Conducting implementation research in ... - Wiley Online Library

1 downloads 1002 Views 145KB Size Report
Mar 20, 2009 - on integrating patient decision support interventions for cancer screening .... receive evidence-based brochure DESIs about prostate and colon ...
doi: 10.1111/j.1369-7625.2009.00579.x

Conducting implementation research in community-based primary care: a qualitative study on integrating patient decision support interventions for cancer screening into routine practice Dominick L. Frosch PhD,*  Kirsty J. Singer BSà and Stefan Timmermans PhD§ *Assistant Professor of Medicine, Division of General Internal Medicine & Health Services Research, Department of Medicine, University of California, Los Angeles (UCLA),  Assistant Professor of Medicine, Palo Alto Medical Foundation Research Institute, UCLA, àStaff Research Associate, Division of General Internal Medicine & Health Services Research, Department of Medicine, UCLA and §Professor, Department of Sociology, UCLA, Los Angeles, CA, USA

Abstract Correspondence Dominick L. Frosch PhD Assistant Professor of Medicine Division of General Internal Medicine & Health Services Research Department of Medicine University of California, Los Angeles (UCLA) 911 Broxton Avenue Los Angeles CA 90024 USA E-mail: [email protected] Accepted for publication 20 March 2009 Keywords: decision support interventions, implementation

Background Despite a growing body of evidence supporting the efficacy of patient decision support interventions (DESI), little is known about their implementation in community-based primary care practices. Objective The goal of this study was to explore the feasibility of integrating the use of DESIs for cancer screening in primary care practices serving patients from diverse backgrounds and learn more about the potential barriers and facilitators of integration. Setting 12 community-based primary care practices in metropolitan Los Angeles. Main variables studied Qualitative field notes documented the roles played by staff and physicians in accomplishing project goals, the impact of the programmes on the clinical work-flow in the practices and other noteworthy observations. Results Practices that were better able to integrate the project had adequate clinic infrastructure, a relatively well-matched patient pool, and positive work and patient care environments. The remaining identified components, including staff facilitation and the physicianÕs role accounted for higher level differences between the clinics, acting as barriers and facilitators that distinguished practices that were able to work independently from those that required more assistance and, to a lesser extent, those clinics that did and those that did not meet the project goals. Discussion and conclusions This study suggests that implementation of DESIs to be used immediately before a consultation is feasible if the practice infrastructure can provide sufficient basic accommodation and physician and staff are dedicated to patient care goals that are implicit in the use of these tools. Overall, the physicianÕs role appeared to be the most important factor in determining whether project integration was successful.

 2009 Blackwell Publishing Ltd Health Expectations, 14 (Suppl. 1), pp.73–84

73

74 Conducting implementation research in community-based primary care, D L Frosch, K J Singer and S Timmermans

Introduction Over the last decades, we have witnessed an increasing emphasis on patient participation in clinical decision making through what is often called Ôshared decision makingÕ.1 A basic challenge in engaging patients in shared decision making is the information imbalance between physician and patient. Physicians typically know the options that are available to address a clinical problem, but often donÕt have the time to provide patients with detailed information about these. Patients on the other hand often lack knowledge about what options are available to them or how they might value the expected outcomes associated with different options.1 To address this problem, researchers have devoted significant effort to developing patient decision support interventions (DESIs) that are intended to correct this information imbalance and prepare patients for subsequent shared decision making with their physicians.2 A systematic review of randomized trials of DESIs found that they increased patient knowledge and realistic expectations about clinical options, and lowered decisional conflict. Individuals who reviewed a DESI were less likely to remain undecided and often made different decisions after reviewing the DESI.3 Despite the rapid growth of research on DESIs and evidence supporting their use, implementation of these interventions in clinical practice remains in its infancy.4 A few studies have explored the process of implementing DESIs in specialty and hospital care. Lack of time on the part of physicians and allied healthcare staff and competing priorities that make it challenging to provide DESIs to patients are consistent themes described in these studies.5–8 Many DESIs focus on clinical problems that are commonly addressed in primary care. Although numerous studies have evaluated the efficacy of DESIs in primary care settings at the patient level, little data are available to document the barriers and facilitators to implementing them in routine clinical care.3,9,10 The purpose of the present study was to explore the feasibility of integrating the use of DESIs

about two preference sensitive cancer screening decisions – prostate and colon cancer – in community-based primary care practices serving patients from diverse backgrounds.11 Our goal was for participating practices to independently identify eligible patients and provide them DESIs for review immediately prior to a clinical consultation. We focused primarily on small practices because independent solo ⁄ 2-physician practices make up 36% of primary care practices in the United States and most Americans receive their health care in non-academically affiliated practices.12 A better understanding of the factors that impact implementation is a critical intermediate step that can inform large scale implementation studies.

