Expert decision making in physical therapy - a survey of practitioners.

Physical therapists are assuming increasing independence in making patient care decisions. Understanding the dimensions of expert decision making will help current practitioners improve their skills and educators prepare students more effectively. It seemed appropriate to begin an investigation of expert behavior by asking the experts themselves what they do or believe they do. There has been considerable research into clinical decision making, information-processing strategies, and differences between expert and novice behaviors in the past two decades. Research has been done in many fields, including medicine, education, nursing, and physical therapy, and the findings generalize across areas of study. Decision making is influenced by knowledge, the way experience has structured that knowledge, the type or format of the decision task, the limitations in human information processing, and the social and contextual elements of the decision.(1-14) Decision making has generally been found to include (1) the use of critical cues or forceful features for promoting the recognition of specific clinical patterns and (2) the early generation of hypotheses for organizing the acquisition and interpretation of information.(2,4,5) Preliminary evidence suggests that physical therapists use similar reasoning processes.(12-17) CUrrent research does not support the popular belief that clinicians collect a complete, routine database before deciding about the patient's problems. Experts, when compared with novices in the same field, exhibit a superior structuring of knowledge into clinically relevant patterns that are unlocked by key cues in the decision environment. Patterns stored in memory enable the expert to recognize meaningful relationships and generate likely hypotheses.(7,9,10,14,17) Recently published models(18-20) attempt to help practitioners organize key pathological concepts as a guide to decision making. Psychological research indicates that human information processing is subject to bias introduced by the presentation of the task and by strategies used for selective attention and interpretation of the environment.(3) If the patient referral, for example, contains a specific diagnosis, the diagnosis has been shown to be a biasing factor in both medicine in England(2l) and physical therapy in Australia.(22,23)

Cognitive style, which can be defined as an individual's preferred way of thinking and organizing information, has also been studied for its effect on decision making. McKenney and Keen(24) developed a paradigm of cognitive style that was used by Bork(12) in a study of cognitive style influences on decision making by physical therapy students. McKenney and Keen's(24) paradigm reflected four styles of cognition, two related to the data gathering phase and two to the information-processing phase of decision making. The data-gathering styles were defined as (1) receptive, a style generally characterized by suspending judgment until all possible data have been collected, paying attention to detail, and attending to the implications of each piece of data individually, and (2) preceptive, a style characterized by moving from one section to another, seeking and responding to cues and patterns as a guide to data gathering. The information-processing styles were defined as (1) systematic, characterized by a consciously methodical approach, defining problems and constraints early, performing an ordered search for information, and completing one step before progressing to the next, and (2) intuitive, characterized by keeping the total problem in mind and considering alternatives simultaneously. The intuitive person may move from one thing to another, relying on cues and hunches. The results of Bork's(12) study of physical therapy students suggested that cognitive style influenced clinical evaluation performance. The ability to operate in the intuitive mode was associated with a better performance in history taking and physical assessment, whereas students who operated primarily in the preceptive mode were less likely to accurately determine a simulated patient's problems. In nursing, Hayes-Roth and HayesRoth(25) suggested that systematic decision making might be effective in solving simple problems but that opportunistic decision making, that is, responding to the stimuli as they occurred, might be more effective in complex situations. Few studies have specifically examined the relationship of cognitive style preference to decision making.

A study of expert clinicians in the United States and Australia was undertaken as the first stage in the process of describing expert behaviors. The study was designed as a preliminary investigation into the nature of datagathering and information-processing phases of expert clinical decision making. A second purpose was to compare the reported decisionmaking processes of therapists in a country with direct client access to physical therapy (Australia) with the reported decision-making processes of therapists in a country with limited opportunity for direct client access (United States). Specific questions addressed by the study were: 1. What are important information

sources for expert physical therapists

in the United States and

Australia? 2. Is there a particular cognitive style

preference among expert physical

therapy practitioners in each

country, 3. Is cognitive style preference influenced

by country of practice, sex,

or major practice area? Method Subjects Subjects were selected through a nomination process. Elected national, state, section, or special interest group officers holding comparable positions in both countries were asked to nominate individuals whom they considered to be expert clinicians and who were involved in direct patient care activities at least 25% of the time. Over 800 nominations, which included 700 individual names, were received from US officers. All individuals named more than once and a random sample of the remaining nominees were used to obtain a sample of 404 individuals. In Australia, over 500 nominations, which included 384 individual names, were received and became the group selected to receive the questionnaire.

