Economic burden and cost determinants of deep vein thrombosis during 2 years following diagnosis: a prospective evaluation
A portion of this work was presented at the 52nd Annual Meeting of the American Society of Hematology held on 4–7 December 2010 in Orlando, Florida.
Susan R. Kahn, Center of Clinical Epidemiology and Community Studies, Sir Mortimer B. Davis Jewish General Hospital, 5790 Cote des Neiges Rm. H420.1, Montreal, Quebec, Canada H3T 1Y9.
Tel.: +1 514 340 8222 4667; fax: +1 514 340 7564.
Summary. Background: Few studies have evaluated the long-term economic consequences of deep vein thrombosis (DVT). None of them have incorporated prospectively collected clinical data to ensure accurate identification of incident cases of DVT and DVT-related health outcomes of interest, such as post-thrombotic syndrome (PTS). Objectives: To prospectively quantify medical and non-medical resource use and costs related to DVT during 2 years following diagnosis, and to identify clinical determinants of costs. Methods: Three hundred and fifty-five consecutive patients with acute DVT were recruited at seven Canadian hospital centers. Resource use and cost information were retrieved from three sources: weekly patient-completed cost diaries, nurse-completed case report forms, and the Quebec provincial administrative healthcare database (RAMQ). Results: The rate of DVT-related hospitalization was 3.5 per 100 patient-years (95% confidence interval [CI] 2.2–4.9). Patients reported a mean (standard deviation) of 15.0 (14.5) physician visits and 0.7 (1.2) other healthcare professional visits. The average cost of DVT was $5180 (95% CI $4344–6017) in Canadian dollars, with 51.6% of costs being attributable to non-medical resource use. Multivariate analysis identified four independent predictors of costs: concomitant pulmonary embolism (relative increase in cost [RIC] 3.16; 95% CI 2.18–4.58), unprovoked DVT (RIC 1.65; 95% CI 1.28–2.13), development of PTS during follow-up (RIC 1.35; 95% CI 1.05–1.74), and management of DVT in the inpatient setting (RIC 1.79; 95% CI 1.33–2.40). Conclusions: The economic burden of DVT is substantial. The use of measures to prevent the occurrence of PTS and favoring outpatient care of DVT has the potential to diminish costs.
Deep vein thrombosis (DVT) is a common and serious vascular condition . The clinical burden of DVT often extends beyond the initial event, with increased risks of recurrent venous thromboembolism (VTE), bleeding while on anticoagulant therapy, and development of the chronic post-thrombotic syndrome (PTS).
Few studies have evaluated the long-term economic consequences of DVT. Studies to date have relied on retrospective administrative claims database analysis [2–5] or Markov modeling , and assessed medical costs only. Non-medical costs, including productivity losses that impact on patients, their families, and society, have not been adequately considered. None has incorporated prospectively collected clinical data to ensure accurate identification of incident cases of DVT and DVT-related health outcomes of interest, such as PTS.
We conducted a study to prospectively quantify medical and non-medical costs associated with DVT during the 2 years subsequent to diagnosis. A secondary study aim was to identify the principal determinants of costs following DVT. Specifically, we hypothesized that the development of PTS was associated with higher healthcare resource use and costs.
Materials and methods
The Venous Thrombosis Outcomes (VETO) study is a prospective multicenter cohort study whose primary objectives were to determine the incidence and risk factors for the development of PTS in patients with acute symptomatic DVT and the effects of PTS on health-related quality of life [7,8]. Assessment of the economic burden of DVT and PTS was another prespecified objective.
Patients were recruited from among consecutive patients with acute DVT seen in the emergency departments, outpatient clinics and inpatient wards of seven university-affiliated hospitals in Quebec and one hospital in Ontario from April 2001 to September 2004. Patients were eligible to participate if they had symptomatic DVT that had been objectively diagnosed within the preceding 7 days . Patients were excluded if thrombosis was limited to the intramuscular calf veins, life-expectancy was ≤ 3 months, or they were unable to complete a questionnaire, were geographically inaccessible or unwilling to return for follow-up visits, or were unable or refused to provide informed consent. Ethics approval was obtained from the relevant committees at each participating hospital center. Written informed consent was obtained from all patients prior to study entry. The attending physicians made decisions regarding duration of anticoagulant therapy and use of compression stockings.
