Self-rated health among adults in Southern Brazil

MÉTODOS: Estudo transversal, de base populacional, com amostra de 2.051 adultos de 20 a 59 anos de Lages, SC, em 2007. Foram aplicados questionários domiciliares para obter dados sobre auto-avaliação da saúde, condições socioeconômicas e demográfi cas, tabagismo, de estilo de vida e morbidades auto-referidas. Foram aferidos pressão arterial, peso, altura e circunferência abdominal. A análise multivariável foi realizada por regressão de Poisson, ajustada pelo efeito do delineamento amostral e estratifi cada por sexo.


INTRODUCTION
Self-rated health has been used in population health surveys because it can be easily applied and it has a high level of validity and reliability.Such assessment is a marker of inequalities among population subgroups, it shows high levels of predictive values of morbimortality and it enables international comparisons to be made. 11,21Individuals with a negative self-rated health had a relative risk of death almost two times higher than those who considered their health as excellent. 7lf-rated health is a subjective measure that combines physical and emotional aspects and one's level of satisfaction with life.Individual perception of health is an important indicator per se, once individual levels of well-being can infl uence quality of life. 19opulation-based studies on self-rated health are important to know health conditions and to monitor them throughout time.They allow the assessment of effectiveness of health policies, actions and services.Selfrated health can be incorporated into the health surveillance system due to its relatively easy feasibility.
Studies on self-rated health are recommended by the World Health Organization (WHO) 3 and some population-based studies performed in Brazil have followed the WHO recommendations, such as the 2003 World Health Survey, 20 the health module of the 1998 and 2003 National Household Sample Survey (PNAD) 1,5,6 and the state of São Paulo Health and Life Condition Survey.a In view of the importance of such population-based research, the present study aimed to analyze factors associated with self-rated health in adults.

