| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Research Reports |
DL Riddle, PT, PhD, FAPTA, is Professor, Department of Physical Therapy, Medical College of Virginia Campus, Virginia Commonwealth University, Richmond, VA 23298-0224 (USA) (dlriddle{at}vcu.edu).
BE Hillner, MD, is Professor, Department of Internal Medicine/General Medicine & Primary Care, Medical College of Virginia Campus, Virginia Commonwealth University
PS Wells, MD, Canada Research Chair in Thromboembolic Disease, is Professor, Department of Medicine and Clinical Epidemiology Unit, University of Ottawa and the Ottawa Health Research Institute, Ottawa, Ontario, Canada
RE Johnson, PhD, is Associate Professor, Departments of Biostatistics and Family Practice, Virginia Commonwealth University
HJ Hoffman, MS, is Graduate Assistant, Department of Biostatistics, Virginia Commonwealth University
WA Zuelzer, MD, is Associate Professor, Department of Orthopaedic Surgery, Virginia Commonwealth University
Address all correspondence to Dr Riddle
Submitted September 3, 2003;
Accepted February 25, 2004
 | Abstract |
|---|
Key Words: Diagnosis • Venous thrombosis
| Introduction |
|---|
 TopAbstract Introduction Materials and MethodsResultsDiscussionConclusionAppendixReferences |
|---|
Patients who have had major orthopedic surgery or trauma of the lower extremities are among those at highest risk for VTE.5 For example, PDVT is reported to occur in 66% of patients with isolated lower-extremity fractures who were seen in a trauma unit.6 Deep vein thrombosis also can occur in patients following discharge from the hospital and in nonhospitalized patients. Warwick and colleagues7 reported that 64% of VTE complications among 1,162 patients following hip arthroplasty occurred following discharge from the hospital.
Because of decreasing lengths of hospitalization for patients following surgery, some authors1,8 have suggested that when these patients are seen in outpatient settings, their risk for PDVT is elevated relative to when hospital stays were longer. We suspect, therefore, that physical therapists in outpatient settings may now be more likely to see patients with undiagnosed PDVT as compared with when hospital stays were longer.
Physical therapists in outpatient settings routinely screen patients with musculoskeletal problems for potentially serious disorders such as PDVT.9 The clinical diagnosis of PDVT, however, has traditionally been thought to be fraught with error, and physicians have relied heavily on radiologic or laboratory diagnostic tests.10 The Homans sign, for example, is one of the more commonly used clinical tests for detecting PDVT, but the test has essentially no diagnostic value.11,12 Many researchers13–17 have developed methods for more accurately identifying outpatients who may have PDVT. One of the more common approaches to improving diagnosis is by use of a clinical decision rule (CDR). Clinical decision rules quantify the individual contributions that components of the medical history and physical examination make to a diagnosis.18
In a series of studies,15–17 Wells and colleagues found that patients could be categorized into low-, moderate-, and high-risk groups based on their CDR scores (Tab. 1). The CDR consists of 9 medical history and physical examination findings that, in our opinion, are simple to obtain. Patients with scores of 0 or less had a 3% (95% confidence interval [CI]=1.7%–5.9%) probability of PDVT, scores of 1 or 2 indicated a 17% (95% CI=12%–23%) chance of PDVT, and those who scored 3 or higher had a probability of PDVT of 75% (95% CI=63%–84%). The findings were confirmed by other researchers.19–21
|
We determined whether physical therapists indicated they would contact the referring physician about the hypothetical patients' condition. Extensive literature suggests that, because of the risk of life-threatening PE, any patient suspected of having PDVT should have formal diagnostic testing.1,10,22
Our final purpose was to determine whether physical therapist characteristics of years of clinical experience, type of practice setting, board certification status, and geographic region affected the judgments of PDVT probability or referral to the patient's physician. Specialty certification has been shown to influence outcome in other medical specialties.23–25 Geographic region also has been shown to contribute to practice variation for other disorders26,27 and may affect judgments about PDVT made by physical therapists.
| Materials and Methods |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionConclusionAppendixReferences |
|---|
Subjects
Of the 8,358 eligible physical therapists, 1,189 therapists were listed as being board certified in orthopedic physical therapy and 7,169 were not board certified. Because we were interested in comparing results of physical therapists who were and were not board certified, we accounted for board certification status in our stratified sampling procedure. We randomly sampled 750 physical therapists who were not board certified (10% of the population) and 750 physical therapists who were board certified (63% of the population). Figure 1 gives a description of how the samples were chosen.
|
|
Our research team wrote the vignettes, initially without the input of one of the coauthors (PSW), who was one of the developers of the CDR. We did not include that coauthor in the vignette writing because we wanted to determine if the probability estimates made by the research team agreed with those of that individual, whose ratings we considered to be the standard for the vignettes. We designed the vignettes to reflect orthopedic clinical practice and the types of outpatients who are generally considered to be at risk for PDVT. For example, 5 of the 6 vignettes described outpatients who had recently had orthopedic surgery.5 After the vignettes were written and probability estimates were made based on consensus of the other research team members, they were sent to the coauthor who was a developer of the CDR (PSW) for an independent estimate of PDVT probability. Probability estimates agreed in all cases.
