tailored nutrition education - Nature

4 downloads 0 Views 78KB Size Report
education and to study the impact of written computer-tailored nutrition education ... computer-tailoring, expert individualised nutrition education can be realised ...
European Journal of Clinical Nutrition (1999) 53, Suppl 2, S78±S82 ß 1999 Stockton Press. All rights reserved 0954±3007/99 $12.00 http://www.stockton-press.co.uk/ejcn

Dutch research into the development and impact of computertailored nutrition education J Brug1* 1

Netherlands Open University, Department of Social Sciences, Heerlen, The Netherlands

Objective: The aim of the present paper is to describe the essential elements of computer-tailored nutrition education and to study the impact of written computer-tailored nutrition education in comparison to general written nutrition advice. Design: The impact of computer-tailored nutrition education was studied in three randomised trials as compared to general nutrition education. The data of the three studies were taken together and re-analysed. Subjects: Random samples of employees at two work sites and a self-selected sample of the Dutch adult population. The total number of subjects was 1309. Response rates were between 45% in one of the work sites and 88% in the self-selected sample. Interventions: Subjects in the experimental condition received computer-tailored nutrition education. With computer-tailoring, expert individualised nutrition education can be realised through an automated process for relatively large target groups with relatively low costs per person. Main outcome measures: The impact on changes in fat intake were studied. The use and appreciation of the intervention was also assessed. Results: Subjects who received computer-tailored advice had a lower mean fat score at post-test, adjusted for baseline intake levels. Subjects who received tailored advice were more likely to have read and discussed the nutrition advice. Conclusion: The results point to the conclusion that printed computer-tailored nutrition education is superior to general written nutrition education. Sponsorship: Financial support was provided by the Netherlands Cancer Society. Descriptors: diet; computer-tailoring; personalised feedback

Introduction Despite recommendations from health authorities about healthy nutrition practices to prevent chronic diseases, many people in Europe and the US consume too much fat (Willett, 1994). In order to stimulate people to adopt a healthier diet with less fat, nutrition education campaigns have been conducted with often limited effects on dietary behaviour (Contento et al, 1995). In the present paper, an innovative nutrition education technique is described and the joint results of three experiments to study its effect on fat reduction are discussed. Because many people need to change their diets, dietary interventions to improve public health should have the potential to reach populations using (cost)-effective approaches. In a review of the literature on effectiveness of nutrition education, three criteria for effective dietary interventions were outlined (Contento et al, 1995: (1) attention should be given to personally relevant motivators and reinforcers for the people in the target group, (2) personalised self-evaluation techniques should be used, and (3) people in the target group should have the opportunity to actively participate in the intervention. Not surprisingly, interpersonal nutrition counselling proved to give *Correspondence: Johannes Brug, Netherlands Open University, Department of Social Sciences, P.O. Box 2960, 6401 DL HEERLEN, The Netherlands. E-mail [email protected]

good results in inducing dietary change. Face-to-face counselling, however, requires well trained counselling and nutrition experts and is therefore relatively expensive and time consuming. Therefore, it is not realistic to expect that interpersonal nutrition counselling could be available to large, healthy population groups to stimulate preventive dietary changes. As well as this, only people who are aware of their need to change will be interested in nutrition counselling and in the Netherlands many people are not aware of their high fat intake (Brug et al, 1994). Since many people should adopt healthier dietary habits, mass media minimal contact interventions are often used. Unfortunately, these campaigns often have little impact on dietary change. It seems that nutrition education is caught between mass media campaigns that have the potential to reach many people but have little impact on dietary behaviour change, and face-to-face nutrition counselling, often effective among participants but with little potential to reach many people and therefore with only a minor impact on the population as a whole. Attempts have been made to try to incorporate characteristics of interpersonal counselling in interventions that can possibly reach many people. Computer-tailored nutrition education is such an attempt. Computer-tailoring enables provision of personally tailored feedback for members of relatively large target populations (Brug et al, 1999). This article will describe the process of providing people with computer-assisted personalised dietary feedback and the joint results of three randomised trials

