Estimativas da cobertura de mamografi a segundo inquéritos de saúde no Brasil Estimates of mammography coverage according to health surveys in Brazil

MÉTODOS: A partir das coberturas por mamografi a em mulheres de 50 a 69 anos, com e sem telefone fi xo, observadas na Pesquisa Nacional por Amostra de Domicílios (PNAD) 2003, calcularam-se as razões entre elas e o respectivo coefi ciente de variação. A razão de cobertura foi multiplicada pela cobertura estimada pelo sistema de Vigilância de Fatores de Risco e Proteção para Doenças Crônicas por Inquérito Telefônico (VIGITEL), permitindo estimar a cobertura entre mulheres sem telefone em 2007. Essas estimativas foram aplicadas à população de mulheres, com e sem telefone, obtidas a partir da PNAD 2006, obtendo-se assim as estimativas fi nais para as capitais.


INTRODUCTION
The higher values obtained by VIGITEL were attributed to the fact that the survey had been conducted in households with a fi xed telephone line and, thus, in a higher-income population.In fact, data from the Instituto Brasileiro de Geografi a e Estatística (IBGE -Brazilian Institute of Geography and Statistics) show that, differently from other richer countries, Brazil shows a positive association between having a fi xed telephone line and family income.b In addition, studies on 2003 PNAD data indicated that having a mammogram performed was associated with family income and health plan coverage, important factors of access to health services. 3,6e difference among the population surveys refers to the geographic units of analysis.In the PNAD, estimates are only valid for the metropolitan areas of nine states and Federal District; in the remaining 17 states of the country, estimates are valid for the group of selfrepresentative cities.More recent estimates, obtained by VIGITEL, correspond to the Brazilian state capitals and Federal District.
There are also methodological differences among the surveys.In addition to the type of interview (home or by telephone), one individual per household is randomly selected for interview in the VIGITEL; in the PNAD, data on residents who were absent at the time of interview are provided by other individuals, usually other residents of household and, very rarely, non-residents.In the specifi c case of data on mammography reported by women aged between 50 and 69 years, responses were provided by the interviewees themselves in 75% of cases.
The present study aimed to estimate mammography coverage, based on two population surveys.PNAD data d,e are collected from a probability household sample, in the whole country, except for a rural area of Northern Brazil.Sampling involves one, two or three selection stages: cities, census tracts and households (private households and dwelling units in collective households).In the metropolitan areas (MA), sampling plan is stratifi ed into cities, and subsequently grouped into census tracts and households.Cities that do not belong to the metropolitan area are classifi ed as "selfrepresentative" and "not self-representative".The former are selected with probability 1 of belonging to the sample; not self-representative cities go through the stratification process and are selected, in each stratum, with a probability proportional to the resident population obtained from the Censo Demográfi co de 2000 (2000 Demographic Census).In the second stage, census tracts are selected with a probability proportional to the number of existing households (based on the 2000 Census), while, in the third stage, households are selected with equiprobability.To complement the selection process, household units are selected from the "new construction records".The PNAD sampling design enables representative estimates to be made for Brazil, Units of the Federation (UF) and nine metropolitan areas (Belém, Fortaleza ) and Federal District (FD).Weights (expansion factors) published with microdata are adjusted so that the expanded totals of the sample coincide with the totals of the projection for the resident population for 2003. 8GITEL data were obtained using a telephone interview survey conducted to monitor the frequency and distribution of risk and protective factors for noncommunicable chronic diseases, in the population aged ≥18 years, in the Brazilian state capitals and Federal District.The system establishes a minimum sample of 2,000 individuals per city.The sample is obtained from a systematic random selection of 5,000 fi xed telephone lines per city.For each eligible line, after obtaining consent from residents to participate in the interview, those aged 18 years or older are numbered and one of them is randomly selected to be interviewed.Refusals to participate in the monitoring system corresponded to 4.8% of the eligible lines.Telephone interviews were conducted by the VIGITEL system between July and December 2007.

