Published ahead of print on July 31, 2003, doi:10.1164/rccm.200303-367OC
© 2003 American Thoracic Society Cost-effectiveness of Ultrasound in Preventing Femoral Venous Catheter–associated Pulmonary EmbolismDivision of Pulmonary and Critical Care Medicine, Department of Medicine, Duke University Medical Center, Durham; Division of Pulmonary and Critical Care Medicine, Department of Medicine; and Department of Health Policy and Administration, School of Public Health, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina Correspondence and requests for reprints should be addressed to Christopher Cox, M.D., M.P.H., Division of Pulmonary and Critical Care Medicine, Department of Medicine, Duke University Medical Center, Box 3221 Durham, NC 27710. E-mail: christopher.cox{at}duke.edu
Femoral central venous catheter use is complicated by a high risk of deep venous thrombosis despite antithrombotic prophylaxis. Although some have recommended screening for femoral catheter–associated thrombosis to prevent pulmonary embolism (PE), this strategy's economic implications are unclear. Therefore, we used a decision model to evaluate the potential cost-effectiveness of a Doppler ultrasound–based screening strategy versus no ultrasound in averting thromboembolic complications associated with femoral catheters. The base-case analysis included a hypothetical cohort of 60-year-old medical patients treated for acute respiratory failure. The perspective was that of the health care payor, and the primary outcomes were quality-adjusted life expectancy, PE, and PE–associated deaths. The ultrasound strategy cost $8,688/quality-adjusted life-year (QALY) gained, $5,305/PE averted, and $99,286/PE death averted. The best- and worst-case scenarios, calculated in multiway sensitivity analyses by varying in-hospital mortality, deep venous thrombosis prevalence, and ultrasound accuracy, ranged from $1,170/QALY to $35,342/QALY, respectively. Probablistic analyses, in which variables with uncertain values were varied randomly within their ranges, demonstrated median costs of $12,793/QALY (interquartile range $8,176/QALY, $20,648/QALY). In summary, ultrasound screening may improve outcomes among the critically ill with femoral venous catheters at acceptable costs and could complement venous thrombosis primary prevention programs.
Key Words: cost-effectiveness • venous thrombosis • ultrasonography, Doppler • critical illness Almost half of the 31 million patients admitted to U.S. intensive care units (ICUs) annually receive central venous catheters (1). Although femoral vein placement of central venous catheters is associated with an incidence of deep venous thrombosis (DVT) that ranges from 8.5 to 26.2% (2–7) as well as a high risk of catheter-related bloodstream infections (3), this location may be used in as many as 20% of central vein catheterizations (8) because of perceived advantages relative to subclavian or internal jugular vein sites. These advantages include comparative ease of insertion, need for no postinsertion confirmatory imaging tests before use, low risk of immediate life-threatening complications, and convenience at an alternative location when upper extremity catheters are contraindicated or upper extremity veins already cannulated by other lines (9). The baseline prevalence of proximal femoral DVT has been reported to be as great as 30% among the critically ill (10). Although the use of sequential compression devices and low-dose anticoagulants reduces DVT incidence by approximately 50% (11), between 13 and 24% of patients still may develop DVT on such regimens (8, 10). Among patients with proximal femoral DVT, pulmonary embolism (PE) can be demonstrated in as many as 50% (12). Because PE mortality increases with the severity of hemodynamic and cardiopulmonary derangements at the time of presentation (13), the critically ill who require mechanical ventilation are at risk for especially poor outcomes (14). Even in the absence of documented PE, prospective studies have demonstrated that critically ill patients with DVT have higher mortality rates than those who do not (15). The clinical diagnosis of DVT in mechanically ventilated patients is complicated by common problems such as lower extremity edema and patients' inability to verbalize suggestive symptoms because of endotracheal intubation or sedation. In light of the poor accuracy of physical examination in detection of DVT (16), diagnostic testing is required. However, contrast venography, the criterion standard of DVT diagnosis, is rarely used in an ICU setting because of its invasive nature, need for patient transport to a radiology suite, and risk of nephrotoxicity (17). Bedside Doppler ultrasound is a more popular imaging modality (17) used to detect DVT among the critically ill because it is relatively inexpensive, noninvasive, yet provides reliable results that may significantly change the treatment course in many critically ill patients (15). Although ultrasound is highly accurate for detecting symptomatic proximal DVT (18), it is less sensitive for detecting DVT when suggestive clinical signs or symptoms are lacking (19), which is the more likely scenario among ICU patients. Nevertheless, some have suggested that ultrasound screening may provide a benefit for medical ICU patients (6, 7, 15). However, it is uncertain if basing the use of highly efficacious anticoagulant treatment (20) on ultrasound's lower positive predictive value among critically ill patients would lead to excessive costs and unacceptable treatment complications such as bleeding and thrombocytopenia. Because of the frequency with which clinicians use femoral venous catheters, the high incidence of DVT associated with femoral vein catheter placement (3), and the potentially important adverse outcomes of venous thromboembolic disease among the critically ill, we developed a decision model to assess the potential cost-effectiveness of routine lower extremity Doppler ultrasound performed in the setting of femoral catheter use in preventing PE and PE-associated death. We included sensitivity analyses to address the variability in baseline probabilities, costs of healthcare, as well as the accuracy of ultrasound testing.
