HERC: Research Guide to the Managerial Cost Accounting National Cost Extracts
Attention A T users. To access the menus on this page please perform the following steps. 1. Please switch auto forms mode to off. 2. Hit enter to expand a main menu option (Health, Benefits, etc). 3. To enter and activate the submenu links, hit the down arrow. You will now be able to tab or arrow up or down through the submenu options to access/activate the submenu links.
 

Research Guide to the Managerial Cost Accounting National Cost Extracts

Menu
Menu
Suggested Citation

Phibbs CS, Barnett PG, Fan A. Research Guide to the Managerial Cost Accounting National Cost Extracts. Guidebook. Menlo Park, CA. VA Palo Alto, Health Economics Resource Center; February 2015.

Disclaimers
 

All tables for the Managerial Cost Accounting (MCA) National Data Extracts (NDE) guidebook are saved in an Excel file. Download the tables here.

Many URLs are not live because they are VA intranet only. Researchers with VA intranet access can access these sites by copying and pasting the URLs into their browser.

For a list of VA acronyms, please visit the VA acronym lookup on the VA intranet at http://vaww.va.gov/Acronyms/fulllist.cfm.

1. Overview

The U. S. Department of Veterans Affairs (VA) uses the Managerial Cost Accounting Office (MCA; formerly Decision Support System) 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 provided by VA. MCA follows practices known as activity-based cost accounting and these methods are designed to provide precise cost estimates.

This document reports on the four NDEs that the MCA refers to as the core NDEs: the inpatient discharge (DISCH), inpatient treating specialty (TRT), observation treating specialty (OBS), and outpatient (OUT, OUT2) files. The goal of this handbook is to describe the contents of the MCA core NDEs and to provide instructions on how they may be used for research. Unlike a typical data dictionary or technical manual, this handbook provides task-oriented directions for using the core MCA NDE’s. It focuses on five major topics:

  1. Accessing NDE data files
  2. The types of cost data that are included
  3. Characterization of records, variables, and facilities included in the NDEs
  4. Linking cost information in the MCA databases to clinical information in the VA utilization databases
  5. Outliers

1.1. Outpatient Extracts

The outpatient NDE consists of one record for each unique clinic encounter. If a patient has multiple encounters at a single clinic within the same day at the same VA station number, the outpatient NDE will consolidate those encounters and report only one record for that clinic. If a patient receives services from multiple clinics, there is a separate record for each clinic the patient visits. The outpatient data consist of nearly 100 million records, distributed among several files.

1.2. Inpatient Extracts

There are two views of the inpatient data: discharge and treating specialty. The discharge view (DISCH) has one record for each hospital discharge in each fiscal year. This discharge file includes the entire cost of these stays, even if they began before the beginning of the fiscal year.

The second view is by treating specialty (TRT). This view separates the inpatient stay into segments based on treating specialty (the type of unit where care was provided, also known as the bed section). A separate record is recorded for each segment of the stay per fiscal period (month). The treating specialty extract includes only utilization from a single fiscal year. It includes costs incurred by patients who have not yet been discharged. The TRT NDE is a cost report produced quarterly with monthly records. If a patient stays in the hospital more than one month (called a fiscal period in this extract), the treating specialty NDE will include multiple records, one for cost incurred in each month.

The observation treating specialty (OBS) NDE contains observation records in the TRT format (with a layout very similar to the inpatient treating specialty file). Observation records represent care in observation units that provide extensive services that should not exceed a 24-hour period. These are considered outpatient services by MCA, but the Patient Treatment files (PTF) has inpatient records for this care. As observation records are also included in the OPAT file, they should not be counted twice.

1.3. Cost Fields

Each NDE contains the total cost of the encounter and fields to identify the patient, the location of service, and the date of the encounter. In the inpatient extracts, subtotals are provided for laboratory, pharmacy, surgery, radiology, nursing and all other care. Each of these subtotals is further subdivided into fixed direct, variable direct and indirect costs. In the outpatient NDE, cost subtotals are absent due to the creation of the outpatient intermediate product department (OIPD) extract. Instead, the outpatient NDE contains total cost variables such as the total fixed direct and total fixed indirect costs and the grand total costs, which sums the total costs of all cost categories.

1.4. Data Access

MCA NDEs are stored in the VA Corporate Data Warehouse (CDW)/VA Informatics and Computing Infrastructure (VINCI)

For more information on permission to request these data see Section 2 or visit the VHA Data Portal website (VA intranet only: http://vaww.vhadataportal.med.va.gov/).

1.5. Linkage to Clinical Data

Because the MCA NDEs do not contain detailed clinical information such as ICD-9/10 diagnosis, researchers often need to merge the NDEs to the VA health care encounter files, including the Medical SAS Inpatient and Outpatient files. Chapter 7 of this guidebook describes the methods of merging each type of NDE file to the associated encounter file and presents some of the problems in merging these databases. Examples of these mergers are contained in HERC Technical Reports available at the same web site under the heading ‘Technical Reports.’

1.6. Updates

The 2015 update of this guidebook removes content now covered by the MCA NDE Technical Guide and the NDE Layout Specifications, or the VIReC website and documentation. We have updated chapter 2 to describe the new data access requirements and refer to other documentation. Chapters 3 to 7 have been updated to report the NDE variables as they are in CDW, including both SAS and SQL names.

For more information on the SAS to SQL crosswalk, see:

  • MCA NDE Technical Guide and NDE Layout Specifications (VA intranet only: http://vaww.dss.med.va.gov/nationalrptg/nr_extracts.asp)

2. Permission to Use MCA National Extracts

The VHA Data Portal provides up-to-date information on accessing VA datasets. Visit the VHA Data Portal web site at (Intranet-only: http://vaww.vhadataportal.med.va.gov/) for additional information. .

2.1. Access

Summary reports generated from the NDE files are available from the MCA Reports web site (https://mcareports.va.gov/) or via the VSSC portal (http://vssc.med.va.gov) under Resource Management > MCA Portal. Researchers can also customize the reports for a specific medical condition, facility, or both. The web site must be accessed with Microsoft Internet Explorer; other web browsers may not be fully compatible.

In addition to the MCA web reports, the MCA NDEs are now available in SAS and SQL formats at the VA Corporate Data Warehouse (CDW).

For more information on NDE data available at CDW, see the VHA Data Portal page on NDEs (VA intranet only: http://vaww.vhadataportal.med.va.gov/DataSources/MCANDEs.aspx). Documentation on the contents of each NDE at CDW can be found on the VIReC website (VA intranet only: http://vaww.virec.research.va.gov/CDW/Documentation.htm).

2.2. Documentation

There are three main sources of information on MCA Data: the VHA Data Portal, the Managerial Cost Accounting Office, and the VA Information Resource Center (VIReC).

The VHA Data Portal is a collaborative website from the VHA National Data Systems (NDS), the VA Information Resource Center (VIReC), the VA Informatics and Computing Infrastructure (VINCI), and the VHA Data Quality Program. Use the toolbar at the top of the page to navigate through information regarding data sources, data access, tools, resources, and training information from the collaborating departments.

MCA annually updates the NDE Technical Guide which provides an overview of each NDE and details any changes to variables. MCA also provides the NDE Metadata and NDE Layout Specifications.

For more information on documentation, see:

  • VHA Data Portal (VA intranet only: http://vaww.vhadataportal.med.va.gov/)
  • MCA NDE Technical Guide, NDE Metadata and NDE Layout Specifications (VA intranet only: http://vaww.dss.med.va.gov/nationalrptg/nr_extracts.asp)
  • VA Information Resource Center (VIReC) CDW summary documentation (VA intranet only: http://vaww.virec.research.va.gov/CDW/Documentation.htm

3.  Cost Data in the National Extracts

Information on the variables within each NDE is listed in the NDE Layout Specifications document, available on the MCA website. The contents of each NDE can be found on the VIReC website.

