Socioeconomic Status and Psychosocial Mechanisms of Lifestyle Change in a Type 2 Diabetes Prevention Trial

ann. behav. med. (2009) 38:160 165 DOI 10.1007/s12160-009-9144-1 RAPID COMMUNICATION Socioeconomic Status and Psychosocial Mechanisms of Lifestyle Change in a Type 2 Diabetes Prevention Trial Nelli Hankonen, M. Soc. Sc. & Pilvikki Absetz, D.Psych. & Ari Haukkala, Ph.D. & Antti Uutela, Ph.D. Published online: 8 December 2009 # The Society of Behavioral Medicine 2009 Abstract Background Little is known about psychosocial mechanisms that may underlie differences in lifestyle change between socioeconomic groups. Purpose The purpose of this study is to examine how educational level influences middle-aged participants (N=385) psychosocial responses to the GOAL Lifestyle Implementation Trial. Methods The measurements of self-efficacy and planning for healthy lifestyle were conducted pre-intervention () and post-intervention (, 3 months), and measurements of exercise and healthy eating as outcomes at and at 12 months (T3). Results Psychosocial determinants at and their changes were mostly similar, irrespective of educational levels. barriers self-efficacy was enhanced slightly less (p=0.08) among the low-ses. levels as well as pre post-intervention changes in exercise self-efficacy predicted 12-month changes in exercise, and diet coping planning predicted changes in dietary fat intake. The associations were similar across all SES groups. Conclusions Enhancing self-efficacy and planning is similarly effective among intervention participants regardless of education level. N. Hankonen (*) : P. Absetz : A. Uutela Department of Lifestyle and Participation, National Institute for Health and Welfare, Helsinki, Finland e-mail: nelli.hankonen@thl.fi A. Haukkala Department of Social Psychology, University of Helsinki, Helsinki, Finland Keywords Socioeconomic status. Health behavior change... Health promotion Introduction Low socioeconomic status (SES), defined in terms of education, income or occupation, is related to poorer health [1]. Unhealthy lifestyle contributes to the development of type 2 diabetes (D), for example, and cardiovascular diseases, as well as largely explaining educational level differences in health [2] and mortality [3]. In the United States, life expectancy disparities between educational groups are steep among men but are also growing among women [4]. In Finland, health inequalities between educational groups are larger than those between income groups [5]. The underlying mechanisms for SES inequalities in health have not been examined systematically but suggested candidates have been poorer access to preventive services [6] and selective reach by public health programs [7]. Yet, even in interventions with equal reach, responses might differ depending on SES. This could mean differences in processes of change, in outcomes, or in both. Socioeconomic status may influence the processes of change at different phases of an intervention. In the present study, we focus on the influence of SES on post-intentional behavior, i.e., the processes after forming the decision to enter an intervention and pursue lifestyle change. Optimistic self-beliefs and sense of control, as well as self-regulation skills, facilitate health behavior change [8, 9]. Specifically, perceived self-efficacy and planning seem to be among the best psychosocial predictors of a variety of health behaviors [10]. relates to perceived self-regulatory capacity to deal with barriers for maintaining the behavior

ann. behav. med. (2009) 38:160 165 161 change in the long term. planning links suitable coping responses to anticipated risk situations to prevent relapse, increasing the prospects of long-term change [9]. The relationships between SES, lifestyle, and health are partly mediated by psychosocial resources [11]. SES is associated with a sense of control [11, 12]: The better educated report greater sense of control over their lives, health [13], and health behaviors [14]. Education might also enhance people s ability to effectively use self-regulation skills, i.e., the use of rational planning and self-monitoring [15]. Given the associations of self-efficacy and selfregulation with SES, attempts to change these cognitions might be more successful among those with higher SES. Mediating