Polich, Ann L. MD, MPH; Etherton, Gale M. MD; Knezevich, Jon T. PharmD, BCPS; Rousek, Justin B.; Masek, Chris M.; Hallbeck, M. Susan PhD
Hospital-acquired venous thromboembolism (VTE) is a common and preventable adverse event that most patients are at risk of developing during their hospital stays. VTE prophylactic anticoagulation (chemoprophylaxis) is the preferred pharmacological assignment for reducing risk of VTE, but it is underused in current practices involving risk stratification (RS) for VTE prevention. We carried out this study to determine whether a protocol that eliminates the RS step is more likely to lead residents to evidence-based VTE assignment than the currently used RS protocol.
VTE is a collective term referring to deep venous thrombosis (DVT) and pulmonary embolism (PE). The average annual incidence of VTE has been estimated at 117 per 100,000.1 Furthermore, the risk of VTE is known to double for individuals each decade over 40 years of age.1 Each year, it is estimated that there are two million VTE events, of which 10% are fatal.2 Hospitalized patients account for 50% of all cases of VTE, and when such patients are assessed for VTE prophylaxis (consisting of pharmacological and nonpharmacological treatment), less than 5% of them are deemed “low risk.”3 VTE prophylactic anticoagulation (chemoprophylaxis) is the preferred preventive treatment for the avoidance of VTE, reducing the incidence of DVT and PE by 50% and 65%, respectively.3 The Agency for Healthcare Research and Quality (AHRQ) estimates that the diagnosis of DVT and PE costs an additional $10,000 and $20,000 per hospitalization, respectively.3 Once diagnosed with VTE, 30% of patients have a recurrence and significant morbidity secondary to postthrombotic syndrome.3
Despite overwhelming evidence in favor of VTE chemoprophylaxis, it continues to be underused in the hospital setting.3,4 In a DVT free-registry study, only 42% of patients diagnosed with VTE had prior chemoprophylaxis.3 Because of the disparity between safe, effective, evidence-based guidelines for VTE prevention and actual hospital practice, the Joint Commission, AHRQ, the Surgical Care Improvement Project, and Medicare are all leading initiatives to increase prophylaxis to reduce incidence of VTE in U.S. hospitals.5 The American Public Health Association regards the underuse of appropriate DVT prophylaxis as a public health crisis.6
Traditionally, the risk for developing VTE has been ascribed to surgical patients, but, in reality, 70% to 80% of fatal PEs occur among hospitalized medical patients.7 Hospitalized patients are at particular risk for developing VTE because they frequently have medical conditions that promote clot formation, such as stroke, chronic obstructive pulmonary disease, infection, malignancy, congestive heart failure, and immobility.8,9 Overall, the risk of hospital-acquired VTE ranges from 10% to 80%, depending on the type and severity of medical illness.2
Further research has shown that the risk for VTE in hospitalized patients is nearly universal.3 Despite this, many hospitalized patients do not receive acceptable therapy.4 The gap that exists between proven VTE prevention measures and practice may be exacerbated by a process that includes steps such as RS.10 There are many VTE risk assessment models but few validated, prospective VTE protocol studies.
Traditionally, the VTE assignment process can be divided into three steps. The first step is assessment of the risk of developing VTE. The second step is exclusion of those for whom chemoprophylaxis would be contraindicated, and the third step is determining the appropriate type and dose of chemoprophylaxis. Each step in the process is a branch point in a decision tree. In the present study, the process of moving through these steps is referred to as VTE assignment.
It remains unclear whether identifying or stratifying individual risk is a necessary first step to determine whether VTE chemoprophylaxis is appropriate. In the present study, we challenge the assumption that RS is necessary. In our study, a traditional RS protocol was compared with a streamlined protocol where the RS step was eliminated. The non-RS protocol assumes that all patients are high risk except those who are <40 years old2,11 and have an estimated hospital stay <48 hours.10 By eliminating the RS step, the process is simplified and, thus, may improve the ability of residents to comply with recommended treatment guidelines. It is generally accepted that simplification of a complex process improves the ability of an individual to complete the process and minimizes chances of introducing unintended error.