Dutch research into the development and impact of computer-tailored nutrition education J Brug

conducted in the Netherlands to test the impact of a computer-tailored intervention on fat reduction. Computer-tailoring Although it lacks direct social support and social interaction, computer-tailored nutrition education mimics some attributes of person-to-person nutrition counselling since it provides participants with personalised information. There is evidence that computer-tailored health education can be offered to larger groups of people with less costs compared to interpersonal counselling (Strecher et al, 1994; Velicer et al, 1993). The process of generating computer-tailored messages is summarised in Figure 1: People are surveyed with a baseline questionnaire, and the survey results are entered into a data ®le. Computer software links the data with a feedback source in the form of a nutrition education message archive (documented in a word processing ®le) containing appropriate feedback for each survey response. The software consists of algorithms to select the feedback segments from the archive and assemble them into a predetermined format, such as a tailored nutrition education letter. Nutrition education expertise is needed both to determine the correct feedback information and to formulate the decision rules on which the computer program is based. The computer program generates individualised nutrition education letters for each individual, based on his or her personal survey results. There are different options for input, computer operation, and output of computer-tailoring. The survey, for example, can be a self-administered written questionnaire, a telephone survey, or a questionnaire that is administered

Figure 1 The tailoring process.

interactively with a computer. Data can be entered manually, but surveys can also be scanned automatically, or respondents can enter their answers directly into a computer. Similarly, the output can take many forms. For example, feedback can be provided directly on the computer screen or it can be provided in personal letters or newsletters. In computer-tailored interventions that have been reported to date, variables such as individual behaviour (dietary fat consumption, servings of fruits and vegetables per day, etc.), socio-demographic variables (sex, age, etc.), health status (cholesterol levels, blood pressure, etc.), and psychosocial factors like attitudes, self-ef®cacy expectations, perceived threat and readiness for change have been used to individualise the feedback (Brug et al, 1996, 1998; Burling et al, 1989; Campbell et al, 1994; Van Beurden et al, 1990). The characteristics to which the information is tailored, the tailoring variables, should be relevant and important for the behaviour and=or the behaviour change that is targeted with the tailored intervention. Three criteria should be met to include a possible tailoring variable in the screening questionnaire and thus use it as a basis for individualised feedback: (1) as mentioned before the characteristic should be important for behaviour change, (2) it must be possible to survey the characteristics in a valid, reliable way with a set of closed-ended questions, and (3) meaningful differences in feedback must be possible for different levels of these characteristics. It is therefore important to apply theoretical insights about determinants of (health) behaviours and processes of behaviour change as guidelines in this selection (Brug et al, 1999; Strecher et al, 1994; Velicer et al, 1993), and when available, validated instruments should be used for screening. As was argued in the introduction of this article and has been described in an earlier article (Brug et al, 1999), computer-tailored. nutrition-education incorporates criteria for effective nutrition education such as providing personalised self-assessment and feedback, and responds to relevant motivators and reinforcers (Contento et al, 1995). Different health behaviour and communication theories also suggest why computer-tailored nutrition education might be more effective than general nutrition education. Since computer-tailoring provides each person with only the information selected for his or her personal situation and characteristics, the nutrition messages contain less redundant information. People are therefore more likely to pay attention to the message. Attention is essential for the nutrition education to have an impact (McGuire, 1985). Since subjects are limited in the amount of information that can be processed, it is important to provide them only with information that is worth the processing effort (Rudd & Glanz, 1990). Furthermore, according to the elaboration likelihood model (Petty & Cacioppo, 1986) involvement with the topic and contents of a health education message is an important determinant of the effort people want to invest in reading, comprehending and processing it. In computer-tailoring, involvement with the health education messages is increased by using the subject's name and other recognisable personal characteristics in the feedback. Finally, in computer-tailored interventions, the nutrition education can be tailored to the personal level of motivation of subjects. According to stages-of-change models, health interventions tailored to motivational stages are more likely to be successful (Prochaska & Diclemente, 1992; Weinstein, 1988).

