To examine the relationship between smoking and work-family conflict among a sample of New England long-term care facility workers.
Data were collected using in-person, structured interviews from workers in four extended care facilities.
There was a strong association between smoking likelihood and work-family conflict. Workers who experienced both stress at home from work issues (i.e., work to home conflict) and stress and work from personal issues (i.e., home to work conflict) had 3.1 times higher odds of smoking compared to those who did not experience these types of conflict. Workers who experienced home to work conflict had an odds of 2.3 compared to those who did not experience this type of conflict, and workers who experienced work to home conflict had an odds of 1.6 compared to workers who did not experience this type of conflict.
The results of this study indicate that there is a robust relationship between work-family conflict and smoking, but that this relationship is dependent upon the total amount of conflict experienced and the direction of the conflict.
Tobacco use is the foremost cause of preventable death and illness in the United States. Tobacco use, primarily in the form of cigarette smoking, is responsible for 1 in 5 deaths, or about 440,000 Americans every year,
“Work-family conflict” refers to the expectations, demands, skills, or knowledge associated with one domain (e.g., work) affecting the other domain (e.g., family); with the term “conflict” implying that the two domains compete for the individual's time and energy in a negative interaction.
Finally, much of the work performed by workers in long-term care facilities is both physically and emotionally demanding, provides relatively low wages, and is likely to be associated with adverse health consequences.
This paper examines the relationship between smoking and work-family conflict among workers in four New England long-term care facilities. We assess both the direction of work-family conflict (i.e., work-to-home versus home-to-work) and the effects of overall conflict (i.e., experiencing both forms of conflict), as we hypothesized that experiencing conflict in both directions may influence smoking likelihood more than experiencing conflict in one direction alone. We also examine the contributions of work and home factors as potential confounders since each may be separately associated with both smoking and work-family conflict.
The cross-sectional data examined in this study were collected in four extended care facilities located in the Boston Metro area. This research was part of a larger study that examined how workplace policies, practices, and attitudes influence the cardiovascular health of employees. This setting was chosen because many employees of extended care facilities earn lower wages, are racially and ethnically diverse, and experience high levels of job strain.
Smoking status was assessed with the question “Do you smoke cigarettes every day, some days, or not at all?” This three-option response was dichotomized into yes or no by combining the “every day” and “some days” responses into “yes” and “not at all” into “no.” We chose to dichotomize this variable because we were interested in smoking status and, for this study, less concerned about smoking intensity.
Work-family conflict was assessed with the following statements: “In the last month, I was preoccupied with my work while I was at home.” (work-to-home conflict) and “In the last month, I was preoccupied with personal responsibilities while I was at work.” (home-to-work conflict) Both items were adapted from standard measures.
To assess the effect of bi-directional work-family conflict (i.e., experiencing both home-to-work and work-to-home conflict), a third variable was created out of the two directional work-family conflict variables. This variable had three possible response categories. Participants who reported experiencing either type of conflict rarely or never were coded as having “no conflict at all”, participants who reported experiencing one type of conflict often or sometimes, but the other type of conflict rarely or never were coded as experiencing “uni-directional conflict”, and participants who reported experiencing both types of conflict often or sometimes were coded as experiencing “bi-directional conflict”. These cut-points were chosen because we were interested in investigating the relationship of work-family conflict, either in both directions or regardless of direction, to tobacco use.
Individual characteristics that were assessed include: gender, age, race/ethnicity, nativity, level of education, annual household income, and alcohol consumption. Standard measures were used to assess gender, age, and race/ethnicity. Race/ethnicity was transformed into a dichotomous variable that represented white race/ethnicity versus all other race/ethnicity categories. Nativity was assessed with the question: “In what country were you born?” Responses for any country other than the United States were coded as “foreign born” and those born inside the US were coded as “native born”. Level of education was assessed with the question, “How much schooling have you had?” The response categories were combined into a dichotomous variable: four-year college graduate and graduate degree versus those with less education. To assess income, respondents were asked about their yearly household income from all sources. Response categories were collapsed into the following 4 categories: <$30,000, $30,000–$49,999, $50,000–$69,999, and $70,000 or more. To assess alcohol consumption, participants were asked the average number of days they had alcohol and, on those days, the average number of drinks they consumed. Responses were combined to reflect average number of alcoholic drinks per day.
