descriptive statistics and frequencies. We conducted a linear regression to examine whether demographic variables predicted stress. Eight regression models were conducted to determine the relationship of health behaviors to one another, with seven linear regressions conducted for physical activity, nutrition, and sleep indicators. One logistic regression was conducted for smoking. Each health outcome was regressed on the others, with stress as a predictor in the model and controlling for key demographic variables. These included age, region of birth, English proficiency, and vehicle ownership. All analyses were conducted using SPSS version 25 (24).

Results

Demographics

A total of 252 drivers were included (see Table 1). The majority of surveys were administered in English (55%), followed by Bengali (16%), Urdu (12%), Arabic (10%), Punjabi (5%), Spanish (2%), Hindi (1%), and French (.5%). Drivers’ mean age was 44.92 years (SD = 11.21). The majority of drivers were foreign-born (98%), with most from South Asian countries (61.5%). Most drivers drove yellow cabs (55%), followed by green cabs (42%). A small percentage of drivers drove black cabs/limousines, livery cabs, and gypsy cabs (3%). The majority of drivers reported that they did not own their vehicles (61%). On average, drivers reported living in the United States for 18 years (SD = 10.05), with a minimum of a 1.5 years and a maximum of 56 years. Most drivers drove the day shift (61%), 22% drove the night shift, and 17% varied shifts. The largest percentage of drivers had a high school degree or some college education (44%), followed by college degree or post-college schooling (43%), and less than a high school degree (13%). Most drivers were married (82%).

Table 1.

Descriptives (n = 252)

Variables	Level	n (%)
Age (M (SD))		44.92 (11.21)
Years in the US (M (SD))		17.94 (10.05)
Region of birth	South Asia	155 (61.5)
	Latin America	27 (10.7)
	Africa	31 (12.3)
	Other	39 (15.5)
Vehicle type	Yellow cabs	139 (55.2)
	Green cabs	105 (41.7)
	Other	8 (3.2)
Vehicle ownership	Doesn’t own car	154 (61.1)
	Owns car	96 (38.1)
Education	Less than a high school degree	33 (13.1)
	High school degree or some college education	110 (43.7)
	College degree and more	109 (43.3)
Marital status	Married/partnered	204 (81.9)
	Never married	26 (10.4)
	Divorced/separated/widowed	19 (7.6)
Work shift	Day	153 (61)
	Night	56 (22.3)
	Varies	42 (16.7)
English proficiency	Not at all/not well/well	186 (73.8)
	Very well	65 (25.8)
Stress (M (SD))		13.91 (6.87)
	Low	111 (45.5)
	Moderate	126 (51.6)
	High	7 (2.9)
Physical Activity	Moderate PA (M (SD))	112.54 (167.91)
	Recommended moderate PA per week (150 min>)	69 (27.4)
	Vigorous PA (M (SD)	56.4 (116.99)
	Recommended vigorous PA per week (75 min>)	67 (26.6)
Nutrition	Fruit consumption	.70 (.76)
	Vegetable consumption	1.82 (1.16)
	Sugar consumption	8.31 (6.52)
Sleep	Sleep quantity (M (SD))	6.57 (1.18)
	Less than 7 hours	126 (52.7)
	7–9 hours	110 (46)
	Over 9 hours	3 (1.3)
	Sleep disturbances (M (SD))	25.38 (18.52)
Smoking	Currently	45 (17.9)
	In the past	34 (13.5)

Stress and Health Behaviors

Stress

On the 10-item perceived stress scale, drivers reported their perceived stress (M = 13.91, SD = 6.87, range = 0–33). Forty-six percent of drivers had scores of 13 or below, indicating low stress. 52% of drivers had scores indicative of moderate stress, and 3% had scores indicating high levels of stress. English proficiency significantly predicted stress; drivers who were English proficient reported lower stress (M = 11.98, SD = 6.86) than drivers who were not English proficient (M= 14.57, SD = 6.79).

