The background survey, conducted from May 2001 through November 2002, provided population-based data, including social contextual factors and sociodemographic data. LIUNA provided contact information for 3,239 members randomly selected from the union roster. Of those, 1,005 could not be contacted. Members were eligible to participate if their membership was current, they were not retired or on disability, were working in construction, and could complete the survey in English or Spanish. Researchers reached 2,234 workers, of whom 477 (21%) were ineligible to participate because they did not meet the inclusion criteria. Assuming 21% (215) of those not contacted were ineligible, we estimate the total number of ineligibles to be 692. Thus, we attempted to contact 2,547 eligible workers.

A total of 1,108 of the 2,547 eligible respondents completed the survey (44% response rate). To calculate the response rate to the background survey, we assumed that 21% (215) of those we were unable to contact were also ineligible (477/2,234 = 21%). Thus, we estimate the total number of ineligibles to be 692 (477 + 215) and the total eligible to contact to be 2,547 (3,239 − 692). The response rate among eligible members was 44% (1,108/2,547). We collected 1,109 surveys; 1 survey was excluded because of incomplete data.

Sociodemographic variables including sex, race/ethnicity, educational level, and native language were measured. We also assessed work/family spillover, a term used to describe work roles affecting participation in family roles. We assessed this construct by using a 6-item scale (Cronbach's α = 0.74) (21). In this scale, respondents were asked to rate their agreement with statements about their work negatively affecting their family (eg, missing meals or feeling distracted at home because of work).

We invited the 1,108 background survey respondents to participate in the randomized trial from November 2002 through December 2003. Of the 915 that remained eligible (194 retired or had lapsed memberships), 673 (74%) consented to participate, completed the baseline survey, and were randomized into the study. Six months later, 582 participants (86%) completed the final survey. Data collection was completed in July 2004.

Cigarette use was determined by self-report on the basis of recommendations by the Society for Research on Nicotine and Tobacco (21). Smoking cessation was measured by the prevalence of smokers and recent quitters (in the last 6 months) at baseline who reported not smoking on the final survey. Fruit and vegetable consumption was measured with a screening tool used in the National Cancer Institute's 5-A-Day for Better Health projects (22). Correlations between intake measured by 24-hour diet recall and the all-day screener have been estimated at 0.66 for men and 0.51 for women (23). The following food groups were included: 100% fruit juice, lettuce salads, fried potatoes, other white potatoes, cooked dried beans, fruits, and other vegetables. Total servings of fruits and vegetables per day were computed by summing the servings per day of each, excluding fried potatoes. Additional variables measured on the baseline survey included marital status, number of minor children in the household, intention to change fruit and vegetable consumption, and whether eating at work helps cope with stress.

We used data from the 300 participants who were randomized to the intervention group and completed the final survey. Analysis began with the creation of a new smoking status variable with 3 categories: nonsmoker (nonsmoker at baseline and final), smoker (smoker at final irrespective of baseline smoking status), and quitter (smoker at baseline, not final). We then examined the relationship between this smoking status variable and change in fruit and vegetable consumption by using analysis of covariance. To explore potential confounders to this relationship, we examined relationships between change in fruit and vegetable consumption and variables including sex, race/ethnicity, education, native language, living with a partner, children in household, work/family spillover, work hours, intention to increase fruit and vegetable consumption, and eating at work to cope with stress, controlling for baseline consumption.

We then constructed a multivariable model examining the relationship between smoking status and change in fruit and vegetable consumption, controlling for all variables individually associated with change in consumption. Significance was set at P < .05. We added all variables into the model as a group. The final model assessed the relationship between smoking status and change in fruit and vegetable consumption, controlling for baseline consumption and all potential confounding variables individually associated with change in consumption. Analyses were conducted by using SAS statistical software version 9.1 (SAS Institute, Inc, Cary, North Carolina).

Discussion groups were originally conducted during the formative phase of TFH (November 2001-April 2002) (24). Because each discussion group included discussion about smoking and fruit and vegetable consumption, we examined these data to help interpret the survey findings by gaining insight into the links between these 2 behaviors from the workers' perspective.

We recruited a nationally representative sample of LIUNA members, through LIUNA's nationwide training centers, using purposive stratified sampling (25). The sampling strategy was constructed to ensure variation in the overall sample on the following variables: smoking status, sex, race/ethnicity, and geographic region. The final sample included 88 participants in 16 discussion groups (median number of participants per group, 5; range, 3-9).

We designed 2 semistructured topic guides to facilitate the groups. The first was used to explore members' perceptions of being a laborer, the union, work relationships, and health-related concerns. The second, used after prototype intervention materials were developed, was used to elicit participants' reactions to these materials. All sessions were recorded and transcribed. Although the primary purpose of the discussion groups was formative, both topic guides elicited discussion around experiences with smoking and healthy eating, yielding data relevant to this investigation.