Health care systems have begun to use Lean and Six Sigma engineering principles to eliminate wasteful and unnecessary work (or steps) and address the need for a streamlined flow of work to reduce delays, improve quality of care, and control costs.12 These approaches rely on the organization's frontline staff (e.g., residents) to identify and make improvements.12 Many processes in health care reflect the old tradition whereby the physician determined the treatment processes without the availability of data and considered the process to be valid if the patient outcome was acceptable. The traditional RS protocol is an example of such a process. To date, no validated studies have been done in which each step in the process was evaluated to see if it is necessary. The evidence supports VTE prophylaxis; what it does not support is how best to accomplish this from a Lean health care perspective.
Goldhaber and Tapson3 validated a VTE prevention protocol that included RS as part of the process. A 2009 study in an Australian teaching hospital noted that the incidence of VTE dropped after the initiation of a risk assessment pocket card tool.13 For both of these studies, it could be argued that introduction of any protocol would have improved outcomes when none existed before. It does not validate the need for RS as a first step.
Similar to traditional RS protocols, the non-RS protocol in this study follows American College of Chest Physician (ACCP) guidelines for VTE prophylaxis14 and assesses a given individual's medical risk for receiving anticoagulation. For example, there are subsets of patients who receive no benefit from VTE chemoprophylaxis while hospitalized, such as those who are already therapeutic on chemoprophylaxis at the time of admission. Additionally, there are patients who would have adverse outcomes if VTE chemoprophylaxis were initiated on admission (e.g., patients with intracranial hemorrhage).
Once it is determined that there are no contraindications to chemoprophylaxis in general, the next step is to determine the appropriate type and dose of chemoprophylaxis. An error in selection and/or dosing of chemoprophylaxis can lead to adverse clinical outcomes. Currently available VTE protocols require physicians to adjust the selection and the dose of the drug available based on recognition of contraindications of those drugs offered. This adds a level of complexity such that unintentional errors could be made. Both protocols include patient information such as weight and creatinine clearance in their algorithms. The non-RS protocol adds an additional step to guide the resident physician in the selection of the proper chemoprophylaxis based on the patient's weight and creatinine clearance. Our institution, the Nebraska Western Iowa (NWI) Veterans Affairs (VA) Health Care System, uses computerized physician order entry into the electronic medical record. In this study, protocols are written as they would be expressed in the physician order entry process, thus mimicking the process by which residents enter orders for VTE prophylaxis.
VTE chemoprophylaxis has been shown to be underused in hospitalized patients.3,4 We hypothesize that the simplified non-RS VTE assignment protocol will improve appropriate assignment and use of VTE chemoprophylaxis and improve compliance with recommended guidelines. We also hypothesized that physicians will prefer a simplified protocol to one that is more complex.
We conducted the study in 2009 at the NWI VA resident clinic in Omaha, Nebraska. VA health care systems have become the largest provider of health care training in the United States, and they respond rapidly to improved models of clinical decision making because of their commitment to innovation and high-quality patient care.15 More than half of all physicians in the United States have trained at a VA health care system during their careers,16 which makes the VA an ideal place to test VTE assignment protocols. We assumed that if an alternative protocol used in this study would increase compliance with VTE chemoprophylaxis guidelines, then the large influx and throughput of residents in the VA would likely spread that protocol beyond the VA.
We invited 110 medical residents to participate on a voluntary basis in this study, which was Web based and administered in the VA outpatient clinic mentioned above. Forty-one (20 first-year, 7 second-year, and 14 third-year residents) participated, resulting in 41 matched pairs for crossover analysis between the two prototype protocols. This sample size was found to be sufficient from a prestudy power analysis that found that a minimum of 36 matched pairs were required to produce a power rating of 0.9 (effect size = 0.5, alpha = .05). Medical residents were ideal participants for this study because they are involved in the majority of admissions, write admission orders, and, thus, determine VTE assignments.
