Because we were unable to locate a comprehensive conceptual model of clinicians' “orientation” toward integrative medicine, we used a grounded, qualitative method, using semi-structured interviews, to develop a conceptual model of integrative medicine (Bernard 2002
). We then developed a survey, based on the interview results, to measure this construct. Lastly, we administered the revised survey instrument to a convenience sample of conventional and CAM health care practitioners to assess its reliability and validity. The UCLA Institutional Review Board approved this study.
We used semi-structured interviews to elicit relational data across practitioners because the interviewer can deeply probe and openly explore practitioner attitudes and behaviors toward integrative medicine. This method was selected rather than conducting focus groups in order to maximize participant response variance (Bauman and Adair 1992
). In addition, focus groups suffer from lack of independence among participants and unequal participation. In contrast, the semi-structured interview is able to elicit complete and independent data from all participants.
In-depth interviews were conducted using a standard interview guide (Bauman and Adair 1992
), consisting of open-ended questions with probes for clarification and additional detail. Following the layout of a “funnel” interview, the conversation started with broad topics and then probed each response. The interview began with the standard “grand tour” (Bernard 2002
) question by asking: “What does integrative medicine mean to you?” If, for example, the respondent answered that integrative medicine is defined by using the best of traditional Chinese medicine and western medicine, then the subject was asked more focused questions, such as “What types of practitioners would provide this care? How would these two types of medicine be mixed?” The interviewer continued to ask narrower questions until the informant exhausted all responses or the topic changed.
We conducted semi-structured interviews with a purposive sample of 50 health care practitioners: 13 physicians, 13 physician acupuncturists (physicians who also completed acupuncture training), 12 chiropractors, and 12 acupuncturists. Two investigators (A. H. and G. R.) reviewed the 50 transcripts and agreed that theoretical saturation was reached. Equal distribution across clinician groups aimed to yield maximal variance of provider attitudes and behaviors concerning integrative medicine. The mean duration of each interview was 30 minutes (range 20–55 minutes). Interview audiotapes were transcribed. The validity of the transcription process was assessed by reviewing the match of audiotape to transcript.
Free Pile Sorting
To identify key dimensions of integrative medicine, we used an exploratory technique (Lincoln and Guba 1985
) in which transcripts were read to uncover “core statements” that represent key constructs of a practitioner's orientation toward integrative medicine. We printed these core statements onto cards. All core statements were laid on a large table. Three investigators (A. H., G. R., and N. W.) free pile sorted the core statements into similar groups or themes; this initial pile sorting was followed by a group discussion to produce consensus. After multiple iterations of free pile sorting with these pairs of researchers, we found four domains and 11 subdomains representing the key dimensions of clinicians' orientation toward integrative medicine.
A Conceptual Model of Integrative Medicine at the Provider Level
Our qualitative analysis uncovered four key domains of integrative medicine: provider attitudes, knowledge, referral, and practice. Provider training and practice setting also emerged as important factors in determining clinicians' orientation toward integrative medicine. Based on the relationships among provider attitudes, provider knowledge, provider behaviors, and other factors, we developed a conceptual model of integrative medicine at the provider level that was grounded in the interview results ().
Model of Integrative Medical Care at the Provider Level
We developed a 63-item initial version of the survey to measure clinicians' orientation toward integrative medicine based on the key domains of integrative medicine. Survey items were constructed directly from phrasing found in the core statements. Response categories to survey items assessing clinicians' attitudes toward integrative medicine consisted of a 4-point categorical response scale (strongly agree, somewhat agree, somewhat disagree, or strongly disagree). Response categories to survey items about clinicians' practice of integrative medicine consisted of a 5-point categorical response scale reflecting the frequency of the behavior (never, rarely, sometimes, often, or always). Items asking about sources of education had response options on a 5-point scale, ranging from “more than once a week” to “never.”
