Abstract

Objective

to validate self-reported body weight of Programa Academia da Saúde (PAS) users in Belo Horizonte, MG, Brazil, and to identify factors associated with weight error.

Methods

self-reported body weight, obtained by telephone interview, was compared to measured weight; we used Student’s t-test, ANOVA, Lin’s concordance correlation coefficient, the Bland-Altman method and Kappa coefficient; women’s self-reported weight was corrected according to measured weight using multiple regression.

Results

441 users participated; weight self-reported by men was valid (error=0; p=0.15); overweight classification concordance was 94.3% (Kappa=0.88); errors were greater in the case of self-reported weight by women over 30 years old and overweight women (-0.8kg; error≠0; p<0.01); after correction using multiple regression, self-reported weight validity was satisfactory (error=0; p=0.99).

Conclusion

self-reported weight of men can be used for research and health surveillance, but for women correction is required.

Body Weight; Validation Study; Health Programs and Plans; Self Report

Introduction

Growing prevalence of excess weight, obesity and other noncommunicable chronic diseases (NCDs) have a heavy impact on Public Health and highlight the importance of studies of the population’s nutritional status.¹1. Swinburn BA, Kraak VI, Allender S, Atkins VJ, Baker PI, Bogard JR, et al. The global syndemic of obesity, undernutrition, and climate change: The Lancet commission report. Lancet [Internet]. 2019 Jan [cited 2020 May 6]; 393(10173):791-846. Available from: https://doi.org/10.1016/S0140-6736(18)32822-8
https://doi.org/10.1016/S0140-6736(18)32...

Self-reported measurements of body weight are frequently used due to the convenience and low cost of obtaining information, whether by telephone or by online questionnaire.²2. Wright FL, Green J, Reeves G, Beral V, Cairns BJ, Million Women Study collaborators. Validity over time of self-reported anthropometric variables during follow-up of a large cohort of UK women. BMC Med Res Methodol [Internet]. 2015 Oct [cited 2020 May 6];15:81. Available from: https://doi.org/10.1186/s12874-015-0075-1
https://doi.org/10.1186/s12874-015-0075-... Their use, however, is subject to greater error than measurements taken by examiners.⁴4. Bes-Rastrollo M, Valdivieso JRP, Sánchez-Villegas A, Alonso A, Martínez-González MA. Validación del peso e índice de masa corporal auto-declarados de los participantes de una cohorte de graduados universitarios. Rev Esp Obes [Internet]. 2005 nov [citado 2020 maio 6];3(6):352-58. Disponível em: https://www.researchgate.net/publication/255632320_Validacion_del_peso_e_indice_de_masa_corporal_auto-declarados_de_los_participantes_de_una_cohorte_de_graduados_universitarios_in_Spanish
https://www.researchgate.net/publication... Be that as it may, above all it is important to analyze the validity of self-reported information.³3. Carvalho AM, Piovezan LG, Selem SSAC, Fisberg RM, Marchioni DML. Validação e calibração de medidas de peso e altura autorreferidas por indivíduos da cidade de São Paulo. Rev Bras Epidemiol [Internet]. 2014 jul-set [citado 2020 maio 6];17(3):735-46. Disponível em: https://doi.org/10.1590/1809-4503201400030013
https://doi.org/10.1590/1809-45032014000...

