movement to preparation) and providing social support (assisting movement to action and maintenance). Since consumption of 5 or more servings a day might ...
:;~ . ;.
. . e .,:rX f.~ ~. .,7-i':;
. , . *,~-.t .q~ . ~:'*.
. . ij;. . ~ ~. . . . . . j WINIE,
NEW
,
JS,.f~~~...... ... ...... .. .. ... ... .... . . . . . . .
"'
;;.
.... ,...........
.... .....
;5 ,a
Final Results of the Maryland WIC 5-A-Day Promotion Program
'*- ;'~~~~~~~~~~~~~~~... .. .
.';.l', . S. P; t
~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~.
.. :'':.'.......''....
..........
;. . . . . . . . . .
_i
Stephen Havas, MD, MPH, Jean Anliker, PhD, RD, Dorothy Damron, MS, Patricia Langenberg, PhD, Michael Ballesteros, MS, and Robert Feldman, PhD
Introduction
In 1998, an estimated 1 228 600 new of cancer will be diagnosed in the United States, and an estimated 564 800 US residents will die of cancer.' Cancer accounts for 23% of all deaths in the United States, of which 35% or more may be attributable to unhealthy diets.2 Various dietary components have been associated with increased incidence and mortality rates from cancer.35 The scientific rationale for increasing consumption of fruits and vegetables in order to prevent cancer is particularly compelling. Reviews of up to 200 studies reveal that the vast majority have found a strong association between higher intakes of fruits and vegetables and a reduced risk of many types of cancer.67 Based on this evidence, the National Cancer Institute (NCI) instituted the national 5-A-Day for Better Health Program (modeled after a California Department of Health Services program8) in 1991 to encourage Americans to eat 5 or more servings of fruits and vegetables daily.9 A 1991 national survey found a mean adult consumption level of 3.5 servings; only 23% of the population was consuming 5 or more servings. I The NCI also funded nine 5-A-Day research studies whose purpose was to develop, implement, and evaluate interventions in community-based settings to increase individuals' consumption of fruits and vegetables." One study, the Maryland WIC 5-A-Day Promotion Program, focused its interventions on women served by the Special Supplemental Nutrition Program for Women, Infants, and Children (WIC). The federally funded WIC program, involving approximately 7.1 million lowincome participants, operates in all 50 states. Low-income populations have a higher risk of cancer and often have poor diets. The 1986 Continuing Survey of Food Intakes by Indicases
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~......
~~~~~~ .. ...... ~~. ~
-August I998-, VL. 88, No. g
viduals reported that low-income women ate the fewest fruits and vegetables, with a mean intake of 2.9 servings.'2 The Second National Health and Nutrition Examination Survey (NHANES II) produced the same finding; 31% of low-income women reported consuming no fruit over 4 days, in comparison with 12% of high-income women, and lowincome women consumed the fewest vegetables.'3 A 1991 national survey also found that individuals at low income levels consumed the fewest fruits and vegetables.'0 Despite nutrition education and supplemental foods being offered, the diets of WIC participants are often inadequate as well.'4"15 One study assessing pregnant WIC participants found that adolescents consumed an average of 2.1 servings of fruits and vegetables each day and that women more than 35 years of age consumed an average of 2.6 servings.'4 Although widespread, the WIC program has been used infrequently as a setting for research. No research prior to ours has focused on nutrition education aimed at reducing WIC participants' risk of developing chronic diseases. In this paper we review the design of our project, present our final results, and discuss the significance of our findings.
Stephen Havas, Dorothy Damron, Patricia Langenberg, and Michael Ballesteros are with the Department of Epidemiology and Preventive Medicine, University of Maryland School of Medicine, Baltimore. Jean Anliker is with the Department of Nutrition and Food Science, and Robert Feldman is with the Department of Health Education, University of Maryland at College Park. Requests for reprints should be sent to Stephen Havas, MD, MPH, Department of Epidemiology and Preventive Medicine, University of Maryland School of Medicine, Baltimore, MD 21201 (e-mail: shavas(epi.umaryland.edu). This paper was accepted February 17, 1998.
American Joumal of Public Health 1161
Havas et al.