Procedures

Survey instrument development. We developed an instrument to gather data on expert physical therapists' preferred sources of information and to measure the experts' selfperception of their decision-making behaviors, focusing on cognitive style preference. The instrument evolved from a multistep process that included interviews, categorization of the interview statements and scoring procedure, pilot testing, and final construction of the instrument. interviews. We tape-recorded interviews with eight physical therapists working in Georgia in different practice settings. Interviews were conducted by both researchers, with one doing the questioning and the other making notes and monitoring the tape recorder. The purpose of the interviews was to develop a set of "real-world" statements reflecting the characteristics of the different cognitive styles described by McKenney and Keen.(24) Respondents were first asked to describe their clinical decision making processes and then asked to recall specific situations that illustrated simple decision making and difficult decision making. An unstructured format was used to provide respondents with the opportunity to describe their decision-making processes in their own words. Categorization. We then screened the interview statements, identifying some that matched the cognitive style descriptions and others that did not fit. The latter appeared to be statements of affect or belief (eg, statements 63 and 67 of the Appendix) or statements about specific knowledge requirements for decision making (eg, statement 55 of the Appendix). To offset the limitations of the small and geographically discrete interview sample, we added statements gathered from our collective experience. The first draft of the survey contained six logically derived components: one for each of the four cognitive styles, one defined as affect, and another defined as knowledge.

Scoring. We developed individual scores by averaging the responses to items within each category (affect, knowledge, receptive, preceptive, systematic, and intuitive).

Pilot test A sample of 21 practicing physical therapists in the United States and 20 physical therapists in Australia from different practice settings was used to evaluate the pilot survey instrument, which contained three sections. The first section requested demographic information, such as sex and years since graduation. The second section asked respondents to indicate the value of sources of information, using a four-point numerical scale ranging from very valuable" to "of no value," with a fifth point if the information was not available. The third section required responses to 55 statements on a four-point scale ranging from "strongly agree" to "strongly disagree." Respondents were also asked to comment on the clarity of the items.

We performed a reliability analysis for each category, followed by a factor analysis. We eliminated some statements and moved others to different categories if the factor analysis indicated a better fit and if the statement had face validity in the new category. We retained 48 items, which were randomized for inclusion in the final survey instrument. At the end of this phase, the alpha value for each category was above .6, with one category (systematic) above .7.

We decided to proceed with the study, but to perform more reliability and factor analyses before analyzing the final data. Comments on clarity were used to revise all three sections. The final survey instrument was similar in structure to the pilot questionnaire and is depicted in the Appendix.

Data Analysis Responses to the survey instrument were coded as indicated on the questionnaire. Data analysis included the following.

Demographics. Frequencies were calculated to provide a description of the two samples. The years since graduation were collapsed into three major categories for general descriptive purposes (ie, 0-10, 11-20, 20+), but five categories (ie, 0-5, 6-10, 1115, 16-20, and 20+), were retained for analyses of variance (ANOVAS) related to cognitive style. Places of employment included private practice, hospitals, rehabilitation centers, and other areas, as listed in Section I of the questionnaire Appendix). We identified six major practice-area groups for the purpose of analysis. General orthopedics, manual therapy, and sports physical therapy were combined to form an orthopedics group; adult and pediatric neurology were combined to form a neurology group. General practice, geriatrics, and cardiopulmonary physical therapy remained as initially established. The sixth group, education, comprised physical therapy educators and practitioners whose professional role was primarily patient education (ie, childbirth educators and consultants to industry).

Sources of information. Frequencies were computed for each source of information by country, and frequency tables, cross-tabulated with place of employment and major practice area, were then generated. Chisquare analysis was not performed because the numbers in some practice groups were extremely small and we preferred to retain qualitative differences at this level.

