Older people’s attitudes towards deprescribing cardiometabolic medication

A survey study was performed using the linguistically validated Dutch version of the revised Patients’ Attitudes Towards Deprescribing (rPATD) questionnaire [48, 49]. In addition to the rPATD questionnaire, an adapted version of the rPATD was included which was modified to investigate the attitudes towards deprescribing of a specific medication group instead of deprescribing in general, as proposed by Edelman et al. [48]. In this adapted part the items on appropriateness and concerns about stopping medication for four cardiometabolic medication groups were included. No additional validation was done for these specific items. Participants were included who were (1) 70 years or older, (2) were able to read and write in Dutch, and (3a) used a sulfonylurea and/or insulin, or (3b) used a statin and at least one antihypertensive but no insulin or sulfonylurea. The ability to read and write in Dutch was implicitly assumed since only surveys in Dutch were offered. Based on inclusion criteria 3a/3b, participants received specific rPATDs about those medication groups. This resulted in the following groups of patients completing the medication specific rPATD questions for (1) sulfonylurea only, (2) insulin only, (3) sulfonylurea and insulin, (4) antihypertensive medication and statin. No formal sample size calculation was made but using a rule-of-thumb for regression analysis (N ≥ 50 + 8 m, where m is the number of determinants) we would need at least 82 participants [50]. A pragmatic recruitment strategy was used aimed at recruiting at least 300 participants through five community pharmacies across the Netherlands. They were identified by the community pharmacist and a researcher using dispensing data from the pharmacy information system. The following Anatomical Therapeutic Chemical (ATC) classification system codes were used to assess medication use: statin (C10AA), antihypertensive medication (C02/C03/C07/C08/C09), sulfonylurea (A10BB) and insulin (A10A) [51]. Potential participants were invited using an email with a link to the online questionnaire (Qualtrics XM), or by mail with a paper version in case no email address was available. No reminders were sent. A question on informed consent was included in the beginning of the questionnaire and was collected from all participants. The online version was password protected and only one survey could be completed per IP address to ensure no duplicate surveys were completed. Participants were unable to skip questions but they could abort the process of completing the questionnaire. Items were not randomised, three rPATD items per page were shown and participants could go back to change their answers. Usability and the technical functionality of the questionnaire were tested by several researchers not related to this study. All participants were offered a 10,- gift card. Data were collected from June until October 2019 and were stored on a password protected user account.

The rPATD questionnaire consists of two global statements and four factors each containing five statements regarding medication and deprescribing of medication in general. Each statement can be answered on a 5-point Likert scale ranging from strongly disagree to strongly agree. The global statements refer to overall satisfaction about medication: ‘Overall, I’m satisfied with my current medicines’ and willingness to stop medication: ‘If my doctor said it was possible, I would be willing to stop one or more of my regular medicines’. The four factors cover ‘burden’, ‘appropriateness’, ‘concerns about stopping’ and ‘involvement in decision making’ in relation to deprescribing [49]. For the medication specific part, the items related to “appropriateness’ and ‘concerns about stopping’ were adapted by replacing the word “medication(s)” with either “insulin”, “antihypertensive medication”, “cholesterol lowering drug” or a specific generic name of the sulfonylurea the participant was taking. For example, “I would like to try stopping the insulin I use to see how I feel without it” in the appropriateness factor. The specific generic name of sulfonylureas was provided because it was anticipated that many participants would not be able to recognize sulfonylureas as a medication group. It was explained that diuretics or “water tablets” were also considered to be antihypertensive medication.

The patients’ age, sex, and the number of medications used (1–5, 6–10 or more than 10 medications) were self-reported. Information on the therapeutic area was derived from the medication dispensed by the pharmacists and self-report. Two groups were distinguished, that is, patients receiving sulfonylurea and/or insulin with or without cardiovascular medication (diabetes area) and patients receiving a statin and at least one antihypertensive medication but no insulin or sulfonylurea (cardiovascular area).

Participants were excluded from analyses when less than 50% of the total rPATD questions were completed. For each of the four factors from the rPATD the average of the five statements was calculated. The answers of the appropriateness factor were scored inversely to obtain the factor score [49]. Although the factor scores of the rPATD have no official cut-off points, we have categorised the average factor scores into disagree (1–2.5), neutral (2.6–3.5) and agree (3.6–5.0) for the descriptive statistics, as previously proposed [48]. Linear regression was used to assess the influence of the age, sex, therapeutic area and number of medications on appropriateness, concerns about stopping, involvement and burden. The answer of the two global questions were categorised into agree, neutral and disagree and ordinal logistic regression was used to assess the influence of the age, sex, therapeutic area and number of medications.

T-tests or Wilcoxon rank-sum tests depending on the normality of the data were used to assess differences in ‘appropriateness’ and ‘concerns about stopping’ between medication groups. Paired tests were used for within-participant comparisons between attitudes towards antihypertensive medication and statins. Unpaired tests were used for between-participant comparisons of attitudes towards the other medication groups. For the comparison between insulins and sulfonylureas, when participants completed the rPATD for both insulin and sulfonylurea, only the insulin scores were used to prevent a mixture of within and between participant comparisons. Chi-squared tests and unpaired t-test were used to compare participants that completed the online versus the paper survey on age, sex and number of medications. Cohen’s d was used to calculate effect sizes of differences between the medication groups. Bonferonni corrections were used to correct for multiple testing. Stata® version 14.2 was used to analyse the data.