Although several studies have shown the validity of self-reported weight,⁴4. Bes-Rastrollo M, Valdivieso JRP, Sánchez-Villegas A, Alonso A, Martínez-González MA. Validación del peso e índice de masa corporal auto-declarados de los participantes de una cohorte de graduados universitarios. Rev Esp Obes [Internet]. 2005 nov [citado 2020 maio 6];3(6):352-58. Disponível em: https://www.researchgate.net/publication/255632320_Validacion_del_peso_e_indice_de_masa_corporal_auto-declarados_de_los_participantes_de_una_cohorte_de_graduados_universitarios_in_Spanish
https://www.researchgate.net/publication... the quality of this information has not yet been investigated among health service users who routinely participate in initiatives such as the Health Academy Program - HAP (Programa Academia da Saúde, PAS, in portuguese) a Primary Healthcare service of National Healthcare System (in portuguese: Sistema Único de Saúde, SUS) aimed at health promotion and health care, especially for patients with NCDs.¹²12. Brasil. Ministério da Saúde. Portaria GM/MS no 719, de 7 de abril de 2011. Institui o Programa Academia da Saúde no âmbito do Sistema Único de Saúde. [Internet]. Diário Oficial da União, Brasília (DF), 2011 abr 8 [citado 2020 maio 6];Seção 1:52. Disponível em: https://bvsms.saude.gov.br/bvs/saudelegis/gm/2011/prt0719_07_04_2011.html
https://bvsms.saude.gov.br/bvs/saudelegi... People who use PAS have peculiar characteristics such as apparently being more disposed to looking after their health and taking part in six-monthly checkups, including having their weight measured. These aspects can interfere directly in reporting of body weight, resulting in more precise measurements.¹³13. Mendonça RD, Horta PM, Santos LC, Lopes ACS. The dietary profile of socially vulnerable participants in health promotion programs in a Brazilian metropolis. Rev Bras Epidemiol [Internet]. 2015 Apr-Jun [cited 2020 May 6];18(2):454-65. Available from: https://doi.org/10.1590/1980-5497201500020013
https://doi.org/10.1590/1980-54972015000...

Valid body weight measurements self-reported by PAS users can contribute to epidemiological studies being conducted in a simpler and less costly manner, enabling greater agility and efficiency in longitudinal monitoring of health service users and surveillance by telephone of those who stop using the service. Self-reporting of weight can reduce the need for human resources, time and equipment, favoring the sustainability of health surveillance actions.³3. Carvalho AM, Piovezan LG, Selem SSAC, Fisberg RM, Marchioni DML. Validação e calibração de medidas de peso e altura autorreferidas por indivíduos da cidade de São Paulo. Rev Bras Epidemiol [Internet]. 2014 jul-set [citado 2020 maio 6];17(3):735-46. Disponível em: https://doi.org/10.1590/1809-4503201400030013
https://doi.org/10.1590/1809-45032014000...

The objective of this study was to validate self-reported body weight of users of the Programa Academia da Saúde - PAS – and to verify factors associated with errors in self-reported weight.

Methods

A validation study was conducted in order to compare self-reported body weight information with measured body weight of PAS users in Belo Horizonte, in the state of Minas Gerais, Brazil.¹⁴14. Menezes MC, Costa BVL, Ferreira NL, Freitas PP, Mendonça RD, Lopes MS, et al. Percurso metodológico de ensaio comunitário controlado em serviço de saúde: pesquisa epidemiológica translacional em Nutrição. Demetra [Internet]. 2017 [citado 2020 maio 6];12(4):1203-22. Disponível em: https://doi.org/10.12957/demetra.2017.28451
https://doi.org/10.12957/demetra.2017.28...

The data were obtained by means of a community randomized controlled trial (CRCT), conducted between March 2013 and March 2018, with the objective of performing nutritional intervention and assessing its effectiveness. Further details of this research can be found in Menezes et al.¹⁴14. Menezes MC, Costa BVL, Ferreira NL, Freitas PP, Mendonça RD, Lopes MS, et al. Percurso metodológico de ensaio comunitário controlado em serviço de saúde: pesquisa epidemiológica translacional em Nutrição. Demetra [Internet]. 2017 [citado 2020 maio 6];12(4):1203-22. Disponível em: https://doi.org/10.12957/demetra.2017.28451
https://doi.org/10.12957/demetra.2017.28... We used baseline sociodemographic and height data gathered at the CRCT (time 0 – 2013/2014), body weight measurements obtained through telephone interviews (carried out 36 months after baseline – 2016/2017) and then validated in face-to-face interviews. Landline and cell telephone numbers were obtained during the baseline interviews.

