Original Article

Developing and Testing an Instrument to Measure the Factors Affecting the Salt Restriction Behaviors among Women

Introduction

A healthy diet is one of the most important preventive factors fornon-communicable diseases (NCDs) including cardiovascular diseases ¹ . High salt intake is known as the main determinant of blood pressure at the individual and population levels ² . Moreover, there is evidence showing the association between high salt intake and stroke, left ventricular hypertrophy, the progression of kidney disease, kidney stones ³ , osteoporosis ⁴ , some types of cancers, respiratory diseases ⁵ and obesity ⁶ . Therefore, salt reduction has been suggested to reduce the risk of high blood pressure and cardiovascular diseases. There is a dose-response relation between sodium reduction and blood pressure-lowering ⁷ .

The mean salt intake in the world is 9 to 12 gr/day, indicating about twice recommended level ¹ . In Iran, the mean salt intake in Iran was 9.52 g/day and intake salt among about 40% of adults was at least two times higher than recommended level ⁸ . In light of context and priorities, Iran developed the action plan by looking for 4.30% relative reduction in average salt intake among the population ⁹ .

Population-level behavior change programs may be moderately effective in reducing salt intake. These theories inform interventions by suggesting mechanisms underlying paths between intervention components and outcomes. The added value of theoretically based interventions to reduce population salt intake has been supported by the previous studies ^10-12 . The Theory of Planned Behavior (TPB) is one of the widely used frameworks for designing dietary programs ^13-16 . This theory highlights, that human behavior is determined mainly by his/ her intention that in turn, is a function of his/her attitude, subjective norms and perceived behavioral control ¹⁷ . Several interventions grounded in TPB have demonstrated positive outcomes in creating significant changes in reducing salt consumption ¹⁸ .

So far, instruments have been provided to measure psychosocial determinants of salt consumption among patients with hypertension or heart disease ^19,20 . As people with specific health-related conditions follow a specific diet, the uses of these instruments cannot accurately reflect the status of constructs and behavior in the general population. Thus, there is a need for a brief scale to assess psychosocial determinants of salt consumption in the community. Moreover, in some studies, several aspects of salt intake behaviors have been integrated as a whole behavior ²¹ or only one salt-related behavior has been investigated ²² . In the field of salt intake, we identified three different behaviors including adding salt during cooking, adding salt at the table, and purchasing salty food. We believe that considering different aspects of salt intake and developing a brief scale to assess psychosocial determinants of salt intake in the community may contribute to an improved understanding of this behavior and evaluating the related interventions. Similar to many other nations, Iranian women are mainly responsible for home cooking. Therefore, the purpose of this study was to develop and test a new TPB- based instrument to measure the factors influencing salt intake behaviors in Iranian women who are responsible for family cooking.

Methods

This study received ethics approval from Institutional Review Board of the Hamadan University of Medical Sciences.

To develop the questionnaire, we followed the guidelines described by Ajzen ²³ and Francis et al ²⁴ . Applying these guidelines has been suggested to reduce the complexity faced in developing TPB's instrument ²⁵ . Accordingly, we used a mixed-method approach including the following steps: (a) conducting an elicitation study to extract the women's common underlying beliefs; (b) reviewing literature to identify the potential items measuring the TPB constructs related to reducing salt intake behaviors (c) conducting a cognitive interview to inform making changes in primary questionnaire and (d) conducting confirmatory factor analysis (CFA).

Step 1: Item generation

To generate appropriate items that fit the TPB constructs, we followed both deductive and inductive methods. Accordingly, we conducted a literature review to identify survey instruments assessing TPB constructs on different behaviors related to salt intake. We also considered instruments measuring variables similar to TPB constructs under different names. Then, through a semi-structured interview (n=30), we assessed women's common underlying beliefs and thought processes of how they justify their salt intake related behaviors. We used this method to review and refine the generated items. The initial questionnaire included 81 items measuring TPB constructs of our desired behaviors (salt intake during cooking, salt in taking while eating at the table and buying salty foods). These items consisted of ten items for behavioral beliefs, ten items for outcome evaluation, 13 items for normative beliefs, 13 items for motivation to comply, ten items for control beliefs, ten items for perceived power, 12 items for intention, and three items for behaviors. These items were scored on a five-point Likert scale (one = strongly disagree, five = strongly agree).

