Decision-analytic models are widely used to determine the cost-effectiveness of 1 treatment or health care program compared with another. The authors constructed a 2-state Markov model similar to that used by Romanelli et al20 in a cost-effectiveness analysis of extracellular matrix versus standard care for the treatment of VLUs. The model was implemented using Microsoft Excel (Microsoft Corporation). The Markov modelling approach is useful for estimating costs and consequences for chronic conditions that change over time; it assumes that a cohort of patients begins in a defined health state and then patients transition over time to other health states at defined intervals (cycles). The model used in the current study had time horizons of 12 weeks and 26 weeks and a cycle length of 1 week (to reflect wound assessment practice).21 In this model, a hypothetical cohort of patients began in the open ulcer health state, and at the end of each weekly cycle a proportion of the cohort moved into the closed ulcer health state according to a constant transition probability. Closed ulcer is an absorbing state; patients cannot return to the open ulcer health state.
This methodology assumes these transitions occur at the start of each cycle, whereas in reality the transition could occur at any time within the cycle. For this reason, a correction often is made under the assumption that the transition occurs halfway through the cycle. Because in this case the cycle length was short (ie, 1 week), this correction was not deemed necessary.
It is standard practice to define the perspective of an economic model. This is particularly important in determining what costs are included in the model as well as their values. In this case, the perspective was the US payer; this study employed treatment costs from an analysis of Medicare claims data.22 Sometimes costs and effects are discounted to take account of the time at which the costs and effects are evaluated. In this study, because the time horizon of this model was less than 1 year, neither costs nor effects were discounted.
Two (2) arms of the model, tNPWT and sNPWT, were calculated separately for a combination of both VLU and DFU ulcer types in the proportions reported by Kirsner et al19 (37.3% DFU). The model assumptions are shown in Table 1. Open ulcer to closed ulcer transition probabilities were calculated using ulcer closure rates reported from the recent 161-patient RCT.19,23 Ulcer recurrence was not included in the model; therefore, the transition probability for closed ulcer to open ulcer was set to zero. Delhougne et al22 recently estimated the cost per day for sNPWT and tNPWT from the payer perspective using US national Medicare claims data, assuming 2016 Medicare rates. Their analysis included claims data from Medicare Parts A and B over a wide range of health care settings and services. These unit costs were inflated to 2018 costs using health care inflation indices (US Bureau of Labor Statistics24) and multiplied by 7 to estimate the weekly costs of treatment for sNPWT and tNPWT (see Table 1).
The costs over the defined timescale were summed to give a total cost of treatment for each arm of the model. The difference (the incremental cost) was calculated as the total cost for sNPWT minus the total cost for tNPWT. Effectiveness was calculated 2 ways: 1) the incidence of healing at a defined point (eg, 12 or 26 weeks), and 2) the total number of open ulcer weeks. The incremental effectiveness was calculated as sNPWT effectiveness minus tNPWT effectiveness.
The analysis described, using point values of all the model inputs, is known as the base case analysis. Further analyses were subsequently performed to investigate the sensitivity of the model to the input parameters and to develop an understanding of the uncertainty in the results. The base case analysis included results at 12 weeks (the timescale of the study by Kirsner et al19) and extended to 26 weeks.
One-way sensitivity analysis. In any modelling approach, the values used for the model parameters, such as the various probabilities and unit costs, are subject to uncertainty. In order to examine the effects of this uncertainty, sensitivity analyses have become an essential part of economic evaluation. The simplest way to perform this analysis is to test the effect of the input parameters one by one (“one-way sensitivity analysis”).
Following the development of probabilistic methods, one-way analysis may be perceived to be less relevant. However, it can offer an insight into the relative sensitivity of the results to the model inputs. Setting the limits of each variable to a defined percentage of the point value allows the sensitivity to be easily compared across the variables. Therefore, in this study, each input value was changed by ±20%, one input at a time, while keeping all the other inputs constant.
Scenario analyses. Additional scenario analyses also were included as follows:
1) a 26-week time horizon where, instead of continuing with NPWT, any ulcers remaining unhealed at 12 weeks were changed to standard dressings;
2) a 26-week time horizon where, instead of continuing with NPWT, any ulcers remaining unhealed at 4 weeks were changed to standard dressings; and
3) as the base case but with the daily costs of tNPWT and sNPWT equal at the daily cost of sNPWT ($62.36).
Scenarios 1 and 2 require a weekly cost of treatment with standard dressings to be assigned to the closed ulcer health state for any weeks after the switch to standard dressings. Nherera et al25 conducted an economic evaluation to compare the use of standard care (including compression, debridement, and foam dressings) with cadexomer iodine with standard care alone for the management of chronic VLUs. This analysis estimated the weekly cost of standard wound care for VLUs from the US payer’s perspective to be $238 (2014 prices). The current authors used this as an estimate of the weekly treatment cost of lower extremity ulcers (VLUs and DFUs) using standard dressings, inflating the cost to 2018 prices using US Bureau of Labor Statistics24 data to result in a weekly cost of $265.02, which was used for the first 2 scenarios. These scenarios also required a weekly transition probability from open ulcer to closed ulcer for standard treatment of hard-to-heal ulcers. In an observational study of 52 hard-to-heal wounds (including VLUs, DFUs and other wound types), Dowsett et al26 estimated that 4 out of 52 of these wounds would heal in a 26-week period. These data were converted to a weekly transition probability of 0.00308 and used in the current first 2 scenarios.
Probabilistic sensitivity analysis. While the one-way sensitivity analysis changes the input values one at a time, a probabilistic approach changes all the values of the model parameters simultaneously according to predetermined distributions. This creates a set of input values that are used to run the model. This process is repeated multiple times with new values of the model parameters being drawn each time in order to create many different scenarios and build up a picture of the uncertainty in the results.
The probabilistic analysis used effectiveness data19 combined across both studied wound types (DFU and VLU). The relative risk of ulcer closure was drawn from a lognormal distribution, the baseline risk of healing from a beta distribution, and the weekly costs of treatment from gamma distributions. A total of 10 000 iterations of the model were run.
Threshold analysis. A threshold analysis was conducted to estimate the daily cost of treating a wound using tNPWT during which the use of sNPWT was cost-neutral with respect to tNPWT. In this analysis, the daily cost of tNPWT was reduced until the incremental cost became zero.
Data collection. Data were collected into Microsoft Excel spreadsheets for analyses.