Methods Practice recruitment & study goals The study protocol was reviewed and approved by the UCLA Institutional Review Board. Potential practices were identified using the American Medical Association Masterfile, a database of practicing physicians. Our database was limited to family practitioners, internists and geriatricians whose practices were located in low-income zip codes of metropolitan Los Angeles. The first step in the recruitment process consisted of the project research assistant calling practices and requesting to speak to the physician or the practice manager. The research assistant provided an overview of the study explaining that we were interested in learning more about using DESIs in primary care practices. DESIs were described as brochures and videos that educate patients about their cancer screening options and can reduce the amount of time physicians spend educating patients, so that more of the clinical encounter can be focused on discussing patient questions and preferences. Our decision to focus both on simple brochure DESIs and more complex video DESIs reflects the diversity of media that have been used in developing these tools.2 By having all participating practices use both types of DESIs, vs. assigning some to use only brochures and

 2009 Blackwell Publishing Ltd Health Expectations, 14 (Suppl. 1), pp.73–84

Conducting implementation research in community-based primary care, D L Frosch, K J Singer and S Timmermans 75

others only video tools, we were able to explore potential differences in utilizing the tools depending on their complexity. Given its exploratory stage, we set modest patient enrollment goals for the project. Practices were told that they would be asked to enrol a total of 20 patients who were clinically eligible for prostate or colon cancer screening in two phases. During the first phase, practices would receive evidence-based brochure DESIs about prostate and colon cancer screening, developed by the US Centers for Disease Control and Prevention and the University of Texas, and would be asked to enroll 10 patients.13 During the second phase, the practices would receive video DESIs about prostate and colon cancer screening developed by the Foundation for Informed Medical Decision Making (FIMDM), as well as a $300 budget to purchase video equipment for showing the DESIs. Practices were asked to enroll another 10 patients during Phase II. We focused on cancer screening decisions for several reasons. First, cancer screening decisions are typically made in the context of primary care. Second, a 1997 California law requires physicians to offer men, undergoing a physical examination that includes the prostate, information about diagnostic procedures related to prostate cancer.14 Finally, at the time we began this project, the prostate cancer screening programme was the only DESI developed by FIMDM that was also available in Spanish. Although the colon cancer screening programme was not available in Spanish, we included this programme to avoid limiting our focus to male patients. Practices were told that the goal was to provide the DESIs to patients at the time of an office visit, to review before the consultation with the physician. Although this approach is more challenging to implement than allowing patients to take DESIs home, it provides more control in ensuring that patients actually review the tool. In a recent study of colon cancer screening decision aids that were mailed to eligible patients from an academic primary care practice, few reported reviewing the DESI they received.15

All practices were instructed that the mix of prostate and colon cancer screening could be determined by clinical needs of their patients. Hence, no specific goals were set with regard to how many patients considered either screening service in the context of the project. Finally, practices were told that project staff would consult with them to help determine the most efficient ways of integrating the use of DESIs into their clinical work-flow. The expectation was that practices would independently identify eligible patients and provide DESIs to review prior to a consultation. Following the initial conversation, practices that expressed an interest in participating in the study were sent a written description of the study by fax that reiterated the study purpose and goals. Once the written description of the project was received and interest in participating confirmed, a meeting was scheduled with the physician and practice manager. During this meeting, the purpose, goals and expectations for participating in the study were again reviewed and the self-reported ability of the practices to meet the project goals assessed. Of the practices with which meetings were scheduled, two declined participation and one was deemed to not have an appropriate pool of potentially eligible patients. Physicians who agreed to participate signed a project consent form. Physicians received $400 for participating in the project. Each physician also identified a staff member who would assist with the project. Practice staff also signed a consent form and received $250 for participating in the project. A total of 201 practices were contacted and 12 (6%) agreed to participate in the study. Of the 12 practices that agreed to participate, 10 were solo-practitioner offices and two were community health centres, in which one physician each participated in the project. Data collection The data evaluating patient responses to the DESIs are reported in detail elsewhere.16 Prior to commencing, four project staff members were trained by the investigators in the use of