METHODS
The fi nal weight (expansion factor) attributed to those interviewed by VIGITEL in each city resulted from the multiplication of three factors: the inverse of the number of telephone lines of the household; the opposite of the number of adults in the interviewee's household; and the post-stratifi cation weight.The latter was used to make the socio-demographic composition of the VIGITEL sample equal to that of the total adult population of the city, based on the 2000 Demographic Census, considering distribution by sex, age group and level of education.
For the estimates of the adult population of the 27 cities, the fi nal weight was multiplied by a fourth weighting factor, which considered the differences between the cities' population contingent and the similar number of individuals (about 2,000) studied by VIGITEL in each city.a Mammography coverage was calculated using the proportion of women aged between 50 and 69 years who had had a mammogram performed in the three years preceding the interview.
Mammography coverage levels in women, living in households with or without a fi xed telephone line in the metropolitan areas or self-representative cities (for the 18 states that did not have a metropolitan area), were calculated for 2003.For 2007, coverage levels in the state capitals were calculated with VIGITEL data.
In the two surveys, mammography coverage was calculated using the complex samples module of SPSS 15.0, considering the sampling design, expansion weights for the population and 95% confi dence intervals.
Ratios among coverage levels and respective coefficients of variation were calculated, based on mammography coverage levels in women aged between 50 and 69 years, in the three years preceding the interview, with or without a fi xed telephone line, observed in the 2003 PNAD.Coverage ratio was multiplied by the coverage estimated by VIGITEL, enabling the estimation of coverage in women without a fi xed telephone line in 2007.These estimates were applied to the population of women, with and without a telephone, obtained from the 2006 PNAD, thus obtaining the fi nal estimate for the state capitals and Federal District.Confi dence intervals of these last estimates were calculated based on the upper and lower 95% confi dence interval limits of mammography coverage ratios in women with or without a fi xed telephone line in 2003.

RESULTS
According to VIGITEL data, mammography coverage in the three years preceding the interview was 76.1%.The lowest coverage level was observed in Macapá (AP) (57.6%) and the highest one, in Vitória (ES) (88.7%).Coverage was lower in the Northern region capitals, where the highest value was observed in Manaus (AM), 70.8%.In the Northeastern region, coverage varied from 64.3%, in Fortaleza, to 84.4%, in Salvador.Coverage was above 80% in Vitória, Belo Horizonte, Florianópolis, Porto Alegre and Goiânia (GO).In São Paulo, Cuiabá (MT), Campo Grande (MS) and Curitiba, coverage was about 80%.The lowest coverage levels were found in Rio de Janeiro and Brasília, 69.7% and 73.5%, respectively (Table 1 in João Pessoa.Among the metropolitan areas studied in the PNAD, the differences in fi nal estimates were greater in Fortaleza, Belém and Recife.The greatest differences were found in João Pessoa, Maceió and Aracajú (SE), where coverage ratios calculated with PNAD data show high coeffi cients of variation (Table 2).
A 10.4% difference is observed among estimates, when comparing adjusted estimates with those from the 2003 PNAD, for the population living in metropolitan areas or self-representative cities.

