Cost-effectiveness of dehydrated human amnion/chorion membrane allografts in lower extremity diabetic ulcer treatment

The prevalence of diabetes is increasing in the US; estimated to be >10% of the population with calculated annual costs of $327 billion in 2017.¹ A key driver of costs for patients with diabetes is lower extremity diabetic ulcers (LEDU) which present a substantial financial burden to payers and a disutility burden to patients. Medicare alone spends nearly $20 billion annually on diabetic-related ulcers.² Patients with an LEDU face challenges with mobility, the risk of infection, amputation, decreased quality of life (QoL) and a shortened lifespan, all of which are exacerbated following amputations.^3,4 Unfortunately, it is estimated that up to 85% of amputations are avoidable with a holistic multispecialty team approach that incorporates innovative treatments and adherence to treatment parameters.⁵

Dehydrated human amnion/chorion membrane allograft (DHACM, EPIFIX, MIMEDX Group Inc.) is an innovative advanced wound care treatment. It belongs to a class of advanced treatments (AT), also known as skin substitutes or cellular and/or tissue products (CTPs), which are comprised of cellular and acellular dermal substitutes predominantly derived from human placental tissues and animal tissues. A previous retrospective study of the Medicare population found reduced frequency of amputations and decreased healthcare utilisation for patients that received AT.⁶ Surprisingly, <10% of LEDU episodes receiving AT were following parameters for use (FPFU), defined as initiation of AT within 30–45 days of an LEDU diagnosis and routine AT applications every 7–14 days during the episode of care. Episodes that were FPFU had statistically better outcomes.⁶

Meta-analyses and large retrospective studies have found that skin substitutes used as an adjunct to standard of care (SOC) are more effective than SOC alone in closing hard-to-heal LEDUs.^6,7,8 Additionally, a randomised controlled trial (RCT) using DHACM found a statistically significant higher probability of wounds closing relative to NAT or a comparator AT product, a bilayered skin substitute.^9,10

Improved time to wound closure and decreases in amputations provide better outcomes for patients and lower overall expenditures. The Agency for Healthcare and Research Quality (AHRQ) reviewed the effectiveness of ATs in RCTs and found shortcomings in most RCTs for skin substitutes, such as lack of real-world comorbidities, poor documentation of wound chronicity, short study periods, and a lack of follow-up for wound recurrence.⁷

There are limited economic evaluations of LEDU treatment. Given that the most widely used AT is DHACM among Medicare LEDU episodes,⁶ DHACM was compared to NAT. Several attempts to evaluate the economic impact of LEDUs in large populations have recently been undertaken. Research in Canada⁸ and the UK^11,12 determined neutral and positive impacts, respectively, based on their national willingness to pay (WTP) thresholds. However, these nations have vastly different healthcare delivery and payment systems than the US. The present work extends previous findings and addresses AHRQ concerns by reviewing the outcomes of over one million Medicare patients with a hard-to-heal LEDU with common comorbidities during a five-year (2015–2019) period. The results were used to create a cost-effectiveness analysis (CEA).

Thus, this study seeks to evaluate the cost-effectiveness and budget impact of using DHACM FPFU in treating LEDU compared with NAT from the US healthcare sector perspective. The cost-effectiveness was analysed at a WTP threshold of $100,000 USD per quality adjusted life year (QALY), and the budget impact was evaluated in the difference in per member per month spending.

Methods

Retrospective cohort design

This retrospective study design was developed as previously described⁶ with the following changes:

• The dataset included an additional year of the Medicare Limited Data Standard Analytic Hospital and Outpatient Department Files, 2015–2019

• The run-in period was increased from 60 to 90 days to better reflect the average Medicare episode (Fig 1)

• The definition of an amputation was updated to include all LEDU amputations that occurred on any claim that was billed while the patient was being treated for a LEDU

• Exclusion criteria for this study was updated (Table 1)

• The propensity model incorporated geographic and socioeconomic variables in this analysis to ensure that episodes within each group shared similar geographic and economic distributions, and to minimise differences in health coverage and care (Table 2).

Table 1. Criteria applied to identify eligible LEDU patients/episodes

Criteria	Rationale	Patients excluded, n	LEDU patients, n
Meta-group exclusions
ICD-10 coded diagnosis as a patient with foot ulcer4	Consensus definition	9,649,219	1,250,908
Confirmed diagnosis of Diabetes with a LEDU4	Consensus definition	37,294	1,213,614
LEDU episode started after 10/1/2015	Study focus criteria	124,508	1,089,106
Exclusions based on the wound
ICD-10 diagnosis coded as an ulcer above the knee4	Consensus definition	8963	1,080,143
No defined wound size during run in period	Study focus criteria	762,665	317,478
Wound depth at bone during run in period	Study focus criteria	20,234	297,244
Multiple wounds identified during run in period	Study focus criteria	88,756	208,488
Exclusions based on timeline or confounding patient and treatment complications
LEDU resolved after 10/2/2019	Period of the Medicare dataset	24,961	183,527
Episodes with no outpatient claims data	Period of the Medicare dataset	672	182,855
NAT episodes resolved within 90 days	Not a chronic foot ulcer	89,077	93,778
Patients receiving hemodialysis (only stage 5)4	Confounding comorbidity	13,400	80,378
Patients that died within 90 days of the last clinic visit	Confounding comorbidity	7027	73,351
Patients with no payment or demographic info	Include validated claims	1130	72,221
Patients treated with products outside the scope of study	Confounding treatment	2310	69,911

LEDU—lower extremity diabetic ulcer; ICD-10—international classification of disease-10; NAT—no advanced therapy

