ABSTRACT:

The Health, Wellbeing and Aging Study (SABE) began in 2000 under the coordination of the Pan American Health Organization as a multicenter study developed in seven urban centers of Latin America and the Caribbean, to determine the health and living conditions of elderly in this region. In Brazil, the study was developed in the city of São Paulo, where 2,143 individuals (cohort A) aged 60 years or older were selected through probabilistic sampling. In 2006, 1,115 of these individuals were interviewed a second time, and a new probabilistic cohort of individuals aged 60 to 64 years was added (cohort B; n=298). Thus the SABE Study - Brazil was transformed into a multi-cohort longitudinal study, with the objective of identifying changes that occur in the aging process among different generations. In 2010, a longitudinal follow-up was developed with cohorts A and B, with the addition of a new cohort of individuals aged 60 to 64 years (cohort C; n=355). Thethree surveys (2000, 2006 and 2010) involved the application of a questionnaire, anthropometric evaluation and functional tests, with the subsequent inclusion of blood collection for the evaluation of biochemical, immunological and genetic variables, as well as an accelerometer for the measuring of caloric expenditure.

Keywords:
Longitudinal studies; Demographic aging; Health of elderly; Epidemiology

INTRODUCTION: INSTITUTIONAL BASES OF THE SABE STUDY

The Health, Wellbeing and Aging Study (SABE) was initially devised by the Pan American Health Organization/World Health Organization (PAHO/WHO) as a multicenter survey to profile the living and health conditions of the elderly in seven urban centers in Latin America and the Caribbean. It was conducted in Bridgetown (Barbados), Buenos Aires (Argentina), Mexico City (Mexico), Havana (Cuba), Montevideo (Uruguay), Santiago (Chile) and São Paulo (Brazil), from October 1999 to December 2000. SABE is the second multi-center survey sponsored by PAHO, the first being the Encuesta de Necesidades de los ancianos (ENA), developed during the 1980s in 12 urban areas also in Latin America and the Caribbean.

The study design, sampling plan, instruments and field operations, as well as the collection and organization of the data, were carried out by the Center for Demography and Ecology of the University of Wisconsin-Madison. At various stages the SABE Study benefited from guidance provided by a body of external consultants who worked in coordination with PAHO/WHO and the Center for Demography and Ecology.

The team of researchers who conducted the SABE Study counted on components such as the PAHO team in Washington, D.C., headed by Dr. Martha Pelaez, at that time the aging coordinator of the institution; some researchers at the Center for Demography and Ecology at the University of Wisconsin-Madison, coordinated by Prof. Dr. Alberto Palloni, responsible for the study design and logistics organization; teams of consultants from both the United States as the involved countries; and, finally, country-based teams made up of senior researchers and their associates.

The primary objective of the SABE Study was to assess the health status of the elderly to project the needs that are likely to result in the rapid growth of this population. The secondary objective was to broaden the dialogue between public health research and the study of aging in order to strengthen interdisciplinary work¹1. Albala C, Lebrão ML, León Díaz EM, Ham-Chande R, Hennis AJ, Palloni A, et al. The Health, Well-Being, and Aging ("SABE") survey: methodology applied and profile of the study population. Rev Panam Salud Publica 2005; 17(5-6): 307-22..

METHODS

In Brazil, the SABE Study was developed in the city of São Paulo, which, although did not have the largest proportion of elderly people in the country, represented - and still represents - the highest absolute number of elderly people, and also the elderly population with greater diversity as a result of immigration and internal migration.

POPULATION

The study population of the first collection (cohort A) was composed of individuals aged 60 years or older residing in the urban area of the city of São Paulo in the year 2000. According to the Brazilian Institute of Geography and Statistics (IBGE), in 1996 the city had 836,223 elderly people, corresponding to 8.1% of the total population.

SAMPLE

In order to comply with the statistical analysis plan proposed by PAHO, the initial sample size - 1,500 elderly - was distributed according to the strata defined by gender and age group. The proportional sharing criterion led to the results presented in the fifth column of Table 1, according to Equations 1 and 2:

Where:

• n1 = initial value of the sample size in each age group;
• f = initial sample fraction; and
• POP = number of elderly in each h-th stratum.

Where:

• f  = initial sample fraction;
• n  = final value of the sample size in each age group; and
• N  = census population.