The purpose of the group analysis was to explore relationships between smoking cessation and fruit and vegetable intake. Data analysis began by reading each transcript. Codes were applied to transcripts in QSR NVIVO version 1.2 (QSR International, Inc, Cambridge, Massachusetts) by using a 2-stage coding process: deductive coding followed by thematic (inductive) coding. For deductive coding, a codebook was developed to classify data into 3 topics: diet, smoking, and experiences/beliefs about behavior change. Data in each segment were further analyzed by using thematic coding, which explored themes that linked dietary habits with smoking habits or behavior change experiences or beliefs that linked the 2 behaviors. No a priori codebook was used during this stage. The first author (AH) assigned codes to emergent themes, queried codes, and reviewed and summarized the data.

Participants were mostly male, non-Hispanic whites, and had completed a high school diploma or equivalent (Table 1). Participants varied by the smoking status groups on 2 factors: native language and intention for changing fruit and vegetable consumption. A comparison in language groups suggests that more participants whose first language was not English were nonsmokers (86%) than were their peers whose first language was English (63%). Nonsmokers also reported higher levels of intention to increase fruit and vegetable consumption than did smokers at baseline.

The relationship between smoking status and change in fruit and vegetable consumption controlling for baseline intake was significant (P = .02) (data not shown). Nonsmokers and quitters showed an average increase in fruit and vegetable consumption of 1.92 servings (95% confidence interval [CI], 1.37-2.45) and 2.47 (95% CI, 0.75-4.19) compared with smokers, who increased their mean fruit and vegetable consumption by 0.58 servings (95% CI, −0.22 to 1.38).

We examined the association of each potential confounding variable with mean change in fruit and vegetable intake, controlling for baseline intake (Table 2). Sex, education, native language, eating at work to cope with stress, and intention to change fruit and vegetable consumption exhibited relationships with change in fruit and vegetable consumption. We compared the multivariable model containing smoking status with the significant variables (Table 2), controlling for baseline intake (Table 3). When other variables were controlled, fruit and vegetable intention and smoking status remained significant. The R ² for this model was 0.15.

In the multivariable model, we suspected that fruit and vegetable intention might be over-controlling, so we removed it. As in the full model, smoking status remained a significant predictor of change in fruit and vegetable consumption. Overall, the relationships with the remaining variables changed little and the R ² decreased to 0.12.

Like the intervention participants, the 88 discussion group participants were primarily white non-Hispanic (64%), men (84%), and reported completing high school or equivalent (54%). Seventy-eight percent of discussion group participants identified as current smokers. They ranged in age from 20 to 63 years (24). Four themes emerged that potentially linked smoking cessation with increased fruit and vegetable consumption: 1) concern for overall health, 2) concern about weight gain, 3) taste preferences, and 4) behavioral compensation, specifically, the need for quitters to keep their hands busy.

Group participants presented improving overall health as a reason for wanting to quit smoking and improve eating habits. In response to questions about why participants might want to eat more fruits and vegetables, they said, "Eating better will make you healthier," and "It's healthier food than anything, even the meats."  Responses to parallel questions about why participants wanted to or had quit smoking included, "I quit because I know it's bad for my health," and "That's why I decided I needed to let [smoking] go because it wasn't making my health no better." One explanation for why quitters changed both smoking behaviors and fruit and vegetable consumption during the same intervention is that changing both behaviors targets overall health improvement.

The second theme linking smoking cessation with fruit and vegetable consumption was concern with weight gain during smoking cessation. In discussions about desire to quit smoking, workers mentioned being worried about gaining weight if they quit. Participants' statements included, "Everybody that quits smoking gains weight," and

Another theme that linked smoking with fruit and vegetable consumption among those with previous quit attempts was that food tasted better during the quit attempt. Participants said, "Food tastes better," "You can't even taste food when you smoke. When you quit smoking for 2 or 3 weeks, you can taste the food better already. Cigarettes take away taste," and "Food even tastes completely different when you quit." This theme may indicate that when someone is not smoking, improvement in fruit and vegetable taste may promote increased consumption.

The fourth theme that linked smoking with fruit and vegetable consumption was behavioral compensation. Discussion group participants discussed needing something to do with their hands during quit attempts. Sample comments include, "That's part of the whole pattern, is the handling," "You quit smoking. What do you do with your hands?" and "I don't smoke for the nicotine. I smoke to have something to do with my hands." Because fruit and vegetable consumption requires handling, especially preparing and eating raw fruits and vegetables, increasing consumption might have offered a substitute for keeping a smoker's hands busy during smoking cessation.