Cognitive style preference. we reevaluated category reliability for the combined sample and for each country separately. Each category was scored by calculating the mean score of the items in the category. The scores of subjects who responded to fewer than 75% of the items in a category were dropped from the calculations of that category. The scores of subjects who responded to 75% or more, but less than 100%, of the items were calculated as the mean of the items answered. We then performed a principal components analysis using the parallel-analysis method to determine whether the instrument was actually measuring different factors.(26) Parallel analysis is reported to be the most consistently accurate method for determining the number of major components to retain.(27) Analysis of variance. Before performing the ANOVAS, open responses ("other") were reviewed individually and either assigned to another response or dropped from the analysis for that variable. The two national samples were analyzed separately using a one-way ANOVA to identify within-nation differences in cognitive style attributable to sex, years since graduation, place of employment, or major practice area. The Tukey's Honestly Significant Difference (HSD) Test procedure was used to identify differences between groups at the significance level of .05. For each category of cognitive style, two-way ANOVAs were used to compare the means of the Australian and American therapists by sex, years since graduation, place of employment, and major practice area.

Results

We examined the data for a response set effect within the two countries. There was greater variance in the Australian therapists' responses, indicating greater heterogeneity in the sample, as compared with the American therapists' responses.

Demographics

Usable responses were received from 348 (86%) of the American nominees and from 290 (76%) of the Australian nominees. Fifty-six percent of the American respondents and 76% of the Australian respondents were female. Figure 1 outlines the experience characteristics of each sample, showing years since graduation and percentage of time currently spent in direct patient care. Table I depicts the distribution of the respondents across major practice areas for each country and indicates that the two national samples are comparable in composition (except for the education group, in which childbirth educators and consultants to industry were mainly represented in the Australian therapist sample). Approximately half of both samples were employed in private practice (US therapists, 49%; Australian therapists, 52%). Hospitals employed 29% of the US therapists, compared with 18% of the Australian therapists, and rehabilitation centers employed 11% of the US therapists, compared with 21% of the Australian therapists. Of the remaining places of employment, 5% of the US therapists worked in a school system, 4% in community care, and 2% in physical therapy education; 7% of the Australian therapists worked in community/ day care centers and as consultants to industry and 2% in physical therapy education. Different practice characteristics and health care structures made the samples less comparable in terms of this variable.

Sources of information

There was a great deal of similarity in the value placed on various sources of information between the two countries. Not surprisingly, therapists in both countries and in all types of practices valued their own assessment more than any other source of information (US therapists, 88%; Australian therapists, 89%).

Physicians' orders were generally considered of limited value by therapists in both countries (US therapists, 58%; Australian therapists, 53%). A greater percentage of Australian therapists compared with American therapists reported that physicians' orders were not available (12% versus 4%, respectively) Fig. 2), because referral is not required for treatment in Australia. The value of physicians' orders varied with area of practice; more Australian therapists (60%) involved in cardiopulmonary care, for example, valued physicians' orders than did American therapists (46%). The percentage of Australian therapists working in orthopedics and neurology who valued physicians' orders (28% and 25%, respectively) was considerably lower than was that of American therapists (46% and 45%, respectively), and more Australian therapists than American therapists reported the nonavailability of physicians' orders.

The reported value of direct communication with the physician is illustrated in Figure 3. Most Australian therapists found direct communications with the physician of value (56%-80%); the responses from American therapists were similar (58%-68%), except for therapists working in geriatrics. in geriatrics, the majority of American therapists found direct communications with the physician of limited value (53%).

The reported value of other sources of information was somewhat practice- and employment-specific. Most respondents valued the patient's past medical history (75% overall); however, 20% of the therapists in private practice and 14% in home health care did not have it available. Overall, 26% of American therapists and 57% of Australian therapists found radiographs a valuable source of information. Sixty-seven percent of American cardiopulmonary therapists, but only 34% of the orthopedic therapists, found radiographs they read themselves quite valuable.

