Data source. Medicare is the federal health insurance program for people 65 or older, certain younger people with disabilities, and people with end-stage renal disease. Medicare has 4 distinct parts (A, B, C, and D) to help beneficiaries cover certain health care services; Parts A and B are of consequence to this analysis. Medicare Part A covers inpatient hospital stays, skilled care in a skilled nursing facility up to 100 days, medically necessary hospice care, and some HHA care. Part B covers certain Qualified Health Plan services and procedures, HOPD care, some medical supplies, and preventive services.
This analysis used the 2012–2014 national Medicare Limited Data Set Standard Analytic Files, including Denominator (enrollment), hospital inpatient and outpatient, DME, carrier (physician), and HHA claims. The 5% sample data (nationally representative sample of beneficiaries) was used to study traditional NPWT episodes, while the 100% sample (all Medicare beneficiaries) was used for disposable NPWT episodes to increase the number of disposable episodes. These files are claims data sets that contain information on care provided to Medicare beneficiaries and on patient characteristics such as age, gender, enrollment history, medical diagnoses, dates and places of service, and payment amounts. Claims data are based on the bills submitted by providers to payers representing services provided, provider setting of care, and patient comorbidities. Although claims data often are utilized to approximate cost of service, the cost identified is not direct medical cost.
Construction of episodes of care. NPWT episodes were identified from the data using service dates on Medicare claims. Traditional NPWT patients were identified using the Current Procedural Terminology/Healthcare Common Procedure Coding System (CPT/HCPCS) codes A6550, A7000, E2402, 97605, and 97606; disposable NPWT patients were identified with the CPT/HCPCS codes A9272, G0456, G0457, 97607, and 97608 (see Table 1). Medicare-denied NPWT claims and home health claims not directly related to nursing care were excluded from analysis.
For traditional NPWT, episodes began with the first observed DME monthly pump rental claim (HCPCS E2402) and continued until a break of >31 days occurred between rentals. To ensure the episode of care was followed to its completion, a break of 31 days was used based on DME monthly pump rental policies. The episode end date was estimated beyond the last pump rental by assuming all supplies were used at the standard Medicare allowed use rates of 15 dressings (HCPCS A6550) and 10 canisters (HCPCS A7000) per month. For disposable NPWT, episodes began with the first observed HOPD NPWT claim (HCPCS G0456/G0457 during this period) and continued until a break of >31 days occurred between claims. The same 31 day cutoff as based on traditional NPWT monthly pump rental policies was used to maintain consistency in evaluating the end of disposable NPWT episodes. For this analysis, neither group episode payment analysis started in the acute inpatient setting.
The episode end date for paid claims (ie, nondenied) was estimated using 1 of 2 methods to ensure the full length of discrete episodes was captured. First, for episodes with more than 1 HOPD NPWT bill, the average interval between device changes (for that episode) was added to the last HOPD NPWT bill. Second, for episodes with just 1 HOPD NPWT bill, the average device change interval for all patients in the data file was added. Patients with disposable or traditional NPWT episodes were mutually exclusive.
Episode costs (total payments) were constructed from relevant wound diagnosis and treatment claims on the billing forms within the episode and included DME, physician, or HOPD paid claims lines with any relevant CPT/HCPCS codes (see Table 1). Cost was defined as total payment including both Medicare program payments and beneficiary coinsurance and deductible liabilities. To ensure a proportion of the HHA bill associated with NPWT was captured and arrived at a more accurate episode cost, the portion of HHA bills that overlapped the NPWT episode was prorated. Because HHA bills do not have line-item payment amounts, HHA claim total payments were prorated based on charges, excluded line items other than nursing visits (eg, physical therapy), and then selected all nursing visits falling within the NPWT episode. Using this method, some HHA cost not directly related to NPWT may have been captured, but the principal diagnoses on HHA claims overwhelmingly indicated wound care as the main reason for the HHA care.
Risk adjustment and sensitivity analyses. Several analyses were performed to test the robustness of the total mean payments results. First, the influence of wound type and comorbidities on episode cost was examined. Episodes of care were stratified by episode type (traditional versus disposable) and wound type (surgical wounds, generic open wounds, skin ulcers, diabetic ulcers, and circulatory disease wounds). Mean cost within strata were calculated, and the means were weighted by the all-episodes proportion of cases by wound type. Episode costs also were assessed for different wound types.
Second, ordinary-least-squares (OLS) regression was used to predict episode costs as a function of episode type, patient demographics, and diagnoses from the claims (diagnosis-based risk adjustment). Comorbidities were identified using diagnoses reported on 12 months of physician and hospital claims, aggregated into Clinical Classification System disease categories.19
Third, aspects of Medicare payment policy were handled. The presence of traditional NPWT qualifies any homebound patient for home health care; all HHA costs and recomputed episode costs were removed. Medicare updates its payment rates on an annual basis through a rulemaking process to reflect updated costs and evolving payment policies. Similarly, Medicare adjusts DME fee schedule rates through its competitive bidding process among DME providers. Therefore, because Medicare payment rates for DME items changed in response to its recent competitive bidding initiatives, all claims were priced using 2016 Medicare rates20,21 and recomputed episode costs.
Statistical analyses. For categorical variables, statistical significance was assessed using chi-squared for comparisons between patients who used traditional and patients who used disposable NPWT. For continuous variables, statistical significance was assessed using Mann-Whitney U tests. The claims data analyses were conducted using Statistical Analysis System (SAS) software, version 9.4 (SAS Institute, Cary, NC).