Table 1. Criteria applied to identify eligible LEDU patients/episodes

View as image HTML

Table 2. Propensity model parameters and values

Step	Summary of stepwise selection				Pr>Chisq	Variable description
	Effect entered	DF	Number in	Score Chi-square
1	WOUND_SM	1	1	8676.8015	<0.0001	Small LEDU at start of treatment (wound size ≤20cm2)
2	FAT	1	2	395.8924	<0.0001	LEDU depth of wound at start of treatment = Fat
3	MAC_JE	1	3	361.5123	<0.0001	Medicare Administrative Contractor = JE Noridian (covers CA, HI, NV, AS, GU, NMI)
4	MAC_JN	1	4	303.774	<0.0001	Medicare Administrative Contractor = JN FCSO (covers FL, PR, VI)
5	MUSCLE	1	5	254.0663	<0.0001	LEDU depth of wound at start of treatment = Muscle
6	ANKLE	1	6	174.972	<0.0001	LEDU location at start of treatment = Ankle
7	LEG	1	7	137.0951	<0.0001	LEDU location at start of treatment = Leg
8	ARF1	1	8	98.7776	<0.0001	Indicates the first claim in an episode that indicated a foot ulcer diagnosis: 1=IP claim was first claim; 0=OP claim was first claim
9	MAC_JH	1	9	84.8814	<0.0001	Medicare Administrative Contractor = JH Novitas (covers AR, CO, NM, OK, TX, LA, MS)
10	AGE	1	10	69.9221	<0.0001	Age at start of treatment
11	MAC_JM	1	11	66.2101	<0.0001	Medicare Administrative Contractor = JM Palmetto (covers NC, SC, VA, WV)
12	ARF13	1	12	60.0055	<0.0001	Diagnosed with lymphoedema in the 0–60 days prior to foot ulcer diagnosis
13	CCIDEM	1	13	37.0012	<0.0001	Dementia diagnosis 60 days prior to or any time after foot ulcer diagnosis (variable used in calculating weighted CC score)
14	ARF4	1	14	33.3772	<0.0001	Episodes with previous minor amputation (2014 to episode start)
15	YEAR_2018	1	15	30.631	<0.0001	Episode started in 2018
16	C11	1	16	29.3634	<0.0001	Diagnosed with non-pressure ulcer within 0–60 days of foot ulcer diagnosis
17	C15	1	17	25.9016	<0.0001	Treated with total contact cast (offloading) within 0–60 days of foot ulcer diagnosis
18	FOOT	1	18	27.4124	<0.0001	LEDU location at start of treatment = Foot
19	CALF	1	19	42.1297	<0.0001	LEDU location at start of treatment = Calf
20	CHF	1	20	24.8656	<0.0001	Congestive heart failure diagnosis 60 days prior to or any time after foot ulcer diagnosis (variable used in calculating weighted CC score)
21	HPPLEGIA	1	21	24.2051	<0.0001	Hemiplegia or paraplegia diagnosis 60 days prior to or any time after foot ulcer diagnosis (variable used in calculating weighted CC score)
22	C1	1	22	22.5416	<0.0001	Diagnosed with Charcot foot within 0–30 days of foot ulcer diagnosis
23	THIGH	1	23	19.5569	<0.0001	LEDU location at start of treatment = Thigh
24	MAC_JF	1	24	16.9932	<0.0001	Medicare Administrative Contractor = JF Noridian (covers AK, AZ, ID, MT, ND, OR, SD, UT, WA, WY)
25	MAC_J5	1	25	19.0191	<0.0001	Medicare Administrative Contractor = J5 WPS (covers IA, KS, MO, NE)
26	SKIN	1	26	18.6226	<0.0001	LEDU depth of wound at start of treatment = Skin
27	C7	1	27	13.7084	0.0002	Diagnosed with polyneuropathy within 0 - 30 days of foot ulcer diagnosis
28	ARF14	1	28	14.0648	0.0002	Diagnosed with renal insufficiency in the 0–60 days prior to foot ulcer diagnosis
29	YEAR_2015	1	29	14.2765	0.0002	Episode started in 2015
30	ARF2	1	30	12.9921	0.0003	Episodes with previous lower extremity ulcer (with and without a previous diabetes diagnosis) (2014 to episode start)
31	C5	1	31	13.6551	0.0002	Diagnosed with Infection within 0–30 days of foot ulcer diagnosis
32	ARF8	1	32	9.1874	0.0024	Femoral artery bypass/angioplasty treatment in the six months prior to foot ulcer diagnosis
33	RACE_1	1	33	8.2799	0.0040	Race = 1 indicates White; 0 indicates All Others
34	ESRD5	1	34	8.4543	0.0036	Diagnosed with ESRD Stage 5, but not on dialysis
35	ARF9	1	35	5.3395	0.0208	Diagnosed with protein calorie malnutrition (PCM) in the 0–60 days prior to foot ulcer diagnosis
36	DIABC	1	36	5.3964	0.0202	Diabetes, with complications, diagnosis 60 days prior to or any time after foot ulcer diagnosis (variable used in calculating weighted CC score)
37	SEX_1	1	37	5.093	0.0240	Sex = 1 indicates male; 0 indicates female
38	HMO	1	38	5.1212	0.0236	Enrolled in Medicare HMO during LEDU treatment
39	UNSPECIFIED_DEPTH	1	39	4.6365	0.0313	LEDU depth of wound at start of treatment = Unspecified
40	ARF15	1	40	4.4869	0.0342	Diagnosed with neuropathy in the 0–60 days prior to foot ulcer diagnosis
41	ARF7	1	41	4.2323	0.0397	Diagnosed with peripheral vascular disease diagnosis in the 0–60 days prior to foot ulcer diagnosis
42	C4	1	42	5.0103	0.0252	Diagnosed with gangrene within 0–30 days of foot ulcer diagnosis
43	YEAR_2016	1	43	3.8878	0.0486	Episode started in 2016

DF—Degrees of freedom; Pr>Chisq—probability greater than Chi-squared; LEDU—lower extremity diabetic ulcer; ESRD—end stage renal disease; HMO—health maintenace organisation

Table 2. Propensity model parameters and values

View as image HTML

Episodes were modelled with SAS (version 9.4, SAS, US) using a stepwise regression model (forward and backward) to identify the most statistically relevant covariates. Scores from the final propensity model were used to match episodes within four groups. Differences in variables were presented as p-values with statistical significance defined as <0.05. Kaplan–Meier curves were used to represent amputation, emergency department (ED) visits, admissions and readmissions risks. To establish initial wound sizes, patient claims within 60 days of the treatment start date were reviewed for related debridement healthcare common procedure coding system (HCPCS)/current procedural terminology (CPT) codes and these codes were used to assess wound size (AT application codes were used if debridement information was absent). Wounds were classified as small (≤20cm²), medium (21–100cm²) or large (>100cm²).