mechanisms might also be different across socioeconomic strata. To date, no studies have examined whether psychosocial mechanisms involved in behavior change develop differently depending on SES over the course of an intervention, or whether SES moderates their effects on behavior. In the GOAL Lifestyle Implementation Trial to prevent D [16], lifestyle outcomes and risk factor changes did not differ by education. Yet, it is possible that educational groups differ in psychosocial mechanisms leading to lifestyle change. Few studies have thus far focused on the impact of changes in determinants. Here, however, we will test whether initial changes in cognitions (from baseline to the 3-month follow-up, ) predict long-term changes in behavior (from baseline to the 12-month follow-up, T3). We examine SES differences in the psychosocial processes among participants of the GOAL Trial, with the following research questions: Q1. Do levels of self-efficacy and planning at baseline () differ by SES? Q2. Do both SES groups benefit from the intervention equally as much in terms of changes ( ) in self-efficacy and planning? Q3. Does SES moderate the associations of: (a) Post-intervention levels () of self-efficacy and planning with 12-month changes ( T3) in diet and exercise? (b) Pre post-intervention changes ( ) in selfefficacy and planning with 12-month changes ( T3) in diet and exercise? Method Intervention Setting and Participants In the GOAL Lifestyle Implementation Trial to prevent type 2 diabetes [16, 17], nutritional and physical activity goals shown as effective in the Finnish Diabetes Prevention Study [18] were incorporated into a social cognitive and self-regulatory behavior-change model [19, 20]. Three hundred eighty-five participants (men, N=103; women, N=282, age 50 65 years) took part in the intervention, which consisted of six 2-h sessions. Five of the sessions took place between the baseline assessment () and the 3-month follow-up (), with one booster session at 8 months. The sessions targeted participants self-efficacy beliefs and planning skills for adoption and maintenance of healthy diet and physical activity, using behavior change techniques such as goal-setting, self-monitoring, and planning [17]. The program was carried out in a primary health care setting in Southern Finland with trained public health nurses acting as group facilitators. (See http://www. palmenia.helsinki.fi/ikihyva/inenglish.html.) At, 380 participants with education information responded, 168 with primary, and 212 with secondary/ higher (college or university degree) education. Intervention exposure and study drop-out at 3-months was similar (5.4%, 4.7%) in both groups, but drop-out at 12-months was SES-dependent (19.6% in low-ses, 11.3% in high-ses, p=0.03). Measures and Measurements Measurements were made before the intervention (), after the intensive intervention phase at three months (), and at 12 months follow-up (T3). Psychosocial determinants of physical activity (PA) and diet were measured with mailed questionnaires at and. Five-item scales were used for barriers self-efficacy for diet (α=0.95) and exercise (α=0.94). Possible responses ranged from 1 (very certain I cannot) to 4 (very certain I can). planning was measured with four items for diet (α=0.91) and four items for exercise (α=0.92) [9]. Responses ranged from 1 (definitely false) to 4 (definitely true). With regard to outcome measures, nutrient intake was measured with a 3-day food record at and T3 and analyzed by a licensed dietician using Nutrica software. The GOAL Trial aimed to improve participants diet focusing on the following established nutritional objectives [21]: (1) less than 30% of total energy intake from fat, (2) less than 10% of total energy intake from saturated fat, and (3) at least 15 g of fiber/1,000 kcal. The inter-correlations of these variables ranged between 0.49 and 0.87. We analyzed data only for the total fat intake objective. was measured as average minutes per day, monitored over a 1-week period, with every 10 min of activity recorded in a diary at and T3. The measure entails moderate/vigorous type of physical activity. Socioeconomic status was measured as the highest level of education attained and divided into two categories: primary education (LSES) and secondary/higher education (HSES).

