Medical Expenditure Panel Survey
Medical Provider Component (MEPS-MPC)

1. Introduction

The Medical Expenditure Panel Survey (MEPS) has been conducted by the Agency for Healthcare Research and Quality (AHRQ) each year since 1996. MEPS is a set of large‑scale surveys of families and individuals, their medical providers, and their employers across the United States. MEPS collects data on specific health services, including frequency of use, costs, and sources of payment for services, and on the cost and scope of health insurance covering U.S. workers.

This report describes the methodology of the 2024 Cycle of the MEPS Medical Provider Component (MPC)¹. The MEPS MPC collects data from Hospitals, Office‑Based Doctors (OBD), Home Health Agencies, Institutions² (such as long‑term care facilities) and Pharmacies reported by MEPS Household Component (HC) respondents, as well as doctors who provide services for patients in Hospitals but bill separately from the Hospital (referred to as Separately Billing Doctors or SBDs). (See Section 2.1 for additional information about provider types.) The MEPS HC is conducted by Westat, Inc. and the MEPS MPC is conducted by RTI International and DLH (formerly known as SSS).

For each cycle, providers for the MPC sample are identified in three rounds of HC data collection for two HC panels (see Table 2‑1). Overall, the HC panel design features five core rounds of interviewing over the course of two full calendar years. The HC collects data from a sample of families and individuals in selected communities across the United States, drawn from a nationally representative subsample of households that participated in the prior year’s National Health Interview Survey (conducted by the National Center for Health Statistics, Centers for Disease Control and Prevention).

During the household interviews, the HC collects detailed information for each person in the household including demographic characteristics, health conditions, health status, use of medical services, charges and source of payments, access to care, satisfaction with care, health insurance coverage, income, and employment.

The 2024 MPC cycle was conducted by RTI International and DLH under the fourth option year of the 2021‑2024 contract awarded by AHRQ to RTI in 2021. RTI completed data collection for Hospitals, Office‑Based Doctors, Home Health Agencies, and Separately Billing Doctors (SBDs), while DLH completed data collection for Pharmacies.

¹ Following convention, the 2024 MPC refers to the data collected about calendar year 2024 which are matched with data from the 2024 Household Component (HC) of MEPS. Data collection for 2024 MPC began in January 2025 and continued through January, 2026 (see Section 3.4).

² Institution data collection not conducted during the 2024 cycle.

Return to Table of Contents

1.1 Changes from 2023 MPC to 2024 MPC

Prior to data collection beginning, a number of recommended Contact Guide, Event Form, and system changes were submitted to AHRQ for review and approval. This included the routine change of updating the reference year from 2023 to 2024. A change was also made to allow points of contact (POCs) to submit records via encrypted email attachments. This update to allow records to be sent by email impacted not only receipting systems, but also the Contact Guide. Additionally, another system update involving the Contact Guide made it easier for records receiptors to decrypt electronic records received in the mail on encrypted media. These Contact Guide and system changes included:

• An update in the Contact Guide that offered POCs email as an option for sending records to the MPC. Updated and new Contact Guide items provided POCs with the email option (in addition to existing options of fax, electronic portal, and mail), along with the MPC email address and instructions for encrypting the records. The MPC receipting systems were also updated to provide a queue for records received by email.
• The addition of a field in the MPC receipting system that displays a password captured within the Contact Guide for use in decrypting files sent in the mail on encrypted electronic media (e.g., CDs). Displaying the password in the receipting system provided efficiency for receiptors. The password was pulled from a Contact Guide field (where it was collected from the POC) to display it within the receipting system, so receiptors had the password available on the same screen as receipting functions.

A change was also made to the Event Forms:

• The addition of Telehealth and Drive‑through as location of service options for global fee events at item A5_WHSPC in the Hospital Event Form. These options were added in the 2020 cycle as locations for the main event in each Event Form; this update made the options available for global fee dates of service.

Detailed information about item wording and instrument flow was provided to AHRQ in Deliverable OP4‑11, MPC 2024 Final Data Collection Instruments.

Since March 2020, many large‑scale, nationally based research projects, including the MEPS MPC, have faced consistent staffing challenges due to significant changes in the economy and the workplace. In response, the MPC team shifted data collection staff to a hybrid work environment with a blend of remote‑ and onsite‑based work beginning in the 2021 cycle and have continued with this model during the 2024 cycle.

Return to Table of Contents

2. Preparations for the 2024 MPC

This section describes the 2024 MPC provider sample and preparations for data collection, including grouping patient‑provider pairs by provider, grouping providers for the purpose of contacting facilities, and updating locating information.

Return to Table of Contents

2.1 Sample Preparations

Respondents in the HC are asked to identify all medical providers associated with healthcare services received by each member of the household for the reference period associated with the time period of the interview date. Thus, the basic sample unit in the MPC is a patient‑provider pair (referred to as a “pair”) where the patient is a member of a household participating in the HC and the provider is identified in the household survey as one associated with a medical event, that is, an office visit, a Hospital stay or visit to an outpatient or emergency department, a prescription for medicine, or other healthcare event. To facilitate the MPC contacting medical providers household members are asked to sign an authorization form (AF) indicating their agreement to allow providers to release information about the event to the MPC. This form is compliant with the Health Insurance Portability and Accountability Act (HIPAA) implemented in 2003.

Within the HC, the term “medical provider” is intended to include any type of practitioner contacted by the household for what the household considers to be healthcare—hospitals, clinics, long‑term care institutions, HMOs, medical doctors and doctors of osteopathy, dentists, home care providers, optometrists, podiatrists, chiropractors, psychologists, and other practitioners.

