Cost-effectiveness of ruling out deep venous thrombosis in primary care versus care as usual

Authors

  • A. J. TEN CATE-HOEK,

    1. Departments of Clinical Epidemiology and Medical Technology Assessment and Internal Medicine, Maastricht University Medical Center, Maastricht
    Search for more papers by this author
  • D. B. TOLL,

    1. University Medical Center Utrecht, Julius Center for Health Sciences and Primary Care, Utrecht
    Search for more papers by this author
  • H. R. BÜLLER,

    1. Academic Medical Center, Departments of Vascular Medicine and General Practice, University of Amsterdam, Amsterdam
    Search for more papers by this author
  • A. W. HOES,

    1. University Medical Center Utrecht, Julius Center for Health Sciences and Primary Care, Utrecht
    Search for more papers by this author
  • K. G. M. MOONS,

    1. University Medical Center Utrecht, Julius Center for Health Sciences and Primary Care, Utrecht
    Search for more papers by this author
  • R. OUDEGA,

    1. University Medical Center Utrecht, Julius Center for Health Sciences and Primary Care, Utrecht
    Search for more papers by this author
  • H. E. J. H. STOFFERS,

    1. Departments of Epidemiology and General Practice, School of Public Health and Primary Care (CAPHRI), Maastricht University Medical Center, Maastricht, the Netherlands
    Search for more papers by this author
  • E. F. Van Der VELDE,

    1. Academic Medical Center, Departments of Vascular Medicine and General Practice, University of Amsterdam, Amsterdam
    Search for more papers by this author
  • H. C. P. M. Van WEERT,

    1. Academic Medical Center, Departments of Vascular Medicine and General Practice, University of Amsterdam, Amsterdam
    Search for more papers by this author
  • M. H. PRINS,

    1. Departments of Clinical Epidemiology and Medical Technology Assessment and Internal Medicine, Maastricht University Medical Center, Maastricht
    2. Departments of Epidemiology and General Practice, School of Public Health and Primary Care (CAPHRI), Maastricht University Medical Center, Maastricht, the Netherlands
    Search for more papers by this author
  • M. A. JOORE

    1. Departments of Clinical Epidemiology and Medical Technology Assessment and Internal Medicine, Maastricht University Medical Center, Maastricht
    2. University Medical Center Utrecht, Julius Center for Health Sciences and Primary Care, Utrecht
    Search for more papers by this author

Manuela A. Joore, Department of Clinical Epidemiology and Medical Technology Assessment, University Hospital Maastricht, PO Box 5800, 6202 AZ Maastricht, the Netherlands.
Tel.: +31 433875434; fax: +31 433875559.
E-mail: m.joore@mumc.nl

Abstract

Summary. Background: Referral for ultrasound testing in all patients suspected of DVT is inefficient, because 80–90% have no DVT. Objective: To assess the incremental cost-effectiveness of a diagnostic strategy to select patients at first presentation in primary care based on a point of care D-dimer test combined with a clinical decision rule (AMUSE strategy), compared with hospital-based strategies. Patients/Methods: A Markov-type cost-effectiveness model with a societal perspective and a 5-year time horizon was used to compare the AMUSE strategy with hospital-based strategies. Data were derived from the AMUSE study (2005–2007), the literature, and a direct survey of costs (2005–2007). Results of base-case analysis: Adherence to the AMUSE strategy on average results in savings of €138 ($185) per patient at the expense of a very small health loss (0.002 QALYs) compared with the best hospital strategy. The iCER is €55 753($74 848). The cost-effectiveness acceptability curves show that the AMUSE strategy has the highest probability of being cost-effective. Results of sensitivity analysis: Results are sensitive to decreases in sensitivity of the diagnostic strategy, but are not sensitive to increase in age (range 30–80), the costs for health states, and events. Conclusion: A diagnostic management strategy based on a clinical decision rule and a point of care D-dimer assay to exclude DVT in primary care is not only safe, but also cost-effective as compared with hospital-based strategies.

