Abstract
This study reviews the clinical and workforce impact of a suggested protocol designed for the management of suspected acute deep venous thrombosis (DVT) in patients seen after standard vascular laboratory business hours. The protocol included the use of Wells score, D-dimer and a single dose of therapeutic anticoagulant to defer venous duplex ultrasound (VDU) testing until routine business hours unless contraindicated. Information was collected on medical history, physical exam and the timing of any diagnostic studies and treatment provided. Over 15% of studies done after-hours were deemed unnecessary by our protocol and in every individual the results were negative for an acute DVT. There were no adverse events from a one-time dose of anticoagulant. Limiting emergency VDU coverage to evaluate for acute DVT based on a management protocol can eliminate unnecessary after-hours VDU testing without having a negative impact on patient care.
Introduction
The diagnosis of a suspected acute deep venous thrombosis (DVT) often follows an algorithm including determination of the Wells clinical score and a serum D-dimer level prior to requesting a venous duplex ultrasound evaluation (VDU) to confirm the diagnosis. 1 Although an accepted approach to providing safe cost-efficient care, it becomes an even more important approach after business hours when a vascular technologist is not as readily available.
Previous studies have demonstrated that limiting after-hours VDU access, either by terminating after-hours availability or via a mandated protocol prior to calling in a technologist, can be done safely and yields a substantial cost-savings.2,3 Furthermore, a bridging dose of anticoagulant has been shown to be effective and safe in allowing patients to delay VDU testing until the following morning, during standard business hours.
Within two institutions in our health-care system, we have differing access to after-hours VDU testing. At a 170-bed suburban hospital, 24-hour VDU testing is available via a technologist on call from 16:30 until 7:30 on weekdays and, on weekends, from 16:30 Friday until 7:30 the following Monday. The second institution is an 800-bed tertiary referral center and in this institution a vascular technologist staffs the laboratory from 7:30 until 22:00 Monday through Friday with no testing available from 22:00 until 7:30 the following morning. On weekends and holidays, they are on call from 8:00 until 19:00.
In both hospitals, the relevant physicians were provided a protocol (seen in Figure 1) for DVT evaluation that includes Wells clinical scoring, D-dimer serum testing with a bridging dose of anticoagulant (enoxaparin 1 mg/kg twice daily), unless contraindicated, followed by next morning VDU testing when indicated.
4
With this protocol a definitive diagnosis is not made after-hours but presumptive treatment is initiated when a DVT is suspected and not ruled out by a Wells score of 2 or less and a negative D-dimer test. The duplex exam is accomplished the following morning during business hours to confirm or rule-out the diagnosis. At present, the protocol as suggested requires no vascular specialist specific approval prior to calling in a vascular technologist during times of call coverage.
Suggested protocol for the management of after-hours suspected acute DVT. DVT, deep venous thrombosis; VDU, venous duplex ultrasound
We hypothesized that the protocol was not being routinely followed in practicality, but if it were to be utilized it would provide a significant cost-savings without negatively impacting patient care. In addition, we wished to evaluate whether a truncated night call coverage policy had a negative impact on patient care. While the protocol is directed towards outpatients where the data is more clear on the utility of D-dimer testing in the evaluation of a suspected acute DVT, we also reviewed inpatient data in the same fashion to determine the safety of such a protocol in that population.
Methods
Patients were selected from two hospitals within Indiana University Health – an 800-bed tertiary-quaternary care center and a 170-bed full-service suburban hospital. Patients were included from both the inpatient units and the emergency departments.
The vascular laboratory database, In Record Time (In Record Time LLC, Fenton, MI, USA), was searched by procedure name for patients who had an urgent (stat) VDU of a lower extremity ordered from June 2006, when our protocol was instituted, until October 2010. The larger hospital was served by a private vascular laboratory until May of 2008 with the same protocol but, unfortunately, no data was available within our database to evaluate those patients. Charts were reviewed on any patient who had a test ordered or performed between 16:00 and 7:00 on weekdays, a test performed between 7:00 and 9:30 (to catch any patients whose study was held over until the morning when those studies are performed as the top priority beginning at 7:00 by the vascular technologists), and one performed at any hour of a weekend or holiday.