Eligibility for the MPC is restricted to services rendered in a hospital or by a medical doctor or doctor of osteopathy (MD or DO) or under the supervision of a MD or DO. The MPC excludes services provided by dentists, optometrists, psychologists, podiatrists, chiropractors, and other kinds of healthcare practitioners who do not provide care under the supervision of a MD or DO. Care provided by home care agencies is an exception to this criterion; the sample design includes all care provided through a home care agency. Pharmacies reported as sources of prescription medicines obtained by household respondents make up a fifth group of MPC pairs generated from the MEPS‑HC. However, the MPC excludes pharmacies that provided durable medical equipment (DME) only and no prescriptions. Finally, additional pairs identified during the MPC data collection as SBDs are identified in medical records obtained from Hospitals and Institutions.

In summary, provider types included in the MPC are:

Hospitals—Providers associated with an inpatient stay as well as hospital outpatient department or emergency room

Institutions—Long‑term care providers

Pharmacies—Pharmacies (corporate and non‑corporate) where household respondents obtained or purchased prescription medicines

Office‑Based Doctors (OBDs)—Physicians (MDs and DOs) associated with non‑Hospital care.

Home Health—Providers associated with care provided in the home of the household respondent, including either healthcare (Health Agencies) or other services excluding healthcare (Non‑Health Agencies)

Separately Billing Doctors (SBDs)—Providers added to the MPC sample during abstraction of medical and patient account records of Hospitals and Institutions. Charges and payments for their services are not included in the Hospital or Institution financial records and must be obtained by contacting the offices of the SBDs.

Sample files in the 2024 MPC

The HC contractor prepared pair data from the computer assisted personal interview (CAPI) survey instrument used in the HC. For Non‑Pharmacy pairs, the file includes pairs with eligible dates of utilization (that is, calendar year 2024). In the file for Pharmacy pairs, the events (prescriptions) are not dated. Files for all provider types include the AF signed by the household respondents. AHRQ subsampled OBDs at the HC Reporting Unit (RU) level and delivered the extracted MPC sample files to RTI. The 2024 MPC OBD subsampling rate was 58%. Table 2‑1 summarizes design features of the HC related to the MPC.

Input to the MPC sample was provided in six separate files:

1. Records in the main sample file were identified at the pair (PAIRID) level. All other files used to construct and load the sample were merged with this file. This file identified the MPC cases loaded into the Integrated Data Collection System (IDCS) Control System (CS) and tracked throughout the MPC data collection period. For the purposes of data collection in the MPC, the CS tracked at the event level, pair level, and provider level. During the matching process, the MPC data collected was linked back to the pairs from this original HC sample file.
2. The person file contained identifying information for every household member associated with a pair in the main sample file. The file can be merged with the main sample using the person ID (PERSID).
3. Provider contact information is contained in the NPI provider directory used by HC interviewers and the monthly non‑matched files delivered by Westat containing providers not found in the NPI directory. For providers identified in the NPI directory, the provider ID (PDDIRID) is the NPI ID (NPIPRVID) from the NPI directory. For providers not identified in the directory, the provider ID (PDDIRID) is the PROVID assigned by Westat in the monthly files of non‑matched providers. Both files contain provider name and contact information. For the non‑matched providers, the contact information is the provider name and address that was provided by the HC respondent. The contact information was then loaded into the control system as part of the MPC case.
4. The Pharmacy directory file can be merged with the main sample file using PHADIRID (same as PDDIRID) so that the name and contact information of the Pharmacy can be loaded as part of the Pharmacy case.
5. Beginning with the 2017 HC, a Pharmacy NPI directory was used by the HC interviewers to assign IDs to pharmacies. If a match was found, a pharmacy NPI ID was assigned to the pharmacy reported by the HC respondent. The pharmacy NPI directory was delivered with the sample files and was merged with the main sample file using the Pharmacy NPI ID (NPIPHAID).
6. Beginning with the 2018 cycle, RTI developed code for assigning pharmacy chain codes by searching for text strings in the pharmacy names.

MPC Sample Delivery from Household Component

For the 2024 MPC, Westat extracted the sample files used for inclusion in the MPC sample in four waves. Westat delivered the Pharmacy sample files directly to RTI. The Non‑Pharmacy files were first delivered to AHRQ for OBD subsampling and then forwarded to RTI for processing. The waves of sample files were delivered to RTI in January (Wave 1), April (Wave 2), June (Wave 3), and July (Wave 4). A total of 43,140 pairs were in the 2024 MPC sample files delivered to RTI: 31,794 (73.7%) in Wave 1 of sample delivery; 5,246 (12.2%) in Wave 2; 2,405 (5.6%) in Wave 3; and 3,695 (8.6%) in Wave 4.

Beginning with the 2020 cycle, pairs with Veterans Administration (VA) providers were held back from the MPC sample releases. The data for these pairs were extracted from agency databases by the Health Economics Resource Center (HERC). The 2024 MPC sample contained a total of 635 VA pairs across provider types. Because of this and other sample processing steps, the number of pairs in the sample files delivered to RTI in the paragraph above is not equal to the number of pairs fielded.

The following data elements were included in the MPC sample in order to identify each pair:

• Unique person and Provider IDs used to link the data collected through the MPC back to the household‑generated data for the matching process
• Identifying information of the household member, such as name, address, gender, and date of birth, parent name if person under age 18, spouse name (if married), and policy holder name for insured persons
• Identifying information about each provider, such as name, address, and telephone number
• At the pair level, the number of each type of event identified for the person for that provider and any other HC variables necessary to assign priority flags (see section 2.2 below).

These data elements are necessary to define a pair, a key data collection unit of the MPC. The extracted file records were sorted so that all pairs for a provider were listed together, thereby creating provider‑level records. (For more information about the data elements included in the extraction files, see the deliverable OP4‑7 – Consolidated Non‑SBD Sample Preparation and Implementation Report.)

