In this study we have examined the use of prescription drugs issued in primary health care according to age, gender, income or education after adjustment for multimorbidity level. Our main findings were that age increased the odds ratio of having prescription drugs, despite adjustment for multimorbidity level. We found that males had lower odds ratios of having prescription drugs compared to females. The differences in the socioeconomic groups were also substantial, where people with the highest income level had the lowest prescription drug use and people with the second to lowest income level had the highest prescription drug use. People with the lowest educational level had the highest prescription drug use.
Age
The finding that age increased both the odds ratio of having prescription drugs and the rate of prescription drug use has previously been shown, when the total prescribing of drugs has been examined [3]. In this study we have only examined the prescription drugs issued in primary health care, and despite the fact that we have adjusted for multimorbidity level, higher age seems to lead to higher use of prescription drugs.
GPs may have different approaches to treatment of patients with a new diagnosis depending on age. It is probably more likely that younger people with a new diagnosis of for example hypertension are recommended changes in lifestyle, while older people may be more likely to be recommended a prescription drug. This may partly explain higher odds ratios of prescription drug use among the elderly. Some diagnoses are progressive, for instance diabetes mellitus, where worsening is expected with higher age [21]. Under these circumstances increasing prescription drug use among elderly is expected and may partly explain the higher rate of prescription drug use among the patients. The prescribing cascade may help to explain, why the elderly use more prescription drugs despite adjustment for multimorbidity level. The prescribing cascade is described to start as an adverse drug reaction that is misinterpreted as a new diagnosis. A new drug is prescribed to treat this "new" diagnosis, and at worst an adverse drug reaction against the new prescription drug is once again misinterpreted as a new diagnosis and another drug is prescribed [22]. With higher age and more diagnoses it is more likely that the elderly are put at risk of this prescribing cascade. The above may to some extent explain why the elderly have higher odds ratios and rate of prescription drug use, which calls for better quality in drug treatment in the elderly.
The difference in odds ratio of having prescription drugs according to age is less prominent, when prescription drugs issued in primary health care are compared with prescription drugs in the total population [8]. Since many elderly with chronic diseases are treated in primary health care in Sweden, we would rather expect that the differences would be greater in primary health care compared to studies where the total population was examined. It could be that GPs have a more holistic approach and that GPs are better at evaluating prescription drugs [23]. It could also be that GPs to a lower extent tend to follow treatment guidelines [24].
Gender
Males had significantly lower odds ratio of having prescription drugs compared to females, despite adjustment for multimorbidity level. This situation has been shown before in studies carried out on the total population [6],[25], but rarely with adjustment for multimorbidity level and rarely only in a primary healthcare population.
Females tend to utilise health care more often than males [26], which may partly explain the gender difference. Females also tend to seek more preventive care than males [27], which may further contribute to the gender difference. There is a gender difference in which diagnoses females and males are diagnosed with, and a former study indicates that this gender difference in morbidity may partly explain the gender difference in odds of having prescription drugs [28]. Former studies have shown that there is variability between how physicians prescribe drugs and it is possible that this may affect the gender difference [29]. It is puzzling that the odds ratio for males in the population to have at least one prescription drug is quite low (OR 0.66), while males among patients in parallel have a higher rate of prescription drug use (IRR 1.06). This could be interpreted as the barrier to initiate prescription drug treatment being higher for males. It is also possible that there is a gender difference in compliance to drug treatment, which has been indicated in some studies [30],[31].
The gender difference regarding odds ratios of having prescription drugs is less distinct when prescription drug use issued in the primary health care is compared with the total prescription drug use [8]. This could partly be explained by the fact that in Sweden, at the time of the study, oral contraceptive drugs were not prescribed by GPs in primary health care, but mainly issued by midwives belonging to secondary care.
Income
Individuals and patients with the highest level of income had the lowest odds ratio and rate of prescription drug use issued in primary health care. Individuals with the second to lowest income had the highest odds ratio of having prescription drugs.
Despite adjustment for multimorbidity level there are substantial differences between the lowest and highest levels of income in drug use in primary health care. This is interesting since theses differences cannot be explained by differences depending on multimorbidity level between the different levels of income. Individuals with the lowest level of income had not the highest but next to highest odds ratio of having prescription drugs. This condition may be interpreted as if the individuals with the lowest level of income were unable to afford the prescribed drug, which has been seen before [6]. In Sweden, there is a high cost threshold system for prescription drugs, which implicates that the patients do not pay more than a defined amount for prescription drugs, SEK 2200 (EUR 248), annually. Despite this benefit system, it may be interpreted in this study as both individuals and patients with the lowest level of income did not purchase their prescribed drugs issued in primary health care.
The difference between the different income levels in odds ratio of having prescription drugs was larger when prescriptions issued by only GPs were examined compared to the differences seen in another of our studies where the prescription of drugs in the total population was examined [8]. This indicates that there is a social gradient in the way primary health care is provided.
Education
Individuals and patients with the lowest level of education had the highest odds ratio and rate of prescription drug use issued in primary health care.
Former studies have shown that individuals with lower socioeconomic status appear to a lesser extent to act on information regarding health risks, e.g. smoking [32]. In many of our common chronic diseases, e.g. diabetes, hypertension and hyperlipidaemia, lifestyle changes are a first point of action, and if individuals with lower socioeconomic status do not act on the physicians' recommendations of lifestyle changes, this may lead to both higher odds ratio and rate of drug use. Utilisation of health care differs according to socioeconomic status with a higher consultation rate among individuals with low socioeconomic status [33]. This may lead to increased odds ratio of having prescription drugs among individuals with lower socioeconomic status.
The difference between the different educational levels in odds ratio of having prescription drugs was also larger when prescription drugs issued by only GPs was examined compared to when the prescription of drugs in the total population was examined [8].
Interactions
When data were analysed for each educational level the rate of prescription drug use among patients in every income level followed the same pattern as in Model 1. This means that also at different educational levels the income level still affects the rate of prescription drug use among patients.
Individuals with the lowest income level had the lowest odds ratio of having prescription drugs if they belonged to educational level 2 or above. This means that the poorest individuals use drugs to a lesser extent if they have completed only primary school. This elucidates further that there is a social gradient in the way primary health care is provided.
Limitations
ACG Case-Mix uses diagnoses to calculate multimorbidity level. It is, therefore, dependent on the quality of registration of diagnoses. The recording of diagnoses was not validated in this study, but a former study in Sweden has shown that 75% of the population have at least one diagnosis-registered encounter with a GP during a three-year period [34].
Even if all the prescription drugs in this study were issued in primary health care, some of the prescriptions were probably originally initiated in secondary care. That is, some of the prescriptions are probably iterations from secondary care and hence do not necessarily entirely reflect the prescribing of drugs in primary care.
In this study we examined the prescription drugs that were collected from the pharmacies and not the prescription drugs that were actually prescribed by the doctors. If the compliance was inadequate, the collection of drugs does not adequately reflect the prescribing of drugs.
We were not able to assess illicit drug use, and these drugs were not included in this study.
It would have been interesting to include the level of general practitioners in the multi-level analysis but in the Swedish Prescribed Drugs Register we had access only to data down to Primary Health Care Centre level.