162 ann. behav. med. (2009) 38:160 165 To calculate body mass index (BMI) as an indicator of obesity, height and weight were both measured at and T3 by study nurses. Data Analyses First, to be sure that the factor structure of self-efficacy and planning variables was similar in both SES groups, we tested whether factor loadings and residual variances were identical among the SES groups, with a series of confirmatory longitudinal factor analyses. For Q1 & Q2 The SES differences in levels and their changes were examined with a series of longitudinal multiple-group factor analyses using nested models, testing the equality of variables mean estimates across the SES groups. The changes were estimated using latent change score () models [22, 23]. The variable was regressed with a weight of 1 on the variable and a was estimated similarly with a weight of 1 on variable. This model also allowed for the estimation of means. To investigate the SES differences in self-efficacy change and planning change, the means were estimated for both groups separately. This model was then compared to a model assuming the means to be equal. For Q3 The predictive associations of self-efficacy and planning on exercise and diet were tested with two different sets of models: (a) a level-model with post-intervention levels of self-efficacy and planning as predictors of T3 change in exercise/diet, and (b) a change-model with changes in self-efficacy and planning as predictors of T3 change in exercise/diet. In the level-models, selfefficacy and planning predict variables, which in turn predict T3 changes in behavior. In the change models, the spredict T3 changes in behavior. The models tested were as follows: Ia: Post-intervention () levels of self-efficacy and planning Change in exercise ( T3) Ib: Changes ( ) in self-efficacy and planning Change in exercise ( T3) IIa: Post-intervention () levels of self-efficacy and planning Change in diet ( T3) IIb: Changes ( ) in self-efficacy and planning Change in diet (-T3) In all cases, the model was first calculated for the whole sample (M1). In the next step, a multiple-group SEM was estimated (M2), with the two path coefficients (self-efficacy behavior change, and planning behavior change) allowed to differ between the SES groups. Finally, to test for the possible moderation of the associations by SES, a third model was specified, constraining the regression coefficients to be equal across the SES groups (M3). As M2 and M3 were nested, we were able to compare them using the χ 2 -difference test to establish whether the constrained model, M3, had a statistically significantly (p<0.05) worse fit compared to M2 (with varying estimates for SES groups). A loss in fit would indicate that the regression estimates of the SES groups would be different from each other. We used CFI, TLI, and RMSEA and SRMR to assess the overall model fit. SPSS (version 15.0) was used to obtain descriptive statistics, and Mplus (version 5.2) for the Structural Equation Modeling (SEM) analyses. Results Factor loadings were identical among LSES and HSES (p<0.05). However, residual variances were not. The latter restrictions were not imposed in further analyses. As a measure of validity for the exercise measure, the baseline correlation with BMI was found to be negative and significant (r= 0.22, p=0.001). Descriptive statistics are presented in Table 1. Q1 & Q2: Differences and Changes in Psychosocial Factors At, the levels of exercise barriers self-efficacy and exercise coping planning were equal across the SES groups (see Table 1). At, barriers self-efficacy had increased among the high-ses but not among the low-ses. However, this difference was only borderline significant (χ 2 -difference test between the nested models, p=0.08). Overall, exercise selfefficacy increased but very slightly (p<0.05, mean of the latent change score=0.07, SD=0.57). planning increased during the intervention in all educational groups (p<0.001, M=0.25, SD=0.72). Levels or changes of diet determinants showed no SES differences. Overall on average, diet self-efficacy increased slightly (p<0.05, M=0.07, SD=0.53) and coping plans slightly more (p<0.001, M=0.37, SD=0.60). To retain the most parsimonious models, we imposed the mean constraints also on the later structural models. Q3: Determinants of Long-Term Changes in and Diet The four models (Ia Ib, IIa IIb) are shown in Fig. 1. I a: The overall model (M1) showed that high selfefficacy predicted increases in exercise (standardized