Due to the low population density, samples from age groups of 75 years and over were extended, by multiplying the numbers obtained at n1 by 2.5. Also, in order to compensate for excess mortality in relation to the female population, the male samples were adjusted to the same amount as the female sample. The last two columns of Table 1 present the expected minimum number of interviews (n2) and the respective final sampling fractions (f2) in each stratum.

The sample was reached by two procedures: draw of 1,500 elderly and free composition for the extended groups.

In order to calculate the minimum number of 5,882 households to be drawn and to obtain the 1,500 desired interviews, Equation 3 was used:

Where:

• d  = minimum number of households to be drawn.

The ratio 10/3 is the inverse of 3 elderly for every 10 residences, and 0.85 is the expected rate of success of the operation of locating and conducting the interviews in the permanent private households drawn.

In this case, the two-stage conglomerate sampling method was used under the probability proportional to size (PPS) criterion²2. Silva NN. Amostragem probabilística: um curso introdutório. 2a ed. São Paulo: EDUSP; 2001. 128 p.. The permanent register of 72 census tracts, available at the Department of Epidemiology of the School of Public Health, was considered the first stage sample. Figure 1 shows the distribution of census tracts by geographical areas of the city³3. Silva NN da, Cunha TN da, Quintanilha JA. Amostra mestra e geoprocessamento: tecnologias para inquéritos domiciliares. Rev Saude Publica 2003; 37(4): 494-502..

This sample was taken from the National Survey by Household Sampling (PNAD) 1995, composed of 263 census tracts drawn under the criterion of probability proportional to the number of households. The sampling fraction of the first stage was then calculated by Equation 4, resulting in Equation 5:

Where:

• f1  = fraction of second stage sampling;
• Di  = number of households in each sector; and
• D  = number of households in the city of São Paulo in 1991.

Where:

• f1  = fraction of second stage sampling;
• Di  = number of households in each sector; and
• D  = number of households in the city of São Paulo in 1991.

The minimum number of households drawn in the second stage, calculated by the mean (5.882/72=81.69), was approximated to 90. The sampling fraction of the second stage is equal to Equation 6, and the final fraction of sampling, expressed by Equation 7, corresponds to the probability that a household belongs to the sample drawn³3. Silva NN da, Cunha TN da, Quintanilha JA. Amostra mestra e geoprocessamento: tecnologias para inquéritos domiciliares. Rev Saude Publica 2003; 37(4): 494-502.. The updating of the addresses was carried out by the PNAD 1998 lists.

Where:

• f₂ = fraction of second stage sampling; and
• D_i = number of households in each sector.

Where:

• f  = final fraction of sampling;
• f₁ = sampling fraction of the first stage; and
• f₂ = fraction of the second stage.

The total number of addresses of each census tract was divided into segments of ten households and, in each sector, nine segments were drawn. Ninety households per sector were visited. All individuals considered eligible according to the research objectives were identified and invited to participate in the interviews. At the end of the first phase of the field work, 1,568 interviews had been conducted. Chart 1 shows the composition of thesample according to the draw stages and the results that remained within the limits set in the planning phase.

WEIGHTING

To compensate for unequal probabilities

The updating of the mailing lists introduced changes in the drawability in the second stage (f₂), resulting in different sampling fractions (fj) among the 72 census tracts. This problem was compensated by the use of weights calculated by the inverse of fj, obtained by Equation 8 ⁴4. Kish L. Selected papers. New York: John Wiley and Sons; 2003.

Where:

• p_j = sample weight relative to stratum j;
• D_j = number of households registered in the 1991 census;
• d = number of households to be visited in stratum j; and
• D_j ’ = number of households registered in the household counting operation performed at the time of the sample draw, in each j-th census tract.

To compensate for lack of response

To adjust the distribution of the sample to the composition of the population according to age and gender, weights (p_h) were calculated for each stratum, using equation 9:

Where:

• p_h = weight calculated by stratification a posteriori in stratum h;
• N_h = number of elements in the h-th stratum recorded by the last demographic census; and
• p_h.j = sum of sample weights (p_j) and total of elements recorded in each stratum h

Final weight added to the element file

Figure 2 shows an excerpt of the file containing the final sample of elements (elderly), the fields referring to the sampling plane and the final weights obtained by their product (p_j .p_h). Every five records in the file have the same weight corresponding to the effect of the design (sector), but depending on the stratum to which it belongs, this factor is changed. Inthe last column are the values of the final weight (pf), which should always be considered in the calculation of the final estimates, and their respective confidence intervals.