Methods Planning, Pretesting, and Pilot Testing Interventions Relatively little was known about our target audience's knowledge, attitudes, and practices concerning fruits and vegetables. Thus, we initially undertook audience research involving focus group discussions and central intercept interviews. 6 These procedures were designed to gain a better understanding of WIC participants' shopping, preparation, and eating practices concerning fiuits and vegetables, along with their perceptions of these foods, perceived barriers to increased consumption, and reactions to potential messages and interventions.'7 We next proceeded to a pilot test. We had not originally planned to implement such a test, but, after discovering that the WIC program had a very high participant turnover rate,'8 we realized that our planned yearlong intervention program would not be effective. We therefore developed a more intensive, 6-month intervention and pilot tested it before conducting the same program at a large number of sites.'9 Following the pilot test, we slightly refined the interventions (described subsequently) to boost recruitment, attendance, dietary change, and survey completion rates for the full-scale study.
Theoretical Underpinnings of the Program Our project was based on Prochaska and DiClemente's stage model of change.20 This model postulates that individual behavior change is a dynamic process that involves a series of stages: precontemplation, contemplation, preparation, action, and maintenance. Different intervention strategies are used at different stages to facilitate movement to a more advanced stage.
Design of the Program Using a randomized crossover design, we implemented a multidimensional program among women served by 16 WIC sites located in Baltimore City and 6 Maryland counties. Eight sites were randomized initially to intervention status and 8 to control status for phase 1 of the full-scale intervention. Recruitment of intervention and control participants then began at the 16 sites. Written informed consent was obtained from all parficipants under a protocol approved by the institutional review boards of both the University of Maryland, Baltimore, and the Maxyland Department of Health and Mental Hygiene. To be eligible, women had to be enrolled in the WIC program 1162 American Journal of Public Health
or have children enrolled; also, they were required to be at least 18 years of age and to have the intention of remaining enrolled at the site for at least 6 months. Four months after completion ofphase 1, intervention sites became control sites, and vice versa. Phase 2 recruitment then began. Because persons enrolled in phase 1 were ineligible to participate in phase 2 and only members of our staff conducted the interventions, no significant contamination effects occurred at the sites. Thus, we had 16 sites that served as controls for themselves. We designed the program to minimize disruption of WIC procedures and to maximize the limited opportunities available for reaching the WIC clientele. We hired peer educators and trained them to implement the program. These peer educators, most ofwhom were participants in the WIC program, were responsible for all contacts with participants. The intervention consisted of 3 components: (1) nutrition sessions conducted by peer educators, (2) printed materials and visual reminders, and (3) direct mail. The control site participants experienced the normal WIC program, which generally includes less than 10 minutes of nutrition education at the bimonthly voucher pickup. The intervention components were designed to actively involve participants and to help them progress through the stages of change. Recruitment messages provided information (assisting movement from precontemplation to contemplation); nutrition sessions focused on building skills (assisting movement to preparation) and providing social support (assisting movement to action and maintenance). Since consumption of 5 or more servings a day might be perceived by many as unrealistic, we emphasized eating more fruits and vegetables, rather than a specific number, and allowed participants to set their own goals. Our primary goal was for intervention participants to increase their consumption of fruits and vegetables by at least one half serving, a reasonable goal given the difficulty involved in achieving change in community intervention programs. A recent report indicates that such an increase, if maintained, could result in an 8% lower cancer incidence rate.2' Secondary goals included positive movement in terms of stage of change, improved attitudes, increased self-efficacy, and a decreased number of perceived barriers.
Nutrition Education The peer educators delivered 2 types of nutrition education to intervention participants: (1) brief messages regarding increasing fruit and vegetable consumption given at
the time of enrollment in our program and (2) a series of 3 group discussion sessions. Over the 6-month intervention period, the peer educators led 3 nutrition education sessions, each lasting about 45 minutes. Participants were encouraged to attend all 3 smallgroup sessions, which usually took place immediately before the regular bimonthly voucher distribution days. Child care was usually provided by another peer educator. The sessions were designed to actively involve participants in discussing their views, redefining those views, and setting personal goals to reflect changes.2" The first session focused on self-assessment, the value of eating fruits and vegetables, and personal goal setting for eating more fruits and vegetables. The second session focused on identifying and overcoming perceived barriers. The third session stressed maintenance strategies. Each session included a food demonstration (e.g., a vegetable stir-fly) to build participants' skills and to allow them to try new foods.