ann. behav. med. (2009) 38:160 165 163 Table 1 Means (M) and standard deviations (SD) of background factors, health behaviors and psychosocial determinants by education a LSES primary education, HSES secondary and higher education b = self-reported minutes of moderate-to-vigorous intensity physical activity per week c Estimates from SEM models Total N=385 LSES a N=168 HSES a N=212 P value M (SD) M (SD) M (SD) Years in school 9.60 (3.9) 8.0 (2.2) 10.7 (4.2) <0.001 Dietary fat (%) 29.9 (6.27) 29.7 (6.24) 29.9 (6.31) 0.76 Dietary fat (%) T3 29.2 (6.19) 28.3 (5.74) 29.9 (6.49) 0.02 b 81.4 (111.8) 66.5 (104.1) 89.0 (115.6) 0.38 T3 94.4 (131.1) 85.7 (129.7) 104.0 (135.3) 0.88 Body mass index 32.8 (5.1) 32.8 (4.7) 32.9 (5.2) 0.35 Body mass index T3 32.4 (5.0) 32.4 (4.9) 32.4 (4.8) 0.43 Psychosocial factors c Diet barriers self-efficacy 2.8 (0.59) 2.8 (0.62) 2.8 (0.57) 0.68 Diet coping plans 2.0 (0.65) 2.1 (0.61) 2.0 (0.66) 0.17 barriers self-efficacy 2.8 (0.62) 2.8 (0.63) 2.8 (0.63) 0.91 coping plans 2.0 (0.68) 2.0 (0.67) 2.0 (0.69) 0.42 regression coefficient=0.20, p=0.004). However, coping planning did not have any significant effect on exercise change. Next, a multiple-group model with freely varying regression parameters (M2) was compared against the model assuming the regression parameters to be the same in both SES groups (M3), the χ 2 -difference test indicated that the model fit of M3 did not decrease (compared to M2), suggesting similar associations for both groups (χ 2 difference test: Δχ 2 =1.8, Δdf=2, p=0.41). I b: M1 showed that change in exercise ( T3) was predicted by changes ( ) in exercise selfefficacy, but not exercise planning. Comparison of M2 and M3 (Δχ 2 =0.84, Δdf=2, p=0.66) indicated similar regression coefficients for both SES groups. II a: Reduction in dietary fat was predicted by high level of diet coping planning (β= 0.18, p=0.004), but not diet barriers self-efficacy (M1). Again, χ 2 difference test (Δχ 2 =3.69, Δdf=2, p=0.16) showed no significant differences between M2 and M3. II b: Change in diet was not predicted by changes in selfefficacy or planning. Although the regression coefficients seem larger for the high-ses group, they did not yield significance. The associations were similar for both SES groups (Δχ 2 =3.20, Δdf=2, p=0.20). Discussion The intervention brought about mostly similar and beneficial changes in self-efficacy and planning across the socioeconomic groups studied. This implies that the socioeconomically more privileged did not gain significantly more psychological benefit from the intervention than other groups. The only near-significant SES difference was that among those with primary education, exercise barriers selfefficacy increased less. Their belief in their own ability to maintain the behavior change was thus more difficult to enhance. This is a cause for concern, since in terms of disease prevention maintenance of lifestyle change is a crucial objective. The lower exercise self-efficacy of those with lower education can be attributed to real or perceived obstacles, to different values and beliefs regarding health behavior, and to less support in general [14]. However, diet self-efficacy changes showed no SES differences. SES did not moderate the psychosocial effects on health behavior change outcomes either (in all cases the more parsimonious model M3 was retained). Possible explanations include this study s focus on the volitional phase of health behavior change, where the intention for lifestyle change has already been formed. The pre-action phase might show a stronger socioeconomic patterning. Indeed, it has been suggested that those with high socioeconomic status are in more-advanced stages of change [24] and thus more willing to enter lifestyle interventions. Also, we studied high-risk individuals who may not be comparable to the general population within their socioeconomic strata. Change in exercise was predicted by self-efficacy. In contrast, the most important predictor for diet was coping planning. The reasons for this difference might lie in the type or essence of the behaviors: Healthy eating (especially avoiding fatty foods) requires preparation of coping plans to tackle food temptations in order to avoid relapse, whereas exercise requires a more proactive approach in taking action. In the latter, strong confidence in one s ability to persist might be more important than preparing to cope with risk situations. We investigated the effects of both psychological changes and post-intervention levels. This analytical approach indicates that despite the results seeming similar overall, the slight discrepancies between SES groups might