At baseline, demographic and clinical characteristics were collected. At follow-up visits (1, 4, 8, 12 and 24 months), data on use of anticoagulants were collected. In addition, the study nurse performed a standardized assessment for PTS using the Villalta scale, a validated clinical PTS measure that grades the severity of five patient-rated symptoms and six nurse-rated clinical signs. Patients were classified as having developed PTS if the Villalta score for the same leg as that of the index DVT was ≥ 5 on at least two visits starting at the 4-month visit or later, or was ≥ 5 at the final follow-up visit. Severity of PTS was defined as follows: 5–9, mild PTS; 10–14, moderate PTS; and ≥ 15 or the presence of a venous ulcer, severe PTS .
During follow-up, suspected recurrent VTEs were evaluated with objective diagnostic testing, using previously described published algorithms , and bleeding events were tracked. All VTE events, bleeding events and deaths were reviewed by the study adjudication committee. If a patient reported hospitalization during follow-up, a chart review was conducted to determine whether the hospitalization was related to VTE or its treatment.
Overview of economic analysis
The economic analysis was carried out from a societal perspective with a ‘Sum_Diagnosis_Specific’ approach , which tracked downstream resource use and costs related to DVT. Medical costs included hospitalizations, physician and non-physician visits, prescription and over-the-counter medications, elastic stockings, and assistive devices. Non-medical costs consisted of lost productivity caused by absence from work, assistance, and transportation. The study time horizon extended to 2 years after diagnosis of DVT. Resources consumed by each patient were valued by the use of individual patient-level information obtained from cost diaries and administrative databases. These costs were inflated by 16.3% to reflect year 2010 costs . When individual-level costing information was unavailable, costs were estimated by applying unit costs to the corresponding resource (Table 1). A discount rate of 5% was applied for costs occurring in the second year . All costs are reported in Canadian dollars (year 2010 average US exchange rate 1.03) . The analysis was restricted to study participants enrolled in Quebec, as Ontario residents were not covered by the same healthcare fund.
Table 1. Allocation of unit costs for resource valuation (representative of year 2010)
| Nurse, $/visit*||48.12||MSSS, FIQ|
| Physiotherapist, $/visit||44.28||Statistics Canada|
| Other healthcare professionals†, $/visit‡||40.92||CD|
| LMWH, $/day§||23.25||RAMQ|
| UFH, $/day¶||11.14||RAMQ|
| Warfarin, $/day§||0.64||RAMQ|
| OTC drugs, $/purchase||8.18||CD|
| Stockings, $/purchase|| ||CD|
| Prescription||60|| |
| Non-prescription||23.5|| |
| Assistive devices, $/purchase or rental||16||CD|
| Transportation, $/trip|| || |
| Private car**||8.2||CAA|
| Public transport||3.5||CD|
| Private disabled transport‡‡||88.52||Medicar|
| Loss of productivity, $/day||155.89||Statistics Canada|
| Assistance, $ per hour||22.27||Statistics Canada|
Data sources for the economic analysis
Administrative databases All Quebec residents are eligible for coverage under the provincial healthcare fund. Although this fund covers the cost of hospitalizations and physician visits, only some residents are covered for the cost of medications: those ≥ 65 years old, welfare recipients, and workers and their family members who do not have access to a private drug insurance plan.
Information on resource utilization and costs was retrieved from two provincial government healthcare Quebec databases. The database of the Régie de l’assurance maladie du Québec (RAMQ) collects data on outpatient prescription drugs dispensed by community pharmacies and inpatient and outpatient physician visits. The Med-Echo database provides information on all hospital admissions in Quebec. Each Quebec resident has a unique health insurance identifier number used to link individual-level information from the RAMQ and Med-Echo databases to the clinical data. Authorization to receive nominative data from the databases was obtained from the Commission d’accès à l’information du Québec and study patients.
Cost diary (CD) A patient-completed CD was developed specifically for the VETO study. Patients were asked to report resource utilization and expenditures related to DVT at the end of each week during the first year of follow-up, and then for three randomly chosen months during the second year. The following resources were captured in the CD: non-physician visits, over-the-counter medications, elastic compression stockings, assistive devices, transportation, and parking. The number of lost workdays and hours of assistance were also reported.
Some aspects of our CD’s validity were previously established . In brief, the coverage rate (defined as the ratio of the number of weeks covered by returned diaries divided by the expected total number of weeks that should be covered during the study period for each patient) was high, with 77.4% of patients reporting data for more than 90% of the study follow-up period. The mean number of physician visits reported by patients in the CD was comparable to that contained in the administrative RAMQ database (12.8 vs. 13.7, P = 0.89, respectively).