The study was conducted in the context of the PAS. This program has infrastructure, equipment and qualified professionals. In Belo Horizonte, the program’s units offer guided physical exercise and health promotion, maintenance and recovery actions in partnership with the Family Health Strategy .¹⁵15. Lopes ACS, Ferreira AD, Mendonça RD, Dias MAS, Rodrigue RCLC, Santos LC. Estratégia de promoção à saúde: programa academia da cidade de Belo Horizonte. Rev Bras Ativ Fís Saúde [Internet]. 2016 jul [citado 2020 maio 6];21(4):379-84. Disponível em: https://doi.org/10.12820/rbafs.v.21n4p%25p
https://doi.org/10.12820/rbafs.v.21n4p%2... The PAS was chosen as the scenario for the study because it is an important element of the Chronic Diseases Care Network and is a privileged space for developing interventions to promote healthy lifestyles, in addition to its health surveillance potential.¹⁶16. Ministério da Saúde (BR). Academia da saúde [Internet]. Brasília: Ministério da Saúde; 2014 [citado 2020 maio 6]. Disponível em: https://bvsms.saude.gov.br/bvs/publicacoes/academia_saude_cartilha.pdf
https://bvsms.saude.gov.br/bvs/publicaco...

Forty-two out of the fifty PAS units operating in the municipality at the time the study began were considered eligible for the sample. The inclusion criteria for the participating units were: not having been involved in food and nutrition studies in the last 24 months; and being located in areas with a medium or high health vulnerability index (HVI), prioritized for having this service in the municipality. HVI is an indicator comprised of socio-economic and environmental variables. It ranges from 0 to 1: the higher the score, the greater health vulnerability is. Risk is classified as being very high, high, medium or low.¹⁷17. Nahas MIP. Experiência de construção e perspectivas de aplicabilidade de índices e indicadores na gestão urbana da qualidade de vida: uma síntese da experiência de Belo Horizonte. In: Seminário sobre Indicadores de Sustentabilidade, realizado no âmbito do Projeto "Redistribuição da População e Meio Ambiente: São Paulo e Centro-Oeste" [Internet]; 2000 abr 10-11; Campinas. 2000 [citado 2020 maio 6]. 24p. Disponível em: https://www.sigas.pe.gov.br/files/06132017020740-construcao.de.indicadores.sociais.pdf
https://www.sigas.pe.gov.br/files/061320...

The participating units were selected by sampling by clusters. Sampling was stratified in each of Belo Horizonte’s nine administrative regions. Two units were randomly selected in each geographic stratum, resulting in a total of 18 (42.8%) participating units, considered to be representative of the municipality’s medium, high and very high HVI units, with a 95% confidence interval and error of less than 1.4%, calculated a posteriori .¹⁴14. Menezes MC, Costa BVL, Ferreira NL, Freitas PP, Mendonça RD, Lopes MS, et al. Percurso metodológico de ensaio comunitário controlado em serviço de saúde: pesquisa epidemiológica translacional em Nutrição. Demetra [Internet]. 2017 [citado 2020 maio 6];12(4):1203-22. Disponível em: https://doi.org/10.12957/demetra.2017.28451
https://doi.org/10.12957/demetra.2017.28...

At each selected PAS unit all service users aged 20 or over who had regularly attended the service activities during the last month (attending at least one lesson) were invited to take part in the study. The exclusion criteria were (i) being pregnant and (ii) having a disability making it impossible to answer the interview questions.¹⁴14. Menezes MC, Costa BVL, Ferreira NL, Freitas PP, Mendonça RD, Lopes MS, et al. Percurso metodológico de ensaio comunitário controlado em serviço de saúde: pesquisa epidemiológica translacional em Nutrição. Demetra [Internet]. 2017 [citado 2020 maio 6];12(4):1203-22. Disponível em: https://doi.org/10.12957/demetra.2017.28451
https://doi.org/10.12957/demetra.2017.28...

All participants with telephone numbers recorded at baseline were contacted for CRCT reassessment. Four attempts were made to call each available telephone number on different days (Monday to Saturday) and at different times (morning, afternoon and night). The interview could also be arranged for a specific time depending on the service user’s needs. If contact was not made after four attempts (telephone unavailable or busy or calls directed to the message box), the research team checked the PAS registry to see if there was another contact number and, if there was, a further four attempts were made. If there was no answer the user was excluded from the study.