METHODS
A cross-sectional study was performed in the city of Lages, in the state of Santa Catarina, Southern Brazil, between May and October 2007.In 2005, the population of this city totaled 166,733 inhabitants, 97.4% of whom lived in its urban area.b In 2000, Lages had a municipal human development index (HDI-M) of 0.813, coming in 74 th place among the 293 cities of the state of Santa Catarina and 315 th place among the 5,564 Brazilian cities. c The representative sample was comprised of 2,051 adults aged between 20 and 59 years (in complete years), living in an urban area (52% of the city's total population).b The present study is part of a broader health survey with several outcomes investigated, where a specifi c sample size calculation, suffi cient to estimate the prevalence of self-rated health, was made for each outcome.A 95% confi dence level, prevalence of fair and poor self-rated health of 25%, d sampling error of 3.5% and a design effect equal to two were adopted.A total of 10% was added to the sample size to compensate for losses and refusals and 20% to control possible confounding factors in the multivariable analysis, totaling 1,531 individuals.A 95% confi dence level, power of 80%, expected outcome (negative self-rated health) in the non-exposed group of 20%, ratio between the number of exposed and non-exposed individuals of 20:80 and minimum relative risk to be detected of 1.6 were adopted.The required minimum sample size to analyze these associations was 722 individuals.Analyses were stratifi ed by sex.The EpiInfo software, version 6.04, was used to calculate sample size.Sample was selected using a two-stage cluster sampling scheme.First, 60 of the 186 census tracts of Lages were randomly selected using a random sampling scheme without replacement.Next, a block was randomly selected and, on it, a corner was thus selected as the starting point of households to be visited, with the fi eld work beginning clockwise.A total of 34 individuals and 17 households were included per census tract.
Exclusion criteria included institutionalized individuals (those staying in prisons, hospices, or hospitals), amputees, bedridden individuals, those with casts, those who could not remain in the adequate position for the required anthropometric measurements and those who were considered handicapped, thus unable to answer the questionnaire.Losses corresponded to residents in randomly selected households that were visited at least four times by the research team, including a minimum of one visit on weekends and another at night time, when the examiner/interviewer could not locate the selected person or when they refused to participate.Household visits were conducted by ten pairs of interviewers, all supervised.Field interviewers were trained and remained unaware of the study objectives.Data were collected using face-to-face interviews and included the taking of participants' blood pressure in two moments (at the beginning and end of the application of questionnaire) and anthropometric measurements.Training and a pre-test of the questionnaire with 30 adults aged between 20 and 59 years, living in an area covered by a local health unit, were conducted.The pilot study was performed with 90 individuals in one of the randomly selected census tracts.This research project was widely promoted by the local media (most popular radio station, television and local newspaper), aiming to increase adherence.Data collection quality control was performed in 10% of the sample, using telephone interviews conducted by one of the supervisors.
The outcome of the study was self-rated general health, obtained through the following question: "Compared to people of your age, how would you rate your health in general?"."Excellent", "very good" and "good" responses were grouped in the positive self-rated health category, whereas "fair" and "poor" responses were grouped in the negative self-rated health category.Independent variables included socio-demographic variables, health-related behavior and habits, blood pressure and anthropometric measurements, and self-reported health morbidities.Socio-demographic variables were: age (20 to 29, 30 to 39, 40 to 49 and 50 to 59 years); level of education (≤ 4, 5 to 8, 9 to 11, ≥12 complete years); per capita family income in reais, calculated as the sum of the previous month income, divided by the number of residents in the house (one minimum wage was equivalent to R$ 380.00 or approximately U$ 211.00 at the time of this study), and categorized according to quartiles (0.026-0.500; 0.510-0.880;0.890-1.580;1.590-19.740);self-reported ethnicity (black, mixed, white, Asian, indigenous); and marital status (with a partner and without a partner).
Health-related behavior and habits included smoking (non-smoker, former smoker, and current smoker; 22 level of physical activity (suffi cient, ≥150 minutes/ week, and insuffi cient, <150 minutes/week), measured by the short version in Portuguese of the International Physical Activity Questionnaire (IPAQ); 10 and problems with alcohol, surveyed with the Cut Down, Annoyed, Guilty, Eye-opener (CAGE) Questionnaire, previously validated in Brasil. 15Individuals who showed CAGE values ≥ 1 were those considered to have problems associated with alcohol use.Blood pressure levels were measured at the beginning and end of questionnaire application (at least ten minutes) and the second measurement was considered.Measurements were carried out with the interviewee in a sitting position, with feet on the fl oor, left arm resting on a table at the level of the heart and palm facing upward.Electronic equipment with a digital reading system (Techline®), which had been adequately calibrated, was used to measure blood pressure levels.Systolic and diastolic pressure levels were categorized into normal and high.Individuals with a systolic pressure >140 mmHg (PAS>140 mmHg) and/or diastolic pressure >90 mmHg (PAD>90 mmHg), or hypertensive individuals who had been regularly using anti-hypertensive drugs were considered to have high blood pressure levels. 4Body weight was measured with individuals wearing light clothes, barefoot, in an upright position, with the feet together and arms extended alongside the body, with the palms facing the legs.Weight was measured using portable digital scales (Tanita®) with a 0.1 kg accuracy and 150 kg capacity; participants were weighed only once.Height was taken by fi xing an inelastic measuring tape on a vertical surface without a skirting board, at a height of 100 cm from the fl oor, with the use of adhesive tape.Individuals were measured without their shoes, wearing nothing on their heads, standing with their heels together and these, gluteal region, shoulders and head touching the vertical surface of the wall, and looking straight ahead while breathing in.Weight and height were used to calculate the body mass index (BMI).Individuals were categorized as follows: eutrophic (BMI < 25 kg/m 2 ), overweight (25.0 kg/m 2 ≤ BMI ≤ 29.9 kg/m 2 ) or obese (BMI ≥ 30 kg/m 2 ). 22Measurement of the natural line with the smallest waist circumference was taken with the use of the same measuring tape, with the individual standing and while breathing out.Cut-off values for normality were ≤ 102 cm in men and ≤ 88 cm in women, considering individuals with values higher than these to have abdominal obesity. 18rticipants were asked about a medical diagnosis of diabetes mellitus, self-reported wheezing in the last 12 months, breathlessness while walking fast on a straight surface or gentle slope, and diagnosis of chronic bronchitis. 17e occurrence of episodes of toothache in the six months prior to interview and the number of natural teeth present in each jaw (ten natural teeth or more; fewer than ten natural teeth; no natural teeth) were also assessed.
A variable that combines all self-reported general health and oral conditions was created and categorized as follows: none, one, two, or three or more self-reported morbidities.
Data were entered in duplicate in the Epi-Info software, version 6.04, by previously trained keyboarders.After verifying the reliability of data, statistical analyses were carried out using the STATA statistical package, version 9.0.All analyses were adjusted for the sample design effect and weighted by frequency of sex, and weights were determined by the ratio between the proportions of sexes in the city population, obtained from the Instituto Brasileiro de Geografi a e Estatística (IBGE -Brazilian Institute of Geography and Statistics) and the sample.Descriptive statistics of variables of the sample population were performed and chi-square and linear trend tests were used, when appropriate, to estimate the associations between the outcome and each independent variable.Multivariable analysis was performed using Poisson regression, with the estimation of prevalence ratios for negative self-rated health and respective 95% confi dence intervals.Analyses followed a theoretical model of determination, divided into three hierarchical groups of variables.The fi rst, more distal group was comprised of socioeconomic and demographic variables that, hypothetically, infl uenced group 2 variables -behavioral factors.These, in their turn, infl uenced group 3's clinical variables, which, fi nally, had an infl uence on the outcome of this study.Variables with p<0.20 in the bivariate analysis were selected for the multivariable analysis.First, the variables in group 1 were included in the analysis.Next, the variables in group 2, adjusted among themselves and for group 1 variables, were included in the multiple model.Finally, group 3 variables were included in the model and adjusted among themselves and for groups 1 and 2 variables.
In both sexes, there was a greater proportion of younger individuals, with higher income and level of education and whose ethnicity was white or Asian, who rated their health as positive than those who were poorer, less educated and either black or indigenous.Among men, the absence of a partner was associated with positive self-rated health, whereas ethnicity did not show statistical signifi cance.The highest proportion of positive self-rated health was observed in men with level of education equal to or higher than 12 years (91.1%), while the lowest proportion was among women with four or less years of study (44.7%) (Table 1).
More non-smoking and eutrophic individuals rated their health as positive.Among men, only smoking was associated with negative self-rated health.The proportion of positive self-rated health was higher in those who had never smoked, followed by former smokers and, last of all, current smokers.To be physically active did not achieve statistical signifi cance with positive self-rated health in men.Among women, in addition to smoking, nutritional status and waist circumference were associated with negative self-rated health.Women who were overweight and obese and those with abdominal obesity considered their health to be negative (Table 2).
The lowest proportions of individuals who rated their health as positive were found in adults with high blood pressure levels, those who reported having diabetes, wheezing, breathlessness, bronchitis and a lower number of natural teeth present.The higher the number of morbidities, the lower the proportion of individuals with positive self-rated health (Table 3).Table 4 shows the crude and adjusted prevalence ratios between negative self-rated health and variables in men.Among socioeconomic and demographic variables, the following were associated with negative self-rated health: per capita income, level of education and age.Poorer, less educated and older men had higher prevalences of negative self-rated health, when compared to richer, more educated and younger ones.Smoking and physical activity lost their association with each other when adjusted for socioeconomic and demographic variables.Among the morbidities analyzed, the following continued to be associated with negative self-rated health after adjustment: having high blood pressure levels and reporting wheezing, whereas breathlessness was on the borderline of association.The number of morbidities was strongly associated with negative self-rated health.Prevalence ratios varied between 2.3 and 5.0 in men with one morbidity and those with three or more morbidities, respectively.
Among women, negative self-rated health was strongly associated with per capita household income, level of education and age.Prevalence of negative self-rated health was higher in poorer, less educated and older women and those who had abdominal obesity.These associations remained after adjustment for more distal variables.Among associated morbidities in the crude analysis, only high blood pressure levels, diabetes, wheezing and breathlessness remained associated with the outcome after adjustment.Like men, the number of morbidities was strongly associated with negative self-rated health (Table 5).

