This study was conducted at a single Veterans Affairs (VA) medical center with seven associated outpatient clinic facilities. Approximately 39 providers practice in these facilities. To understand the current status and best practices of outpatient medication reconciliation and the potential role for secure messaging in medication reconciliation, the study was designed in two parts: direct observations of medication reconciliation and secure messaging workflow in the primary care clinics, followed by in-depth interviews with providers.

Providers were recruited for this study from January 18, 2012 to May 30, 2012. To be eligible, providers practiced ambulatory primary care and had at least one year of clinical experience at the VA. A nurse informaticist (TM) assisted the study team in identifying the two clinic sites most active in the use of secure messaging. A list of eligible providers at these sites was generated and randomly scrambled in their order; several additional providers were randomly added to this list so that providers from the largest clinic site (and third most active in secure messaging) were represented. Providers were then contacted sequentially via email regarding their willingness to participate in an interview. In this process, only one provider declined. Interviews were conducted until theoretical saturation was reached; a total of 15 providers participated.

The inter-professional study team included two physicians (LH and SRS), two research assistants (JC and TM), a pharmacist (AP), a project coordinator (MS), and a systems engineer (CA).The VA Boston Institutional Review Board approved this study.

To confirm that providers were already engaging in medication management via SM and to better understand the team-based approach to SM triage in primary care, we conducted observations of how primary care staff and providers managed SM at the two largest clinics within the VA Boston Healthcare System. Four individuals on the study team (TM, AP, MS, and CA) led the observations of staff members (providers, nurses, and pharmacists) within two primary care clinics. To minimize observers’ bias, LH and CA created an observation protocol to guide each observer to report on the role and purpose of the staff member replying to the SM, as well as detailing SM workflow and actions taken to process each SM (see Multimedia Appendix 1). On the day prior to the observation, we contacted clinic staff members via telephone or email to explain the intended process of observation. We obtained in-person verbal consent from staff members on the day of observation. Each clinic site was observed once for 2-3 hours by two members of the study team, during which time each observer took field notes while staff members viewed and handled the SM in their inbox at the time. Specifically, we observed how staff members managed secure messages in their inbox, noting how the staff members routed messages to other members of the clinical team, how they referenced other information to respond to patients’ inquiries (eg, medication list or schedule of upcoming visits, both of which are found electronically in a location separate from the secure messaging portal), and how they responded by reply message or by telephone to the patients’ requests.

To characterize medication reconciliation practices, we conducted in-depth interviews with 15 eligible primary care providers. An interviewer (LH) obtained informed consent and conducted the interviews using a semi-structured interview guide that we developed (see Multimedia Appendix 2). The interview questions consisted of both closed-ended and open-ended questions (eg, “Are adverse drug events a significant cause of morbidity/mortality for your patients?” and “What does medication reconciliation mean to you?”). Probing questions were added to the script and improvised during the interview to enrich providers’ responses. Each interview lasted approximately 40 minutes. Interviews were structured around three main domains of interest: (1) clinician perspectives on medication reconciliation and adverse drug events, (2) practice of medication reconciliation in the ambulatory setting, and (3) use of, and potential for, secure messaging in medication reconciliation.

Providers were encouraged to make additional open-ended comments about their experiences. A total of 14 interviews were audio recorded; for one interview, the clinician’s responses were transcribed verbatim by hand per request. All recorded interviews were transcribed (JC, TM, and MS) for subsequent coding and analysis.

A co-investigator (CA) compiled handwritten notes from roughly 3 hours of direct observations based on the protocol (Multimedia Appendix 3). We incorporated findings from the structured observations of primary care practices in our analyses of the interview content.

