We used a qualitative research approach to develop in-depth understanding about perspectives of physicians [20,21]. Face-to-face, semistructured, in-depth interviews were conducted with family physicians to elicit their perceptions of and experiences with the computer-assisted HRA. As all potential participants worked at the same clinic, individual interviews were preferred over focus groups. This resulted in scheduling of the interviews at times convenient to each participant and ensured participants’ confidentiality [22,23]. Physicians were eligible to participate if they had seen at least five patients who had participated in the randomized trial. This purposeful sampling allowed information-richness in relation to the studied phenomenon [24,25]. For an exploratory study with a homogeneous sample, five to eight participants are generally considered sufficient [26,27]. We considered our sample homogeneous because all participants were physicians working at the same clinic, and all had used the computer-assisted health-risk assessment tool. The study procedures were approved by the ethics review boards of the University of Toronto and the hospital with which the family practice clinic where the study took place was affiliated.

Ten eligible physicians (seven females and three males) participated in the interviews which were conducted between October and November of 2005. The average age of participants was 46 years (range 32-64 years). Participants had been in practice for 16 years on average (range 1-30 years) and reported practicing 30 to 50 hours per week. At the trial site, physicians’ weekly number of hours ranged from 16 to 40. Eight of the physicians reported seeing female patients at 50% or more of visits.

All interviews were conducted by the first author at a place and time convenient to the physician.

Participant physicians provided written consent and completed a one-page demographic questionnaire before the interview. No monetary incentives were offered. The interviewer used the principle of “ask, wait, and probe” and a semistructured interview guide with open-ended questions [23]. All interviews were audio recorded and field notes were taken.

The interview guide was constructed jointly by the research team and clinical collaborators (a family physician, a nurse, and a social worker) from the study site. This was informed by our literature review on the modes of inquiry for psychosocial health risks [28-33]. We identified dual barriers for the face-to-face encounter of physicians and patients. Patient barriers included feelings of embarrassment, fear of physician’s rejection or reaction, concerns about confidentiality, and lack of physician’s direct inquiry [34-36]. Physician barriers included discomfort, fear of patient’s negative reactions, lack of time, priority of the acute problem, and lack of familiarity with resources [10-13]. At the same time, computer-assisted HRA was identified as having potential to address many of these barriers on the patient (eg, desire of “direct questioning” by provider in a nonjudgemental manner) and provider side (eg, time efficiency, referral information, and anonymity). This informed the development of topic areas and probes for the interview guide (eg, sensitivity of certain health risks and visit time).

The use of open-ended questions and probes in the guide allowed defining the research area without restraining the expressions of participants [22,23]. The guide was revised after the first two interviews and included four key questions: (1) What do you think of your experience with the computer-assisted HRA? (2) How would you describe its potential across various risks and visits? (3) Would you recommend such computer-assisted HRA in a family practice setting? and (4) What factors are important for its implementation in a family practice setting?

The interviews with the physicians were taped, transcribed verbatim, and from the transcriptions, Word files were prepared. The techniques of analytic induction and constant comparison were used to code and analyze the transcripts [37]. Analytic induction, originating from Znaniecki’s work [38], entails the development of concepts and testing of propositions from the data by a systematic examination of the similarities between various social phenomena (eg, acceptance or rejection) and processes (eg, agent-agent and agent-structure relations) while emphasizing the research context and negative cases that challenge an emerging finding and thus lead to analytical refinement [39,40]. Also, the analytical method of constant comparison, derived from the grounded theory approach [41], was used within and between interviews [42]. Informant statements were compared for thematic and/or conceptual similarities within and between interviews (eg, What do these quotes have in common? What is unique?)

Coding involved collating and analyzing all of the transcribed data related to the emerging themes and concepts informed by our literature review. An initial coding scheme was developed jointly by the research team after all members had read the collected information [43,44]. This initial coding scheme was then applied to the text data by assigning symbols to represent each category. The sorted and coded data was then read and read again by the first and last author to refine the analysis and to control researcher bias [37]. For rigor, attention was directed to the range and diversity of experiences, meanings, and perceptions along with a search for disconfirming evidence. For example: What does this tell about how participants developed their perspectives? How do these relate to each other? How are emerging relationships confirmed or disconfirmed within data?