The research team interviewed 2,371 users by telephone and 15% of them were selected to take part in the self-reported body weight validation study. The sample was calculated based on previous validation studies,⁷7. Miranda AES, Ferreira AVM, Oliveira FLP, Hermsdorff HHM, Bressan J, Pimenta AM. Validação da síndrome metabólica e de seus componentes autodeclarados no estudo CUME. REME [Internet]. 2017 [citado 2020 maio 6];21:e1069. https://doi.org/10.5935/1415-2762.20170079
https://doi.org/10.5935/1415-2762.201700... with the aim of achieving a 95% data confidence level, capable of representing the PAS users attending the units participating in the study and enabling validation of self-reported weight. The sample was stratified for each PAS unit in the study according to user attendance at the service (attending at least one lesson in the last month), sex (female; male), age range (in years: 20-29; 30-59; 60 or over) and nutritional status (low weight/good nutrition; overweight/obesity]).¹⁸18. World Health Organization - WHO. Preventing chronic diseases: a vital investment [Internet]. Geneva: World Health Organization; 2005 [cited 2020 May 6]. Available from: https://www.who.int/chp/chronic_disease_report/en/
https://www.who.int/chp/chronic_disease_... The random selection of the sample was performed so as to contain, when possible, two people in each established class, with the aim of minimizing possible bias in self-reported answers in relation to the characteristics of the participants. The sample power was 99%, calculated a posteriori .

Random selection of sample participants was done with the aid of an online program (www.sorteiador.org). If a user refused to take part, was not contacted after three telephone call attempts or did not attend three appointments, they were replaced by means of a new draw using the same stratum.

Data were collected face-to-face at time 0, and at face-to-face validation at the user’s registered PAS unit; as well as by telephone interview 36 months after initial assessment.

The following baseline (time 0) variables were used:

1. sex (male; female);

2. age (in years);

3. schooling (in years of study);

4. marital status (married/common law marriage; separated/single/widowed);

5. occupation (retired/pensioner, unemployed; housewife; other); and

6. self-assessment of health (very poor/poor/regular; good/very good).¹⁴14. Menezes MC, Costa BVL, Ferreira NL, Freitas PP, Mendonça RD, Lopes MS, et al. Percurso metodológico de ensaio comunitário controlado em serviço de saúde: pesquisa epidemiológica translacional em Nutrição. Demetra [Internet]. 2017 [citado 2020 maio 6];12(4):1203-22. Disponível em: https://doi.org/10.12957/demetra.2017.28451
   https://doi.org/10.12957/demetra.2017.28...

Height was measured a single time using a portable Alturexata® stadiometer (up to 220cm, 0.5cm precision); and body weight was measured using Marte® PP 180 digital weighing scales (up to 180kg, 100g precision). The scales were placed on a level surface and participants were weighed without their shoes and wearing light clothes.²⁰20. Ministério da Saúde (BR). Secretaria de Atenção à Saúde. Departamento de Atenção Básica. Orientações para a coleta e análise de dados antropométricos em serviços de saúde: norma técnica do sistema de vigilância alimentar e nutricional – Sisvan [Internet]. Brasília: Ministério da Saúde; 2011 [citado 2020 maio 6]. 76 p. Disponível em: https://bvsms.saude.gov.br/bvs/publicacoes/orientacoes_coleta_analise_dados_antropometricos.pdf
https://bvsms.saude.gov.br/bvs/publicaco...

During the telephone interview at 36 months after baseline, participants were asked about their attendance at PAS activities (“ Do you still attend the Health Fitness Center Program? ”) and self-reported body weight (“ What is your current weight? ”).

At the face-to-face validation stage the following data were collected: recent attempts to lose weight since the last study assessment (yes; no); and body weight measured using the same procedures as at baseline.

Multiple linear regression was used to estimate corrected body weight of women, based on a formula that considered the characteristics associated with error in self-reported weight. This included information on baseline nutritional status, age and self-reported weight during the telephone interview. The characteristics of each participant were multiplied by the coefficients obtained through multiple linear regression, added together in order to obtain the corrected measurement. Measured body weight – at the face-to-face data validation interview – and height – measured at database – were used to obtain the measured body mass index (BMI):

_measuredBMI = measured weight at validation (kg)/height measured at baseline (m)²

Classification as ‘overweight’ (yes; no) was done in a differentiated manner, according to the age of the participants: adults (20-59 years) with overweight= BMI≥25kg/m²; and elderly (60 years or over) with overweight = BMI≥27kg/m².¹⁸18. World Health Organization - WHO. Preventing chronic diseases: a vital investment [Internet]. Geneva: World Health Organization; 2005 [cited 2020 May 6]. Available from: https://www.who.int/chp/chronic_disease_report/en/
https://www.who.int/chp/chronic_disease_...