Over ten hours of formal interviews produced 254 pages (double-spaced) of transcription. All transcripts were de-identified prior to coding. We conducted a content analysis informed by grounded theory, with the goal of developing a theoretical framework for using secure messaging for medication reconciliation [27,28]. The initial coding scheme was developed by a team (LH, JC, and SRS) from a sample of three of the interview transcripts: LH and JC independently reviewed these transcripts, annotating important themes, and then met with SRS to review the themes and generate a coding scheme. The newly defined coding scheme was then applied to the remaining transcripts (JC and TM) with regular team meetings to discuss discrepancies and to refine the coding scheme. Discrepancies were resolved by consensus. All coded transcripts were reviewed to ensure that the quotes selected were relevant and accurate to the established coding scheme (LH and TM). We used NVivo 8 for coding and analysis of the interview data [29].

Our objectives were to ascertain which members of the clinical team were viewing secure messages from patients, how messages were forwarded among staff members, and the workflow for resolving the requests or issues raised in the patient’s message. We also wanted to confirm that SM was being used to address issues related to medication management—a topic we hoped providers could address in detail during the in-depth interview. In observations of SM use within primary care, we observed that the first member of the clinical team to view each of the 42 observed patient-initiated SM were divided between 5 primary care providers (physicians or nurse practitioners), 4 licensed practicing nurses, 3 registered nurses, and 2 pharmacists. Among 42 SM managed by staff members during our structured observations, the most common subjects of patients’ secure messages related to medication management (20 messages) and general medical issues (7 messages). Among the 42 messages observed, the individual initially retrieving the message most commonly responded or completed the message immediately (27 messages) and less commonly reassigned the message to a more appropriate person (9 messages). In a minority of cases, there was a delay of more than two business days before responding to the patient’s message (6 messages). The observed workflow, illustrated in Figure 1, provided a foundation for creating relevant in-depth interview questions. Appreciating the team-based structure of SM triage and confirming that medication management was already being conducted were common experiences upon which providers could further elaborate during the interview.

We interviewed 15 eligible primary care providers about their medication reconciliation practices and perspectives on the use of secure messaging. Twelve of the providers interviewed were female and 10 worked at the largest primary care clinic within the local VA system. Eleven providers saw 10 or fewer patients in a typical workday. All providers reported having staff to assist with patient-care responsibilities and all providers deferred to a hospitalist for inpatient care of their patients. Providers consistently reported that medication reconciliation, based on their own definition, was performed less than 25% of the time prior to a patient encounter. Only one provider noted that her practice tracked adverse events. All but one provider recognized that the Computerized Patient Record System (CPRS), the VA’s electronic health record, included computerized decision support to improve medication safety. Comparing providers interviewed for our study with the demographics of other providers working at these same three clinic sites, providers in our study were more likely to be a physician, female, and practice at the two clinic sites closest to VA Boston’s hospital; study participants reflected the larger population of providers in this setting in terms of deferring to an inpatient physician for care of inpatients (Table 1).

Four overarching themes characterized providers’ experiences and perspectives with medication reconciliation in the primary care clinic (summarized in Table 2). First, providers shared the belief that good quality medication reconciliation could improve outcomes. Second, providers felt that achieving quality medication reconciliation among medically complex patients was often challenging. Third, providers identified systems-level obstacles that prevented high-quality medication reconciliation. Last, providers described opportunities they believed could improve medication reconciliation after hospital discharge. Each theme is examined further, using providers’ quotes to illustrate subthemes.

We identified 3 main themes with respect to providers’ SM use (summarized in Table 3): (1) the use of SM improved workflow, (2) the existence of patient and provider-level barriers to use of SM, and (3) providers believed in the potential of using SM to improve medication reconciliation in ambulatory primary care. Each theme is described in detail, with selected quotes exemplifying each theme.

