You are on page 1of 2

European Journal of Epidemiology 16: 19±26, 2000.

Ó 2000 Kluwer Academic Publishers. Printed in the Netherlands.

Sociodemographic factors related to self-medication in Spain

Adolfo Figueiras1, Francisco Caama


no1 & Juan JesuÂs Gestal-Otero1,2
1
Department of Preventive Medicine and Public Health, University of Santiago de Compostela; 2Preventive Medicine Service,
University Hospital of Santiago, Santiago de Compostela, Spain

Accepted in revised form 11 October 1999

Abstract. To identify the sociodemographic factors self-medication in the sample was 2.5% during the
associated to self-medication (i.e. use of non-pre- two weeks previous to the interview. Undesirable self-
scription medicines) and undesirable self-medication, medication is twice as common among persons older
a cross-sectional study was carried out using a sample than 40 years, as compared to persons younger than
(n = 20,311) representative of the population of 27 years. Undesirable self-medication prevalence is
adults of 16 years of age and older in Spain. Multi- 53.0% higher among those who live alone as com-
variate Cox's regression was used. The prevalence of pared to those who live with their partner (95%
self-medication in the sample was 12.7% during the con®dence intervel (CI): 15.2±103.2) and 36.8%
two weeks preceding the interview. Self-medication is higher among students as compared to full-time
more prevalent among women, persons who live workers (95% CI: 1.9±83.5). People over 40 years of
alone, and persons who live in large cities. For per- age, people living alone, and students should be
sons who reported acute disorders, self-medication the priority target populations for public health edu-
prevalence was higher among those with higher cation programs aimed at improving the quality of
educational levels. The prevalence of undesirable self-medication behavior.

Key words: Cross-sectional studies, Non-prescription, Prevalence, Proportional hazards models, Self-
medication

Introduction Various de®nitions of SM can be found in the


literature. Some authors de®ne SM as the use of
Self-medication (SM) can be understood as a modern any drug not prescribed by a licensed health prac-
style of self-care which is becoming increasingly im- titioner, while others widen their de®nition to in-
portant in developed countries. Factors such as a clude any alteration to the recommended dose
decrease in prescription coverage by the social secu- guidelines undertaken by the patient [11]. Other
rity services, an increase in the use of alternative authors consider SM any use of medicines which,
remedies, availability of over-the-counter health although prescribed by a physician, were given due
products, and an increasing presence of pharmaceu- to direct pressure from the patient [12]. For this
tical products in the media, are probably among the study, we consider SM as the use of any non-pre-
causes of the rise of SM [1, 2]. The World Health scription medicine. We also noted a lack of studies
Organization has indicated that SM is acceptable, but evaluating undesirable self-medication (USM) ±
that it must be correctly taught and controlled until it de®ned by us as the use of medicines that, due to
is fully integrated into social behavior [3]. their nature, should not be used without medical
Despite the increasing relevance that SM is supervision ± and of studies identifying the popu-
acquiring in developed countries, there are few lation groups in which USM occurs with higher
studies which shed light upon the subject [4±6]. frequency. Data on USM would be very useful for
When reviewing the available literature, it becomes the design of health education programs to improve
clear that previous studies on the sociodemographic medicines use.
characteristics of SM do not control for potential The objectives of this study are: to determine the
confounding variables through the use of multivar- prevalence of SM and USM in the Spanish popula-
iate statistical analysis [7, 8], and when they do so, tion over 15 years of age, and to identify sociode-
the results are dicult to interpret [9, 10]. In addi- mographic and medical factors related to SM and
tion, the associations between SM and some factors USM. While the study was designed to be explor-
are generally not strong, which is why the studies do atory, we formulated the preliminary hypothesis that
not always have the sucient statistical power to the prevalence of SM is higher among populations of
detect them. higher social and economic status.
20