S79

Dutch research into the development and impact of computer-tailored nutrition education J Brug

S80

Methods Dutch studies to assess the impact of computer-tailored nutrition education Three randomised trials were conducted in the Netherlands to test, among other things, the impact of computer-tailored nutrition education on changes in fat intake. The ®rst study was conducted among employees of an oil company (Brug et al, 1996). The second study was conducted among employees of a regional health care organisation (Brug et al, in press), while the third experiment was done among a self-selected sample of healthy volunteers (Brug et al, 1998). The design and results of these three studies have been reported separately. Since in the three studies very similar computer-tailored interventions were used, in this article comparable data from the three studies were analysed together to summarise the overall results of this computer-tailored intervention on fat reduction. Study design All three studies were pretest-post-test randomised trials in which subjects were allocated to either a computer-tailored intervention or general nutrition education. The present secondary analyses were aimed to detect differences in impact between computer-tailored nutrition education and general nutrition education on fat intake. The questionnaires and intervention Computer-tailored nutrition education requires three related components: an instrument to assess variables on which the tailored feedback will be based, a message source ®le with feedback messages tailored to all possible screening results, and a computer program that selects speci®c feedback messages for each respondent from the source ®le (Figure 1). The screening questionnaire Self-administered questionnaires were used to survey participants at baseline. The questionnaires for the three studies were very similar. Only results based on identical measures are included in the present analyses. The questionnaires took about ®fteen minutes to complete. The ®rst part of the questionnaires included questions about the participants' sex, age, height, weight, and education. The second part was a validated food-frequency questionnaire (25 items) which assesses fat scores with a range between 12 and 60 (Van Assema et al, 1992). This score is the result of questions about the frequency of use and portion size of the 12 main fat sources in the Dutch diet. This fat score instrument was developed for use in intervention research in which short instruments, that are easy to administer and process, are essential in order to avoid low response rates and to ensure practical applicability of the intervention. Inclusion of such a limited number of food groups proved to be suf®cient to rank respondents according to individual fat intake and to detect changes in individual fat scores, but it does not allow computation of percent energy from fat since non-fat energy sources were not included in the fat score instrument (Van Assema et al, 1992). In the third part of the questionnaire, a number of psychosocial variables were assessed. They differed rather substantially between the different studies. Results on these variables are therefore not included in the present paper.

Post-test questionnaires Post-test questionnaires were similar to the baseline questionnaire. Questions about the participants' sex, age, education, and other background factors were excluded from the post-tests, additional questions about participants' reactions to the nutrition information letters were included. The message source ®le The message source ®le consisted of 223 different feedback messages. Different dietary feedback messages were written for various categories of dietary fat intake. For example, the source ®le included messages for people eating more fat than is recommended but less than most of their peers, for people eating more fat than is recommended as well as more than most of their peers, and for people who were eating according to the recommendations for fat. The messages were also tailored to the participants' self-rated fat intake: Messages for participants who were unaware of their high fat scores, differed from those for participants who were aware of this. Further messages were included that addressed seven different important dietary fat sources in the Dutch diet (milk and milk products, meat and meat products, gravy and sauces, spreads, cheese, hot snacks, and sweet snacks), for which low fat alternatives for high fat choices were suggested. Participants received only information about fat sources that were important in their personal diet. Respondents were advised to change those dietary behaviours that were not in accordance with recommendations and to sustain and if possible further improve their dietary behaviours that already met the recommendations. In some cases, further messages were included in the source ®le that addressed respondents with negative attitudes and self-ef®cacy expectations toward the recommended dietary changes. These messages addressed the most prevalent and salient negative bene®ts about eating less fat. The three studies differed in the amount of psychosocial feedback that was given. A typical computer-tailored feedback letter would start with an introduction on the importance of healthy nutrition, proceed with information (in words as well as in a graph) about the respondent's individual fat score as compared to recommended intake and peer group average intake levels. Subsequently, the respondent's main fat sources were presented with low fat alternatives, and attitudinal and self-ef®cacy information was given. The tailoring computer program A computer program was written in Turbo Pascal which linked individual screening results to speci®c feedback messages from the source ®le. The program consisted mainly of a number of `IF-THEN statements' which were the decision rules for the selection of speci®c feedback messages from the source ®le for individual respondents based on their answers on the screening questions. The computer program regulated the production of personal feedback letters from the selected messages. General nutrition education letters The general, non-tailored nutrition education letters provided information about fat similar to the information given in lea¯ets and brochures from the Dutch Nutrition Education Bureau. Both tailored and general information letters were printed on identical paper. The participant's name was printed in the letter's introduction and in the ®nal paragraph

