Study design and subjects
This consisted of a cross-sectional survey covering a sample of 4008 subjects representative of the non-institutionalized Spanish population aged 60 years and over. Before conducting the interview, informed consent was obtained in all cases from subjects or cohabiting next-of-kin. The study was approved by the Clinical Research Ethics Committee of the "La Paz" University Hospital in Madrid, Spain.
Study subjects were selected through probabilistic multistage cluster sampling, stratified by region of residence and size of town. Census sections were selected at random in each cluster, followed by individual households where information was obtained from residents. Data were collected on a total of 450 census sections in Spain, with subjects being selected in two sex and three age (60–69 y, 70–79 y, 80 y and over) strata. Individuals aged 80 years and over were over-sampled to assure enough number of subjects for a meaningful analysis. Subjects were replaced for interviews only after 10 failed visits by the interviewer or because of subject's incapacity, death, institutionalization or refusal to participate. There was an overall study response rate of 71%. Reasons for non-response were 'impossible to locate after several attempts' (17%), 'refused to be interviewed' (6%) and the rest of motives (6%). Given the study's sample design, subjects were assigned a weighting coefficient according to their sex, age, region and size of town of residence, which allowed for reconstructing the characteristics of the Spanish population in the analysis.
Data were collected from October 2000 to February 2001 by home-based personal interview using a structured questionnaire, followed by a physical examination to measure blood pressure and anthropometric variables. Field work was undertaken by interviewers who underwent standardized training. Of the 4008 subjects interviewed, 3030 (75.58%) furnished complete information on all variables of interest for this study. The variable accounting for the largest amount of incomplete data was cognitive deficit, which could not be calculated for 414 subjects. Compared to persons who furnished complete information on the study variables, subjects who did not were more frequently women (64% versus 53%), and older (mean of 73.2 versus 71.3 years).
To obtain information on health services use individuals were asked whether they had sought medical advice in the preceding month or had received a home medical visit in the last year. They were also asked about hospital admission in the previous year, influenza vaccination in the most recent immunization campaign, and the number of medications currently being taken. Lastly, a variable of overall utilization was created which encompassed the above five variables. Individuals were deemed to have made overall use of health-care services if they used at least one of the above services.
The variables studied as possible contributing factors to gender differences in the utilization of health-care services were classified into three groups, in accordance with Andersen's classic model, i.e., predisposing, enabling and need factors . Predisposing factors are defined, not as direct causes of utilization, but rather as determinants of the propensity to use such services; need can be objectively established by the medical practitioner and/or perceived by the patient; and enabling factors are conditions that enhance availability and access to services . The following were studied: among the predisposing factors, age and head-of-family status; among the need factors, lifestyles variables (e.g., tobacco and alcohol consumption, sedentariness, and hypertension) were studied as indicators of objective health-needs amenable through clinical preventive services, either professional advice and counseling or drug treatment. These services could serve both for primary or secondary prevention of chronic disease (e.g. ischemic heart disease). Also, as need factors we included chronic diseases, functional status and cognitive deficit as indicators of curative and rehabilitation services, and health-related quality of life (HRQL) as an indicator of the perceived need which may contribute to seeking a specific health service; and lastly, among the enabling factors, educational level, marital status, head-of-family employment status, and social network.
Lifestyle variables on which information was obtained were tobacco use, physical activity and alcohol consumption. Moderate drinkers were defined as men who consumed ≤30 g and women who consumed ≤20 g of alcohol daily. Heavy drinkers were those who exceeded the limits of moderate alcohol consumption in each gender. In addition, weight and height were measured using standardized procedures . Body mass index (BMI) was calculated as weight in kilograms divided by the square of the height in meters (kg/m2), and subjects were classified in three groups: low and normal weight (<25 kg/m2), overweight (25–29.9 kg/m2) and obese (≥30 kg/m2). Blood pressure was determined in a standardized manner , and subjects were deemed to be hypertensive where systolic blood pressure was ≥140 mm Hg, diastolic blood pressure was ≥90 mm Hg, or on current antihypertensive drug treatment.
Moreover, information was obtained on self-reported chronic diseases diagnosed by the physician, and specifically: asthma and chronic bronchitis, ischemic heart disease, stroke, arthritis, cataracts without treatment, diabetes mellitus, Parkinson's disease, cancer (at any site), and depression with need for treatment. These diseases are rather prevalent in people aged 60 years and older, and are important enough to be sure that people can be aware of their diagnosis by a physician. Subjects were classified into four groups: absence of disease, presence of one, two, and three or more chronic diseases.
Functional status was assessed through limitations in instrumental activities of daily living (IADL), as determined by Lawton and Brody's test . This assesses limitations in the following 8 activities: using the telephone, going shopping, preparing meals, doing household chores, washing clothes, traveling independently, responsibility for own medication, and handling money. The absence of limitation in any given activity scores 1 point. Subjects deemed to be IADL-independent are women scoring 8 and men scoring 5 points, since the scores for "preparing meals", "doing household chores" and "washing clothes" are excluded in the case of men. A lower score means that the subject manifests some type of dependence, with 0 points indicating the maximum degree of dependence.
Cognitive deterioration was evaluated with the Cognition Mini-exam [26, 27], which is a Spanish adaptation and validation of the original Mini-Mental Status Examination . It is made up of 11 items that evaluate the following cognitive areas: orientation, memory registration, short-term recall, attention, calculation, and language comprehension and expression. The maximum score attainable is 30 and individuals who obtain a score = 22 are deemed to suffer from a cognitive deficit.
HRQL provides subjective information on the health-status of the individuals, which has been shown to correlate with morbidity. As a result, HRQL served as an indicator of perceived need of health-care services . HRQL was measured using the Spanish version of the SF-36 questionnaire [30, 31]. This questionnaire is made up of 36 items, which assess the following 8 HRQL components or scales: physical functioning, role-physical, bodily pain, general health, vitality, social functioning, role-emotional, and mental health. Subjects' answers to any given item receive a numerical score which, after coding, is ranked on a scale of 0 to 100, so that the higher the score the better the HRQL. After aggregating the scores for each scale, by assigning a coefficient to each of them, summary physical and mental HRQL indices were calculated.
Finally, to assess social network, subjects were asked with whom they usually lived, how often they saw family members other than those with whom they cohabited, and how often they saw friends or neighbors.
Relative differences between women and men in the utilization of health-care services were summarized using odds ratios (OR) and their 95% confidence limits (CL) obtained from unconditional logistic regression, in which the principal independent variable was gender and the dependent variables were each of the health-care services studied. First, crude OR were calculated; then models were adjusted for groups of variables (predisposing, need and enabling factors). All the independent variables were modeled using dummies. The contribution of each group of variables to gender differences in the use of each health service was evaluated by comparing the OR of health services use of women versus men before and after adjustment for such groups of variables.
Analyses were performed using the SAS package, version 8.02 (2001) .