Decision Model We performed a cost-effectiveness analysis by adopting a decision model approach and standard methodology (21) to estimate life expectancy and costs for patients with acute respiratory failure who require mechanical ventilation (Figure 1) . In our model, either an ultrasound strategy incorporating unilateral duplex Doppler examination of the proximal veins of the lower extremity catheterized by a femoral central venous line or no ultrasound was chosen. Patients found to have DVT, defined as a partial or complete occlusion of the proximal veins of the lower extremity (3) by an ultrasound performed after 7 days of catheterization, received a continuous infusion of unfractionated heparin at a dose of 30,000 U a day for 5 days (18 U/kg/hour for a 75 kg patient), followed by coumadin for 3 months.
Patients We analyzed outcomes among a hypothetical cohort of 1,000 patients with acute respiratory failure requiring mechanical ventilation. This population, of whom 75.1% were white, 12.5% Hispanic, and 12.4% African American, approximated the demographics of the most recent U.S. census (22). We assumed 61.3% were male, 38.7% were female, and the mean age was 60 years based on a recent international epidemiologic study of 5,183 patients requiring mechanical ventilation (23). We also conducted analyses for separate cohorts composed of 40-, 50-, 70-, and 80-year-old patients, as well as age-stratified analyses for males and females (each modeled as a composite of white, Hispanic, and African American survival rates). To account for the increased mortality observed among persons with respiratory failure in the first 2 years after hospital discharge, we adjusted age-appropriate survival estimates by the mean survival reported in cohort studies of persons who required mechanical ventilation for more than 2 days (24, 25). In addition, we adjusted survival of patients who were found to have DVT or PE as reported in a recent population-based study with over a decade of follow up (26), though it is assumed that all recurrent thromboembolic complications occurred within the first year after discharge because the presence of a venous catheter was a perceived reversible risk factor for thrombosis. We assumed an extra 8% decrement in survival for those who had either a DVT or PE that was undiagnosed and an additional 4% decrement in survival for those who were correctly diagnosed and treated (26). After this initial period of lower than average survival (see Table 1E in the online supplement), we assumed later mortality would parallel age-, ethnicity-, and sex-appropriate rates from reported 1998 census data (27).
Outcomes The primary outcome we assessed was quality-adjusted life expectancy. We adjusted survival by incorporating patients' values for health states, or utilities. These utility values, based on time-tradeoff methods and obtained from a large cohort of critically ill patients (28), were 0.73 for in-hospital survival, 0.79 from discharge until 2 months, and 0.88 afterward. We calculated incremental cost-effectiveness ratios, ratios of the differences in costs and the differences in effectiveness of the ultrasound and no ultrasound strategies. We measured incremental effect as the difference in quality-adjusted survival (measured in quality-adjusted life-years [QALYs]) between the two strategies. Secondary outcomes included the incidence of PE, PE-associated death, major fatal and nonfatal gastrointestinal bleeding, heparin-induced thrombocytopenia and heparin-induced thrombocytopenia–associated death. We analyzed outcomes from the perspective of the healthcare payor and discounted healthcare costs and benefits at a rate of 3% per year (29).
Model Assumptions
Probability and Cost Estimates
Sensitivity Analyses
Base-Case Analysis In the base-case analysis, performing screening ultrasound compared with no imaging after femoral catheter removal resulted in a small increase in overall costs from $16,409 to $16,538 (Table 2) . This slightly higher cost was associated with a gain of 0.016 QALYs (5.9 quality-adjusted life-days) and incremental costs of $129. The incremental cost-effectiveness ratio for 60-year-old male and female patients, reflecting costs per QALY gained by ultrasound screening followed by appropriate anticoagulant therapy, was $8,688 for the average simulated cohort member. In secondary analyses, use of ultrasound was associated with overall incremental costs of $5,305/catheter-related PE averted and $99,286/PE death averted. On the basis on our analyses, we would have to perform 38 ultrasounds to avert one PE, 694 to cause one major bleed, 719 to avert one PE death, and more than 100,000 to cause one bleeding-related death.