BMI was also calculated using self-reported weight (_{self-reported}BMI = self-reported weight [kg]/height measured at baseline [m]²) and corrected weight (_correctedBMI = corrected weight [kg]/ height measured at baseline [m]²), always using height measured at baseline.

The data were collected by Nutrition course undergraduates and health professionals, trained beforehand to use the instruments and conduct the interviews, under the guidance of a field supervisor and the principal investigator. Training courses were held every six months with the entire team, field manuals and data collection logistic were developed for different moments of the study.¹⁴14. Menezes MC, Costa BVL, Ferreira NL, Freitas PP, Mendonça RD, Lopes MS, et al. Percurso metodológico de ensaio comunitário controlado em serviço de saúde: pesquisa epidemiológica translacional em Nutrição. Demetra [Internet]. 2017 [citado 2020 maio 6];12(4):1203-22. Disponível em: https://doi.org/10.12957/demetra.2017.28451
https://doi.org/10.12957/demetra.2017.28...

During face-to-face data collection, data consistency was checked by the field supervisor who gave the questionnaire back to the interviewer in cases of inconsistency. During data collection by telephone, the supervisor reviewed the answers and accompanied the telephone calls.

The analyses were performed in two stages. In the first stage, we identified the sociodemographic and health characteristics of individuals with the biggest errors between self-reported weight and measured weight. Self-reported data was then corrected when necessary. All tests were performed using the Data Analysis and Statistical Software (Stata) version 14.0 taking a 5% (p<0.05) significance level.

The Kolmogorov-Smirnov normality test was used to verify distribution of the numerical variables and the data were presented in the form of averages and 95% confidence intervals (95%CI); except for the time interval between interviews, which was presented in the form of medians and interquartile ranges (p25-p75). The participants’ sociodemographic and health characteristics were presented according to frequency distribution.

When analyzing the data, individuals were removed who had very large differences between measured weight and self-reported weight, i.e. those above the third quartile or below the first quartile, based on the Interquartile Range [IQR] method.²¹21. Tukey JW. Exploratory data analysis [Internet]. London: Addison-Wesley; 1977 [cited 2020 May 6]. Available from: https://www.stat.berkeley.edu/~brill/Papers/EDASage.pdf
https://www.stat.berkeley.edu/~brill/Pap... This information was later corrected.

The following were calculated to validate the data: variation in self-reported weight and measured weight errors (error = self-reported weight – measured weight); and variation in erros analized according to participant characteristics.

Student’s paired t test was used to check for differences between self-reported weight and measured weight; and Student’s t test for single samples was used to check whether average weight error was equal to zero. The ANOVA analysis model was used to compare average weight error according to participant characteristics (age; schooling; occupation; marital status; self-assessment of health; previous attempts to lose weight; nutritional status).

Following this, multiple linear regression (MLR) was applied in order to identify characteristics associated with weight error. The tested variables were input to the model in blocks in the following order: sociodemographic characteristics (age; occupation; schooling; marital status); general characteristics (participation in the PAS; self-assessment of health); and characteristics related to weight (nutritional status; satisfaction with weight; previous attempts to lose weight). Variables associated with weight error (p<0.05) were input to a MLR model in order to estimate corrected weight. With the aim of checking the quality of the models, the variance inflation factor (VIF) test was performed to remove the possibility of multicollinearity between independent variables. Residue normality, homoscedasticity and independence analyses were also performed.