 2009 Blackwell Publishing Ltd Health Expectations, 14 (Suppl. 1), pp.73–84

76 Conducting implementation research in community-based primary care, D L Frosch, K J Singer and S Timmermans

ethnographic field methods using a widely cited guide on writing ethnographic notes.17 For every consultation with participating practices, staff recorded detailed qualitative field notes. Observations were noted down on forms with six pre-structured categories: (i) the date, time and place of observation, (ii) observations related to the roles played by staff and physicians in accomplishing project goals and the impact of the programmes on the clinical work-flow in the practices, (iii) the patients present during the period of observation, (iv) salient interactions between staff-patients-and researchers, (v) presence of external parties (e.g. pharmaceutical representatives), and (vi) an open residual category of other noteworthy observations. Research staff expanded their notes at the end of each consultation resulting in approximately 400 pages of field notes that provide the basis for this analysis. The length of project staff consultations varied depending on the practiceÕs ability to independently work toward the project goals. In practices that werenÕt able to work independently, consultations required more time, creating more opportunities to gather observational data. To ensure adequate observational data for practices that were able to work independently, project staff still observed multiple occasions when patients were identified and provided a DESI. The research staff thus gathered observational data at each practice. Observations lasted an average of 2–3 h and took place during both phases of intervention. To safeguard the uniformity of field notes and lower the likelihood of observational biases, observers received feedback about the quality and quantity of field notes at weekly team meetings. Data analysis The focus of our data analysis was to use the qualitative field notes to develop an explanatory model that would illuminate the barriers and facilitators to identifying eligible patients and providing a DESI to review prior to a consultation with a physician, and meeting the project goals of recruiting 20 patients. We inductively

constructed a conceptual explanatory model to account for variation in the practicesÕ ability to meet the project goals. Drawing upon the qualitative grounded theory methodology,18 the three primary investigators derived the elements of the model through an iterative coding process of open-, axial- and selective coding of the observational field notes.19 In a first round of open coding, the research team distinguished a list of the possible themes and elements that played a role in meeting project goals. Open coding is a process of categorizing data in salient themes. The three researchers independently coded a full set of field notes of a random selection of four practices and collectively constructed a preliminary coding scheme. Then, during rounds of axial coding, the various elements were put together in the model and checked for redundancy and clarity. Axial coding refers to a set of procedures used to put categories distinguished during open coding together by making connections between categories and concepts. At this stage, each researcher independently coded a different subset of the remaining practices. To validate the explanatory power of the model and its fit with the data, each team member independently engaged in selective coding – coding procedures used to develop and validate conceptual models. Finally, during subsequent team meetings, the researchers used the model to go through the observational notes of all the individual practices to verify whether the model explained the siteÕs ability in meeting project goals. This process led to further fine tuning of conceptual categories and to distinguish the key characteristics explaining the practicesÕ ability to reach the goals and their variability in doing so independently. Through this process of gradual conceptual refinement of coding categories with additional data, the three team members reached a robust consensus about the varying levels of ability and independence for each individual research site. According to the grounded theory methodology, inductively derived conceptual models will gain theoretical sensitivity when elaborated, modified and incorporated in future research.19

 2009 Blackwell Publishing Ltd Health Expectations, 14 (Suppl. 1), pp.73–84

Conducting implementation research in community-based primary care, D L Frosch, K J Singer and S Timmermans 77

Results Practice reasons for participating in the project We achieved a low recruitment rate of 6% of practices contacted, as noted above. The most common reason cited by practices that agreed to participate was that they identified patient education as an important goal for primary care. Several physicians also cited previous positive experiences with research and indicated a desire to contribute to science. By contrast, among practices that declined participation, the most common reasons cited were lack of time, being too busy and general lack of interest. A few practices cited negative previous experiences with research and one practice indicated that research detracted from patient care. Practice ability to meet project goals During the conduct of the project, we observed significant variability in the practicesÕ ability and independence in meeting the project goals. Contrary to the expectations articulated during recruitment meetings, few practices were ultimately able to work completely independently in

meeting project goals. Rather, 10 of 12 practices required continuous hands-on assistance from project staff to meet the agreed upon goals (see Table 1, practices 2 & 4–12). In some practices, this translated into project staff providing the DESI to patients after clinic staff had identified them as eligible (practices 2 & 4–6), whereas in other practices, project staff had to remind clinic staff continually to check whether patients coming into the practice were eligible, as well as provide the DESI to patients (practices 7–12). Table 1 lists the medical specialty of each respective practice and the median household income of the neighbourhood in which the practice was located. Practices are ranked in the table by how many total staff consultations were necessary, with practices that were unable to complete the project goals listed at the bottom of the table. Practices that were able to work independently generally required fewer consultations. The average number of consultations required did not differ significantly between the two phases of the project. The mean number of consultations during Phase I for practices that completed both phases of the project was 10.89 (SD = 5.86). The mean number of consultations during Phase II was 11.56 (SD = 6.93; t(8) = )337, P = 745).