The protocols and the related scenarios
The two assignment protocols compared in this study consisted of the current RS protocol and a proposed protocol (non-RS protocol) that eliminates the RS step. (See the Supplemental Digital Appendix at http://links.lww.com/ACADMED/A63 for detailed algorithms showing all steps of each protocol.) An online survey tool was used to translate the two protocols into an electronic online menu-driven data-collection system. Six scenarios were presented in each protocol in randomized orders within each protocol. For example, the first scenario (Scenario A) states “70 y/o WM with h/o DJD admitted with hematemesis and melena. Has been taking NSAIDs for pain control. Weight 100 kg. Creatinine clearance 75 mL/min. You are writing admission orders for this patient.” The appropriate (evidence-based) VTE assignment for this scenario was nonpharmacological treatment (antiembolic stockings) because of the present contraindications for chemoprophylaxis. (See the Appendix for a list of all six scenarios and appropriate assignments.)
The scenarios were presented individually with the appropriate protocol and hosted on LimeSurvey V. 1.82+ (6825; Hamburg, Germany), an online survey application service, on a controlled site inaccessible to the general public. LimeSurvey is an open-source survey tool used to collect survey data online. The software precisely time stamped each response. All scenarios required approximately 10 minutes to complete and had a one-time use (random number) token to access the tool to prevent multiple accesses and ensure confidentiality. The survey measured time to complete each scenario and the final assignment (appropriate or inappropriate).
Validity and reliability of scenario performance measures
The content validity and the clarity of the six scenarios were evaluated by a team of experts. The team consisted of two experienced internal medicine physicians and one experienced clinical pharmacist. The scenarios were created to reflect common clinical situations that the residents would encounter when caring for hospitalized patients. The assignment recommendations by the expert panel were consistent with the eighth edition of the ACCP guidelines.14
On being informed of the aims of the research, the voluntary nature of participation, and institutional review board approval from the NWI VA Health Care System, each resident was given a handout sheet containing a Web address for the online data-collection tool, initial scenario number to be completed (to identify randomized order used), and a one-time use token (random number) to access the tool. Signing on was considered consent.
The welcome page of the online survey reminded the participants of the aims of the research:
The purpose of this exercise is to compare two different assessment and treatment methodologies. You will be presented six patient scenarios at admission; two different menu-driven systems (protocols) will walk you through the assessment and treatment options for venous thromboembolism. You will use each scenario with each assessment and treatment menu. There are 15 questions in the survey.
Participants were free to begin and/or complete the survey at their discretion at the VA Outpatient Resident Clinic. A study coordinator (G.M.E. and A.L.P.) was physically available to the residents if any questions, concerns, or comments arose.
Each resident was assigned a randomized scenario set consisting of six hypothetical patient scenarios to be assessed by the two different protocols: the RS protocol and the non-RS protocol. Half the randomized scenario sets started with the RS protocol and ended with the non-RS protocol; the other half of the randomized scenario sets began with the non-RS protocol and ended with the RS protocol. The scenario order was randomized within the protocol.
Two 6-point Likert scale questions were asked after completion of each specific protocol. The residents were asked, “If you had to use this protocol, would this format help you in determining appropriate anticoagulation assignment for VTE?” The residents recorded “1” if they strongly disagreed, recorded “6” if they strongly agreed, and recorded “2” through “5” for less strong agreement or disagreement, “3” being a neutral response.
Data collected included time to complete protocols—using time stamp of mouse click (finish − start time)—VTE assignment selected, scenario identifier, year of postgraduate medical residency training, and the Likert scale response for each protocol. The statistical tests used in the analyses were an analysis of variance (ANOVA), McNemar and Pearson chi-square test, and the Mann–Whitney U test. All statistical tests were performed using SPSS (V17; Chicago, Illinois) using a .05 level of significance.