Introduction

The yearly incidence of a first episode of venous thrombosis is 100/100 000 persons; two-thirds of these patients have deep venous thrombosis (DVT) [1]. As DVT is a potentially life-threatening disorder, current practise is to refer all patients presenting with complaints suspected of being DVT, to specialized diagnostic services for objective testing. These services are readily available, use non-invasive tests (such as ultrasonography and D-dimer testing) and provide the referring physician with the assurance that a DVT is not missed [2,3]. However, numerous studies have revealed that 80–90% of these referred patients do not have deep venous thrombosis [3,4]. Therefore, it would be ideal to safely exclude DVT at initial presentation in a large proportion of these patients, thereby avoiding referral, and hence saving costs. The recent introduction of rapid point-of-care D-dimer assays combined with a specific clinical decision rule makes it possible to realize a diagnostic work-up in a primary care setting [5–7].

In the AMUSE study (Amsterdam, Maastricht, Utrecht Study on venous thromboEmbolism), a recent management study of a large series of 1002 consecutive patients seen in primary care, the safety and efficiency of excluding DVT by the combination of a clinical decision rule and a point-of-care D-dimer assay, was evaluated [8] The clinical decision rule that was used was derived from earlier studies in primary care [7], but was not yet validated in a true management study. For each patient suspected of DVT a score was calculated using the clinical decision rule and patients were managed accordingly. The clinical decision rule included a rapid point-of-care D-dimer assay result (Clearview Simplify D-dimer assay®; Inverness Medical, Bedford, UK) [6,9,10].

Those with a score ≤ 3 were not referred for ultrasound, and received no anticoagulant treatment, but were instructed to contact their general practitioner in case of worsening symptoms. Patients with a score ≥ 4 were referred for ultrasound. All patients visited their general practitioner between day 5 and day 9 for re-evaluation. Three months after entering the study, all patients received a questionnaire addressing signs and symptoms of (recurrent) venous thromboembolism.

The management study showed that based on the use of a simple clinical decision rule and a point-of-care D-dimer assay the need for referral to secondary care of patients with clinically suspected DVT was reduced by 50%. Moreover, a low subsequent risk for venous thromboembolic events was found (1.4%). The findings of this management study indicate that primary care physicians now have a simple tool available to safely refute the diagnosis of deep venous thrombosis in a large proportion of their patients. The present study focuses on the long-term cost-effectiveness of a diagnostic strategy, as evaluated in the above-mentioned management study, as compared with usual care (based either on ultrasound alone or on ultrasound following an in-hospital rule).

Methods

Background

Decision modeling is an analytical method used to describe the essential events that occur over time. In the context of health economic evaluation, a decision analytical model uses mathematical relationships to define a series of possible consequences that could flow from a set of alternative options under evaluation [11]. A Markov model is a modeling approach in which the health condition under investigation is divided into a finite number of mutually exclusive health states. The time horizon of the model is divided into cycles. In each cycle, people may move from one health state to another. The probability of making such a movement is referred to as transition probability. Costs and health effects (quality of life) are incorporated into the model as a mean value per health state, or the transition to a health state. The cost and health outcomes of the model are determined by calculating time spent in a given health state. Markov models are particularly useful when the decision problem involves ongoing risk [12]. For an introduction to decision modeling for health economic evaluation the reader is referred to a series of articles [12–14] and a handbook [11].

Model description

A Markov model was constructed to simulate the course of events in a hypothetical cohort of 1002 persons who present to their primary care physician with signs and symptoms suggestive of DVT of the leg (equal to the cohort included in the prospective study). The cohort had the age (58 years) and sex distribution (375 males; 37%) as well as the proportion of patients with post-thrombotic syndrome (62; 6%) and previous venous thromboembolism (170; 17%) as observed in the prospective study. Future costs and effects were discounted to their present value by a rate of 4% and 1.5%, respectively, according to Dutch guidelines [15]. The cycle length of the model was set to 6 months, with a 5-year time horizon. The model was constructed to compare the expected 5-year costs and health effects of the following diagnostic strategies for suspected DVT.

  • 1The AMUSE rule in primary care, followed by ultrasound in hospital for patients with a score of 4 and higher (referred to as ‘AMUSE rule’).
  • 2Referral to the hospital with ultrasound for all patients (referred to as ‘Hospital’).
  • 3Referral to the hospital with clinical decision rule in hospital followed by ultrasound for patients with a clinical score of 2 and higher and elevated D-dimer (referred to as ‘Hospital rule’).