Step 2: Evaluate the content and face validity

The content validity of the questionnaire was evaluated using a panel of experts. The panel comprises of ten experts in the field of health education and two nutritionists. The relevancy of items was assessed using a 4-point Likert scale (1= not relevant, 2= somehow relevant, 3= quite relevant, 4= highly relevant). To ensure clarity and readability, the panel assessed the items using a 4-point Likert scale (1=not clear, 2=somehow clear, 3=clear, and 4=very clear). To calculate the CVI, the number of experts who chose options 3 and 4 was divided by the total number of experts, and values higher than 0.70 were accepted ²⁶ . To calculate the CVR, the panel was asked to evaluate the necessity of each item. The necessity of the items was assessed using a 3-point rating scale (1=not essential, 2=useful but not essential, 3=essential). Considering the number of experts, the CVR for scale ≥0.56 was considered satisfactory ²⁷ .

To assess face validity, besides testing it with the expert panel, a cognitive interview was performed. To do that, we recruited five women who fitted the criteria of potential participants to study and read the items aloud to identify their understanding of the phrasing of each item by determining the items that they were not clear or were too complex. The expert’s feedbacks and the cognitive interviews led to remove 11 items and revise in two Items, resulting in a pool of 70 items.

Step 3: Evaluating the scale

In the third step, we conducted construct validity and reliability tests. Hence, the convergent and discriminant validity were evaluated using confirmatory factor analysis. Internal consistency reliability was evaluated using Cronbach’s alpha coefficients. Overall, 300 women were recruited and interviewed at the households' homes. Although different recommendations have been suggested to calculate the sample size and power estimates for a confirmatory factor analytic models, a common recommendation is to use the sample sizes of at least 200/5 or ten cases per parameter ^28,29 . We used a cluster sampling method using a national survey sampling framework ³⁰ to select women who were responsible for household cooking. Participants were asked to provide demographic data and complete the TPB questionnaire.

Ethical approval for study was provided by Hamadan University of Medical Sciences Ethical Committee (IR.UMSHA.REC.13970882).

Data analysis

SPSS 24.0 (Chicago, IL, USA), and SmartPLS 3.2.8 were used to assess reliability and validity of constructs (convergent and divergent validity) and confirmatory factor analysis (CFA). First, KMO and Bartlett values were performed to evaluate the adequacy of sample size and sphericity of relationship before confirmatory factor analysis ²⁸ . Then, Questions were removed that did not achieve a factor loading of 0.4 or greater ^31,32 . Cronbach's Alpha, as well as the composite and communality reliability test were used to evaluate construct reliability. We considered values greater than 0.6 as an acceptable cut-off point for Cronbach's Alpha ^33,34 and composite reliability tests ^35,36 . For communality reliability test, the acceptable value for each variable was considered to be greater than 0.5.

The degree of inter-relation for items of each construct was assessed using convergent validity. To do that, the average variance extracted (AVE) and composite reliability (CR) were calculated and AVE> 0.5, and CR>AVE were considered as strong inter-relation of items. To assess the divergent validity of scale, we used the Fornel Larker test ^36,37 . The quality of model was examined by the blindfolding approach, proposed by Wold ^36,38 , so cv-communality and the cv-redundancy were used as an indicator for quality of measurement and structural model, respectively ^37,39 . Moreover, the goodness-of-fit (GOF) was used as a model quality assessment index. To evaluate GOF, the explained variance (R²) was calculated. GOF values ​​are compared with three values ​​of 0.10, 0.25 and 0.36, which show poor, moderate and strong predictive quality, respectively ^37,40 .

Results

Overall, 300 women aged 42.82 ±12.10 were included in the analysis. The majority of them were married (82%), and the housewives (78%). About 38% of women reported completing high school, followed by 35% who had some university education, and 27% with less education. Only 15% of participants described their socio-economic status as high.

