A literature-based Markov model using probabilistic sensitivity analysis via Monte Carlo simulation was developed to project the clinical, QOL, and economic burden of long-term complications of a primary DVT following THRS in the United States. The target population for this analysis was composed of two hypothetical cohorts of patients. Both cohorts were similar in age at 72 years old and in sex at 65% women to the population of all patients undergoing THRS in the United States as reported in the 1995–1996 Health Care Utilization Project (HCUP, Agency for Healthcare Research and Quality, Rockville, MD). However, one of the two cohorts was representative of patients who would have developed and survived a primary DVT following a THRS (i.e., the cases), while the other was representative of patients not experiencing a DVT following THRS (i.e., the controls). By comparing the lifetime costs and survival of cases and controls, the model allows the estimation of the long-term economic cost and survival impact attributable to the development of a primary DVT following THRS.
The analysis was performed from the perspective of payers responsible for the direct medical costs in both the inpatient and the outpatient setting. Indirect costs, such as time loss to seek care and productivity losses, were therefore not included in this analysis. We included time preference by discounting future economic and health consequences at an annual rate of 3%.
The Markov model depicts the natural history of post-DVT complications (i.e., PTS, recurrent VTE, and death) as an evolving sequence of 16 “health states” defined to capture important traits (i.e., QOL and cost) of these complications (Fig. 1). Of these 16 health states, 13 were actually used in the simulation and the additional 3 were included to help frame the clinical problem at hand. Specifically, following THRS, an individual can experience a DVT and survive the short-term period following this event, experience DVT but die from its complications from PE or treatment-related complications, or experience no DVT and therefore survive. Once patients have gone through these three states, the simulation effectively begins with the survivors in the two main groups: the “cases,” patients who survived a primary DVT following surgery, and the “controls,” patients who survived without experiencing a primary DVT following surgery. As time passes, individuals who have survived a primary DVT after surgery (i.e., the cases) can remain in this post-DVT state, develop signs and symptoms of mild-to-moderate PTS, signs and symptoms of severe PTS, or die. Patients who did not experience a primary DVT after surgery (i.e., the controls) are also assumed to experience idiopathic PTS, albeit less frequently. The model established a distinction between the first and subsequent years after the development of PTS to allow for differences in diagnostic and treatment patterns and associated costs that occur at different rates in the first and subsequent years. Once patients entered in the PTS states, they remained in these states until they died or the end of the simulation at the age of 100 years. Because of limited epidemiologic data, the model further assumed no movements of patients between the mild-to-moderate PTS and severe PTS states. Finally, patients also incur costs as a result of VTE events, which were accounted for as isolated clinical events rather than as health states. VTE also occurred relatively more frequently among patients with a history of DVT.
The definition of PTS used in the model followed the clinical-etiologic-anatomic distribution-pathophysiologic (CEAP) dysfunction system . As its name indicates, the clinical classification of the CEAP is based on a combination of etiologic classification (congenital, primary, or secondary), anatomic distribution (superficial, deep, or perforating, including subclassifications for each), and pathophysiologic dysfunction (reflux, obstruction, or both). Table 1 presents the six classes included in the CEAP. In the model, mild-to-moderate PTS and severe PTS refer to clinical classes 1–4 and 5–6, respectively. All patients developing severe PTS for the first time were classified as “severe, open ulcer,” following the assumption that an ulcer must be open before being healed. In addition, although an ulcer may heal (clinical class 5), patients who had developed an ulcer were assumed to remain in the severe PTS health state, alternating between healed and open ulcer status, for the rest of their lives.