The ‘AMUSE rule’ in primary care consisted of a clinical decision rule including a rapid point-of-care D-dimer assay (Clearview Simplify D-dimer assay®; Inverness Medical) for all patients who were suspected of DVT [6,9,10]. Only patients with a score ≥ 4 were referred for ultrasound. In the strategy referred to as ‘hospital’, all patients who were suspected of DVT were sent to the hospital and had an ultrasound examination; no differential risk assessment was performed. For the strategy referred to as ‘hospital rule’, all patients who were suspected of DVT were referred to hospital and at the hospital emergency department a clinical decision rule [16] was applied. Subsequently, a laboratory D-dimer test was performed in patients with a high clinical probability (score 2 and over). If the D-dimer test was elevated, patients were additionally evaluated by means of ultrasound examination.

Current practise in general is characterized by variation and will consist of a combination of the hospital and the hospital rule strategies. Therefore we included a combined strategy based on a weighted average of these two strategies, to reflect current practise.

Model construction

Health states

The health states included in the model were: no history of DVT, post venous thromboembolism (Post VTE), post thrombotic syndrome (PTS), and central nervous system bleeding (CNS bleed). The probability of recurrent VTE is higher for persons that have experienced a prior event of VTE compared with the population risk; therefore the health state Post VTE was included in the model. The health states PTS and CNS bleed were included because these conditions impact on quality of life and costs. The following events were modelled: DVT, pulmonary embolism (PE), major (gastrointestinal) bleed and CNS bleed. The final absorbing state was Death. Figure 1 is a graphical presentation of the model structure.

Figure 1.

 Model structure.

Model assumptions

It was assumed that every patient with DVT or recurrent episodes of DVT and PE was treated for 6 months with anticoagulant medication. The population of patients with DVT roughly consists of patients with a first episode of VTE (80%) and patients with a recurrent VTE (20%) [17]. DVT if provoked is treated for 3 months, DVT if unprovoked is treated for 6 months, and a recurrent VTE for 12 months. In the Netherlands all patients with a first episode of PE are treated for 6 months. Therefore, the treatment period was assumed to be 6 months. The costs and consequences of the events DVT, PE, major bleed and CNS bleed are calculated for 6 months, after this period subjects move to one of the health states: Post VTE, PTS, CNS bleed or Death.

Probabilities

The probabilities associated with the diagnostic strategies and the consequences of treatment for VTE are given in Table 1. Data were derived from the prospective study [8], as well as from the literature. Background mortality was based on age-specific death rates from the Central Bureau of Statistics.

Table 1.   Transition probabilities per 6 months and strategy properties
ParameterMeanSEDistributionSource
  1. *The meta-analysis of Ten Cate-Hoek & Prins, 2005, was extended with data from the contributing studies: Wells et al., 2003; Bates et al., 2003; Anderson et al., 2003; Schutgens et al., 2003; Kearon et al., 2001; Janes & Ashford, 2001. A positive rule implicates ultrasound testing in hospital.

Transition probabilities
 VTE first cycle0.140.1BetaAMUSE [8]
 VTE from No DVT after 1st cycle0.0010.001White, 2003 [29]
 PTS resulting from Post VTE0.030,01Prandoni, 1996 [26]
 Major bleed in treated patients0.02100.0014DirichletLinkins, 2003 [27]
 Fatal bleed in treated patients0.00340.0006
 CNS bleed in treated patients0.00120.0003
 Recurrent VTE from Post VTE0.030.01BetaPrandoni, 1996 [26]
 Recurrent VTE from PTS0.030.01
 PE given VTE0.270.13AMUSE [8]
 Fatality of PE given VTE0.430.07Douketis, 1998 [28]
Strategy properties
 Sensitivity AMUSE strategy0.92650.0223Beta126/136; AMUSE [8]
 Sensitivity hospital strategy0.97700.016085/87; Wells, 2003 [16]
 Sensitivity hospital rule strategy0.96920.061787/812; Meta-analysis*
 Positive AMUSE rule0.50100.0158AMUSE [8]
 Positive hospital rule0.81820.0164Ten Cate-Hoek, 2005 [4,5,16,31–34]

Health effects

The health state utilities assigned to the different health states and events are presented in Table 2. The disutility from having PTS and from experiencing an event DVT was calculated from EQ5D (a generic quality of life measure) data in the prospective study. The quality of life of persons in the health states No DVT and Post VTE was assumed to be equal to the quality of life of persons in the general population. As utility weights, age-specific EQ5D norm values for the general population were used [18]. For the events PE and major bleed and the health state CNS bleed, Time Trade Off, in the absence of EQ5D utilities, patient values from the literature were used [19,20].