Dutch research into the development and impact of computer-tailored nutrition education J Brug

of the tailored letters. In both the tailored letters and the general nutrition information letters, illustrative cartoons and recipes for low fat meals were included. The tailored letters consisted of four to eight pages. The general nutrition information letters were ®ve pages long. Statistical methods Multiple analysis of covariance (mancova's) with fat scores at post-test as dependent variable, intervention (two levels: tailored vs non-tailored) and study population (three levels: two work site populations and self-selected sample) as factors and baseline fat score as co-variate, were conducted to study differences in impact between tailored and nontailored advice. Study population was included as a second factor in order to check for differences in impact of tailoring between the three study populations. Cross tabulation with Chi-square tests and one way analyses of variance were used to assess differences in use and appreciation of dietary advice between the tailored and non-tailored advice. Results Mean fat scores at pretest and post-test with 95% con®dence intervals are presented in Table 1. No signi®cant difference in mean fat score was found at baseline (F(1) ˆ 1.5; P ˆ 0.22). The multiple analysis of covariance (mancova's) revealed a signi®cant intervention effect on post-test mean fat score adjusted for baseline score with a lower mean fat score in the tailored group (F(1) ˆ 24.5; P ˆ 0.00). No signi®cant intervention 6 study population interaction effect was found (F(1) ˆ 1.7; P ˆ 0.20), suggesting no difference in intervention effect between the three study populations. Between baseline and post-test, the mean fat score declined with 5.5% in the tailored group compared to a 1.5% decline in the non-tailored group. Table 2 presents differences in use of the nutrition education letters between respondents in the tailored and the non-tailored groups. People in the tailored group more often read (Chi-square (2) ˆ 0.68; P ˆ 0.00), saved (Chisquare(2) ˆ 0.39; P ˆ 0.00), and discussed (Chisquare(2) ˆ 0.29; P ˆ 0.00) the feedback they received. Table 3 provides information on how the respondents in the two intervention groups appreciated the nutrition information letters. Respondents in the tailored group rated the Table 1 Mean fat scores (range 12 ± 60) at baseline and post-test for the intervention groups

Baseline Post-test

Tailored intervention group (n ˆ 919)

Non-tailored intervention group (n ˆ 390)

28.0 (27.7 ± 28.4) 26.5 (26.3 ± 26.8)

28.2 (27.7 ± 28.7) 27.8 (27.2 ± 28.0)

Table 2 Participants' use of the tailored letters (% yes) Tailored intervention Non-tailored (n ˆ 919) intervention (n ˆ 390) Have you read the complete letter? Have you saved the letter? Have you discussed the letter with others?

96

82

82 64

66 47

Table 3 Participants' opinions of the tailored letters (mean scores and standard deviations) Range ˆ 1 (very negative) to 7 (very positive)

Tailored General feedback feedback (n ˆ 919) (n ˆ 390)

How interesting was the tailored letter? How personally relevant was the nutrition information letter? How credible was the information? How dif®cult or easy to understand was the information?

