Cost-Effectiveness of Prevention and Treatment of the Diabetic Foot: A Markov analysis

A Markov model was built to simulate the health and economic outcomes of optimal care of the diabetic foot in a hypothetical population of diabetic patients. Markov models are well-recognized methods for analyzing clinical and economic consequences of medical decisions, particularly in long-term diseases characterized by repeating risks of events over time. Because of its chronic nature and the recurrent character of its major adverse outcomes, foot ulcers and amputations, diabetic foot disease fulfills these criteria.

In this study, diabetic foot pathology was modeled by 13 health states describing the spectrum of the disease from causes to consequences: three risk health states, six wound type states, and four outcome states were included. Cohorts of 10,000 patients with newly diagnosed type 2 diabetes were followed through the model individually over their lifetime. In the simulations, optimal treatment and prevention were compared with present level of care. The clinical outcomes measured were life expectancy, QALYs, and incidence of foot ulcers and amputations. The economic outcomes studied were mean expected total lifetime cost and incremental cost-effectiveness ratio (ICER).

Health states were defined considering their clinical relevance within the history of the disease and their association with different health outcomes and resource costs (Fig. 1). The first three health states represented pathophysiologic precursors for development of ulcers. The presence or absence of these factors confer different risks for ulceration and, thus, have been included in the model as diabetic foot risk (DFR) states. Based on the classification system of the International Working Group on the Diabetic Foot (IWGDF) (7), these states were defined as: DFR1, no neuropathy; DFR2, sensory neuropathy; and DFR3, sensory neuropathy and deformity or peripheral vascular disease. Previous ulcer or amputation, the fourth risk factor identified by the IWGDF, was represented in the model by the healed and postminor amputation states since the entrance to these states implied the occurrence of a previous episode of an ulcer or amputation.

The six wound health states reflect important characteristics of diabetic foot lesions, namely depth, presence of ischemia, and infection, and they attempt to reconcile two established wound classification systems: Wagner’s (15) and the University of Texas (16) classification systems. The remaining four health states encompass all possible outcomes of the diabetic foot disease: healed, amputation (minor or major), and death.

Patients progressed to other health states, remained in the same state, or died, depending on the associated probabilities for each transition. Six-month transition probabilities were assigned for movement between the health states. Mean age at entry was 61 years, corresponding to the mean age of diagnosis of type 2 diabetes in the Netherlands (17). All patients began the model in DFR1 (88.5%) or DFR2 (11.5%) states, concurring with the prevalence of sensory neuropathy at time of diagnosis of diabetes (18).

Information on probabilities of morbid and mortal events was derived from studies on incidence and prevalence of diabetic foot pathology published from 1996 to 2002. Priority was given to national published articles and prospective cohort studies to obtain the best reflection of the history of the disease in a Dutch setting. To ensure consistency, all international studies used pertained to the same research group: the University of Texas study group. Data derived from the U.K. Prospective Diabetes Study group were only used in the transition to sensory neuropathy. These studies were selected based on comparability with Dutch diabetic population characteristics, sample size, follow-up period, reporting on variables of interest, and overall quality. Transition probabilities and study sources are summarized in Table 1.

Core information was obtained from the only Dutch prospective cohort study on diabetic foot disease (17). An annual ulcer incidence rate of 2.1% and an amputation incidence rate of 0.6% were among the reference country-specific parameters derived from this study and adopted in the model.

Transition probabilities to the different ulcer states were specific for each of the four risk categories. Incidence rates per risk subgroup were calculated using information from two of the few prospective cohort studies reporting ulcer incidence following the risk classification system of the IWGDF (19,20). Ulcer distribution per risk group was derived from local and international studies (21–23). A recategorization of the wound types was performed to meet the health states definition of the model. The progressions and associated probabilities of the wound types were derived from longitudinal studies on clinical outcomes of foot ulcers (22,23). In general, once in a wound state, four possible transitions were allowed: the wound healed primarily, the wound healed after amputation, the lesion remained unhealed, or the patient died. Wound healing probabilities were calculated as the residual of the probability of the other transition(s). The healed and postminor amputation states were considered to have the same wound progressions and associated transition probabilities. In the postmajor amputation state, the transition to a new ulcer was not allowed; in this state, patients either remained in the same state or died.