Wells criteria for predicting the pretest probability of deep vein thrombosis
Adapted from Wells et al. 4
*Unless calf swelling was documented to be greater than 3 cm they were scored as if it was not with the likely underestimation of some scores
The D-dimer test employed in our institution is the quantitative HemosIL D-dimer assay (Instrumentation Laboratory, Bedford, MA, USA). Internal studies by the manufacturer demonstrated a 100% sensitivity and negative predictive value (NPV) with a cut-off value of <230 ng/mL. The specificity with the same cut-off value was 35.9%. A value greater than or equal to 230 ng/mL was considered a positive D-dimer for this study.
The study was approved by the Indiana University Institutional Review Board.
Results
Demographics
Demographics
DVT, deep venous thrombosis
Protocol adherence
D-dimer testing
D-dimer testing at each hospital
DVT, deep venous thrombosis; VDU, venous duplex ultrasound
*This individual had recently been diagnosed with an acute DVT and was already on therapeutic anticoagulation and should have been excluded from our study. Their duplex revealed acute on chronic DVT
VDU results in those who had D-dimer tested
DVT, deep venous thrombosis; VDU, venous duplex ultrasound
Acute DVT incidence
At the institution with unrestricted testing, 10.9% of all duplex ultrasounds were positive for acute DVT (of those with a Wells score ≤0: 2.9% were positive; Wells score 1–2: 11.8%; ≥3: 52.5%), while at the facility with limited hours 19.4% (with respective Wells scores of ≤0: 8.7%; 1–2: 21.3%; ≥3: 59.3%) were positive (P < 0.001). Overall, 7.4% of studies were delayed until the next morning (16.8% of studies at the facility where VDUs were limited and 1% where VDUs were always available). Thirty-two patients who had a VDU performed within our timeframes were diagnosed with a pulmonary embolism. Eighteen of these had an immediate chest computed tomography (CT) scan with duplex imaging when available, while 14 had a VDU-confirmed DVT followed by a chest CT scan. If the CT scan did not show a pulmonary embolism, VDU testing was performed to evaluate for an acute DVT based on technologist availability at the facility. Our vascular database was searched to see if any individual in this study had another VDU performed within our hospital system in the following three months and no individual with an initial negative VDU was noted to have an acute DVT in the subsequent three months.
Anticoagulation
The use of a bridging dose of anticoagulation was more frequently used at the hospital with restricted VDU access (15.4%) compared with that of the hospital with unlimited access (3.8%). A therapeutic dosage of enoxaparin (1 mg/kg twice daily) was used when a bridging anticoagulant was administered. Out of a total of 67 patients given a one-time dose of anticoagulant to protect the patient overnight, there was one minor bleeding event in an individual with hematuria who was subsequently diagnosed with bladder cancer. No other adverse events were noted from the administration of a one-time bridging dose of anticoagulant. Of those treated with a bridging dose of anticoagulation, 18 of 67 (26.9%) had a positive VDU confirming the diagnosis of acute DVT. There were no adverse outcomes noted as a consequence of delaying testing until the next business day.
Inpatient versus outpatient
The overall percentages of positive VDUs did not differ significantly between inpatients and outpatients: 53/344 (15.4%) of inpatients and 76/552 (13.8%) of outpatients. For inpatients with a positive VDU evaluation, the Wells score obtained was, respectively, ≤0 in 9.8% of cases; 1–2 in 15.4% of cases and ≥3 in 35.7% of cases. For outpatients proven to have a positive VDU for acute DVT, the Wells score was, respectively, ≤0 in 3.3% of cases; 1–2 in 16.8% of cases and ≥3 in 69% of cases.
Protocol predictive ability
Based on Wells score, D-dimer level and VDU results, we analyzed the predictive ability of our protocol. Comparisons between the inpatient and outpatient settings used chi-square tests, or Fisher's exact tests when the cell counts were 5 or lower. The overall sensitivity of our protocol was 98%. For inpatients, the protocol sensitivity was 93% and for outpatients 100%. There was no statistically significant difference in sensitivity between the inpatient and outpatient setting (P = 0.27). The overall specificity was 18%, with the specificity higher in the outpatient setting, 20 versus 9% in inpatients (P = 0.0487). The positive predictive value (PPV) of our protocol was 20% for both inpatients and outpatients with no significant difference in PPV between inpatient and outpatient settings (P = 0.99). The false-positive rate, calculated as 1-PPV, was 80%. The NPV was 98%; for inpatients it was 83% and 100% for outpatients with no significant difference in NPV between inpatient and outpatient settings (P = 0.13). The false-negative rate, calculated as 1-NPV, was 2%.