All participating physicians emphasized the on going nature of their relationship with patients. This prefaced both positive and negative perceptions of physicians with respect to the computer-assisted HRA tool. For interpretation of this two-sided aspect of the tool, we used the lens of social exchange theory [45,46]. This theory has been applied at both micro and macro levels to explore the provider-patient relationship [47-49]. According to this theory, health services are provided through an exchange process in the physician-patient relationship. Patients need the physician’s knowledge and competence to restore or maintain their health. Physicians need patients for income, new patient referrals, and the less tangible rewards of compliance, praise, and appreciation. Patients’ relative dependency on physicians is influenced by the degree of their trust in physicians while increasing trust in return enhances the social responsibility of physicians. Limited successes in social exchanges give rise to negative emotions (eg, self-depreciation, guilt, anger, or discomfort with others), which act as internal stimuli to avoid recurrence of these emotions. Therefore, a tool which can change the nature of social exchange between physician and patient and/or physician and institution, such as computer-assisted HRA, can be perceived as helpful or not helpful depending on how it influences the exchange and provokes negative or positive emotions. The inclusion of the theoretical lens (ie, social exchange theory in this study) at the time of interpretation augments rigor and trustworthiness of the findings [50]. We also incorporated the technique of peer audit by sharing our findings with the site collaborators whose feedback refined our interpretation, adding credibility [37,50].

Benefits of the computer-assisted HRA emerged as a dominant theme across all physician interviews and included two subthemes. The tool was unanimously perceived to open dialogue on psychosocial health risks. The HRA or screening was also perceived to serve a general facilitative role in providing patient care in a busy clinic environment. We discuss each subtheme and its subcategories (also see Table 1).

Physicians discussed some concerns or challenges in relation to the computer-assisted HRA. Three subthemes emerged: scope of new risk information, patient readiness, and length of visit. Participants expressed both positive and negative stances (Table 2).

The overall feasibility of the computer-assisted HRA in a family practice setting emerged in physician discussions. Three subthemes characterized concerns about feasibility: general acceptance of the tool; the tool’s fit with the visit; and availability of clinical and organizational resources for its implementation (Table 3).

Perceived benefits of the computer-assisted HRA for psychosocial health risks emerged as a dominant, crosscutting theme. Physicians felt the tool improved patient disclosure and physician detection of psychosocial health risks, consistent with recent effectiveness studies [6-9]. Participants also discussed possible underlying mechanisms of this positive aspect. Physicians attributed improved patient disclosure to the tool’s specific inquiry about these risks and its anonymity in conjunction with the patients’ time for self-reflection. In other words, physicians felt that patients were empowered to disclose and discuss their socially sensitive risks with a higher level of comfort and confidence. This is in accordance with existing research on modes of inquiry [28-31] and “activated patients” [52-55], who receive pre-visit interventions (eg, education about their health risks) to become knowledgeable and willing to discuss their risk status. These activated patients become effective “prompts” for the medical providers, leading to the provision of health preventive and promoting care in medical visits. We have also conducted qualitative interviews with the participant patients to understand their enablement and empowerment; this work is in progress.

In our study, physicians attributed their enhanced detection of patients’ psychosocial risks to the consistency and comprehensiveness in risk screening provided by the tool. These findings imply that computer-assisted HRA enhanced the patient-centeredness of the physician-patient interaction [56], an area that needs further exploration. In brief, future use of the tool in a family practice setting could benefit a large number of patients seeking care for multiple health reasons.

Some concerns or challenges also emerged in relation to the newness of the risk information generated by the computer program, patient readiness to discuss the reported risks, and the increase in visit time. Few physicians discussed the dilemma of providing adequate care to one versus many patients within limited time. Some physicians proactively managed these challenges by assessing a patient’s willingness to discuss the risks, taking notes, and/or setting up follow-up visits. Indeed, strategic management of the perceived barriers is possible in future applications of the tool. For example, physicians should receive comprehensive training on the varying forms of “clinical success” for the management of chronic psychosocial issues so that they don’t feel frustrated when dealing with these complex cases in the time constrains of a medical visit. The training program should emphasize various stages of patient-readiness to take an action [57,58] along with motivational interviewing techniques [59,60] and the physician’s gatekeeper role of making referrals to other services. Management of psychosocial issues often requires diverse health and social resources [61,62], and physicians need not be the sole providers of care. Also, multidisciplinary models of care and the incentive of billing codes for counseling hold potential to address physicians’ time concerns.