Data were obtained from studies reporting on mortality of diabetic patients in the Netherlands (24,25). Mortality rates of the risk states were age specific. Age ranges per risk categories were calculated based on mean duration of diabetes reported in a previous study (19). Risk group 1 patients were assigned the mortality rate of the diabetic population in the Netherlands aged 55–64 years (newly diagnosed diabetic patients). Risk groups 2 and 3 were assigned the mortality of diabetic patients aged 65–74 years, and risk group 4 patients were assigned the mortality of diabetic patients aged ≥75 years (24). Similarly, these mortality rates were assigned to the wound states according to severity and risk category.

The treatment effects of conventional glycemic control and IGC simulated in the model were based on U.K. Prospective Diabetes Study results (18). The neuropathy rates reported for both groups were incorporated into our model in the transition from DFR1 (no neuropathy) to DFR2 (sensory neuropathy).

OFC according to guidelines includes professional protective foot care, education of patients and staff, regular inspection of the feet, identification of the high-risk patient, treatment of nonulcerative lesions, and a multidisciplinary approach to established foot ulcers. These strategies have been reported to decrease the incidence of LEA by 49–85% (6). Because the major goal of these interventions is to decrease the occurrence of foot pathology, the effect of OFC in the model was incorporated by reducing the probabilities of development of foot lesions. To overcome the uncertainty around the decrease in frequency of foot lesions required to attain a decrease in LEA, the probabilities describing the transitions to the different wound states were varied across a range of 10–90% and the effects on LEA rates were recorded.

The clinical and economic outcomes of the two intervention strategies contained in the guidelines (IGC and OFC) were analyzed separately and in combination to determine their individual weight and additive impact on the results.

Resource use associated with the various health states and transitions was obtained from a published study comparing standard and intensive treatment protocols in general practices in the Netherlands (26,27) and from an economic study on the treatment of diabetic foot ulcers (28). Only direct medical costs were considered. The costs of each health state were determined by the health care utilization associated with that state and included expenses such as labor, medication, laboratory, materials (shoes, insoles, contact casts), and procedure (diagnostic tests, debridement, bone resection) (Table 1 and Fig. 3). Costs used in the model were expressed as 1999 U.S. dollars and discounted by 3%.

The utility weights needed to calculate QALY were derived from a Dutch study on foot ulcers and amputation state valuations (29). Utility weights ranged from 0 to 1, where 0 represented a quality of life equal to death and 1 represented perfect health (Table 1).

Utilities were discounted by 3%.

Sensitivity analyses were performed on effectiveness of preventive foot care strategies (level of foot lesion reduction), utility weights, and costs. The utility weights were varied according to the results from a previous study on quality of life and foot ulcers (30). These values were adapted to meet our ulcer categorization. Costs were changed by varying the length of stay of inpatient care. The length of stay reported for inpatient care of foot ulcers in the U.S. was used as a reference in the sensitivity analysis (31). This model was constructed using TreeAge (DATA 3.5) software.

The lifetime health and economic outcomes for patients under standard care and guidelines care are summarized in Table 2. Results are shown for the three types of guidelines care scenarios considered: IGC, OFC, and IGC + OFC. For the cohorts under OFC intervention, results are shown for the lowest (10%) and highest (90%) level of ulcer reduction simulated.

The health outcomes results of the cohort following standard care were comparable to figures reported for diabetic patients in the Netherlands. The undiscounted mean life expectancy was 14 years (discounted mean life expectancy, 8.1 years). In the 10,000 patients followed until death, a total of 1,780 ulcer episodes occurred, corresponding to a cumulative ulcer incidence of 17.8% and an annual ulcer incidence of 2.2% (mean annual ulcer incidence for the Netherlands is 2.1%) (17). The number of amputations observed was 362 (250 major and 112 minor), corresponding to a cumulative incidence of 3.6% and an annual incidence of 0.4% (mean annual amputation incidence reported for the Netherlands is 0.6%) (17). Mean lifetime costs totaled $2,142.

All cohorts of patients simulated for the different scenarios of guidelines care resulted in improved life expectancy, QALYs gained, and reduced incidence of foot ulcers and LEA compared with standard care. The largest effects on these outcomes were obtained when patients received IGC + OFC. When comparing the independent health effects of the two guidelines strategies, OFC resulted in a greater reduction in ulcer and amputation rates than IGC. Moreover, patients who received IGC + OFC showed approximately the same LEA incidence as patients who received OFC alone. The LEA decrease obtained was proportional to the level of foot ulcer reduction attained.