DISCUSSION
According to the methodology used, mammography coverage in women aged between 50 and 69 years, in 2007, would be about 70%, whereas this coverage was reported by less than 50% of women in this age group, in 2003, according to the 2003 PNAD b and the WHS. 9 The variation in mammography coverage among cities suggests a multiple order of determination, once richer cities or those where the proportion of population with a private health plan is greater did not necessarily show higher coverage levels, as would be expected.the post-stratifi cation adjustment, as used in VIGITEL.However, in the case of mammography, the literature shows that this adjustment is insuffi cient to correct selection bias.a In the present study, the methodology to estimate mammography coverage in the cities considered the differences in reporting mammography between the populations with and without a fi xed telephone line, as well as the local fi xed telephone line coverage.For this reason, the greatest differences between the estimates calculated and those from VIGITEL were found in cities where either the mammography coverage ratio was low (João Pessoa), the fi xed telephone line coverage was low (Maceió), or both were low (Fortaleza, Aracajú, Natal [RN] and Porto Velho).
As PNAD data do not have representativeness for state capitals, the estimates calculated for coverage in women with or without a fi xed telephone line are based on populations living in the nine metropolitan areas and Federal District or in self-representative cities for the remaining states, where the capital and greater cities are included.Estimates made for the remaining states are more accurate, because, in these cases, a higher number of observations are involved.In addition, cities included in the metropolitan areas have characteristics similar to those of the capital and their geographical proximity can enable sharing of physical and human resources of the health system, refl ecting comparable mammography probabilities in relation to the capital.In contrast, self-representative cities can be located in any part of the state and constitute a less homogeneous stratum than metropolitan areas, resulting in less accurate estimates.
The fi rst PNAD data analyses were made with cities that were comparable to state capitals, given the size of their population.In three of them, it was not possible to distinguish which city was the capital, once two or three cities in each state included similar-sized populations.In the remaining states, however, differences in mammography coverage between women with and without a fi xed telephone line were small, except for the three state capitals in the Center-West region, where there were differences in mammography coverage rates from 11% to 22% between women with and without a fi xed telephone line, despite low coeffi cients of variation.
For the group of geographical areas considered, there was a difference of about 10% The impact of the increase in equipment availability for mammography use must be viewed with caution, once a medical request is necessary to perform a mammogram.
In the literature, it has been reported that requests for a mammogram are less frequent than women would like them to be. 4 An analysis of National Health Interview Survey data on women aged 40 years or older, who had not had a mammogram performed in the last two years in the United States, indicated that there was no medical recommendation to perform a mammogram in 80% of cases, despite the majority of these women having a health plan and regularly using some service. 5pulation surveys are an essential tool to monitor mammography coverage and factors associated with its performance. 1,7VIGITEL data are obtained and spread rapidly, constituting a subsidy to reorient the health care and promotion policy.However, their estimates tend to overestimate coverage, because they are based on the population of women who live in households with a fi xed telephone line.PNAD data enable coverage for the total population of women to be estimated, but the sampling design does not allow this to be more accurately achieved for the state capitals.In contrast, despite the use of self-representative cities as proxies of state capitals, the sample size is insuffi cient to work with specifi c population groups, such as women aged between 50 and 69 years.This occurs especially in areas with low demographic density, such as Northern Brazil.
The spread in use of mobile telephone lines in the last years indicates that data obtained from telephone surveys can become an ever more important source of information about morbidity, and use of and access to health services in urban areas.In 2008, PNAD and VIGITEL conducted data collections again.However, the limitations of these surveys will still require adjustments according to methods similar to that used in the present study.It is recommended that data collected with VIGITEL be adjusted according to national surveys, whose samples enable population-based information about the use of and access to health services in major Brazilian cities or at least in the state capitals to be obtained.
Data used in the present study come from the 2003 d and 2006 e PNAD and from the 2007 VIGITEL.c

Figure .
Figure.Correlation between fi xed telephone line coverage in metropolitan areas or self-representative cities and differences in mammography coverage estimates, according to VIGITEL and 2003 PNAD.

Table 1 .
Mammography coverage in women aged between 50 and 69 years in the three years preceding the population surveys.2007VIGITEL and 2003 PNAD.2003PNAD data for the population of women living in metropolitan areas or self-representative cities show coverage of 59.4% for mammography reported in the three years preceding the interview.Although coverage values were lower, some consistencies could be observed between VIGITEL and 2003 PNAD results: low coverage levels in the Northern region capitals; the highest coverage in the Salvador MA and the lowest in the Fortaleza MA, compared to other Northeast region capitals; and the lowest coverage in the Rio de Janeiro MA, compared to the state capitals of the Southeastern, Southern, and Center-West regions.On the other hand, in contrast to VIGITEL data, in the Northeast region, coverage in the state of Paraíba and in the city of Maceió (AL) was almost as low as that of the city of Fortaleza; coverage in Vitória (ES) was not the highest among all the locations studied; coverage levels in the cities of the Southern region were similar to each other; coverage in the Federal District was higher than in other cities of the Center-West region.The highest coverage levels were found in Salvador, Belo Horizonte, São Paulo and the ).a Ministério da Saúde.Secretaria de Vigilância em Saúde.VIGITEL Brasil 2007: Vigilância de fatores de risco e proteção para doenças crônicas por inquérito telefônico.estimativas sobre freqüência e distribuição sócio-demográfi ca de fatores de risco e proteção para doenças crônicas nas capitais dos 26 estados brasileiros e no Distrito Federal em 2007.Brasília; 2008.

Table 2 .
Estimate of mammography coverage in women aged between 50 and 69 years, according to 2007 VIGITEL and 2003 PNAD population surveys.
between the mammography coverage estimates calculated for the state capitals in the present study and those obtained from the 2003 PNAD.The concentration of resources in the state capitals may partly explain this difference, once coverage estimates in 2003 were made for metropolitan areas or self-representative cities, rather than state capitals.On the other hand, IBGE data indicate that there was an increase in mammography availability between 2002 and 2005, in the health sector as a whole and in the Sistema Único de Saúde (SUS -Unifi ed Health System), which could have enabled greater service use by women.According to the Sistema de