The smaller hospital averaged 10 callbacks per month for lower extremity duplex exams. The larger hospital which only has a technologist on call on weekends averaged 23 callbacks for 35 exams per month.
Discussion
Multiple studies have evaluated the outpatient evaluation and management of suspected acute DVT. A meta-analysis published in 2006 determined the prevalence of DVT in the low, moderate and high-risk outpatients, as determined by Wells scoring, to be 5.0, 17 and 53%, respectively. 5 Our outpatient data of 3.3, 16.8 and 69%, respectively, closely reflects these findings and helps to validate the reliability of the Wells clinical predictor model in our outpatient population. 1 Other investigations have demonstrated the safety of ruling out DVT in an outpatient with a low Wells score and negative D-dimer.4,6 Another study determined the safety of ruling out DVT in the outpatient setting with a moderate clinical suspicion in addition to those with a low suspicion based on their Wells score if the quantitative D-dimer is negative. 7 Other studies have since reinforced this management strategy.8,9
With only 37.7% of outpatients having had a D-dimer tested, it is clear that the suggested protocol is not being utilized to its full capacity. At the hospital with limited testing only 10.5% of patients had a D-dimer tested. There appeared to be an impression by the managing physicians that D-dimer testing was not helpful in a time-efficient manner to impact the treatment of acute DVT. Since a stat D-dimer in our system is available within one hour, that impression may be without validity. Nevertheless, the result for the patient is the same. As an alternative, these physicians appear to prefer to either perform a rudimentary duplex exam themselves or give a bridging dose of anticoagulant and have them tested by VDU in the morning. To improve the utilization of the protocol, the algorithm for managing acute DVT will likely have to be made mandatory and actively supported by the physicians who manage the patients to change the culture by which a possible acute DVT is evaluated. Alternatively, the hospital with unlimited access to a technologist has a culture wherein physicians do not feel a pressure to follow a potentially cost-effective and patient-safe protocol so that their threshold for ordering an ultrasound at any time is very low. Defensive medicine may play a role in the low threshold for VDU testing in both hospital settings. One option to aid in protocol utilization that we are pursuing currently would be a specific order set available online with the Wells clinical score readily available for calculation and providing for D-dimer level testing to be positive prior to ordering a VDU. However, even in those who had a D-dimer tested, this finding was not consistently used to guide further management decisions in accordance with the protocol.
Our results suggest that the protocol as provided would function to safely determine if a VDU is necessary. All 18.8% of outpatients who had D-dimer tested did not require VDU by our protocol (Wells ≤0 or 1–2 and negative D-dimer); however they all underwent VDU testing which was negative in every individual. If the remaining 62.3% of outpatients had a D-dimer tested many more VDUs could have been safely eliminated. The patient risk would be negligible and the cost savings significant particularly regarding manpower usage.
Previous work has suggested that in the inpatient settings, the NPV and sensitivity of D-dimer remain sufficiently high to safely rule out acute DVT but this must be modified by adjusting for those factors which impact the D-dimer results. Factoring in those patient characteristics which affect the D-dimer does make a useable algorithm more difficult for the clinician to follow. The factors that falsely elevate D-dimer levels include malignancy, trauma, recent surgery, ischemic heart disease, stroke, inflammation, liver disease, recent major bleeding, sepsis, increasing age and pregnancy.10–12 Heparin therapy and a very short delay between symptom onset and D-dimer testing may cause a falsely normal D-dimer level.13,14 As a result, the proportion of patients who can have a D-dimer aid in the diagnosis of acute DVT is much smaller in the inpatient setting.
While our protocol was intended only for outpatient use, we collected inpatient data to determine its possible utility for implementation in an inpatient setting. D-dimer was tested in only 13.1% of inpatients leading to sample sizes too small from which to draw conclusions. Of the inpatients who had a low or moderate probability of DVT and a negative D-dimer, all except one had a negative VDU. The NPV of 83% in the inpatient population was skewed due to one individual who tested positive for an acute DVT who had been diagnosed with a DVT 18 days prior and started on warfarin. Were this protocol to be implemented for inpatients, this patient would be outside any protocol since the patient was known to have an acute DVT and was being treated with anticoagulation which could result in a negative D-dimer result. 14 It would be hoped that the clinical manifestation of acute DVT were moderated by therapy which may explain the moderate Wells criteria noted in this patient.