Medical resource use and costs
We only considered the cost of DVT-related hospitalizations. These were identified in the Med-Echo database by use of a principal ICD-9 (International Classification of Disease, 9th edition) diagnostic code (Table S1) and/or a nurse-completed case report form that reported VTE-related hospitalization (DVT, pulmonary embolism [PE] or bleeding episode related to anticoagulation). Hospitalization costs were based on an index called Niveau d’Intensité Relative des Ressources Utilisées provided in the Med-Echo database. Costs were computed by multiplying this index by the calendar year’s adjusted average cost of acute-care hospitalization in Quebec. The cost of baseline hospitalization for patients included in the study as inpatients was restricted to hospital stays for which DVT or PE were listed as the principal discharge ICD-9 diagnostic code.
DVT-related physician visits were identified in the RAMQ database by the use of specific ICD-9 codes (Table S1). Costs were calculated from the service fees paid to physicians by the RAMQ. Non-physician visits and associated costs were assessed according to patients’ specifications in the CD.
The pharmacologic management of DVT in most patients involves administration of a parenteral anticoagulant (heparin), in conjunction with and followed by daily oral warfarin . Specific drug identification numbers were used to identify in the RAMQ database each medication related to the treatment of DVT dispensed during follow-up and corresponding costs. Costs of anticoagulants for patients not covered by the RAMQ drug plan were calculated by multiplying the average daily price of each drug class (Table 1) by the duration of treatment reported in the CD.
Non-medical resource use and costs
Personal or family transportation costs were calculated according to the travel distance provided by patients in the CD multiplied by a flat rate of $0.47 per kilometer . Ambulance, private disabled transport, taxi, public transportation and parking costs were based on patient specifications in the CD.
Costs of lost productivity were estimated with the friction-cost method . The friction period, based on the duration of unemployment according to Statistics Canada’s Labor Force Survey, was set at 18 weeks . Costs were calculated by multiplying the number of lost workdays as reported in the patient diaries by the average daily wage of Canadian employees in 2010 . We did not seek to take into account the specific job position and income of patients in our study, as this is not currently recommended practice in Canada, partly because of ethical concerns about valuation of medical treatments on the basis of social position, and partly because of statistical concerns regarding overadjustment to employment and income data from a specific clinical sample . We performed a series of sensitivity analyses with varying assumptions related to cost allocations. Lost productivity was valued with the minimum Canadian wage ($9.50 per hour) instead of the average wage. We also stratified the average Canadian daily salary by age category (< 25, 25–55 and > 55 years) and sex. The costs of assistance were estimated by use of the average hourly wage of employees in Canada in 2010 .
For the descriptive analysis of study participants and resource use, we calculated means (standard deviations [SDs]) and proportions, where appropriate. Despite the skewness in distribution of costs, results are presented as means. The non-parametric bootstrap method was used to generate 95% confidence intervals (CIs) around mean cost . The association between patient characteristics and costs was tested with univariate and multiple regression analyses, with 11 clinically relevant candidate variables. Candidate variables included in our analysis included age, sex, body mass index, patient status at inclusion (inpatient or outpatient), concomitant PE, DVT localization (proximal or distal), history of previous VTE, feature of DVT (transient risk factor associated DVT, cancer-related DVT, or unprovoked DVT), and development of PTS during follow-up.
Costs were log transformed in the multivariate models. The regression coefficient (on the log[cost] scale) and corresponding 95% CI were then exponentiated back to the cost scale . The effect of individual determinants of costs is presented as the relative increase in cost (RIC) and corresponding 95% CI. Statistical analyses were performed with sas version 9.1.3 (SAS Institute, Cary, NC, USA).
Patient characteristics and clinical outcomes during follow-up
Of 387 patients enrolled in the VETO study , 359 were residents of the province of Quebec. Four did not have a health insurance identifier number; hence, 355 (92.8% of 387) patients were included in the analysis.