The objective (dependent) variables collected for the ANOVA consisted of the protocol and scenario completion times and final scenario assignment. The independent variables tested were protocol used, scenario used (for scenario completion time), and training year. Two different protocols to determine chemoprophylaxis assignment were used, one using RS (the RS protocol) and the other eliminating risk RS (the non-RS protocol). There were a total of six different scenarios (A, B, C, D, E, or F). The participant training variable was the current year of medical residency (three levels; first-, second-, or third-year). The resulting full-factorial ANOVA was a 2 (protocol used) × 6 (scenario used) × 3 (education level). Tukey post hoc tests were performed on significant effects, and simple-effects F tests were performed on all significant interactions. Completion time outliers found in the scenarios were identified as any scenario completion time lasting greater than 180 seconds (which occurred when residents were called away during testing or multitasked) and discarded from evaluation in 18 of the 492 total completions (41 residents × 6 scenarios × 2 protocols).
The McNemar chi-square test (dependent data use) was used to test for significant differences between the VTE assignments among the protocols (two levels), scenarios (six levels; individually tested) and first-assigned protocol (two levels; RS or non-RS). The appropriate (evidence-based) VTE assignment was determined by the residents' responses to the drug dosing question in each of the given scenarios. The Pearson chi-square test (independent data use) was used to test for significant differences between the VTE assignments (two levels; appropriate or not appropriate) among the different education levels (three levels) of participants.
The Likert scale questions were graphed by the frequencies of response 1 (strongly disagree) through response 6 (strongly agree) by protocol. The two responses by each resident regarding the Likert scale questions were then tested using the Mann–Whitney U test to verify whether the responses were significantly different among protocols.
Protocol completion time
Residents completed the non-RS protocol (M = 307.8 seconds, SD = 88.14) significantly faster than the RS protocol (M = 431.5 seconds, SD = 126.74), F(1, 70) = 20.97, P < .001. The non-RS protocol was completed 1.4 times faster than the RS protocol.
Scenario completion time
The average scenario completion time using the non-RS protocol (M = 52.8 seconds, SD = 29.41) was significantly less than when using the RS protocol (M = 79.6 seconds, SD = 35.20), F(1, 473) = 72.00, P < .001. Similar results were found when the scenario completion times were categorized into their six respective scenarios, F(11, 462) = 9.54, P < .001. A Tukey post hoc test was conducted on the scenario completion time (see Figure 1) by the residents, and all but scenarios D and F differed significantly (further Tukey test results can be seen in Table 1).
Sixteen residents were assigned the non-RS protocol first, and 25 residents were assigned the RS protocol first. The specific protocol assigned first to the residents did not have a significant effect on the appropriate (evidence-based) scenario assignment. There was also no significant difference in resident year (first, second, or third year) among scenario assignments at the .05 level of significance for either protocol.
The significance of the McNemar chi-square test indicated that assignments prescribed using the non-RS protocol (84% correct) had a 33% greater percentage of appropriate assignments than those using the RS protocol (63% correct), χ2(1, N = 246) = 34.68, P < .001. The same test was then conducted on the six paired scenarios (A, B, C, D, E, and F in each protocol) to find which pairs generated significantly different appropriate scenario assignment data (see Table 2). Scenarios A, B, C, and F were found to have significantly more appropriate assignments for the non-RS protocol than for the RS protocol.
The six-point Likert scale statement was completed by residents after completion of each of the two protocols. Their responses to the statement did not produce a statistically significant difference between the two protocols. However, there were more residents who strongly agreed that the non-RS protocol (13 participants) would help them in determining the appropriate VTE assignment compared with the RS protocol (10 participants).
The results of this study support the original hypothesis that the elimination of RS (via the non-RS protocol) would improve appropriate assignment and use of VTE chemoprophylaxis and would improve compliance with recommended guidelines. The study did not produce significant results to support the hypothesis that residents would prefer a simplified protocol.
Simplification of VTE assessment
From a systems point of view, introducing complexity into a process increases the likelihood of error. In this study, simplification of the VTE assignment is evidenced by a shorter time to complete the non-RS protocol as compared with the more traditional RS protocol. Another indication of ease of use was noted by the result that there were no statistical differences in the appropriate VTE assignment by year of residency training: On average, first-year residents were able to arrive at the same percentage of appropriate VTE assignment as their third-year counterparts. Such a streamlined protocol design will reduce opportunities for error in a setting where physicians are multitasking and/or lack experience.
