Abstract

INTRODUCTION

The conditions that trigger the greatest proportion of health outcomes occur within the sociocultural environment in which individuals live, at a collective or contextual level. Over the last few decades, there has been renewed public health interest²⁷27. Santos SM. Desigualdades socioespaciais em saúde: incorporação de características de vizinhança nos modelos de determinação em saúde. In: Barcellos CC, organizador. A geografia e o contexto dos problemas de saúde. Rio de Janeiro: Abrasco; 2008. p.165-92. in the importance of the social and geographical context, especially in places where people live, and its impact on health. Scales for assessing neighborhood characteristics have been devised in order to pick up processes that occur in ordinary residential areas in which the population shares similar physical and social environmental conditions under which routine activities take place.¹⁸18. Macintyre S, Ellaway, A, Cummins S. Place effects on health: how can we conceptualise, operationalise and measure them? Soc Sci Med. 2002;55(1):125-39.

The effects of different neighborhoods on health have been shown, for example, in relation to the quality of life of the elderly population,⁴4. Baulfour JL, Kaplan GA. Neighborhood environment and loss of physical function in older adults: evidence from the Alameda County Study. Am J Epidemiol. 2002;155(6):507-15. DOI:10.1093/aje/155.6.507
https://doi.org/10.1093/aje/155.6.507... lifestyle habits,¹⁵15. Ellaway A, Anderson A, Macintyre S. Does area of residence affect body size and shape? Int J Obes Relat Metab Disord. 1997;21(4):304-8. self-reported health²² and cardiovascular diseases.¹⁰10. Diez-Roux A, Nieto FJ, Caulfield L, Tyroler HA, Watson RL, Szklo M. Neighborhood differences in diet: the Atherosclerosis Risk in Communities (ARIC) Study. J Epidemiol Community Health. 1999;53(1):55-63. ^{, 11}11. Diez-Roux AV, Merkin SS, Arnett D, Chambless L, Massing M, Nieto FJ, et al. Neighborhood of residence and incidence of coronary heart disease. N Engl J Med. 2001;345(2):99-106. DOI:10.1056/NEJM200107123450205
https://doi.org/10.1056/NEJM200107123450... ^{, 24}24. Ross CE. Walking, exercising, and smoking: does neighborhood matter? Soc Sci Med. 2000;51(2):265-74. Stressful factors present in the physical and social environment of the neighborhood have been shown to be positively associated with higher prevalence of diabetes,³3. Auchincloss AH, Diez-Roux AV, Mujahid MS, Shen M, Bertoni AG, et al. Neighborhood resources for physical activity and healthy foods and incidence of type 2 diabetes mellitus: The Multi-Ethnic Study of Atherosclerosis. Arch Intern Med. 2009;169(18):1698-704. DOI:10.1001/archinternmed.2009.302
https://doi.org/10.1001/archinternmed.20... obesity,²¹ acute myocardial infarction,⁶6. Chaix B, Lindström M, Rosvall M, Merlo J. Neighbourhood social interactions and risk of acute myocardial infarction. J Epidemiol Community Heath. 2008:62(1):62-8. DOI:10.1136/jech.2006.056960
https://doi.org/10.1136/jech.2006.056960... smoking¹³13. Echeverría S, Diez-Roux AD, Shead S, Borelle LN, Jackson S. Associations of neighborhood problems and neighborhood social cohesion with mental health and health behaviors: the Multi-Ethnic Study of Atherosclerosis. Health Place. 2008;14(4):853-65. DOI:10.1016/j.healthplace.2008.01.004
https://doi.org/10.1016/j.healthplace.20... and depression.¹³13. Echeverría S, Diez-Roux AD, Shead S, Borelle LN, Jackson S. Associations of neighborhood problems and neighborhood social cohesion with mental health and health behaviors: the Multi-Ethnic Study of Atherosclerosis. Health Place. 2008;14(4):853-65. DOI:10.1016/j.healthplace.2008.01.004
https://doi.org/10.1016/j.healthplace.20... As well as diseases that can be medically diagnosed, studies have also addressed the influence of the neighborhood on subclinical markers of chronic diseases¹⁷ and on variations in the levels of cortisol,¹² a hormone related to stress.