The following workplace factors were assessed: job control/demand attributes, shift worked, job flexibility, total hours worked per week, and occupation. Control/demand job attributes were assessed with a 12-item questionnaire.
The following home/family factors were assessed: whether or not the participant was married or living with a partner, the number of people that were supported by that individual's income, and how many children 18 or younger were living at home. To assess marital status and number of children living at home, participants were asked, first, how many people lived with them, and, second, to give specific information about each household member. This information was summarized into a) marital status/living with a partner, and b) number of children living at home. To assess the number supported by the respondent's income, each participant was asked, how many people are currently supported by their income. Responses to this item were grouped into the following categories: 1 person, 2 people, and 3 or more people.
To investigate the association between work-family conflict and smoking likelihood, we built three separate logistic regression models, one model for each of the three work-family conflict variables. The reference category for home-to-work conflict was “no home-to-work conflict”, the reference category for work-to-home conflict was “no work-to-home conflict”, and the reference category for overall conflict was “no conflict at all” (with uni-directional conflict and bi-directional conflict as two categories that were compared to the reference category). We chose to create three separate models (one for work-to-home conflict, one for home-to-work conflict, and another for overall conflict) because, based upon the work-family conflict literature, the direction of the conflict is important and is almost always considered separately.
The following variables were considered as potential confounders in the relationship between smoking and work-family conflict: age, education, gender, nativity, race/ethnicity, alcohol use, marital status, number of children, annual household income, number of people supported by income, Karasek job control/demand attributes, occupation, job flexibility, hours worked per week, and work shift. If inclusion of that variable modified the beta coefficient by more than 5%, it was included as a covariate in the model. Thus, the following variables were included as covariates in all models: education, nativity, race/ethnicity, number of children, number supported by income, Karasek job control/demand attributes, and work shift. Further, all models were adjusted for the fixed effect of the work site by including three `dummy' variables to represent the four work sites in each model (work site refers to the specific extended care facility that the employee worked at).
A total of 439 subjects were included in the analysis. Participants were mostly female (82.5%) and older, as about half (49.7%) were between 40 and 64 years old (see
The prevalence of smoking was 19.3% (86/439) in this sample of nursing home workers. Among those who smoked 23.3% reported experiencing neither work-to-family nor home-to-work conflict, 43% reported that they experience conflict in one direction (i.e., either work-to-home or home-to-work) but not in both directions, and 33.7% reported experiencing conflict sometimes or often in both directions. Among those who smoked 65.1% worried about personal matters while at work (i.e. home-to-work conflict), and 45.3% worried about work while at home (i.e., work-to-home conflict).
The unadjusted relationship between work-family conflict and smoking likelihood was assessed first, then three separate multivariate logistic regression models were created to further investigate the relationship. Each model adjusted for the effects of education, nativity, and race/ethnicity, number of children, number supported by income, job control and demand, work shift, and worksite.
The odds of smoking for those who reported experiencing only one type of conflict was 1.54 (95% CI, 0.85–2.78) compared no conflict at all, and the odds of smoking for those who reported experiencing both types of conflict was 2.11 (95% CI, 1.12–3.97) compared to no conflict at all. The odds of smoking for participants who reported experiencing home-to-work conflict, compared to those who reported no such conflict, was 1.90 (95% CI, 1.16–3.10). The odds of smoking for those who reported work-to-home conflict, compared those who reported no such conflict was 1.26 (95% CI, 0.79–2.03).
The first logistic regression model that controlled for all relevant confounders assessed the relationship between the 3-level directional work-family conflict variable and smoking likelihood (see
The second logistic regression model assessed the relationship between home-to-work conflict (i.e., feeling preoccupied with personal responsibilities while at work) and smoking likelihood, controlling for all relevant confounders. The results of this model indicated that, on average, for participants who experienced home-to-work conflict, the odds of smoking was 2.3 compared to participants who reported no such conflict (95% CI: 1.31–4.10).
The final model assessed the relationship between work-to-home conflict (i.e., feeling preoccupied with work while at home) and smoking likelihood. This model indicated that, on average, for participants who experienced work-to-home conflict, the odds of smoking was 1.55 compared to those who reported no such conflict (95% CI: 0.89–2.69).
