This is an open-access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in the Journal of Medical Internet Research, is properly cited. The complete bibliographic information, a link to the original publication on http://www.jmir.org/, as well as this copyright and license information must be included.
The HealthValues Healthy Eating Programme is a standalone Internet-based intervention that employs a novel strategy for promoting behavior change (analyzing one’s reasons for endorsing health values) alongside other psychological principles that have been shown to influence behavior. The program consists of phases targeting motivation (dietary feedback and advice, analyzing reasons for health values, thinking about health-related desires, and concerns), volition (implementation intentions with mental contrasting), and maintenance (reviewing tasks, weekly tips).
The aim was to examine the effects of the program on consumption of fruit and vegetables, saturated fat, and added sugar over a 6-month period.
A total of 82 females and 18 males were recruited using both online and print advertisements in the local community. They were allocated to an intervention or control group using a stratified block randomization protocol. The program was designed such that participants logged onto a website every week for 24 weeks and completed health-related measures. Those allocated to the intervention group also completed the intervention tasks at these sessions. Additionally, all participants attended laboratory sessions at baseline, 3 months, and 6 months. During these sessions, participants completed a food frequency questionnaire (FFQ, the Block Fat/Sugar/Fruit/Vegetable Screener, adapted for the UK), and researchers (blind to group allocation) measured their body mass index (BMI), waist-to-hip ratio (WHR), and heart rate variability (HRV).
Data were analyzed using a series of ANOVA models. Per protocol analysis (n=92) showed a significant interaction for fruit and vegetable consumption (
Results suggest that the program helped individuals to increase their consumption of fruit and vegetables and to sustain this over a 6-month period. The observed reduction in fat and sugar intake suggests that monitoring behaviors over time is effective, although further research is needed to confirm this conclusion. The Web-based nature of the program makes it a potentially cost-effective way of promoting healthy eating.
A diet that is high in saturated fat and added sugars and low in fruit and vegetables is associated with a range of chronic diseases, including cardiovascular disease, cancer, and diabetes [
One way of promoting a more healthy diet is via Internet-based intervention. This has a range of potential advantages [
Although these results offer a useful first step in understanding the efficacy of Internet-based health promotion interventions, most of them draw on the same set of behavior change theories to guide content development. In particular, social cognitive theory, the theory of reasoned action / planned behavior, and the transtheoretical model are frequently used [
This paper describes the initial evaluation of a new, fully automated Internet-based healthy eating intervention: the HealthValues Healthy Eating Programme. This program differs from previous Web interventions in its use of novel behavior change techniques. In developing the HealthValues Programme, we used a more bottom-up approach, employing a selection of distinct, brief interventions that have been shown to influence behavior. There are a wide range of such techniques in the research literature, but these often fail to be translated into practice. As such, the strategies we selected can be viewed as a starting point rather than a comprehensive selection.
The first strategy involved asking individuals to spend 5 minutes thinking about why the value of health is important or unimportant to them. There is evidence that social values (eg, equality, helpfulness) often lack cognitive support. In other words, although individuals believe them to be important, they have not necessarily thought about why they are important [
The second and third strategies asked individuals to spend 5 minutes considering (1) their desires and aspirations in relation to their health together with how achieving these would make them feel and (2) their concerns in relation to their health alongside how failing to avoid these would make them feel. These strategies map onto techniques commonly employed in motivational interviewing (MI) [
The fourth strategy consists of implementation intentions with mental contrasting. Implementation intentions are specific plans of when, where, and how someone will change their behavior. They are believed to work by (1) increasing the accessibility of the situational cue that is relevant to the target behavior and (2) increasing the efficiency with which one performs the target behavior in the presence of the situational cue [
To enhance the efficacy of the implementation intentions, we also utilized evidence about moderators by including a number of other features. These were the use of an “if...then...” format [
The fifth strategy was the use of tailored dietary feedback in conjunction with standard health promotion advice [
Finally, the program also incorporated weekly tips during the last phase. These were primarily aimed at maintaining user engagement [
Drawing on the model of action phases [
Given that this study served as an initial test of the program, there were no comparable studies on which to base sample size calculations. That said, our sample size was informed by our previous research that examined the effects of 1 of the intervention components (thinking about reasons for values) on eating behavior over a 7-day period [
Participants were recruited using both online and print advertisements in the local community. These included posters and flyers in local shops and community facilities, and advertisements on social media sites, email networks, and in local newspapers. The advertisements stated that the study team were looking for individuals to test a new online healthy eating program and noted that individuals would be reimbursed for participation. The study’s website address (which included a full participant information sheet) was included in the advertisement. See
As inclusion criteria, we stipulated that participants be aged 18 or older and able to comply with the study procedures (ie, attend the laboratory appointments and complete the weekly online sessions). Other exclusion criteria were pregnancy, being out of the country for more than 3 weeks during the study period, another household member already participating, and participation in a previous related study. A total of 159 individuals contacted the study team during the recruitment period. Of these, 38 decided not to take part or failed to respond to subsequent communications and 21 did not meet inclusion criteria.