The contextual characteristics that potentially influence health can be grouped into socioeconomic, physical and psychosocial dimensions.²⁶26. Santos SM, Chor D, Werneck GL, Coutinho ESF. Associação entre fatores contextuais e auto-avaliação de saúde: uma revisão sistemática de estudos multinível. Cad Saude Publica. 2007;23(11):2533-54. DOI:10.1590/S0102-311X2007001100002
https://doi.org/10.1590/S0102-311X200700... Secondary databases have comprised the main source of contextual information in recent studies. However, individuals' responses (primary data) to relevant questions about processes that occur in the neighborhood and also the definitions of measurements inherent to the contextual level⁹9. Cummins S, Macintyre S, Davidson S, Ellaway A. Measuring neighbourhood social and material context: generation and interpretation of ecological data from routine and non-routine sources. Health Place. 2005;11(3):249-60. DOI:10.1016/j.healthplace.2004.05.003
https://doi.org/10.1016/j.healthplace.20... ^{, 19}19. Mitchell R, Gleave S, Bartley M, Wiggins D, Joshi H. Do attitude and area influence health? A multilevel approach to health inequalities. Health Place. 2000;6(2):67-79. DOI:10.1016/S1353-8292(00)00004-6
https://doi.org/10.1016/S1353-8292(00)00... (e.g., the definition of "neighborhood" itself) still form gaps in the knowledge of this field.

With the objective of investigating the relationships between self-reported characteristics of the neighborhood and occurrences of cardiovascular diseases and diabetes, which constitute the main outcomes of the Longitudinal Study for Adult Health (ELSA-Brasil),²2. Aquino EM, Barreto SM, Benseñor IM, Carvalho MS, Chor D, Duncan BB, et al. Brazilian Longitudinal Study of Adult Health (ELSA-Brasil): objectives and design. Am J Epidemiol. 12;175(4):315-24. DOI:10.1093/aje/kwr294
https://doi.org/10.1093/aje/kwr294... specific scales were included in its baseline questionnaire. Five domains were studied: social cohesion, environment suitable for physical activity; availability of healthy foods; safety in relation to crime; perceived violence; and one question about victimization.

These scales were available and had been validated only in English.²⁰20. Mujahid MS, Diez-Roux AV, Morenoff JD, Raghunathan T. Assessing the measurement properties of neighborhood scales: from psychometrics to ecometrics. Am J Epidemiol. 2007;165(8):858-67. DOI:10.1093/aje/kwm040
https://doi.org/10.1093/aje/kwm040... ^{, 25}25. Sampson RJ, Raudembush SW, Earls F. Neighborhoods and violent crime: a multilevel study of collective efficacy. Science. 1997;277(5328):918-24. DOI:10.1126/science.277.5328.918
https://doi.org/10.1126/science.277.5328... The present article describes the process of cross-cultural adaptation performed within the scope of ELSA-Brasil, and estimates for test-retest reliability.

METHODS

Instrument Selection

A bibliographic review on this topic showed that no validated instrument for measuring contextual characteristics existed among Brazilian studies published up to 2007. Among studies in other countries, two highly important studies that used self-reporting scales for measuring neighborhood characteristics relating to cardiovascular diseases were identified: the Project on Human Development in Chicago Neighborhoods (PHDCN)²⁵25. Sampson RJ, Raudembush SW, Earls F. Neighborhoods and violent crime: a multilevel study of collective efficacy. Science. 1997;277(5328):918-24. DOI:10.1126/science.277.5328.918
https://doi.org/10.1126/science.277.5328... and the Multi-Ethnic Study of Atherosclerosis (MESA).²⁰20. Mujahid MS, Diez-Roux AV, Morenoff JD, Raghunathan T. Assessing the measurement properties of neighborhood scales: from psychometrics to ecometrics. Am J Epidemiol. 2007;165(8):858-67. DOI:10.1093/aje/kwm040
https://doi.org/10.1093/aje/kwm040... In these studies, the neighborhood scales presented good reproducibility (intraclass correlation coefficient ranging from 0.73 to 0.91), which formed an important condition to be included in ELSA-Brasil.