4.00 (3.86 ± 4.15) 3.12 (2.94 ± 3.29) 3.68 (3.53 ± 3.83) 3.04 (2.87 ± 3.21) 4.14 (4.00 ± 4.28) 4.86 (4.66 ± 5.06) 6.34 (6.27 ± 6.41) 6.38 (6.28 ± 6.47)

nutrition information as signi®cantly more interesting (F(1) ˆ 46.9; P ˆ 0.00) and personally relevant (F(1) ˆ 25.1; P ˆ 0.00). No difference in understandibility was found (F(1) ˆ 0.4; P ˆ 0.54). The general nutrition education letters were rated as more credible (F(1) ˆ 31.3; P ˆ 0.00). Discussion The rationale for tailored nutrition education is based on the assumption that responding to individual dietary behaviour, needs and beliefs of subjects in the target population, will increase message relevance and therefore result in higher attention and motivational impact. From research described above, it appears that personalised dietary and psychosocial feedback is indeed more likely to be read, and seen as personally relevant compared to standard, nontailored materials. The printed computer-tailored intervention that was studied also appears to have greater effects in motivating people to reduce their fat intake compared to non-tailored messages. Nevertheless, the general nutrition education intervention material was given higher credibility. Secondary analysis revealed that this result could be attributed to respondents in the tailored group with a large discrepancy between their self-rated fat intake and the actual fat score that was communicated in the tailored feedback who doubted the credibility of the feedback they received. The evidence for the impact of computer-generated feedback presented in the present study is based on selfreports from study participants. Furthermore, only the short term impact of computer-tailoring was assessed. In general, no studies using more objective criteria like blood parameters or assessing long term effects of computer-tailoring in nutrition education are available at present (Brug et al, 1999; De Vries & Brug, 1999). For further proof of the superiority of computer-tailored nutrition education such studies are needed. There is a great variety of variables that have been used as a basis for generating tailored feedback. In our studies and other studies on computer-tailoring that have been conducted to date, feedback has been provided on dietary intake levels, dietary patterns, psychosocial factors like outcome expectancies and self-ef®cacy expectations, and=or stages of change (Brug et al, 1999). It is yet unclear what speci®c feedback elements are necessary in order to promote dietary changes although it seems justi®ed to conclude that feedback about the level of risk behaviour alone is not suf®cient to motivate people to start changing their diet. People at risk tend to respond negatively to this information and might therefore not succeed in altering their risks when no practical information for risk reduction is provided (Bowen et al, 1994). The present body of

S81

Dutch research into the development and impact of computer-tailored nutrition education J Brug

S82

knowledge about computer-tailoring indicates that it is important to supplement risk information with behavioural theory-based personalised suggestions on how to reduce risks and, possibly, psychosocial information, that is information tailored to psychosocial behaviour determinants like attitudes, self-ef®cacy expectations, etc. to motivate subjects to start contemplating and carry out behaviour changes (Brug et al, 1999). Computer-tailoring is a promising intervention approach in work site and point-of-choice settings. In these settings participants can be reached repeatedly for surveys and feedback. Periodical health examinations that are conducted in many worksites, for example, offer good opportunities to include computer-tailored nutrition education. In such a setting iterative feedback can be provided that is not only tailored to the results of the most recent screening but also to changes in behaviour, psychosocial variables or risk indicators between different screening moments (Brug et al, 1998). There is evidence that such iterative, longitudinal feedback improves the impact of computer-tailored nutrition education and=or prevents relapse (Brug et al, 1998). In point of choice settings, computer-tailored feedback offers the opportunity to provide subjects with immediate feedback about their dietary choices that can then lead to immediate action. Finally, general practices have been used as intermediaries for computer-tailoring (Campbell et al, 1994). Research has shown that general practitioners are regarded as the most trustworthy source of information on diet and health by the general public, whereas general practitioners often feel unable to provide their clients with nutrition education (Hiddink et al, 1995). Adding computer-tailoring capability to medical of®ces may enable general practitioners to provide their clients with effective nutrition education without having to invest much of their limited amount of time. Similarly, computer-tailoring systems offer the opportunity for non-nutrition experts to provide people with expert personalised nutrition information.

References Bowen DJ, Fries E, Hopp HP (1994): Effects of dietary fat feedback on behavioral and psychological variables. J. Behav. Med. 17, 589 ± 604. Brug J, Campbell M & Van Assema P (1999): The application and impact of computer-generated personalised nutrition education: A review of the literature. Pat. Educ. Couns. 36, 145 ± 156.