EFFECTS OF DESIGN

The sample draw plan defines the census tract as the primary sampling unit ⁵5. Kish L. Survey Sampling. New York: John Wiley and Sons; 1965. 274 p.. Asthe results of Tables 2 and 3 show, there is variability between the number of interviews achieved in each sector and between the estimated results for some sociodemographic variables ⁶6. Skinner CJ, Holt D, Smith TMF. Analysis of complex survey. Chichester: John Wiley and Sons; 1989..

As indicated by the results, the values of the effects of the design (deff) for gender only are close to 1, since the composition of the elderly population within the census tract in this characteristic is natural and is not impacted by cultural or economic influences. However, for the other variables, all the results show deff greater than 1, indicating possible similarities in the internal distributions and differences between the census tracts. The extreme value for the economic variable is confirmed by several studies carried out on the effect of the design in national and international epidemiological surveys⁴4. Kish L. Selected papers. New York: John Wiley and Sons; 2003.^,55. Kish L. Survey Sampling. New York: John Wiley and Sons; 1965. 274 p..

Thus, the final sample resulted in 2,143 people aged 60 years or older. It was composed of two segments: the first one, resulting from draw, corresponded to the probability sample formed by 1,568 interviews; the other one, made up of 575 residents in the districts where the previous interviews took place, accounts for the increase made to compensate the mortality in the population of more than 75 years, and to complete the desired number of interviews in this age group.

THE CONTINUATION...

Five years after the first data collection, it was decided to revisit the people interviewed in 2000 - which only occurred in 2006 - thus materializing the original SABE Project proposal to be a longitudinal study in all participating countries. For various reasons, this continuation could not be carried out in all countries, in the manner foreseen.

In 2006 and 2010, the second and third collections of the cohort A were constituted. Initially a search was made in the databases of municipal and state deaths. Next, the elderly were located through the addresses already known because of the previous interview. In the case of changes, information was sought in the vicinity and in nearby commercial establishments, such as pharmacies, greengrocers and real estate offices.

Of the 2,143 elderly interviewed in 2000, 1,115 were located and re-interviewed in 2006 and 748 in 2010 (cohort A). The differences can be seen in Table 4.

NEW COHORTS: B AND C

It is believed that the characteristics of elderly cohorts are very different from each other. For this reason, it is important that the surveys follow the cohorts over time in order to capture the variations of responses to the patterns associated with those changes. Longitudinal studies that continue to add new cohorts at the base of the age scale and to be representative of the study population are the ones best able to respond to these questions⁷7. Nacional Research Council. Preparing for an Aging World: The case for Cross-National Research. Washington, D.C.: National Academies Press (US); 2001..

Thus, in parallel to the research with cohort A, new cohorts were introduced, composed of people aged 60 to 64 years, living in the urban area of the city of São Paulo in 2006 and 2010.

In 2004, the number of people aged 60 to 64 in the city was 310,694, corresponding to 2.9% of the population, according to preliminary estimates of population totals released by IBGE⁸8. Instituto Brasileiro de Geografia e Estatística (IBGE). Estimativas de população para 2008. 2008. [Internet]. Disponível em: http://www.ibge.gov.br/home/estatistica/populacao/estimativa2008/default.shtm (Acessado em 05 de junho de 2017).
http://www.ibge.gov.br/home/estatistica/... .

The sample for cohort B was composed of 400 individuals, which defines the sampling fraction 400/310,694=0.0012874. In order to calculate the sample size, the prevalence of hypertension in 2000 was equal to 50%⁹9. Lebrão ML, Laurenti R. Saúde, bem-estar e envelhecimento: o estudo SABE no Município de São Paulo. Rev Bras Epidemiol 2005 Jun; 8(2): 127-41.; the design effect equal to 1.5; and margin of error of 7%. Corrections to the expected rate of response (75%) and mortality equal to 2%⁴4. Kish L. Selected papers. New York: John Wiley and Sons; 2003.^,55. Kish L. Survey Sampling. New York: John Wiley and Sons; 1965. 274 p. were introduced.