Concordance between the two information sources – self-reported weight and measured weight – was assessed using Lin’s concordance correlation coefficient (CCC).²²22. Lin LI. A concordance correlation coefficient to evaluate reproducibility. Biometrics [Internet]. 1989 Mar [cited 2020 May 6];45(1):255-68. Available from: https://www.jstor.org/stable/2532051?seq=1#metadata_info_tab_contents
https://www.jstor.org/stable/2532051?seq... The method proposed by Bland & Altman²³23. Bland JM, Altman DG. Comparing methods of measurement: why plotting difference against standard method is misleading. Lancet [Internet]. 1995 Oct [cited 2020 May 6];346(8982):1085-87. Available from: https://www.thelancet.com/pdfs/journals/lancet/PIIS0140-6736(95)91748-9.pdf
https://www.thelancet.com/pdfs/journals/... was used to identify errors and differentiation patterns between self-reported values and measured values. The Kappa concordance index was used to verify differences between classification of overweight for (i) self-reported versus measured body weight and (ii) corrected versus measured body weight.

The study was conducted in accordance with the norms for research involving human beings after having been submitted to and approved by the Federal University of Minas Gerais Research Ethics Committee (COEP/UFMG) – 0537.0.0203.000-11; 52683916.0.0000.5149 – and the Belo Horizonte City Hall Research Ethics Committee – 0537.0.0203.410-11A; 52683916.0.3001.5140. All participants were informed about the study and signed a Free and Informed Consent form.

Results

Validation data collection took place in an interval of no more than two months (2016-2017) following the telephone interview (average interval = 23.2 days [95%CI 22.4;24.1]). There was no relationship between weight error and time interval between obtaining self-reported and measured weight (p=0.94). This validation study found a predominance of female participants (85.7%), participants in the 30-59 year age group (54.2%), with 9-11 years of schooling (30.8%), men and women who were retired or pensioners (37.6%) and married or with common law partners (64.9%) ( Table 1 ).

The average difference between participants’ self-reported weight and measured weight was -0.7kg (95%CI -1.1;-0.4): men underestimated their weight by 0.6kg (95%CI -1.3;0.2), and women by 0.8kg (95%CI -1.1;-0.4) ( Table 2 ).

Concordance was substantial for the general population (CCC=0.97) and when stratified by sex (CCC=0.97 for women; CCC=0.98 for men). Distribution of the self-reported weight error among men was found to be close to zero (p=0.15), while for the general population and specifically for women, there was deviation in relation to zero: p<0.01 ( Table 2 ). As such, we investigated factors associated with weight error among women.

Weight concordance, according to the women’s characteristics, showed greater variations among those aged 20-29 and with unfavorable self-perception of health ( Table 3 ). Average self-reported weight error varied according to occupation, attempt to lose weight and nutritional status.

In the multivariate analysis of the women’s data, only age and nutritional status remained associated with weight error ( Table 4 ). The final model determination coefficient (R²) showed that 98.1% of the variation in women’s self-reported weight error was explained by these variables. In general, the older the women, the greater the weight error in comparison to women aged 20-29 ( Table 4 ).

Based on these values, we were able to calculate weight corrected according to nutritional status and age, using multiple linear regression:

Figure 1 shows the scatter plot prepared using the Bland & Altman method. Taking the concordance limits, corrected weight for women improved precision, increasing the CCC value from 0.97 (95%CI 0.96;0.97) to 0.99 (95%CI 0.98;0.99), with distribution of the new error close to zero (p=0.999).

Overweight prevalence among women, calculated as per the BMI obtained based on self-reported weight and measured weight, was 65.5% and 63.6%, respectively, showing 94.3% concordance (Kappa=0.878). After correction of self-reported weight, overweight prevalence was 63.9% and concordance with measured weight was 94.6% (Kappa=0.883).

Discussion

Thus study consisted of validating body weight reported by PAS users during a telephone interview, compared to measured weight in a face-to-face interview, together with analysis of factors associated with self-reported weight error. The results show the validity of self-reported weight for men. For women, self-reported weight was sufficient to assess nutritional status but statistical correction was needed in order to use the continuous measurement. The self-reported weight error identified among women was associated with nutritional status and age.

