Background
Adverse drug events (ADEs) are associated with increased healthcare utilization, costs, and morbidity [
1,
2]. One approach to prevent ADEs is to discontinue medications that are outdated, not indicated, or of limited benefit relative to risk for a particular patient [
3]. This activity, also known as deprescribing, has been defined as a “systematic process of identifying and discontinuing drugs in instances in which existing or potential harms outweigh existing or potential benefits within the context of an individual patient’s care goals, current level of functioning, life expectancy, values, and preferences” [
4].
While discontinuing a medication can be considered “doing less,” it often requires more provider effort than simply continuing the status quo. Literature supports the clinical feasibility and safety of medication discontinuation but does not necessarily provide information about how to identify which medications can be deprescribed or how to proceed in typical practice [
5]. One example is the “Good Palliative-Geriatric Practice” framework developed by Garfinkel et al., but this effort was a feasibility study and the method appears to be labor intensive, making its routine use difficult [
6]. Others have described the success and failures of prescribing interventions but give little direct guidance to clinicians; [
7,
8] one exception has been developed in New Zealand [
9].
Qualitative research exploring prescribers’ understanding and approach toward discontinuation revealed concerns about the negative effects of inappropriate medication use and overall support for the idea of discontinuing unnecessary medications [
10]. However, clinicians also discussed the many factors that impede their ability to deprescribe, including patient complexity, clinical uncertainty, and shared management with other healthcare providers, all of which can contribute to “clinical inertia” around medication discontinuation [
10,
11].
The development of clinical interventions that are integrated into the clinical workflow may facilitate appropriate deprescribing decisions. However, to increase the likelihood of success of any such intervention, the interest and support of the end-users (in this case, the prescribing clinicians) need to be garnered. To that end, we undertook the present study to survey a national sample of primary care providers and clinical pharmacy specialists in the US Department of Veterans Affairs (VA) to determine preferences for interventions that would improve their ability to discontinue medications.
Results
A total of 411 clinical providers completed the online questionnaire. After accounting for unreachable individuals (
n = 25), the response rate was 17% (411/2475). Non-responders were more likely than responders to be physicians, as compared with NP/PAs or pharmacists, but were otherwise similar with respect to age, gender, and geographic region. Details regarding respondent demographics are in Table
1. Regarding data quality, the median percent of missing responses per item on the substantive questions was 11.7 (range 0.01–16%); the median percent missing on the demographic questions was 17.2 (range 16–19%).
Table 1
Self-Reported Respondent Demographics
Age, years |
≤ 49 | 128 (37) |
50–59 | 125 (37) |
≥ 60 | 89 (26) |
Gender |
Male | 162 (48) |
Female | 174 (52) |
Race |
Non-Hispanic White | 243 (72) |
Other | 93 (28) |
Provider Type |
Physician | 304 (74) |
Nurse Practitioner or Physician Assistant | 68 (17) |
Clinical Pharmacy Specialist | 39 (9) |
Clinic sessions per week |
< 8 | 159 (52) |
≥ 8 | 175 (48) |
Number of years working in VA |
≤ 4 | 99 (29) |
5–9 | 84 (25) |
≥ 10 | 158 (46) |
Prior experience working in non-VA setting | 268 (79) |
With respect to medication discontinuation, clinicians rated themselves on the higher end of the comfort scale, with 73% scoring 7 or higher out of 10. A total of 11% agreed or strongly agreed that medications are overused, and 27% rarely or never encountered uncertainty about the indication for a medication. Additional details about respondents’ beliefs can be seen in Table
2.
Table 2
Respondent Beliefs and Perceptions
Self-rated comfort with medication discontinuation (0–10 scale) |
≤ 6 | 92 (27) |
7–8 | 138 (40) |
≥ 9 | 116 (34) |
Beliefs about medication overuse | |
Agree/Strongly Agree | 44 (11) |
Neutral/Disagree/Strongly Disagree | 355 (89) |
Perceptions of patients’ ability to manage their own health |
More Ability | 119 (33) |
Less Ability | 246 (67) |
Frequency of experiencing uncertainty about the indication for a patient’s medication |
Never/rarely | 107 (27) |
Sometimes | 209 (53) |
Often/usually | 80 (20) |
Experience with discontinuation of medications initiated by other providers |
Never/rarely | 51 (14) |
Sometimes | 295 (83) |
Often/usually | 10 (3) |
Experience with patient activation |
Low | 144 (36) |
Medium | 134 (33) |
High | 126 (31) |
Among the 326 respondents who provided rankings, the top choice for a change that would help improve their ability to discontinue medications was “Requiring all medication prescriptions to have an associated ‘indication for use.’” This preference was followed by “Assistance with follow-up of patients as they taper or discontinue medications is performed by another member of the Patient Aligned Care Team (PACT)” and “Increased patient involvement in prescribing decisions” (Table
3). This combination of options, albeit in varying rank order, was the most commonly selected, regardless of prescriber demographics, experience, or beliefs; overall, 250 (77%) of respondents who provided rankings included at least one of these in their top three highest ranked choices.