Flow of participants through the study.
The study received ethics approval from Swansea University Psychology Department Ethics Committee. Informed consent was collected by researchers at the first laboratory assessment (described subsequently). Although the study was a randomized controlled trial design, given its exploratory nature, the trial was not registered.
Laboratory measures were taken at baseline (February to April 2012), and at 3 months (May to July 2012) and 6 months (August to October 2012) postbaseline by GJB and a second research assistant, both of whom were blind to group allocation. Following baseline assessment, GJB emailed KT details of each participant’s dieting status and fruit and vegetable consumption. KT then allocated participants to an intervention or control (“monitoring”) group using a stratified block randomization protocol on the basis of dieting status (dieting versus nondieting) and fruit and vegetable consumption (≥5 a day versus <5 a day). Block size was 2 and random numbers were generated in Excel. KT then emailed the participant details of their user ID and password and they were informed of their group allocation the first time they logged on. Although participants were not blind to group allocation, they were informed that both the “experimental” group and the “monitoring” group would monitor eating behaviors and that this had been shown to be useful for reaching health goals. Participants in the control group were offered the opportunity to complete the program tasks at the end of the study.
All participants were asked by automated email to log onto the study website every week on 24 separate occasions to complete measures (intervention and control group) and program tasks (intervention group only). Each session could be accessed 6 days after completion of the previous session. Once the session became available, the participant was sent an email asking them to log in to complete it. Up to 3 automated reminders were emailed 2, 4, and 6 days later to participants who failed to complete the session. After completion of each session, the participant was sent an automated email thanking them and reminding them to log in again the following week. If participants failed to log in for 3 weeks, GJB attempted to contact them by phone and then email to establish whether they still wanted to participate in the online sessions and, if not, to assure them that we would still be keen for them to attend the laboratory assessments.
Each participant received £10 (approximately US $17) for attending the first laboratory session, £25 (US $42) for the second, and £50 (US $84) for the third. Additionally they received £2 (US $3) per session for completing the first 10 online sessions, £2.50 (US $4) per session for completing the next 10 online sessions, and £5 (US $8) per session for completing the last 4 online sessions. Thus, participants could receive up to £150 (US $253) for completing all laboratory and online sessions. Money for completing the online sessions was given at the final laboratory assessment and amounts allocated were indicated in emails sent to prompt, remind, and thank participants. In a further effort to limit attrition, participants received small gifts (a fabric bag and a mouse pad) at the first and second laboratory assessments. These were branded with the HealthValues logo.
Primary outcome measures were intake of (1) saturated fat, (2) added sugar, and (3) fruit and vegetables. These were assessed in a laboratory using the Block Fat/Sugar/Fruit/Vegetable screener, a 55-item food frequency questionnaire (FFQ) adapted from a longer version that has been shown to have good reliability and validity [
Secondary outcome measures were BMI, waist-to-hip ratio (WHR), heart rate variability (HRV), smoking status, smoking frequency, quantity of alcohol consumed, binge drinking, physical activity, dietary behaviors, and additional online assessments of saturated fat, added sugar, and fruit and vegetable intake. BMI, WHR, and HRV were assessed in the laboratory by trained researchers. These physiological measures provide an objective assessment of health status [
Alcohol consumption was measured in the laboratory using a questionnaire designed to capture episodes of binge drinking as well as typical drinking behaviors [
Smoking was assessed in the laboratory by asking participants whether they smoked cigarettes and, if yes, the number they usually smoked either per day, per week, or per month. Scores were recorded into number smoked per week.