In this study across multiple ambulatory clinics within a large VA health care system, we conducted direct observations of medication reconciliation and secure messaging workflow in primary care and interviewed primary care providers about their experiences with medication reconciliation. Our findings confirm providers’ perception that medication reconciliation has the potential to improve medication safety. Providers highlighted a number of patient-level obstacles hindering high-quality medication reconciliation, emphasizing the difficulty in achieving accurate medication reconciliation among complex or elderly patients. Providers identified limited time and support for medication reconciliation as key barriers. While the majority of providers felt that the task of medication reconciliation was the responsibility of the primary care provider, almost all providers favored shifting this often time-intensive task to support staff or pharmacists and suggested that the process could be optimized by being conducted prior to the visit, either virtually or in person.

With respect to secure messaging, providers commonly expressed positive experiences and reported that its use in primary care improved workflow, largely via team-based message triage, but felt that making it easier for patients to sign up for and use SM could foster more frequent provider use and increase patient adoption. We found that providers were optimistic about the potential use of SM for medication reconciliation, primarily seeing such a system as an opportunity to decrease ADEs.

Traditionally, medication reconciliation has been recognized as the responsibility of the prescriber, usually a physician. Our finding that most providers would choose to relinquish this role to pharmacists or share the responsibility with clinical support staff is noteworthy and supports prior literature demonstrating provider support for the role of pharmacists in medication reconciliation [13,30]. This perspective may reflect the increasing pressures in primary care, including a workforce shortage resulting in increasing patient care demands [31], pay-for-performance programs [32], as well as high levels of physician burnout [33]. Future research should focus on providers’ definitions of medication reconciliation and their perception of their role and the role of other team members in executing it.

Our findings underscore the ongoing challenge of effective medication reconciliation in primary care, particularly for medically complex patients and during transitions in care. Our study suggests that a tool designed to facilitate virtual (non-face-to-face) medication reconciliation via SM would likely benefit providers in primary care (Figure 2).

There is little literature describing effective or novel medication reconciliation practices in primary care [34]. One recent study demonstrated provider interest in a patient self-service kiosk linked to the EHR for medication reconciliation [35], while other studies have shown that existing SM users within the VA system have lower rates of health care utilization [36]. Surveys have consistently found high patient interest and satisfaction ratings when given the option to communicate with providers via SM [37-39]. At a time when large health care organizations are investing in patient portals and Web-based health management [8,40], our study contributes the perspectives of primary care providers to the existing literature that, to date, has focused on patients’ willingness and organizational-level readiness to embrace Web portals and SM to manage care.

There were several limitations of our study. First, interviews were conducted among a small number of individuals (mainly female physicians) at three clinics within a single institution, possibly limiting generalizability. Future research should examine perspectives from other members of the clinical team. One of the interviews was transcribed by the interviewer instead of being recorded (given the wishes of the interviewee), potentially impacting the data collected. However, this interview was conducted over a longer time frame in order to mitigate possible inaccuracies and to transcribe complete quotes. Second, because these interviews were conducted ahead of a pilot study recruiting patients to use a medication reconciliation tool within secure messaging, it is possible that some providers were alerted to this intervention by patients participating in focus groups, priming their perceptions of SM use for medication reconciliation prior to the interview. Future research should examine providers’ perspectives in the proactive use of SM for a range of topics in primary care. Last, all providers were part of a large, integrated national health care system with a common EHR, established patient portal, and computerized provider order entry linked to pharmacy dispensing software, resulting in potentially distinctive provider perspectives on medication reconciliation and SM compared to providers practicing in clinics or private offices without such integration, and thus limiting generalizability.

With the pursuit of medication reconciliation as a National Patient Safety Goal [5,30,41-43], novel approaches to accurate medication reconciliation will be vitally important to improving medication safety and systems of care within primary care. Our study found that primary care providers, on the frontlines of patient safety and chronic illness management, recognize the importance of complete and accurate medication reconciliation. Providers favor having their professional colleagues, such as nurses and pharmacists, assume primary responsibility for medication reconciliation and express enthusiasm about aligning and shaping SM for the purposes of medication reconciliation. At a time of organizational readiness for patient-led online health management [8,40], future studies should focus on the design and implementation of SM or other Web-based tools for medication reconciliation.