Sensitivity Analyses Results of one-way sensitivity analyses are shown in a tornado diagram (Figure 2) . The variables most sensitive to changes in incremental cost-effectiveness ratios in the model were probability of PE in the setting of DVT without anticoagulation, ultrasound sensitivity, the probability of catheter-associated DVT, and the probability of in-hospital death. Costs per QALY gained by ultrasound screening decreased significantly as either the sensitivity or specificity of ultrasound improved. For example, varying sensitivity from 33 to 87% changed incremental cost-effectiveness ratios from $20,370 to $4,852/QALY. Likewise, as specificity increased from 74 to 100%, costs per QALY fell from $11,428 to $7,224. Also, the probability of DVT was an important factor in our calculations, with costs per QALY increasing from $4,330 to $18,350 coincident with a decrease in DVT prevalence from 33 to 11%. Finally, we found the competing effect of ultrasound and physical examination sensitivity for DVT diagnosis on both costs/QALY (Figure 3) and PEs averted (Figure 4) was important. As physical examination accuracy improved and ultrasound accuracy declined, the number of PEs averted by the ultrasound strategy declined substantially, whereas costs/QALY gained increased.
Overall, complication rates, costs of diagnostic testing, costs of hospital care, and number of days required for treatment- and illness-associated complications had relatively small effects on incremental cost-effectiveness ratios. Varying daily ICU and ward costs by 50% produced variation less than $1,500/QALY, as did varying length of stay for complications of bleeding, PE, and heparin-induced thrombocytopenia. Only ultrasound costs had an impressive effect on incremental cost-effectiveness ratios, with a cost of $389 associated with $15,119/QALY and a cost of $151 associated with $4,852/QALY gained. Using base-case assumptions, ultrasound costs would have to exceed $1,200 to result in costs/QALY greater than $50,000.
In two-way sensitivity analyses, we found both ultrasound performance and patient characteristics had significant effects on incremental cost-effectiveness ratios. When we varied ultrasound sensitivity and specificity from minimal (33 and 74%) to maximal levels (87 and 100%), incremental costs decreased from $20,829 to $3,765 per QALY. Low sensitivity increased costs/QALY much more than low specificity. This was because the relative costs and mortality of PE, the incidence of which was increased in situations of low sensitivity, were greater than that of major bleeding and thrombocytopenia, which were increased in incidence preferentially under conditions of low specificity. Likewise, when the probability of death for cohort members was very low (15%) and the prevalence of DVT high (33%), costs per QALY of ultrasound were found to be $3,618. Costs increased to $43,075 with low DVT prevalence (11%) and high expected mortality (75%). Physical examination accuracy had a less important effect on outcomes and costs. At the highest tested levels of sensitivity (57%) and specificity (88%), the incremental cost-effectiveness ratio was slightly less than $18,000/QALY. We also analyzed the effect of varying the probabilities of DVT and PE simultaneously on incremental cost-effectiveness ratios and the number needed to treat (see Table 2E in the online supplement). At a typical probability of both DVT (20%) and PE (30%), the number of ultrasounds needed to prevent one PE is 18 times that needed to cause one bleed. Although one-third fewer ultrasounds are required to avert either a PE or PE death as DVT prevalence increases from 10 to 30%, the cost/QALY gained with this prevalence change decreases more sharply (by In three-way sensitivity analyses, we evaluated the worst-case scenario of a low prevalence of DVT (11%) and both poor sensitivity (33%) and specificity (74%) of ultrasound. In this case, an ultrasound screening strategy cost $28,736/QALY gained. On the other hand, the best-case scenario of high DVT prevalence (33%) as well as excellent sensitivity (87%) and specificity (100%) was associated with overall costs of $1,170/QALY. By incorporating high in-hospital mortality (50%), low testing accuracy (sensitivity 33% and specificity 74%), and low prevalence of DVT (11%), in four-way sensitivity analyses, we found that costs per QALY gained were $35,342. We varied all values found in Table 1 in Monte Carlo probabilistic n-way sensitivity analyses. The ultrasound strategy was associated with median costs of $12,793/QALY (interquartile range $8,176/QALY, $20,648/QALY) and resulted in costs of less than $44,000/QALY in 95% of simulations. The physical examination–based strategy was more cost-effective in less than 1% of simulations. We also analyzed the cost-effectiveness of the intervention by examining incremental cost-effectiveness ratios by age for males and females alone, as well as for the average cohort member (see Table 2E in the online supplement). The intervention resulted in costs of $8,862/QALY gained for 60-year-old males and $8,124/QALY gained for 60-year-old females. For an average cohort member, costs ranged from $6,834/QALY gained for 40-year-olds to $15,211/QALY gained for 80-year-olds. Incremental cost-effectiveness ratios for males and females