Physical activity was assessed online at sessions 1, 8, 12, and 24 using the short version of the International Physical Activity Questionnaire (IPAQ) [
In addition to the laboratory assessments, saturated fat, added sugar, and fruit and vegetable consumption were also assessed online at sessions 1, 8, 12, and 24 using a validated UK FFQ [
To compute daily intake of saturated fat and added sugar, the proportions of these macronutrients in each of the 63 foods were calculated based on data provided by the British Food Standards Agency [
Two additional questions were used in the calculation of fruit and vegetable consumption. These were the number of portions of fruit (excluding fruit juice), and the number of portions of vegetables (excluding potatoes, beans, and lentils) eaten on a typical day during the previous week. Examples of portions were provided. These scores were combined with scores from items relating to fruit juice and beans/lentils from the FFQ to compute daily servings of fruit and vegetables. In-line with UK guidelines, juice and beans/lentils were counted as a maximum of 1 serving a day each.
Dietary behaviors were assessed at the start of each of the 24 online sessions using a questionnaire that was developed for the project. This consisted of 17 items associated with standard dietary advice related to consumption of saturated fat, added sugar, and fruit and vegetables (eg, reducing the number of teaspoons of sugar added to hot drinks, cereals, and desserts; replacing red meat with white meat or fish). The items were a mix of quantitative (eg, number of high fat snacks during the previous week) and categorical (eg, type of milk usually consumed). To reduce respondent burden, after the first session participants were presented with their responses from the previous session and asked to simply adjust their answers where they had made a dietary change. The questionnaire was scored by calculating the number of positive versus negative changes made since the previous session (–17 to +17).
All online questionnaires were tested for usability before the study. Questionnaires and items were presented in the same order for each participant and participants needed to complete all items before progressing to the next screen. Adaptive questioning was used for the IPAQ.
Details of participants’ gender, age, level of education, and first language were collected at the first online session.
Data relating to potential mediators (habits, intentions, self-efficacy, anticipated emotions), moderators (need for affect, need for cognition, behavioral approach system sensitivity, behavioral inhibition system sensitivity, environmental change), and process measures (poststudy feedback questionnaires and telephone interviews) were also collected, but these are not discussed in the present paper.
The intervention was tested for usability before the study. At all sessions, intervention components were delivered after assessment measures. The intervention components are detailed in
Baseline characteristics of the 2 groups were compared using
Per protocol analysis was conducted on all primary and secondary outcomes by including only those participants who completed all 3 laboratory assessments as well as 12 or more of the 24 online sessions (for laboratory measures) or all 24 online sessions (for online measures). Although the samples for such analyses are subject to bias, they are an important means of examining intervention efficacy in exploratory trials. A series of 3 (time) × 2 (group) mixed ANOVA models were used to examine effects on laboratory-based measures whereas 4 (time) × 2 (group) ANOVA models were used for online measures. Analyses were conducted with outliers (defined as 3.5 SDs from the mean) both included and excluded. Fisher exact test was used to examine smoking status and chi-square test was used for binge drinking status.
To examine the effects of the individual intervention strategies employed in the motivational phase, change scores were calculated using the dietary behaviors questionnaire. These were computed using figures from the session in which the strategy was employed and 2 sessions later (eg, change between sessions 1 and 3, see
Analysis of baseline characteristics showed that the intervention and control groups were well matched across a range of variables (see
Baseline characteristics of the intervention and control groups (N=100).
Variable | Control group (n=50) | Intervention group (n=50) |
|
Gender (female), n (%) | 42 (84) | 41 (82) | .79b |
Age (years), mean (SD) | 37.7 (13.2) | 41.1 (14.1) | .21c |
BMI (kg/m2), mean (SD) | 28.1 (5.8) | 27.1 (5.7) | .40c |
Dieting status (dieting), n (%) | 11 (22) | 12 (24) | .81b |
Education level (degree level or higher),a n (%) | 29 (58) | 34 (68) | .86b |
First language (English/Welsh), n (%) | 49 (98) | 45 (90) | .09b |
Ethnic background (white British), n (%) | 42 (84) | 34 (68) | .32b |
a Highest level of educational attainment coded as GCSEs, A-levels, degree (or equivalent), still studying or other.
b
c Chi-square test.