Table 2.   Utilities and disutilities for health states and events
UtilitiesMeanSEDistributionSource
  1. *The age-specific utility of the health states ‘No DVT’ and ‘Post VTE’ was subtracted by this disutility to obtain the utility for the health state PTS.

Health states
No DVT & Post VTE
 Age 18–190.940.02BetaAge-specific norm values, Kind, 1999 [18]
 Age 20–240.940.02
 Age 25–290.930.03
 Age 30–340.930.03
 Age 35–390.910.04
 Age 40–440.910.04
 Age 45–490.850.04
 Age 50–540.850.04
 Age 55–590.800.04
 Age 60–640.800.04
 Age 65–690.780.04
 Age 70–740.780.04
 Age 75–790.730.04
Disutility PTS*0.020.03BetaAMUSE [8]
CNS bleed0.330.01Van Dongen, 2004 [19]
Events
Deep venous thrombosis0.670.03BetaAMUSE [8]
Pulmonary embolism0.620.01Van Dongen, 2004 [19]
Major bleedEqual to pulmonary embolismBased on Locadia, 2004 [20]

Costs

All unit costs were based on actual costs or standard unit costs from the Dutch Cost Manual [21]. Costs were calculated to their 2004 value using price index figures from the Central Bureau of Statistics. Volumes of medical consumption were based on the prospective study and the literature, as well as on expert opinion. In Table 3 the mean costs for the three diagnostic strategies and for the health states and events are listed.

Table 3.   Summary of cost parameters per cycle of 6 months or per event
ParameterMean valueSourcesUncertainty
  1. *A beta pert distribution was used in the probabilistic sensitivity analysis.

Diagnostic strategies
 AMUSE strategy€168Various, see AppendixFixed
 Hospital strategy€251
 Hospital rule strategy€227
Travel for diagnosis
 Travel to GP€3Various, see AppendixSee Appendix
 Travel to hospital€7
Health states
 PTS€3247Various, see AppendixMinimum €140, maximum €10 580*
 CNS bleed€28 419Costs of nursing home admissionFixed
Events
 Incident PTS€3367Various, see AppendixMinimum €273, maximum €10 670*
 DVT€1322See Appendix
 PE€4210
 Major bleed€4211Minimum €1688, maximum €11 497*
 CNS bleed€11 281Bergman et al., 1995 [24,30]Fixed

The costs of the strategies included the costs of medical care and travel costs to the GP and/or the hospital. In the AMUSE strategy two GP consultations, the D-dimer point-of-care test, GP time to perform D-dimer testing, and, in case of referral based on a positive rule, ER visit, ultrasound and in-hospital laboratory procedures were included. In the hospital strategy only one GP consultation was included, while all patients received an ER visit and ultrasound. In the hospital rule strategy the ultrasound was limited to patients with a positive rule including a hospital D-dimer test. It was assumed that subjects in the health states No DVT and Post VTE did not experience any costs related to DVT. Subjects in the health states PTS and CNS bleed did experience costs.

To calculate the costs of diagnosing DVT after the initial presentation of complaints (in the following Markov cycles), it must be taken into account that of all patients suspected of DVT who present themselves to the general practitioner, only one out of seven is actually diagnosed with DVT [8]. Hence, the ratio ‘suspected DVT to documented DVT’ is 7:1. The incidence of VTE as found in the literature was multiplied by this ratio to obtain the number of patients that enter the diagnostic process, and thus generate costs. Details of the cost calculations are listed in the Appendix.