We then conducted a pilot survey study, in part, to determine if therapists and orthopedic surgeons found the vignettes to be credible descriptions of the types of patients seen in orthopedic practices. We sent the pilot survey questionnaire to 12 physical therapists with 5 or more years of clinical experience. Only experienced therapists were chosen because we wanted the therapists to use their clinical experience as a basis for judging credibility of the vignettes. The therapists were clinical instructors for the Department of Physical Therapy, Virginia Commonwealth University, and working in outpatient orthopedic clinics in the Richmond, Va, area. Ten of 12 therapists completed the pilot survey.
After reading each vignette, the physical therapist was asked to answer 2 questions. The first question asked the therapist to estimate the probability that the patient had symptomatic PDVT of the lower extermity. The 3 probability options were "low," "moderate," and "high" and were operationally defined for the therapists in accordance with the work of Wells et al.16 In our survey, we defined low, moderate, and high probability for the participants in the following way: low probability indicated that the therapist suspected that the probability of PDVT was 5% or less, moderate probability indicated that the probability of PDVT was greater than 5% but less than 25%, and high probability was a 25% or greater likelihood of PDVT. The second question asked was, "Would you contact the referring physician today about this patient's condition?"
We also sent the pilot survey questionnaire (with the question about physician referral deleted) to 8 orthopedic surgeons who were faculty members of the Department of Orthopaedic Surgery, Virginia Commonwealth University. We sought input from orthopedists because the vignettes addressed surgical procedures and we wanted this information to reflect actual practice. Six of the orthopedists completed the pilot survey. After completing the pilot survey, the orthopedists and the physical therapists were asked whether they found the vignettes to be credible descriptions of the types of patients seen in their practices. All clinicians indicated that they found the vignettes to be credible.
We concluded, based on our pilot work, that our vignettes had probability estimates that agreed with those of the developer of the CDR and were credible (clinicians reported that the vignettes were indicative of patients seen in their practices). The survey questionnaires were then sent to the randomly selected sample of therapists. The questions asked in the survey were identical to those asked in the pilot study. Therapists participating in the pilot study were not enrolled in the main study.
Data Analysis
Survey responses were reported as percentages and displayed in bar graphs. These data included sample proportions of physical therapists who answered "low probability," "moderate probability," and "high probability" to each of the vignettes. Sample proportions of therapists who would and would not contact the referring physician that day also were calculated.
Because we had 2 vignettes for each probability category, we examined the intratester reliability of therapists' probability judgments. Instead of using a conventional approach to examining reliability with repeated judgments of the same subject (or, in this case, the same vignette), we determined the extent to which therapists agreed on their probability estimates of each pair of vignettes we assigned to the same category of probability. For the probability estimates of low, moderate, and high probability, intratester reliability was analyzed by calculating the percentage of agreement for the 2 vignettes in each category.
Multiple logistic regression was used to determine the effect of the following variables on each question asked in the survey: place of practice (hospital-based practice or outpatient-based practice), board certification status (yes or no), number of years of clinical experience (above or below median of 11 years), and region of the country (US census regions West, Midwest, Northeast, and South). The first analysis determined if any of the therapist characteristic variables influenced the responses to the question that asked therapists to estimate the probability of DVT for each of the 6 scenarios. A therapist's answer was considered "correct" if it matched the probability level as determined by the CDR for the given vignette. The second analysis determined if any of the therapist characteristic variables influenced the decision about whether to contact the referring physician. The "correct" decision regarding physician notification was "yes" for all scenarios. For both sets of analyses, we also tested all 2-way and 3-way interactions.
The significance level for the regression analyses was set at P
.05. We did not adjust for the experiment-wise error rate because we considered these regression analyses to be exploratory in nature. Because we calculated 2 logistic regression models (one for each question asked in the survey) for each of the 6 vignettes, we considered a variable as important only if it was significant in a majority (
4) of vignettes. Variables that were found to be significant in less than a majority of vignettes were, in our opinion, not important.
Because we oversampled the physical therapists who were board certified, we adjusted our analysis using sampling weights in order to reflect the numbers of board-certified physical therapists in the population. The sampling weights include factors that correct for reductions in sample size due to survey nonresponse. Our sample size was a fairly large proportion of the population size—10% and 63% for non–board-certified and board-certified physical therapists, respectively. We therefore adjusted sample variance estimates using finite population corrections (subjects were sampled without replacement, which is designed to reduce sampling error).30
To evaluate potential nonresponse bias, we first calculated the mean age, proportion of board-certified physical therapists, proportion of physical therapists within each practice type, and proportion of physical therapists within each region of the country for the full population. We then calculated 95% confidence intervals based on the responding sample. If a population mean or proportion did not fall within the corresponding 95% confidence interval, the sample was deemed biased on that characteristic.