Measurement Instruments

In ELSA-Brasil, questions that assessed the participants' perceptions regarding psychosocial and physical characteristics of the environment of their neighborhood were included. At the beginning of the section containing these questions, the interviewees were guided to think of their neighborhood as: "the general area surrounding their home where you usually perform routine activities such as going shopping, going to the park or visiting neighbors".²⁵25. Sampson RJ, Raudembush SW, Earls F. Neighborhoods and violent crime: a multilevel study of collective efficacy. Science. 1997;277(5328):918-24. DOI:10.1126/science.277.5328.918
https://doi.org/10.1126/science.277.5328...

Before answering the first set of scales, the interviewee was guided to choose the best answer among the following options: 1 - thoroughly agree; 2 - partially agree; 3 - neither agree nor disagree; 4 - partially disagree; and 5 - thoroughly disagree. These options were presented on a card, in relation to each item. This format takes into account the following scales: 1) social cohesion, with five items; 2) environment suitable for physical activity, with nine items; 3) availability of healthy foods, with four items; and 4) safety, with four items (Table 1).

Likewise, before the items on the scale about perceived violence (with five items), the interviewee was guided to answer how often the events described had happened during the past six months, according to what he/she knew about his/her neighborhood, and not just what had been witnessed or suffered, choosing the best answer among the following options: 1 - often; 2- sometimes; 3 - rarely; or 4 - never. Lastly, the interviewee was asked how long he/she had live in the neighborhood and about personal victimization: "1. Has anyone ever subjected you to violence (robbery, fighting, sexual violence or kidnapping), or has this happened to anyone who lives in your household in this neighborhood, during the time that you have being living in this house?", with "yes" or "no" as the answer options.

Instrument Adaptation Process

The cross-cultural adaptation of the scales in English was performed in accordance with the methodology proposed by Herdman et al,¹⁶16. Herdman M, Fox-Rushby J, Badia X. A model of equivalence in the cultural adaptation of HRQoL instruments: the universalist approach. Qual Life Res.1998;7(4):323-35. as already used in other studies.¹1. Alves MGM, Chor D, Faerstein E, Lopes CS, Werneck GL. Versão resumida da "job stress scale": adaptação para o português. Rev Saude Publica. 2004;38(2):164-71. DOI:10.1590/S0034-89102004000200003
https://doi.org/10.1590/S0034-8910200400... ^{, 8}8. Chor D, Werneck GL, Faerstein E, Alves MGM, Rotenberg L. The Brazilian version of the effort-reward imbalance questionnaire to assess job stress. Cad Saude Publica. 2008;24(1):219-24. DOI:10.1590/S0102-311X2008000100022
https://doi.org/10.1590/S0102-311X200800... This allows evaluation of six types of equivalence between the instruments in the original language and in their translations: 1) conceptual; 2) between items; 3) semantic; 4) operational; 5) measurements; and 6) functional.

In order to obtain the semantic equivalence, the translation from English into Portuguese was performed by two independent Brazilian translators who were fluent in English. Both translations were analyzed, together, by the translators and researchers from ELSA-Brasil. For each sentence, the translators evaluated the degree of difficulty that they had in performing the translation. This analysis ended with production of a consensus version of the translation, which was forwarded to a third professional for back-translation. This back-translation was performed by an American translator who was fluent in Brazilian Portuguese. This English version was compared with the original scale by two Brazilian specialists who were fluent in English, one with experience in using scales in the field of epidemiology, and the other with experience of the topic of collective effectiveness, accompanied by two researchers from ELSA-Brasil.