Table 1. PTS by the CEAP classification system
|No visible or palpable signs of venous disease||0|
|Telangiectases, reticular veins, malleolar flare||1|
|Edema without skin changes||3|
|Skin changes ascribed to venous disease (e.g., pigmentation, |
venous eczema, lipodermatosclerosis)
|Skin changes as described above with healed ulceration||5|
|Skin changes as described above with active ulceration||6|
Data and Assumptions
Natural history and clinical outcomes. Data from Prandoni et al.  were used to estimate the incidence rate of PTS following a DVT (i.e., among cases). This 8-year observational prospective study followed 528 consecutive patients with a first episode of symptomatic, venographically confirmed DVT to monitor the development of complications of a primary DVT, such as PTS, recurrent VTE, and mortality. The rate of idiopathic PTS among controls was assumed to be similar to that of the general population and was taken from a population-based study in the United States .
Patients that developed leg ulcers because of venous insufficiency were considered to have severe PTS. Accordingly, all patients in whom severe PTS developed for the first time were considered by definition to have an open ulcer. Although ulcers can heal, the underlying condition remains, as the damage to the venous system is serious and irreversible. Therefore, from the second year onward, patients were assumed to remain in the severe PTS state, with alternating periods of time during which they would have open and healed ulcers. Specifically, at any given time after the first year with severe PTS, 31.5% and 68.5% of patients would have an open and a healed ulcer, respectively.
In addition, patients in the model were at risk for recurrent VTE. Prandoni et al.  provides the incidence rate of recurrent DVT and PE in cases. It was assumed that 20.8% of recurrent VTEs would be PEs . Finally, patients who did not develop a DVT following surgery were assumed to experience idiopathic DVT and PE at the incidence rates observed in the general population (i.e., 160 DVT per 100,000 and 70 PE per 100,000) .
Survival curves were constructed to model the life expectancy of patients in both groups. Patients who survive THRS surgery typically have a very favorable life expectancy [20,21]. In contrast, there is evidence that the long-term survival of patients who have experienced a DVT is compromised [3,9,16]. In the baseline model, survival curves for the first 15 years of the simulation were modeled based on an observational study of long-term complications and survival after acute DVT . This study found that the survival of patients who had an acute DVT was shorter than for age- and sex-matched controls that did not experience a DVT. To complete survival curves up to the age of 100 years, US life tables  were used, explicitly assuming patients in both groups would experience the life expectancy of the general population. In the sensitivity analyses, a scenario in which no difference in survival was assumed between cases and controls was conducted. In this scenario, all patients had a life expectancy identical to the age- and sex-matched control group as reported in Bergqvist et al.  for the first 15 years, and, for subsequent years, patient survival was based on the US life tables provided by Anderson .
Health-related quality-of-life adjustments. Life expectancy was adjusted for quality of life by using health state utilities (Table 2). Utilities represent an individual patient's preferences for a given health state and are scaled from 0 to 1. Quality-adjusted life years (QALY) are calculated by multiplying the time spent in a given health state by the utility value of that state. For the health state associated with no complications after DVT, weighted age- and sex-specific utilities obtained using the Quality of Well-Being Scale from a community sample of adults were used . QALY weights for mild-to-moderate PTS and severe PTS were based on standard gamble utilities obtained from healthy volunteers . Decrements in utility for recurrent VTE and treatment complications were expressed in days lost equivalent to the length of hospital stay .
Table 2. Utility weights
|Age- and sex-specific estimate|
Estimation of costs. The costs for diagnosis and treatment of PTS and VTE were estimated in several steps. First, patient care protocols were defined by the literature to specifically identify the key, direct cost resource use items required for managing patients with PTS. The cost was expressed in US$2000 for each of these protocols and was then determined. This was accomplished by first estimating the type and amount of resources that would be used with each protocol. Resource use items included physician/ nurse office visits, diagnostic tests, medical supplies, medications, hospitalizations, and surgeries. The annual cost of these complications was then obtained by multiplying the amount of resource use by their unit costs. Appendix A details the patient care protocols and associated costs for PTS.