DISCUSSION
A total of ¾ of the population studied rated their health as positive, with signifi cant differences according to sex, age, income and level of education.Older, poorer and less educated individuals and women considered their health to be worse.The higher the number of self-reported comorbidities, the greater the proportion of individuals with negative self-rated health.The magnitude of effect of morbidities in women was higher than in men.These results are in accordance with those from the World Health Survey performed in Brazil and the 1998 National Household Sample Survey. 1,5,20,21vels of education interfere with the perception of health, understanding of information about health promotion and prevention, adoption of healthy lifestyles, adherence to therapeutic treatments and use of health services.e By selecting a response, the participant indicates their general health status, considering the physical, social and psychological/mental dimensions.Studies suggest that the physical dimension seems to have more infl uence on self-rated health.The effect of self-rated health on the risk of death varies according to the level of education and income.Poor self-rated health is strongly associated with mortality in adults with high level of education and/or income. 8he results of this study agree with those of Dachs, 5 who found that 70% of individuals belonging to the fi rst decile of income had a positive self-rated health, whereas this value totaled 87% among those in the upper decile.Higher income is associated with the acquisition of assets and products, including medications, greater access to leisure activities and greater work autonomy, aspects related to health.Individuals with a higher level of education reported multidimensional aspects such as feelings of physical and psychological well-being, being physically active and not having diseases, when compared to those who were less educated.These tend to assess their health by associating it with physical and functional aspects. 12e prevalence of negative self-rated health increased signifi cantly with age, as exemplifi ed by what occurred with the results of the State of São Paulo Multicenter Health Survey (ISA-SP).f In the present study, younger individuals considered their health to be good.Among these, there was a different in perception according to sex; men rated their health in a more positive way.The proportion of positive self-rated health in women in the 30-to-39-year age group was similar to that of men in the 40-to-49-year age group.[16] The percentages of dissatisfaction with one's own health increased with age in women, thus corroborating the results of Brazilian 1,21 and international studies. 2,13,g Smoking, physical activity, diabetes, symptoms of bronchitis and number of natural teeth present lost statistical signifi cance when adjusted for groups of distal variables, suggesting that these factors act as mediators between social conditions and self-rated health.
The major sources of bias in cross-sectional studies are associated with selection and reverse causality or temporal bias.The hypothetical theoretical model adopted simulates a determination model, although it is not possible to guarantee that the chain of determination established actually occurred.This theoretical exercise is preferable to the mere entry of variables according to mostly statistical criteria.In the present study, a representative sample of the population aged between 20 and 59 years of the city and a response rate higher than 90% were obtained, in addition to the sample being homogeneously distributed among age groups.However, the proportion of women in the sample was Canada, the alternatives were the same adopted in the present study, 21 whereas the longitudinal study in electrical and gas workers in France (GAZEL study) used a scale with eight categories, ranging from very good to very poor. 9These methodological differences hinder comparisons among studies.
Older, poorer and less educated individuals and women rate their health as fair or poor.The higher the number of self-reported morbidities, the greater the proportion of individuals with negative self-rated health; among women, the effect of morbidities is greater.It is recommended that self-rated health be integrated with the health surveillance system, once it can be obtained with relative simplicity when population surveys are performed.
This research project was approved by the Research Ethics Committee of the Universidade do Planalto Catarinense, under protocol 001/2007, in November 2006.Participants in this study signed an informed consent form.