Descriptive and inferential statistics for intention-to-treat analyses (without outlier adjustment) are shown in
Means (SDs) and results from ANOVA models for intake of saturated fat, added sugar, and fruit and vegetables at baseline, 3 months, and 6 months in the intervention and control groups, for the intention-to-treat analysis.
Variable and time | Group, mean (SD) | Effects for time | Effects for time × group | ||||||
|
Control |
Intervention |
|
|
Partial η2 |
|
|
Partial η2 | |
|
|
|
35.9 | <.001 | 0.27 | 0.8 | .36 | 0.01 | |
|
Baseline | 21.4 (8.9) | 19.7 (9.6) |
|
|
|
|
|
|
|
3 months | 17.3 (8.3) | 16.1 (7.7) |
|
|
|
|
|
|
|
6 months | 15.9 (6.6) | 15.7 (9.9) |
|
|
|
|
|
|
|
|
|
8.6 | .004 | 0.08 | 0.2 | .62 | 0.00 | |
|
Baseline | 47.6 (34.0) | 43.2 (42.0) |
|
|
|
|
|
|
|
3 months | 36.7 (30.4) | 30.3 (25.5) |
|
|
|
|
|
|
|
6 months | 38.5 (37.6) | 30.5 (37.0) |
|
|
|
|
|
|
|
|
|
0.0 | .98 | 0.00 | 3.1 | .08 | 0.03 | |
|
Baseline | 3.6 (1.5) | 3.7 (1.7) |
|
|
|
|
|
|
|
3 months | 3.5 (1.9) | 3.8 (1.7) |
|
|
|
|
|
|
|
6 months | 3.3 (1.5) | 3.9 (1.6) |
|
|
|
|
|
|
Descriptive and inferential statistics for continuous primary and secondary outcome measures collected at laboratory sessions are shown in
Means (SDs) and results from ANOVA models for laboratory-assessed primary and secondary outcomes at baseline, 3 months, and 6 months in the intervention and control groups, for the per protocol analyses.
Variable and time | Group, mean (SD) | Effects for time | Effects for time × group | ||||||
|
Control |
Intervention |
|
|
Partial η2 |
|
|
Partial η2 | |
|
|
|
28.7 | <.001 | 0.24 | 1.2 | .23 | 0.01 | |
|
Baseline | 21.0 (8.9) | 19.3 (8.9) |
|
|
|
|
|
|
|
3 months | 16.7 (8.0) | 16.2 (7.3) |
|
|
|
|
|
|
|
6 months | 15.5 (6.4) | 15.7 (9.6) |
|
|
|
|
|
|
|
|
|
7.2 | .009 | 0.07 | 0.1 | .76 | 0.00 | |
|
Baseline | 46.7 (34.3) | 42.3 (43.0) |
|
|
|
|
|
|
|
3 months | 35.8 (30.5) | 30.4 (26.1) |
|
|
|
|
|
|
|
6 months | 37.2 (38.1) | 30.4 (38.5) |
|
|
|
|
|
|
|
|
|
0.3 | .57 | 0.00 | 4.0 | .048 | 0.04 | |
|
Baseline | 3.6 (1.4) | 3.7 (1.7) |
|
|
|
|
|
|
|
3 months | 3.4 (1.7) | 3.8 (1.7) |
|
|
|
|
|
|
|
6 months | 3.4 (1.5) | 4.1 (1.6) |
|
|
|
|
|
|
|
|
|
1.6 | .20 | 0.02 | 0.2 | .69 | 0.00 | |
|
Baseline | 6.4 (5.6) | 6.3 (6.2) |
|
|
|
|
|
|
|
3 months | 6.8 (7.2) | 6.7 (6.9) |
|
|
|
|
|
|
|
6 months | 7.2 (7.5) | 6.7 (7.3) |
|
|
|
|
|
|
|
|
|
11.2 | .001 | 0.11 | 0.1 | .93 | 0.00 | |
|
Baseline | 28.4 (5.8) | 27.0 (5.9) |
|
|
|
|
|
|
|
3 months | 28.3 (5.9) | 26.8 (5.7) |
|
|
|
|
|
|
|
6 months | 28.0 (5.9) | 26.6 (5.9) |
|
|
|
|
|
|
|
|
|
7.2 | .009 | 0.07 | 0.0 | .71 | 0.00 | |
|
Baseline | 0.82 (0.09) | 0.82 (0.09) |
|
|
|
|
|
|
|
3 months | 0.81 (0.09) | 0.82 (0.09) |
|
|
|
|
|
|
|
6 months | 0.81 (0.08) | 0.81 (0.08) |
|
|
|
|
|
|
|
|
|
1.4 | .25 | 0.02 | 2.0 | .13 | 0.02 | |
