The U. S. Department of Veterans Affairs uses the Managerial Cost Accounting Office (MCA) for fiscal management and to determine the cost of patient care. National Data Extracts (NDEs) have been created to facilitate access to workload and cost information. These extracts report costs of inpatient and outpatient encounters, reflecting actual resource use within the VA. These data contain rare outlier observations with cost inconsistent to the care provided.
HERC has developed SAS code to flag and adjust outlier observations in the Managerial Cost Accounting Pharmacy National Data Extract (the MCA PHA NDE). The PHA file contains prescription-level records which include detailed cost information on each prescription filled, both inpatient and outpatient.
In order to identify costs that are likely to be truly erroneous, rather than legitimately expensive medications, strict criteria are applied to define outliers. The product cost of the medication and the dispensing cost are considered separately. In consultation with MCA and after review of cost distributions by VA Class, product cost outliers are defined as costs greater than 120% of the maximum price of all drugs in the VA drug class, as shown in the drug price list from Pharmacy Benefits Management Services. This criterion is designed to select only gross outliers. Since dispensing costs are typically low and do not vary much across VA drug class, the outlier cut-off for dispensing costs is set at a fixed $100. (Note that median dispensing costs by VA drug class range from $3.56 to $9.95 in FY15.) Absolute values are used in order to identify both positive and negative outliers. Out of 243 million records in the FY15 PHA file, 33,410 product costs (0.014%) and 158,635 dispensing costs (0.065%) are classified as outliers by these criteria.
Both product and dispensing outlier costs are adjusted by replacing the outlier value with the median cost for the VA drug class (and retaining the sign of the original cost). Although this code is designed to adjust the prescription-level records in the MCA PHA dataset, with some additional coding to restrict to outpatient fills and summarize by dispensing date, the prescription-level adjustments can also be used to adjust the daily pharmacy cost records in the MCA Outpatient NDEs.
This work was conducted as part of a project funded by the Geriatric & Extended Care Data & Analysis Center (GECDAC). Researchers interested in obtaining this code should contact HERC at firstname.lastname@example.org.
HERC researchers recommend using the Primary Care Management Module (PCMM) or Reengineered Primary Care Management Module (RPCMM) in the Corporate Data Warehouse (CDW) to identify a patient’s primary care provider. RPCMM is the current system used at most, but not all, sites. PCMM is the legacy system that is still in use. In general, if you are interested in years prior to fiscal year 2016 (FY16), use the PCMM file; if you are interested in FY16 or later, use both the PCMM and RPCMM.
PCMM/RPCMM list an assignment between the primary care provider (PCP) and the patient. Each record in PCMM/RPCMM lists the date the assignment started (RelationshipStartDate) and the date the assignment ended (RelationshipEndDate), if applicable. Relationship end dates may be missing for two reasons: the relationship has not ended, or the ending date was never posted to the dataset. In general, patients can have only 1 assigned primary care provider at a time, unless they have multiple residences.
Researchers working PCMM/RPCMM should note that both domains include records for care assignments other than primary care (such as case managers). It is important to distinguish these, and not regard them as primary care assignments, when filling in missing values for the end date for the primary care relationship. To identify a list of PCPs in the PCMM table PatientProviders, try the following logic: (ProviderRole="PC ASSIGNMENT" OR "PC ATTENDING") AND (TeamPurpose="Primary CARE" OR "PRIMARY CARE _ HBPC"). The resulting list allows you to identify a provider using the variable PrimaryProviderSID.
To identify current PCP assignments in the CDW PCMM/RPCMM domain table, only use records where: [(CurrentProviderFlag=YES) AND (RelationshipEndDate=(.))].
When you have identified a list of providers in PCMM/RPCMM, you can use the variables PrimaryProviderSID and sta3n to identify these providers in other files. The combination of PrimaryProviderSID+sta3n from the CDW PCMM.PatientProviders table can be crosswalked to the CDW Staff table as StaffSID+sta3n. (Essentially, PrimaryProviderSID is the same as StaffSID for a given sta3n).