For more information see:

  • NDE Layout Specifications (VA intranet only: http://vaww.dss.med.va.gov/nationalrptg/nr_extracts.asp)
  • VIReC NDE contents (VA intranet only: http://vaww.virec.research.va.gov/CDW/Documentation.htm)

3.1. How MCA Estimates Cost

MCA extracts data from the VA accounting system, Financial Management System (FMS), and the VA payroll system (PAID). FMS and PAID track expenditures by Budget Object Code (BOC). The Budget Object Codes distinguish the type of expense, identifying specific job categories (e.g., physicians, nurses, etc.), or type of supplies or equipment. These systems also track expenditures by the service, an administrative entity such as nursing, laboratory, or medicine. Neither the Budget Object Code nor the service corresponds to a particular location where patient care is provided. Data must be entered into MCA to allocate costs to cost centers defined by their function.

This allocation of cost from FMS and PAID is done by the Account Level Budgeter (ALB). Costs are assigned to Account Level Budget Cost Centers (ALBCC). These cost centers consist of patient care departments such as primary care clinics, intensive care wards, or psychiatric units, as well as overhead departments, such as administration or environmental services. For the payroll data, MCA maintains a table that allocates each employees time to specific ALBCCs on a percentage basis. Employees can be allocated to more than one ALBCC. This is a “living” table and MCA uses current allocations when it assigns costs. The frequency of how often this table is updated varies by medical center, and by labor type and cost center within medical centers.

Data on employee activities are used to allocate expenses. The payroll expense of physicians is allocated using individual time reports completed by each physician. Some medical centers use time reports for all employees. At other medical centers, the allocation of the non-physician labor cost is based on periodic reports made by managers. The ALBCC report includes detail on each type of cost, including the Budget Object Code (BOC). This code distinguishes the labor type, such as physicians, residents, and nurses. The costs of nursing labor are allocated based on the unit or clinic where the nurse usually works. For inpatient units, these costs are adjusted by the average amount of time that the nurses on each unit float to other units.

In the next step, costs are distributed to patient care departments and then to intermediate products. This is carried out in the MCA Department Cost Manager (DCM).

The costs of a few ALB cost centers, called Exempt Accounts, are not carried from ALBCC to DCM. Exempt accounts represent costs that have no corresponding workload. An example is work costs from care that is purchased from non-VA providers where workload is not collected.

Costs of overhead are distributed to direct departments in a “step down” method. The MCA step down restricts the cost of some overhead departments so that they are only distributed to the corresponding patient care departments.

DCM tracks labor costs using three categories for employee labor and one category for contract labor. The variable labor categories include VL1 (technicians, social workers, and trainees), VL2 (nursing), VL4 (providers including physicians, dentists, psychologists, and residents), and VL5 (contracted labor).

DCM also extracts information on the workload produced by each department. This workload is a count of the number of units of each intermediate product produced by that department. An intermediate product is a specific service or product used in a hospital stay or outpatient visit. Examples include: a chest x-ray, a day in the medical ward, or a 15-minute block of time in the operating room.

As many as 13 different types of costs are tracked by MCA. For each type of cost, a separate Relative Value Unit (RVU) is used to distribute each type of department cost to the intermediate products made by that department. MCA computes two different intermediate product costs: a standard cost, based on expected department cost and workload, and an actual cost, based on the department’s actual cost and workload. VA provides sites with a national template of RVUs that medical centers may modify to reflect local conditions.

Relative values for labor costs are expressed in minutes. For example, the relative value for registered nursing labor is the number of minutes of nursing labor ordinarily required to make that product. Non-RN nursing labor is computed separately. Because of differences in operating structure, service volume, and management methods, the number of minutes allocated to the same service varies across medical centers.

To find the nursing labor cost in a product, MCA multiplies the expected minutes of nursing labor (the relative value) by the mean cost of nursing labor per relative value unit. This mean cost is found by dividing the department’s nursing labor cost by its nursing labor workload. The workload is the sum of the expected minutes of nursing labor required to produce all of intermediate products of the department. Previously these costs were calculated monthly and would fluctuate based on patient census and nurse workload. Now these costs are calculated monthly on a year-to-date basis, in order to smooth out the swings in workload (e.g., 19 workdays per month versus 22 workdays per month, low workload in November and December, etc.).

The Clinical Cost Manager (CCM) finds the number of intermediate products used in each health care encounter (e.g., in an outpatient visit or hospital stay). It multiplies the number of products used in the encounter by the cost of each product. The cost of all products is summed to find the total cost of the encounter.

3.2. Cost Data Reported in the NDEs

MCA National Data Extracts report the total actual cost of each encounter. Inpatient NDEs also report cost sub-totals or the costs incurred in a group of departments. The designers of the NDE assigned MCA departments to six mutually exclusive groups: bedday of care (nursing ward or residential), surgery, laboratory, radiology, pharmacy, and all others. Prior to FY04, cost sub totals were also reported in outpatient NDEs. Table 1 lists the departments associated with each of these cost categories. Individual product records that do not match any in the table are assigned to the “All Other” category.

Surgery cost includes costs such as pre-op, recovery, the operating suite and the recovery room only on the day of surgery. It does not include the cost of surgical clinics (for outpatient care) or the daily cost of surgical wards (for inpatient care). The bedday of care costs (formerly known as nursing costs) include the cost of operating regular acute-care wards and long term care units, but should not include any physician costs. These costs also include the bedday costs for Psychiatric Residential Rehabilitation Treatment Programs (PRRTP) and other residential treatment programs where nursing staff may not be assigned. The sum of the costs reported in the six department groups is equal to the total cost of the encounters.

Starting in FY11, the “40**” departments in RRTP will be removed because only administrative departments exist under that combination. “44**” departments will be added which correlate to the new treating specialties “1L”, “1K” and “1M”.

Each cost sub-total is divided into three categories: fixed direct costs, variable direct costs, and indirect costs. Direct costs are those that are directly attributable to a patient care department. Costs that are incurred regardless of the volume of services provided are considered fixed costs. Costs that vary with the volume of services provided are variable costs. Variable costs consist of supplies and the cost of labor that might be released if workload decreased. Indirect costs are the costs of overhead departments such as housekeeping, engineering, and administration. Because indirect costs are fixed in the short-term, the category of variable indirect costs does not exist. Total cost is the sum of fixed direct (FD), fixed indirect (FI), and variable direct costs (VD).

3.3. Cost Information in Current Year File

NDEs for the current year include information from the beginning of the fiscal year up to the current month. For example, the March extract contains cost information from October 1st to March 31st. Because a new cumulative extract is created each month, the accuracy of the cost estimate increases as the fiscal year progresses. Cost estimates for earlier months in the fiscal year are less accurate because MCA does not revisit previous months to reassign costs.

There may be some change in costs as new cumulative extracts are created during the fiscal year. MCA finds the actual cost of intermediate products by dividing the total cost of a department by its total workload. Each cost type for each workload product is expressed in relative value units. Since cost and workload change as the year progresses, the unit cost of an intermediate product may change monthly as the year progresses. A particular intermediate product reported with one cost in the February extract may have a slightly different cost in the March extract.

A final extract is created at the end of the federal fiscal year representing the period from October 1st through September 30th. The final extract reflects the year-to-date average costs of the fiscal year.