Return to Table of Contents

2.2 Sample Maintenance

In order to facilitate data collection, RTI sorted providers into contact groups, that is, groups where several providers share the same contact information (e.g., telephone number, practice name, street number, and provider name). Potential groups were carefully reviewed to confirm that grouping was appropriate. In the formation of contact groups, provider identification numbers and other detailed information from the HC were preserved to assure accurate linkages back to the initial sample files. During the MPC data collection, the IDCS enabled contact groups to change as facilities could be restructured, bought out by other entities, or change location of the medical and/or patient account records.

Contact Groups

All pairs were assigned to contact groups. A pair was assigned to a contact group first by checking whether the provider in the 2024 MPC sample was in a previous cycle’s MPC sample. If so, the pair was assigned to the provider’s most recent contact group. Providers not found in a previous MPC sample were grouped to form a new contact group based on the provider’s contact information. An automated process grouped pairs by telephone number, address fields, and a SOUNDEX program in SAS to identify similar practice or provider names.

As in prior cycles, before delivery of sampled pairs, Westat checked for duplicate pairs based on unique identification numbers assigned to each person (PERSID) and provider (PROVID). The sample preparation process at RTI included further checking for duplicate pairs by searching the sample files for pairs that had the same PERSID and NPI identifier but a different PROVID. When duplicate pairs were identified, one pair was assigned a code that indicated the pair had been merged. This merged code was used to prevent the pair from being fielded. The other pair was fielded for data collection.

An additional check searched pairs within the same RU for instances where pairs had the same provider telephone number (reasoning that in these situations, providers with the same telephone numbers might be the same individual). Suspected duplicate providers were confirmed through manual review of provider names and addresses and, if associated with the same person, merged as above.

HMO providers were grouped together and assigned to a small team to coordinate contacts with common corporate offices rather than with the individual providers. This grouping facilitated efficient contacts for recruiting HMO providers into the study and helped to make records abstraction more consistent and efficient.

Provider Type Classification

Provider type classification in the MPC is critically important operationally for several reasons. Because Hospital events are likely to be associated with high expenditures, it is important to track provider type participation to ensure that Hospital providers are responsive to the survey. Hospitals are often complex environments, especially for data collection projects, and thus the MPC data collection instruments are designed to assist the data collection staff in dealing with multiple points of contact within the Hospital and with potentially more complicated medical records and patient account information. The MPC Hospital data collection forms are also designed to facilitate the collection of SBD information associated with Hospital events.

Provider type was assigned at both the pair level and the provider level. The initial provider type for the pair was assigned during the HC interview when the household respondent identified the type of medical events associated with a medical provider. During sample processing, the household provider type is updated. First, labs and dialysis centers, imaging centers, and surgery centers are assigned a Hospital provider type. Second, providers are assigned a Hospital provider type if they were in a Hospital contact group in the previous wave.

Note that the provider type assigned during the HC could have been incorrect because of a household respondent’s misunderstanding about a provider’s status. Typically, this occurred when a household respondent confused Hospital and Office‑Based Doctors. Efforts were made to correct the classification during sample preparation and during the field period.

Following the sorting of provider pairs into contact groups, RTI reviewed the composition of contact groups to see if provider classification at the pair level was consistent within contact group. Inconsistencies, such as an OBD pair in a Hospital contact group, were resolved by creating a new contact group, so that all providers within a contact group were consistent.

In addition, during data collection, staff periodically learned that the provider type was incorrect and the field was updated so that the appropriate event form could be administered. The most common change was to a Hospital provider from another provider type, typically an OBD provider. This provider type change was important so that the appropriate Hospital Event Form could be used to collect SBD information. Updating provider type was uncommon among other provider types.

As a result of such provider type changes during sample preparation and during data collection, in the 2024 MPC, the count of Hospital pairs increased by 1,106 pairs, an increase of 11% between the count of Hospital pairs in the HC sample and the count at the close of the field period. Among changes to Hospital provider type, 28% occurred during sample preparation and 72% during data collection. The overall count of Home Health pairs increased by 46, an increase of approximately 6%. The overall count of OBD pairs decreased by 1,152 (6.5%).

Priority Code

Patients were flagged as high priority if they had one or more in‑patient hospital stay for any length of time. Cases (contact groups) with patients having a high priority flag are contacted and worked earlier by the data collection staff when working MPC cases. Once the priority flag is set at the person‑level, it is rolled up to the provider and contact group levels. That is, any contact group that includes at least one high‑priority person will be a high‑priority contact group, regardless of the characteristics of other persons associated with that contact group. These priority cases are closely tracked and monitored during MPC data collection using production reports that track the progress of completing these priority cases.

Fielding the 2024 MPC Sample

The 2024 MPC sample (consisting of Hospital, OBD, Pharmacy, and Home Health pairs identified in the HC) was fielded in four waves following the receipt of each wave from Westat and AHRQ. Given the HC data collection procedures, it is possible for a pair to be included in more than one wave of the MPC sample. Before fielding each subsequent wave, the sample was reviewed to identify pairs that had been included in an earlier wave. When a pair in the new wave matched a pair from an earlier wave and the same event types were reported in both (or all four) waves, the pair was not fielded in the later wave. If different event types are reported, the case is reviewed to determine whether additional data collection is necessary. (Fielding the SBD sample is discussed in Section 3.1 below.)

Return to Table of Contents

2.3 Integrated Data Collection System

The MPC IDCS supported the 2024 MPC data collection and tracking requirements. Its main purposes were to:

• Manage and update the provider contact information
• Collect updated information into one central database
• Produce reports for project staff as well as AHRQ, updating data collection progress at the event, pair, and provider level
• Provide a secure model to contain information with RTI’s Enhanced Security Network
• Produce data files for the matching process.

The IDCS is a Windows .Net MVVM based system that facilitated obtaining Points of Contact, call scheduling, contact information, appointment times, and event/status information. This system was tightly integrated with Blaise-based MEPS MPC Event Forms for data capture over the phone or by record abstraction.