Descriptive analysis for this article was generated using SAS software, Version 8 of the SAS System for Windows.
Logistic regression analysis was generated using SUDAAN,
a statistical software program designed to analyze data from complex multistage sample surveys and cluster-correlated data. Graphics were generated using Microsoft Excel 2002.
| Results |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionConclusionAppendixReferences |
|---|
For the 2 low-probability vignettes, physical therapists overestimated the probability of PDVT 71% of the time in one vignette and 10% of the time in the other vignette. For the 2 moderate-probability vignettes, physical therapists underestimated probability 21% of the time for one vignette and 26% of the time for the other vignette. Therapists overestimated probability for the 2 moderate-probability vignettes 28% and 37% of the time. For the 2 high-probability vignettes, therapists underestimated PDVT probability 87% and 64% of the time (Fig. 2). Figure 2 also summarizes population estimates for each of the probability categories. For example, for vignette H1, 35% of the population, or approximately 2,900 therapists, would rate the probability of PDVT in this high-probability vignette as low, whereas approximately 1,100 therapists in the population would rate the probability as high.
|
|
For all analyses, we tested both 2-way and 3-way interactions, but we did not find them to be significant. Years of clinical experience was examined as a continuous variable or as a dichotomous variable (greater than the median or less than or equal to the median) in all analyses, with no differences in results. We therefore presented the results for clinical experience using the simpler dichotomous variable.
To examine nonresponse bias, we compared characteristics of physical therapists who completed the survey (n=969) with those of the population of physical therapist members of APTA's Orthopaedic Section (n=8,358). We examined the following variables: median years of clinical experience, board certification status, types of practice settings, and regions of the country. For all variables, the population mean or proportion fell within the corresponding 95% confidence interval of the sample.
| Discussion |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionConclusionAppendixReferences |
|---|
Because the potential consequences of missing a PDVT (eg, PE) are great, the literature supports conducting formal diagnostic testing even when the risk of PDVT is determined to be low.1,10 Given the considerable mortality and morbidity associated with PDVT, we contend that physical therapists should contact the referring physician whenever PDVT is suspected. The CDR can be used to aid in the identification of people who are at risk for PDVT and for estimating the likelihood of PDVT.
If our vignettes reflect the types of patients seen by physical therapists and the survey data reflect the clinical decisions physical therapists make, our data suggest that physical therapists may not be contacting the referring physician about a patient's PDVT risk as frequently as they should. This finding may be partially explained by therapists' inability to consistently estimate a patient's risk for PDVT.
To illustrate how discrepancies in probability estimates may influence patient care, we examined the impact of the physical therapists' responses to question 1 (DVT probability estimate) on their responses to question 2 ("Would you contact the referring physician?"). Figure 4 illustrates the patterns of responses for those therapists who indicated that they would contact the referring physician stratified by the responses to question 1 (low, moderate, or high probability). For one high-probability vignette (vignette H2), 18% of the therapists indicated that they would contact the referring physician when they estimated the probability of developing PDVT to be low. This finding contrasts with 89% and 99% of the therapists who would contact the referring physician when estimates of probability were moderate or high, respectively. These data suggest that for patients with a high probability of developing PDVT, physical therapists who underestimate PDVT probability are more likely not to contact the referring physician as compared with therapists who do not underestimate probability. Underestimates of PDVT probability appear to be the more serious errors because they may lead the therapist to falsely conclude that physician referral is not needed. Use of the CDR developed by Wells and colleagues15–17 appears to have potential to improve the accuracy of physical therapists' judgments of PDVT probability and physical therapists' decisions regarding the need for physician referral.
|
Evidence exists to suggest that with training, reliability of probability estimates made by nurses and physicians achieves an acceptable level (kappa=.75).16 Research is needed to assess the reliability of physical therapists' probability estimates using the CDR on patients rather than with vignettes and to determine if training in use of the CDR affects reliability.
We found no evidence to indicate that physical therapists' characteristics of years of clinical experience, board certification status, practice setting, or region of the country influenced the results. Clinical judgments related to DVT diagnosis appear not to be influenced by these variables. Because of the random sampling procedure used in our study, the estimates obtained appear to be generalizable to the population of physical therapist members of APTA's Orthopaedic Section.