The purpose of this study was to investigate the relationship between work-family conflict and smoking among a sample of nursing home workers. We found that work-family conflict does significantly contribute to smoking likelihood and that when this conflict is experienced in both directions (i.e., being preoccupied with personal matters while at work and preoccupied with work while at home), the influence on smoking is the greatest. This finding indicates a possible gradient. Such that, when work-family conflict is experienced in both directions (i.e., from home-to-work and from work-to-home), it increases the likelihood of smoking more than experiencing one type of conflict alone.
When each direction of conflict was examined separately, we found that conflict has differential effects on smoking likelihood depending on the direction of the conflict. We discovered a robust relationship between home-to-work conflict and smoking likelihood, as participants who were preoccupied with personal matters while at work were significantly more likely to smoke, but that the relationship between smoking and work-to-home conflict (i.e., being preoccupied with work while at home) was much weaker. This finding highlights the importance of directionality in the smoking/work-family conflict relationship. Most research that deals with the work-family interface addresses the bi-directional nature of conflict by regarding work-to-home conflict as separate and distinct from home-to-work conflict, and generally considers work-to-home conflict and home-to-work conflict as two theoretically separate and distinct concepts.
As there is a literature that reports an association between smoking and workplace factors, including policies such as smoking bans,
A key strength of this study is the ability to look at the relationship between work-family conflict and tobacco use, as this important topic has received little attention in either the public health or the sociology literatures. Other strengths of our study include a high response rate, the inclusion of a very diverse group of predominantly low-wage workers, and a clear gradient in the results. A limitation of this research is its cross-sectional design, and because there is no time component in the design, it is impossible to attribute a causal relationship to the associations between work-family conflict and smoking. It is possible that smoking is causing the work-family conflict. The economic burden and health consequences of smoking can contribute to increased stress at home, as tobacco use can significantly affect family budgets, due to the cost of cigarettes and the increased cost of healthcare that result from tobacco use.
Tobacco-control policies have heavily influenced smoking rates in the US. The findings of this study show a relationship between smoking and work-family conflict and identifies a possibly fruitful area for tobacco intervention and control as well as workplace policies related to reducing work-family conflict. However, because there is very little research on this topic, and the research that has been conducted uses small samples and cross-sectional designs, the next step in the smoking/work-family conflict link is replication in studies with larger samples and study-designs that contain a time component. Should the findings presented in this study be replicated, there is potential for workplace policies and programs to alleviate work-family conflict and by doing so decrease smoking rates among their workers.
This research was conducted as part of the Work, Family and Health Network (
This study was approved by the Institutional Review Board at the Dana Farber Cancer Institute, Boston, MA.
STATEMENT OF CONTRIBUTIONS All of the authors have made a substantial and direct intellectual contribution to the manuscript. Dr. Nelson conceptualized the study, performed the analysis, and wrote the manuscript. Dr. Li contributed to the analysis plan, provided feedback during previous iterations of the analysis and aided in the interpretation of results. Dr. Sorensen contributed to the conceptualization of the study, the interpretation of results, and provided comments on previous drafts of the manuscript. Dr. Berkman contributed to the conceptualization of the study, the interpretation of results, and provided comments on previous drafts of the manuscript. All authors approved the final version of the manuscript.