Physicians perceived the success of computer mediated patient interactions in light of the perceived benefits and concerns or challenges of this type of interaction. Interactions were perceived as successful when patients shared health risk information (they disclosed and were ready to talk) and when the information was new to the physicians and led to the provision of care, provided time was available. The interactions were perceived as partially successful when patients disclosed but were not ready to talk, and/or physicians did not have enough time to adequately deal with the reported risks. Based on Lawler’s social exchange theory and related research [63,64], exchanges with low success generate negative emotions, such as frustration leading to low self-efficacy. These internal stimuli in turn lead to motivations to avoid recurrence of negative feelings, consistent with the social cognitive theory [65,66]. Accordingly, physicians in our study sought to reduce the likelihood of partially successful exchanges. Because of the preventive focus of computer-assisted HRA, physicians recommended the use of this tool for periodic health exams and/or follow-up visits where time and patient-readiness were not seen as undermining factors. Also, prevention is a built-in focus of periodic health exams. It seems that system level supportive mechanisms are needed to enhance physicians’ confidence in their ability to manage psychosocial health risks [67].

Physicians discussed the need for clinic and organization-related resources to implement the use of computer-assisted HRA in the future. At the practice level, physicians emphasized the management of patient flow through early appointments and staff training. Patient privacy and confidentiality were viewed as important for completion of the computer surveys, but lack of space and technological skills were considered logistical limitations. These findings provide practical insights for future initiatives on computer-assisted HRA in a family practice setting and have theoretical implications for advancing understanding about diffusion of innovations [51,68]. At the organization level, physicians pointed toward the need for greater financial support or mitigation of time investment. Indeed, institutional prioritization is salient for health-promotion and disease-prevention orientation in medical settings [69]. Health care service institutions could incorporate effective computer-assisted HRA tools as part of their quality initiatives because of their potential to detect socially-sensitive health risks (eg, poor mental health, partner violence, or substance abuse) in a timely and efficient manner. This could be especially beneficial in assisting vulnerable populations who are exposed to higher risks of psychosocial issues. Thus, from the population health perspective, system level adoption of such programs could play an important role in addressing health inequities.

Participant physicians also wished for the merging of clinical practice guidelines, institutional goals, and local patient needs in the identification of health risks for assessment or screening. This reflects not only physicians’ multiple roles at various levels of health care (individual, institutional, public health) but their desire to have coherence within policies. Lack of consistency in policies across health sectors is often reported as a barrier to timely screening practices [70]. There is a need to actively involve clinicians, public health experts, health care administrators, and policy makers to establish locally tailored coherent screening guidelines.

Future direction for implementation should draw from the tenets of diffusion of innovation theory [51]. For example, early efforts could focus on settings with characteristics of enthusiastic “early adopters” who use the data on an innovation to make their own careful adoption decisions. Their success then creates a domino effect where a larger majority adopts the innovation at a pace quicker than the average.

The emerging eHealth tools can contribute to new models of care by linking clinic care and self care. For example, computer-assisted HRA (augmenting clinic care) could be offered in conjunction with “virtual clinics” and “e-messaging” to patients, supporting self care [71-73]. This evolving area holds potential to improve timely access to health care with fewer errors, leading to patient empowerment and cost savings.

Some limitations in the design of our study warrant caution for the interpretation of the results. Participant physicians who volunteered in a randomized controlled trial of the computer-assisted HRA practiced at an inner-city, hospital-affiliated, academic family practice clinic. The views of these physicians may not represent the views of physicians practicing at different sites. The qualitative nature of the study may limit the applicability of results to wider clinical settings. We used several strategies for rigor and trustworthiness (as described in the methods section), such as a jointly agreed upon initial coding-scheme and a theoretical lens along with peer audit for the interpretation. This increases our confidence in the transferability of findings. The data were collected in 2005, but we don’t anticipate much change in the studied population. Perhaps patients today are more likely to be acquainted with computer-assisted tools in medical settings or electronic health records than they were five years ago.

Participant physicians perceived computer-assisted HRA as a useful tool in family practice, particularly for the detection and discussion of psychosocial health risks. Physicians displayed a general acceptance of the computer HRA tool and indicated its greater feasibility for periodic health exams and/or follow-up visits than for all visits. Physician training on psychosocial issues should address physicians’ concerns about patient readiness and visit time by emphasizing the varying forms of “clinical success” for the management of chronic and complex psychosocial issues. Future research is needed to examine the best ways to implement this program in diverse clinical settings and patient populations. From a public health perspective, computer-based HRA in a family practice could mean timely psychosocial risk identification and access to care for many people.