Cost-effectiveness analysis study design

This study implemented a hybrid model approach, first using a decision tree and then a Markov model to analyse the cost-effectiveness of DHACM FPFU in treating LEDUs. The population for the model were patients with a confirmed diagnosis of diabetes that had a hard-to-heal lower extremity ulcer (ulcer did not heal within 90 days using other treatments), who did not have any confounding complications, and who survived for at least one year after treatment initiation. The analysis was divided into two treatment arms. The intervention was the use of DHACM FPFU, which was compared to NAT.

The total time horizon for the CEA was five years. A one-year time horizon was assumed for the decision tree and used to calculate the budget impact. The Markov model had a four-year time horizon, with one-month cycles. All reported costs were inflated to represent 2021 US dollars, when applicable, using a 2.5% annual inflation rate. Effectiveness was measured in QALYs. The model calculated an incremental cost-effectiveness ratio (ICER), net monetary benefit (NMB), and budget impact of DHACM FPFU compared with NAT. Costs and QALYs beyond one year were discounted at 3% annually. A US healthcare sector perspective was used in accordance with the guidelines established by the Second Panel on Cost-Effectiveness in Health and Medicine.¹³ All modelling was performed using Microsoft Excel (Microsoft Corp., US) and the analyses were conducted in 2021.

Model structure—semi-Markov Model structure

The starting cohort either received DHACM FPFU for a LEDU or NAT (Fig 2a). Individuals then transitioned through the tree to: resolved LEDU; unresolved LEDU; minor amputation; or major amputation. Those who underwent a minor amputation were also eligible to receive a subsequent major amputation. After either type of amputation, patients who did not undergo an additional amputation were eligible to transition to either a closed ulcer or an unresolved ulcer. Since all patients included in the study survived at least one year from the initiation of treatment, we assumed that all patients survived throughout the one-year time horizon. The budget impact and interim one-year cost-effectiveness were calculated following the completion of the decision tree.

After the decision tree, patients transitioned into one of the six possible health states (resolved after treatment, unresolved LEDU, resolved after minor amputation, unresolved after minor amputation, resolved after major amputation, or unresolved after major amputation). The starting health state of the Markov model corresponded to the final health state of the decision tree. The Markov model was a four-year model (years 2–5) to complete the five-year time horizon of the CEA (Fig 2b). The model featured monthly cycles and individuals transitioned at the end of each cycle. Individuals in the Markov model only transitioned to death.

Clinical inputs and transition probabilities

The probabilities in the decision tree were derived from the retrospective Medicare analysis. The number of patients by treatment arm who had an LEDU close after treatment, minor amputation or major amputation informed the probabilities of these decision tree branches. Time to closure and time to amputation informed how long individuals would face the costs of post-amputation care and post-closure care (Table 3). The Markov model transition probabilities were derived from the mortality rate following a diabetic ulcer, minor amputation and major amputation, as described in published literature.¹⁴

Table 3. Model parameters

Parameter	Base case	Lower bound	Upper bound	Distribution	Source
Clinical
DHACM—Decision tree
Time to LEDU closure, months	2.95	2.76	3.15	Gamma	Medicare analysis
Unresolved LEDU rate, %	2.37	1.35	3.67	Beta	Medicare analysis
Minor amputation rate, %	3.05	1.74	4.71	Beta	Medicare analysis
Time to minor amputation, months	4.20	3.99	4.41	Gamma	Medicare analysis
Major amputation rate, %	1.00	0.57	1.55	Beta	Assumed
Time to major amputation, months	5.43	5.00	5.67	Gamma	Medicare analysis
Unresolved after minor amputation rate, %	15.00	8.43	23.04	Beta	Assumed
Unresolved after major amputation rate, %	50.00	25.89	74.11	Beta	Assumed
Major amputation after minor amputation rate, %	6.00	3.41	9.256	Beta	Assumed
Unresolved after major and minor amputation rate, %	50.00	25.89	74.11	Beta	Assumed
NAT—Decision tree
Time to LEDU closure, months	2.76	2.55	2.99	Gamma	Medicare analysis
Unresolved LEDU rate, %	4.07	2.32	6.28	Beta	Medicare analysis
Minor amputation rate, %	5.42	3.08	8.37	Beta	Medicare analysis
Time to minor amputation, months	2.06	1.88	2.26	Gamma	Medicare analysis
Major amputation rate, %	3.05	1.74	4.71	Beta	Medicare analysis
Time to major amputation, months	2.38	2.21	2.56	Gamma	Medicare analysis
Unresolved after minor amputation rate, %	15.00	8.43	23.04	Beta	Assumed
Unresolved after major amputation rate, %	50.00	25.89	74.11	Beta	Assumed
Major amputation after minor amputation rate, %	6.00	3.41	9.256	Beta	Assumed
Unresolved after major and minor amputation rate, %	50.00	25.89	74.11	Beta	Assumed
Markov model
Resolved after treatment to death rate, %	0.76	0.43	1.17	Beta	(Armstrong, 2020)1
Unresolved LEDU to death rate, %	0.76	0.43	1.17	Beta	(Armstrong, 2020)1
Resolved after minor amputation to death rate, %	1.28	0.73	1.98	Beta	(Armstrong, 2020)1
Unresolved after minor amputation to death rate, %	1.28	0.73	1.98	Beta	(Armstrong, 2020)1
Resolved after major amputation to death rate, %	1.72	0.98	2.66	Beta	(Armstrong, 2020)1
Unresolved after major amputation to death rate, %	1.72	0.98	2.66	Beta	(Armstrong, 2020)1
Costs and healthcare utilisation
DHACM utilisation
DHACM treatments	4.80	4.55	5.05	Gamma	Medicare analysis
Emergency department visits	0.37	0.21	0.58	Gamma	Medicare analysis
Inpatient days	1.10	0.79	1.45	Gamma	Medicare analysis
Readmission days	0.30	0.16	0.48	Gamma	Medicare analysis
Outpatient visits	11.90	10.72	13.14	Gamma	Medicare analysis
NAT utilisation
Emergency department visits	0.62	0.35	0.96	Gamma	Medicare analysis
Inpatient days	2.50	2.06	2.98	Gamma	Medicare analysis
Readmission days	1.40	0.95	1.93	Gamma	Medicare analysis
Outpatient visits	11.90	10.75	13.11	Gamma	Medicare analysis
Utilisation costs (2021 $USD)
DHACM treatments	1307	1276	1338	Gamma	Medicare analysis
Emergency department visits	8292	7298	9347	Gamma	Medicare analysis
Inpatient days	2980	2728	3242	Gamma	Medicare analysis
Readmission days	2097	1943	2257	Gamma	Medicare analysis
Outpatient visits	319	254	392	Gamma	Medicare analysis
Amputation costs (2021 $USD)
Minor amputation	13,216	11,987	14,503	Gamma	Medicare analysis
Major amputation	23,435	10,404	41,667	Gamma	Medicare analysis
Monthly aftercare costs (2021 $USD)
Diabetes care (no amputation)	883	505	1366	Gamma	(ADA, 2018)2
Inpatient rehabilitation (minor amputation)	34	19	52	Gamma	(Mackenzie, 2007)3
Prosthetics (minor amputation)	326	186	504	Gamma	(Mackenzie, 2007)3
Physical and occupational therapy (minor amputation)	210	120	325	Gamma	(Mackenzie, 2007)3
Outpatient visits (minor amputation)	413	236	639	Gamma	(Mackenzie, 2007)3
Inpatient rehabilitation (minor amputation)	279	160	432	Gamma	(Mackenzie, 2007)3
Outpatient rehabilitation (major amputation)	256	147	397	Gamma	(Franklin, 2014)4
Prosthetics (major amputation)	670	383	1036	Gamma	(Mackenzie, 2007)3
Physical and occupational therapy (major amputation)	403	230	623	Gamma	(Mackenzie, 2007)3
Outpatient visits (major maputation)	293	167	453	Gamma	(Mackenzie, 2007)3
Infections (major amputation)	239	136	369	Gamma	(Franklin, 2014)4
Home health (major amputation)	567	324	877	Gamma	(Rice, 2014)5
Health utilities
Diabetes	0.800	0.288	0.999	Beta	(Sullivan, 2006)6
Chronic skin ulcer disutility	0.027	0.016	0.042	Beta	(Sullivan, 2006)6
2nd chronic condition disutility	0.094	0.053	0.145	Beta	(Sullivan, 2006)6
3rd chronic condition disutility	0.088	0.050	0.135	Beta	(Sullivan, 2006)6
Amputation disutility	0.280	0.154	0.427	Beta	(Beaudet, 2014)7

