This cross-sectional study is based on a large sample of older patients with polypharmacy recruited for the European randomised controlled multicentre trial PRIMA-eDS. Results suggest that frailty, multimorbidity, and obesity as well as lower physical and mental health composite scores on the SF-12 are independent risk factors for excessive polypharmacy. Also, in multivariable analysis, the country or even the region plays an important role. Old age alone (≥85 years) does not seem to increase the risk of polypharmacy and may even be associated with lower risk. In our study sample, sex, educational level, and smoking status apparently do not contribute to excessive polypharmacy.
Interpretation and comparison with existing literature
It is easily understandable that > 8 diagnoses contribute to excessive polypharmacy because guideline-adherent treatment of multiple diseases will inevitably lead to a large number of drugs being prescribed. This has also been shown by other studies with no polypharmacy as a comparison to excessive polypharmacy [
20,
21,
24].
In our study, being frail/terminally ill is significantly associated with excessive polypharmacy. Saum et al. [
33] and Herr et al. [
34] showed that taking ≥10 drugs compared to taking ≤4 drugs is significantly associated with frailty [
24]. Morley et al. [
35] described frailty as a risk factor for a medication increase as physicians lack a concept on how to treat frail old people and thus often start medication. However, the question of causality remains unsettled as polypharmacy may also lead to frailty [
36], and polypharmacy as well as frailty may be a result of multimorbidity. Studies suggested that there is a dose-response relationship between the number of drugs taken and the risk of being frail [
33,
37].
Regarding health-related quality of life, a lower physical and mental health composite score indicating worse functioning within these health domains are significantly associated with excessive polypharmacy. Jyrkka et al. found moderate and poor self-reported health to be risk factors for excessive polypharmacy compared to no polypharmacy [
20]. Polypharmacy patients usually suffer from several diseases and we expect them to have a reduced health-related quality of life due to illness. Here again, causality may not be easily determined as polypharmacy could also lead to a decrease in health-related quality of life e.g. due to adverse effects of drugs.
It is not surprising that obesity is associated with excessive polypharmacy. Obesity has been shown to lead to an increased use of drugs [
38] and can result in chronic diseases especially with advancing age [
39]. However, a causal inference cannot be made as chronic diseases caused by obesity may be associated with excessive polypharmacy and it might be that obesity is an intermediate variable or a confounder.
We found that being ≥85 years of age is a protective factor against excessive polypharmacy. This was also found by Kim et al. [
21] when comparing older patients with and without polypharmacy, and by Onder et al. [
24] who detected an inverse correlation between polypharmacy and increasing age. However, Jyrkka et al. [
20] found ≥85 years to be a risk factor for excessive polypharmacy compared to no polypharmacy, and two further studies [
19,
25] showed being ≥80 years of age to be a risk factor. One explanation for the decreased use of drugs might be that due to a limited life expectancy of these very old people preventive medications are stopped in order to improve the patients’ current well-being [
40]. However, whether this really happens is questionable. Age does not influence patients’ priorities in taking preventive medication and reducing adverse events [
41], and GPs find deprescribing of preventive medication difficult [
42]. Another interpretation could be that excessive polypharmacy patients die earlier and do not reach the very old age.
We did not find that sex was significantly associated with excessive polypharmacy. The literature is conflicting here [
19,
20,
24,
25]. In this study there was no significant association between educational level and excessive polypharmacy. In the literature it has been shown that educational level had an impact on health, however, this effect appeared to decrease with age and was not significant anymore in adults ≥51 years [
43].
Smoking contributes to the burden of disease. Surprisingly, smoking was not associated with excessive polypharmacy in older polypharmacy patients, but there were very few smokers among the patients in our study.
There was a slightly significant association between excessive polypharmacy and the study centre Germany 1 when compared to Germany 2. One possible explanation could be the differences between the two settings. Germany 1 recruited patients in a more rural setting in the former Eastern part of Germany while Germany 2 recruited patients in the large metropolitan area of the highly industrialised Ruhr-region. These differences cannot be explained by the variables recorded in this study and deserve further investigation. The study centre in the UK was significantly associated with excessive polypharmacy compared to the study centre Germany 1 in the multivariable analysis. A sensitivity analysis showed that the UK was significantly associated with excessive polypharmacy compared to all other centres. Interestingly, the univariable analysis showed a slightly divergent result which was not significant (OR 0.92; 95% CI 0.70–1.22).