Likert scores between the two protocols (RS and non-RS) were not statistically significant, which indicates that the residents did not tend to favor the protocol that was objectively easier for them to use. We theorize that this is most likely due to the familiarity of the residents with the traditional RS protocol.
It seems that the simplified non-RS protocol improved the ability of the residents to reach the appropriate drug and dose for VTE chemoprophylaxis (84% correct) when compared with the traditional protocol (63% correct). Thus, simplification of the protocol and added decision support increased compliance with ACCP guidelines. Improving compliance with evidence-based medicine is a core competency outlined by the Accreditation Council for Graduate Medical Education (ACGME).17 The non-RS protocol design was guided by the recommendations of the ACCP for VTE prophylaxis.14
Increase in appropriateness of VTE assignment
Statistically significant results showed that residents arrived at the appropriate VTE assignment 33% more frequently when using the non-RS protocol than when using the RS protocol. The non-RS protocol provides the patient's creatinine clearance and weight to help guide the physician to the appropriate medication choice. This minimizes error and provides a forcing function for the appropriate dose adjustments and selection for chemoprophylaxis. The non-RS protocol reduces the burden of memorizing and recalling the specific dose adjustment and weight-based calculations for chemoprophylaxis.
Relevance of study
This study seems to be the first of its kind in comparing whether or not the standard RS protocol is optimal during the evidence-based assignment of VTE in hospitalized patients by residents. The importance of this study is reflected in the results, which show that efficiency can be improved by a protocol that uses decision guidance and that leads the resident to the appropriate VTE assignment. Because VTE chemoprophylaxis assignment is known to be underused in hospitalized patients,3,4 a systematic approach that focuses on residents' interface with the computerized protocol and improves their performance will likely improve appropriate use. The information provided from this study will help in the design of an efficient, safe, and user-friendly VTE order menu for residents, so that VTE chemoprophylaxis can be universally and appropriately applied to the inpatient population. In addition, once the non-RS protocol has been introduced to the residents, it is likely that the ACCP guidelines for anticoagulation14 will be spread throughout the medical community via the residents as they leave the VA to practice at affiliate hospitals or in private practice.
Simplifying and streamlining a well-established process such as VTE prophylaxis to make it easier for the ordering resident physician to get the appropriate treatment for the patient is a model for quality improvement. Our study demonstrates that even well-accepted processes can be improved on, and, in doing so, it encourages residents to think about other well-established processes and how these might be improved on to provide more efficient and better-quality clinical care for our patients. This reinforces ACGME core competencies17 in practice-based learning and improvement as well as systems-based practice.
Although this is a limited study population, if our study results are extrapolated to the two million DVT patients per year,2 then application of the non-RS protocol would increase correct VTE assignment of chemoprophylaxis by >400,000 patients per year, prevent approximately 200,000 VTE events, and reduce the number of deaths by 20,000. RS for VTE prevention in hospitalized patients is an antiquated approach. Current data support the near universality of risk for VTE when a patient is hospitalized. As a majority of patients at risk for ischemic heart disease should receive aspirin for primary prevention, this same concept applies in the hospitalized patient when receiving chemoprophylaxis. The challenge will be to change the current practice of using RS in VTE assignment, a practice that is both familiar to and commonly used by physicians.2,3,10,14
During the investigation, several ACGME core competencies were addressed by the resident participants. The most evident competency demonstrated by the residents was that of “practice-based learning and improvement.” Additionally, participation in the trial allowed for the residents to have a better understanding of “systems-based practice” in the determination of a more efficient, evidence-based approach that uses a physician order entry computer system. One can hypothesize that the effects of using the system described would translate to improved “patient care.” Finally, residents' participation allowed for improvement in their medical knowledge. Using a non-risk-stratifying VTE prophylaxis pathway allowed for guidance in selection of the most appropriate pharmacologic option for a given patient scenario, which educated the practitioner each time the protocol was completed as to what thought processes were needed to complete the task at hand based on evidence-based medicine.