DHACM—dehydrated human amnion/chorion membrane; LEDU—lower extremity diabetic ulcer; NAT—no advanced therapy;

1 Armstrong DG, Swerdlow MA, Armstrong AA et al. Five year mortality and direct costs of care for people with diabetic foot complications are comparable to cancer. J Foot Ankle Res 2020; 13:16

2 American Diabetes Association. Economic costs of diabetes in the U.S. in 2017. Diabetes Care 2018;41:917–928;

3 MacKenzie EJ, Jones AS, Bosse MJ et al. Health-care costs associated with amputation or reconstruction of limb-threatening injury. J Bone Joint Surgery 2007; 89:1685–1692;

4 Franklin H, Rajan M, Tseng CL et al. Cost of lower-limb amputation in U.S. Veterans with diabetes using health services data in fiscal years 2004 and 2010. J Rehab Research Develop 2014; 51:1325–1330;

5 Rice JB, Desai U, Cummings AKG et al. Burden of diabetic foot ulcers for Medicare and private insurers. 2014; 37:651–658;

6 Sullivan PW, Ghuschyan V. Preference-based EQ-5D index scores for chronic conditions in the United States. Medical Decision Making 2006; 26:410–420.

7 Beaudet A, Clegg J, Thuresson PO et al. Review of utility values for economic modeling in type 2 diabetes. Value in Health 2014; 17:462–470

Table 3. Model parameters

View as image HTML

Costs

The decision tree, costs (year one) were first applied based on the treatment and resulting healthcare resource utilisation a patient received. In the DHACM FPFU arm, patients accrued costs based on the average number of AT treatments, ED visits, inpatient visits, readmissions, and outpatient visits DHACM patients utilised in the retrospective analysis. Conversely, individuals in the NAT arm faced the same types of costs and utilisation as the DHACM FPFU arm but did not accrue costs associated with advanced treatment (AT). Patients in both arms who had a minor or major amputation faced a one-time amputation cost. Following the amputation, patients amassed monthly costs for post-amputation care for the remainder of the year. Similarly, patients with a closed LEDU after treatment faced a monthly cost for diabetes care following the closure (Table 3). The decision tree costs were calculated for NAT and DHACM groups (Table 4), and used to calculate the budget impact and cost effectiveness (Table 5).

Table 4. Decision tree costs and utilisation

Parameter	Unit cost (2021 $USD)	Utilisation count	Total cost*	Source
DHACM treatment costs
Treatments	1307	4.80	6273	Medicare analysis
Emergency department visits	8292	0.37	3106]	Medicare analysis
Inpatient days	2980	1.10	3278	Medicare analysis
Readmission days	2097	0.30	629	Medicare analysis
Outpatient visits	319	11.90	3801	Medicare analysis
Total			17,086
NAT treatment costs
Treatments	—	—	—	Medicare analysis
Emergency department visits	8292	0.62	5144	Medicare analysis
Inpatient days	2980	2.50	7449	Medicare analysis
Readmission days	2097	1.40	2936	Medicare analysis
Outpatient visits	319	11.90	3801	Medicare analysis
Total			19,329

* Total cost is the product of the utilisation count multiplied by unit cost;

DCAHM—dehydrated amnion/chorion membrane; NAT—no advanced therapy; all calculations have been rounded to the nearest second or third place

Table 4. Decision tree costs and utilisation

View as image HTML

Table 5. Base case cost-effectiveness and budget impact results

Cost-effectiveness results per patient
	Year one	Years 2–5	Years 1–5
Cost of DHACM, $USD	25,677	34,315	59,992
Cost of NAT, $USD	29,347	35,422	64,769
Cost difference, $USD	(3670)	(1108)	(4777)
QALYs of DHACM	0.785	2.516	3.301
QALYs of NAT	0.772	2.481	3.252
QALY difference	0.013	0.035	0.048
ICER ($/QALYs)	Dominant	Dominant	Dominant
NMB at $100,000 WTP threshold, $USD	5004	4621	9625
Budget impact for one million members in year one
Cost difference for 5980 people at risk*14, $USD			21,944,742
Cost difference per one million members in a health plan, $USD			21.94
Savings per member per month, $USD			1.83