Patient care protocol for mild-to-moderate PTS. Limited information is available in the literature regarding standard care protocols for patients with mild-to-moderate PTS. While several surgical procedures are available, conservative management is largely preferable . Diagnosis of mild-to-moderate PTS includes a clinical evaluation by a physician and ultrasonography of the vascular system in legs .The cornerstone of therapy is indefinite use of Grade 2 elastic compression stockings to reduce venous hypertension, improve tissue microcirculation, and prevent progression to severe PTS and open venous ulcers . Some patients may benefit from surgical intervention early on to prevent progression to severe PTS . While perceptions of managing PTS are slowly changing, the American Venous Forum recognizes that for many years venous disease has been considered the “stepchild” of vascular surgery, resulting in less than optimal care for many patients, such as clinic care by medical students for open ulcers . For primary care practitioners, training in the diagnosis and management of venous disease is lacking and diagnosis/intervention may not occur until patients progress beyond the mild/moderate stage, at which point they are then referred to a vascular surgeon . Therefore, because the symptoms of PTS often go unrecognized or are ignored for months or even years, we assumed that only about 50% of mild-to-moderate PTS patients would actually be identified and receive care for PTS. In sensitivity analyses, this proportion ranged from 25% to 75%. We also used the clinical expertise and experience of two of our authors, J.A.C. and A.T.C., to supplement the gaps in the literature such as selection of specific procedure codes for applying costs and specific frequency of some resources used.
For the initial work-up of mild-to-moderate PTS, 50% of patients would receive a level 5 (CPT 99215) physician office visit and a duplex ultrasound scan of the venous anatomy. A small proportion of patients equaling 7.5% without palpable pedal pulses would also receive an arterial Doppler as part of the differential diagnostic evaluation. Once diagnosis was confirmed, 50% of patients would receive Grade 2 elastic compression stockings, with 30 to 40 mmHg pressure at the ankle, as first-line management. During follow-up care in the first year, 50% of the patients would receive approximately 4 level 3 (CPT 99213) follow-up office visits. Some patients would also receive level 1 nurse visits (CPT 99211) for education or skin care. In addition, 20% of patients would receive vein legation and stripping.
In subsequent years, routine follow-up care consisted of regular physician office visits and appropriate nurse visits, periodic reevaluation with duplex ultrasound and arterial Doppler, and use of Grade 2 compression stockings.
Patient care protocol for severe PTS. For severe PTS, the comprehensive patient care protocol to achieve healing of an open ulcer was based on published guidelines for the management of venous leg ulcers, supplemented with expert panel opinion for utilization frequencies of various resources and selection of procedure codes for costing purposes [6,28]. All patients with severe PTS were assumed to seek and receive care.
Components of care included physician office visits, nurse visits for wound care, compression bandages (Unna's paste boot or long- or short- compression bandages), medications such as antibiotics, home health care, hospitalization, outpatient surgical debridement of wounds, and skin grafting . For costing severe, open-ulcer PTS, a study by Marston et al.  was used. This study outlined the care and associated costs required to heal an open ulcer based on its size. Of those patients with open ulcers, 36% had an ulcer size of < 5 cm2, 37% were size 5 to 20 cm2, and 27% were size > 20 cm2.
For patients with open ulcers, 75% received a prescription for a fluoroquinolone to manage infection on an outpatient basis as part of wound care. Hospitalization for severe cellulitis and/or uncontrolled edema was required in approximately 7% of severe, open-ulcer PTS patients . In some rare instances, patients were assumed to undergo amputation (0.4%), skin graft (2%), and debridement (2%). After ulcer healing in the first year, 10% of severe PTS patients would receive vein legation and stripping procedures and about 5% would receive subcutaneous endoscopic perforator surgery (SEPS).
Follow-up care for patients with healed ulcers included 4 level 4 (CPT 99214) physician office visits per year. Duplex ultrasound scan was repeated annually for all severe PTS patients and arterial Doppler was used for follow-up care in 20% of patients. All patients received Grade 2 elastic compression stockings as the primary maintenance measure after the ulcer healed.