For the household draw, the method of sampling by conglomerates in two stages, under the PPS criterion, was used. The units of first and second stages were, respectively, the census tract and the household. Forty sectors were drawn in the first stage, observing the criterion by which the number of primary sampling units (PSUs) should be ≥30, and the minimum number of interviews per sector equal to 400/40 = 10.

The minimum number of households drawn by sector was 118, calculated by Equation 10, whose terms correspond to 10 interviews per sector. The ratio 10/1 is the inverse of 1 elderly for every 10 households, and 0.85 is the expected rate of success of the location and interviewing operation in the permanent households that were selected.

Where:

• D  = number of households drawn by sector.

For all of the 40 sectors, 4,720 households were drawn.

At the end of the process, 375 people were identified within the established criterion (60 to 64 years), and 298 interviews were performed (Table 5).

For the determination of sample of the cohort C, the same process was used, considering the population of 413,563, aged 60 to 64 in 2010, according to census data. The result is presented in Table 6.

WEIGHTING

The applied methods of estimation and weighting corresponded to the use of simple, unbiased estimators and reason estimators, which considered, in all cases, the calculation of the expansion factors of the sample design employed - that is, the selection probabilities applied in each step and the no answers. The variables gender and age were also considered, given the differences in mortality between men and women and the over dimensioning of the population aged 75 years or over. At the end, a factor was obtained for each individual in the sample, summarizing all the elements mentioned in the sample design and the strata considered.

In addition to the longitudinal analysis and the monitoring of the population, the SABE study design allows cross-sectional analyzes in each new collection (every five years on average), being the representative sample of the population in the year of interest (2006 or 2010). For this purpose, weight adjustments were performed and are described below.

ANALYSIS OF ELDERLY POPULATION IN 2006 (A₀₆+B₀₆)

The 2006 total sample consisted of the cohort A elderly interviewed in 2000, located and re-interviewed in 2006, and the new cohort B.

The weights due to sample design, absence of response and unequal probabilities in selection refer to the sample of 2000. Thus, the elderly of the A₀₆ cohort represent an elderly population of São Paulo referring to 2000, who did not move, did not die and agreed to be interviewed in 2006. That made it a bit distant from the city’s senior population in 2006.

In order to compensate for unequal probabilities in the sectors, it would be very difficult to change the weights, since they refer to the households counted at the time of the sample draw in 2000, in each census tract. On the other hand, they could not simply be discarded, which is why they were retained.

To compensate for the lack of response, the losses between 2000 and 2006 (deaths, refusals, no location etc.) can be considered as lack of response. Based on the 2006 population estimates, Equation 11 was applied, in which the denominator weights refer to each of the 1,115 re-interviewed, calculated in each stratum:

Where:

• p_h = final value of the sample weighting;
• N_h = population in the stratum in question; and
• p_h,j = sample weights in stratum h, sector j.

In the age group of 60 to 64 years, which corresponds to the B₀₆ contingent, we did not substitute the weights as in the others. The final weights were only adjusted for the new population estimates.

Thus, the new weights allow the almost exact reproduction of the age structure of the city in 2006, leaving no doubt about the representativeness of the expanded sample.

ANALYSIS OF ELDERLY POPULATION IN 2010 (A₁₀+B₁₀+C₁₀)

In order to study the elderly population living in the city of São Paulo in 2010, the A, B and C cohorts can be added with due weight. A similar procedure to that described above was performed with the total sample of 2010, composed of the elderly interviewed in 2000, 2006 and 2010 from cohort A, and of the respondents in 2006 from cohort B, who were located and re-interviewed in 2010, and by the new cohort C.

The weights due to sample design, lack of response and unequal probabilities in selection refer to the 2000 and 2006 samples. Thus, the elderly of the A₁₀ and B₁₀ cohorts represent the elderly population of São Paulo in 2000 and 2006 that did not move, did not die and agreed to be interviewed in 2010. This makes the population a little bit distant from that of 2006.

In order to compensate for unequal probabilities in the sectors, it would be very difficult to change the weights referring to the households counted at the time of the draw of the 2000 and 2006 samples in each census tract, which is why the weights were maintained.