Baseline characteristics of the 355 patients comprising the study cohort are shown in Table 2. The mean age was 56.5 years, 50% were males, and 67% were outpatients. DVT was located in proximal venous segments in 58% of cases. Concurrent symptomatic PE was present in 15% of patients. More than one-third (36%) of the study population was retired. Of the 355 study patients, 60% were covered by the RAMQ drug plan. Patients in the drug plan were significantly older (62.2 vs. 48.0 years, P < 0.001), more likely to be employed (70% vs. 18%, P < 0.001) and to have comorbid conditions such as cancer, lung or heart disease (40% vs. 21%, P < 0.001) than non-covered patients.
Table 2. Baseline characteristics of study participants (n = 355)
|Demographic characteristics|| |
|Age, mean (SD)||56.5 (14.7)|
|Men, n (%)||178 (50.1)|
|Inpatients, n (%)||116 (32.7)|
|BMI (kg/m2), mean (SD)||27.5 (5.4)|
|Employment status, n (%)|
| Employed||181 (51.0)|
| Laid off||16 (4.5)|
| Retired||127 (35.8)|
| Student||3 (0.8)|
| Homemaker||28 (7.9)|
|Clinical characteristics, n (%)|
| Current smoker||65 (18.3)|
| Hypertension||90 (25.4)|
| Diabetes||28 (7.9)|
| High cholesterol||72 (20.3)|
| Asthma||35 (9.9)|
| Chronic obstructive lung disease||14 (3.9)|
| Angina or myocardial infarction||34 (9.6)|
| Stroke||11 (3.1)|
| Congestive heart failure||12 (3.4)|
|Characteristics of DVT|
| Localization of DVT, n (%)|| |
| Proximal||206 (58.0)|
| Distal||149 (42.0)|
| Type of DVT, n (%)|| |
| Cancer-related*||46 (13.0)|
| Transient risk factor†||145 (40.9)|
| Unprovoked‡||164 (46.2)|
| Concurrent symptomatic PE, n (%)||53 (14.9)|
| Previous DVT or PE, n (%)||77 (21.7)|
| Previous ipsilateral DVT, n (%)||39 (10.9)|
The mean (SD) durations of heparin and warfarin therapy were 7.4 days (6.0 days) and 5.4 months (2.5 months), respectively. During the 2-year follow-up period, 31 objectively confirmed recurrent VTE events occurred among 30 patients, and 64 bleeding episodes were observed among 45 patients. Thirty-six patients died during follow-up. PE was a contributory cause of death in six patients. The cumulative incidence of PTS was 48% (95% CI 42–55%) at 24 months. Among patients who developed PTS, 84 (59%) were classified as having mild PTS, 36 (25%) as having moderate PTS, and 23 (16%) as having severe PTS.
The results for medical and non-medical resource utilization are shown in Table 3. The rate of DVT-related hospitalization during 24 months was 7.0 per 100 patients (95% CI 4.4–9.7). Of 26 DVT-related hospitalizations, four were related to bleeding events, all gastrointestinal in origin. On average, there were 15.0 (14.5) physician visits and 0.7 (1.2) visits to other healthcare professionals during follow-up. Patients reported buying a mean of 1.2 (2.1) pairs of compression stockings. On average, 12.7 (9.2) transportation episodes to attend healthcare visits were reported. Patients reported missing 12.1 (39.8) workdays in 2 years. An average of 38.6 hours (138.0 hours) of assistance was required during follow-up.
Table 3. Average deep vein thrombosis (DVT)-related resource utilization during 2 years after diagnosis of DVT
| Hospitalization, per 100 patients (95% CI)*||3.9 (1.9–6.0)||1.4 (0.2–2.6)||1.7 (0.3–3.0)||7.0 (4.4–9.7)|
| Physician visits (SD)||8.9 (10.1)||3.8 (5.9)||2.2 (4.7)||15.0 (14.5)|
| Non-physician visits (SD)†||0.2 (0.4)||0.1 (0.2)||0.4 (0.6)||0.7 (1.2)|
| Stockings (SD)||0.5 (0.7)||0.3 (0.9)||0.4 (1.3)||1.2 (2.1)|
| Assistive devices (SD)‡||0.13 (0.39)||0.03 (0.18)||0.02 (0.31)||0.19 (0.54)|
| Transportation (SD)§||6.5 (4.2)||3.6 (3.8)||2.7 (4.6)||12.7 (9.2)|
| Loss of productivity, days (SD)||7.8 (18.2)||3.5 (19.6)||0.9 (9.8)||12.1 (39.8)|
| Assistance, hours (SD)||29.3 (115.6)||6.3 (26.8)||3.1 (18.2)||38.6 (138.0)|
| Relatives||26.2 (112.4)||5.3 (25.4)||2.6 (16.8)||34.2 (129.7)|
| Hired||3.1 (14.2)||0.9 (6.5)||0.5 (4.6)||4.5 (18.7)|
The average total per-patient cost during the 2-year period was $5180 (95% CI $4344–6017), with 51.6% of costs ($2677; 95% CI $2101–3253) being attributable to use of non-medical resources (Table 4). More than two-thirds of all resource consumption occurred during the first 4 months after diagnosis. The largest components of costs were loss of productivity ($1543) and hospitalizations ($1504, baseline and follow-up hospitalizations).