Table 1 .
Positive self-rated health, according to demographic and socioeconomic characteristics in adults.City of Lages, Southern Brazil, 2007.

Table 2 .
Positive self-rated health, according to health-related habits, nutritional status and waist circumference in adults.City of Lages, Southern Brazil, 2007.
a Chi-square test

Table 3 .
Positive self-rated health, according to high blood pressure levels and self-reported morbidities in adults.City of Lages, Southern Brazil, 2007.

Table 4 .
Association between negative self-rated health and demographic, socioeconomic variables, health-related types of behavior and self-reported morbidities in adult men.City of Lages, Southern Brazil, 2007.Distal variables, adjusted among themselves; b Intermediate variables, adjusted among themselves and for group 1 variables; c Proximal variables, adjusted among themselves and for groups 1 and 2 variables; d Proximal variable, adjusted for groups 1 and 2 variables. a

Table 5 .
21sociation between negative self-rated health and demographic and socioeconomic variables, health-related types of behavior and self-reported morbidities in adult women.City of Lages, Southern Brazil, 2007.Proximal variables, adjusted among themselves and for groups 1 and 2 variables; d Proximal variable, adjusted for groups 1 and 2 variables.higherthanthat of the population, thus leading to the decision of performing global analyses weighted by sex.All data collection instruments adopted in this study have been validated and used in Brazil and abroad.Likewise, measure instruments (weight, height and blood pressure) were calibrated.The data collection team was trained and standardized and this collection underwent quality control, which contributed to the internal validity of the study.The literature on the theme adopts different forms of outcome categorization.The Inquérito de Saúde do Estado de São Paulo (for individuals aged more than 60 years) used the Medical Outcome Study 36-item Short Form Health Survey (SF 36), validated in Brazil.According to Theme-Filha,21studies based on health interviews in Barcelona, Spain, and the 1998 PNAD used the following categories: very good, good, fair, poor and very poor, similarly to those adopted by the 2003 World Health Survey in Brazil (very good, good, average, poor and very poor).In the 1996 National Health Interview Survey and Current Population Survey, conducted in the United States, and in the Health and Social Survey, conducted in Quebec, a Distal variables, adjusted among themselves; b Intermediate variables, adjusted among themselves and for group 1 variables; c