Physical activity

Drivers reported an average of 113 minutes of moderate activity per week (SD = 167.91) and 56 minutes of vigorous activity per week (SD = 116.99). However, almost 40% of drivers indicated that they had not engaged in any moderate activity over the past month and most drivers (63.9%) indicated that they had not engaged in any vigorous activity over the past month. With all other predictors held constant, age significantly negatively predicted vigorous exercise (β = −.28, p=.001), such that being older was associated with less vigorous exercise. English proficiency significantly predicted both moderate exercise (β = .22, p=.010) and vigorous exercise (β = .33, p<.001). Those who were English proficient reported more minutes of moderate exercise (M = 151.92, SD = 218.12) than those were not (M = 99.38, SD = 144.94). Similarly, those who were English proficient reported more minutes of vigorous exercise (M = 110.68, SD = 148.91) than those who were not (M = 37.73, SD = 97. 44).

Nutrition

Drivers reported low rates of fruit and vegetable consumption. The average total fruit consumption per day was .70 cups (SD = .76). Mean total vegetable consumption per day was 1.82 cups (SD = 1.16). Mean sugar consumption per day was 8.31 teaspoons of added sugar (SD = 6.52). There were no significant predictors of fruit and vegetable consumption. However, smoking was a significant predictor of mean sugar consumption (β = .166, p= .037), such that those who reported smoking currently reported higher sugar consumption (M = 9.95, SD = 7.64) than those who did not currently smoke (M = 7.84, SD = 6.15).

Sleep

On average, drivers reported sleeping 6.57 hours per night (SD = 1.18, range: 3.5 – 10 hours) over the past four weeks. 53% of drivers obtained less than 7 hours of sleep, 46% of drivers obtained 7–9 hours of sleep, and 1% of drivers reported more than 9 hours of sleep nightly. The mean for the nine-item sleep disturbances domain was 25.38 (SD = 18.52). Perceived stress was a significant positive predictor of sleep disturbances (β = .525, p<.001).

Smoking

17.9% of the sample reported being current smokers. With regard to region of birth, South Asians were 11.56 times more likely to currently smoke than drivers from Other regions. The odds for smoking were 9.9% lower for per unit of stress. Higher sugar consumption was related to higher odds of smoking, such that reporting a one unit (one teaspoon) increase in sugar consumption indicated an increase in the likelihood of being a current smoker by a factor of 1.10.

Co-occurrence or Independence of Health Behaviors

Except for the relationship between sugar consumption and smoking, there were no other significant relationships between any of the indicators of the examined health behaviors, suggesting that these health behaviors may occur independently in this sample.

Discussion

This study describes health behaviors among NYC taxi drivers and examines the relationships among perceived stress, demographic factors, and health behaviors. The majority of drivers indicated moderate or high stress. Poor nutrition and low rates of physical activity were pervasive across our entire sample, and smoking rates were higher than national rates for men of similar ethnicities, whereas sleep quantity and disturbances were comparable to averages reported in prior work (25).

Overall, drivers reported poor nutrition and low rates of physical activity. The majority of drivers reported low fruit and vegetable consumption. It is currently recommended that adults eat 1.5–2 cups of fruit and 2–3 cups of vegetables daily (26). Only 15 drivers reported eating 1.5 or more cups of fruit per day, whereas 79 drivers reported eating two or more cups of vegetables per day. Almost half of the drivers reported no moderate physical activity and over 60% of drivers reported no vigorous physical activity over the past month. Only 27% reported obtaining 150 minutes or more of moderate physical activity and 27% reported 75 minutes or more of vigorous physical activity per week, in line with CDC recommendations (27). Because nutrition and physical activity habits contribute to cardiovascular health, and especially because poor cardiovascular health has been documented in this population (4, 6, 28), these rates are concerning. Age negatively predicted vigorous exercise, supporting evidence that as adults grow older, they are less likely to report physical activity (29, 30). Drivers who reported speaking English very well also reported more moderate and vigorous physical activity. It is possible that being English proficient presents more opportunities for participating in physical activity; health education and advertising (e.g., gym ads) may also be more accessible to these drivers.

Smoking was predictive of mean sugar consumption, which supports previous literature (31). Smoking and sugar consumption each confer high risk for multiple diseases, and together this risk compounds, in part due to inflammatory cytokines. Smoking and sugar together have been shown to contribute to inflammatory bowel disease (32, 33), Crohn’s disease (34), diabetes (35), heart disease (36), and cancer (37); they pose further risk for those who already exhibit chronic disease (e.g., diabetes) or for groups particularly prone to chronic disease such as cardiac illness (i.e., taxi drivers).