Analyses

We compared the cost-effectiveness of three diagnostic strategies: the AMUSE strategy, the hospital strategy and the hospital rule strategy. Incremental cost-effectiveness ratios (iCERs) were calculated, dividing the incremental costs by the incremental quality adjusted life years (QALYs). ICERs were calculated by comparing each strategy with the next most effective strategy. Whether a strategy is deemed efficient depends on how much society is willing to pay for a life year in perfect health, which is referred to as the ceiling ratio. In the Netherlands an informal ceiling ratio of €80 000 ($107 400) per QALY exists [15]. This is, however, a maximum ceiling ratio that applies when there is a high burden of disease. Although this may not directly apply to DVT, the complications of (missed) DVT, such as PE and PTS, and the side-effects of the treatment of DVT, such as CNS bleed, can be considered as serious conditions. The National Institute for Health and Clinical Excellence in the United Kingdom uses a ceiling ratio between £20 000 and £30 000 per QALY [22], which is roughly €40 000 ($53 700).

Parameter uncertainty surrounding the iCERs was handled probabilistically. This means that we assigned distributions to the model parameters, to reflect the second-order uncertainty in the estimation of that parameter [23]. Measures of variance were retrieved from the prospective study, the patient cohort or published literature and, if no other source was available, from expert opinion; see Table 1 for the assigned distributions. Parameter values were drawn at random from the assigned distributions, using Monte Carlo simulation with 5000 iterations. To illustrate the results of the simulation, cost-effectiveness acceptability curves (CEACs) were calculated [24,25]. For different ceiling ratios, the net monetary benefit was calculated for each strategy by subtracting the costs from the effects, multiplied by the ceiling ratio. The CEACs show the probability that a diagnostic strategy has the highest net monetary benefit, and thus is deemed cost-effective, given different ceiling ratios.

As uncertainty exists, there is always a chance that the ‘wrong’ decision will be made [11]. The EVPI is the expected value of obtaining perfect knowledge of the ‘true’ values of all parameters. We calculated the total EVPI by subtracting the net monetary benefit of the organizational format we would choose under conditions of uncertainty, from the net monetary benefit of the optimal decision we would make if we knew the ‘true’ parameter values.

One-way sensitivity analyses were performed to test the consistency of the results. Included in the one-way sensitivity analyses were the discount rate and age, as well as the following (partially) fixed cost parameters: the health states PTS and CNS bleed, and the events DVT, PE, major bleed and CNS bleed. Furthermore, we calculated the costs and the sensitivity of the AMUSE diagnostic strategy for the cost-effectiveness threshold of €40 000 ($52 924) per QALY.

Results

Base case analysis

The AMUSE strategy had both slightly lower costs and less QALYs than both ‘care as usual’ strategies. The hospital strategy was the most effective, and had the highest costs. The iCER of the hospital strategy vs. the hospital rule strategy amounts to €89 956 ($120 766). This indicates that, even based on a maximum threshold of €80 000 ($107 400), the hospital rule strategy is preferred to the hospital strategy. Therefore, we compared the AMUSE strategy to the hospital rule strategy. This resulted in on average a cost saving of €138 ($185), and a QALY loss of 0.002. The iCER is €55 753 ($74 848). If usual care consists of an equal mix of the hospital and hospital rule strategy, the iCER of the AMUSE strategy vs. usual care is €58 622 ($78 700) (see Table 4). Adopting a strict health care perspective by excluding the travel costs from the analyses did not alter the results (iCER AMUSE strategy vs. hospital rule strategy: €55 436).

Table 4.   Results of the cost-effectiveness analysis
Diagnostic strategyLife yearsQALYsCosts Incremental QALYsIncremental CostsIncremental cost-effectiveness ratio
  1. *As compared with the AMUSE strategy.

AMUSE4.87233.8532€3589    
    }−0.0025−€138€55 753
Hospital rule4.87743.8557€3727    
    }−0.0005−€41€89 956
Hospital4.87823.8562€3768    
Combination of hospital (50%) and hospital rule (50%)4.87783.8559€3747 −0.0027*−€158*€58 622*

The probabilistic sensitivity analysis showed that the simulation results of the cost and QALY outcomes of the diagnostic strategies were comparable (data not shown). The incremental costs and QALYs from the Monte Carlo simulation comparing the AMUSE strategy and the hospital rule strategy indicate cost savings [95% confidence interval €186–€115 ($246–$152) saved] at the expense of a QALY loss (95% confidence interval 0.0081 QALY lost to 0.0005 QALY gained; Fig. 2). The cost-effectiveness acceptability curves show that the hospital strategy has the lowest probability of being cost-effective for a threshold of the iCER up to €48 000 ($63 509). The AMUSE strategy has the highest probability of being cost-effective as long as society demands less than €80 000 ($105 848) compensation for a QALY loss. For a threshold of €40 000 ($52 924), the probability that the AMUSE strategy is cost-effective is 66%, while for a threshold of €80 000 ($105 848) it is 37%, Fig. 3.

