4.6. Statistical methods

Statistical analyses were performed using the SAS statistical software package (SAS Institute Inc. 1990), release 6.1 to 8.2. The results on categorial variables were presented as cross-tabulations in the papers I–IV. Chi-square tests were used to compare the categorial variables between groups of persons. Fisher´s two-sided probability test was used when the Chi-square test was not valid. Student´s t-test was used for continuous variables. The statistical significance of the results was presented as p-values, and p-values of less than 0.05 were considered as statistically significant.

Paper I. Drug use (%) was first cross-tabulated with age and sex. The mean number of drugs per study person and per drug user was counted. The prevalence of the use of each medication category was calculated among all study persons, in the different age and sex groups and among the persons with polypharmacy. The persons with polypharmacy were compared to those with 1 to 5 drugs in use, and the odds ratios (95% confidence interval - CI) were calculated to assess the probability of polypharmacy among the drug users by comparing the prevalences of the persons with polypharmacy by sex in 1998–99 to those in 1990–91 as a function of time (Table 4a-b) and women to men in 1990–91 and separately in 1998–99 (Table 5a-b). All comparisons were made in the different age groups.

The drug users with 0 to 5 drugs were compared with the persons with polypharmacy in terms of the sociodemographic factors using the Chi-square test. After that, a multivariate model was created by means of stepwise logistic regression analysis in order to assess the probability of polypharmacy using the following explanatory variables: age, sex, marital status, basic education, previous occupation, living alone, smoking, and uses of home nursing services in both surveys and, in 1998–99, additionally chronic morbidity, self-perceived health, life satisfaction, and frequency of alcohol consumption.

Drug users with 0, 1–5, and > 5 drugs (polypharmacy) were first cross-tabulated with the different items of ADL, mobility, or IADL, to find out the prevalence of persons with difficulties or dependency in activities. In covariance analysis, the mean number of drugs was calculated after adjustment for age and sex among the persons with difficulties or dependence and without difficulties or dependence in activities, in order to find out if a greater adjusted mean number of drugs (95% CI) was independently associated with poorer physical functional abilities after controlling for age and sex.

Paper II. The prevalences of different psychotropic use patterns in community were compared with the national prescription data provided by Social Insurance Institution by age and sex. Psychotropic use (any use and regular use) (%) was cross-tabulated with age and sex in community. Concomitant (≥ 2, ≥ 3, or ≥ 4) psychotropic use was calculated (%), as was also the use of different psychotropics (%) among the young elderly (64–71 years old), who were different persons in 1990–91 and 1998–99.

Paper III. The classification of drugs by their sedative properties was applied only in 1998–99 to a cross-sectional survey of the home-dwelling elderly subjects in Lieto, because the classification was made around that time. It was not correct to make a similar classification in the early 1990s, because some drugs used at that time were from decades ago. The sales licences of some drugs had expired by that time, e.g., the sales licence of a respiratory drug consisting of diprophyllinum and amobarbital (Cardiofyllin comp®, sedative drug) had expired in 1990 (Source: National Agency for Medicines). The purpose was to make as current a classification as possible, which could be updated and utilized in the future years.

Paper IV. Persons with no sedative drugs (no drug users and drug users), with some sedative drugs (sedation sum score 1–2), and with polysedation (sedation sum score ≥ 3) were compared for sociodemographic factors. The groups (mainly with no sedatives and polysedation) were compared in cross-tabulation using the Chi-square test. After that, a multivariate model was created by means of stepwise logistic regression analysis in order to assess the probability of polysedation by using as explanatory variables age, sex, marital status, basic education, previous occupation, living alone, smoking, use of home nursing services, chronic morbidity, self-perceived health, life satisfaction, and frequency of alcohol consumption.

The physical functional abilities (ADL, mobility, or IADL) of the persons were first cross-tabulated with the sedation sum scores (0, 1, 2, 3, 4, ≥ 5) to show the trend of possible difficulties or dependence in abilities versus sedative drug use. Covariance analysis, after adjustment for age and sex, was used to compare the adjusted mean sedation sum scores (95% CI) between persons with or without difficulties or dependence in activities (ADL, mobility, or IADL).