Participant Characteristics by Exposure Category
| Overall | No Conflict | Uni-directional Conflict | Bi-directional Conflict | ||||
|---|---|---|---|---|---|---|---|
| (n=439) | (n=142) | (n=184) | (n=113) | ||||
| n | n | % | n | % | n | % | |
| Smokers | 86 | 20 | 23.3% | 37 | 43.0% | 29 | 33.7% |
|
| |||||||
| Gender | |||||||
| Male | 77 | 29 | 37.7% | 25 | 32.5% | 23 | 29.9% |
| Female | 362 | 113 | 31.2% | 159 | 43.9% | 90 | 24.9% |
|
| |||||||
| Age | |||||||
| 18–25 yrs | 65 | 24 | 36.9% | 32 | 49.2% | 9 | 13.9% |
| 26–39 yrs | 137 | 38 | 27.7% | 66 | 48.2% | 33 | 24.1% |
| 40–64 yrs | 218 | 72 | 33.0% | 79 | 36.2% | 67 | 30.7% |
| 65+ | 18 | 8 | 44.4% | 6 | 33.3% | 4 | 22.2% |
|
| |||||||
| Race/ethncity | |||||||
| Non-Hispanic black | 156 | 49 | 31.4% | 55 | 35.3% | 52 | 33.3% |
| Non-Hispanic white | 186 | 57 | 30.6% | 87 | 46.8% | 42 | 22.6% |
| Non-Hispanic other | 64 | 26 | 40.6% | 26 | 40.6% | 12 | 18.8% |
| Hispanic | 33 | 10 | 30.3% | 16 | 48.5% | 7 | 21.2% |
|
| |||||||
| Nativity | |||||||
| Foreign-born | 242 | 82 | 33.9% | 93 | 38.4% | 67 | 27.7% |
| US-born | 197 | 60 | 30.5% | 91 | 46.2% | 46 | 23.4% |
|
| |||||||
| Education | |||||||
| < High School | 67 | 24 | 35.8% | 22 | 32.8% | 21 | 31.3% |
| High School/GED | 129 | 44 | 34.1% | 57 | 44.2% | 28 | 21.7% |
| Some College | 174 | 48 | 27.6% | 77 | 44.3% | 49 | 28.2% |
| College Grad/Graduate degree | 69 | 26 | 37.7% | 28 | 40.6% | 15 | 21.7% |
|
| |||||||
| Annual Household Income | |||||||
| < $30,000 | 92 | 27 | 29.4% | 36 | 39.1% | 29 | 31.5% |
| $30,000–$49,000 | 107 | 36 | 33.6% | 47 | 43.9% | 24 | 22.4% |
| $50,000–369,000 | 73 | 28 | 38.4% | 32 | 43.8% | 13 | 17.8% |
| $70,000+ | 122 | 37 | 30.3% | 53 | 43.4% | 32 | 26.2% |
|
| |||||||
| Married | 247 | 81 | 32.8% | 99 | 40.1% | 67 | 27.1% |
|
| |||||||
| Number supported by income | |||||||
| 1 person | 44 | 18 | 40.9% | 16 | 36.4% | 10 | 22.7% |
| 2 people | 91 | 32 | 35.2% | 41 | 45.1% | 18 | 19.8% |
| 3 or more people | 304 | 92 | 30.3% | 127 | 41.8% | 85 | 28.0% |
|
| |||||||
| Children 18 or younger | |||||||
| 0 | 211 | 78 | 37.0% | 86 | 40.8% | 47 | 22.3% |
| 1 | 111 | 25 | 22.5% | 56 | 50.5% | 30 | 27.0% |
| 2 | 88 | 33 | 37.5% | 31 | 35.2% | 24 | 27.3% |
| 3+ | 29 | 6 | 20.7% | 11 | 37.9% | 12 | 41.4% |
|
| |||||||
| Job Control/Demand Category | |||||||
| Low Strain | 136 | 50 | 36.8% | 57 | 41.9% | 29 | 21.3% |
| High Strain | 111 | 28 | 25.2% | 52 | 46.9% | 31 | 27.9% |
| Passive | 85 | 29 | 34.1% | 32 | 37.7% | 24 | 28.2% |
| Active | 107 | 35 | 32.7% | 43 | 40.2% | 29 | 27.1% |
|
| |||||||
| Shift-Worked | |||||||
| Day | 270 | 90 | 33.3% | 112 | 41.5% | 68 | 25.2% |
| Evening | 101 | 27 | 26.7% | 49 | 48.5% | 25 | 24.8% |
| Night | 68 | 25 | 36.8% | 23 | 33.8% | 20 | 29.4% |
Logistic Regression: Odds Ratios (OR) and 95% Confidence Intervals for Smoking Likelihood
| Experiences Uni- or Bi-Directional | Experiences Home to Work Conflict | Experiences Work to Home Conflict | |||
|---|---|---|---|---|---|
| (n=439) | OR (95% CI) | (n=439) | OR (95% CI) | (n=439) | OR (95% CI) |
| No Conflict at all (ref.) | 1.00 | No (ref.) | 1.00 | No (ref.) | 1.00 |
| Uni-directional Conflict | 1.46 (0.75–2.85) | Yes | 2.32 (1.31–4.10) | Yes | 1.55 (0.89–2.69) |
| Bi-directional Conflict | 3.11 (1.48–6.56) | ||||
All models include the following covariates: education, nativity, race/ethnicity, number of children, number supported by income, Karasek job control/demand attributes, work shift, and work site.
The term `uni-directional conflict' designates the category of participants who experience either work-to-home conflict or home-to-work conflict, but not both. The term `bi-directional conflict' designates the category of participants who experience both work-to-home conflict and home-to-work conflict.