The strategy used consisted of three times: 1) Comparison of the back-translated scale with the original by two independent evaluators according to two criteria: a) whether there was any change in the meaning, in general terms, through dichotomous evaluation (altered/unaltered meaning); b) quality of the back-translation, in literal terms, according to evaluation using the Likert scale: 1 = Very poor, 2 = Poor, 3 = Fair, 4 = Good and 5 = Excellent; 2) Comparison between the evaluations by the two specialists in order to identify contrasts, for example a situation in which one evaluator attributed grade 2, and the other, grade 4, for the same evaluated item; 3) Individual evaluation of the item, when one of the evaluators attributed a grade less than or equal to 3 (fair, poor or very poor) or considered that the meaning of the item had been changed (differing from the original).

The translation resulting from this process was subjected to probing in order to clarify any doubts (with workers from another public institution with characteristic similar to those of ELSA-Brasil), along with three stages of pretests and two pilot studies in conjunction with other sections of the questionnaire. The final format of the scales was obtained from these stages and was included in the questionnaire.⁷7. Chor D, Alves MGM, Giatti L, Cade NV, Nunes MA, Molina MCB, et al. Questionário do ELSA-BRASIL: desafios na elaboração de instrumento multidimensional. Rev Saude Publica. 2013;47(Supl 2):27-36.

Design of the test-retest reliability study

ELSA-Brasil is a cohort study on approximately 15,000 workers at six Brazilian teaching and research institutions, both with active and with retired status, aged between 35 and 74 years of age.²2. Aquino EM, Barreto SM, Benseñor IM, Carvalho MS, Chor D, Duncan BB, et al. Brazilian Longitudinal Study of Adult Health (ELSA-Brasil): objectives and design. Am J Epidemiol. 12;175(4):315-24. DOI:10.1093/aje/kwr294
https://doi.org/10.1093/aje/kwr294... A total of 261 volunteer participants from six ELSA-Brasil investigation centers were included in the test-retest reliability study on the neighborhood scales, in accordance with previously established quotas regarding sex, age and educational level.

Data-gathering for the "test" was performed during the processes of interviewing/examining the study participants. Each participant was invited to answer again the sections of the questionnaire that were included in the reliability study, in order to fulfill the defined sampling quota. The second application of questions (retest) of these sections was performed by the same interviewer, seven to 14 days after the first application.

Statistical Analyses on Instrument Reliability

The responses to the test and retest were inserted in a computerized database through double independent data entry in the EpiInfo software, with subsequent correction of inconsistencies.

Two components were evaluated in the test-retest reliability analysis: the internal consistency of each domain, by estimating Cronbach's alpha; and the temporal stability of the measurements, by means of the intraclass correlation coefficient (ICC) for the scores and the kappa (k) and prevalence-adjusted kappa (ka) indexes for the question on victimization (dichotomous yes/no). This was done according to the following characteristics of the participants: sex, age (35-54 or 55-74) and educational level (elementary, high school or higher education).

The ICC was applied to the scores, both from the test and retest, resulting from the sum of the responses obtained in each item for each scale. Some items from the scales received reverse coding (e.g., items 4 and 5 of the social cohesion scale, as shown in Table 1). For the social cohesion, environment suitable for physical activity, availability of healthy foods, and safety in relation to crimes scales, the greater the score was, the worse the quality of the set of characteristics of that domain also was. For the perceived violence scale, the greater the score was, the lower the frequency of occurrences of violence was. The reliability of the individual items of each scale was estimated using quadratic weighted kappa (κp).

The cutoffs suggested by Byrt et al⁵5. Byrt T, Bishop J, Carlin JB. Bias, prevalence and kappa. J Clin Epidemiol. 1993;46(5):423-9. DOI:10.1016/0895-4356(93)90018-V
https://doi.org/10.1016/0895-4356(93)900... for classifying the level of stability of the responses were used: weak (0 to 0.20); mild (0.21 to 0.40); reasonable (0.41 to 0.60); good (0.61 to 0.80); very good (0.81 to 0.92); and excellent (0.93 to 1.00).

RESULTS

In this section, the stages for semantic, items, operational and measurement equivalence are described.

The items that comprise each scale, in the original version and the version adapted for Portuguese, are presented in Table 1.