Resources on PCMM/RPCMM can be found on the VIReC CDW Documentation intranet site (Intranet-only: https://vaww.virec.research.va.gov/CDW/Documentation.htm) and the CDW MetaData Reports Page on the CDW SharePoint site (Intranet-only: https://vaww.cdw.va.gov/Pages/CDWHome.aspx).
For answers to more commonly asked questions about identifying providers in VA administrative data, visit the webpage Identifying Providers in VA Administrative Data
There are two main sources of encounter-level cost data in VA administrative data: Managerial Cost Accounting (MCA) data and Health Economics Resource Center (HERC) Average Cost data. VA researchers and operations users who need to evaluate costs may find themselves asking, “Which source of VA cost data is best for my project?” These datasets are created for different purposes and are built using different methodology, so understanding the characteristics of both datasets is important before deciding which to use.
We provide a brief summary to help researchers and operations users determine which VA cost data source is best for their project needs.
Purpose of the data
MCA data is created using a local, activity-based method and is the official managerial cost accounting system for the entire VA. Each VAMC determines the relative value units of products, the components of encounters, and then uses staff labor mapping along with RVUs to assign costs to encounters. MCA data contain the cost of each VA hospital stay, outpatient visit, and dispensed outpatient prescription. Since MCA data are designed for management, users will find specific cost information for encounters. MCA is the most frequently used source of data for VA economic research.
HERC Average Cost data represents the estimated cost of each VA encounter. These estimates are created by distributing VA aggregate costs to patient encounter-level utilization using the hypothetical Medicare payment for each service. HERC Average Cost data include national estimates which eliminates the geographic variations in cost for similar encounters. HERC Average Cost data are designed for research; the purpose of these data is for users to quickly understand the average cost of all VA health care services and compare costs to systems outside VA (e.g. Medicare).
Advantages of MCA data
- MCA costs estimate reflect facility differences in the production of health care services
- MCA includes pharmacy cost data
- MCA includes state nursing home stays
- MCA data are considered more precise than HERC data
Advantages of HERC Average Cost data
- HERC data provides costs that are consistent across similar encounters and across VAMCs
- HERC Average Cost data allow for comparison to non-VA healthcare costs
MCA data is best when
- You need local cost estimates
- You want to compare costs between VA facilities, providers, or procedures (Make sure to control for geographic wage differences, if applicable)
- When conducting implementation work where you want to focus on variable costs rather than total costs
HERC data is best when
- You want to generalize VA costs to non-VA health systems
- You want data that is already cleaned and summarized
- The precise costs of specific services are not important
Consider the specific needs and goals of your project. If possible, use both cost data sources; one as the primary data source and one as a sensitivity analysis.
More information is available in the Cost-Effectiveness Analysis Course VA Costs: HERC versus MCA.
Although they share the same name, Relative Value Units (RVUs) created for the Centers for Medicare and Medicaid Services (CMS) are not the same RVUs used in the VA Managerial Cost Accounting (MCA) system.
MCA is an activity-based costing system. Activity-based costing combines activity reports, financial data, workload, and intermediate products used in encounters and hospitals stays (Azoulay 2007). These cost estimates are regarded as a more accurate measure of resource use than cost-adjusted charges (Ross 2004; Udpa 2006). RVUs are essential part of MCA and other activity-based costing systems.
MCA RVUs are used to find the cost of intermediate products, specific health care services such as a cholesterol test, a chest x-ray, or a 20-minute clinic visit. MCA tracks the cost of each department and the intermediate products it produces. Each intermediate product has an RVU. The total cost of the department is divided by number of RVUs it produces, giving the cost per RVU, so that the cost of each intermediate product can be found. The cost of a hospital stay or outpatient visits is found by adding together the cost of all the intermediate products used in that stay (or visit).
MCA tracks each type of cost separately (e.g., labor, supplies, management overhead, contract services, and as many as 13 different cost types). Each type of cost has its own set of RVUs. MCA has a standard set of RVUs, but these may be tailored to reflect the specifics of the cost of production at each medical center or department within the medical center. The result is an estimate of costs based on the cost of the department and the intermediate products it produces. MCA cost estimates thus reflect the variations in cost that result from differences in staff and supply costs, efficiency, and the mix of intermediate products used in providing care. The resulting cost estimate reflects resources used in providing care, providing researchers with the opportunity to study variations in cost. For example, HERC researchers have used MCA to find the difference cost that result from different techniques for Coronary Artery Bypass Graft (CABG) surgery.