were all within $1,000/QALY of the cohort member values. We found that a conservative estimate of the long-term effect of venous thrombosis on survival altered incremental cost-effectiveness ratios significantly. Without correcting for a postdischarge decrement in survival due to DVT and PE, the ultrasound strategy cost $9,875/QALY. By raising the mortality rate for all patients who developed DVT, the cost per QALY gained with an ultrasound strategy changed little. However, when the mortality rate among false negatives (DVT or PE present but not diagnosed) relative to true positives (DVT or PE present and diagnosed) was increased, incremental cost-effectiveness ratios declined significantly. For example, increasing the mortality rate among the false-negative patients from 8 to 10% while keeping the true-positive rate unchanged at 4% halved the base-case cost per QALY gained with an ultrasound strategy. When the false-negative rate was further increased by another third to 12%, the cost per QALY gained decreased to less than $1,500. These results are important because the long-term outcomes of those inadequately treated for DVT or PE are unknown and could be underestimated in this way by our model. Overall, our results suggest that an ultrasound-based intervention could have beneficial effects that persist for decades after the in-hospital screening period.
The critically ill have a high incidence of proximal femoral DVT (15) despite antithrombotic prophylactic regimens (8) that is increased further with femoral vein catheterization (3) and the requirement for mechanical ventilation (33). For these reasons, routine use of femoral central venous catheters in this population should be discouraged. However, because clinicians place millions of central venous catheters annually, a significant number of femoral catheters will be among these. In this light, some have suggested that an ultrasound-based screening strategy should be considered for the prevention of PE among the critically ill who have femoral catheters (7, 15). Using decision analysis modeling based on the probabilities most relevant to this population, we found that a Doppler ultrasound–based treatment strategy performed after removing femoral catheters was associated with potential costs of $8,688/QALY gained, $5,305/PE averted, and $99,286/life saved from PE death. By basing anticoagulation therapy on ultrasound results, 263 nonfatal and 14 fatal PEs potentially could be averted at the expense of 14 major episodes of gastrointestinal bleeding and 13 episodes of heparin-induced thrombocytopenia for every 10,000 ultrasound tests performed. Our results may be helpful to physicians engaging in decision-making because recent consensus group statements have not addressed the diagnosis and management of thromboembolic disease among the critically ill (34, 35). Despite the imperfect accuracy of ultrasound for detecting proximal femoral DVT in our target population, our findings support physicians' current clinical practice of using ultrasound as a first-line screening test for the presence of proximal DVT in a critically ill patient (33). The potential cost-effectiveness of the ultrasound screening strategy relative to standard of care is comparable with other commonly accepted interventions across a variety of clinical characteristics and health care costs. Base-case costs of $8,688/QALY are less than those required to provide mechanical ventilation to seriously ill patients at low risk of in-hospital death ($37,000/QALY) (36), treat two-vessel coronary artery disease with bypass surgery ($61,000/QALY) (37), or provide hemodialysis for critically ill patients with a better than 40% prognosis of surviving 6 months ($90,000/QALY) (38) (all values in 2001 U.S. dollars). Our costs per PE averted were lower than that found by others analyzing patients with trauma at high risk for PE ($59,000 per PE averted) (39), likely because of higher ultrasound costs used in their model ($1,300 in 2001 U.S. dollars), omission of costs associated with increased length of hospitalization for complications of undiagnosed DVT and PE, and the fact that these studies examined screening all patients—not just those with femoral venous catheters—for DVT. Our finding that high likelihood of in-hospital mortality had a significant effect on incremental cost-effectiveness ratios also complements the findings of others (40). The ultrasound screening strategy was cost-effective relative to a work-up based on physical examination because the incidence of PE and PE-associated death was reduced by 49%, yet the additional costs were only $129. But should clinicians change their practice on the basis of this analysis? We recognize that not all DVTs and PEs are believed to be clinically important among mechanically ventilated patients. However, nonfatal PEs may have different implications among those with acute respiratory failure than healthy, ambulatory persons (41). We believe it is also important to consider potential problems associated with nonfatal venous thromboembolism relevant to the critically ill that we did not address in our base-case analyses. For those who experience a PE and survive, pathophysiologic impairments may include increased work of breathing, increased dead space ventilation with resultant