3.4. Costs for Prior Years Utilization

MCA costs are based on unit costs of intermediate products computed in the same fiscal year as the year of the file. For example, costs in the FY11 file would be based on FY11 costs, not FY10 costs. The discharge file contains information on hospital stays that ended in the current fiscal year. Some of these stays began in a previous fiscal year. The cost of utilization that is from a prior fiscal year is computed using the unit costs of the fiscal year in which the patient was discharged.

3.5. Cost Information for New and Merged Facilities

When two facilities are merged, the legacy facility becomes a division of the primary facility. The new legacy facility’s station number is the primary facility’s station number followed by a suffix. For example, the medical center identification number (the variable called “STA3N”) only contains 3 digits of a station number. Cost information for the legacy facility is reported under the old station number before the integration and under the primary facility’s STA3N number after the integration. If the merger occurs at the beginning of a fiscal year (i.e. October 1), the legacy facility’s old station number will disappear from the new fiscal year and its cost information will be under the primary facility’s station number. However, if two facilities merge in the middle of a fiscal year, encounters that occurred before the integration will be recorded under the legacy facility’s old station number and encounters that occurred after the merge will be under the primary facility’s station number. In this case, the legacy facility’s old station number in the NDE files appears until the month of the integration. New and merged facilities that have taken place since 1998 are reported in Table 2.

3.6. Utilization Not Reported

HERC evaluated whether utilization was recorded for all stations for each fiscal period in the PTF and MCA files. In 2001 and 2002, MCA did not report utilization for a few stations during the last months of the fiscal year, suggesting that processing at these sites was incomplete. Since 2003, MCA has consistently reported utilization for every station for every month of the fiscal year, with two exceptions. The MCA treating specialty file reported stays for Salisbury, NC only during the first month of FY03. This station reported inpatient care in every month in the PTF and MCA discharge files during that year. No stays were reported for New Orleans, LA in the last month of FY05 since this medical center was shut down because of Hurricane Katrina in late August 2005. Caution is urged in analyzing data from New Orleans after hurricane Katrina. As of FY08, station 629 was still closed but community based nursing homes associated with station 629 post data on occasion.

3.7. Comparability of MCA Costs to Costs of Non-VA Providers

Analysts often want to compare VA costs to non-VA costs. A common source of non-VA cost estimates is Medicare, and comparisons must consider differences in what is reported between MCA and Medicare. VA hospitals include the costs of physician services while Medicare does not since they are reimbursed separately from hospitals. VA physician services costs, however, do not include the cost of physician malpractice liability as these costs are covered by the U.S. Justice Department settlement payments. In calculating indirect costs, MCA includes VA central office and national centers operating costs in addition to hospital administrative costs while Medicare hospitals use only hospital administrative costs. Also, the VA does not include financing costs related to capital acquisitions since these expenses are covered by the U.S. Treasury while Medicare adds in these costs.

Differences in Cost Determination between VA and Medicare Hospitals
Cost Type VA Medicare
Physician services Included in hospital costs Excluded from hospital costs
Indirect costs VA central office and national center operating cost plus hospital administrative costs Only hospital administrative costs
Capital acquisitions Financing costs excluded Financing costs included

Moreover, a recent analysis comparing the costs of hospital discharges from VA hospitals versus Medicare national average costs for the same discharges adjusted for differences in capital, physician costs, and wage differentials by geographic area and found higher costs among a majority of VA hospitals. Another analysis limited to care for acute myocardial infarction found lower costs in VA hospitals compared to Medicare hospitals. These cost differences at VA hospitals may reflect differences in provider practice patterns, patients’ disease severity, the large amount of care provided by medical residents in VA hospitals, and other structural differences between VA and non-VA hospitals.

3.8. Data Changes to Key Variables in FY 2008 MCA Core NDEs

The Managerial Cost Accounting Office recognized that several key variables (e.g., total cost, length of stay, etc.) were missing from the FY08 National Data Extracts (NDEs). These variables are now included in the NDE files at CDW. However, researchers can easily compute them. The variable names listed below are the SAS file variables.

For more information on variables see:

  • MCA Layout Specifications (VA intranet only: http://vaww.dss.med.va.gov/nationalrptg/nr_extracts.asp)

3.8.1. Total Cost Variable in the MCA Discharge (DISCH) NDE

The total cost variable, DCST_TOT, is the sum all the fixed direct (variable names end in “_FD”), fixed indirect (variable names end in “_FI”), and variable direct (variable names end in “_VD”) subtotals. For convenience, the following SAS code may be copied and pasted into the program.

DCST_TOT=sum(DAO_FD, DLAB_FD, DNUR_FD, DPHA_FD, DRAD_FD, DSUR_FD, DAO_FI, DLAB_FI, DNUR_FI, DPHA_FI, DRAD_FI, DSUR_FI, DAO_VD, DLAB_VD, DNUR_VD, DPHA_VD, DRAD_VD, DSUR_VD);

3.8.2. Total Cost Variable in the MCA Treating Specialty (TRT) NDE

The total cost variable, TCST_TOT, in the Treating Specialty NDE can be derived using the following SAS code.

TCST_TOT=sum(TAO_FD, TLAB_FD, TNUR_FD, TPHA_FD, TRAD_FD, TSUR_FD, TAO_FI, TLAB_FI, TNUR_FI, TPHA_FI, TRAD_FI, TSUR_FI,    TAO_VD, TLAB_VD, TNUR_VD, TPHA_VD, TRAD_VD, TSUR_VD);

3.8.3. Length of Stay Variable in the Treating Specialty File

To derive the Treating Specialty length of stay variable, TRT_LOS, use the following SAS code.

if TRTIN ne . then TRT_LOS=max(TRTOUT-TRTIN,1);

If a record consists of a non-missing value in the TRTIN variable (date of entry into a treating specialty segment), then the treating specialty length of stay is the difference between TRTOUT (date of exit from treating specialty segment) and TRTIN. If this difference is zero, then TRT_LOS will equal one because an inpatient stay is at least one day.

3.8.4. Budget Object Code (BOC) in the MCA Account Level Budget Cost Centers (ALBCC) Report

The Account Level Budget Cost Centers (ALBCC) report includes detail on each type of cost, including the Budget Object Code (BOC). This code distinguishes the labor type, e.g., physicians from nurses. In FY 2008, this variable was dropped from the ALBCC. However, it can be derived by taking the first four characters of the ALB account variable, ALBACCT, using the following SAS code:

            BOC=substr(ALBACCT,1,4)

3.8.5. Cost Center Variable in the ALBCC Report

The Cost Center variable, CC, was also been dropped in the FY 2008 ALBCC. Cost centers consist of patient care departments, such primary care clinics, intensive care wards, or psychiatric units, as well as overhead departments, such as administration or environmental services. To derive this variable, use the following SAS code:

CC=substr(ALBACCT,5,3);

3.8.6. Changes to the Diagnosis Related Group (DRG) Variable

Effective October 1, 2007, the Centers for Medicare and Medicaid Services (CMS) adopted a new Diagnosis Related Group (DRG) classification system called MS-DRG. In response to this new system, MCA has replaced variables such as DRG and ADMITDRG with DRGMS and ADDRGMS, respectively. This makes MCA consistent with the FY2008 PTF, which also uses the MS-DRG system.


4. Outpatient Cost Extract

The outpatient cost (OUT, OUT2) National Data Extracts consists of information on all VA outpatient visits, as well as the cost of outpatient laboratory, pharmacy, ancillary services, and other care not tied to a specific outpatient visit.