Principal Components Analysis Seven factors were identified. We compared the items in each factor for congruence with the logical categorization that we had previously imposed. Although the reliability of the cognitive style categories had been satisfactory in the pretest, regrouping some items and eliminating others in accordance with the principal components analysis strengthened the statistical basis for the survey instrument without altering the logical premises on which it was based. We retained the four categories of cognitive style (receptive, preceptive, systematic, and intuitive) and identified two other categories (physician dependency and holism), which will not be reported in this study. The seventh factor identified minor components, including some universal value statements, and was not retained. Our final categories, their alpha values, and the related instrument items for both national samples are shown in Table 2.

Analyses of Variance The one-way ANOVA indicated that cognitive style preferences were not influenced by years since graduation, except for Australian therapists graduated for 5 years or less, who responded significantly less positively to the receptive category (F=3.51, df=4, P<.0l). The one-way ANOVA also revealed differences related to place of employment. Private practitioners in both the United States and Australia identified significantly less positively with the preceptive style than those employed in hospitals. The private practitioner group also responded significantly more positively to the systematic style and less positively to the intuitive style than those employed in rehabilitation centers. Other comparisons were not significant. Effects of major practice areas within countries. Differences in means of the five major practice groups in both countries are illustrated in Figure 4. The education group is not reported in detail because its composition was not comparable between countries. The Tukey's HSD Test procedure identified the following groups to be significantly different at the .05 level. Some of the differences were shared between countries, whereas others were country-specific. In the United States, the cardiopulmonary physical therapy group responded more positively to the preceptive style than the orthopedics and geriatrics groups, the orthopedics group responded more positively to the systematic style than the general practice and cardiopulmonary physical therapy groups, and the general practice group responded more positively to the systematic style than the neurology group. In Australia, the general practice group responded more positively to the preceptive style than the orthopedics group; the neurology group responded more positively to the intuitive style than the orthopedics group, although the mean response was not in the positive range; and the orthopedics group responded more positively to the systematic style than the geriatrics group. In the combined sample, the orthopedics group responded more positively to the systematic style than the neurology and general practice groups, and the neurology group responded more positively to the preceptive style than the orthopedics group. Results of the two-way ANOVAs are reported in Tables 3 and 4. There were no significant differences for country or sex for receptive data gathering or between countries for systematic information processing. Female physical therapists in both countries reported significantly greater identification with the preceptive mode of data gathering than male physical therapists, and the US sample overall responded significantly more positively to this category than the Australian sample. Male physical therapists in both countries responded significantly more positively to the systematic mode of information processing than did female physical therapists. American female physical therapists responded positively to the intuitive category, but the two-way interaction was not significant. American physical therapists overall had a greater affinity for the intuitive category than the Australian physical therapists.

Analysis by major practice area showed significant differences in all categories. The Australian orthopedics group responded least positively to the preceptive category, but the interactive effect was not significant. The neurology groups in both countries did not respond positively to the systematic approach, nor did the geriatrics group in Australia. Major practice area was significant for the receptive category, with the orthopedics and cardiopulmonary physical therapy groups in both countries responding more positively to this style than the geriatrics and education groups. Country effects remained the strongest predictor of identification with the intuitive style. The Australian neurology group had the most positive identification with the intuitive style; an interactive effect was also noted for this group.

Discussion

Sources of information Physicians'orders. More than half of the therapists in both countries considered the physicians' orders of limited value, which may reflect the therapists' levels of independence or the content of the referral letters. We expected that a group of expert clinicians capable of specifying patients' problems (diagnoses) and making independent treatment decisions would find physicians' orders of limited value; however, it would be interesting to determine what therapists expect from the referral. In Australia, Twomey(28), suggested that the patients' radiographs, the results of special tests, and a request for physical therapy were all therapists required when treating patients with vertebral problems or similar nonspecific diagnoses. Our results are congruent with studies in Australia(22) and Canada(29) that have demonstrated dissatisfaction with the information received from medical practitioners about the patients' medication and overall health status. Value of direct communication. The value placed on direct communication with the physician was not surprising. Considering that Australian physical therapists have direct access, however, it is worthy to note that direct communication with physicians is equally, if not more, important to them as compared with American physical therapists. The higher value placed on direct communications by the Australian therapists in cardiopulmonary care may reflect a more direct involvement with acute respiratory care. Dennis22 reported that, despite direct access, the majority of patients (67.2%) came to physical therapists in private practice in Victoria (Australia) via physician referral and that clinicians reported varied strategies to maintain and strengthen direct communications. The clinicians also wanted to educate physicians about the skills and values of the physical therapist.