Table 4. Mean deep vein thrombosis (DVT)-related costs (in Canadian dollars) during 2 years after diagnosis of DVT
| At baseline||954||–||–||954|
| During follow-up||305||117||127||550|
| Physician visits||244||107||59||411|
| Non-physician visits||12||10||5||27|
| Prescription drugs||219||154||27||398|
| Anticoagulants (parenteral)||150||68||1||219|
| Anticoagulants (oral)||38||28||8||74|
| OTC drugs||< 1||1||1||2|
| Assistive devices||5||2||0||6|
|Subtotal||1792 (1286–2297)||439 (286–592)||273 (153–393)||2503 (1921–3086)|
| Loss of productivity||1276||211||53||1543|
|Subtotal||2097 (1600, 2566)||421 (272, 558)||160 (88, 237)||2677 (2101, 3253)|
|Totals||3888 (3163–4614)||860 (641–1078)||432 (271–593)||5180 (4344–6017)|
We performed a series of sensitivity analyses, varying critical assumptions related to cost allocations. Valuing the time spent for assistance and lost productivity with the minimum Canadian wage ($9.50 per hour)  instead of the average wage ($22 per hour) decreased non-medical costs by 51.5% ($2677 to 1299). Valuing productivity losses with the average Canadian daily salary stratified by age category (< 25, 25–55 and > 55 years) and sex decreased non-medical costs by 0.9% ($2677 to 2654) .
Independent determinants of cost profiles
Univariate analyses Concomitant symptomatic PE at baseline, cancer-related DVT and unprovoked DVT (as opposed to transient risk factor-associated DVT) were associated with significantly higher medical costs (P < 0.001, P = 0.02, and P = 0.01, respectively; Table S2). In contrast, age > 65 years and unprovoked character of DVT were associated with significantly lower total costs (P = 0.02). This could reflect the way in which loss of productivity has been valued in our study. Indeed, we valued time lost from work but not the time for which patients were unable to perform other daily activities, hence inducing a decrease in patient-reported loss of productivity and total costs in elderly people. Development of PTS during follow-up was not significantly associated with higher medical or total costs, whereas being included as an inpatient at baseline was associated with higher total costs. Healthcare visits and prescription medication costs were significantly higher in patients who developed PTS (Table 5; P < 0.05).
Table 5. Costs, by category, in patients who developed post-thrombotic syndrome (PTS) as compared to patients without PTS†
| Hospitalization during follow-up||582||456||125 (− 360 to 610)|
| Healthcare visits||483||396||87 (4–170)*|
| Prescription medications||518||315||203 (1–406)*|
| Anticoagulants (parenteral)||310||164||146 (− 40 to 331)|
| Anticoagulants (oral)||96||62||34 (7–61)**|
| Analgesics||114||90||24 (− 15 to 64)|
| Other devices||146||90||56 (− 8 to 120)|
|Subtotal||1716||1239||477 (− 150 to 1150)|
| Transportation||318||276||41 (− 41 to 125)|
| Loss of productivity||1647||1628||21 (− 745 to 787)|
| Assistance||853||949||− 96 (− 774 to 583)|
|Subtotal||2811||2845||− 33 (− 1216 to 1150)|
|Total||4527||4083||444 (− 934 to 1822)|
Multivariate regression models Table 6 shows the results of multivariate regression models that used medical costs (models 1 and 2) and total costs (models 3 and 4) as dependent variables. Models 1 and 3 were tested with candidate baseline predictors only. Models 2 and 4 were tested with candidate baseline predictors and also took into account development of PTS during follow-up. We identified four independent predictors of medical costs: concomitant PE (RIC 3.16; 95% CI 2.18–4.58), unprovoked DVT (RIC 1.65; 95% CI 1.28–2.13), development of PTS during follow-up (RIC 1.35; 95% CI 1.05–1.74), and management of DVT in the inpatient setting (RIC 1.79; 95% CI 1.33–2.40). A relative increase in cost of 1.65 for unprovoked DVT means that costs were increased by 65% for patients with unprovoked DVT as compared with transient risk factor-associated DVT. Concomitant PE at baseline, management of DVT in the inpatient setting and development of PTS during follow-up were also independently predictive of higher total costs in all models. Age > 65 years was associated with lower total costs.