Content and face validity

The first round of content validity evaluation revealed one item with CVI in clarity category less than 0.70 and ten items with a CVR less than 0.56. However, after modification, the CVI and CVR reached an acceptable level. Moreover, nine items were described as “unnecessary" by the experts and removed from the scale. Four items were merged because of semantic similarity. Moreover, some wording corrections were done in accordance with suggestions provided by experts and women. The final instrument contained 70 items including:

Behavior 1: 22 items for adding salt during cooking consisted of three items for precursors of attitude behavioral beliefs and outcome evaluation, three items for precursors of subjective norms (normative beliefs and motivation to comply), three items for precursors of perceived behavioral control (control beliefs and perceived power), and three items for behavioral intention (Table 1). The behavior was measured using a single item that asked” How many times have you cooked salty food in the past week?”

Behavior 2: 26 items for adding salt at the table consisted of four items for precursors of attitude, four items for precursors of subjective norms, three items for precursors of perceived behavioral control, and three items for behavioral intention (Table 1). The behavior was measured using a single item that asked “How many times have you added salt to the food at the table in the past week? “

Behavior 3: 22 items for purchasing salty foods consisted of three items for precursors of attitude, three items for precursors of subjective norms, three items for precursors of perceived behavioral control, and three items for behavioral intention (Table 1). The behavior was measured using a single item that asked “How many times have you bought salty foods in the past week? “

KMO and Bartlett test

The Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy was 0.74, 0.73, and 0.75 for behavior 1, 2 and 3, respectively. KMO values above 0.7 indicate adequate sampling ²⁸ . Bartlett’s test of sphericity was significant (P <0.000) for all three behaviors which indicated the variables were not uncorrelated.

Construct validity

The CFA was used to evaluate how well the theoretical framework behind instrument fitted the data. Results of CFA are shown in Table 1. The measurement models fit by using CFA has been shown in .

Close

Figure 1. Modified measurement model in standard coefficient estimation mode (Behavior 1)

Close

Figure 2. Modified measurement model in standard coefficient estimation mode (Behavior 2)

Close

Figure 3. Modified measurement model in standard coefficient estimation mode (Behavior 3)

According to homogeneity test, 14 items including three items of behavioral beliefs, three items of outcome evaluation, two items of normative beliefs, two items of motivation to comply questions, two items of control beliefs question, two items of perceived power were removed due to a factor loading less than 0.4. The measurement model was modified based on the remaining items.

Table 1 shows that convergent validity was confirmed by AVE values above 0.50 for constructs of three desired behaviors. Results of Fornell and Larcker tests showed that AVE square root of each construct was greater than existing correlations with other constructs, so the discriminant validity was confirmed (Table 2). Cronbach’s alpha values ranged from 0.65 to 1.00, therefore, the internal consistency reliability of all constructs proved adequate. In addition, the values of composite reliability for all constructs were within the acceptable range (more than 0.70) (Table 1).

The average value of cv-redundancy of the structural model for intentions was 0.17(values for intention1, intention 2, and intention 3, were 0.104, 0.220, and 0.190, respectively). This value for the structural model of behaviors was 0.15(values for behavior1, behavior 2, and behavior 3, were 0.03, 0.24, and 0.17, respectively). The obtained values indicate that the structural models were predictive ^37,39 .

Our model explains a small part of the variance of intention and behavior with an average global R² of 0.19 and 0.17, respectively. As PLS path modeling naturally lacks fit indices such as GFI, AGFI, and RMSEA (differently from SEM-ML) the goodness-of-fit (GOF) acts as an index for validating the model. We calculated GOF for each desired behavior by the following formula:

$$GOF=\sqrt{\overline{Communality}*\overline{R^2}}$$

Our results showed that the GOF index for behavior 1 was 0.250, for behavior 2 was 0.414 and for behavior 3 was 0.374. Based on guidelines suggested by Wetzels et al ⁴¹ GOF= 0.10, 0.25, and 0.36 should be considered as small, medium and large. Therefore, the overall predictive power of the model for behavior 1 was moderate, and for behavior 2 and behavior 3 the GOF value of the model was exceeding the large cut-off point indicating that the explaining power of the model was substantial for these behaviors.