Other sources of information. The much higher values reported by the cardiopulmonary physical therapy groups in both countries may reflect the use of radiographs for treatment decisions, whereas the orthopedics groups may use radiographs to rule out diagnostic alternatives. Therapists in both countries showed considerable flexibility in using sources of information based on availability. Cognitive Style Preferences Overall. The total sample's identification with receptive data gathering and systematic information processing suggests a response set based on valuing the scientific approach and believing it is appropriate in physical therapy evaluation, but it may also reflect actual practice. Positive response was strongest for the receptive category. Statistically, the most strongly positive items in the subscale were collecting information to confirm findings, checking out ideas, and gradually building a picture of the patient's problems. Researchers suggest that expert decision makers are subject to logical errors, including a redundancy phenomenon, in which clinicians continue gathering data to substantiate findings after there is sufficient evidence for a conclusion.(3,22) Efforts at cost containment may not be compatible with the receptive approach (for example, cost of procedures, duplication of tests), and clinicians may need to review their own cognitive style preferences in relation to the demands of the health care system. Redundancy in evaluation may require further investigation. The statistical differences between countries are interesting, but we believe they are due to the greater heterogeneity of the Australian sample. Differences by major practice area. All subscales showed effects for practice area. These findings are congruent with literature suggesting that problem structure evokes cognitive behavior.(2,30,31) Hammond et al(3O) described task characteristics likely to induce intuitive or analytical processing, which may explain our findings. if the task offers a large number of cues simultaneously, it is hard to decompose into discrete parts, and, if measurement is perceptual, it favors intuitive processing. Tasks with fewer and sequential cues, which can be decomposed into discrete parts and measured objectively, favor an analytical (systematic) approach. This finding raises the question of whether patient care tasks in different areas of physical therapy differ in these dimensions. Our profession espouses the scientific method and analytical thinking but may also need to consider the influence of task structure on cognitive strategies.

Understanding the relationship between cognitive style preference and clinical practice may help clinicians gain an improved perspective on their own performance. As we bring aspects of our activities into conscious awareness, we are better able to determine our own strengths and weaknesses and thus reduce the potential for error. As our understanding of the decision-making process increases, so will our ability to make decisions more efficiently. Educationally, there are important implications for the relationship between cognitive style and clinical decision making. With an understanding of a student's cognitive style preference, faculty can guide the student to select more effective learning strategies. Faculty can also structure case studies and learning experiences to elicit desired approaches to decision making.

Future Studies

Considerably more study is needed on the effects of cognitive style preference and the physical therapy task on decision making. Hammond et al(30) suggest that performance is most accurate when task attributes are matched with cognitive attributes. We hesitate to suggest that therapists choose their respective practice areas because they are attuned to different cognitive styles; rather, we believe that most clinicians, faced with a differently structured problem type, utilize a different cognitive strategy. This hypothesis could be the subject of further investigation. We also plan to use the scale with new graduates to compare their cognitive style preferences with those of expert groups.

Conclusions

A study of expert decision-making behaviors in the United States and Australia revealed an overall preference for one's own assessment as a source of information, for the receptive style of data gathering, and for the systematic style of information processing. Significant differences were found for major practice areas in both countries, which suggests different cognitive approaches for different task structures. Country effects were also found between American and Australian therapists. Acknowledgment We would like to thank Harry Davis of the Medical College of Georgia's Department of Research, Statistics, and Computers for his invaluable assistance in research design and data analyses.