In the stage of evaluating semantic equivalence, among the 28 items analyzed, only five presented divergence or doubt among the evaluators. Two of them were in the social cohesion domain: 1- "This is a close-knit or unified neighborhood"; 2- "People in this neighborhood generally DON'T get along with each other". In the first case, the item was translated as "very cohesive". However, according to surveys, the meaning of this term was not clear enough for the interviewees. Regarding the second item, the evaluators considered that negative expressions ("people... DON'T get along with...) might confuse the responses of the interviewees. Nevertheless, the phrasing was kept, in accordance with the original conception of the scale by Sampson et al,²⁵25. Sampson RJ, Raudembush SW, Earls F. Neighborhoods and violent crime: a multilevel study of collective efficacy. Science. 1997;277(5328):918-24. DOI:10.1126/science.277.5328.918
https://doi.org/10.1126/science.277.5328... and interviewers were given the guidance that negative expressions should be emphasized, and the word DON'T was highlighted in the questionnaire.

In the item "There are many fast-food snack-bars in my neighborhood", in the "availability of healthy foods" domain, it was decided to explain the desired meaning in the Portuguese version, while retaining the English expression "fast food", since this is used colloquially in social groups similar to the population of ELSA-Brasil.

Finally, two items in the "safety" domain deserved special attention in several stages: 1- "My neighborhood is safe from crime"; and 2- "Violence is a problem in my neighborhood". The first of these was initially translated as "My neighborhood is crime-free". However, the evaluators concluded that the literal translation would give a very radical idea about the absence of crime, which should be made explicit in this case. Thus, another sentence that transmitted the meaning more properly in the Brazilian context was chosen. The other item was initially literally translated, including the word NOT. The evaluators' opinion concerning difficulties with negative expressions was confirmed in surveys and pretests. Thus, it was decided to change the expression into an affirmative one since there was a further reason for doubts among the interviewees in this case, since they would give contradictory responses to these two items, which were presented one after the other. Thus, for example, they agreed that their neighborhood was safe in relation to crimes, and at the same time disagreed that violence was not a problem in the same neighborhood.

The average age of the respondents in the test-retest reliability study was 52.5 years (standard deviation 8.7); 49.4% (129) were women, and 50.6% (132) were men. The mean length of time living in the neighborhood was 18.1 years (standard deviation 15.1).

Descriptive statistics on the stage of equivalence of measurements on the scales are presented in Table 2. Cronbach's alpha ranged from 0.64 (social cohesion) to 0.84 (availability of healthy foods, after exclusion of item 4, as shown in Table 1, due to the poor performance presented by this item).

The reproducibility of the overall scores ranged from good to very good. The scales with the highest stability values were environment suitable for physical activity and availability of healthy foods, followed by perceived violence, safety and social cohesion (Table 3). In the stratified analyses, the intraclass correlation coefficients were above 0.78 for all strata of sex, age, and educational level. No important variations or consistent patterns could be identified according to the strata of these characteristics. The question about victimization presented an overall kappa index (κ) of 0.66, with notable variation among the strata, with the lowest performance in the stratum of elementary educational level (κ = 0.36). After adjustment according to prevalence, the general index of this question presented the greatest temporal stability (overall κa = 0.71) with reduction of the difference among the strata of educational level (elementary κa = 0.56; high school κa = 0.70; higher education κa = 0.73).

The weighted kappa statistics (κp) for individual evaluation of the items ranged from 0.30 to 0.42 for sentences with statements with negative expressions (not), and from 0.54 to 0.74 for the other sentences.

DISCUSSION

According to the scheme proposed by Herdman et al,¹⁶16. Herdman M, Fox-Rushby J, Badia X. A model of equivalence in the cultural adaptation of HRQoL instruments: the universalist approach. Qual Life Res.1998;7(4):323-35. the evaluation of the five types of equivalence (conceptual, among items, semantic, operational and measurement) suggested that there is functional equivalence between the scales in English and in Portuguese.