DHACM—dehydrated human amnion/chorion membrane; NAT—no advanced therapy; QALY—quality adjusted life year; ICER—incremental cost-effectiveness ratio; NMB—net monetary benefit; WTP—willingness to pay;

* Model assumes 13% prevalence of diabetes and 4.6%4 prevalence of lower extremity diabetic ulcer in patients with diabetes; all calculations have been rounded to the nearest second or third place

Table 5. Base case cost-effectiveness and budget impact results

View as image HTML

Throughout the Markov model (years 2–5), patients faced monthly costs dependent upon in which health state they resided. Monthly costs were based on amputation status: no amputation, minor amputation, or major amputation. No distinction in terms of costs were made based on whether the LEDU was closed or unresolved during the one-year decision tree, yielding a conservative estimate.

Utilities

Health utility weights were assigned based on the QoL by health state in both the decision tree and Markov model. The utility weights were used to develop the resulting QALYs. All utility weights were collected from published literature. Baseline diabetes, the presence of a hard-to-heal skin ulcer, and other chronic condition utilities and disutilities were based on EQ-5D Index Scores collected in the US Medical Expenditure Panel Survey.¹³ The disutility assigned by an amputation event was from a systematic literature review.¹⁵ Baseline diabetes utility reflected individuals with diabetes and no other chronic condition, including an LEDU. Disutilities were applied based on the presence of an LEDU, occurrence of a minor or major amputation, and whether the LEDU resolved following an amputation. The disutility associated with an amputation was only applied to individuals who had an amputation during the one-year decision tree. After moving into the Markov model, individuals only faced the disutility associated with the presence of multiple chronic conditions (Table 3).

Sensitivity analyses

The uncertainty of the base case results, for both the one-year and five-year time horizons, was tested using univariate and probabilistic sensitivity analyses (PSA). For the univariate sensitivity analyses, baseline parameter values were varied, one at a time, to upper and lower bounds based on the reported standard error and a designated distribution. For the PSAs, parameter values were varied simultaneously based on the reported standard error and designated distribution with over 10,000 Monte Carlo simulations. A conservative assumption of a standard error of 25% of the baseline value was used for parameters with no reported standard error. Beta distributions were used for variables ranging between 0.0 and 1.0, while gamma distributions were used for positive values >0.0 with no upper bound.

Budget impact analysis

The budget impact of treating all eligible LEDU patients with DHACM FPFU was calculated assuming a one million member health plan. The prevalence of diabetes was 13%¹ and the prevalence of LEDU among individuals with diabetes was 4.6%.¹⁴ A one-year time horizon was assumed for the budget impact. The one-year cost savings per patient was multiplied by the risk pool to determine the overall cost-savings for individuals at risk. The cost-savings per plan member in year one and per member per month were then estimated.

Ethics statement

All data were previously collected, deidentified, and were available through the Center for Medicare & Medicaid Services (CMS). This research was also subject to the Medicare limited data set (LDS) data use agreement (DUA). Contact with beneficiaries is not permitted under the LDS DUA. All analysis and reporting of Medicare data was performed in compliance with relevant laws and institutional guidelines approved by the CMS.

Data availability

Data that support the findings of this study are available from the CMS. Restrictions apply to the availability of these data—used under a data use agreement between MIMEDX Group Inc. and CMS for the current study—and therefore are not publicly available. However, data may be available from the authors upon a reasonable request and with permission from CMS.

Results

Retrospective cohort results

In the full dataset, 10,900,127 patients had a confirmed diabetes diagnosis, within which 1,213,614 had a confirmed diagnosis of LEDU (Fig 3). A total of 19,910 episodes that received AT were propensity-matched to the same number of NAT episodes to establish propensity-matched Group 1. Propensity-matched Group 2 was based on the subset of 5528 episodes (27.8%) where DHACM was used exclusively for treatment. Propensity-matched Group 3 is the subset of 1829 episodes from Group 1 that were FPFU (9.2%). Lastly, propensity-matched Group 4 is the subset of episodes that were FPFU and used DHACM (n=590, 2.9%) (Fig 3). DHACM was the most frequently used AT product (33%) upon review of 16,735 episodes from propensity-matched Group 1 (Fig 4), more than double the next most used product. The demographic characteristics were similar across propensity-matched groups 1–4 (Tables 6–8).

Table 6. Demographics of over one million Medicare patients with LEDU

LEDU patient counts	Population	SD
Mean age, years	69.8	11.8
			Percentage
	Patients, n	Episodes, n	Patients	Episodes
Sex
Male	703,043	1,124,230	58.1	59.0
Female	507,653	782,370	41.9	41.0
Total*	1,210,668	1,906,600
Race
0 Unknown	11,699	18,404	1.0	1.0
1 White	933,594	1,470,078	77.1	77.1
2 Black	182,364	289,290	15.1	15.2
3 Other	15,248	22,878	1.3	1.2
4 Asian	13,218	18,866	1.1	1.0
5 Hispanic	38,593	59,411	3.2	3.1
6 North American Native	16,182	27,763	1.3	1.5
Total*	1,210,668	1,906,600
Socioeconomic variables
Medicaid dual enrolments	399,497	608,557	33.0	31.9
HMO enrolment	23,831	33,068	2.0	1.7
Patients and episodes, n*	1,210,668	1,906,600
Missing demographic data	5042	5107

* Missing demographic data—not all claims had data for sex, age and race;

LEDU—lower extremity diabetic ulcer; SD—standard deviation

Table 6. Demographics of over one million Medicare patients with LEDU

View as image HTML

Table 7. Demographics of 19,910 Medicare episodes with an LEDU. Propensity-matched group 1