Survey items were evaluated by conducting cognitive interviews with 10 practitioners who had not participated in the semi-structured interview. This technique is designed to increase the quality of self-reported data obtained through questionnaires by querying respondents about the meaning of questions and using feedback to refine item wording. Respondents were encouraged to verbalize their thought processes aloud as they comprehended and responded to survey items. Information obtained from these interviews also identified skip pattern inconsistencies and difficulties with question order. Based on the cognitive interviews, we revised the items and produced a 56-item version for the field test.
Field Test of Survey
The revised 56-item version of the survey, supplemented with questions about demographics, practice setting, and training, was mailed to a convenience sample of 294 practitioners. This sample included the 50 practitioners who had participated in the interviews plus an additional 71 primary care physicians, 58 physician acupuncturists, 55 chiropractors, and 60 acupuncturists. The survey was accompanied by a letter of introduction, a $10 honorarium, and an information sheet. Participants provided consent to participate by completing the survey instrument.
We used multitrait scaling analysis to evaluate the field test items (Hays and Hyashi 1990
). In multitrait scaling analysis, item-scale correlations are examined to evaluate whether each item correlates more highly with its hypothesized scale (corrected for item overlap with the scale) than with the other scales. We eliminated items with low correlations with their hypothesized scale (domain) and combined scales with poor item discrimination. Specifically, we eliminated four items with correlations less than 0.30 with their hypothesized scales (Stewart, Hays, and Ware 1992
), eight redundant items that overlapped with items having higher item-scale correlations, and 14 items that were correlated weakly with all scales. Several scales were collapsed because of weak item discrimination. Items derived from the pile sorting domains of Comanagement and Practice style were collapsed into a single “Readiness to refer patients” subscale; items from two domains focusing on proficiency and training were collapsed into a “Learning from alternate paradigms” subscale.
The final instrument consisted of 30 items comprising five scales: awareness and openness to working with practitioners from other paradigms (10 items), readiness to refer patients to other paradigms (7 items), learning from alternate paradigms (5 items), patient-centered care (3 items), and safety of integrative medicine (5 items). We conducted all subsequent psychometric analyses on the 30-item (IM-30) survey.
Mean scores for the IM-30 as well as for each subscale were transformed linearly to a possible range of 0–100, with higher scores indicative of greater orientation toward integrative medicine. We calculated the mean, median, standard deviation, skewness of scale, and percentage of participants scoring the minimum (floor) and maximum (ceiling) for each item and scale. Internal consistency reliability was estimated for each subscale and for the 30-item scale using Cronbach's coefficient α
). For each subscale, we evaluated item discrimination (extent to which items correlate most with the scale they are designed to measure) by calculating the percentage of times that items in the subscale correlated significantly higher (at least two standard errors higher) with the hypothesized subscale (correcting for overlap) compared with other scales (Hays and Hyashi 1990
). Analyses were conducted using SAS 8.0
(Cary, NC, USA) and STATA 8.0
(College Station, TX, USA).
To assess construct validity, we examined the associations of the IM-30 score with items assessing dual training, practice setting, practitioner type, and self-perceived “integrativeness.” We hypothesized that practitioners who had received formal training in at least two healing paradigms would score higher on the IM-30 than “single-trained” practitioners. These dual-trained practitioners include physician acupuncturists (e.g., Western medicine and traditional Chinese medicine) and CAM practitioners (e.g., chiropractic and traditional Chinese medicine). We also hypothesized that practitioners who work in an integrative medicine setting would score higher on the overall IM-30 scale than those who work in a solely conventional or CAM setting. Moreover, we hypothesized that chiropractors and acupuncturists would score higher on the IM-30 total scale than physicians because prior work has shown that they are more open and ready to refer to practitioners from other medical paradigms (Coulter 1991
; Barnes 2003
). Lastly, we hypothesized that practitioners who self-perceived a higher level of “integrativeness” would score higher on the IM-30 scale compared with practitioners who rated themselves as less integrative.