Table 3
Preferences for Interventions to Improve Ability to Deprescribe Medications
All Respondents |
All (326) | Indication for use | Assistance with follow-up | Patient involvement |
Age |
<49 years (118) | Indication for use | Assistance with follow-up | Information exchange with all pharmacies |
50–59 years (120) | Indication for use | Assistance with follow-up | Patient involvement |
≥60 years (83) | Indication for use | Patient involvement | Assistance with follow-up |
Gender |
Male (151) | Indication for use | Assistance with follow-up | Patient involvement |
Female (165) | Assistance with follow-up | Indication for use | Patient involvement |
Race |
White (235) | Assistance with follow-up | Indication for use | Patient involvement |
Non-White (81) | Indication for use | Patient involvement | Clinical decision support |
Provider type |
Physician (243) | Indication for use | Assistance with follow-up | Patient involvement |
Nurse Practitioner or Physician Assistant (53) | Indication for use | Assistance with follow-up | Patient involvement |
Clinical Pharmacy Specialist (30) | Indication for use | Information exchange with all pharmacies | Improved information exchange with providers located at community facilities |
Number of clinic sessions per week |
≤7 (148) | Assistance with follow-up | Indication for use | Patient involvement |
8–10 (167) | Indication for use | Patient involvement | Assistance with follow-up |
Number of years working in VA |
≤4 (94) | Indication for use | Information exchange with all pharmacies | Assistance with follow-up |
5–9 (80) | Assistance with follow-up | Indication for use | Patient involvement |
≥10 (146) | Indication for use | Assistance with follow-up | Patient involvement |
Prior experience working in a non-VA setting |
Yes (250) | Indication for use | Assistance with follow-up | Patient involvement |
No (69) | Assistance with follow-up | Patient involvement | Indication for use |
Self-rated comfort with medication discontinuation (0–10 scale) |
Low, 0–6 (83) | Assistance with follow-up | Indication for use | Information exchange with all pharmacies |
Medium, 7–8 (126) | Assistance with follow-up | Indication for use | Patient involvement |
High 9–10 (108) | Indication for use | Assistance with follow-up | Information exchange with all pharmacies |
Beliefs about medication overuse |
Neutral/disagree/strongly disagree (288) | Indication for use | Assistance with follow-up | Patient involvement |
Agree/strongly agree (38) | Patient involvement | Indication for use | Information exchange with all pharmacies |
Perceptions of patients’ ability to manage their own health |
Less (223) | Indication for use | Assistance with follow-up | Patient involvement |
More (101) | Indication for use | Patient involvement | Assistance with follow-up |
Frequency of experiencing uncertainty about the indication for a patient’s medication |
Never/Rarely (83) | Indication for use | Assistance with follow-up | Information exchange with all pharmacies |
Sometimes (173) | Assistance with follow-up | Indication for use | Patient involvement |
Often/Usually (68) | Indication for use | Patient involvement | Assistance with follow-up |
Experience with discontinuing medications initiated by other providers |
Never/rarely (46) | Patient involvement | Indication for use | Clinical decision support |
Sometimes (271) | Indication for use | Assistance with follow-up | Patient involvement |
Often/usually (8) | Assistance with follow-up | Indication for use | Patient involvement |
Experience with patient activation |
Low (117) | Indication for use | Assistance with follow-up | Information exchange with all pharmacies |
Medium (108) | Assistance with follow-up | Indication for use | Patient involvement |
High (97) | Patient involvement | Indication for use | Assistance with follow-up |
There were a few exceptions to this combination of preferred interventions. “Improved information exchange with all VA and non-VA pharmacies to confirm medication reconciliation” was selected more often than “increased patient involvement” among respondents who were younger (age ≤ 49 years), had worked in the VA for ≤4 years, had never/rarely encountered indication uncertainty, had low or high self-rated comfort, or had lower experience with patient activation. “Improved information exchange” was also chosen by respondents who believed medications were overused, rather than selecting “assistance with follow-up of patients.” Providers who reported rarely discontinuing medications by other clinicians or who were non-white preferred “Use of clinical decision support tools, such as drug-drug alerts, within the electronic health record” instead of “assistance with follow-up.” Finally, clinical pharmacy specialists ranked having an “indication for use” first, followed by “improved information exchange,” but then ranked “Improved information exchange with providers located at community facilities” third.
Fifty-one respondents (14%) wrote additional suggestions or comments. The majority were variations on the 15 proposed interventions, expressing support for division of labor among PACT members, improving medication reconciliation, and the benefit of additional time with patients. Others requested explicit clinical guidelines and criteria for deprescribing.
Discussion
We surveyed a national sample of physicians, nurse practitioners, physician assistants and clinical pharmacists in VA to assess their preferences for interventions to support deprescribing. Including the indication for use on all prescriptions was the highest ranked choice to support clinical providers’ efforts to discontinue medications, suggesting that this intervention may have wide acceptance among many clinicians across VA. Other highly rated interventions included increased teamwork to monitor patients after deprescribing and including patients in the decision-making process.