The National Sleep Foundation recommends that adults obtain 7–9 hours of sleep per night (38). Over half of the drivers in our sample obtained fewer than 7 hours of sleep per night, and at least 16% of drivers reported 5 hours or fewer. Researchers have indicated the long-term health consequences of poor sleep, including increased risk of cancer (39), cardiac illness (40, 41) and car accidents (42). A large body of literature has identified a relationship between stress and sleep (43). We did find that stress was a significant positive predictor of sleep disturbances. Prior research has indicated that fatigue (11) and sleep apnea (12) are concerns for the taxi driving population, and a related study examining the links between stress and sleep behaviors among NYC taxi drivers found high rates of snoring and shortness of breath (44).

Finally, approximately 18% of our sample reported currently smoking cigarettes and an additional 13% reported smoking in the past. Rates of smoking in the United States have decreased from 20% in 2005 to 15%, and to 9% in non-Hispanic Asians (45), lower than current smoking rates in this sample (18%). In our sample, South Asian drivers reported the highest rates of current smoking in comparison to drivers from other ethnic/racial groups. Smoking is the number one preventable cause of cancer (46) and is a CVD risk factor (47), and should therefore be a primary intervention target for this group. Research has also suggested that smokeless tobacco (i.e., gutka and paan) is commonly used by South Asians; however, there is prevalent skepticism in the community about its perceived harm (48). We were unable to investigate the use of smokeless tobacco in the current study due to low completion rates of this section of the survey. Lower stress was related to higher odds, albeit only slightly, of reporting current smoking. Smokers frequently cite stress as their reason for tobacco use, and recent work indicates that smoking may decrease arousal and stress on a physiological level (49). However, diary data indicate that smoking is related to higher daily stress (50), and nationally representative data suggests that smokers more frequently report more stress in comparison to non-smokers and those who quit smoking (51). The relationship between stress and smoking is complex, and more longitudinal data is needed to disentangle causal mechanisms. Higher sugar consumption was also strongly predictive of smoking, which as previously discussed, poses multiplicative health risks.

In summary, the results of this study reveal areas for health improvement within this population and compel the use of separate interventions for the primary intervention targets of nutrition, physical activity, and smoking.

Strengths and Limitations

Our sample was diverse, which allowed us to investigate sociodemographic predictors of health behaviors. The survey was also available in several languages, to reflect the languages of the drivers. Study limitations include its cross-sectional design, which did not allow us to consider causal mechanisms, the convenience sampling strategy, which may have biased the results, and the relative paucity of night shift drivers, who may experience heightened sleep issues, for example. We also conducted this study between 2014–2016, during a period when the taxi medallion values in New York City were plummeting and the ride-sharing industry was gaining popularity. In 2013, taxi medallions were worth as much as $1.30 million and in 2018 as little as $160,000 (52). However, we did not explicitly ask about financial strain or concerns around the industry; as such we are unable to determine whether and how this might have had an impact on drivers’ reports of perceived stress. Additionally, we relied on self-reported measures of health behaviors. When self-reporting sleep duration, participants may account for their time spent in bed, rather than actual time spent sleeping, which inflates average estimates. Studies have noted that social desirability biases can lead to participants underreporting poor health behaviors, such as tobacco use. In particular, there may have been religious and/or cultural implications for these health behaviors, which could have precluded participants from accurately reporting them.

Implications for Interventions

The taxi drivers’ health behavior profile shows that poor nutrition, minimal physical activity, and smoking rates are primary targets for intervention with this group. Research indicates that the greatest predictor of healthy behaviors, robustly, is one’s environment (53), and even with the best individual behavior change, people are better able to make healthy choices when those opportunities are available (54). Partnering with community stakeholders is therefore also critical. Garages and/or the NYC Taxi and Limousine Commission could sponsor the availability of healthy food options and provide more physical activity opportunities (a few of the garages have exercise equipment, the airport holding lots have space for walking and other activities but drivers are required to be with their taxis when they are queued up to be dispatched) or partner with nearby gyms and restaurants to offer incentives and discounts for healthy eating and exercise. Empowering drivers to find alternative ways to exercise and to increase fruit and vegetable consumption is important, as is changing the environment via policy changes.