Methods evaluated health disturbances. The acute disorder


variable measures the presence of acute disorders
Design, sample and data collection which would have imposed restrictions on free-time
activities during the two weeks prior to the interview
This study has a cross-sectional design, and the study (fever, headaches, diarrhea, fainting, coughing, colds,
participants were adults of 16 years of age and older ¯u, wounds and lesions, urinary irritations, menstrual
included in the Spanish National Health Survey of problems, etc.). The chronic disorder variable refers
1993. This sample is comprised of 20,311 persons to the presence of chronic disorders which may have
distributed in a non-proportional manner throughout limited the subject's main activity ± work or study ±
the various regions of Spain. The type of sample used in the previous year (hypertension, diabetes, hype-
was multi-stage, strati®ed by clusters, with random rcholesterolemia, heart disease, asthma, allergies and
proportional selection of the di€erent units of the stomach ulcers).
sample in the case of municipal areas, and simple
random sampling in the case of sections. Individual Data Analysis
sampling was carried out through the selection of
routes and quotas of age and sex. Face-to-face in- Univariate (prevalence), bivariate (prevalence ratio
terviews were carried out by especially trained inter- (PR)) and multivariate (adjusted PR) analyses were
viewers, between 15 February and 3 March 1993. The performed. Multivariate analysis was ®rst carried out
data collection instrument was based on the ques- using logistic regression [14]. We excluded from the
tionnaire used in the Spanish National Health Survey logistic models variables that had no e€ect, and were
of 1987. Details on the methodology of the Spanish not confounders or e€ect modi®ers of the other
National Health Survey have been published else- independent variables. We used the Hosmer±Leme-
where [13]. show test to determine the goodness of ®t of the
models to the data [15].
Variable de®nitions The odds ratios (ORs) obtained by applying lo-
gistic regression to cross-sectional data are prevalence
Three dependent variables were considered: thera- ORs which are an overestimation of the prevalence
peutic drug use (TDU), SM, and USM, all three di- ratio. Thus, once the de®nitive models were deter-
chotomous. TDU was de®ned as the self-reported use mined, Cox's proportional hazard regression was
of any pharmaceutical product in the two weeks prior applied, taking as an event the presence of the e€ect
to the interview, and SM as the use of any non-pre- (TDU, SM and USM), and assuming an equal ®cti-
scription medicine, during the two-week period prior tious follow-up period for all the participants. In this
to the interview. Self-medication using drugs from the way, and as Lee [16] demonstrated, a non-biased
following groups: anti-in¯ammatories (for rheuma- estimation of the PR and its con®dence interval (CI)
tism); cardiovascular disease drugs; antibiotics; anti- was obtained.
depressants and tranquilizers; and hypolypemiants Interactions between independent variables were
and hypoglucemiants, would not be appropriate analyzed using an additive model. We ®rst calculated
medical practice. Therefore, we de®ned USM as the the PRs for each of the combinations of interacting
reported use of at least one non-prescribed medicine variables using the jointly unexposed group as a ref-
from any of those drug groups, during the two weeks erence category. Then, we calculated the PR among
prior to the interview. Despite Spanish legislation those with both exposures that is attributable to their
requiring dispensation of pharmaceutical products interaction (proportion attributable to interaction or
with the pertinent physician's prescription, it has long AP) [17]. The con®dence intervals for AP were based
been standard practice that pharmacists can dispense on Wald-type statistics using approximate variance
prescription medicines at their own discretion. estimators [18].
Taking previous studies as a starting point, we se-
lected sociodemographic and medical variables from
the Spanish National Health Survey which could be Results
related to SM. Some of these variables were analyzed
in the same way as they appeared in the questionnaire In the two weeks prior to the interview, nearly half
(sex, living arrangements, and education level); others (44.8%) of the sample under study took some form of
were recorded by grouping categories (habitat size medication, 12.7% self-medicated, and 2.5% self-
(number of inhabitants in the municipality of resi- medicated in an undesirable fashion. Table 1 shows
dence), and occupation). The age variable was cate- the percentages of TDU and SM by therapeutic drug
gorized into quartiles and used in the logistical model type. The two most widely used drug groups during
as a dummy variable to avoid assuming linearity in that period were medicines for colds, ¯u, throat or
the age e€ect. The acute disorder and chronic disor- bronchial infections (14.2% of the sample), and an-
der variables are dichotomous and were generated algesics or antipyretics (9.8%). Among those who
from a group of items in the questionnaire that self-medicated, these two drug-groups also showed

You might also like