Design

The current study was nested in the LESS Study [31], which is a cross-sectional anonymous survey-study that evaluates the overall willingness to deprescribe and the barriers and enablers towards the willingness to deprescribe in older Swiss individuals with multimorbidity and polypharmacy. This manuscript reports the results of patients’ attitudes towards deprescribing related to PIMs.

Study population

Sixty-four general practitioners (GPs) from the German-speaking part of Switzerland recruited primary care patients for involvement in this study. All of them were located in different GP offices. Eligible patients were ≥70 years old with multimorbidity and polypharmacy. Multimorbidity was defined as the presence of ≥3 chronic diseases, with chronic diseases being present for at least six consecutive months [32]. Polypharmacy was defined as the concurrent use of ≥5 long-term medications [33,34]. GPs were instructed to consecutively screen eligible patients and recruit 5 participants, reporting the number of patients screened, to reduce the risk of selection bias. The questionnaire was completed by a total of 306 patients, 6 of whom were excluded based on missing information about prescribed medication. Patients anonymously filled in the survey and handed it back to the practice nurse to limit the chance of social desirability bias.

Questionnaire

For this study, we used data from 300 questionnaires on demographic status like age, gender, living situation, help with medication intake, involvement in medication self-management and education level. As for willingness to deprescribe, we used data from the revised Patients’ Attitudes Towards Deprescribing questionnaire (rPATD). This is a validated and reliable tool that has been applied in multiple studies [1,35–38]. It contains 22 questions on a 5-point Likert scale, ranging from “strongly agree” to “strongly disagree” which relate to beliefs and attitudes about their medications and deprescribing [37]. The rPATD was translated into German as previously described [31].

Medication appropriateness

In the present study, we used the self-reported list of prescribed medications and medication dosages for the assessment of PIMs. Self-reported medication is proven accurate and valid for long-term medication in the general population [39,40] and was chosen in this case to specifically focus on which medications patients report they take. The self-reported medication list was checked for inconsistencies (e.g. spelling errors) before analysis of PIMs was performed. In case of uncertainty regarding self-reported medication (e.g. due to poor or unreadable handwriting), in consultation with a GP researcher, the best applicable option was chosen [41]. Next, each medication was coded according to the WHO ATC-coding system. For the assessment of medication appropriateness, a selection of the AGS Beers 2019 criteria was used [42]. The AGS Beers list is the most commonly used tool for assessment of PIMs worldwide [43]. Since data on medical conditions was limited in this study, we used only the criteria that were applicable without clinical information (52 of 97 criteria). A list of the included criteria is added in S1 Appendix. Criteria were excluded based on weak strength of evidence as defined in the AGS Beers list (n = 8) or lack of information (n = 37). Application of a subset of the Beers 2019 criteria is in line with previous studies that used subsets of the Beers criteria for assessing medication appropriateness [29,44–46].

Willingness to deprescribe

Our main outcome was the willingness to deprescribe in relation to medication appropriateness. We therefore analysed data from the rPATD where patients were asked if they are satisfied with their current medications and if they are willing to deprescribe if their doctor said it was possible, along with 20 other questions grouped into four factors: involvement, burden, appropriateness and concerns about stopping, as described elsewhere [1].

Ethics

Ethical approval for this study was obtained from the Ethics Committee of the Canton of Bern, Switzerland (Ref. 2017–02188). All patients provided written informed consent before participating in the study.

Statistical analysis

Before the analysis, consistency checks were performed on the complete data set including the AGS Beers criteria and uncertainties were resolved by consensus of two researchers. Descriptive results were presented in frequencies, proportions, means and standard deviations (SD), and 95% confidence intervals (CI) were appropriate. Hypothesis testing for categorical variables was done using Chi-squared tests and simple linear regression for continuous variables when normally distributed. Patients with at least one PIM in their medication list where grouped to ‘PIM yes’, all others to ‘PIM no’ (exposure).

The individual scores (n = 22) of the rPATD showed a non-normal distribution. For the multivariate model we therefore dichotomized each of the 5-point Likert questions as well as factor scores according to the median as done previously [1]. Individual scores equal to or higher than the median were placed in the “high score” group, whereas scores below the median were placed in the “low score” group. In a multivariate model with different components of the rPATD as the outcome (satisfaction, willingness to stop, involvement, burden, appropriateness, concerns about stopping) [1], we calculated odds ratios (OR) and adjusted for age, gender and number of medications. To account for possible clustering of answers from patients from the same GP, we chose a mixed-effects model with the individual GP as random-effects. In a sensitivity analysis, we repeated the same models but with number of PIMs as the exposure instead of PIM yes vs. no. Significance level was set at <0.05. Data analyses were performed using STATA version 15.2 (Stata Corp, College Station, TX, USA).

Summary

In this study, we found PIM to be prevalent (54%) in a consecutive sample of older patients with multimorbidity and polypharmacy in a primary care setting. PIM use was found to be higher in patients with more prescribed medications compared to less. Interestingly, females were more frequently prescribed a PIM, mostly benzodiazepines or other CNS-active drugs, than males. We observed no difference in patients’ attitudes and willingness to deprescribe in patients with or without PIM. Our findings suggest that willingness to deprescribe is equally high in patients with and without PIM. There was also no difference in the adjusted analysis in burden, appropriateness or involvement factor scores, but participants taking PIM had lower concerns about stopping. This may therefore indicate that older adults with polypharmacy are not aware of whether they are taking potentially inappropriate medications or not. Therefore, efforts to increase awareness of the concept of PIM may be beneficial to shared-decision making about deprescribing in regular practice. Our study has also highlighted some areas that could be targeted, such as long term use of benzodiazepines in females.