4.1. Outpatient Extract Files

Outpatient cost data are grouped in one file per each VA network (VISN). MCA OUT data continue to consolidate (into a single record) multiple visits having the same clinic stop on the same day at the same station number. That is, MCA OUT and OUT2 allow only one record for a unique station-patient-day-clinic stop combination.

Starting in FY11, the OUT NDE file includes the low-cost encounters previously included in OUT2. The low cost outpatient NDE files (OUT2) contain information on outpatient encounters that were assigned low cost by MCA. “Low cost” data refer to those outpatient encounters that are assigned costs between -$1 and $1. The OUT2 NDE file now only contains encounters with an absolute cost of zero. The low cost files, which have supplemented MCA outpatient extracts since FY02, are important in identifying VHA outpatient encounters.

Records with the daily cost of outpatient pharmacy are kept in the same extract as other OUT records. That is, all clinic stops (including pharmacy pseudo clinic stops 160 and 161), encounters and outpatient costs (including pharmacy costs) are included in the OUT file.

Records with pseudo clinic stops 160 and 161 include medications dispensed in the outpatient clinics, medications dispensed by the outpatient pharmacy, and medications mailed to the patient by the Consolidated Mail Outpatient Pharmacy (CMOP). The total cost of a pharmacy record includes the total cost of all medications dispensed, but does not include any information about what medications were dispensed. Prescription-level pharmacy data are found in the PHA NDE, one of the clinical NDEs.

4.1.1. Negative Values in MCA Pharmacy Data

MCA PHA NDE consists of information from three VistA data sources: Outpatient, IV and Unit Dose packages. Some records in the inpatient and outpatient MCA pharmacy data may contain negative quantity and/or cost values for prescriptions. All three packages allow returns thereby creating transactions with negative balances. The following are a few explanations of why negative balances exist in the MCA pharmacy data:

Records with negative values in the VistA IV and Unit Dose packages may appear as outpatient records in the NDE. This explains why there is a large number of negative values in the MCA Outpatient Pharmacy record. The MCA VistA extracts check all records against the Patient Movement File (#405) as they are created. The Application Program Interface (API) obtains the patient's internal entry number (DFN) and date/time of the occurrence and checks to see if the patient was admitted at the time. If so, the MCA extracts mark that record as an inpatient record, otherwise they mark it as an outpatient record. If either IV or Unit Dose returns from a ward are recorded in their respective VistA Pharmacy packages after the patient has been discharged, MCA will mark the record as outpatient.

MCA outpatient pharmacy records with negative balances may also emerge from Pharmacy IV and Unit Dose returns made on a different date than the date the prescriptions were issued. MCA creates a separate encounter for each combination of SSN + Date + Primary Stop Code. Since returns are processed at a date or time after discharge, MCA records them as an outpatient transaction. This is assuming that at large hospitals, there are separate inpatient and outpatient sections of the pharmacy. For instance, if a patient receives multiple prescriptions on a given day and one of them is returned on a different day, MCA will create two separate encounters: one from the multiple issues on the first day and a separate one for the return because the return was made on a different day.

Medications that are issued in individual-dose amounts to patients for consumption on the same day, such as those from the ward, are pulled a day in advance. Given that a large hospital will have large quantities of prescriptions to fill and will have to allow time for quality control, the pharmacy technician has to prepare a day’s supply of prescriptions prior to the day of patients’ consumption. Many of the ward medications are returned after the patient has been discharged. Since they process at a date/time after discharge, MCA records them as an outpatient transaction.

Consolidated Mail Outpatient Pharmacy (CMOP) undeliverable medications that are returned and turned back into stock also contribute to negative quantities in the MCA pharmacy data.

4.2. Variables

For more information on variables see the MCA NDE Layout Specifications (Intranet-only: http://vaww.dss.med.va.gov/nationalrptg/nr_extracts.asp).

4.2.1. Flag Variables

The MCA outpatient National Data Extract is based on a sequential extraction of outpatient services from the VA electronic medical record (VistA). MCA uses flag variables to identify the primary source of information. Data from the NPCD are indicated by setting the variable NPCD to “Y”. Other sources are indicated with flags suggesting data came from clinic appointments (SAS: CLI/SQL: CLIFLAG), the prosthetics database (PROS/PROSFLAG), or other types of utilization that could not be assigned to an encounter (UTIL/UTILBUILTFLAG).

Due to the creation of the No-Show NDE the no-show flag field was removed from the outpatient and outpatient intermediate product department (OIPD) NDE, effective FY08. Flag variables are presented in the DSS NDE Technical Guide in the OUT section.

4.2.2. Cost Variables

The outpatient extract contains cost variables that total fixed direct (OCST_FD/ACTDIRCOST) and fixed indirect (OCST_FI/ACTINDIRECTCOST) costs. The total cost of each outpatient visit (OCST_TOT/ACTTOTCOST) is the sum of fixed direct (OCST_FD/ACTDIRCOST), fixed indirect (OCST_FI/ACTINDIRCOST), and variable direct costs (OCST_FD/ACTDIRCOST). Indirect costs are always fixed, thus there are no variable indirect costs.

For years prior to FY05, MCA also includes variables with sub-totals of the outpatient cost incurred in groups of departments: laboratory (OLAB), nursing (ONUR), pharmacy (OPHA), radiology (ORAD), surgery (OSUR), and all other (AO). These cost sub-total fields were dropped from FY05 and subsequent years because they were made unnecessary by the outpatient intermediate product department (OIPD) extract. Analysts may use the OIPD extract to find the cost incurred in each department during any outpatient visit. The IPD extract is documented in a separate HERC guidebook, see Guidebook for the MCA Intermediate Product Department Files on the HERC guidebooks page at http://www.herc.research.va.gov/include/page.asp?id=guidebooks.

Also contained in this extract are variable direct costs. These include variable labor categories 4 and 5 (VL4 and VL5) in the surgery (SUR), radiology (RAD), and all other (OTH) departments. These are the costs of providers, including physicians, psychologists, residents, dentists, etc. Variable labor category 4 (VL4) represents the cost for employee providers and VL5, the cost for contracted services. VL4 is the cost for employee providers and VL5 is the cost for contracted services. You can find the costs using the following formulas:

VL4 = Encounter Department Variable Direct Cost x (A/B),

where: A=Total Department Variable Labor Cost (4)

B=Total Department Direct Variable Cost (variable labor + variable supply + variable other)

 

VL5 = Encounter Department Variable Direct Cost x (A/B),

where: A=Total Department Variable Labor Cost (5)

B=Total Department Direct Variable Cost (variable labor + variable supply + variable other)

Note: *** denotes a 2 or 3 digit code representing one of three departments: RAD, SUR or AO.

In FY11, additional cost variables were added to include fields that would represent variable cost. Currently only DIRECT COST, INDIRECT COST and TOTAL COST are included in these NDEs with no way for users to compute variable cost. VAR DIR COST was added to indicate the cost directly related to patient care workload that fluctuates with the volume of products. This variable is computed from CHGDTL records. VAR SUPP COST was also added to indicate the cost of supplies directly related to patient care that fluctuates with the volume of products.

In the initial versions of the MCA outpatient NDE, the nursing cost category (see Table 1) was not included. This was because the nursing category in MCA was designed to capture all of the regular daily costs associated with inpatient units, of which nurses are a major component. In FY02, nursing (ONUR) was added to the outpatient files. The MCA NDE Technical Guidebook indicates that the purpose of the nursing cost category in the outpatient NDE is to capture the costs associated with contract nursing homes and state veterans homes and observation cases.