hypercapnea and hypoxemia, and right ventricular overload (42). These changes are likely to reduce weaning success and therefore increase days of mechanical ventilation, complications, and healthcare costs (43). We calculated that for every day that liberation from the mechanical ventilator was delayed by complications of PE, the cost per QALY gained by the ultrasound strategy decreased by $680. As the incidence of respiratory failure and demand for ICU beds rises in the coming years, this finding could have even greater significance. We also believe that the cost-effectiveness of the ultrasound strategy should be viewed in relation to DVT prevalence. We found that the probability of proximal femoral DVT had a more significant effect on incremental cost-effectiveness ratios than improvement in the accuracy of either ultrasound or physical examination. This suggests that ultimately, a more effective method of primary prevention of DVT may be the most efficacious and cost-effective strategy in reducing the burden of thromboembolic complications among the critically ill (11). Despite the potential benefits of ultrasound screening, there are logical concerns with its widespread implementation. First, we do not suggest that the use of ultrasound screening could make the placement of femoral catheters "more safe." Because placement of central venous catheters in upper extremity locations rather than the femoral vein has been shown to decrease the risk of venous thrombosis (3), avoiding lower extremity placement of catheters would therefore be the most efficacious strategy to avoid catheter-associated venous thrombosis. There is also the concern for bleeding related to more frequent use of anticoagulants among the critically ill. However, the two main factors in our model related to this concern, the rate of major bleeding and the specificity of ultrasound (low specificity leading to more false-positive tests and thus more frequent use of heparin), demonstrated relatively minor added costs and adverse effects across a range of values in sensitivity analyses. Even when the probability of major bleeding was as high as 5% in our model, more than double that reported in a recent clinical trial of heparin for acute PE (44), the cost/QALY gained using an ultrasound strategy was less than $12,000. Clearly, however, our model may be an oversimplification of the complexities clinicians experience when treating a critically ill patient with multiple comorbidities. It is important to consider our results in the context of the assumptions included in our decision model. First, cost-effectiveness analyses such as ours that address the critically ill often must make a number of assumptions in base-case calculations and test wide ranges of values in sensitivity analyses because of the relative lack of population-specific data (45). Indeed, although we report results stratified by age and sex in secondary analyses, we have used a composite cohort member in our base-case analyses in an effort to make our results applicable to a broader audience. Some readers may not agree with this methodology, preferring a base case defined by more specific demographics as well as fewer assumptions related to thromboembolic disease. However, such outcomes' data may not become available soon because of physicians' perceived risk of enrolling critically ill patients in clinical trials examining PE incidence and treatment that involve contrast imaging, the gold standard of diagnosis. Nevertheless, we have used estimates for PE incidence and PE death in our model that are significantly lower than past cost-effectiveness analyses (39, 46, 47) to bias the model against an ultrasound-based strategy. Also, we did not address specifically the relationship between duration of catheterization and the likelihood of thrombosis in our decision model. However, there is controversy regarding the importance of in situ days of catheterization in the pathogenesis of thrombosis, with some demonstrating clot as soon as 1 day (5). Therefore, it is not clear that there is a true "safe period" for catheterization of the femoral vein. Finally, our model may actually underestimate the full impact of catheter-associated DVT because we did not incorporate complications of catheter-related bloodstream infection, which is associated with both catheter-related thrombosis among the critically ill (48) and with increased in-hospital mortality and costs (49). Overall, we believe one of the unique benefits of our model is that it can demonstrate potential outcomes based on varying clinical probabilities and costs over a wide range of values that clinicians could use to aid decision making in the absence of a definitive randomized clinical trial.
Conclusions
C.E.C. has no declared conflict of interest; S.S.C. has no declared conflict of interest; A.K.B. has no declared conflict of interest.
Supported by a National Research Service Award Clinical Research Grant during the study (C.E.C). This article has an online supplement, which is accessible from this issue's table of contents online at www.atsjournals.org Received in original form March 12, 2003; accepted in final form July 30, 2003
This article has been cited by other articles:
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
TOP
ABSTRACT
METHODS
75%). 