The components of the IDCS are described in the following paragraphs.

Case Management System (CMS)

The CMS provided oversight and control over the MPC sample by tracking pending and final disposition for individual cases and for the aggregate sample. The CMS imported the provider sample files and arranged information about providers and patients into contact groups to facilitate provider recruiting efforts and data collection. For individual cases, the CMS tracked the completion of data collection by individual medical events, patients, providers and provider practices (contact groups), providing production supervisors and project staff a tool for measuring progress in completing the varied data collection units in the MPC. The CMS triggered the production of materials (including AFs) faxed, mailed, emailed, or sent via the web portal to providers. It notified data collection staff that these materials had been sent to providers and generated notices for follow‑up. At the aggregate level, the CMS produced daily standard or customized reports to track performance of the data collection activity. The CMS was used to monitor production of cases completed via record abstraction as well as by telephone.

Contact Guide

The Contact Guide was programmed as an aid for recruiting providers across all provider types. The Contact Guide was used to record contact information for several points of contact within a provider organization (e.g., a group practice or Hospital) and results of each contact. The Contact Guide included the capability to generate packages of materials, including copies of a patient’s signed AF that were then faxed, emailed, or mailed to providers or made available via the web portal for download. The Contact Guide interacted with the CMS to prompt follow‑up contacts with providers after an appropriate time (24 hours for faxed/emailed material or material sent via the web portal; 5 days for mailed material).

Event Forms

Event Forms were used for collecting information either during telephone calls with providers or by abstracting medical or patient account records. The Event Forms were designed to be adaptable to the particular format of medical and patient account records maintained by providers. The Event Forms featured edit checks on individual items and were also programmed to alert users to inconsistencies that may be resolved either with telephone respondents or by further investigation in the records. As each Event Form was completed, it was checked for critical items and, if missing, the Form was flagged for follow‑up.

Completion of Event Forms was tracked automatically in the CMS to record progress in completing information about medical events, patients, providers, and provider contact groups.

Assignment Transfer

The Assignment Transfer System was used to assign cases among the data collection staff. It was also used to reassign a reluctant provider to a more skilled negotiator on the data collection team or to balance and adjust workloads following staffing changes. Results of all previous call attempts or entered data were accessible to the new user.

Automated Fax/Mail/eMail/Web portal

Prior to data collection and using the contact information collected from the provider during initial contact, providers were sent (by fax, mail, email, or web portal) the following materials:

• Cover sheet
• Cover letter providing general information about the study from the U.S. Department of Health and Human Services and AHRQ
• Brochure that addresses commonly asked questions about the MEPS MPC study
• Patient List of all MEPS HC respondents who reported receiving services from the provider
• AF for each patient on the Patient List
• Return form used by the respondent when they preferred to fax, mail, email, or send via the web portal their medical and patient account records for abstraction. The fax and web portal return cover sheet contained pre‑printed information for faxing/transmitting records. The mail return form includes a pre‑printed mailing label for the provider to send via mail.

Return to Table of Contents

2.4 Enhanced Security Network

All files containing personally identifiable information (PII) or protected health information (PHI) were stored and managed within the FIPS‑Moderate Enhanced Security Network (ESN), a network developed by RTI to meet the security requirements of NIST SP 800‑53, Rev.4, Recommended Security Controls for Federal Information Systems and Organizations at the Moderate level (http://nvlpubs.nist.gov/nistpubs/SpecialPublications/NIST.SP.800-53r4.pdf). A key IDCS security feature provided access to the MEPS MPC desktop based on the login attributes assigned to individual users.

Return to Table of Contents

2.5 Recruiting and Training

Data Collection Specialists (DCSs) were the “front‑line” staff charged with contacting medical providers and abstracting medical event data from medical and payment records. Abstracting this information could be completed either over the telephone in interviews with provider staff or by abstracting records sent in by providers. Separate training modules were conducted to emphasize the different skills necessary to complete data collection in either mode. Although some DCSs developed expertise in one mode, many DCSs were cross‑trained for either telephone or records abstraction methods.

Return to Table of Contents

3. Data Collection

In the 2024 MPC, the project team continued to follow a core protocol for collecting information from providers. The protocol was customized in the Event Forms to address the unique challenges of each provider type. Project procedures were designed to make data collection as efficient as possible for the providers and DCSs.

As noted in Section 2.1, the pairs in the sample files were sorted by provider. In addition, providers who appeared to work in the same practice were sorted into contact groups to minimize the number of contact attempts with individual providers.

As part of the initial communication with each contact group, the DCS identified appropriate POCs to facilitate data collection completion. The Contact Guide was designed to enable DCSs to record the outcome of each contact attempt and to give supervisors and project staff the ability to review the provider group contact history prior to subsequent contact attempts. DCSs were assigned a set of provider contact groups to establish rapport with contacts in each group. If any cooperation or staffing issues arose, cases were reassigned to refusal converters or another DCS. During initial contacts, DCSs performed several tasks:

• Introduce the study
• Confirm the provider groupings in the initial assignment
• Identify the provider staff to fulfill the requests
• Obtain fax numbers, addresses, or emails for sending project materials
• Negotiate the manner in which data collection would proceed
• Determine whether the facility charged a fee for providing records.

Depending on the size and complexity of the provider practice, these tasks may have been completed in a single call or over several calls with different POCs.

Return to Table of Contents

3.1 Provider Recruitment and Data Collection Procedures

While the MPC includes data collection procedures common to all provider types, each provider type also has unique features and specific procedures DCSs are required to follow. The sections below describe the MPC data collection protocols and procedures for each provider type.