Table 1 Description of participating clinics and efficiency in meeting project goals

Medical Practice specialty 1 2 3 4 5 6 7 8 9 10 11 12

Family Medicine Family Medicine Community Health Center Internal Medicine Internal Medicine Family Medicine Internal Medicine Family Medicine Family Medicine Internal Medicine Community Health Center Family Medicine

Total number Total days Number of Neighbourhood Completed Number of to meet staff consultations staff consultations of staff project median consultations project goals – Phase II – Phase I goals income $50 610 $27 471 $32 168

Yes Yes Yes

3 4 7

6 6 8

9 10 15

103 47 189

$27 $33 $22 $33 $22 $27 $51 $24

Yes Yes Yes Yes Yes Yes No No

9 10 13 19 14 19 6 12

12 13 11 6 14 28 N ⁄A N ⁄A

21 23 24 25 28 47 6 12

130 133 91 167 210 172 N ⁄A N ⁄A

No

17

N ⁄A

17

N ⁄A

471 656 246 997 043 591 275 207

$26 884

 2009 Blackwell Publishing Ltd Health Expectations, 14 (Suppl. 1), pp.73–84

78 Conducting implementation research in community-based primary care, D L Frosch, K J Singer and S Timmermans

Explanatory model and variation between levels of ability and independence Drawing on the observational notes, the team was able to identify several key components that contributed to the relative abilities of the clinics to meet the project goals (see Fig. 1). These include the clinic infrastructure, the available patient pool from which to identify potential participants, the patient care and work environments, staff facilitation of the project and the physicianÕs role. Below, we define the components of the model and use quotes from the field notes to illustrate the differences between clinics of varying efficiency. Clinic infrastructure While the project sought to find best ways to implement DESIs within the existing infrastructure of all participating clinics, the research effort required certain basic accommodations in terms of space and an available stream of patients from which to identify participants. A combination of clinic infrastructure – that is, the resources clinics had to work with – and the utilization of those resources by staff members produced varying degrees of success at integrating project logistics.

Clinics with greater ability and independence were characterized by infrastructures that were able to accommodate the logistics of having patients review a DESI before a consultation: ÔThe video process runs very efficiently because of the space provided in the clinic, which allows for multiple patients to watch the video at the same time.Õ (Clinic #2) The set-up of the clinic is such that all the patients pass through the hall next to the reception desk, ...setup of the video room makes it very easy for patients to walk in and fill out forms and then watch video right away. After seeing the video they go into the exam room and wait for the doctor. (Clinic #4)

On the other hand, in clinics that were less able and those that were unable to meet the project goals, lack of adequate infrastructure created significant impediments to identifying potentially eligible patients: …high reception desk all around and surrounded by plexiglass. I am not able to interact with her as I am with the clinic managers at the other clinics. I always have to remind her to let me know if any patients are eligible. (Clinic #7) [Staff] office is in the back of the clinic so she does not get to see the flow of patients…Clinic appeared

INFRASTRUCTURE PHYSICIAN

PROJECT LOGISTICS

PATIENT CARE ENVIRONMENT PHYSICIAN'S ROLE WORK ENVIRONMENT

PROJECT/DESI INTEGRATION STAFF FACILITATION PATIENT POOL

PATIENT CHARACTERISTICS

STAFF

LOCAL POPULATION

Figure 1 Explanatory model.

 2009 Blackwell Publishing Ltd Health Expectations, 14 (Suppl. 1), pp.73–84

Conducting implementation research in community-based primary care, D L Frosch, K J Singer and S Timmermans 79 to be extremely busy…all patients are walk-in. (Clinic #11)

Patient pool Patients are generally drawn from the local population and share its demographic composition and characteristics. Patients may, however, comprise a particular subgroup of the local population. The size of the available patient pool is a product of these patient characteristics and the volume of patients visiting the practice at any given time. It is important to have a patient pool that is well matched with the eligibility criteria for the DESIs and of sufficient size to meet the project goals. Although all clinics in the project understood the patient eligibility criteria when they agreed to participate, there was a significant variability in the patient populations that impacted the practicesÕ ability to reach the project goals. Generally, more able practices served patient populations that matched the eligibility criteria well: …most of the patients are elderly, mostly Englishspeaking… (Clinic #1) …practice consists of all Geriatric population and mostly African American patients. (Clinic #5)

The limited availability of the DESIs in Spanish impeded progress in some clinics. However, it is important to note that practices #2 and #4 were in East Los Angeles, a predominately Spanish-speaking neighbourhood. [Patients are] mostly female, under age 50, many children…Spanish spoken exclusively. (Clinic #11)