|
Baseline | 45.0 (20.1) | 49.6 (19.7) |
|
|
|
|
|
|
|
3 months | 46.4 (20.1) | 47.8 (18.7) |
|
|
|
|
|
|
|
6 months | 46.1 (17.9) | 43.1 (15.2) |
|
|
|
|
|
|
|
|
|
1.4 | .24 | 0.02 | 2.9 | .06 | 0.03 | |
|
Baseline | 28.9 (14.6) | 33.1 (19.6) |
|
|
|
|
|
|
|
3 months | 19.3 (15.3) | 30.5 (16.3) |
|
|
|
|
|
|
|
6 months | 30.2 (15.4) | 25.8 (12.9) |
|
|
|
|
|
|
a For alcohol consumption, n=46 due to questionnaire completion error.
b For alcohol consumption, n=44 due to questionnaire completion error.
c For alcohol consumption,
For smoking status, there were 91 participants who provided data on smoking at all 3 laboratory assessments and completed at least 12 of the online sessions. At each of the 3 time points there was no difference in the proportion of smokers in the experimental group compared to the control group at baseline (control: n=6, experimental: n=2,
Analysis of binge drinking included 90 participants who provided data on alcohol consumption at all 3 laboratory assessments and completed at least 12 of the online sessions. Again, at each of the 3 time points, there was no difference in the proportion of individuals who engaged in binge drinking in the experimental group compared to the control group at baseline (control: n=25, experimental: n=23; χ2
1=0.0,
Descriptive and inferential statistics for secondary outcome measures collected during the online sessions are shown in
Means (SDs) and results from ANOVA models for secondary outcomes assessed online at sessions 1, 8, 12, and 24 in the intervention and control groups, for the per protocol analyses.
Variable and session | Group, mean (SD) | Effects for time | Effects for time × group | ||||||
|
Control |
Intervention |
|
|
Partial η2 |
|
|
Partial η2 | |
|
|
|
7.8 | .006 | 0.08 | 0.6 | .43 | 0.01 | |
|
1 | 24.4 (9.9) | 26.0 (15.4) |
|
|
|
|
|
|
|
8 | 22.3 (10.6) | 21.4 (13.0) |
|
|
|
|
|
|
|
12 | 21.2 (10.4) | 21.7 (11.4) |
|
|
|
|
|
|
|
24 | 22.4 (10.0) | 21.5 (9.1) |
|
|
|
|
|
|
|
|
|
8.41 | .005 | 0.10 | 2.0 | .16 | 0.02 | |
|
1 | 47.8 (43.6) | 57.32 (74.5) |
|
|
|
|
|
|
|
8 | 34.4 (22.7) | 34.4 (32.3) |
|
|
|
|
|
|
|
12 | 31.7 (21.4) | 32.1 (25.3) |
|
|
|
|
|
|
|
24 | 39.8 (27.0) | 31.8 (19.4) |
|
|
|
|
|
|
|
|
|
5.6 | .02 | 0.06 | 5.5 | .02 | 0.06 | |
|
1 | 4.9 (2.1) | 5.0 (2.0) |
|
|
|
|
|
|
|
8 | 5.2 (2.4) | 6.0 (2.3) |
|
|
|
|
|
|
|
12 | 5.3 (2.8) | 6.1 (2.2) |
|
|
|
|
|
|
|
24 | 4.9 (2.3) | 6.2 (2.7) |
|
|
|
|
|
|
|
0.2 | .67 | 0.00 | 0.2 | .69 | 0.00 | |||
|
1 | 2857 (2320) | 2432 (1626) |
|
|
|
|
|
|
|
8 | 2534 (2290) | 2138 (1522) |
|
|
|
|
|
|
|
12 | 2932 (4270) | 2420 (1966) |
|
|
|
|
|
|
|
24 | 2985 (3525) | 2350 (2344) |
|
|
|
|
|
|
a For physical activity n=39 due to participants coding “don’t know.”
b For physical activity n=37 due to participants coding “don’t know.”