Number
DFU patient counts	NAT		AT		NAT		AT FPFU		AT not FPFU
Mean age	69.6		69.5		69.3		70.7		69.4
	Patients	Episodes	Patients	Episodes	Patients	Episodes	Patients	Episodes	Patients	Episodes
Sex
Male	11,765	12,040	11,515	11,999	1108	1111	1034	1039	1122	1125
Female	7708	7870	7647	7911	716	718	789	790	701	704
Race
1 White	16,220	16,578	15,908	16,506	1514	1516	1553	1556	1517	1521
2 Black	2063	2110	1976	2071	196	198	153	156	186	187
4 Asian	153	157	125	127	16	16	13	13	*	*
5 Hispanic	474	489	554	581	41	41	52	52	54	54
6 North American Native	166	170	167	173	16	17	18	18	19	19
3 Other	220	225	209	216	23	23	15	15	15	15
0 Unknown	178	181	223	236	18	18	19	19	22	23
Total	19,473	19,910	19,162	19,910	1824	1829	1823	1829	1823	1829
Socioeconomic variables
Medicaid dual enrolment	6971	7137	6121	6360	686	688	529	531	567	568
HMO enrolment	316	317	325	332	40	40	27	27	25	25
Patients and episodes	19,473	19,910	19,162	19,910	1824	1829	1823	1829	1823	1829
Percentage
	Patients	Episodes	Patients	Episodes	Patients	Episodes	Patients	Episodes	Patients	Episodes
Sex
Male	60.4	60.5	60.1	60.3	60.7	60.7	56.7	56.8	61.5	61.5
Female	39.6	39.5	39.9	39.7	39.3	39.3	43.3	43.2	38.5	38.5
Race
1 White	83.3	83.3	83.0	82.9	83.0	82.9	85.2	85.1	83.2	83.2
2 Black	10.6	10.6	10.3	10.4	10.7	10.8	8.4	8.5	10.2	10.2
4 Asian	0.8	0.8	0.7	0.6	0.9	0.9	0.7	0.7	*	*
5 Hispanic	2.4	2.5	2.9	2.9	2.2	2.2	2.9	2.8	3.0	3.0
6 North American Native	0.9	0.9	0.9	0.9	0.9	0.9	1.0	1.0	1.0	1.0
3 Other	1.1	1.1	1.1	1.1	1.3	1.3	0.8	0.8	0.8	0.8
0 Unknown	0.9	0.9	1.2	1.2	1.0	1.0	1.0	1.0	1.2	1.3
Socioeconomic variables
Medicaid dual enrolment	35.8	35.8	31.9	31.9	37.6	37.6	29.0	29.0	31.1	31.1
HMO enrolment	1.6	1.6	1.7	1.7	2.2	2.2	1.5	1.5	1.4	1.4

* Cell sizes of <11 people or episodes are suppressed as per Center for Medicare & Medicaid Services policy;

LEDU—lower extremity diabetic ulcer; AT—advanced treatment; NAT—no advanced treatment; FPFU—following parameters for use; DHACM—dehydrated human amnion/chorion membrane

Table 7. Demographics of 19,910 Medicare episodes with an LEDU. Propensity-matched group 1

View as image HTML

Table 8. Demographics of 5528 Medicare episodes with an LEDU. Propensity-matched group 2

Number
DFU patient counts	NAT		DHACM		AT		NAT		DHACM FPFU		AT FPFU		DHACM not FPFU
Mean age	69.5		69.6		69.5		69.1		70.5		70.3		69.0
	Patients	Episodes	Patients	Episodes	Patients	Episodes	Patients	Episodes	Patients	Episodes	Patients	Episodes	Patients	Episodes
Sex
Male	3374	3397	3238	3316	3281	3332	361	361	329	329	336	339	339	339
Female	2124	2131	2156	2212	2176	2196	229	229	260	261	251	251	250	251
Race
1 White	4578	4601	4514	4619	4465	4522	505	505	515	515	486	487	494	494
2 Black	594	597	521	539	602	612	49	49	40	41	60	62	55	55
4 Asian	42	42	37	38	36	36	5	5	*	*	*	*	*	*
5 Hispanic	125	125	153	160	167	168	14	14	17	17	16	16	19	20
6 NA Native	48	49	49	50	54	54	2	2	*	*	*	*	*	*
3 Other	60	62	60	62	65	66	10	10	*	*	*	*	*	*
0 Unknown	52	52	60	60	68	70	5	5	*	*	*	*	*	*
Total	5498	5528	5394	5528	5457	5528	590	590	589	590	587	590	589	590
Socioeconomic variables
Medicaid dual enrolment	1951	1961	1750	1799	1718	1734	205	205	172	173	175	176	191	192
HMO enrolment	108	108	91	92	89	91	–	–	–	–	42	42	–	–
Patients and episodes	5498	5528	5394	5528	5457	5528	590	590	589	590	587	590	589	590
Percentage
DFU patient counts	NAT		DHACM		AT		NAT		DHACM FPFU		AT FPFU		DHACM not FPFU
	Patients	Episodes	Patients	Episodes	Patients	Episodes	Patients	Episodes	Patients	Episodes	Patients	Episodes	Patients	Episodes
Sex
Male	61.4	61.5	60.0	60.0	60.1	60.3	61.2	61.2	55.9	55.8	57.2	57.5	57.6	57.5
Female	38.6	38.5	40.0	40.0	39.9	39.7	38.8	38.8	44.1	44.2	42.8	42.5	42.4	42.5
Race
1 White	83.3	83.2	83.7	83.6	81.8	81.8	85.6	85.6	87.4	87.3	82.8	82.5	83.9	83.7
2 Black	10.8	10.8	9.7	9.8	11.0	11.1	8.3	8.3	6.8	6.9	10.2	10.5	9.3	9.3
4 Asian	0.8	0.8	0.7	0.7	0.7	0.7	0.8	0.8	*	*	*	*	*	*
5 Hispanic	2.3	2.3	2.8	2.9	3.1	3.0	2.4	2.4	2.9	2.9	2.7	2.7	3.2	3.4
6 NA Native	0.9	0.9	0.9	0.9	1.0	1.0	0.3	0.3	*	*	*	*	*	*
3 Other	1.1	1.1	1.1	1.1	1.2	1.2	1.7	1.7	*	*	*	*	*	*
0 Unknown	0.9	0.9	1.1	1.1	1.2	1.3	0.8	0.8	*	*	*	*	*	*
Socioeconomic variables
Medicaid dual enrolment	35.5	35.5	32.4	32.5	31.2	31.4	34.7	34.7	29.2	29.3	29.8	29.8	32.4	32.5
HMO enrolment	2.0	2.0	1.7	1.7	1.6	1.6	0.0	0.0	0.0	0.0	7.2	7.1	0.0	0.0