Moreover, other patient factors could also lead to challenges, even in clinics in predominately English-speaking neighbourhoods: [Patients] range from kids getting immunizations to elderly patients.…many elderly patients coming in, but not being seen by the doctor. Quite a few patients came in for medication pickup. (Clinic #7)

Work environment Both the physician and the staff contribute to the work environment, although the physician may have a more significant role in setting the

tone for the practice. A well-organized and amiable work environment enabled staff participation in meeting the project goals. On the other hand, a jaded physician or dissatisfied staff members were far less likely to be motivated to make the special effort to integrate project goals into the workflow. Clinics that worked independently had positive work environments characterized by team work: Overall, it seemed that the staff were quite content working here. [There is a] light relationship between doctor and his staff as they tease each other a lot. (Clinic #1)

In other clinics, we observed varying levels of tension between physician and staff and among staff: Doctor likes to test his staff to make sure they know what theyÕre doing…Doctor was telling her [staff member] to prove that she knew how to give shots and to show what part of the body they should be given. Staff member was upset because doctor was disputing with her in front of the patient and she commented to us that the patient was uncomfortable in that situation... (Clinic #6) Staff member had been laid off because the clinic manager told doctor that she didnÕt do any work. Some tension between them and the other staff members are aware of it…Doctor said hello and told us right away that staff member was no longer working at the clinic, that she had stopped working without calling. [Another] staff member just stopped coming because she wasnÕt getting along with the other co-workers. (Clinic #8) It was evident that the staff relationship at this clinic is contentious. [Based on the observation of arguments between staff] (Clinic #12)

Patient care environment The quality of care is largely determined by the physician and staff, but the patientÕs experience of the visit will also be affected by the atmosphere in the waiting room and aspects of the clinic infrastructure, such as the length and predictability of the wait time. The patient care environment is in turn influenced by the rapport between the physician, staff and the patients. Rapport is likely to be better in practices with

 2009 Blackwell Publishing Ltd Health Expectations, 14 (Suppl. 1), pp.73–84

80 Conducting implementation research in community-based primary care, D L Frosch, K J Singer and S Timmermans

established patients who may feel loyal toward the physician. Such loyalty increases the influence of the physicianÕs implicit or explicit support of the project goals. In more capable clinics, we observed physician–patient and staff–patient relationships that were characterized by strong rapport and patientsÕ sense that they felt respected and well taken care of:

autonomy and investment in the goal of providing DESIs to patients prior to a consultation. In clinics that worked independently, staff was able to work proactively in identifying eligible patients and providing them the DESI and project materials:

[Doctor] talked in detail with the patient after he had viewed the video…he really loves explaining things to patients. (Clinic #1)

In less capable clinics, staff did not complete the necessary tasks independently, but were nevertheless actively engaged in identifying eligible patients:

[Patient] said heÕs been coming to this clinic for many years and is very pleased with the staff and the way they treat their patients. (Clinic #4)

By contrast, less capable clinics and those that were unable to complete the project were characterized by poor rapport and patient frustration due to long wait times: Pharma reps came in with catered lunch for the clinic. Doctor talked to them for at least 45 min, while his patients were out waiting in exam rooms. Staff seemed to be getting frustrated because the patients were complaining to her about the wait. Doctor seems to enjoy talking to the reps and always spends upwards of half an hour or more if they bring food. (Clinic #9) [Patient] then went to the front desk and told [staff], Ô[Doctor] always comes late. HeÕs gonna lose his patients like that. (Clinic #10)

Staff facilitation Ultimately, the integration of the project into the routine workflow of the clinic relied on staff facilitation. Facilitation can mean simply helping identify potentially eligible patients and creating space within the workflow to accommodate the demands of the project. At a basic level, staff needs to be accessible and not overburdened with other work. Clear role definitions are useful in enabling efficient processes within the existing infrastructure. Ideally, particularly as the goal of this project was to work toward implementation of the DESIs, staff would be able to identify eligible patients and provide DESIs independently. This requires staff that are motivated to take initiative and with sufficient

[Staff] had already attached the brochure and questionnaires to the patientsÕ charts. (Clinic #1)

[Staff member] has been very valuable in helping identify [eligible] patients, because he is able to express to the patient the benefit of receiving the information. (Clinic #4)

In other clinics, staff were significantly less engaged or overwhelmed by competing demands: I was not told about any patients by the staff or doctor. I had a feeling as I did on previous occasions that there are patients who do qualify but we are not made aware of them by the staff. (Clinic #7) Staff was too busy to screen potentially eligible patients …some patients have asked about the decision aids but staff was not aware of what to do. (Clinic #11)