We found evidence for a valid CDR in the general medical literature regarding DVT diagnosis,15,19 and we determined the extent to which physical therapists' judgments agreed with this evidence. Some studies31,32 have demonstrated that a gap exists between research evidence and the practice of medicine, but we are not aware of data showing the extent of knowledge transfer to physical therapist practice. Knowledge transfer is especially difficult when attempting to take evidence from one specialty area and applying it to another area of practice.33 Therefore, we suspected that there would be a disparity between the evidence and current practice. We believe our results describe the magnitude of this disparity.
Clinical Application of the CDR to Outpatients With Musculoskeletal Disorders
The CDR of Wells and colleagues15–17 was designed for use when there is any suspicion, based on medical history, signs, or symptoms, that the patient may have PDVT. Suspicion, in our experience, usually arises when a patient's signs or symptoms are inconsistent with or out of proportion to the disorder for which the patient is being managed. The CDR has not been validated as a general screening tool on all patients independent of their signs or symptoms, and therefore we do not recommend indiscriminate use of the CDR.
In our experience, diagnosing PDVT is especially difficult in outpatients following lower-extremity surgery or trauma. These patients may have clinical findings (eg, lower-extremity swelling or calf tenderness) that would increase the CDR score, but these findings also may be considered part of the routine recovery following surgery or trauma and therefore unrelated to DVT. When working with patients who have a lower-extremity injury or surgery, physical therapists first need to determine whether use of the CDR is appropriate. If the clinician cannot confidently rule out the possibility of PDVT based on medical history and examination data, then we recommend that the CDR be used to further refine the patient's degree of risk for PDVT. We believe that physical therapists should be conservative and use the CDR whenever any question exists about the presence of PDVT.
In vignette L1, for example, the patient had calf swelling of greater than 3 cm and calf tenderness, which counted for 2 points on the CDR. Calf swelling and tenderness, however, also may be considered by the physical therapist to be attributable to chronic knee problems and not to a DVT. We believe the therapist has 2 options in this scenario. If, after collecting history and examination data, the therapist cannot confidently rule out DVT, we recommend that the CDR be used. The patient would score 2 points on the CDR (1 point each for the calf swelling and tenderness). In addition, the therapist would likely consider an alternative explanation (alternative diagnosis of a chronic knee injury) for the lower-extremity swelling and tenderness, which reduces the patient's score by 2 points. In this case, the CDR would indicate a score of 0 and a probability of DVT of approximately 3%. The second option is to consider the swelling and tenderness to be entirely attributable to the knee injury and not even consider PDVT as a possibility. For this second option, the therapist would not use the CDR. If the therapist has any suspicion that the signs or symptoms are out of proportion to usual clinical findings, then there is a risk of PDVT. We then recommend using the CDR so that the potential risk of PDVT is not ignored.
Limitations
It is unclear to what extent patient vignettes actually reflect clinical practice. It is possible that the decisions made by physical therapists while completing the questions for the vignettes vary from those actually made during clinical practice. Recent evidence exists to support the use of case simulations for examining some clinician behaviors.34,35 Peabody and colleagues35 determined whether case simulations were a legitimate method for measuring the process of care compared with actual clinical practice. The authors compared the data collected by a large group of physicians on a series of patients with data collected while the physicians completed case simulations that were identical to the characteristics of one of the patients seen in the clinic. Data from vignettes with diagnoses as varied as low back pain, diabetes, and coronary artery disease were found to be valid predictors of actual clinical performance for decisions related to clinical examination and diagnostic testing as well as intervention.35 Use of vignettes appears to be appropriate for investigating the process of care provided by physicians in clinical practice. It is unclear whether the same is true for physical therapists. We also have no data that allow us to generalize our findings to patients with disorders unrelated to the musculoskeletal system or to physical therapists who are not APTA Orthopaedic Section members. Future research should address the generalizability of our results.
| Conclusion |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionConclusionAppendixReferences |
|---|
| Appendix |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionConclusionAppendixReferences |
|---|
|
| Footnotes |
|---|
The results of this study were presented, in part, at the Combined Sections Meeting of the American Physical Therapy Association; February 12–16, 2003; Tampa, Fla.
This study was supported by a grant from the Agency for Healthcare Research and Quality (#RO3 HS13059–01).
* Orthopaedic Section, American Physical Therapy Association, 2920 East Ave South, Suite 200, La Crosse, WI 54601. 
SAS Institute Inc, PO Box 8000, Cary, NC 27511.
Research Triangle Institute, 3040 Cornwallis Rd, Research Triangle Park, NC 27709-2194.
Microsoft Corp, One Microsoft Way, Redmond, WA 98052-6399.
| References |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionConclusionAppendixReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  D. U Jette, K. Ardleigh, K. Chandler, and L. McShea Decision-Making Ability of Physical Therapists: Physical Therapy Intervention or Medical Referral Physical Therapy, December 1, 2006; 86(12): 1619 - 1629. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. L Craik Never Satisfied Physical Therapy, November 1, 2005; 85(11): 1224 - 1237. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