Table 6. Multivariate linear regression analysis of predictors of log-transformed 2-year deep vein thrombosis (DVT)-related costs†
|Age (reference < 65 years)||–||–||–||–|
| ≥ 65 years||0.84 (0.56–1.24)||0.77 (0.52–1.14)||0.50 (0.33–0.76)**||0.44 (0.29–0.67)**|
|Inpatient at baseline||1.84 (1.37–2.47)**||1.79 (1.33–2.40)**||1.75 (1.30–2.35)**||1.68 (1.23–2.30)**|
|Provoking feature of DVT|
| Transient (reference)||–||–||–||–|
| Cancer||1.20 (0.79–1.81)||1.20 (0.76–1.88)||0.74 (0.48–1.14)||0.84 (0.52–1.34)|
| Unprovoked||1.65 (1.28–2.13)**||1.65 (1.28–2.13)**||0.97 (0.74–1.28)||0.97 (0.74–1.28)|
|Associated PE||3.32 (2.29–4.82)**||3.16 (2.18–4.58)**||2.46 (1.66–3.64)**||2.29 (1.55–3.39)**|
|PTS during follow-up||–||1.35 (1.05–1.74)*||–||1.45 (1.12–1.87)**|
Sensitivity analyses with different assumptions for loss of productivity (minimum and average daily wage adjusted for age and sex) were conducted, and we obtained similar results in models 3 and 4. We obtained similar results in multiple regression analysis models including and excluding all patients who died (VTE was a contributory cause of death in six patients, three of whom reported being employed at the time of inclusion in the study).
In this prospective multicenter cohort study assessing costs and cost determinants of DVT, the average per-patient DVT-related cost during 2 years after diagnosis was $5180. We found that more than 50% of costs were attributable to non-medical resource use. Variables that predicted higher costs during the 2 years after DVT diagnosis included concomitant symptomatic PE at baseline, unprovoked DVT, development of PTS during follow-up, and management of DVT in the inpatient setting.
Our study has several strengths. Index DVT was objectively and prospectively identified in a well-defined clinical cohort. The 2-year follow-up period allowed sufficient time to observe the occurrence of complications such as PTS, which was assessed with a validated clinical measure. Linkage of prospectively collected clinical and economic data to administrative databases provided a powerful way of identifying determinants of costs with adjustment for potential confounders. Estimation of non-medical costs with a validated CD  provided a more comprehensive perspective on the cost of DVT. Allowing treating physicians to make decisions about the prescription of stockings and the duration of anticoagulation is likely to be more reflective of real-life management of DVT than protocol-guided evaluation of costs.
The costs of DVT have been studied in two US retrospective observational studies that used claims databases. Spyropoulos et al. identified hospital claims containing DVT or PE as the primary or secondary diagnosis from 1998 to 2004. VTE-related medical costs ranged from US$7594 to $10 804 during the year following diagnosis . The real-world medical costs of DVT may have been overestimated in this study. Indeed, only inpatients with DVT were included, whereas, in our study, two-thirds of subjects were outpatients at enrollment, a proportion that more closely reflects the current management of DVT. An analysis of US healthcare plan administrative claims from 1995 to 2005 reported DVT-related medical costs among 1500 patients discharged from hospital with a diagnosis of DVT or PE . Average per-patient costs were estimated at US$7712 for the initial hospitalization and US$2101 per year for patients readmitted to hospital during follow-up. Higher rates of hospital readmissions during follow-up (24.9% vs. 7.0%) and longer duration of initial parenteral therapy (13.0 vs. 7.4 days) were reported than in our study.