PhysicianÕs role In addition to staff facilitation, the physician plays in important role in ensuring project integration and meeting goals. The physician may play an active role, identifying eligible patients and explaining the value of DESIs to them. The physician may also play a more passive, supportive role, communicating to staff the importance of facilitating, and checking in on progress, accommodating the integration of DESIs into the clinical workflow, and taking care not to interrupt patients reviewing a DESI. In more capable clinics, the physician played an active role in identifying eligible patients and in some cases encouraging them to review a DESI:

 2009 Blackwell Publishing Ltd Health Expectations, 14 (Suppl. 1), pp.73–84

Conducting implementation research in community-based primary care, D L Frosch, K J Singer and S Timmermans 81 Doctor always comments to patients that the videos are informative and will help them understand the screening. He encourages his patients to watch them whenever they are eligible. (Clinic #2) She informed me that tomorrow they will have a lot of patients who will qualify. She looked up the patients scheduled to come in tomorrow in her laptop. As she went through the list, she was able to tell me which patients would qualify based on age, whether they had a previous screening, and whether or not they have a history of prostate or colon cancer. She was also able to tell me which patients could communicate well...suggests that she knows her patients really well. (Clinic #5)

By contrast, in less capable clinics, the physicians showed little or no engagement in identifying eligible patients: Doctor was not at all concerned with project progress. He didnÕt ask about the project. He sometimes says IÕm the ÔcounselorÕ and will ask if I can help patients fill out forms like disability claim. (Clinic #9) [Doctor] also suggested that…the patients are not interested in patient education and are generally apathetic…He stressed that the patients do not view us as people who are interested in helping them… (Clinic #12)

In summary, we found that capable project integration required adequate clinic infrastructure, a relatively well-matched patient pool, and positive work and patient care environments. The remaining identified components, including staff facilitation and the physicianÕs role seemed to account for higher level differences between the clinics, acting as barriers and facilitators that separated capable clinics from those that were able to work completely independently, to a lesser extent, less capable clinics from those that did not meet the project goals. The former group might, therefore, be considered foundational features, while the latter set can be understood as moderating factors.

Discussion To date, the majority of research on DESIs has been conducted in academic medical practices.9,16 Our focus on primary care physicians in community settings arose from the growing

recognition by researchers that new interventions often do not work as well outside institutional and ⁄ or research contexts.20 Physicians practicing in academic settings are either in training or are teachers and ⁄ or researchers: thus, they may differ systematically from nonacademic practitioners in their individual-level propensity to adopt potentially practiceimproving changes. Broader implementation of DESIs in medical practice requires that we study their use in non-academic settings. Our practice recruitment rate of 6% was disappointingly low. There may be several reasons for this. A systematic review has examined factors associated with recruitment of community physicians into health services research projects. Working with influential physicians, who are well known in their professional community, to identify potential participants is generally associated with significantly higher participation rates than we achieved without such assistance. Other studies that worked without physician recruiters achieved similarly low participation rates to ours.21 Our low recruitment rate may also reflect physiciansÕ perception of the potential burden on clinical staff that would result from participation in our project.21 The reasons cited for not participating in our project were similar to those reported in published literature: lack of time, interest and resources.21 These are also commonly identified barriers in previous studies of DESI implementation, including specialty care settings.5,7,8,22 Others have also noted that although significant progress has been made in developing and evaluating DESIs, the Ôtipping pointÕ of widespread adoption in clinical practice has not yet been reached.4 The physicians who chose to participate in our project could therefore be viewed as Ôearly adoptersÕ who, by definition, constitute a minority of the physician population of interest.23 In that sense, these individuals may be ideal collaborators for reaching larger groups of community physicians for future projects. Overall, we found that foundational features, including the clinic infrastructure, the patient pool and the work and patient care environments determined whether integration of patients reviewing a DESI about cancer screening