4.2.3. Utilization and Diagnostic Variables

The national extract includes basic information about the quantity of health services utilized by patients during their encounter. The unit for each cost category is defined differently.

Starting in FY12, additional fields were added to the outpatient extract files. Provider NPI (PROVNPI/PROVNPI) and Primary Care NPI (PCPNPI/PCPNI) provide additional NPI information for the existing NDE fields Provider (PID_DSS/PROV) and primary care provider (PCP_DSS/PCPROV). 15 secondary provider fields were added to the VistA CLI extract as an enhancement during FY12 conversion to help report the multiple providers that can be associated with a single encounter.

A new variable PS (period of service) was created in FY12 and added to the outpatient (OUT) and inpatient (DISCH and TRT) extracts. This new field helps identify the Active Duty cohort in MCA web reports, cubes and dashboards.

For more information see:

  • NDE Layout Specifications (VA intranet only: http://vaww.dss.med.va.gov/nationalrptg/nr_extracts.asp)

4.3. Non-VA Long-term Care Records

The outpatient extract includes records representing the cost of care from non-VA long-term care facilities. In FY09, about $591.7 million in costs were reported under long-term care clinic stops in the MCA outpatient file in FY09 (see Table 3). This includes care provided by community nursing homes, state veterans homes, and residential facilities.

In FY10, MCA began posting Community Nursing Home (CNH) PTF records as uncosted ENCTR outpatient records using the value 'CNH'. The previous use of Stop Code 649 for CNH was discontinued. This is in line wiht other uses of alphabetic stop code values for when those values are assigned solely within MCA, such as 'PHA' for prescription and 'PRO' for prosthetics items.


5. Inpatient Discharge Extracts

5.1. Discharge File

The discharge extract file (DISCH) provides one record for each hospital stay that ended during the fiscal year. If the hospital stay began in an earlier year, the MCA system finds the cost of a hospital stay by summing up the cost of all intermediate products and services provided during the stay. Unit costs of intermediate products are from the year of discharge, even if the stay began in a different fiscal year. When a stay crosses fiscal years, cost may differ from what is reported in the treating specialty file (described in Chapter 6). The treating specialty files use the unit cost of the year when care was provided.

MCA was implemented throughout VA in 1998. In the first few years of MCA data, cost day are incomplete for stays that began before MCA was implemented at the site.

5.2. Variables

A table of complete variable lists for the NDEs can be found in the NDE Layout Specifications at (VA intranet only: http://vaww.dss.med.va.gov/nationalrptg/nr_extracts.asp).

The inpatient files report costs using six core categories: laboratory (LAB), bedday (formerly nursing), pharmacy (RX), radiology (RAD), surgery (SURG), and all others (AO). The bedday of care (nursing ward or residential) department represents the cost of nursing departments for inpatient stays (nursing staff), plus all of the regular operating costs of the inpatient units. These cost variables are described in detail in Chapter 3. Bedday of care costs are added for inpatient care. Variable direct costs such as variable labor categories 4 and 5 in the surgery (SAS: DSUR_VL4/SQL: SURGVL4 and DSUR_VL5/SURGVL5), radiology (DRAD_VL4/RADVL4 and DRAD_VL5/RADVL5), and all other (DAO/OTH) departments and the variable cost of pharmacy supply (DPHA_VS) are already included in one of the six core cost categories; adding these costs to the totals will double count them. Fields for variable labor and supply costs represent supplemental information about costs already included as variable direct cost. The cost of the radiology department for variable labor category 4 (DRAD_VL4/RADVL4, physician employees) and category 5 (DRAD_VL5/RADVL5 contract physicians) are also reported in the variable direct costs of the department (DRAD_VD/RADVD). Variable labor cost of surgery (physician employees DRAD_VL4/RADVL4, contract employees DRAD_VL5/RADVL5), are included in surgical variable direct costs DSUR_VD/SURVD. The variable supply costs of pharmacy (DPHA_VS/RXVS) are included as variable direct cost of pharmacy (DPHA_VD/RXVD).

The total cost of each inpatient discharge (DCST_TOT/TOTCOST) is the sum of fixed direct (DCST_FD/TOTFD), fixed indirect (DCST_FI/TOTFI), and variable direct costs (DCST_VD/TOTVD).  Indirect costs are always fixed, thus there are no variable indirect costs.  The cost incurred in a specific department (nursing, laboratory, radiology, surgery, pharmacy, prosthetics or all other) cost is the sum of fixed direct, fixed indirect, and variable direct costs of that department.

In FY11, additional provider fields were added to clearly indicate the attending provider at time of discharge, separate from the patient’s primary care provider. New variables include: provider IEN (DISMD/DISCHMD), provider type/person class (DISMDPT/DISMDPT), and provider’s National Provider Index (DISMDNPI/DISMDNPI).

Starting in FY11, OPADMIT/OPADMIT was added to the TRT NDE and DISCH NDE to identify an inpatient admission resulting from a previous outpatient visit. In FY12, period of service (PS/PS) and eligibility (ELIGIBILITY/ELIGIBILITY) was added to the inpatient extracts (DISCH and TRT NDEs) to help identify the Active Duty cohort in MCA web reports, cubes and dashboards.


6. Inpatient Treating Specialty File

6.1. Treating Specialty File

The treating specialty (TRT) file provides detailed information on inpatient care. A record represents a person’s stay per treating specialty per month. Each stay is divided into segments based on the month and treating specialty of the provider responsible for each part of the stay. The treating specialty is ordinarily associated with a location, such as a medical care or surgical ward, or a long-term care unit. It is also called a bed section. The treating specialty NDE is a monthly cost report. If a bed section stay crosses multiple fiscal periods (months), the TRT file will contain one record for each month of the same bed section stay. Records belonging to the same bed section stay can be linked by station (SAS: STA3N/SQL: STA3N), patient identifier, such as scrambled social security number (SCRSSN/SCRSSN), treating specialty start date (TRTIN/TXSPSDT), and treating specialty end date (TRTOUT/TXSPEDT) where treating specialty start date (TRTIN/TXSPSDT) and treating specialty end date (TRTOUT/TXSPEDT) are the bed section admission and discharge dates.

The TRT file includes all care provided during that file’s fiscal year. It does not include the cost of care provided in the previous fiscal year.

  • If a patient has not yet been discharged from the hospital as of the last day (September 30) of the current fiscal year, a census flag is designed to be set to Y on those records having a bed section discharge date of September 30. This indicates that the patient was still in the hospital, counted in the hospital census taken on the last day of the fiscal year. Otherwise, if the patient was discharged from the hospital before the last day of the current fiscal year, the census flag is set to N, as the patient was not in the census.
  • If a patient was admitted to the hospital prior to the beginning of the current fiscal year and is discharged from the hospital in the current fiscal year, the cost of care provided during the previous fiscal year will be reported in the previous year’s treating specialty file. Only that care provided during the current fiscal year will be reported during the current fiscal year’s file.
  • The MCA treating specialty file often reports the cost of a single bed section stay in a single record. However, because the purpose of the MCA treating specialty file is to report the monthly cost of all inpatient stays, it reports the cost of a single bed section stay in two or more records if the start and end dates of the stay span across two or more months. For example, if a stay starts on January 20 and ends on February 5, the NDE treating specialty file would contain two records for the single stay; the first including the cost for the 11 days in January and the second for the four days in February.

TRT files contain data beginning in FY99. The TRT file includes the cost of all inpatient care that was provided during the fiscal year. The discharge file (Chapter 5) includes the cost of stays that ended during the fiscal year. The care reported in these files overlaps, but each file includes cost not included in the other.

