We conducted a retrospective cohort study of a convenience sample of patients from ambulatory care clinics at Veterans Affairs (VA) Boston Healthcare System from April 2010 until July 2011. Patients were former members of the United States military who sought and were eligible to receive care at the VA. Data were collected by fourth-year medical students, who were individually instructed on project processes as part of their Ambulatory Medicine Quality Improvement rotation. One student per month was assigned to the rotation. Immediately prior to a student-led clinical encounter, patients were given a printout of the electronic health record (EHR) listing of their medications and asked to answer two questions: “Which one of your medicines, if any, do you think is the most important? (if none, please write ‘none’)” and “Which one of your medicines, if any, do you think is the least important? (if none, please write ’none’).” If needed, the student assisted the Veterans by reading the questions or writing their responses.

Because the data collection was intended as a quality improvement educational project, informed consent was not obtained. One of us (AL) entered all data into an Excel database, and we analyzed only the first chronologic encounter for each Veteran. Patient responses were entered exactly as written. Specific medications identified by patients were classified into medication classes using VA Drug Class Codes. Other data extracted from the EHR included patient sex, age at visit and number of actively prescribed medications.

Our two primary outcomes were patients’ responses to the “most important” and “least important” questions. Responses to each question were categorized as one of four types: 1) One specific medication – the patient identified one medication only; 2) More than one medication – the patient reported more than one medication, chose medications to treat a diagnosis (e.g. “heart meds”) or wrote “all”; 3) None – the patient wrote “none”; and 4) Did not answer the question – the patient left the response blank or wrote “n/a,” “don’t know,” “uncertain” or “not sure.”

We determined the prevalence of each primary outcome. Frequency counts identified the medication classes involved for responses where specific medications were chosen. Finally, we used chi-square tests to assess for associations between patient factors and choosing one specific medication as most or least important. All analyses were performed with SAS version 9.2 (SAS Institute, Inc) or Excel (Microsoft). Statistical significance was set at alpha < 0.05. This study protocol was approved by the Institutional Review Board of the VA Boston Healthcare System.

Approximately one in three Veteran patients did not identify a most important medication, and more than two in three patients did not identify a least important medication. There are several possible reasons to explain this finding. Patients may not have understood the questions, leading them to choose medications to treat a particular condition – therefore not selecting a single medication – or simply to leave the response blank. Limited health literacy, reflected as poor understanding of their health conditions or of their medications and associated indications, may have contributed to low response rates [8–10]. Without full understanding of their medical problems, the potential consequences of untreated health conditions and the possible benefits and risks of medication for those conditions, some patients may have been unable to make an informed selection from their medication list. This notion highlights the importance of education and knowledge in enabling patients to maintain an active role in their health care decisions.

Another hypothesis for the observed pattern of responses is that the use of the word “important” may not accurately reflect the construct that we were attempting to measure. That is, patients may have perceived medications as important but may not have believed that they are of a necessity or benefit to their future health to warrant full adherence. Others have demonstrated that perceived adverse effects of medications often outweigh preventive benefits of medications [11]. Competing demands, such as financial or social obligations, may further outweigh patient perceptions of importance. Psychometric testing of the best way to assess patient prioritization of medications with reliability and validity will improve measurement in future studies.

Another possible explanation for our findings is that Veteran patients were in fact able to understand and identify a medication yet they were unwilling to share these beliefs with their providers. Patients discontinue prescriptions for a variety of reasons without informing their healthcare provider of this decision [12]. Improving communication between patients and providers can lead to better shared decision making and better adherence [2].

Our study results need to be interpreted in the context of several limitations. We analyzed a small convenience sample of patients from one site within a larger health care system, and the study participants’ responses may not reflect the views of Veterans receiving care elsewhere or the perceptions of non-Veterans. Future research involving multiple settings and patient populations will enable better generalizability. We also had limited information on patients’ comorbidities and other medications, restricting us from appraising patient responses as concordant with clinician opinion. However, these answers still reflect what the patient perceived, and it is recognized that patient beliefs are associated with medication adherence [4]. Additionally, explicit discussions may enable providers to better reconcile known conflicts between what they believe is clinically best and what the patient perceives as important [13, 14]. Finally, the training status of providers (i.e., medical students) theoretically could have influenced patients’ willingness to divulge their prioritization.

In this study population where the mean number of medications was nine, higher than most commonly accepted definitions of polypharmacy [15], a minority of patients were able to express a medication as least important. Better understanding of how patients prioritize their medications and how best to elicit this information will improve patient-provider communication and perhaps lead to discontinuation of medications that both the patient and the clinician feel have less importance.