 2009 Blackwell Publishing Ltd Health Expectations, 14 (Suppl. 1), pp.73–84

82 Conducting implementation research in community-based primary care, D L Frosch, K J Singer and S Timmermans

immediately prior to a consultation was viable. Interestingly, meeting the goals of Phase 2, which focused on a more complex video DESI, did not require significantly more consultations on average across practices. Higher level moderating factors then influenced how capable practices were in integrating DESIs into their workflow. Among these factors, staff facilitation was certainly important, but tended to work well in most cases where the foundational features were solid. The physicianÕs role, however, appeared to be the most important factor in determining a practiceÕs ability to integrate the project. This was due to the physicianÕs unique ability to integrate the project not only into the clinical workflow, but also into the patient care, as well the positive influence of the physicianÕs investment and motivation on both staff facilitation and patient interest. This critically important role of the physician is consistent with other studies of DESI implementation, both in outpatient specialty care and in hospital settings.5,7,8,22 Our data suggest that implementation of DESIs to be used immediately before a consultation is feasible if the practice infrastructure can provide sufficient basic accommodation and physician and staff are dedicated to patient care goals that are implicit in the use of these tools. But it is important to note that only two practices were able to work independently. Thus, our experiences raise doubts about whether the majority of these practices could independently implement a consultation-driven model of DESI provision to patients. While lack of physician and staff dedication to these goals was evident in several practices, the sheer number of competing demands also played a clear and important role. Most practices that participated in this project had limited resources and were vulnerable to increased demands on these caused by staff illness or turn over and other administrative disruptions, such as the implementation of a new Medicare billing system that was observed in one practice. A further challenge to the broader implementation of DESIs is that these efforts will compete for limited physician and staff attention with other quality improvement efforts and payfor-performance measures introduced by health

insurances.24 How practices might prioritize these and what incentives are necessary for successful DESI implementation remains unknown. There are several important limitations to our study. First, we focused on primary care practices in one metropolitan area. As such, our efforts may not reflect the potential barriers and facilitators to using DESIs in other cities and geographic regions. Our sampling of practices was neither random nor representative. Rather, we worked with all practices that expressed an interest and willingness to participate and indicated that they had (or believed they had) an adequate and appropriate patient pool for the cancer screening services that were the focus of our project. Ideally, we would have supplemented our ethnographic observational data with staff and physician interviews or questionnaires. Thus, our conclusions are limited by what was observable and more time was spent in practices that were less capable. Given the large number of practices we had to contact to recruit our pool of participating practices, these could be viewed as potential early adopters of the DESIs, as noted above. In that sense, the challenges we encountered suggest that the challenges to broader implementation of these tools are considerable. Nevertheless, it is important to note that preventive cancer screening services may not have sufficient priority in the face of many other competing clinical and administrative demands. More research on DESIs for other conditions commonly treated in primary care, e.g., chronic low back pain, is clearly needed. Finally, our study did not evaluate a randomized implementation intervention, but rather was exploratory. More research is clearly warranted to determine what factors would causally facilitate integration of DESIs at the time of a consultation in community-based primary care practices. It may not be realistic to expect communitybased primary care practices to provide DESIs to patients to review at the time of a consultation. Alternative implementation models are being tested, such as providing DESIs for patients to review at home.4 However, these models are not without challenges either. If a patient takes a program home, both physician

 2009 Blackwell Publishing Ltd Health Expectations, 14 (Suppl. 1), pp.73–84

Conducting implementation research in community-based primary care, D L Frosch, K J Singer and S Timmermans 83

and patient need to ensure that the decision in question is returned to in a subsequent consultation, a process sometimes referred to as Ôclosing the loopÕ. Whether or not this would be done consistently remains to be seen. A recent study reviewed potential policy measures that may be necessary to move medical practice to the tipping point where use of DESIs becomes routine.4 Our work identified a number of challenges that stand in the way of reaching this point in community-based primary care practices if a consultation-based model is to be implemented. Perhaps most important is the still limited physician buy-in to using these interventions observed in this study, which has also been documented in other published DESI implementation studies.5–8 Arguably, greater physician buy-in will require a change in medical culture that assigns greater importance to patient preferences in clinical decision making. Achieving this cultural change is a significant challenge facing researchers, health insurances and policy makers interested in giving patient preferences greater prominence in making choices about clinical services.

Source of funding

2

3

4

5

6

7

8

9

Supported by a grant from the Foundation for Informed Medical Decision Making, Boston, MA. 10

Conflict of interest Dr. Frosch serves as a consultant for the Foundation for Informed Medical Decision Making. 11

Acknowledgements Gratitude is expressed to Socorro Ochoa, Sandra Contreras, Naveen Dhawan, Carlos Lopez, Irma Ocegueda, and the physicians and clinical staff who participated in this project.

12

References 1 Frosch DL, Kaplan RM. Shared decision making in clinical medicine: past research and future directions.