The treating specialty file includes the cost of stays that were not yet over by the end of the fiscal year. The discharge file excludes these costs. The discharge file includes the total cost of stays that began before the beginning of the fiscal year. The treating specialty file includes only part of their cost – the cost that was incurred since the beginning of the fiscal year. The total cost of each inpatient treating specialty record total cost (TCST_TOT/ TOTCOST ) is the sum of fixed direct (TCST_FD/TOTFD), fixed indirect (TCST_FI/TOTFI), and variable direct costs (TCST_FD/TOTFD). Indirect costs are always fixed, thus there are no variable indirect costs.  The cost incurred in a specific department (nursing, laboratory, radiology, surgery, pharmacy, prosthetics or all other) cost is the sum of fixed direct, fixed indirect, and variable direct costs of that department.

VA provides long-term care, and some patients have exceptionally long stays, of many years duration. Neither file reports the complete cost of stays that began before MCA was implemented at the site.

There is no national file of MCA inpatient data after FY07. For FY08 and later years, only VISN level files are available.  These files began to be created in FY04.

6.2. Variables

NDE Layout Specifications (VA intranet only: http://vaww.dss.med.va.gov/nationalrptg/nr_extracts.asp)

Costs in the treating specialty file are reported by the same groups of departments that are used in the discharge file.

The utilization and diagnosis variables for the inpatient treating specialty file are the same as those in the discharge extract. The treating specialty extract contains two variables with diagnoses: admitting Diagnosis Related Group (ADMITDRG/ADMDRG) and admitting diagnosis (ADMITDX/ADMICD9) and a variable identifying the treating specialty (TRTSP/TXSP). Please note that the admitting DRG is not the bed section admitting DRG. It is the admitting DRG for the entire inpatient stay, which is the same as the admitting DRG in the discharge file for the same inpatient stay. Similarly, the admitting diagnosis is the same as that in the discharge file for entire inpatient stay.

Starting in FY12, period of service (PS/PS) and eligibility (ELIGIBILITY /ELIGIBILITY) were added to the inpatient extracts (DISCH and TRT NDEs) to help identify the Active Duty cohort in MCA web reports, cubes and dashboards.


7. Merger of the MCA NDEs with MedSAS files

Because the MCA National Data Extracts (NDEs) do not contain detailed clinical information such as procedures and diagnosis, researchers often need to merge the NDEs to the VA health care encounter files, including the Medical SAS (MedSAS) Inpatient and Outpatient Files. This chapter describes the methods of merging each of the NDE files to the associated encounter files and discusses some of the problems of merging these databases. Information on how well these databases compared in FY04 and prior years is contained in HERC Technical Reports. The documents are available in the publications section of the HERC Intranet web site. (http://vaww.herc.research.va.gov/publications/default.asp).

With the data transition to CDW, NDS transitioned data from SAS files into SQL tables. Prior to 2013, MCA created one SAS file per NDE per fiscal year. At CDW there is now one SQL table for each NDE, combining all years of a single NDE. With this transition, variable names were harmonized across all years. The Managerial Cost Accounting Office (MCA; formerly Decision Support Office) created a file (NDE Layout Specifications) showing the conversion of SAS to SQL variables and any changes in variable names across years. Details of any changes to variables are recorded in the MCA’s NDE Technical Guide.

The MedSAS Inpatient dataset is sourced from the Patient Treatment File (PTF), which we refer to in the sub-sections below. Until Fiscal Year (FY) 2019, The MedSAS Outpatient dataset was sourced from the National Patient Care Database (NPCD), which is referenced below. Beginning FY 2019, the MedSAS Outpatient dataset is created using data from the Corporate Data Warehouse (Outpatient SQL Files (CDW_SE_IE_SF). This replacement dataset is created using the same business logic as the dataset sourced from NPCD data. Visit the VHA Data Portal Medical SAS page to learn more about MedSAS data: http://vaww.vhadataportal.med.va.gov/DataSources/MedicalSASInpatientOutpatientDataSets.aspx (VA intranet only).

For more information see:

  • MCA's NDE Layout Specifications  (VA intranet only: http://vaww.dss.med.va.gov/nationalrptg/nr_extracts.asp)
  • MCA’s NDE Technical Guide (VA intranet only: http://vaww.dss.med.va.gov/nationalrptg/nr_extracts.asp)

7.1. NDE and PTF Discharge Files

The MCA and PTF discharge files can be joined by the key variables, SCRSSN, ADMITDAY, STA3N and DISDAY. The inpatient data are organized differently in the two databases (MCA and PTF). Discharges from all VA facilities including hospitals, nursing homes, residential programs, and domiciliary, are recorded in a single file in the MCA national data extract. Discharge records in the PTF database are grouped into three files: 1) the PM file for discharges from hospital main bed sections, 2) the XM file for discharges from VA nursing homes, domiciliary and other residential health care bed sections, and 3) the PMO file for discharges from VA hospital observational bed sections. As MCA began classifying (in FY2002) all observational beds as outpatient encounters, discharge records from the PMO file should be excluded from the comparison process in the following sections that describe how the MCA and PTF discharge files were linked.

7.1.1. Merging Variables

The MCA and PTF databases can be merged using the following four common variables: (1) scrambled Social Security Number (SCRSSN/SCRSSN), (2) medical center identification number (3-digit numeric STA3N/STA3N), (3) admission date (ADMITDAY/ADMITDAY), and (4) discharge date (DISDAY/DISDAY).

7.1.2. Community nursing home deletions

Before linking records from the two databases, one should first exclude discharges from community nursing homes that were recorded in the PTF XM file because the MCA inpatient files do not include discharges from community nursing homes. However, MCA records community nursing home stay data in the outpatient cost (OUT) file for each fiscal year; see Section 4.3. The community nursing home stays are identified by the variable STATYP=42 in the XM file.

7.1.3. Deletion of duplicates

Records with the same values for the SCRSSN/SCRSSN, STA3N/STA3N, ADMITDAY/ADMITDAY and DISDAY/DISDAY variables should be considered to be duplicates. After removing the community nursing home stays from the XM file, one should check for (and delete) duplicate records within and between the PM and XM files.

7.1.4. Summary of the comparison between MCA and PTF discharge files

As all but 0.02 percent of the MCA and PTF discharge file records could be matched in FY04, researchers should be able to link these two databases accurately in subsequent years for inpatient discharges. Although PTF contained a few more discharges than the MCA NDE file in FY04, it is unlikely that the missed discharges will affect the results of most studies. Researchers could use the HERC average cost estimates to estimate the cost of the discharges that are found only in the PTF.

7.2. NDE Treating Specialty and PTF Bed Section Files

The MCA treating specialty file reports the monthly cost of each bed section stay. Records from the same bed section in a single hospital stay can be consolidated to give the total cost of each bed section stay and then combined with the PTF bed section files to create a file with cost and clinical information for each bed section stay.

The MCA treating specialty file represents a single hospital stay with one or more different records. At the end of each month, a record is generated for each bed section visited by the patient during the stay. Records are generated even if the patient has not yet been discharged. Records are kept by fiscal year in which the service was provided. A stay that begins in one fiscal year and ended in another fiscal year will have records in more than one year’s MCA treating specialty file.

The PTF bed section file includes one record for each bed section visited by the patient. The record is generated when the patient is discharged. Records are kept by fiscal year in which the patient was discharged from the hospital. The patient may spend time in another bed section before finally being discharged.

The PTF files include diagnoses, demographics, and other information not found in the MCA treating specialty file. The MCA treating specialty records have cost data not found in the PTF. Analysts often want to combine these data sources.