In order to compensate for the lack of response, losses between 2000, 2006 and2010 (deaths, refusals, non-location etc.) can be considered as no response. Withthe 2010population estimates, Equation 12 was applied, in which the weights of the denominator refer to each of the 989 respondents (cohorts A and B), calculated in each stratum:

Where:

• p_h = final value of the sample weighting;
• N_h = population in the stratum in question; and
• P_h,j = sample weights in stratum h, sector j.

The new weights allow the almost exact reproduction of the age structure of the city in 2010, leaving no doubt about the representativeness of the expanded sample.

QUESTIONNAIRES

The original 2000 questionnaire was developed at PAHO with the assistance of experts from various universities and researchers from other similar studies conducted in different parts of the world. The instrument consisted of 11 sections that covered the main aspects related to aging studies: personal information, cognitive assessment, health status, functional status, medications, use and access to health services, social and family support network, income, housing conditions, anthropometry and functional tests (balance, flexibility and mobility). The questionnaires applied in 2006 and 2010 maintained the central axis of the instrument used in 2000¹⁰10. Lebrão ML. SABE - Saúde, bem-estar e envelhecimento. O Projeto Sabe no município de São Paulo: uma abordagem inicial. Brasília: Organização Pan-Americana da Saúde; 2003., and its sections were revised and updated. Also it received the addition of instruments such as: functional independence measure (FIM)¹¹11. Riberto M, Miyazaki MH, Jucá SSH, Sakamoto H, Potiguara P. Validação da Medida de Independência Funcional. Acta Fisiatr 2004; 11(2): 72-6.; sexuality, quality of life assessment through SF-12^® (Quality Metrics License nº R10-060608-36374)¹²12. Gandek B, Ware JE, Aaronson NK, Apolone G, Bjorner JB, Brazier JE, et al. Cross-validation of item selection and scoring for the SF-12 Health Survey in nine countries: results from the IQOLA Project. International Quality of Life Assessment. J Clin Epidemiol 1998 Nov; 51(11): 1171-8.; evaluation of family functionality (Adaptation, Partnership, Growth, Affection, Resolve - family APGAR)¹³13. Duarte YAO. Família: rede de suporte ou fator estressor: a ótica de idosos e cuidadores familiares. São Paulo: s.n.; 2001.; mistreatment; assessment of the overload of family carers’¹⁴14. Zarit SH, Zarit JZ. The memory and behavior problems checklist - 1987R and the burden interview. 1987.^,1515. Scazufca M. Brazilian version of the Burden Interview scale for the assessment of burden of care in carers of people with mental illnesses. Rev Bras Psiquiatr 2002 Mar; 24(1): 12-7.; evaluation of deaths by means of “verbal autopsy”¹⁶16 World Health Organization. Verbal autopsy standards: ascertaining and attributing causes of death. Geneva: WHO; 2007.; and evaluation of institutionalization. An examination of the oral cavity was carried out to verify the dental situation by means of the index of decayed, missing and filled teeth (DMFT) and possible soft tissue lesions.

In Annex 1 the structure of the data collection instrument used in the three periods of study is available, followed by the references related to it.

EVALUATION OF DEATHS

Deaths occurred between 2000, 2006 and 2010 were verified based on the information reported by relatives or neighbors, during a home visit, at the time of the subsequent data collection, in which case having a “verbal autopsy” performed¹⁶16 World Health Organization. Verbal autopsy standards: ascertaining and attributing causes of death. Geneva: WHO; 2007.. Subsequently, these data were compared to the mortality bases available in the Mortality Information Improvement Program (PROAIM), the State Data Analysis System (SEADE) and the Mortality Information System (SIM).

EVALUATION OF INSTITUTIONALIZATIONS

As in the evaluation of mortality, the occurrence of institutionalization between the years 2000, 2006 and 2010 was verified from the information reported by family members or neighbors, during a home visit, at the time of the subsequent data collection, having filled a specific instrument of characterization of the situation. These data allow the analysis of the predictive factors of institutionalization, little known in our country.