C For physical activity,
Portions of fruit and vegetables consumed in the intervention and control groups at the start and end of each program phase.
For analysis of motivational phase strategies, all participants who completed the first 9 online sessions were included (control: n=47; intervention: n=46). Because fruit and vegetable consumption was improved by the intervention, we conducted exploratory analyses examining changes in fruit and vegetable consumption in the intervention and control groups in the 2-week period following the delivery of each of the 4 different program components (see
Portions of fruit and vegetables consumed in the intervention and control groups during the motivational phase.
Results of the per protocol analysis indicated that the HealthValues Healthy Eating Programme brought about significant increases in fruit and vegetable consumption relative to a control group. These equated to approximately 0.75 cups, or 1.3 portions of the recommended 5 or more portions per day. The results also suggested that these increases were brought about primarily by strategies employed in the motivational and maintenance phases of the program, rather than the implementation intentions employed in the volitional phase. Thus, it may be that low fruit and vegetable consumption among this particular group was limited primarily by motivation rather than any difficulties in implementing the behavior; when we increased motivation, it had a direct effect on consumption.
In contrast, although the program was associated with a decrease in saturated fat and added sugar consumption, these effects were comparable to those found in the control condition. Unlike increasing fruit and vegetable intake, which involves introducing additional foods into the diet, reducing fat and sugar entails cutting back. As such, intake may be influenced by additional factors that may not be as amenable to motivational strategies. In particular, consumption of high fat and sugar foods may be habitual and carried out with a degree of automaticity [
The results did, however, show overall reductions in intake of saturated fat and added sugar among both groups by approximately 4.7 and 11.4 grams per day, respectively. These findings are consistent with the physiological data that showed significant reductions in BMI and WHR. Given that our recruitment method targeted individuals who wanted to improve their diet, it is possible that these changes would have occurred even in the absence of study participation. However, this seems unlikely given the general trend for weight to increase over time [
The absence of effects for implementation intentions are at odds with previous non-Internet interventions [
The benefits of participation did not generalize to behaviors that were not directly targeted by the program; there were no significant spillover effects on levels of physical activity, alcohol consumption, smoking, or HRV, either between groups or over time. Although some research has suggested that health improvements may show spillover effects to other health-related behaviors [
In future research, it would be important to trial the program in the absence of incentives for session completion. Given the high rates of attrition in online interventions [
It is also important to examine the effects of the program with different populations. In the current study, we recruited participants who were interested in improving their diet. Thus, they were a group who were already reasonably motivated (as indicated by a baseline mean of 4.16 on a scale of 1 to 5 for intention to eat a healthy diet). It is possible that the motivational strategies would have been more effective among a less motivated group of individuals who might be accessed via workplace settings, for example.
In conclusion, the HealthValues Healthy Eating Programme significantly increased fruit and vegetable consumption among users. Future research comparing different versions of the program should help to identify more accurately the elements that were responsible for this effect. It seems likely that the monitoring component of the study also brought about reductions in intake of saturated fat and added sugar, although further research is needed to confirm this. Given that the program is fully automated, it represents a potentially cost-effective way of promoting healthy eating.
HealthValues Healthy Eating Programme homepage.
Study information sheet provided to participants.
Intervention components used in the HealthValues Programme together with most closely aligned categories of behaviour change technique according to the Behavior Change Technique Taxonomy.[
CONSORT-EHEALTH checklist V1.6.2 [
body mass index
food frequency questionnaire
heart rate variability
metabolic equivalent of task
motivational interviewing
root mean square of successive differences
standard deviation of RR
self-determination theory
waist-to-hip ratio
The research was funded by the Economic and Social Research Council. We thank Clare Clement for help with data collection, Paul Tapper for technical support, and Jennie Davies for nutritional advice.
None declared.