To determine the costs of the entire time the patient stayed in a bed section requires an understanding of the differences between the PTF bed section file and the MCA treatment specialty database. Stays in a bed section are represented by one record in the PTF bed section file, but if that stay spans more than one calendar month, it is presented by two or more records in the MCA treating specialty file. Costs from the monthly records in the treating specialty file must be added together to get the total costs of that bed section stay. If the stay began in one year and ended in another, these records will be found in MCA treatment specialty files for different years.

Bed section stays in the MCA NDE treating specialty extract can be compared with those in the PTF bed section files. MCA includes all records classified by inpatient treating specialty (i.e., bed section) into a single file, whereas the PTF separates bed section records into three files: the main bed section file (PB), the extended bed section file (XB), and the observation bed section file (PBO). Effective FY03, MCA treated stays in observation bed sections as outpatient care. Therefore, records in observation bed sections should be excluded in a comparison of the MCA treating specialty NDE with the PTF bed section files.

The MCA treating specialty file often reports the cost of a single bed section stay in a single record. However, because the purpose of the MCA treating specialty file is to report the monthly cost of all inpatient stays, it reports the cost of a single bed section stay in two or more records if the start and end dates of the stay span across two or more months. For example, if a stay starts on January 20 and ends on February 5, the NDE treating specialty file would contain two records for the single stay; the first including the cost for the 11 days in January and the second for the four days in February. The PTF files, however, only include one record for each single stay. Also, the PTF has a census file that contains stays that are not discharged or transferred by the end of the fiscal year, whereas the NDE treating specialty file includes those records in the last month of a fiscal year. Because of these structural differences between the two data sets in FY04, the number of records (not bed section stays) in the NDE treating specialty file is normally larger than the total number of records (bed section stays) in the three PTF files.

7.2.1. Methods

This section describes adjustments and methods that should be used for a comparison of the MCA treating specialty NDE with the PTF bed section files. We use the term bed section admission date for the date a patient is admitted or transferred to a bed section. The term bed section discharge date refers to the date a patient is discharged or transferred to another bed section.

7.2.2. Census records

Stays that were not discharged at the end of the fiscal year in the NDE Treating Specialty file (census stays) should be dropped, even though there is a PTF census file to match those census stays. Census stays in the NDE treating specialty extract may be excluded by eliminating all records with a value of “Y” (indicating that a patient was still in the hospital at the end of the fiscal year) for the census stay variable (CENSUS/CENSUS).

7.2.3. Consolidation of MCA Treating Specialty Records

Monthly records in the MCA treating specialty file should be consolidated into one record for each unique bed section stay. As previously indicated, the MCA treating specialty file reports the cost of a single bed section stay in two or more records if the start and end dates of the stay span across two or more months. These multiple records have the same values for five variables (SCRSSN/SCRSSN, STA3N/STA3N, TRTIN/TXSPSDT, TRTOUT/TXSPEDT, and TRTSP/TXSP). The treating specialty file therefore should be consolidated into one record per bed section stay using these variables.

7.2.4. Community Nursing Homes

Since the MCA NDE treating specialty extract does not contain data from community nursing homes, community nursing home stays from the PTF XB file should be excluded by eliminating records with “STATYP=42.” The main and bed section files can be merged by STATYP to identify stays in community nursing homes.

7.2.5. Duplicated records

Records within a file with the same values in the five aforementioned variables (SCRSSN/SCRSSN, STA3N/STA3N, TRTIN/TXSPSDT, TRTOUT/TXSPEDT, and TRTSP/TXSP) should be considered duplicates and removed before comparing the treating specialty and PTF files.

7.2.6. Variables used in the ‘MERGE’ statement

The MCA treating specialty and PTF bed section files can be merged on the following five variables: (1) scrambled Social Security Number (SCRSSN/SCRSSN), (2) medical center identification number (3-digit numeric STA3N/STA3N), (3) bed section admission date, (4) bed section discharge or transfer date, and (5) bed section number.

Note that three pairs of variables have different names in the two databases. The bed section admission date is named as “BSINDAY” in the PTF and named as “TRTIN/TXSPSDT” in the treating specialty file. The bed section discharge or transfer date is named as “BSOUTDAY” in the PTF and is named as “TRTOUT/TXSPEDT” in the treating specialty file. The SAS name for bed section number is BEDSECN in the PTF and is TRTSP/TXSP in the MCA file. Table 4 lists the equivalent variables in the two data sets.

7.2.7. Summary of the comparison between MCA Treating Specialty and PTF Bed Section file

The FY04 MCA treating specialty and FY04 PTF bed section files were merged on the five aforementioned variables (scrambled social security number, station number, inpatient bed section admission date, discharge date, and bed section number). Over 73% of the MCA bed section stays matched exactly by those five variables with the PTF database. These files have different rules for setting bed section admission and discharge dates, however. Admission or discharge dates sometimes differ by one day; patients admitted in a prior fiscal year were sometimes assigned the first day of the fiscal year as their admission date. When these differences were adjusted for, the comparison attained was very high, as the PTF files included 97.8% of the stays in the MCA treating specialty file. The MCA treating specialty file included 99.7% of the stays recorded in the acute care (PB) PTF bed section file and 97.4% of the stays in the extended care (XB) PTF bed section file.

7.2.8. Recommendations to researchers

If researchers need to link bed section stays in the MCA and PTF data sets, they should first extract all bed section stays from the MCA and PTF databases for the study subjects using scrambled social security number regardless of other information. Then, researchers can adjust as explained above to conduct a match. If a small percentage of unmatched MCA stays remain, researchers may link them in the two data sets by matching on SCRSSN, bed section number and on any combination of two of the three remaining variables. For any few remaining unmatched records, manual examination is necessary.

7.3. Comparison between MCA Treating Specialty and MCA Discharge File

To validate whether stays and costs reported in the MCA treating specialty file are consistent with stays reported in the MCA discharge file, these files may be compared. The MCA treating specialty file contains records for only that part of the bed section stay that took place during the fiscal year. If a stay began in a previous fiscal year, that part of the stay is not in the current fiscal year. Rather, it is in the earlier year’s MCA treating specialty file. If a bed section stay lasts more than a single fiscal period (month), there will be multiple records for the same stay. The discharge file, however, provides one record for each hospital stay that ended during the fiscal year. Accordingly, the records in the treating specialty file should be summarized so that they would have the same format as the discharge file, with one record per hospital stay.

Stays that had not ended by the end of the fiscal year should be excluded (from the treating specialty file), because such stays are not reported in the discharge file. In addition, stays that began before the first day of the fiscal year should be excluded (from both the discharge and treating specialty files) because costs are included in the discharge file but not in the treating specialty file.

7.3.1. Summary of the comparison between MCA Treating Specialty and MCA Discharge file

When the FY04 MCA treating specialty file was compared with the MCA discharge file, some stays were contained in the discharge file but not in the treating specialty file. Between FY03 and FY04, differences of this type continued to decrease sharply (from 221 to 16). Another type of difference was stays contained in both the discharge and treating specialty files but with costs that differed by more than each of $100, $1000 and $5000. Between FY03 and FY04, differences of this type decreased sharply (e.g., from 2,367 in FY03 to 539 in FY04 for stays in both files and for costs that differed by more than $100).