The Medical Ethics Review Board of the University Medical Center Groningen concluded that the study did not require a Research Involving Human Subjects Act (WMO) approval because it is not a clinical research with human participants as meant in the Medical Research Involving Human Subjects Act.

Age, sex, therapeutic area or number of medications were not associated with the willingness to stop medication and the satisfaction with their current medication (Table 3). All participants with more than ten medications were willing to stop medication and therefore a reliable estimate of odds ratio was mathematically not possible. A Fischer’s exact test showed that there was no association between taking more than ten medications and willingness to stop (p = 0.22). No significant differences were found for the appropriateness, concerns about stopping and involvement factors for any of the explanatory characteristics. Taking more than ten medicines was significantly associated with a higher medication burden.

Both antihypertensive medication and insulins were considered more appropriate in comparison to statins not favouring deprescribing (Table 4). Additionally, insulins were considered more appropriate than sulfonylureas but this difference was not significant after correcting for multiple testing. There were no significant differences in concerns about stopping medication between insulins, sulfonylureas, antihypertensive medication and statins.

Looking at the items underlying the main factors of the rPATD questionnaire, although no statistical testing was done, some variations are of note (Supplementary tables, tables S1, S2, S3 and S4). It seems that fewer participants wanted to try to stop their insulin (12%), antihypertensive (12%) or sulfonylurea (19%) than their statin (32%) or one of their medicines in general (29%) (Table 2). Conversely, less participants were reluctant to stop their statin (38%) or sulfonylurea (38%) than their antihypertensive (46%), insulin (52%) or one of their medicines in general (55%). Furthermore, there were less participants who wanted their doctor to reduce the dose of their insulin (15%) or their antihypertensive (11%) compared to participants who wanted to reduce the dose of their statin (22%), sulfonylurea (20%) or one or more of their medicines in general (26%).

This is a first study assessing attitudes towards deprescribing across different medication groups. A linguistically validated questionnaire was used in a relatively large sample of older Dutch patients recruited from five different locations across the Netherlands. The most important limitation, common for survey studies, is a low response rate. The influence of non-response rate bias is hard to test but has found to be small in other survey studies [61, 62]. Selection bias can be introduced due to the topic of the survey. Patients that are more interested in stopping medication and are more involved in their medication might be more likely to complete the survey. This could result in an overestimation of the involvement score. The generalizability is reduced by excluding participants unable to read and write in Dutch. This is likely to result in an underrepresentation of minority groups and patients with low health literacy. For the comparison of the medication groups a mix of within and between participant’s designs were used. Differences in attitudes found in the between person analyses might be caused by differences in the population instead by differences caused by the medication groups. We did not distinguish between different antihypertensive medication groups, patients that took multiple antihypertensive medications answered the questions for these medication as a group, which might not reflect how they think about these medications. Lastly, we only included demographic and treatment characteristics that are easily accessible to health care providers in our analyses to determine which characteristics influence attitudes towards deprescribing. However, this provides an incomplete picture of which patients’ characteristics may influence attitudes towards deprescribing.

Previous research has already established that older people want to be involved in decisions about changing their medication, and this is also the case for older people with type 2 diabetes or cardiovascular diseases [21‐23]. When health care providers want to discuss deprescribing cardiometabolic medication, it is important to address the perceived appropriateness of the specific medication. Even more so because willingness to deprescribe may not be associated with actual potential inappropriate medication use, indicating that patients are not necessarily aware of which medication is inappropriate [28]. Particularly, an explanation may be needed why medication that was appropriate for the patient in the past, is no longer required or can even be harmful. Many of the same beliefs about medication addressed by physicians to support good adherence can influence willingness to stop medication. This needs to be taken into account when medication is started or intensified. This can be done by communicating already at the start of medication that the medication has to be taken for a long time, but that periodically re-evaluating is needed. This can facilitate deprescribing efforts in the future.

Although the participants using more than ten different medications appeared to be just as satisfied with their medication than those using less, they did perceive a higher burden from the medication and all of them were willing to stop medication. Targeting older patients with a large number of medications for deprescribing interventions could help relieve this burden. An Irish study, however, did not show a reduction in the rPATDs burden factor after a medication review was conducted in a group of 54 older patients, showing that reducing medication burden might be difficult to achieve [63].

In this study, we investigated the association between commonly available patients’ characteristics and general attitudes towards deprescribing and found that such general factors are of limited value to tailor conversations about deprescribing. Future research could focus on other patients’ characteristics that may influence attitudes towards deprescribing, for example, health literacy, relationship with the health care provider, social support, and illness beliefs or awareness, preferably using instruments that can be easily applied in clinical practice. Also, more attention is needed for medication-specific attitudes towards deprescribing.