Hospitals

Data collection procedures were designed to adapt to particular situations in provider facilities while maintaining consistency in the data collected. DCSs typically contacted three Hospital departments: medical records, patient accounts, and the administrative office. After the Hospital received a provider information packet, the DCS re‑contacted the medical records department. Because of the length and complexity of Hospital records and because Hospital providers were often associated with multiple pairs, standard protocol requires sending records. This mode was also a preference so that records were available for quality assurance purposes. In a small percentage of cases (about 11.7% of medical records and 11.3% of patient accounts, see Table 3‑1) data were collected by telephone.

Four key pieces of information were obtained from the Hospital medical records:

• Date(s) of service
• Event type (ER, outpatient, inpatient)
• Diagnoses (ICD‑10 codes), and
• Names and specialties of any health professionals who saw the patient during the Hospital event and who charged for services separately from the Hospital’s billing record (SBDs).

Concurrent with the request for this information, the DCS also contacted the patient accounts department to collect the services provided, charges, and sources and amounts of payment for each event identified. Finally, after records abstraction was completed, a DCS contacted the Hospital’s administrative offices (AO) to obtain the billing status of each health professional identified in the medical records and contact information for confirmed SBDs.

Institutions³

³ Institution data collection not conducted during the 2024 cycle.

The procedures for Institutional care settings were similar to Hospital. The Institution sample consisted of long‑term healthcare facilities, such as skilled nursing or rehabilitation facilities. Non‑profit organizations are excluded.

Office‑Based Doctors (OBDs)

Compared with Hospital providers, the information required from OBD practices was often less complicated. In addition, OBDs were typically associated with fewer pairs per contact group than Hospital providers. For both reasons, OBD data collection was more amenable to telephone data collection and DCSs encouraged OBD providers to give information during the telephone contact when they had few patient records or only a few events to report. The Contact Guide was designed to factor in OBDs who use off‑site billing services. DCSs were trained to collect information from off‑site billing services during their contacts.

Home Health Providers

Data collection for Home Health providers followed the same basic protocol as the OBD sample. In certain cases, the DCSs contacted social service agencies or corporate offices to locate the necessary records. The Home Health Event Form was programmed to conform to Medicare Home Health Prospective Payment System. The system allowed the option of collecting payment data in 2‑month or 1‑month time frames as appropriate.

Pharmacy

For small retail Pharmacies unassociated with a chain, and for Pharmacies associated with small chains, the DCS contacted the Pharmacy to explain the study purpose and determine if patient profiles were available. If they were, the DCS verified the profile contained required data elements. If patient profiles were not available or if the profiles did not contain all required data, the DCS collected the information by telephone or requested supplemental reports from the pharmacist. Pharmacy data were received in any format including hardcopy patient profiles, electronic files with patient profile data, and/or collecting or supplementing the profiles by telephone data collection.

For large retail Pharmacy chains, individual pharmacies were grouped by chain using a unique code. Historical contact information from previous data collection years was reviewed for each chain to develop a contact approach. Specially‑trained negotiators followed up in one of two ways:

• If the corporate office preferred to collect data from the local stores, the data collection followed the small retail model. However, an endorsement from the corporate office was requested to be included with each contact packet.
• If the corporate entity preferred the data request be handled with a regional or central contact, the negotiator facilitated the most efficient method for data collection.

Separately Billing Doctors (SBDs)

Hospital, Institution, OBD, Home Health, and Pharmacy providers were all identified by household respondents during the HC. The balance of the MPC sample consisted of physicians (reported by Hospitals and Institutions) who provided services during a Hospital‑ or Institution‑based event. These events often resulted in charges from providers who may or may not have direct patient contact (e.g., pathologists or radiologists) and whose fees may or may not have been included in the Hospital charge. These charges are a key part of Hospital event costs, and this information can only be obtained from the MPC.

For all doctor names abstracted from the medical record, DCSs contacted the Hospital medical records or professional staffing department to confirm the SBD status. Either working with medical records personnel by telephone or from records, the DCS recorded each provider who provided any services and whose charge might not have been included in the Hospital charge. The DCS then contacted the Hospital’s administrative office to verify the SBD billed separately. If there was any possibility of a separate charge, the DCS obtained complete contact information and created a link within the IDCS to connect the Hospital provider, patient, event type, event date, and SBD. This link is referred to as a node, that is, a unique combination of Hospital, patient, event type, event date, and SBD provider.

Similar to prior MPC cycles, fielded SBD nodes were based on a priority status where higher priorities were expected to yield nodes more likely to be eligible and to be associated with higher charges. Physician’s role, specialty, and location of service were used to define SBD fielding priority. In 2018 and earlier, three priority levels (High, Medium, and Low) were used. The priority categories were revised in 2019 and four levels (High, Medium, Low, and Extra Low) were assigned to the 2024 SBD nodes, as follows.

High priority was assigned when the physician’s role was Active Physician/Providing Direct Care, Don’t know, blank or missing, and the physician specialty and Hospital location of service was one of the following combinations:

• Anesthesiology or Surgery, with any location of service;
• General/Family Practice, Internal Medicine, Psychiatry, or missing specialty with Hospital Inpatient or Institution location of service; or
• OB/GYN or Pediatrics with Hospital Inpatient location of service

Medium priority was assigned when the physician’s role was Active Physician/Providing Direct Care, Don’t know, blank or missing and the physician specialty, location of service, and Hospital event CPT codes were one of the following combinations:

• OB/GYN or Pathology (excluding pathology with pathology CPT codes for the event only in the range 80000‑84999), with Outpatient location of service;
• Other specialty with Emergency Room (ER), Hospital Inpatient, or Institution location of service;
• Pathology with Hospital Inpatient or Institution location of service,
• Pediatrics, Psychiatry, or missing specialty with ER location of service; or
• Radiology with any location of service;

Low priority was assigned when the physician’s role was Active Physician/Providing Direct Care, Don’t know, blank or missing and the physician specialty, location of service, and Hospital event CPT codes were one of the following combinations:

• General/Family Practice or Internal Medicine, with ER or Outpatient location of service;
• Pediatrics, Psychiatry, Other specialty, missing specialty, or Pathology (with pathology CPT codes for the Hospital event in the range 80000‑84999 only), with Outpatient location of service; or
• OB/GYN or Pathology with ER location of service

Low priority was also assigned for all other roles where the physician specialty was Surgery, Radiology, OB/GYN, or Anesthesiology.