Discussion

This study provides preliminary evidence for validity and reliability of developed scale for identifying Iranian Women’s determinants of behaviors related to salt consumption. To the best of our knowledge, there are no tools developed specifically for assessing the psychological constructs of salt consumption behaviors in women who are responsible for cooking. The recent evidence has supported partly the improvements in reductions in salt intake; however, controversy about the effectiveness of interventions largely arises from poor measurements ⁴² . Despite the vast application of the TPB to investigate factors associated with eating behaviors, there have not been many attempts to develop and validate assessment tools for measuring the psychological influencing factors on salt consumption behaviors. Indeed, many studies within this area have used the TPB scale without evaluating its psychometric properties. The literature review, cognitive interview, and content validity supported idea that salt consumption has different aspects and it is necessary to assess distinct behaviors including adding salt during cooking, adding salt at the table consisted, and purchasing salty food.

Within the existing instruments, none have assessed TPB constructs for different aspects of salt consumption behaviors in general population. ME Cornélio et al developed and investigated psychometric properties of an instrument to measure TPB constructs of salt consumption behaviors among hypertension patients. They assessed the scale only using the content validity and reliability ²⁰ .

The unique aspects of our instrument are the measurement of psychological determinants of three different salt intake behaviors rather than an aggregated behavior, including indirect measures of TPB constructs that may help in better identifying specifically targeted points for behavior change and assessing psychometric properties using various methods, including content validity, face validity, construct validity and reliability. Our reason for developing TPB constructs related to different aspects of salt reduction behaviors was to increase sensitivity of measure to capture psychological factors affecting each of these behaviors which are not necessarily similar. We identified the participants' salient beliefs through an elicitation study cited as a valuable stage in constructing a TPB questionnaire ²⁴ . However, in many TPB research, this component is overlooked. The elicitation study helped to inform the scale development and reflected the key socio-psychological factors of salt intake behaviors. Although our model accounted only for 19% and 17% of the variance in intention and behavior a small part of the variance of intention and behavior, respectively, this finding is somewhat in line with those of systematic review reported by McEachan et al ⁴³ . Through reviewing over 200 studies, TPB constructs explained 19% of the variance in behavior and 44% of the variance in intention. One explanation for low R² values for intention may be that women were very "homogeneous" as a group, especially for their attitudes. Another possible explanation may be that we splitting the salt-consumption behavior into three different sub- behaviors.

The development and psychometric testing of scale had limitations. First, employed women were under-represented and 78% of participants were housewives. Therefore, the generalizability of results to employed women is unknown. Moreover, because this scale was developed for use with a sample of the population with specific cultural food practices involving salt intake, additional studies are needed to determine the validity of this instrument in diverse populations. As with any self-reported measurement, our scale relied on participants' respondents and social desirability may have affected the response.

Conclusion

Our scale provides a validated and reliable instrument developed based on TPB for assessing different aspects of salt intake behaviors in women. This tool could be used to evaluate the effectiveness of interventions focusing on salt reduction behaviors.

Acknowledgements

This article is part of a Ph.D. thesis in health education and health promotion approved by Hamadan University of Medical Sciences. The authors thank all participants in this study. Ethical approval for study was provided by Hamadan University of Medical Sciences Ethical Committee whit code IR.UMSHA.REC.13970882.

Conflict of interest

The authors declared no conflicts of interest.

Funding

This work was supported by Hamadan University of Medical Sciences (grant number 9712218019).

Highlights

• One feature of this instrument is to address three different behaviors related to salt intake.

• Results of this study provide a reliable and valid instrument for investigating the salt intake behavior in one Iranian population.

• This instrument would be the basis to evaluate the effectiveness of interventions addressing influencing factors of salt intake behaviors.