Brug J, Glanz K, Van Assema P, Kok G & Van Breukelen GJP (1998): The impact of computer-tailored feedback and iterative feedback on fat, fruit, and vegetable intake. Health Educ. Behav. 25, 357 ± 371. Brug J, Steenhuis IHM, Van Assema P & De Vries H (1996): The impact of a computer-tailored nutrition intervention. Prev. Med. 25, 236 ± 242. Brug J, Steenhuis IHM, Van Assema P, Glanz K & De Vries H (1999): Computer-tailored nutrition education: Differences between two interventions. Health Educ. Res., in press. Brug J, Van Assema P, Kok G, Lenderink T & Glanz K (1994): Self-rated dietary fat intake: Association with objective assessment of fat, psychosocial factors and intervention to change. J. Nutr. Educ. 26, 218 ± 223. Burling TA, Marrotta J, GonzaÂlez R, Moltzen JO, Eng AM, Schmidt GA, Welch RL, Ziff DC & Reilly PM (1989): Computerized smoking cessation program for the worksite: Treatment outcome and feasibility. J. Consult. Clin. Psychol. 57, 619 ± 622. Campbell M, DeVellis BM, Strecher VJ, Ammerman AS, DeVellis RF & Sandler RS (1994): Improving dietary behaviour: The effectiveness of tailored messages in primary health settings. Am. J. Public Health 84, 783 ± 787. Contento I, Balch GI, Bronner YL, Lytle LA, Maloney SK, Olson CM & Swadener SS (1995): The effectiveness of nutrition education and implications for nutrition education policy, programs, and research: A review of research. J. Nutr. Educ. 27, 277 ± 422. De Vries H & Brug J (1999): Computer-tailored interventions motivating people to adopt health promoting behaviours: Introduction to a new approach. Pat. Educ. Couns. 36, 99 ± 105. Hiddink GJ, Hautvast JGAJ, Van Woerkum CMJ, Fieren CJ & Van 't Hof MA (1995): Nutrition guidance by primary-care physicians: Perceived barriers and low involvement. Eur. J. Clin. Nutr. 49, 842 ± 851. McGuire WJ (1985): Attitudes and attitude changes. In: Handbook of social psychology: Vol. II. Special ®elds and application, G Lindzey G, E Aronsen E, (eds) New York: Random House, pp 233 ± 246. Petty RE & Cacioppo JT (1986): The elaboration likelihood model of persuasion. Adv. Exp. Soc. Psychol. 19, 123 ± 205. Prochaska JO & DiClemente CC (1992): Stages of change in the modi®cation of problem behaviours. Prog. Behav. Mod. 28, 184 ± 218. Rudd J & Glanz K (1990): How individuals use information for health action: Consumer information processing. In: Health behaviour and health education: Theory, research and practice. K Glanz, FM Lewis & BK Rimer (eds) . San Francisco (Ca): Jossey Bass, pp 115 ± 139. Strecher VJ, Kreuter M, Den Boer DJ, Kobrin S, Hospers HJ & Skinner CS (1994): The effects of computer-tailored smoking cessation messages in family practice settings. J. Fam. Practice 39, 262 ± 270. Van Assema P, Brug J, Kok G & Brants HAM (1992): The reliability and validity of a Dutch questionnaire on fat consumption as a means to rank subjects according to individual fat intake. Eur. J. Cancer Prev. 1, 375 ± 380. Vam Beurden E, James R, Dunn T & Tyler C (1990): Risk assessment and dietary counselling for cholesterol reduction. Health Educ. Res. 5, 445 ± 450. Velicer WF, Prochaska JO, Bellis JM, DiClemente CC, Rossi JS, Fava JL & Steiger JH (1993): An expert system intervention for smoking cessation. Addic. Behav. 18, 269 ± 290. Weinstein ND (1988): The precaution adoption process. Health Psychol. 7, 355 ± 386. Willett WC (1994): Diet and Health: What should we eat? Science 264, 532 ± 537.