Much as clinical uncertainty can impede a physician from intensifying medication, it can also hinder discontinuation. If a patient is tolerating a medication and the indication (clinical reason for a medication’s use) is unclear, there may be reluctance to stopping it without a compelling reason. In a study by Straand et al. looking at patient-physician agreement about the decision and rationale for discontinuing a medication, up to 30% of the medicines had an unclear indication [
16]. One factor contributing to this uncertainty could be the frequent use of off-label prescribing in practice, which one study found accounts for over 20% of all prescription medications [
17]. Certain medications, such as gabapentin and amitriptyline (an anticonvulsant and antidepressant, respectively) are used off-label up to 80% of the time [
17]. Given that one of the first steps in the deprescribing process is to review the rationale for use so as to identify potentially inappropriate medications, not knowing the indication for a medication is an early obstacle to discontinuation [
3,
8].
Many patients have medications prescribed by multiple providers in separate healthcare systems and filled in different pharmacies [
18]. This system-level complexity also contributes to loss of information about indication for use [
19]. In the Straand study, over half of the medicines discontinued were prescribed by someone other than the discontinuing provider [
16]. Compounding the impact of such information loss is potential reluctance among many clinicians to stop a medicine initiated by another provider [
3,
19]. Knowing the indication for a prescription may provide reassurance to the discontinuing provider that deprescribing is appropriate.
Others have proposed similar interventions, such as including a “review by” date or “planned duration of use” that would prompt review of a medication’s continued appropriateness [
4]. Likewise, the Centers for Medicare & Medicaid Services requires nursing home patients to have an indication for all medications [
3]. One intervention study prompted inpatient providers to document the indication for three specific medications if an appropriate one was not found on the problem list [
20]. This study found including an associated indication at the time of prescribing increased, albeit these were often for off-label or non-evidence based use. Evaluation of this type of intervention in the outpatient setting and determining its effect on future discontinuation is still needed.
Including the indication with prescriptions has known and potential benefits in addition to aiding deprescribing decisions by the clinician. It may allow for therapeutic substitutions, in that the clinician can evaluate if there are new treatment options that were unavailable when the medication was initiated. In situations where prescriptions are handwritten, including a diagnosis allows a pharmacist to confirm that the medication is the one intended, especially when the handwriting is poor [
21]. It may also help with drug utilization reviews [
22]. Related to the findings in our study, adding an indication may increase patient knowledge [
23]. One of the top ranked interventions was to increase patient involvement in decision-making; educating patients at the
initiation of a medication should facilitate their participation in subsequent deprescribing discussions. In a review of 12 intervention studies by Ostini et al., including the patient was a key component of successful deprescribing [
24]. While it is unclear whether the indication should be written in medical or layman’s terms, either approach could potentially educate the patient. Including the indication on the printed prescription and/or the label of the medication packaging might allow for more feedback from the patient about whether the clinical problem still exists or whether the medication ever alleviated symptoms. It has been suggested that conversations about deprescribing may also lead to increased discussion about overall goals of care; [
25] given the high cost of medical care at end of life, providing the indication and consequent deprescribing may have even greater implications on healthcare utilization.
Another highly preferred intervention was involvement of other members of the Patient Aligned Care Team, VA’s patient centered medical home. This finding speaks to the time constraints felt by many PCPs and their desire for workload support [
19]. In a corollary to the increased monitoring of patients often required when initiating or increasing the dose of a medication, discontinuing or decreasing the dose may lead to similar increased requirements for surveillance, communication and office visits. Knowing that this follow-up could be managed by another team member (e.g., a nurse) may make prescribing clinicians more amenable to enacting medication changes. Indirectly, redistributing tasks may free up clinicians to give more time to re-evaluating the continued need for a medication rather than just renewing a prescription. In one study, the majority of outpatient prescriptions were renewals (72%) as compared to initiation of a new medication [
26]. It may also create more time in the clinical visit to have the conversation with the patient about discontinuing a medication, a process which might, in fact, be less time consuming if the patient is more knowledgeable.
These findings are of use to clinical practice managers in helping to guide future changes to clinical practice. The consistency of the preferences among the providers in our study, regardless of provider characteristics, experience or beliefs, indicates that these intervention options are more likely to be accepted by a wider range of providers. Further, including an indication could be integrated into electronic health records and computerized order entry systems. Other health information technologies and clinical decision support systems have been shown to reduce the initiation of inappropriate prescribing [
27,
28]. Requiring the prescriber to select from approved indications and commonly used off-label reasons, with an option to write in alternate rationale, could lead to widespread inclusion of indication for use.
Several limitations to our study should be noted. The survey response rate was low, which could reduce the ability to generalize to all VA clinicians. Multiple factors may have contributed to this, including clinicians’ busy schedules and the ability to only access the survey from behind the institutional firewall. Clinicians have lower response rates than the general population, [
29,
30] and there was no incentive provided. However, other than having a higher proportion of physicians, the non-responders were similar to the responders with regard to age, sex and geographic region. Also, the survey was conducted in VA and may not be generalizable to other health care settings. Future studies outside VA should examine prescribers’ preferences for interventions to support deprescribing.
Acknowledgements
We thank Barbara Bokhour, PhD, for her contributions to the development of the survey instrument. We also thank Rachel Lippin-Foster for her assistance with manuscript preparation.