* Cell sizes of <11 people or episodes are suppressed as per Center for Medicare & Medicaid Services policy;

LEDU–lower extremity diabetic ulcer; AT–advanced treatment; NAT–no advanced treatment; FPFU–following parameters for use; DHACM–dehydrated human amnion/chorion membrane

Table 8. Demographics of 5528 Medicare episodes with an LEDU. Propensity-matched group 2

View as image HTML

On average, AT was initiated about 80 days into an episode of care in contrast to only about 35 days for AT episodes that were FPFU (Fig 1). Episodes using DHACM FPFU also had the shortest average length of treatment. The distribution of initial wound sizes was graphed for propensity-matched Group 1 (Fig 5). The propensity-matched groups were analysed for frequency of healthcare resource utilisation: amputations, ED visits, admissions and readmissions, and were consistent with previously reported results. Episodes from propensity-matched Group 4 that used DHACM or NAT were plotted as Kaplan–Meier curves (Fig 6) demonstrating significant reductions across major amputations and all measured healthcare resource utilisations.

Cost-effectiveness analysis base case results

Compared with NAT, DHACM FPFU in the treatment of LEDU was a dominant strategy, meaning that it was cost-effective at any WTP threshold, for the one-year decision tree, four-year Markov model, and five-year hybrid model (Table 5). In year one, DHACM FPFU provided an additional 0.013 QALYs while saving $3670 per patient. This yielded a NMB of $5003 at a $100,000 per QALY WTP threshold. In years 2–4, those in the DHACM FPFU arm gained an additional 0.035 QALYs while saving $1108 per patient, and yielding a NMB of $4621 at a $100,000 per QALY WTP threshold. The full five-year results net out an increase in 0.048 QALYs and a savings of $4777. At a WTP of $100,000 per QALY, the five-year NMB was $9624.

Sensitivity analysis

The univariate sensitivity analysis illustrated that DHACM FPFU remained cost-effective over NAT for treating LEDU under all analysed scenarios for both the one- and five-year time horizons. In each case, the NMB for DHACM FPFU was greater than $0. For both time horizons, the most sensitive parameters were healthcare resource utilisation and amputation rates (Fig 7).

The PSA demonstrated the robustness of the base case results. Under the one-year time horizon, DHACM FPFU had a 97.6% likelihood of being dominant and a 99.5% likelihood of being cost-effective at a WTP threshold of $100,000 per QALY. Similarly, in the five-year time horizon, DHACM FPFU was dominant in 98.9% of simulations and was cost-effective at a $100,000 per QALY WTP threshold in 99.9% of simulations (Fig 8).

Budget impact results

DHACM FPFU presents a cost-saving alternative to NAT for LEDU treatment in a one-year budget impact (Table 5). The per-member per-month cost savings was estimated to be $1.83 for a one million member plan. These results are based on the estimation that a health plan with one million members would have 5980 LEDU patients eligible for DHACM.

Discussion

DHACM FPFU was associated with lower overall costs in the first year, despite higher treatment costs. This cost saving was due to lower amputation rates and lower healthcare resource utilisation (for example, ED visits, inpatient admissions). Improved QoL in years 2–4 was attributed to DHACM FPFU preventing amputations during the first year. Minor and major amputations led to poorer QoL and higher mortality rates, driving the difference in outcomes.

Costs followed a similar course, as those who avoided amputations also avoided aftercare costs. The true costs and impacts of LEDUs and amputations are likely underestimated in this study. Amputations are also associated with reductions in occupational productivity which extends to family and caregivers who reorganise their lives in support of patients.¹⁶ Additionally, a host of intangible QoL, social and mental health issues are associated with the patient's morbidity.^17,18 Additional value is added when wounds are resolved quicker since shorter treatment times and less healthcare resource utilisation create bandwidth to treat additional patients. The impacts of intangible patient costs and increased availability of hospital resources were not captured in this study.

This retrospective analysis of Medicare files with 2015–2019 data generated outcomes consistent with previous findings, including reductions in amputations, and healthcare resource utilisation.⁶ Additionally, the data strongly supported FPFU, defined as the initiation of AT within 30–45 days of diagnosis and AT applications at 7–14 day intervals after the first application during the episode of care.⁶ While only 9.2% of Medicare LEDU episodes treated with AT were FPFU, the improved outcomes warranted a cost-effectiveness analysis for episodes using DHACM and FPFU. DHACM was the most widely used AT and, as such, provided the greatest number of propensity-matched LEDU episodes for analysis—590 episodes.

Patients with diabetes and LEDU face increased morbidity, mortality and costs. In the US, specifically, diabetes and its comorbidities present a growing challenge for the healthcare system. Opportunities to provide patients and payers with beneficial treatment pathways will only increase in importance in the coming years. The undertaken analyses explored the clinical benefit, cost-effectiveness and budget impact of using DHACM FPFU in the treatment of LEDU compared with NAT.

The statistical significance of results in this study showed that outcomes were unlikely due to chance. The actual treatment effect or clinical significance and impact on the patient and the healthcare system have greater importance but is potentially a more subjective result.¹⁹ The ‘extent of change, whether the change makes a real difference to subject lives, how long the effects last, consumer acceptability, cost-effectiveness, and ease of implementation’ have been reported as indications of clinical significance.²⁰ By these standards, salvaging 3% more limbs in 590 episodes (propensity-matched group 4) and similar results in propensity-matched groups of over 10,000,⁶ by using a cost-effective and patient-friendly product is clinically significant.

The current study shows that within the Medicare population over 90% of episodes that were treated with DHACM were not FPFU. Given an annual incidence of DFUs in the US Medicare population of 6–7%,²¹ combined with the commercially insured incidence of 4.1% (employer-sponsored insurance for patients aged 18–64 years),²¹ there is a tremendous opportunity to improve patient outcomes while reducing costs. The analysis presented here calculates nearly $22 million annual savings in a typical one million person plan. As with any dominant cost-savings strategy the savings will increase for a larger population or a population with a higher incidence of LEDUs. Given that these improvements in outcomes increase quality of life, reduce costs to patients as well as all facets of the healthcare system, industry adoption of FPFU makes excellent clinical and economic sense.