References

1. World Health Organization. Healthy diet. WHO Web Site; 2020 [updated 29 April 2020; cited 28 August 2020]; Available from: http://www.who.int/news-room/fact-sheets/detail/healthy-diet.
2. Waśkiewicz A, Szcześniewska D, Szostak-Węgierek D, Kwaśniewska M, Pająk A, Stepaniak U. Are dietary habits of the Polish population consistent with the recommendations for prevention of cardiovascular disease?—WOBASZ II project. Kardiol Pol 2016; 74(9):969-77.
3. Suckling RJ, He FJ, MacGregor GA. Altered dietary salt intake for preventing and treating diabetic kidney disease. Cochrane Syst Rev 2010; 8(12):CD006763.
4. Park S, Joung J, Cho Y, Sohn SY, Hur K, Kim J. Effect of high dietary sodium on bone turnover markers and urinary calcium excretion in Korean postmenopausal women with low bone mass. Eur J Clin Nutr 2015; 69(3):361-6.
5. Penney S. Dropping the Salt: practical steps countries are taking to prevent chronic non-communicable diseases through population-wide dietary salt reduction. Public Health Agency of Canada. 2009.
6. Johnson C, Raj TS, Trieu K, Arcand J, Wong MM, McLean R. The science of salt: a systematic review of quality clinical salt outcome studies June 2014 to May 2015. J Clin Hypertens 2016; 18(9):832-9.
7. Huang L, Trieu K, Yoshimura S, Neal B, Woodward M, Campbell NR. Effect of dose and duration of reduction in dietary sodium on blood pressure levels: systematic review and meta-analysis of randomised trials. BMJ 2020; 368:m315.
8. Rezaei S, Mahmoudi Z, Sheidaei A, Aryan Z, Mahmoudi N, Gohari K. Salt intake among Iranian population: the first national report on salt intake in Iran. J Hypertens 2018; 36(12):2380-9.
9. Peykari N, Hashemi H, Dinarvand R, Haji-Aghajani M, Malekzadeh R, Sadrolsadat A. National action plan for non-communicable diseases prevention and control in Iran; a response to emerging epidemic. J Diabetes Metab Disord 2017; 16(1):3-10.
10. Trieu K, McMahon E, Santos JA, Bauman A, Jolly K-A, Bolam B. Review of behaviour change interventions to reduce population salt intake. Int J Behav Nutr Phy 2017; 14(1):17-32.
11. Cornelio ME, Godin G, Rodrigues RC, de Freitas Agondi R, Alexandre NM, Gallani M-CB. Effect of a behavioral intervention of the SALdável program to reduce salt intake among hypertensive women: A randomized controlled pilot study. Eur J Cardiovasc Nur 2016; 15(3):e85-e94.
12. Irwan AM, Kato M, Kitaoka K, Ueno E, Tsujiguchi H, Shogenji M. Development of the salt‐reduction and efficacy‐maintenance program in Indonesia. Nurs Health Sci 2016; 18(4):519-32.
13. Mullan B, Wong C, Kothe E. Predicting adolescent breakfast consumption in the UK and Australia using an extended theory of planned behaviour. Appetite 2013; 62:127-32.
14. Mullan B, Xavier K. Predicting saturated fat consumption: Exploring the role of subjective well-being. Psychol Health Med 2013; 18(5):515-21.
15. Close MA, Lytle LA, Chen D-G, Viera AJ. Using the theory of planned behavior to explain intention to eat a healthful diet among Southeastern United States office workers. Nutr Food Sci 2018; 48(2):365-74.
16. de Bruijn G-J, Kroeze W, Oenema A, Brug J. Saturated fat consumption and the Theory of Planned Behaviour: Exploring additive and interactive effects of habit strength. Appetite 2008; 51(2):318-23.
17. Ajzen I. The theory of planned behavior. Organizational behavior and human decision processes 1991; 50(2):179-211.
18. Rahimdel T, Morowatisharifabad M, Salehi-Abargouei A, Mirzaei M, Fallahzadeh H. Evaluation of an education program based on the theory of planned behavior for salt intake in individuals at risk of hypertension. Health Educ Res 2019; 34(3):268-78.
19. Rodrigues M, Rabelo-Silva E, Fuchs F, Fuchs S, Moreira L. Validity and reliability of the dietary sodium restriction questionnaire in patients with hypertension. Eur J Clin Nutr 2017; 71(4):552-4.