Patient care protocol for VTE. Acute inpatient care costs for recurrent VTE (DVT and PE) were based on the average Medicare reimbursement for DRG 128 and 78, respectively. Follow-up outpatient care included oral anticoagulation therapy for three months [2,30] with warfarin (Dupont Pharmaceuticals, Wilmington, DE), using an average of 6 mg per day to maintain the international normalized ratio between 2 and 3, and an average of 9.25 prothrombin time laboratory tests and two outpatient follow-up visits.
Costing of patient care protocols. For inpatient care, costs consisted of the average Medicare DRG payment  plus the physician fees for the number of days of inpatient follow-up care based on the mean length of stay published with the DRG. Physician fees were based on designated CPT-4 codes. Medicare reimbursements for physician fees were gathered from Medicode's National Fee Analyzer .
For all outpatient care, such as vascular laboratory testing, office visits, and minor surgeries, costs consisted of the Medicare reimbursement rates for physician or laboratory fees (based on CPT-4 codes) plus the outpatient facility fee for that procedure. Outpatient facility fees were gathered from the Federal Register using Medicare Ambulatory Payment Classification (APC) codes . Home healthcare equipment costs were from the Centers for Medicare and Medicaid Service's (formerly Health Care Financing Administration) Durable Medical Equipment, Prosthetics/Orthotics, and Supplies (DMEPOS) Fee Schedule . Jobst Compression stocking costs were estimated using retail prices. Drug costs were estimated using average wholesale prices (AWP) from the RedBook .
For severe PTS with open ulcers, a literature-based estimate of the comprehensive cost to heal ulcers was used . The average cost of out-patient care to heal an open ulcer (adjusted to US$2000 prices) was $1372 for an ulcer size of <5 cm2, $2045 for a size of 5 to 20 cm2, and $5567 for a size of >20 cm2. This study included Medicare reimbursement levels of physician fees for evaluation and nonsurgical procedures; blood, vascular, and other labs; and home health care. Hospital costs were used to estimate costs of wound dressing materials. No surgery was performed until ulcers healed. The hospital cost component of inpatient SEPS was also based on the literature .
The model projects the cumulative incidence of PTS and VTE, the average number of years of life spent in the various health states, the life expectancy and the quality-adjusted life expectancy of the patients in the case and control groups. The annual costs associated with each health state were estimated based on the treatment protocols. These cost estimates were subsequently used to project the lifetime costs of cases and controls and the subset of patients who developed PTS.
All analyses were performed using probabilistic sensitivity analysis via Monte Carlo simulation to account for parameter uncertainty [13,37]. This method generated 5000 runs or trials of the model using a different set of values for the model parameters selected from the assumed distribution of the parameters for each run. All variables in the model were included in the Monte Carlo simulation, with the exception of mean age of surgery and discount rate, which remained held at their baseline values, as recommended by Briggs . In general, the exact distribution of the value taken by the parameter was unavailable. In these cases, triangular distributions were used. In triangular distributions, the likeliest value, that is, the baseline value, falls between the minimum and maximum values, assumed to be ± 50% from baseline, forming a triangular-shaped distribution, which shows that values near the minimum and maximum are less likely to occur than those near the likeliest value [39,40]. However, in a few instances with sex-specific utility and utility for time spent with PTS, log normal distributions were used, following the transformation proposed by Doubilet et al. . The results of the 5000 simulations were then analyzed to determine mean and uncertainty intervals by taking the 2.5 and 97.5 percentile values to represent endpoints (for a 95% interval) for the outcomes of the model. Following Briggs , we use the term “uncertainty interval” as a generic term rather than employing the frequently used “confidence interval” or the Bayesian equivalent “credible interval.” We conducted additional univariate and multivariate sensitivity analyses to further 1) determine the robustness of the results and conclusions; 2) identify the parameters that contributed the most to the results; and 3) identify important model uncertainties. This was accomplished by testing the impact on results of adding or subtracting 50% from the baseline estimate of key parameters. Finally, the model was run using alternative discount rates (0% and 5%).
Finally, a separate scenario analysis was conducted to test the impact of setting survival for all cases equal to that of controls, as opposed to assuming that cases have a shorter life expectancy than controls.