13

American Journal of Preventive Medicine, 1999; 17: 285–294. Elwyn G, OÕConnor A, Stacey D et al. Developing a quality criteria framework for patient decision aids: online international Delphi consensus process. BMJ, 2006; 333: 417. OÕConnor AM, Bennett C, Stacey D et al. Do patient decision aids meet effectiveness criteria of the international patient decision aid standards collaboration? A systematic review and meta-analysis Medical Decision Making, 2007; 27: 554–574. OÕConnor AM, Wennberg JE, Legare F et al. Toward the Ôtipping pointÕ: decision aids and informed patient choice. Health Affairs (Project Hope), 2007; 26: 716–725. Holmes-Rovner M, Valade D, Orlowski C, Draus C, Nabozny-Valerio B, Keiser S. Implementing shared decision-making in routine practice: barriers and opportunities. Health Expectations, 2000; 3: 182–191. Le´gare´ F, Ratte´ S, Gravel K, Graham ID. Barriers and facilitators to implementing shared decisionmaking in clinical practice: update of a systematic review of health professionalsÕ perceptions. Patient Education and Counseling, 2008; 73: 526–535. Silvia KA, Sepucha KR. Decision aids in routine practice: lessons from the breast cancer initiative. Health Expectations, 2006; 9: 255–264. Silvia KA, Ozanne EM, Sepucha KR. Implementing breast cancer decision aids in community sites: barriers and resources. Health Expectations, 2008; 11: 46–53. Gattellari M, Ward JE. A community-based randomised controlled trial of three different educational resources for men about prostate cancer screening. Patient Education and Counseling, 2005; 57: 168–182. Protheroe J, Bower P, Chew-Graham C, Peters TJ, Fahey T. Effectiveness of a computerized decision aid in primary care on decision making and quality of life in menorrhagia: results of the MENTIP randomized controlled trial. Medical Decision Making, 2007; 27: 575–584. Wennberg JE, Fisher ES, Skinner JS. Geography and the debate over Medicare reform. Health Affairs (Project Hope), 2002; Suppl Web Exclusives: W96– W114. Available at: http://content.healthaffairs.org/ cgi/content/full/hlthaff.w2.96v1/DC1, accessed 1 December 2008. Liebhaber A, Grossman JM. Physicians moving to mid-sized, single-specialty practices. Center for Studying Health System Change Tracking Report, 2007. Available at: http://www.hschange.org/ CONTENT/941/941.pdf, accessed 1 December 2008. Chan EC, Haynes MC, OÕDonnell FT, Bachino C, Vernon SW. Cultural sensitivity and informed

 2009 Blackwell Publishing Ltd Health Expectations, 14 (Suppl. 1), pp.73–84

84 Conducting implementation research in community-based primary care, D L Frosch, K J Singer and S Timmermans

14

15

16

17

18

decision making about prostate cancer screening. Journal of Community Health, 2003; 28: 393–405. Frosch DL, Kaplan RM, Felitti V. The evaluation of two methods to facilitate shared decision making for men considering the prostate-specific antigen test. Journal of General Internal Medicine, 2001; 16: 391–398. Lewis CL, Brenner AT, Griffith JM, Pignone MP. The uptake and effect of a mailed multi-modal colon cancer screening intervention: a pilot controlled trial. Implementation Science: IS, 2008; 3: 32. Frosch DL, Legare F, Mangione CM. Using decision aids in community-based primary care: a theorydriven evaluation with ethnically diverse patients. Patient Education and Counseling, 2008; 73: 490–496. Emerson RM, Fretz RI, Shaw LL. Writing Ethnographic Fieldnotes. Chicago, IL: University of Chicago Press, 1995. Bryant A, Charmaz K. The Sage Handbook of Grounded Theory. London: SAGE Publications, 2007.

19 Strauss A, Corbin J. Basics of Qualitative Research. Newbury Park, CA: SAGE Publications, 1990. 20 Ruffin MT, Gorenflo DW. Interventions fail to increase cancer screening rates in community-based primary care practices. Preventive Medicine, 2004; 39: 435–440. 21 Asch S, Connor SE, Hamilton EG, Fox SA. Problems in recruiting community-based physicians for health services research. Journal of General Internal Medicine, 2000; 15: 591–599. 22 Stapleton H, Kirkham M, Thomas G. Qualitative study of evidence based leaflets in maternity care. BMJ, 2002; 324: 639. 23 Rogers EM. Diffusion of Innovations, 5th edn. New York, NY: Free Press, 2003. 24 Rosenthal MB, Landon BE, Howitt K, Song HR, Epstein AM. Climbing up the pay-for-performance learning curve: where are the early adopters now? Health Affairs (Project Hope), 2007; 26: 1674–1682.

 2009 Blackwell Publishing Ltd Health Expectations, 14 (Suppl. 1), pp.73–84