164 ann. behav. med. (2009) 38:160 165 Fig. 1 Ia: Post-intervention () levels Change in exercise ( T3), I b: Changes ( ) Change in exercise ( T3). IIa: Post-intervention () levels Change in diet ( T3). II b: Changes ( ) Change in diet ( T3). For each parameter, three estimates are shown: the overall estimate (from M1), then the low-ses/high-ses estimates (from M2). The variables of interest are emphasized with gray. For representation clarity, some of the parameters (e.g., factor indicators, correlations, residual correlations) have been omitted from the figure. The regression paths yielding non-significant coefficients are represented with dashed lines (although allowed to be estimated in the models). For the diet models IIa and IIb only the section of interest is displayed, although they are identical to the exercise models Ia and Ib. latent change score. p<0.10, *p<0.05, **p<.01, ***p<0.001. Pre-intervention, Postintervention, 3-month follow-up, T3 12-month follow-up Ia.32***.44***/.24***.21***.18*/.24*** Ib.32***.44***/.24***.16**.18*/.15*.53***.58***/.50***.42***.36***/.46*** -.43*** -.36***/-.48***.23***.26**/.20** -.52*** -.50***/-.54***.35***.34***/.35****.15*.10/.20*.03.01/.07.20**.23*/.17.03 -.06/.12 -.44*** -.38***/-.48*** Change Change T3.21**.18*/.24*** -.41*** -.33***/.-.44*** T3 IIa.37***.38***/.36***.03.15/-.05 -.18** -.14/-.19* Dietary fat change IIb.25***.29**/.22** -.03.00/-.07 -.05.07 /-.13 Dietary fat change have remained unexplained with results based on one model only: For example, the effect of self-efficacy on behavior in Ia is slightly stronger for the low-ses, but the opposite is true for the Ib model. Additionally, it should be noted that while the coefficients from psychological factors to 12-month behavior changes were low-to-moderate in size, they are however quite high for a real-world intervention with a 9-month time lag in measurement. This study has limitations. First, participants were selfselected into the study and might not be representative of their respective socioeconomic group or even the high-risk individuals in their group. However, the data represents a real-world situation with people who seek lifestyle counseling from primary health care. This yields the study a strong external validity. Second, the results may be gender-specific. Socioeconomic gradients in men s mortality and morbidity rates have been found to be steeper and more consistent than those of women [25]. Larger samples allowing, domain and gender differences should be explored in future studies. Third, all behavior measures are self-reported. In an earlier efficacy trial on preventing D diet was measured identically and predicted an incidence of D at follow-up [21]. The self-report measure of exercise was correlated with the BMI, suggesting validity also. Fourth, the only indicator of SES used here was educational attainment level, while other often-used indicators include occupation and income.

ann. behav. med. (2009) 38:160 165 165 Still, it remains that education is a robust indicator of SES. Furthermore, the participants were representative of the same aged general population in terms of education [26]. Fifth, there is an overlap of 3 months during which both psychological and behavior changes take place, challenging the uni-directionality of causality. However, many studies thus far have dealt with simultaneous changes instead of temporally distant changes. In our data, though, we include the 9-month outcome-change subsequent to the determinant change. To conclude, our findings suggest that if equal access to a lifestyle change program is guaranteed, persons with different educational backgrounds are likely to proceed to behavior change through similar psychosocial mechanisms. References 1. Adler NE, Boyce WT, Chesney MA, et al. Socioeconomic status and health: The challenge of the gradient. Am Psychol. 1994; 41: 15-24. 