At institutions of any size, it appears safe to minimize or eliminate on-call VDU studies when a one-time dose of anticoagulant would bridge someone to VDU as first-case during the next standard business hours. Of the 92.6% of studies that were performed after-hours, 32.1% were completed late on weekdays or late weekend when they could have waited until business hours had they been given a 12-hour bridging anticoagulant dose. No serious adverse events were associated with a one-time anticoagulant dose in the patients so treated in our study. Another option to consider would be a 24-hour dose of enoxaparin; however, no patients in our study received this dosage and so we cannot comment on the safety or efficacy of that dosing in this population. This extended dosing would substantially broaden the percentage of patients that would be candidates for our protocol.
The most efficacious staffing arrangement may differ between high-volume centers and those with lesser demand for VDUs. At our busier hospital, there is an evening-shift technologist in-house until 22:00 who performs an average of four VDUs after business hours daily from Monday through Friday. This arrangement was beneficial in two ways; it eliminated the need for at home call and lessened the burden of the day time staff by allowing this technologist to aid in completing the overflow studies from the daytime shift. This helps to ensure that the daytime shift can go home on time thereby improving overall job satisfaction and retention among technologists. The cost for this technologist including benefits is cost-effective with the number of studies performed during the time on call but it may be cost-prohibitive for a smaller volume institution.
At both hospitals on weekends and during holidays, a technologist needs to be available since a bridging dose of anticoagulant, unless given Sunday night, is inadequate to tie someone over until the next standard business day. Both hospitals provide a technologist on-call from home on weekends and holidays. The cost for calling in a technologist is time and one half pay per callback with a minimum of three hours of pay regardless of actual time worked. One option for weekend coverage at the higher-volume hospital is to have a technologist in-house for a truncated business day to eliminate the excess cost of repeatedly calling in a technologist from home throughout the day. This has not been tried at our institution but in our high-volume centers could prove a potential cost–benefit. Another strategy to eliminate unnecessary testing would be to require approval from a physician who oversees the vascular laboratory before a technologist is called in.
There are situations where a bridging dose of anticoagulant is contraindicated (patients with ongoing bleeding, some with recent trauma or surgery, uncontrolled severe hypertension, thrombocytopenia or other bleeding disorders) and this remains a group that is difficult to address. This group constitutes a very small minority in our experience and the cost of after-hours VDUs for these unusual patients would be low. If they are a candidate for an inferior vena cava filter then calling in a technologist for testing is reasonable; however as in all situations, if the results of a study will not immediately or urgently change management it seems reasonable to wait until standard business hours to perform it.
The main weakness of this study is its retrospective nature. While every attempt was made to determine the Wells score accurately, due to differences in documentation, some inconsistencies were inevitable. Documentation is quite standardized in both emergency departments which enabled Wells scores to be accurately recorded in the outpatient setting. In the inpatient setting this was more difficult. This was most pronounced with regard to swelling below the knee where the Wells score criteria is a limb 3 cm larger than the opposite leg. For the sake of our study unless the swelling was measured, which it often was and then accurately documented, it was not assumed to meet the criteria which could have caused an underestimation of some Wells scores. In instances where other documentation was not adequate to determine the Wells score, the patients were excluded from the study. Since our main criteria in finding patients for our study was the timing of a VDU, it is also possible that some patients were missed from our study due to delays in obtaining the next morning VDU. We estimate that the number of such patients would be quite low since the hospital with limited VDU access has a designated area near the emergency department to hold patients overnight for testing that needs to be completed in the morning. It is inevitable that some patients would refuse an overnight stay and some would have VDU testing done elsewhere or in a timeframe more convenient for them. These patients would be missed in our study as well but the numbers are likely to be small.
Conclusions
Our results validate the work of others demonstrating the safety of ruling out an acute DVT in the outpatient setting in patients who have a low or moderate risk based on their Wells clinical score and a negative D-dimer. The protocol including a bridging one-time dose of anticoagulant would be a safe way to manage patients with a suspected DVT seen after-hours. This study was an attempt to determine the real world utility of prospective studies regarding the use of Wells criteria and D-dimer in the evaluation of acute DVT. It demonstrates the challenges in having physicians utilize a protocol but did demonstrate its safety when properly followed. It is hoped that this study provides some support for a more universal adoption of safe and resource-efficient protocols for the diagnosis of acute DVT especially in the outpatient setting. Further investigation with higher volumes is necessary prior to drawing any conclusions about the safety of the protocol in the inpatient setting although our limited results suggest it may be safe at least in individuals at low risk.