Extra Low priority was assigned for all other roles and specialties (that is, the physician role was something other than Active Physician/Providing Direct Care, Don’t know, blank or missing and the physician specialty was Pathologist, Internal Medicine, Psychiatry, Pediatrics, General/Family Practice, or Other).

These criteria for assigning priority status were applied to the 2024 MPC. All High and Medium priority nodes and subsamples of the Low and Extra Low priority nodes were fielded. Low and Extra Low priority nodes were undersampled relative to the High and Medium priority nodes. The sample was constructed such that all nodes in a pair were fielded. Release of SBD pairs emphasized High priority nodes so that SBD providers and billing services would have ample time to respond. Four waves were used in the 2024 SBD cycle.

Prior to SBD sample release and data collection, a computer algorithm identified instances of overlapping OBD and SBD providers. The OBD and SBD provider identification numbers were required to be the same to be considered a match by the computer algorithm. Nodes identified as directly overlapping an OBD were excluded from subsequent subsampling and from SBD data collection. Four situations were considered.

1. Direct node match—As in recent previous cycles, nodes were filled using the overlap pair with an S‑code event (that is, an inpatient, ER, or outpatient event) on the same date at the node. The following situations were also used to automatically link OBD and SBD nodes:
   1. Events where the OBD location of service is a physician’s office and the SBD location is outpatient, dates of service are the same, and charges and payments are not the same;
   2. Events where the SBD location is inpatient and the OBD date of service is within the range of the inpatient stay (excluding first and last day); and
   3. Events where the SBD location of service is either outpatient or inpatient, the CPT4 codes for the OBD are associated with Hospital events and are not used in ambulatory settings, and the date of service is either the same for an outpatient event or within the date range of the inpatient event, including the first and last day of the stay.
      
      In the 2024 cycle, 43 nodes were identified as a direct node match.
      
      
2. Systematic coding of obvious disavowal nodes—For a large proportion of the nodes associated with an OBD pair with various types of specialty services with a date close to or the same as an OBD event, often the role of the SBD is Referring or copied doc. Some examples of this situation are an office visit with an OB/GYN followed closely by a mammogram; an office visit with an internist preceded by a blood panel; and an office visit with an orthopedist followed closely by an x‑ray.
   
   The specifications used to identify the disavowal nodes were as follows:
   • If the OBD overlap pair does not have an S‑code event within 2 weeks plus or minus of the SBD node, and
   • The node is either radiology or pathology (as defined by CPT4 codes that begin with a “7” or “8” or any BETOS code in categories 3‑Imaging or 4‑Tests), and
   • There is a regular OBD event (defined by CPT4 code that begins with a 99 or a BETOS code of M1A or M1B) within 2 weeks plus or minus of the SBD node (i.e., within 14 days before or 14 days after).

   The node was automatically coded as a referring/copied doc when all three of these conditions were met.

   If all OBD events have location of service as physician office, all OBD events have CPT 4 codes that are part of the evaluation/management series, and the SBD role is anything other than department head/follow‑up, the SBD was coded as a referring/copied doc.

   If all OBD events have location of service as physician office, all OBD events have CPT 4 codes that are part of the evaluation/management series, and the SBD role is specified as department head/follow‑up‑doc, then the SBD node was coded as department head/follow‑up doc. In the 2024 cycle, 237 nodes were coded as disavowals.

3. If the overlap pair was a refusal during OBD data collection, the SBD node was automatically coded as a refusal. In the 2024 cycle, 13 nodes were identified as refusals based on a match to a refusing OBD.

4. Nodes were also reviewed to determine if any were abstracted in error. The logic for identifying these was when the OBD location of service is physician’s office, the SBD location is outpatient, the dates of services are the same, and the charges and payments are identical. In the 2024 cycle, 0 nodes were identified as abstracted in error.

Remaining nodes where the SBDs and OBDs were associated with different provider IDs but possibly overlapped were reviewed by senior project staff to determine whether to field the node or not and, if not fielded, the code to describe the node’s status. In the 2024 cycle, 842 nodes that were subsampled to be fielded were reviewed and, of these, 398 (47.3%) were not fielded and resolved as follows:

• Included in an OBD, that is, a direct match that was not identified in the automated process (109 nodes)
• Disavowal (271 nodes)
  • Type 2 Disavowal (30 nodes)
  • Referred or copied physician (203 nodes)
  • Department head or follow‑up (38 nodes)
• Abstracted in error (4 nodes)
• Included in another SBD (5 nodes)
• Included in Hospital bill (6 nodes)
• Node is part of a global fee where charges were captured on another date, that is, node is a leaf. (3 nodes)

These procedures for identifying SBD‑OBD overlap in the manual review were similar to those used for the automated review, except the manual review looked across the entire SBD contact group (instead of being restricted to OBD and SBD providers with the same provider identification number). In addition to these rules, the SBD was coded as abstracted in error if the SBD should not have been recorded during the Hospital stay because the specialty (such as “nurse”) was included in the Hospital event charges.

As a step in the preparation of the SBD sample, attempts are made to match all SBD providers to a National Provider Identifier (NPI) to assign an identification number. In many instances, the provider’s NPI was included in the records and was abstracted into the Event Form. If the NPI was not in the record, DCSs looked up the number in the NPI Registry. SBD providers that could not be associated with an NPI were assigned a unique identifier in the same format as the NPI. The NPI Registry includes both individual and organizational providers.