Studies show the validity of self-reported measurements in different populations.⁸8. Bes-Rastrollo M, Sabaté J, Jaceldo-Siegl K, Fraser GE. Validation of self-reported anthropometrics in the Adventist Health Study 2. BMC Public Health [Internet]. 2011 Apr [cited 2020 May 6];11:213. Available from: https://doi.org/10.1186/1471-2458-11-213
https://doi.org/10.1186/1471-2458-11-213... However, women have been found to underestimate self-reported weight.¹¹11. Pregnolato TS, Mesquita LM, Ferreira PG, Santos MM, Santos CC, Costa RF. Validade de medidas autorreferidas de massa e estatura e seu impacto na estimativa do estado nutricional pelo índice de massa corporal. Rev Bras Crescimento Desenvolv Hum [Internet]. 2009 abr [citado 2020 maio 6];19(1):35-41. Disponível em: http://pepsic.bvsalud.org/scielo.php?script=sci_arttext&pid=S0104-12822009000100004&lng=pt&nrm=iso
http://pepsic.bvsalud.org/scielo.php?scr... In this study, women underestimated their weight by approximately 0.8kg, this being less than that found in the United Kingdom (-2.6kg error) and in Salvador, BA (-0.9kg error), although it was higher than that found for university leavers in the state of Minas Gerais (-0.6kg error).²2. Wright FL, Green J, Reeves G, Beral V, Cairns BJ, Million Women Study collaborators. Validity over time of self-reported anthropometric variables during follow-up of a large cohort of UK women. BMC Med Res Methodol [Internet]. 2015 Oct [cited 2020 May 6];15:81. Available from: https://doi.org/10.1186/s12874-015-0075-1
https://doi.org/10.1186/s12874-015-0075-... Differences between self-reported and measured weight are frequently associated with sex, nutritional status, age range and schooling.²²22. Lin LI. A concordance correlation coefficient to evaluate reproducibility. Biometrics [Internet]. 1989 Mar [cited 2020 May 6];45(1):255-68. Available from: https://www.jstor.org/stable/2532051?seq=1#metadata_info_tab_contents
https://www.jstor.org/stable/2532051?seq... The direct relationship found between greater age and lesser weight validity among women may be related to older women weighing themselves less frequently and not accompanying changes in body weight over time. This difference may also reflect the socially and culturally valued slimness model.¹¹11. Pregnolato TS, Mesquita LM, Ferreira PG, Santos MM, Santos CC, Costa RF. Validade de medidas autorreferidas de massa e estatura e seu impacto na estimativa do estado nutricional pelo índice de massa corporal. Rev Bras Crescimento Desenvolv Hum [Internet]. 2009 abr [citado 2020 maio 6];19(1):35-41. Disponível em: http://pepsic.bvsalud.org/scielo.php?script=sci_arttext&pid=S0104-12822009000100004&lng=pt&nrm=iso
http://pepsic.bvsalud.org/scielo.php?scr...

Good concordance was found for nutritional status classification using self-reported measurements, both for men and women, thus corroborating the findings of other studies. An example of this is a population-based study conducted in a rural area in Northeast Brazil in 2011, notwithstanding it having found less accuracy among the elderly and those with less than four years of schooling. The study also showed that difficulty in accessing health services and places where weight could be measured also resulted in worse reporting.²⁸28. Martins PC, Carvalho MB, Machado CJ. Uso de medidas autorreferidas de altura, peso e índice de massa corporal em uma população rural do Nordeste do Brasil. Rev Bras Epidemiol [Internet]. 2015 mar [citado 2020 maio 6];18(1):137-48. Disponível em: https://doi.org/10.1590/1980-5497201500010011
https://doi.org/10.1590/1980-54972015000...

Validation of self-reported measurements in the 2013 National Health Survey found weight information error for all categories of sex, age and schooling assessed, although the error was lower among the elderly and people with no schooling or incomplete elementary education. The National Health Survey also showed that, despite the error found, self-reported weight can be used as a proxy for values measured in adults.⁵5. Moreira NF, Luz VG, Moreira CC, Pereira RA, Sichieri R, Ferreira MG, et al. Self-reported weight and height are valid measures to determine weight status: results from the Brazilian National Health Survey (PNS 2013). Cad Saúde Pública [Internet]. 2018 maio [citado 2020 maio 6];34(5):e00063917. Disponível em: https://doi.org/10.1590/0102-311x00063917
https://doi.org/10.1590/0102-311x0006391... The researchers found that sensitivity for determining overweight and obesity based on self-reported measurements is greater among men than among women.⁵5. Moreira NF, Luz VG, Moreira CC, Pereira RA, Sichieri R, Ferreira MG, et al. Self-reported weight and height are valid measures to determine weight status: results from the Brazilian National Health Survey (PNS 2013). Cad Saúde Pública [Internet]. 2018 maio [citado 2020 maio 6];34(5):e00063917. Disponível em: https://doi.org/10.1590/0102-311x00063917
https://doi.org/10.1590/0102-311x0006391...