While the MCA estimates can be highly accurate, the method has the potential for occasional errors. Incorrect mapping of costs and products can distribute the cost of a department to the wrong products or to too few products. MCA has worked over the years to develop auditing and cleaning systems, but researchers should verify MCA cost data. MCA National Data Extracts are available at CDW. Record layouts, metadata, and the Technical Guide are available on the MCA Office website (intranet only).
The RVUs developed for the Centers for Medicare and Medicaid Services (CMS) are used to calculate payments to providers. These RVUs do not represent cost, but were designed to compensate providers based on time, mental effort, technical skill, judgement, and stress. Services are characterized by Current Procedure and Terminology (CPT) code, and separate RVUs are created for physician work, practice expense, and malpractice expense. If the researcher knows the CPT coding of a service, CMS
RVUs can be used to estimate the Medicare reimbursement. Medicare reimbursement is sometimes used as a surrogate for costs; it is used that way in the HERC outpatient average cost datasets.
There are limitations with the CMS RVU approach. Notably, the source of CMS RVUs, the Resource Based Relative Value System (RBRVS), was not designed to reflect the actual costs of care, but to set reimbursement as part of a public process that includes comment and input from medical specialty organizations. Critics of the RBRVS reimbursement system say that it is fundamentally flawed because of its reliance on CPT codes to represent provider services. This criticism holds that the CPT coding system can distinguish greater complexity for procedures but not for evaluation and management services (Kumetz and Goodson 2013). MCA does not use RBRVS RVUs, but instead bases its relative values on measures of resource use, such are minutes of time in the operating room or length of the clinic visit.
Given that MCA and CMS RVUs are different entities, the correlation between CMS and MCA is an empirical question. HERC creates the HERC average cost data using RVUs from CMS and other sources which researchers could use to test the correlation. Studies by Chapko et al (2008) and Phibbs and Schmit (2007) indicate that the datasets are highly correlated.
- Azoulay A, Doris NM, Filion KB, Caron J, Pilote L, Eisenberg MJ. The use of the transition cost accounting system in health services research. Cost Eff Resour Alloc. 2007;5:11.
- Chapko M, Liu C, Perkins M, et al. Equivalence of two healthcare costing methods: bottom-up and top-down. Health Econ (2008); 18(10):1188-201.
- Kumetz E, Goodson JD. The undervaluation of evaluation and management professional services: the lasting impact of current procedural terminology code deficiencies on physician payment. Chest (2013); 144(3):740-745.
- Phibbs CS, Schmitt SK. A Comparison of Outpatient Costs from the FY 2001 HERC and DSS National Data Extract Datasets. Technical Report 14. Menlo Park CA. VA Palo Alto, Health Economics Resource Center; 2007.
- Ross TK. Analyzing health care operations using ABC. J Health Care Finance. Spring 2004;30(3):1-20.
- Udpa S. Activity-based costing for hospitals. Health Care Manage Rev. Summer 1996;21(3):83-96.
Diagnosis-Related Groups (DRGs) were created by the Health Care Financing Administration (now CMS) in the early 1980s for inpatient payment. Although VA has included the DRG variable, it does not have the same meaning within VA. VA has a different definition of inpatient stay than CMS. Imagine a VA patient transferred from medical/surgical, to nursing home, back to medical/surgical, then to rehab. If these bedsections were contiguous, then VA considers this one stay and assigns it one DRG. However, CMS would treat this as more than one stay (possibly four) and assign a DRG to each. Additionally, there is a lot that affects costs and resource use beyond DRG. Therefore, we suggest using the MCA or HERC data rather than DRG to understand cost or resource use.
HERC has created guidebooks for researchers interesting in learning more about MCA and HERC data. The inpatient data webpage contains more information on DRG and how that goes into estimating payment.