The study was conducted in a VA health care system, where a majority of the patients are men. The favorable results may be generalizable to the older veteran population (the most at-risk age group1,5), but they may not be generalizable to all patient populations in community health care settings.
The survey in this study was provided to medical residents at an outpatient clinic. The number of participating residents was limited, and those who did participate may have been interested in and more open to a systematic protocol change than the other residents who did not volunteer for this study. One unanticipated outcome was that the residents multitasked, answering pages and taking care of patient-related issues after starting the survey (this occurred in 18 of 492 trials). Whereas the scenarios consisted of hospitalized patients, the testing environment was in an outpatient clinic, which may have influenced the results.
Although the six patient scenarios were fashioned to mirror real-life practice and appropriate VTE assignment based on evidence-derived guidelines, the residents may have had difficulty determining the appropriate assignment because of the complexity of the scenarios (e.g., multiple comorbidities or timing of surgery). These complexities and any inability to interpret the scenario may have influenced the residents' final Likert scores.
Although there were no obvious patient safety issues identified in this study, the study was done using fictitious scenarios and simulated order entry. Real-world application of this non-RS protocol would be needed to validate its safety.
Despite overwhelming evidence in favor of chemoprophylaxis, it continues to be underused in the hospital setting.3,4 Factors that negatively influence the use of VTE chemoprophylaxis among physicians include poor understanding of the significance of risk factors associated with VTE, efficacy doubt, lack of awareness of the recommendations, or preoccupation with the patient's bleeding risk.4 The limited formal pharmacology instruction for residents at the graduate medical education level18 may also contribute to suboptimal VTE assignment. Further reducing VTE chemoprophylaxis assignment is the perception that the guidelines are too complex and application is difficult.4 Thus, the non-RS protocol minimizes the dependence on the physician to recall the contraindications for these drugs.
There is currently a gap between known effective VTE guidelines and their actual use in hospitalized patients. This study has shown that residents reached the appropriate assignment more frequently and efficiently by following a decision-guided VTE protocol, which eliminated traditional RS as the first step. This was achieved through the implementation of the non-RS protocol, which was found to produce 40% faster resident completion times with a 33% improvement in VTE assignment. The non-RS protocol bridges the experience gap by giving the residents the template, reinforcing their current knowledge base, and enhancing VTE assignment training.
Additionally, through decision support, appropriate pharmacologic choices were made by all levels of trainees, which removes concern about potential errors being made by novice practitioners. Furthermore, by incorporating medical residents into treatment algorithms, the likelihood that there will be dissemination of evidence-based practices into the medical community will increase. It should be noted that we are currently in the process of implementing the user-friendly VTE pathway hospital-wide. Further analysis will be conducted after implementation of the protocol to address the underlying question of whether the pathway helped residents' decision-making skills and thereby affected the care of patients on a larger scale.
Based on these results, a user-friendly protocol that eliminates RS may reduce the inpatient incidence of VTE when used by residents and can increase adherence to VTE guidelines. VTE is the number one cause of preventable hospital death in the United States.14 If implemented across the United States, this simplified non-RS protocol could potentially save 20,000 lives per year with a cost savings of $2 billion/year of avoidable hospital costs.
The authors of this study would like to acknowledge the Nebraska Western Iowa Veterans Affairs Health Care System in Omaha, Nebraska, for their access and assistance in this study and the participating residents.
This report is the result of work supported with resources and the use of facilities at the Veterans Affairs Nebraska–Western Iowa Health Care System.
Ethical approval was granted by the Veterans Affairs Health Care System institutional review board.
The contents of this report do not represent the views of the Department of Veterans Affairs or the United States government.
A modified version of the study topic was presented at the 54th annual Human Factors and Ergonomics Society conference in San Francisco, California, in 2010, and a poster of the findings was presented at the VHA eHealth University Annual Conference in Las Vegas, Nevada, in 2010.