The internal consistency estimates on the scales evaluated presented values that were similar to those of reference studies such as the one carried out by Echeverría et al,¹⁴14. Echeverría SE, Diez-Roux A, Link BG. Reliability of self-reported neighborhood characteristics. J Urban Health. 2004;81(4):682-700. DOI:10.1093/jurban/jth151
https://doi.org/10.1093/jurban/jth151... in which Cronbach's alpha ranged from 0.77 to 0.94, and the MESA study, with a range from 0.73 to 0.89,²⁰20. Mujahid MS, Diez-Roux AV, Morenoff JD, Raghunathan T. Assessing the measurement properties of neighborhood scales: from psychometrics to ecometrics. Am J Epidemiol. 2007;165(8):858-67. DOI:10.1093/aje/kwm040
https://doi.org/10.1093/aje/kwm040... considering all scales.

Like in the MESA study,²⁰20. Mujahid MS, Diez-Roux AV, Morenoff JD, Raghunathan T. Assessing the measurement properties of neighborhood scales: from psychometrics to ecometrics. Am J Epidemiol. 2007;165(8):858-67. DOI:10.1093/aje/kwm040
https://doi.org/10.1093/aje/kwm040... the item about fast food was excluded from the scale of availability of healthy foods, which improved its internal consistency. Thus, all future analyses of this domain will be based on scores obtained through summing its first three items.

Considering the cutoff points for the stability levels for responses suggested by Byrt et al,⁵5. Byrt T, Bishop J, Carlin JB. Bias, prevalence and kappa. J Clin Epidemiol. 1993;46(5):423-9. DOI:10.1016/0895-4356(93)90018-V
https://doi.org/10.1016/0895-4356(93)900... the victimization scale presented good stability (0.66), and the social cohesion, safety and perceived violence scales, followed by the availability of healthy foods and environment suitable for physical activity scales presented very good stability (ICC 0.83 to 0.90).

These results are similar to those of American studies, in which high reproducibility was estimated for all scales. In the pilot study developed by Echeverría et al¹⁴ among 48 volunteers living in New York, the ICC ranged from 0.78 to 0.91 (all scales). In the case of the investigation conducted by Mujahid et al²⁰20. Mujahid MS, Diez-Roux AV, Morenoff JD, Raghunathan T. Assessing the measurement properties of neighborhood scales: from psychometrics to ecometrics. Am J Epidemiol. 2007;165(8):858-67. DOI:10.1093/aje/kwm040
https://doi.org/10.1093/aje/kwm040... (MESA), the scales used presented slightly lower coefficients, ranging from 0.60 to 0.88, and were applied by telephone to a sample of 120 individuals.

It should be noted that the subsample of study participants was distributed similarly to the whole population of ELSA-Brasil, with regard to sex, age group and educational level. Thus, it was similar to the population for which the instrument was targeted. This, combined with the results from the stages of semantic, item and operational equivalence, indicates that the equivalence of measurements that was the aim was obtained, and that the estimates of the statistical analyses represented the real reliability of the instruments.

We consider that the favorable results found in the present study, i.e. indicating that self-reported measurements of neighborhood characteristics are reproducible in Brazil, are fundamental for enabling analysis on mechanisms that can explain how the characteristics of neighborhoods can affect health.

The next step in using self-reported measurement scales on neighborhood characteristics is to study correlations between the participants who live in the same neighborhood, by means of ecometric techniques.²⁰20. Mujahid MS, Diez-Roux AV, Morenoff JD, Raghunathan T. Assessing the measurement properties of neighborhood scales: from psychometrics to ecometrics. Am J Epidemiol. 2007;165(8):858-67. DOI:10.1093/aje/kwm040
https://doi.org/10.1093/aje/kwm040... ^{, 23}23. Raudenbush SW, Sampson RJ. Ecometrics: toward a science of assessing ecological settings, with application to the systematic social observation of neighborhoods. Sociol Methodol. 1999;29(1):1-41. DOI:10.1111/0081-1750.00059
https://doi.org/10.1111/0081-1750.00059... This approach will provide estimates of the consistency between the items on each scale, among participants of the same neighborhood, and between various neighborhoods simultaneous

References

Publication Dates

History