Figure 2.

 Cost-effectiveness plane.

Figure 3.

 Cost-effectiveness acceptability curves.

Additional sensitivity analyses

The discount rate and age did not influence the result. Across a range of 30–80 years the iCER of the AMUSE strategy vs. the hospital rule strategy varied with only €1300 ($1720) (data not shown). In the prospective study it was found that the sensitivity of the AMUSE strategy was 0.9265. In the Netherlands, the threshold for cost-effectiveness amounts to a maximum of €40 000 ($52 924)/QALY [15], meaning that the loss of one QALY needs to be compensated for by at least €40 000 ($52 924) savings. Therefore, a calculation was made of which decrease in the sensitivity of the AMUSE strategy would result in this threshold iCER. This was found to be a sensitivity of 0.9032, a decrease of 2%. An increase in the costs of the AMUSE diagnostic strategy of €27–€195 ($36–$258) per patient, resulted in an iCER of €40 000 ($52 924). A wide range of variation in the costs for health states and events did not change our results substantially. The results of the sensitivity analyses are presented in Table 5.

Table 5.   Results of the sensitivity analyses
Parameter in the sensitivity analysisDiagnostic strategyQALYsCostsiCER
Base caseAMUSE3.8532€3589€55 753/QALY
Hospital rule3.8557€3727
UndiscountedAMUSE3.8820€3660€56 436/QALY
Hospital rule3.8845€3801
Sensitivity AMUSE strategy 0.9032 instead of 0.9265AMUSE3.8519€3574€40 000/QALY
Hospital rule3.8557€3727
Costs AMUSE strategy €195 instead of €168AMUSEAs in base case€3628€40 000/QALY
Hospital rule€3727
Costs of PTS high value €10 580 instead of €3247AMUSE€9919€68 767/QALY
Hospital rule€10 089
Costs of PTS low value €140 instead of €3247AMUSE€907€50 240/QALY
Hospital rule€1031
Costs of CNS bleed high value €56 838 instead of €28 419AMUSE€3637€56 706/QALY
Hospital rule€3777
Costs of CNS bleed low value €14 210 instead of €28 419AMUSE€3565€55 277/QALY
Hospital rule€3701
Costs of PTS incident high value €10 670 instead of €3367AMUSE€4025€58 972/QALY
Hospital rule€4171
Costs of PTS incident low value €274 instead of €3367AMUSE€3404€54 466/QALY
Hospital rule€3539
Costs of event DVT high value €2644 instead of €1322AMUSE€3819€60 499/QALY
Hospital rule€3969
Costs of event DVT low value €661 instead of €1322AMUSE€3474€53 379/QALY
Hospital rule€3606
Costs of event PE high value €8420 instead of €4210AMUSE€3607€51 427/QALY
Hospital rule€3735
Costs of event PE low value €2104 instead of €4210AMUSE€3580€57 916/QALY
Hospital rule€3723
Costs of event major bleed high value €11 497 instead of €4211AMUSE€3620€56 386/QALY
Hospital rule€3759
Costs of event major bleed low value €2104 instead of €1688AMUSE€3578€55 534/QALY
Hospital rule€3716
Costs of event CNS bleed high value €22 562 instead of €11 281AMUSE€3592€55 810/QALY
Hospital rule€3730
Costs of event CNS bleed low value €5641 instead of €11 281AMUSE€3588€55 725/QALY
Hospital rule€3725

Expected value of perfect information analysis

The uncertainty surrounding the decision of whether or not to adopt the AMUSE strategy resulted in an EVPI of €32 ($42) per person, given a ceiling ratio of €40 000 ($52 924)/QALY. The incidence of DVT is estimated to be 0.1% [1]. In the Netherlands annually 16 000 persons experience a DVT. The ratio incidence to complaint found in the prospective study was 1:7, which means that 112 000 persons present themselves to the GP with complaints each year. Adopting the AMUSE strategy would affect all these persons, being a total of 873 790 persons in the Netherlands in the next 10 years. This makes the population EVPI €27 million ($35 million), meaning that in the Netherlands perfect information on this topic is worth €27 million ($35 million). Figure 4 shows the EVPI for different ceiling ratios.