Payers and providers who manage patients with diabetes or prediabetes make decisions to adopt coverage for cost-effective care on a regular basis. Disease prevention²² and choice of insulin treatment²³ are two areas of high utilisation resulting in important clinical and economic outcomes. Noting that some cost-effective decisions may add costs, while others can save costs. For context, the insulin analogue, degludec injection is a dominant strategy compared with regular (basal) insulin for patients with Type 1 diabetes, and frequently Type 2 diabetes.^22,23,24,25 Furthermore, analogue versus regular insulin has been proven to be cost effective in a recent study.²⁵ These favourable clinical and economic results likely contribute to the adoption of insulin in clinical practice.²⁶ Similarly, DHACM FPFU is a dominant treatment strategy for patients with diabetes who develop a LEDU since it saves costs and still improves clinical benefits. DHACM FPFU should be adopted in a similar fashion to other treatments that provide clinical and economic benefit for patients with diabetes since it has a positive budget impact on Medicare and potentially other payers.

The evidence supported that DHACM FPFU provided clinical and economic benefits to patients. Thus, policies should be enacted to encourage following evidence-based best practices. Proactively working to resolve the LEDU is better for the patient and healthcare system. Increasing education among all disciplines providing wound care can enhance patient treatment, and help guide physicians or other clinicians who have limited training in wound management. Improved policies targeting healthcare decision-makers, inclusive of both the economics and clinical outcomes, can help encourage better treatment practices. All in all, care that provides patients with reduced morbidity, mortality and costs should always be chosen.

Assumptions and limitations

As with any model, there are several assumptions relied upon and limitations to the model. First, patients being treated with DHACM were assumed to follow parameters for use. As seen in the retrospective analysis, FPFU provided patients with the best outcomes. All practicing medical directors, healthcare institutions and healthcare policy makers should prioritise evidence-based treatment algorithms in an area of medicine that lacks standardised training, such as wound care. However, at this time, only a limited number of patients are treated with DHACM FPFU. Alternative ATs to DHACM are available to providers but given the predominant usage of DHACM, the analysis was limited to DHACM use only.

Additionally, patients were assumed to survive at least one year following the initiation of treatment, regardless of the treatment arm. Therefore, the findings are only generalisable to those whose health is good enough that they can be assumed to survive into the following year. Next, patients were assumed to remain in the health state in which they ended the decision tree throughout the time horizon in the Markov model, unless they transitioned to death. Finally, input populations to the analysis that contained <10 patients could not be reported in order to remain compliant with Medicare's data reporting guidance, and thus needed to be resolved through assumptions. These assumptions were primarily made about transitions after minor and major amputations, regardless of treatment arm (Table 3).

Retrospective studies have many benefits but can suffer if the data is of poor quality, especially when confounding factors are not accounted for, or if appropriate comparison cohorts are unavailable. We recognise that the nature of Medicare claims data allows the possibility of some confounding features to go unnoticed. Thus steps within this analysis were taken to avoid such likelihoods by excluding over 700,000 claims with missing data. To minimise biases, we used propensity-matching to improve the comparability of cohorts. More specifically, in a move to significantly further reduce the risks of selection or treatment bias,²⁷ demographic, economic and geographical features were added to a previous propensity model which also included the Charlson Comorbidity Index.⁶ We note that the propensity-matched cohorts showed similar demographics, wound sizes and trends.

This analysis highlights notable disparities in AT usage across race (Tables 6–8) and geography. The impact of these social factors were accounted for by propensity-matching these variables, as provided in the Medicare data. However, biases in the collection and reporting of race are likely to exist. Likewise, potential impacts of selected providers, clinics, social access equity,²⁸ secondary copayments,²⁸ and quality of treatment are not captured or truly standardised in a retrospective analysis. Selection biases are also possible in any group comparisons. The most stringent selections were applied to propensity-matched group 4, the smallest with 590 episodes. This group was controlled for via FPFU, treatment type (DHACM) and a validated set of propensity variables. The fact that observations within propensity-matched group 4 were seen in larger groups suggests that the scalability of the findings is promising. A prospective randomised controlled trial looking at the impacts of FPFU on cohorts of NAT, AT and DHACM patient outcomes would represent the gold-standard for validating these results.

Nevertheless, the univariate sensitivity analysis and the PSA illustrated the robustness of the base case model results. The results are also similar to a previously conducted study from the UK National Health Service perspective, which found DHACM FPFU to have a high likelihood of being cost-effective at a WTP threshold of £20,000 per QALY.¹⁰

Conclusion

In this study, DHACM FPFU was a dominant strategy compared to NAT and should always be considered, especially when the LEDU fails to close after 30–45 days of standard care. DHACM FPFU provides better clinical outcomes than NAT by reducing major amputations, ED visits, inpatient admissions and readmissions. These clinical gains were achieved at a lower cost, providing excellent value. DHACM FPFU is likely to be cost-effective at any willingness-to-pay threshold.

DHACM presents a beneficial treatment pathway for patients and payers. This study validated that only 9.2% of LEDU episodes followed parameters for use, providers' delayed the first AT until approximately 80 days after diagnosis, and AT reapplication did not occur regularly. Improved education of all providers should be required to better guide proper AT FPFU, thus improving patient outcomes and reducing healthcare costs.

Improving policies supporting appropriate use of ATs, instead of limiting them, is also necessary and in years 1–5 post-treatment will be more cost-effective. Payers may want to consider linking proper usage to reimbursement to encourage the scheduling of regular visits. This retrospective and cost-effectiveness analysis clearly demonstrated DHACM—when used according to defined parameters for use—could guide quality wound care practices that favourably impact patients' QoL and reduce healthcare costs, while saving both limbs and lives.

Reflective questions

• How frequently does an AT provided under Medicare follow parameters for use?

• Where are the $3670 cost savings accrued in year one for LEDU episodes using DHACM FPFU?

• At what willingness to pay threshold are DHACM savings cost effective?

• How does the ICER of DHACM compare to insulin?