The patients in Germany 1 seemed to be frailer and had more diagnoses compared to the patients in the UK. In multivariable analysis the UK resulted in having more excessive polypharmacy. A reversal of effect can result due to the adjustment in the multivariable model. A possible explanation could be the “Quality and Outcomes Framework” (QOF) introduced in 2004 in the UK, which set financial incentives for certain performance indicators (pay-for-performance). Among these were indicators that relate to chronic conditions [
44], some of them directly naming the prescription of certain drugs while other indicators indirectly entailed drug treatment in order to reach the targets [
45]. Studies observed rising prescription rates of drugs indicated by QOF around the time when the framework was implemented, such as lipid-regulating drugs, renin-angiotensin system drugs [
45], ß-blockers or antiplatelet therapy [
46].
Implications
Understanding the health characteristics of an aged population taking several drugs, and investigating factors influencing excessive polypharmacy is highly relevant in times when the geriatric population is growing. This study helps to develop targeted strategies to reduce polypharmacy by identifying factors contributing to excessive polypharmacy. Physicians should especially pay attention to their frail, obese patients that have > 8 diagnoses, check whether all medications are necessary, evidence based and appropriate, and whether there are relevant interactions. To do so, GPs should perform medication reviews for their patients with excessive polypharmacy on a regular basis to optimise these patients’ medication. They should allocate extra time to care for these complex patients which needs to be reimbursed by the health care system.
Strengths and limitations
The major strength of our study is that we examined a very large sample of older patients representing several different health care settings/countries. We collected various parameters in this geriatric study population which gave us a comprehensive overview of demographic, clinical and functional status, and recorded the frailty level to distinguish between the fitter and the less fit ones. A major limitation of our study is that its cross-sectional design does not allow conclusions on causality. Further limitations are that the health-related quality of life was self-reported and all variables in the eCRF were reported by the GP, by practice staff or by a clinical research nurse. Even though instructions to record patient data were the same throughout all settings, we do not know whether the documentation of variables differs in different settings. True drug consumption is difficult to assess. We instructed the GPs to talk to their patients about all drugs they are taking. However, we were not able to verify drug consumption. Also, in this cross-sectional analysis, only patients were analysed who were recruited according to the inclusion criteria of the PRIMA-eDS trial. We therefore could only investigate patient characteristics associated with excessive polypharmacy in comparison to less excessive polypharmacy as patients without polypharmacy were not included in the trial. Furthermore, external characteristics e.g. of the prescribers could not be taken into account, and we did not judge whether medication intake was appropriate or not.
A further limitation of this study is that multiple relationships between variables exist. Multicollinearity is the cause of conspicuous differences between univariable and multivariable analysis. The interpretation for the affected variables should be regarded with caution. Noticeable correlations were found for the relationship between the research centre and the educational level (Cramer’s V = 0.48), or the number of diagnoses (Cramer’s V = 0.40) respectively, as well as between frailty, the two SF-12 scales (physical health composite score Cramer’s V = 0.20 and mental health composite score Cramer’s V = 0.33) and age (Cramer’s V = 0.22).
The physical and the mental health composite scores as well as frailty were identified as risk factors in both univariable and multivariable analysis, but the ORs are smaller in multivariable analysis because of the dependencies. Frailty and health-related quality of life are closely associated [
47], still we retained these variables in our analysis as they measure different concepts.
In univariable analysis, age was not significantly related to polypharmacy. It could be that there is a connection of the variable age with information about the condition of the patient, such as frailty, health-related quality of life, and diagnoses. Frailty [
48], lower health-related quality of life [
49], and a high number of diagnoses [
50] are more common in old age. On the other hand, these factors increase the likelihood of excessive polypharmacy regardless of age [
33,
50,
51]. Therefore, the positive effect of high age may become more apparent when adjusting for these factors.
Problematic is the variable “research centre”, which significantly increases the associations found after adjustment for UK/Manchester. This is mainly attributable to the consideration of the number of diagnoses. After adjustment for this variable, the OR increases from 1.10 to 1.71. We did not want to give up the number of diagnoses as an independent variable, as in the literature this is reported as an important risk factor. The variable “research centre” also seemed essential for the model, since this variable represents a variety of influences, such as the quality of the data collection, country-specific features and so on. Yet, it must be regarded with caution.