Comparison with existing literature

The proportion of participants receiving at least one PIM matches previous studies from several countries worldwide, which generally report prevalence’s between 40 and 80% [6–9,47]. We demonstrated that patients receiving 10 or more medications show a significantly higher risk of PIMs. This confirms findings from previous studies that the number of medications is the most important risk factor for PIMs [4,8,48]. Furthermore, we detected a correlation between gender and prevalence of PIMs. It has been previously reported that females receive a higher number of PIMs on average than males [7,47–50]. It has been suggested that this might be due to females being at a higher risk for developing multiple chronic conditions compared to males. This would imply that they are more susceptible to drug-drug and drug-disease interactions, which challenges appropriate prescribing [47]. Yet, the actual reason for this gender-difference is unknown. Similar to previous studies internationally [29,51], proton-pump inhibitors and benzodiazepines were the most common PIMs identified in our study population.

Interestingly, we found that patients’ reported willingness to deprescribe is not related to PIM use according to the 2019 AGS Beers criteria. As investigated in a recent study on PIM and deprescribing interventions that explored factors associated with deprescribing refusal, likewise, PIM use was not associated with acceptance or refusal of deprescribing [29]. Previous qualitative studies on deprescribing, have reported that patients generally lack knowledge of the potential harms of medications and rely on the GP as a central and prominent figure in decision-making [52,53]. These findings suggest that older adults are not aware of whether their medications are appropriate or not. Reported willingness to deprescribe was equally high in our participants with and without PIM. Furthermore, very few of the individual factors had evidence to support a relationship with the use of PIMs, which stems from the fact that the study might not have been sufficiently powered to detect such differences. The factors ‘burden’ and ‘concerns about stopping’ were associated in the unadjusted analyses, but the burden association was lost in the adjusted analyses (likely because of the confounding nature of PIMs being associated with number of medications). Therefore, it is still unclear if and how use of inappropriate medications influences attitudes and beliefs or vice versa. The overall high willingness of older adults with polypharmacy to deprescribe is promising for further implementation of deprescribing in primary healthcare and is in line with prior studies investigating patients’ attitudes towards deprescribing [1,25–30].

Limitations and strengths

We acknowledge several limitations in our study. First, we gathered information on prescribed medications through self-reported medication lists, which might have affected the completeness of the medication lists. The accuracy of self-reported medication data can vary with medication type and duration; with self-reporting generally being more accurate for long-term medications [39,40]. Specifically, certain medication categories (e.g. psychoanaleptics and analgesics) were previously found to be less reliably self-reported [40]. Therefore, by using self-reported medication lists in this study, our results may be an underestimation of the use of PIMs. Second, the prevalence of PIMs in our sample might be underestimated due to the limited generalizability of the Beers criteria, published by the American Geriatrics Society, as they are based on medications commonly prescribed in the United States. Additionally, the Beers criteria capture ‘potentially’ inappropriate medications and so the assessment of appropriateness is not individualised. Third, the recruitment of study participants by GPs could have introduced selection bias. However, since consecutive sampling was used for participant inclusion–as it was not possible to recruit a random sample–the risk of selection bias was minimised. Lastly, since our sample consisted of Swiss older patients, we do not know if our findings are generalizable to other populations. However, as other countries reported similar attitudes towards deprescribing (e.g. 88% willingness to deprescribe in Australia [1], 89% in Italy [25] and 92% in the USA [27]) this increases the confidence in our findings.

To the best of our knowledge, we were the first to investigate the relation between medication appropriateness and patients’ willingness to deprescribe using validated tools. The Beers criteria is the most widely used tool for medication appropriateness and has proven to be accurate in the assessment of PIM [42,43]. The Beers 2019 version is updated according to the latest evidence and includes drug-drug interactions when assessing PIMs. Lastly, we used the rPATD questionnaire, which is validated and has been used internationally to assess willingness to deprescribe [37,54]. We followed international standards with independent forward and back translation to translate the rPATD into German.

Implications for future practice and research

Although we did not detect an association of PIM and willingness to deprescribe, we did see positive trends of patients on PIM towards, for instance, the perception of having more side effects (14%) and taking medication they no longer need (47%). However, since willingness to deprescribe is equal, this implies that patients’ willingness does not seem to be driven by knowledge that they are on a PIM. This again indicates the need to raise awareness about PIMs in older patients with polypharmacy, especially in females. Clinicians should be encouraged to regularly discuss deprescribing and the fact that the risks and benefits of medication use can change over time. Currently, patient education materials are increasingly being developed that will likely add to patients understanding of PIMs [55]. However, the group most at risk for PIM are vulnerable (oldest-) old patients [13–15] that demonstrate highly varying care wishes and needs, thereby challenging clinicians to provide appropriate care. Hence, in addition to our main finding–medication appropriateness being independent of patients’ willingness to have medication deprescribed–this pleads for an even bigger role of shared decision-making in the deprescribing process.

Future studies should further investigate the relationship between enabling factors of deprescribing and medication appropriateness and whether patients’ attitudes and beliefs about medications may change with education [56]. Furthermore, they should focus on what patients consider inappropriate medication and which medications they would be willing to stop. Specific questions about patients’ awareness about PIMs should be included in future research. As we found PIMs to be more prevalent in females compared to males, gender specific causes of PIM should be assessed in future studies. We also suggest focusing on specific classes and/or categories of medications in future research into PIM and willingness to deprescribe as this has not yet been explored and could be informative for translating into practice. Lastly, future studies should apply multiple PIM assessment tools, as well as comprehensive medication reviews determining actual appropriateness.