Although we did not find differences by region of birth, prior work has suggested differences in the types of physical activity people choose (55). Approximately 60% of our sample consisted of South Asians (i.e., from India, Pakistan, or Bangladesh). South Asians tend to be less physically active than other groups (56, 57) but may actually require more exercise due to higher rates and risk of CVD (58). Culturally specific tailored physical activity interventions for South Asians have been proposed (56, 59), and researchers and program planners working with this population should consider further adapting these interventions so that they can be adopted by drivers with limited available time.

The best outcomes for smoking cessations occur with nicotine replacement therapy (NRT), behavioral counseling, and policy changes (22, 60). NRT and behavioral counseling are provided through the New York State Quitline and the New York City “Help Me Quit” app, although anecdotal evidence has indicated low success with these systems. There are a number of other behavioral counseling resources available for smoking cessation, including smokefree.gov and several empirically supported apps for smartphones (61), although more research is necessary. At the policy level, calling attention to these resources with targeted advertisement has been shown to be helpful; New York City is already seeing benefit to targeted smoking cessation resource ads to Chinese immigrants (62), suggesting potential advantages for tailored outreach with drivers. Our findings also suggest that stress management may be helpful for those with sleep disturbances and for those who smoke. However, as perceived stress was apparently not related to all health behaviors, this study shows that specific health behavior interventions remain necessary.

Conclusion

Our findings indicate that in the taxi driver sample, stress was a significant predictor of smoking and sleep disturbances. Overall, except for the relationship between smoking and sugar consumption, modifiable health behaviors did not generally cluster together or co-occur. However, poor nutrition and physical activity were pervasive concerns across our entire sample, and smoking rates exceeded the national rates for men of similar ethnicities. Demographic differences across stress, physical activity, and smoking were also found. These results have implications for intervention development and implementation. Overall, we propose a framework of specific changes, including targeted interventions for drivers coupled with policy and structural changes. In combination with input from community partners in garages and taxi cab community advisory boards, we recommend a swift rollout of interventions for fast and efficacious behavior change.

Table 2.

Regression Models Predicting Health Behaviors

	Stress		Moderate Physical Activity		Vigorous Physical Activity		Fruit Intake		Vegetable Intake		Sugar Intake		Sleep Quantity		Sleep disturbances		Smoking
Predictors	β	p	β	p	β	p	β	p	β	p	β	P	β	p	β	p	Exp (B)	P
R2	.093	.001	.080	.419	.209	.000	.087	.185	.025	.973	.090	.195	.131	.048	.329	.000
Age	−.124	.061	−.153	.078	−.267	.001	−.140	.088	−.025	.771	−.163	.057	−.087	.318	−.011	.885	1.006	.776
Vehicle ownership	−.109	.103	.113	.180	−.061	.432	−.114	.153	.021	.799	−.005	.953	.039	.639	.059	.405	.785	.614
Region of birth
South Asia	−.021	.816	.006	.962	.027	.803	−.053	.631	.009	.938	−.141	.191	.122	.281	−.188	.057	11.564	.032
Latin America	−.088	.272	.005	.959	.108	.264	.051	.600	−.059	.563	−.102	.293	.022	.832	−.094	.287	4.913	.233
Africa	.131	.100	−.025	.806	−.026	.788	−.007	.938	−.066	.509	.036	.714	−.107	.282	−.026	.762	1.661	.741
English proficiency	−.222	.001	.216	.010	.330	.000	.117	.162	−.009	.922	−.060	.483	.122	.157	−.094	.204	.961	.941
Stress			.117	.233	.056	.537	.121	.205	.022	.820	.119	.208	−.137	.096	.525	.000	.901	.013
Moderate physical activity							−.011	.891	.058	.468	−.046	.557	−.065	.409	−.039	.570	1.001	.508
Vigorous physical activity							−.001	.990	.053	.534	.099	.248	−.104	.223	.020	.786	.997	.235
Fruit			−.023	.777	−.013	.864							.072	.355	−.114	.092	1.233	.449
Vegetable			.055	.485	.052	.474							−.025	.741	.006	.929	.850	.445
Sugar			−.062	.454	.074	.333							.020	.800	−.006	.930	1.097	.010
Sleep quantity			−.101	.239	−.110	.167	.040	.621	−.006	.938	.040	.642					1.456	.087
Sleep disturbances			−.057	.559	−.015	.866	−.158	.101	−.027	.783	−.040	.679					1.027	.090
Smoking			.048	.565	−.088	.259	.097	.208	−.082	.305	.166	.037	.101	.212	.070	.324