Figure 4.

 Expected value of perfect information analysis curve.

Discussion

The model-based cost-effectiveness analysis showed that the AMUSE strategy, which was deemed to be safe clinically, was associated with on average a minimal loss of QALYs (0.002, 95% confidence interval, 0.0081 QALY lost to 0.0005 QALY gained). The cost saving associated with the strategy was also relatively small [€138 ($185), 95% confidence interval, €186 saved to €115 saved ($246–$152)]. This resulted in a marginally acceptable iCER of the AMUSE strategy vs. the next best alternative, the hospital rule strategy of €55,753 ($73 767)/QALY. However, the cost-effectiveness acceptability curves show that the AMUSE strategy has a higher probability of being cost-effective than the alternatives for thresholds up to €80 000 ($105 848). For a threshold of €40 000 ($52 924), the probability that the AMUSE strategy is cost-effective is 66%. Hence, it can be concluded that the AMUSE strategy for the diagnosis of suspected deep vein thrombosis with a point-of-care D-dimer test combined with a clinical decision rule in primary care has the highest probability of being cost-effective.

Some details of our model study require attention. Some costs of health states and events were estimates based on expert opinion or are partially based on assumptions. However, the sensitivity analyses show that for a wide range of variation in these costs our results do not change substantially. The cost-effectiveness outcome is strongly influenced by the sensitivity of the diagnostic strategies. While the sensitivity of the hospital rule strategy was based on a meta-analysis including over 6000 patients, the estimates for the hospital and AMUSE strategies were based on studies including 1100 and 1000 patients, respectively. This resulted in larger uncertainty for the sensitivity of these strategies. In current practise it is likely that hospitals use a mix of the hospital and hospital rule strategy. Therefore, our results can be considered as conservative. Finally, it could be argued that results would be more favourable if a more sensitive D-dimer test combined with the decision rule was used in the AMUSE strategy. However, in that case also the specificity is likely to be lower, resulting in more patients referred to secondary care, and hence less cost savings.

For clinical practise, implementation of the AMUSE strategy results in the exclusion of DVT in approximately 50% of patients in primary care. This may have the added benefit of convenience for the patients, because referral for ultrasound is not necessary, and may enable the general practitioner to direct attention at finding alternative diagnoses without delay. Although the general practitioner, or the practise assistant, spends extra time to perform the D-dimer test and apply the rule (for which extra costs were included in our model), the prospective study proved that implementation in general practise was highly feasible. In order to facilitate successful implementation a reimbursement for the extra time of the general practitioner would be appropriate.

In summary, the AMUSE strategy to exclude DVT in primary care is not only safe, but also has a higher probability of being cost-effective as compared with hospital-based strategies, with or without a rule, in the diagnosis of DVT.

Addendum

Concept and study design: H.R. Büller, M.H. Prins, M.A. Joore, A.W. Hoes, K.G.M. Moons, R. Oudega, H.E.J.H. Stoffers, H.C.P.M. van Weert; data collection: D.B. Toll, E.F. van der Velde, A.J. ten Cate-Hoek; data analysis and interpretation: M.A. Joore, M.H. Prins, A.J. ten Cate-Hoek; manuscript drafting: A.J. ten Cate-Hoek, M.A. Joore, M.H. Prins; Critical writing and final approval of manuscript version to be published: H.R. Büller, M.H. Prins, A.W. Hoes, K.G.M. Moons, R. Oudega, H.E.J.H. Stoffers, E.F. van der Velde, H.C.P.M. van Weert, D.B. Toll, A.J. ten Cate-Hoek, M.A. Joore.

Acknowledgements

We gratefully acknowledge the enthusiastic participation of general practitioners and other health care professionals who contributed to the study. This study was supported by The Netherlands Organization for Scientific Research (ZON-MW), grant number 945-04-009.

Disclosure of Conflict of Interests

The authors state that they have no conflict of interest.

Ancillary