The mean total lifetime costs of a patient under either of the three guidelines care scenarios ranged from $4,088 to $4,386. For patients receiving IGC + OFC, these costs resulted in <$25,000 per QALY gained (relative to standard care). For patients receiving IGC alone, the ICER obtained was $32,057 per QALY gained, and for those receiving OFC alone, this ICER ranged from $12,169 to $220,100 per QALY gained, depending on the level of ulcer reduction attained.

Figure 2 is a graphical representation of the additional costs, QALYs gained, and ICER of the three guidelines care scenarios compared with standard care. All levels of foot ulcer reduction (10–90%) considered in the sensitivity analysis are represented. Also shown are the results of the sensitivity analyses performed on the QALY weights and length-of-stay input parameters. The results of the simulations for the combined scenario (IGC + OFC) were used to illustrate the effects of these variations (curves to the right). Results are shown for both input parameters varied simultaneously and separately (Fig. 3).

ICERs of <$25,000 were obtained for levels of preventive foot care of ≥10% for patients under IGC + OFC and for levels of preventive foot care of >40% for patients receiving OFC alone (Table 2).

Increasing the effectiveness of preventive foot care in patients under OFC and IGC + OFC resulted in more QALYs gained, lower costs, and a more favorable ICER. The results of the simulations for the combined scenario (IGC + OFC) were rather insensitive to changes in utility weights and costing parameters. Similar results were obtained for parameter variations in the other two scenarios (IGC and OFC separately).

The results of this study suggest that IGC + OFC reduces foot ulcers and amputations and leads to an improvement in life expectancy. Greater health benefits are obtained with higher levels of foot ulcer prevention. Although care according to guidelines increases health costs, the cost per QALY gained is <$25,000, even for levels of preventive foot care as low as 10%. ICERs of this order are cost-effective according to the stratification of interventions for diabetes recently proposed (32).

OFC strategy alone is only cost-effective at levels of preventive foot care >40%. In a previous economic study performed in Sweden (13), intensified foot care intervention was cost-effective when a 25% reduction in incidence of foot ulcers and amputations was assumed. This difference in the level of preventive foot care between their study and our study, albeit not large, may be attributed to differences in model architecture, follow-up period, and information used in the models, including country-specific parameters such as incidence of foot complications, prevalence of foot wounds, and costs.

The cost-effectiveness of IGC alone was estimated to be $32,057 per QALY and the reduction in incidence of LEA was estimated to be 4.4%. Our results are comparable to those reported by the Centers for Disease Control and Prevention (CDC) group (8), in which an ICER of $41,384 and an LEA reduction of ∼3% were reported. IGC falls into the category of a possibly cost-effective intervention in the management of the diabetic foot. Although it does not produce significant reduction in foot ulcers and LEA, its effectiveness resides in the slowing of neuropathy progression rates.

Extrapolating our results to a practical situation, if IGC + OFC was to be given to all diabetic patients in the Netherlands, with the aim of reducing LEA by 50% (St. Vincent’s declaration), the cost per QALY gained would be $12,165 and the cost for managing diabetic ulcers and amputations would decrease by 53 and 58%, respectively. From a policy perspective, this is clearly cost-effective and cost saving compared with current care. From a clinical perspective, this goal is achievable considering that LEA reductions of 49–85% have been reported in the literature for strategies containing only foot care preventive measures (6). For OFC alone, these results would be achieved at an ICER of $22,812, which is twice the ratio seen for the combined scenario.

The results of our study are supported by the robustness of the model. The agreement of the epidemiologic outputs with known data on incidence and prevalence of the diabetic foot provides evidence of the model’s ability to adequately reflect disease progression. In building the model, great effort was put into collecting updated, representative, and consistent information; therefore, the conclusions that can be drawn from it are valid and the probability for bias is small. Although the input parameters used were based on a Dutch setting, the model’s structure has worldwide applicability.

In conclusion, guidelines care comprising metabolic and foot care interventions combined is cost-effective and may even be cost saving in the management of the diabetic foot in the Netherlands. A valid and comprehensive Markov model that can be used for future cost-effectiveness analysis of diabetic foot disease prevention has been developed.