2. Laaksonen M, Talala K, Martelin T, et al. Health behaviours as explanations for educational level differences in cardiovascular and all-cause mortality: A follow-up of 60,000 men and women over 23 years. Eur J Public Health. 2007; 18: 38-43. 3. Huisman M, Kunst AE, Bopp M, et al. Educational inequalities in cause-specific mortality in middle-aged and older men and women in eight western European populations. Lancet. 2005; 365: 493-500. 4. Meara ER, Richards S, Cutler DM. The gap gets bigger: Changes in mortality and life expectancy, by education, 1981 2000. Health Aff (Millwood). 2008; 27: 350-360. 5. Cavelaars AE, Kunst AE, Geurts JJ, et al. Differences in self reported morbidity by educational level: A comparison of 11 western European countries. J Epidemiol Community Health. 1998; 52: 219-227. 6. Acheson D. Independent inquiry into inequalities in health: Report. London: HMSO; 1998. 7. Victora CG, Vaughan JP, Barros FC, Silva AC, Tomasi E. Explaining trends in inequities: Evidence from Brazilian child health studies. Lancet. 2000; 356: 1093-1098. 8. Bandura A. : The exercise of control. New York: Freeman; 1997. 9. Sniehotta FF, Schwarzer R, Scholz U, Schüz B. Action planning and coping planning for long-term lifestyle change: Theory and assessment. Eur J Soc Psychol. 2005; 35: 565-576. 10. Schwarzer R, Schüz B, Ziegelmann JP, Lippke S, Luszczynska A, Scholz U. Adoption and maintenance of four health behaviors: Theory-guided longitudinal studies on dental flossing, seat belt use, dietary behavior, and physical activity. Ann Behav Med. 2007; 33: 156-166. 11. Taylor SE, Seeman TE. Psychosocial resources and the SEShealth relationship. In: Adler NE, Marmot M, McEwen BS, Stewart J, eds. Socioeconomic status and health in industrial nations: Social, psychological, and biological pathways. New York: New York Academy of Sciences; 1999: 210-225. 12. Gurin P, Gurin G, Morrison BM. Personal and ideological aspects of internal and external control. Soc Psychol. 1978; 41: 275-296. 13. Ross CE, Wu C-l. The links between education and health. Am Sociol Rev. 1995; 60: 719-745. 14. Clark DO, Patrick DL, Grembowski D, Durham ML. Socioeconomic status and exercise self-efficacy in late life. J Behav Med. 1995; 18: 355-376. 15. Goldman DP, Smith JP. Can patient self-management help explain the SES health gradient? Proc Natl Acad Sci USA. 2002; 99: 10929-10934. 16. Absetz P, Valve R, Oldenburg B, et al. Type 2 diabetes prevention in the "real world": One-year results of the GOAL Implementation Trial. Diabetes Care. 2007; 30: 2465-2470. 17. Uutela A, Valve R, Talja M, Absetz P, Nissinen A, Fogelholm M. Health psychological theory in promoting population health in Päijät-Häme, Finland: First steps toward a type 2 diabetes prevention study. J Health Psychol. 2004; 9: 73-84. 18. Tuomilehto J, Lindstrom J, Eriksson JG, et al. Prevention of type 2 diabetes mellitus by changes in lifestyle among subjects with impaired glucose tolerance. N Engl J Med. 2001; 344: 1343-1350. 19. Schwarzer R, Fuchs R. and health behaviours. In: Conner M, Norman P, eds. Predicting health behaviour. Buckingham: Open University Press; 1996: 163-196. 20. Oettingen G, Honig G, Gollwitzer PM. Effective self-regulation of goal attainment. Int J of Educ Res. 2000; 33: 705-732. 21. Lindström J, Peltonen M, Eriksson JG, Louheranta A, Fogelholm M, Uusitupa M. High-fibre, low-fat diet predicts long-term weight loss and decreased type 2 diabetes risk: The Finnish Diabetes Prevention Study. Diabetologia. 2006; 49: 912-920. 22. McArdle JJ. Latent variable modeling of differences and changes with longitudinal data. Annu Rev Psychol. 2009; 60: 577-605. 23. McArdle JJ, Nesselroade JR. Using multivariate data to structure developmental change. In: Cohen SH, Reese HW, eds. Life-span developmental psychology: methodological contributions. Hillsdale, NJ: Erlbaum; 1994: 223-267. 24. Adams J, White M. Are the stages of change socioeconomically distributed? A scoping review. Am J Health Promot. 2007; 21: 237-247. 25. MacIntyre S, Hunt K. Socio-economic position, gender and health: How do they interact? J Health Psychol. 1997; 2: 315-334. 26. Fogelholm M, Valve R, Absetz P, et al. Rural urban differences in health and health behaviour: A baseline description of a community health-promotion programme for the elderly. Scand J Public Health. 2006; 34: 632-640.