Return to Table of Contents

3.2 Data Abstraction

Once the provider acknowledged receipt of the AFs, the DCS either collected information over the telephone through electronic Event Forms specific to each provider type or planned to receive medical records and patient account information.

Since the 2020 cycle, all MPC abstraction work has been performed using PDF files of the records. Tools that accommodate the abstraction of the PDF records include allowing abstractors to highlight the PDF files and an eANF used for entering abstraction notes. Records that arrived via hardcopy were scanned and converted to PDF format to allow for abstraction. Abstractors were able to access the PDF records and highlight and save the abstracted version for future review. Once all data elements were successfully highlighted, the abstractor could proceed with keying the data elements into the eANF. The data abstracted into the eANF were automatically loaded into the Blaise Event Forms for manual review and verification by the abstractor. These same processes and procedures were used for the 2024 cycle.

Table 3.1 displays the proportion of participating Hospital, OBD, and SBD contact groups⁴ that elected to participate by sending in medical records and patient account information for abstraction. Reflecting the preference for collecting Hospital records for abstraction, in the 2024 cycle most Hospital contact groups, 88.3%, provided medical records for abstraction and 88.7% provided patient account records. In both OBD and SBD contact groups, protocols concerning collecting data by telephone were more flexible than in Hospitals. Close to half (45.3%) of OBD contact groups provided records and 27.3% of SBD contact groups provided records.

The distribution for the 2024 cycle reflects emphasis on Hospital records abstraction, and on telephone data collection for OBDs. Because Hospital records tend to be lengthy and because of the number of patients involved in the record requests, Hospitals generally prefer to participate in the MPC by sending records rather than providing data over the telephone. This is also beneficial from a data quality perspective because the Hospital protocol can result in a great deal of information and availability of records for review is helpful to assure comprehensive and accurate abstraction. In the 2024 cycle, the data collection team had to exhibit more flexibility with regard to completing eligible Hospital pairs by phone, given the constraints some POCs faced with availability of data and access to systems when working remotely.

Information obtained from OBD and SBD contact groups is more straightforward and more amenable to telephone data collection which can be less burdensome to providers as well as a more efficient mode for uncomplicated billing situations.

⁴ Note that these counts and percentages are based on participation at the contact group level, not individual providers. As noted in section 2, contact groups may consist of multiple providers as, for example, a group practice that employs a number of physicians or a healthcare system that may contain several Hospitals. Note as well that contact group is a different metric than the concept of “provider wave” reported in the MPC prior to 2009. In a provider wave, a provider is counted once for each wave of the sample in which it is represented. Table 3.1 reports the percentage of contact groups that provided medical and patient account records.

Return to Table of Contents

Table 3‑1. Percent of Participating Contact Groups Providing Records, 2021‑2024

Return to Table of Contents

3.3 Coding Text Fields Collected in the 2024 MPC

Standard coding systems supported the coding of free text for the following types of data:

• Medical Conditions—verbatim text coded to the International Classification of Disease (ICD‑10); additional classifications of these codes employed a collapsed version of HCUP’s Clinical Classifications Software Refined (CCSRMATCH) during final file preparations
• Medical Procedures and Supplies—verbatim text coded to Berenson‑Eggers Type of Service (BETOS) codes
• Non‑Pharmacy Sources of Payment—coded to AHRQ‑supplied classification (SOP)
• Pharmacy Sources of Payment—coded to AHRQ‑supplied classification (RxSOP)
• Prescribed Medicines—verbatim text coded to the General Product Identifier (GPI‑9)
• Separately Billing Doctors—verbatim text recording name, practice, and location information was used to assign an identifier from the National Provider Identifier Registry (NPI)
• SBD Specialty—Specialties of SBD were coded to a specialty classification
• Location of Service—coded.

Sources of payment (SOP) and SBD information were coded by RTI staff using coding schemes developed and used in previous MPC cycles; sources of payment data (RxSOP) for Pharmacy were coded by DLH (formerly SSS) staff. RTI also completed location of service and CCSR and CCSRMATCH coding as part of file preparations prior to matching. Coding of text descriptions for conditions (ICD‑10), and procedures and supplies (BETOS) was completed by Health Care Resolution Service (HCRS), a firm in Laurel, MD, with extensive medical coding experience. DLH was responsible for coding prescribed drugs. More detailed discussions may be found in Deliverable OP4‑24 2024 Cycle Data Collection Coding and Work Processes Plan and OP4‑26 2024 Cycle MPC to HC Events Matching Plan.

Return to Table of Contents

3.4 Data Collection Schedule

Table 3‑2 summarizes the 2021‑2024 MPC data collection schedules. The MPC sample is provided from the HC in three waves and fielded as such, although in the 2024 cycle there were four waves for Hospital and OBD. Since the 2013 MPC cycle, the SBD sample, developed during MPC data collection, has been fielded in four waves except for the 2021 cycle which was fielded in five waves.

Table 3‑2. MPC Data Collection Schedule 2021‑2024

Following data collection, additional editing of the files preceded file preparation and matching tasks. These steps have been implemented to assure data quality and consistency in the data across survey years.

Return to Table of Contents

3.5 Data Collection Results

Completion Rates

The MPC applies the following criteria to assess or determine whether an event is complete or partially complete (see Appendix C for a full discussion of critical items). The final event level codes determine the final pair disposition.

Criteria for Non‑Pharmacy Providers. For a pair to be considered partially complete, at least one event in that pair had to have a valid response for all critical items, that is, no critical item in that event could contain a don’t know, refusal, or missing response entry. If one critical item in the event had a don’t know, refusal, or missing entry, the event was assigned a new disposition code “final critical item missing.” If all the events in a pair had this new disposition, the pair was considered a partial complete and became eligible for matching. As pairs roll up to the provider level, some providers that would have a final disposition of non‑response under the former criteria would have a final disposition of partial complete under the revised criteria.