Printed Materials and risual Reminders The focal piece for the peer-led group discussions was a photonovella, a colorful, illustrated booklet with a story line. It served as each participant's guidebook for writing thoughts about fruits and vegetables, setting goals, choosing behavioral strategies to achieve those goals, and identifying and overcoming barriers. During the nutrition sessions, peer educators guided participants through exercises in the photonovella, facilitated discussion, and provided social support to help participants achieve their goals. Our program used a series of 5 clue cards to stimulate interest and an exchange of ideas for eating more fruits and vegetables; most of the cards were mailed to participants prior to the nutrition sessions. Each clue card posed a question related to a specific behavior that we were promoting (e.g., 'What is a quick, easy way to combine different vegetables for dinner?'). Participants were asked to write their ideas on the back of the clue card and bring it to the next session, where it was used as a focus for discussion and as an introduction to the food demonstration. A tip sheet with additional ideas and recipes relating to the behavior in question was distributed after the discussion. Other materials included a booklet of recipes submitted by participants, a children's activity book focused on fruits and vegetables, a videotape showing children singing about fruits and vegetables, and a refrigerator magnet with the program's logo. Calendar reminder sheets containing information about future nutrition sessions were August 1998, Vol. 88, No. 8
Maryland WIC 5-A-Day Program
distributed during the first session. Participants were encouraged to attach stickers to their calendar sheets each day they reached their goal. Attractive posters of fruits and vegetables were prominently displayed during all sessions.
sisted over time. This survey was administered 1 year following completion of the phase 1 postsurvey. The survey was conducted on-site for those still enrolled in the WIC program, and then by mail; a final attempt was made via a similar, structured telephone survey.
Direct Mail
Statistical Analyses Since we had limited opportunities for personal contact with WIC clients, we also used direct mail as an intervention. Over 6 months, the peer educators sent participants 4 different, tailored letters, each accompanied by a tip sheet and a clue card. The letters were tailored according to pregnancy status, baseline stage of change, attendance at nutrition sessions, and individual goals for eating more fruits and vegetables.
Evaluation In each phase, all intervention and control site participants completed a self-administered survey upon enrollment (copies of the survey are available from the first author). We attempted to have all intervention and control participants complete an identical postintervention survey (hereafter called the "postsurvey") 2 months after the last nutrition session in each phase. These surveys included questions about factors potentially related to fruit and vegetable intake, including demographics, smoking, stage of change, knowledge, attitudes, self-efficacy, social support, responsibility for food shopping and preparation, and perceived barriers to increasing consumption; scales were used to assess the latter 5 factors. Cronbach ot values for the survey's 5 scales ranged from .80 to .92, indicating high levels of internal response consistency.23 As a means of assessing fruit and vegetable consumption over the previous month, 7 summary questions were used in all NCI-sponsored national surveys and 5-ADay research projects.10"' We pretested the survey using the cognitive interview technique, a think-aloud interview in which the client is asked to read each question aloud and then think aloud when answering the
question.24'5 To maximize completion of the postsurvey, we attempted to reach the participants in 3 different ways. Initially, we sought to reach them at voucher pickup. If that opportunity was missed, we sought to reach them at another time at the WIC site or in the community. If that also failed, we mailed them a survey. Participants were offered $10 as an incentive for completing the postsurvey. We also conducted a follow-up survey of all phase 1 intervention and control participants to assess whether our findings per-
August 1998, Vol. 88, No. 8