In our study, PTS was significantly associated with higher costs. In the multivariate regression analysis, we estimated that medical and total costs for patients who developed PTS during follow-up were 35–45% higher than for those without PTS. Our finding fits well with findings from related published studies. In a retrospective analysis of claim databases, the estimated mean adjusted incremental annualized cost of developing PTS was US$11 667 . However, the algorithm used to identify PTS was non-specific and based on ICD-9 codes related to leg pain, swelling, or varicosities. In a literature-based Markov model, Caprini et al.  estimated average per patient costs for treating mild-to-moderate PTS to be US$839 in the first year and US$341 per year thereafter.
Unprovoked DVT (as compared with transient risk-associated DVT) was associated with higher medical costs. This was not described in previous studies, and might be related to the fact that this patient population is more frequently tested for thrombophilia , is treated with antithrombotic drugs for a longer period of time [23,24], and is at higher risk for VTE recurrence . The higher cost of DVT with concomitant PE has already been described in the Spyropoulos study , and is probably explained by the fact that PE patients are traditionally treated in hospital.
We found that being included as an inpatient at baseline was associated with significantly higher total costs ($5263 vs. $3713, P = 0.03) in univariate analysis and remained significant in multivariate regression analysis (RIC 1.68, 95% CI 1.23–2.30). This indicates that potential savings could be realized with appropriate outpatient treatment of VTE, as many patients are still admitted routinely in the real world.
It is instructive to compare our estimates of the annual medical and non-medical costs incurred by DVT with estimates for other common medical conditions. For example, in a study of the cost of atrial fibrillation, Jonsson et al.  reported higher average total per-patient annual costs of $6697 in Sweden and $5355 in Germany (in 2005 Canadian dollars). Migliaccio-Walle et al.  compared the medical costs of peripheral arterial disease with the costs incurred by myocardial infarction (MI) during the 5 years after diagnosis in Canada. The average annualized postdiagnosis medical cost was $8394 for the year 2002, as compared with $9716 for the MI patients.
Our study has some limitations. Health service use and associated costs in our clinical setting may not be representative of other countries with different healthcare systems. Although the VETO study used broad inclusion criteria, more than half of the potentially eligible patients declined to participate; such patients were significantly older, and more likely to be female, to have proximal rather than distal DVT and to have other comorbid conditions than participants. We were not able to include all components of the costs of warfarin monitoring, as we did not have access to the number of International Normalized Ratio tests performed during follow-up.
In conclusion, the cost of DVT is substantial. Patients with concomitant symptomatic PE at baseline, with unprovoked DVT, who developed PTS during follow-up and included as inpatients had higher costs. Among these four predictors, PTS and inpatient status at baseline are the only ones that are preventable. Better adherence to thromboprophylaxis strategies, providing the optimal intensity and duration of anticoagulation, in order to reduce the risk of recurrent DVT, and the use of elastic compression stockings are important ways to prevent the occurrence of PTS. Favoring outpatient care of VTE for patients with isolated DVT or concomitant low-risk PE should also be promoted to reduce costs.
S. R. Kahn, I. Shrier, and M. Johri: study conception and design; R. Guanella, S. R. Kahn, I. Shrier, T. Ducruet, and M. Johri: analysis and interpretation of data; R. Guanella: drafting of the article. S. R. Kahn, M.-J. Miron, A. Roussin, S. Desmarais, F. Joyal, J. Kassis, S. Solymoss, and J. S. Ginsberg: provision of patients, and collection and assembly of data; T. Ducruet: statistical analysis; S. R. Kahn: obtaining funding. All authors were reponsible for critical revision of the article for important intellectual content and final approval of the article.
S. R. Kahn and I. Shrier are recipients of Clinical Investigator Awards from the Fonds de la Recherche en Santé du Québec. M. Johri is a recipient of a New Investigator Award from the Canadian Institutes of Health Research. J. S. Ginsberg is a recipient of a Career Award from the Heart and Stroke Foundation of Ontario, and holds the David Braley and Nancy Gordon Chair in Thromboembolic Disease at McMaster University.
We thank the VETO study personnel and the patients who participated in the VETO study. We thank L. Desjardins (deceased) for recruiting patients to the study. This paper is dedicated to the memory of Donna L. Lamping PhD.
Disclosure of Conflict of Interests
The VEnous Thrombosis Outcomes (VETO) study was funded by the Fonds de la Recherche en Santé du Québec and by an unrestricted grant-in-aid from GlaxoSmithKline. The study sponsors had no role in the design and conduct of the study, the collection, management, analysis, interpretation and reporting of study data, the writing of the manuscript, or the decision to submit the article for publication.