Criteria for Pharmacy Providers. As with other providers, for a pair to be considered partially completed, it must have included an event where critical items contained valid data. Three additional categories took account of response to three data elements: Patient Amount, Third Party Payment Source, and Third Party Payment Amount.

• If Patient Amount was missing but at least one of the other two variables was complete, the event was assigned to Partial Category A.
• If Patient Amount was complete, but either of the other two variables was missing, the event was assigned to Partial Category B.
• If both Patient Amount and Third Party Payment Source were complete but Third Party Payment Amount was missing, the event was assigned to Partial Category C.

The 2024 MPC cycle target pair‑level completion rates were the same as the 2023 goals, with pair target completion rates of 88% for Hospital, 80% for OBD, Home Health, and Institution, and 85% for Pharmacy providers. The target SBD completion rate goal was 60% of fielded SBD nodes, which was estimated at baseline to be 12,000 completed nodes. Table 3‑3 displays the provider‑level results and Table 3‑4 shows the pair‑level results for the 2021 through 2024 MPC cycles. The pair‑level completion rates increased for OBD and HMO provider types, but dropped for Hospital, Home Care, SBD, and Pharmacy. Institution was not fielded for the 2024 cycle.

Deliverable OP4‑17 MPC Evaluation of 2024 Cycle Non‑SBD and SBD Data Collection Plan addresses key factors that likely contributed to the actual 2024 cycle completion rates.

Return to Table of Contents

Return to Table of Contents

Return to Table of Contents

SBD Node‑Level Results

Table 3‑5 presents SBD node‑level results. A total of 22,525 nodes were released for data collection in the 2024 cycle. Of these, 32.6% were confirmed as ineligible nodes (that is, no charges were recorded for that provider). Of the remaining 15,236 nodes (67.4% of the total), additional information was obtained for 7,202 nodes for a completion rate of 47.3%. Among eligible High priority nodes, the completion rate was 49.6% (n =1,771); among Medium priority nodes, the completion rate was 49.9% (n =5,047); among Low priority nodes, 31.9% (n=327); and among the Extra Low priority nodes, 13.2% (n=57).

Veterans Administration Pairs

Beginning with the 2020 cycle, RTI generated Excel files containing pairs with Veteran Administration (VA) providers for each wave and transmitted those files to AHRQ. Those files were then used by the Health Economics Research Center (HERC) to extract the MPC data for the pairs from the VA files. VA providers that are either state administered or CHAMPVA are excluded from the files.

For the 2024 cycle, there were 331 unique Non‑Pharmacy VA pairs and 292 Pharmacy VA pairs. Data was extracted by HERC for 318 Non‑Pharmacy pairs and 264 Pharmacy pairs, resulting in a Non‑Pharmacy completion rate of 96.1% and 90.4% for Pharmacy.

Refusal Conversion

Table 3‑6 provides additional information about refusal conversion for the 2021‑2024 MPC cycles. The analytic unit in this table is contact group, an operational unit which may consist of several providers who share facilities for medical records and billing (e.g., a medical group practice with several physicians or a healthcare system with several hospitals). The final column in this table displays the percent of initial refusals that were converted to a participating or partially participating contact group (i.e., provided all or some of the requested information). The 2024 MPC cycle refusal conversion rates by provider type were: 44.1% for Hospital, 35.5% for OBD, 16.2% for Pharmacy, 22.4% for Home Health, and 16.1% for SBD.

Return to Table of Contents

Components of MPC Data Collection

Figures 3‑1 through 3‑4 display historical MPC data collection information at the provider level for Hospitals, OBDs, SBDs, and Pharmacies (corporate and non‑corporate). Each graph displays:

• Provider sample size (eligible providers), as a proportion of the eligible sample in 2002
• Provider ineligibility rate, expressed as the complement of the eligibility rate (1 ‑ (Eligibility Rate)) for presentation purposes,
• Final provider completion rate, and
• Final provider refusal rate.

For Hospitals, (Figure 3‑1), the sample size and the provider refusal rate increased from the previous year, the provider completion rate decreased, and the ineligibility rate was about the same.

For Office‑Based Doctors (Figure 3‑2), the sample size decreased from the previous year, the ineligibility rate and refusal rate increased, and the provider completion rate stayed about the same.

For Separately‑Billing Doctors (Figure 3‑3), the sample size, ineligibility rate, and completion rate decreased, and the refusal rate was about the same.

For Pharmacies (Figure 3‑4), the refusal rate decreased, sample size increased, and the ineligibility rate and completion rate stayed about the same.

Return to Table of Contents

Figure 3‑1: Hospital providers ‑ Response factors over time

Figure 3‑2: Office‑Based providers ‑ Response factors over time

Figure 3‑3: SBD providers ‑ Response factors over time

Figure 3‑4: Pharmacy providers ‑ Response factors over time

Return to Table of Contents

Timing

Table 3‑7 presents the hours per completed pair (or node) by provider type for the 2021‑2024 MPC cycles. These timings include telephone and record abstraction as well as recruiting efforts.

Return to Table of Contents

Appendix A: Acronyms and Definitions

Return to Table of Contents

Appendix B: MPC Data Collection Summary Tables

Table B‑1. MPC Sample Sizes, Provider Level, 1996‑2024

Return to Table of Contents

Table B‑1. MPC Sample Sizes, Provider Level, 1996‑2024 (continued)

Return to Table of Contents

Table B‑1. MPC Sample Sizes, Provider Level, 1996‑2024 (continued)

Return to Table of Contents

Table B‑2. MPC Sample Sizes, Pair Level, 1996‑2024

Return to Table of Contents

Table B‑2. MPC Sample Sizes, Pair Level, 1996‑2024 (continued)

Return to Table of Contents

Table B‑2. MPC Sample Sizes, Pair Level, 1996‑2024 (continued)

Return to Table of Contents