20. Cornélio ME, Gallani MCBJ, Godin G, Rodrigues RCM, Mendes RDR, Nadruz Junior W. Development and reliability of an instrument to measure psychosocial determinants of salt consumption among hypertensive patients. Rev Latinoam Hiperte 2009; 17(5):701-7.
21. Ipjian ML, Johnston CS. Smartphone technology facilitates dietary change in healthy adults. Nutrition 2017; 33:343-7.
22. Chen J, Liao Y, Li Z, Tian Y, Yang S, He C. Determinants of salt-restriction-spoon using behavior in China: Application of the health belief model. PloS One 2013; 8(12):e83262.
23. Ajzen I. Constructing a TPB questionnaire: Conceptual and methodological considerations. 2002.
24. Francis J, Eccles MP, Johnston M, Walker A, Grimshaw JM, Foy R, et al. Constructing questionnaires based on the theory of planned behaviour: A manual for health services researchers. Centre for Health Services Research, University of Newcastle upon Tyne; 2004.
25. Oluka OC, Nie S, Sun Y. Quality assessment of TPB-based questionnaires: a systematic review. PloS One 2014; 9(4):e94419.
26. Tilden VP, Nelson CA, May BA. Use of qualitative methods to enhance content validity. Nurs Res 1990; 39(3):172-5.
27. Lawshe CH. A quantitative approach to content validity 1. Pers Psychol 1975; 28(4):563-75.
28. Kline RB. Principles and practice of structural equation modeling. 4th ed. New York: Guilford Publications; 2015.
29. Lee T, Cai L, MacCallum RC. Power analysis for tests of structural equation models. Handbook of structural equation modeling: Guilford Press; 2012.
30. National plan to study the pattern of food consumption. IRIB News Agency; 2019 [updated 28 July 2019; cited 28 August 2020]; Available from: http://www.iribnews.ir/00Qld.
31. Hulland J. Use of partial least squares (PLS) in strategic management research: A review of four recent studies. Strategic Manage J 1999; 20(2):195-204.
32. Hair JF, Ringle CM, Sarstedt M. PLS-SEM: Indeed a silver bullet. Eur J Mark 2011; 19(2):139-52.
33. Taber KS. The use of Cronbach’s alpha when developing and reporting research instruments in science education. Res Sci ED 2018; 48(6):1273-96.
34. van Griethuijsen RA, van Eijck MW, Haste H, den Brok PJ, Skinner NC, Mansour N. Global patterns in students’ views of science and interest in science. Res Sci Educ 2015; 45(4):581-603.
35. Sim J, Wright C. Research in health care: concepts, designs and methods: Nelson Thornes. J Adv Nurs 2002; 39(2):208-10.
36. Hair Jr JF, Sarstedt M, Hopkins L, Kuppelwieser VG. Partial least squares structural equation modeling (PLS-SEM). Eur Bus Rev 2014; 26(2):106-21.
37. Hair Jr JF, Hult GTM, Ringle C, Sarstedt M. A primer on partial least squares structural equation modeling (PLS-SEM): Sage Publications; 2016.
38. Jöreskog KG, Wold HO. Systems under indirect observation: Causality, structure, prediction. North-Holland; New York: Sole distributors for the USA; 1988.
39. Cohen J. Statistical power analysis for the behavioral sciences. 2th ed. New York: Hillsdale NJ: Erihaum; 1988.
40. Tenenhaus M, Vinzi VE, Chatelin Y-M, Lauro C. PLS path modeling. Comput Stat Data An 2005; 48(1):159-205.
41. Wetzels M, Odekerken-Schröder G, Van Oppen C. Using PLS path modeling for assessing hierarchical construct models: Guidelines and empirical illustration. Mis Quart 2009; 1(33):177-95.
42. Johnson C, Santos JA, McKenzie B, Thout SR, Trieu K, McLean R. The science of salt: a regularly updated systematic review of the implementation of salt reduction interventions (September 2016–February 2017). J Clin Hypertens 2017; 19(10):928-38.
43. McEachan RRC, Conner M, Taylor NJ, Lawton RJ. Prospective prediction of health-related behaviours with the theory of planned behaviour: A meta-analysis. Health Psychol Rev 2011; 5(2):97-144.