EVALUATION OF FRAILTY SYNDROME

Frailty is one of the most important syndromes that affect the elderly, and its study has been the target of many international research. Since 2006, this syndrome has been an important outcome of the SABE Study, given the scarcity of information about it at the time. Thelongitudinal design of the study allows the analysis of its main determinants, their impact on the quality of life of the elderly and their families, the care costs involved and the use of social and health services. Among the existing models, the SABE researchers opted for the one proposed by Fried et al. (2001)¹⁷17. Fried LP, Tangen CM, Walston J, Newman AB, Hirsch C, Gottdiener J, et al. Frailty in older adults: evidence for a phenotype. J Gerontol A Biol Sci Med Sci 2001 Mar; 56(3): M146-56.. Thus, it was decided to add in this article the description of the methodology used for the operationalization of the phenotype proposed in the adopted model, which uses the measurement of five components. The presence of one or two indicate the condition of pre-frailty, and the presence of three or more is considered frailty. Tobe classified as “non-frail” it is necessary that the elderly does not present any of the following components:

1. Unintentional weight loss: a subjective question was asked (“In the last year, did you lose three or more pounds of weight without a diet?”). If yes, it was punctuated in this component;

2. Referred fatigue: obtained through two questions from the Center for Epidemiological Studies - Depression (CES-D), validated for Brazilian elderly by Batistoni, Neri and Cupertino (2007)¹⁸18. Batistoni SST, Néri AL, Cupertino AP. Validade e confiabilidade da versão Brasileira da Center for Epidemiological Scale - Depression (CES-D) em idosos Brasileiros. Psico-USF 2010 Apr; 15(1): 13-22.;

• “How often, in the last week, did you feel that everything you did required a lot of effort?”;

• “How often, in the last week, did you feel that you could not carry on with your things?”;

The answers to both questions were: 0= rarely or no time (<1 day); 1= some part of the time (1 to 2 days); 2= a moderate part of the time (3 to 4 days); or 3= the whole time. Elderly who answered 2 or 3 on at least one of the two questions positively scored for the component;

3. reduction in force: the manual grip force was measured by means of the dynamometer. The elderly in the lowest quintile of distribution, stratified by gender and quartile of body mass index, were scored for this component;

4. low walking speed: obtained by the three-meter walking test, part of the Short Physical Performance Battery (SPPB)¹⁹19. Pfeffer RI, Kurosaki TT, Harrah CH, Chance JM, Filos S. Measurement of functional activities in older adults in the community. J Gerontol 1982 May; 37(3): 323-9.. The elderly who were in the highest quintile of the distribution, stratified by gender and the median value of height, were scored for this component;

5. low level of physical activity: identified by the short (translated) version of the International Physical Activity Questionnaire (IPAQ)²⁰20. Guedes DP, Lopes CC, Guedes JERP. Reprodutibilidade e validade do Questionário Internacional de Atividade Física em adolescentes. Rev Bras Med do Esporte 2005 Apr; 11(2): 151-8.^,2121. Benedetti TRB, Mazo GZ, Barros MVG de. Application of the International Physical Activity Questionnaire (IPAQ) for evaluation of elderly women: concurrent validity and test-retest reproducibility. Rev Bras Ciência e Mov 2004; 12(1): 25-34.^,2222. Benedetti TRB, Antunes PC, Rodriguez-Añez CR, Mazo GZ, Petroski ÉL. Reprodutibilidade e validade do Questionário Internacional de Atividade Física (IPAQ) em homens idosos. Rev Bras Med do Esporte 2007 Feb; 13(1): 11-6., which evaluates three specific types of activities: walking, of moderate and vigorous intensity; and leisure, domestic and/or work, in which an estimate of the weekly caloric expenditure (kcal) is obtained. The elderly in the lowest quintile of caloric expenditure, stratified by gender, were scored for this component.

The way force, velocity and physical activity were assessed can be observed in Chart 2.

The SABE Study, like any longitudinal study, had changes along its route, either because in the first phase we were in line with the international coordination of the project and other partners, by the appearance of new instruments or, also, because new partners saw in it an opportunity to develop their research. Thus, the evolution of the study in the three collections so far can be followed in Chart 3.

Figure 3 shows the complete study design. It is worth noting that it allows, in addition to the longitudinal analyzes obtained by the follow-up of the cohorts of interest, to identify the changes that have been occurring in the aging process from the comparison of cohorts A, B and C from 60 to 64 years, since these represent different generations (those born before World War II, those born during the war and those born after it).

All collections were funded by the Fundação de Amparo à Pesquisa de São Paulo (FAPESP) and, at different times, we received assistance from the Ministry of Health, to whom we thank.

References

Annex 1

Publication Dates

History