A differential of this study was that it was conducted in a SUS service, namely the PAS, characterized by regular service user attendance and periodic physical assessments.¹⁵15. Lopes ACS, Ferreira AD, Mendonça RD, Dias MAS, Rodrigue RCLC, Santos LC. Estratégia de promoção à saúde: programa academia da cidade de Belo Horizonte. Rev Bras Ativ Fís Saúde [Internet]. 2016 jul [citado 2020 maio 6];21(4):379-84. Disponível em: https://doi.org/10.12820/rbafs.v.21n4p%25p
https://doi.org/10.12820/rbafs.v.21n4p%2... As PAS attendance is weekly (three times a week on average), we assumed that this implied attendees taking greater care of their health and checking their weight more frequently (at the PAS units studied, weight is measured every six months), so that self-reported measurements would be closer to real measurements. However, this hypothesis was not confirmed for women participants of our study.

With the aim of overcoming this problem, we opted to statistically correct self-reported weight, considering the characteristics that most impacted errors in women’s reports. Concordance between the corrected measurement and measured weight was high, indicating that using this form of correction can be useful for longitudinal weight assessment studies in populations with similar characteristics. Additionally, once weight had been validated, the formula obtained was used to correct the self-reported weight of participants excluded from the analysis, and the concordance correlation coefficient was checked – CCC = 0.853 – for corrected weight, revealing that the correction formula is indeed valid for the population studied (data not shown). The results show the importance of using corrected measurements to get predictions that are more valid, as well as reinforcing the importance of this study.

Another important aspect of this study was the inclusion of infrequent PAS users in the analyses. As it is a health service with continuous participation, although turnover is high,¹³13. Mendonça RD, Horta PM, Santos LC, Lopes ACS. The dietary profile of socially vulnerable participants in health promotion programs in a Brazilian metropolis. Rev Bras Epidemiol [Internet]. 2015 Apr-Jun [cited 2020 May 6];18(2):454-65. Available from: https://doi.org/10.1590/1980-5497201500020013
https://doi.org/10.1590/1980-54972015000... we opted to include them. The fact that differences were not found regardless of frequency of participation in this service may indicate that previous participation in PAS demonstrates continuous care of health. The authors’ hypothesis is that despite not attending the PAS activity routine continuously, previous participation could have made these service users more aware and contributed to promoting their health. This hypothesis suggests that the existence of the PAS is an important strategy for controlling chronic noncommunicable diseases and promoting healthy lifestyles.²⁹29. Sá GBAR, Dornelles GC, Cruz K G, Amorim RCA, Andrade SSCA, Oliveira TP, et al. O Programa Academia da Saúde como estratégia de promoção da saúde e modos de vida saudáveis: cenário nacional de implementação. Ciênc Saúde Coletiva [Internet]. 2016 jun [citado 2020 maio 6];21(6):1849-60. Disponível em: https://doi.org/10.1590/1413-81232015216.09562016
https://doi.org/10.1590/1413-81232015216...

This study also has limitations. The time that elapsed between data collection by telephone and the face-to-face weight validation interview may have had repercussions on real variation in weight. However, if this variation did occur it was of low magnitude, given that the interval of time between the measurements was small and that furthermore no correlation was found between the measurement interval and the weight error.

Validation of body weight measurement of SUS health promotion service users is unprecedented and allows it to be monitored by telephone. This simplifies the logistics of epidemiological studies and of the service itself, by favoring quicker and cheaper health surveillance actions with less need for human resources.

In conclusion, self-reported weight of Belo Horizonte PAS users was valid for men but needed to be corrected for women due to variations associated with nutritional status and age. Following this correction, women’s body weight was found to have a satisfactory level of validity. Obtaining valid self-reported weight measurements simplifies continuous monitoring of health service users, especially infrequent ones, favoring longitudinal epidemiological studies and health surveillance being conducted in a sustainable manner.

References

Publication Dates

History