7.4. Comparison between the MCA Outpatient Extract and the NPCD Database

The MCA outpatient (OUT) extract contains services recorded in the NPCD. The MCA outpatient file is also designed to include many outpatient services that are not recorded in the NPCD database. For example, MCA identified 21% more outpatient services in FY04 (such as prosthetics and addiction severity index tests) that were not recorded in the Austin NPCD. The NPCD outpatient event file (the SE file) includes all encounters to outpatient clinic stops. Linking records in the NPCD SE file with the MCA outpatient extract can generate a combined outpatient file with cost and clinical information, which is useful for VA healthcare studies.

To link records in the MCA outpatient extract with the NPCD, researchers should identify which records in the MCA outpatient extract used NPCD as their data source. Then these MCA records can be matched with records in the NPCD SE file. For more specific information about the methods to link the MCA outpatient extract with the NPCD, researchers are encouraged to consult HERC Technical Reports that are available in the publications section of the HERC Intranet web site. Researchers should also note important information in section 4.1 of this Guidebook about the new naming convention that impacts most NDE data beginning with FY04, low cost MCA data as well as the grouping of outpatient pharmacy cost data with other outpatient (OUT) records.

7.4.1. Summary of the comparison between the MCA Outpatient extract with the NPCD

FY04 MCA outpatient extract records that were marked by the NPCD flag were matched with records in the NPCD SE file. Almost all records with an NPCD flag in the MCA outpatient file had corresponding records in the NPCD SE file. For example, in FY03, 9.4% of the records in the NPCD SE file did not have corresponding MCA records with the NPCD flag equal to “Y”. In FY04, the analysis was expanded by using a file that contains outpatient encounters that were assigned low cost by MCA (“low cost” data refer to those outpatient encounters that are either not assigned costs or assigned costs between -$1 and $1). In FY04, the percentage of non-matched records decreased to 8.8% for normal cost data. However, when MCA records consisted of normal cost and low cost data, the percentage of non-matched NPCD SE records decreased sharply to 0.6%. The significance of this finding is that by including low cost MCA encounters in the FY04 comparison between the MCA and NPCD databases, outpatient utilization thought to be missing from MCA was found, and almost all outpatient care was found to be reported in MCA.

7.5. Summary of the Comparison between the MCA NDEs and the VA NPCD

The MCA national data extracts can be linked almost perfectly with the VA discharge and outpatient data sets after adjustments in database design are made. For example, with the adjustments noted in preceding sections in this chapter, inpatient stays in the FY04 MCA NDEs matched almost perfectly with corresponding records in the FY04 PTF. In addition, in a comparison of the FY04 MCA treating specialty file with the FY04 MCA discharge file, discrepancies decreased sharply for stays contained in both the discharge and treating specialty files but with costs that differed by more than each of $100, $1000 and $5000. For outpatient services, the two databases differed largely in design. More than 90% of the records in the FY04 NPCD event file were linked to the FY04 MCA for cost information whereas MCA allocated 21% of outpatient cost to services other than those recorded in NPCD. The correspondence between the MCA and NPCD databases (particularly for records but also for patients) sharply improved when MCA records consisted of normal cost and low cost data. The significance of this finding is that by including low cost MCA encounters in the FY2004 comparison between the MCA and NPCD databases, outpatient utilization thought to be missing from MCA was found, and almost all outpatient care was found to be reported in MCA.


8. Identification of Outlier Costs

When analyzing VA MCA cost data, you may encounter some records with cost that appear unusually low or high compared to the rest of the data. These cases are often labeled as outliers for being at the extreme of the distribution. The analyst would like to know if the unusually high (or low) cost of a particular MCA record reflects the actual resources used in providing the service or whether it reflects it as an artifact of the MCA system. There are several approaches to evaluating these outlier records.

Identifying outliers is hard if you are looking at the distribution of heterogeneous products. For example, inpatient discharges are inherently heterogeneous: some hospital stays are short and others last a long time. Similarly, a heart bypass is likely to require more intensive services than an appendectomy. Normalization can help make products more homogenous and facilitate the identification of outliers. We often calculate the average cost per day or the average cost per clinic stop (for outpatient care). One can also divide the costs by a relative value unit (RVU), which is a measure of resource intensity. Well-known RVUs include the Medicare DRG weight for inpatient care and the Current Procedure Terminology (CPT) weight for outpatient care. These RVUs are not part of the MCA data, and additional efforts are required to merge them to the MCA data.

Normalizing cost data can help you identify outliers. However, you might want to consider more complex methods and/or statistical models.

8.1. Outpatient Cost Outliers

Any outpatient clinic encounter with a total cost of more than $100,000 was identified as an outlier. In FY 2009, there were 132 outliers among outpatient pharmacy records, compared to 47 outliers identified in FY 2004. These account for an extremely small proportion (close to 0%) of the total number of pharmacy records. There were 174 records from the MCA outpatient file with total costs of more than $100,000. From these, 79% of the encounters involved prosthetics services, compared to 37% in FY 2004.

8.2. Points to Consider When Handling Outliers

Not all outliers are errors. Some outliers are just that—a combination of unusual circumstances that lead to unusually high or low costs. You might need to “drill down” to determine what is causing the outlier. MCA has created intermediate product datasets and these can be used to identity high or low cost products.

You might not be able to determine whether an outlier is an error or not.

MCA costs are based on local factor (input) prices. Labor costs can vary widely across the nation. You may need to adjust for wage differentials when analyzing outliers. Additional information on adjusting for wage differentials is available in a separate guidebook on the HERC guidebook web page (http://www.herc.research.va.gov/include/page.asp?id=guidebooks#STALVL); see Medicare Wage Index for VA Facilities.

8.3. Statistical Evaluation of Outliers

You can compare MCA costs to the HERC Average Cost Dataset and identify differences in absolute or relative terms (%).

Your statistical results may be robust to outliers. You can compare the findings with and without the outliers, and perhaps the findings are not affected by the exclusion. In this case, it may not be worthwhile to determine what is causing the outlier.

Some people turn to methods, such as analyzing medians, or to non parametric methods such as Wilcoxon log rank test. These are more robust to skewed distributions. However, economists generally prefer reporting the conditional mean as the high costs are often providing valuable information about rare but expensive events that represent the real use of health care resources.

If you exclude outlier records from your results, you should consider reporting that in your paper.


9. References

Hendricks AM, Lotchin TR, Hutterer J, Swanson J, Kenneally K; Decision Support System Cost Evaluation Work Group. Evaluating VA patient-level expenditures: decision support system estimates and Medicare rates. Med Care. 2003 Jun;41(6 Suppl):II111-7.

Barnett PG, Chen S, Boden WE, Chow B, Every NR, Lavori PW, Hlatky MA. Cost-effectiveness of a conservative, ischemia-guided management strategy after non-Q-wave myocardial infarction: results of a randomized trial. Circulation. 2002 Feb 12;105(6):680-4.

Comparison of DSS Encounter-Level National Data Extracts and the VA National Patient Care Database: FY2004 King SS, Phibbs CS, Yu W, Barnett PG Technical Report #23, Health Economics Resource Center (November 2007)

Comparison Between DSS National Data Extracts and HERC Average Costs: Aggregate and Person-Level Costs, FY2001 Yu W, Berger M Health Economics Resource Center #13, May 2004

 

Acknowledgements

This guidebook is based in part on earlier versions that were co-authored by Wei Yu, Cherisse Harden, and Samuel S. King. We would like to thank Elizabeth Cowgill, Jennifer Y. Scott, and Jean Yoon for their contributions. Additionally, we would like to acknowledge the Managerial Cost Accounting Office staff for providing a careful review of the document and making many useful suggestions. We gratefully acknowledge the financial support from the VA Health Services Research and Development Service (HSR&D).

Last Updated Date: April 10, 2020