Sample size calculations in this crossover design for number of sites and participants within sites were based on the ability to detect a difference of 0.5 servings between the intervention and control groups. We assumed an intrasite correlation of 0.2 and a pretest-posttest correlation of 0.11; between and within error variances were based on NHANES II data provided by NCI.26 Data from all surveys were analyzed via intention-to-treat analyses.27 Intention-to-treat analysis requires that anyone who enters a trial be included in the final analysis. For any participant in either the intervention or control group who did not complete the final survey, we used the "last value carried forward" method; that is, final survey values for noncompleters were assumed to be equal to their corresponding values on the baseline survey. This assumption that no change had occurred in any variable value was reasonable given the short time frame and lack of strong secular trends. Intention-to-treat analysis results in lower mean change estimates than an analysis that includes only survey completers. Its findings are less likely to be biased, and it provides a better assessment of the public health impact of an intervention. This paper presents data for 15 of the 16 sites in our program. Because the peer educator at 1 of the sites did not follow our quality control guidelines during the control phase, data from that site for both the intervention and control phases were excluded from our analyses. Inclusion of these data would not have changed the statistical significance of any of the findings presented
subsequently. Intervention and control groups were compared initially on demographic and other baseline characteristics by means of paired t tests or Pearson x2 tests. Differences were considered statistically significant when P values of .05 or less were obtained. Analyses of the dietary intake data were based on change between baseline and 8 months. For 1-year follow-up data (phase 1 only), changes from baseline and 8 months were examined. Comparisons between the intervention and control participants (within site) on change in individual consumption and other outcomes were made at both the site and individual levels. In terms of an overall esti-
mate of intervention effect, site may be the most appropriate unit of analysis. The reason is that sites, not individuals, were randomized, and many but not all aspects of the intervention were delivered within sites.2829 Analyses at the site level were based on either site means (for continuous variables) or site proportions (for dichotomous variables), both treated as continuous in analyses of the 15 sites. Paired t tests were used to compare intervention and control groups on mean change within site in terms of intake, attitude, self-efficacy, knowledge, and social support scores. Individual-based analyses were also carried out. Individuals were used as the unit of analysis, but site was included as a random effect, as recommended by Murray.30 Results were similar to those of site-based analyses and are not presented. All analyses were performed with SAS, version 6.11 (SAS Institute Inc, Cary, NC).
Results Recruitment Recruitment took place at the WIC sites during voucher pickup and certification. About 85% of women met our eligibility criteria. Overall, the acceptance rates were 66% (range: 55% to 85%) during the intervention phases and 87% (range: 68% to 100%) during the control phases.
Demographics We deliberately chose sites with diverse clientele to enhance the generalizability of our findings. As a result, the demographic characteristics of participants at the various sites were quite different (Table 1). Overall, the majority were Black, less than 30 years of age, single, and unemployed. Slightly higher percentages of intervention participants than control participants were Black and/or on food stamps, while a lower percentage worked.
Attendance Attendance at the nutrition sessions varied considerably by site. Overall, 19% (range: 8% to 31%) attended all 3 sessions, 14% (range: 9% to 21%) attended 2 sessions, 20% (range: 15% to 27%) attended 1 session, and 46% (range: 31% to 58%) attended no sessions.
Completion of the Postsurvey and 1-Year Follow-Up Survey Overall, 75% of intervention participants completed the postsurvey (range: 60% American Journal of Public Health
1163
Havas et al.
TABLE 1-Demographic Characteristics of Maryland WIC 5-A-Day Promotion Program Participants Intervention (n=1443), % (Range)
Control (n=1679), % (Range) Age, y 18-24 25-29 30+ Race Black White Other Education Less than high school High school graduate More than high school
Other Marital status Single Married
Separated/divorced/widowed Currently employed Currently in school Current smoker On food stamps Pregnant Breast-feeding
pa
41.3 (27.1-55.5) 26.7 (19.7-36.7) 32.0 (20.2-50.0)
39.7 (22.7-50.8) 26.3 (19.2-29.6) 34.0 (24.4-52.1)
.468
53.2 (8.3-93.7) 43.1 (5.6-87.9) 3.8 (0.8-11.3)
58.2 (5.3-96.7) 38.2 (2.2-93.4) 3.6 (0.0-12.6)
.019
19.2 (6.5-45.3) 40.9 (20.0-46.6) 36.9 (11.7-68.9) 3.0 (0.0-7.3)
19.8 40.7 36.1 3.4
53.3 (36.6-81.1) 30.8 (8.8-46.0) 15.9 (9.5-23.9) 35.4 (24.2-50.9) 17.6 (12.0-24.1) 27.3 (8.5-47.4) 42.1 (22.7-66.4) 22.4 (3.7-55.3) 10.4 (5.8-18.9)
56.9 (38.9-83.3) 29.3 (6.7-50.0) 13.9 (6.4-20.6) 28.7 (16.7-40.7) 17.2 (8.7-23.9) 26.0 (15.4-46.7) 45.5 (26.5-89.1) 22.2 (5.6-29.8) 11.1 (3.3-20.2)
(7.1-41.3) (33.0-50.0) (18.2-52.3) (0.8-8.3)
.875
.110
