User Notes

Accessing MEPS Condition, Event, and Prescribed Medicine Records through IPUMS

Topics covered in this user note include:

  1. Condition Records
  2. Event Records
  3. Prescribed Medicine Records
  4. Linking Condition and Event Records
  5. Linking Condition and Event Records to Person and Round Variables
  6. Linking Prescribed Medicine and Event Records
  7. Linking Prescribed Medicine and Condition Records
  8. Additional Resources

Overview

As part of the MEPS Household Component (MEPS-HC) interview, family respondents are asked to report medical conditions, medical events, and prescription medications for each family member. AHRQ uses information collected from these questions to generate the medical condition, event, and prescribed medicine files. IPUMS MEPS offers variables from the medical condition files on the condition record (RECTYPE="X"), office-based medical provider, outpatient medical provider, ER visit, inpatient hospitalization, home health, and other medical expense event files on the event record (RECTYPE="E"), and prescribed medicine files on the prescribed medication-round (RECTYPE="M") and prescribed medication-round fill records (RECTYPE="F"). Data from these records can be linked to each other and with Person and Round level records. Condition, event, prescribed medication-round, and prescribed-medication-round fill records are currently only available in hierarchical extracts.

1. Conditions

MEPS condition data can be used to generate estimates of treated prevalence, but should not be used to make estimates of disease, prevalence of health conditions, or mortality/morbidity (AHRQ MEPS HC-137 2010 Medical Conditions pg. 8). When conducting trend analyses using condition records, users should be aware of ways in which the medical conditions data has changed over time, including updates to the definition of "current conditions" and switching from ICD-9-CM to ICD-10-CM codes. These changing definitions and codes are described below.

Back to Top

1.1. Current Conditions

Conditions reported by the family respondent are only included in the medical conditions data if the condition is considered a current condition. The criteria for what is considered a current condition, and thus included in the medical conditions data, has changed as AHRQ has updated the MEPS survey design.

1.1.1. 1996-2007

For 1996-2006 and Panel 11 in 2007, medical conditions reported by the family survey respondent are included in the medical conditions data if: (1) the condition is linked to a medical event or prescribed medicine during the calendar year, (2) is identified as a priority condition, such as cancer, diabetes, heart disease, and asthma (see PRTYCOND for more information on priority conditions), or (3) is reported during the current year; a condition is considered reported during the current year if it meets one of the following criteria:

  1. Reported in Rounds 1 or 2 by the family respondent in the first year of the MEPS panel (relative year 1)
  2. Reported in Rounds 4 or 5 by the family respondent in relative year 2
  3. Reported in Round 3 and
    1. Was due to an accident or injury, or
    2. More than 50 percent of the person's reference period for Round 3 occurred in the calendar year, or
    3. For persons in relative year 2, the condition was not reported on the previous calendar year's file

1.1.2. 2008-2017

For Panel 12 in 2007 and 2008-2017, medical conditions reported by the family respondent are included in the medical conditions data if:

  1. The condition is linked to a medical event or prescribed medicine during the calendar year, or
  2. The person is currently experiencing the condition; a person is considered to be currently experiencing a condition if the condition meets one of the following criteria:
    1. The condition is selected in the Condition Enumeration section or,
    2. For priority conditions, the age of diagnosis is before the end of the calendar year or the age of diagnosis is unknown. Priority conditions are specific conditions flagged by AHRQ as particularly important to monitor due to their prevalence, expense, or policy relevance. See AHRQ's general topical page on Priority Conditions for more information.

1.1.3. 2018 forward

Beginning with Panel 21 Round 5 and Panel 22 Round 3 (the last round collected in 2017 for both panels), medical conditions reported by the family respondent, including priority conditions, are only included in the medical conditions data if the condition is linked to an event or prescribed medicine during the calendar year. This is a departure from previous approaches to identifying which reported conditions would be included in the medical conditions data. Although they have been excluded from the medical conditions data beginning in 2018, data on priority conditions that are not linked to a medical event in the current calendar year are still included on the person-level data (e.g., CHEARTDIEV, DIABETICEV).

These changes to the definition of current condition affect the number of condition records included on the MEPS public use files and will impact trend analyses.

Back to Top

1.2. Medical Condition Coding

From 1996 to 2015, medical conditions in MEPS are coded with ICD-9-CM diagnosis (ICD9CODE) and procedure codes (ICD9PROC). Beginning in 2016, medical conditions in MEPS are coded with ICD-10-CM diagnosis codes (ICD10CODE) and procedure codes were discontinued. Users should be aware that ICD-9-CM and ICD-10-CM diagnosis codes that seem comparable may hide large underlying differences. The CDC notes that ICD-9-CM codes are very different from ICD-10-CM codes; there are nearly five times as many diagnosis codes for ICD-10-CM compared to ICD-9-CM and nearly 19 times as many procedure codes. Additionally, conditions have been grouped differently and some have been renamed. At this time IPUMS MEPS does not provide a crosswalk between ICD-9-CM and ICD-10-CM codes.

Figure 1.2.1. below, from a case study by the Healthcare Cost and Utilization Project, shows the impact the change from ICD-9-CM to ICD-10-CM codes can have on analyses attempting to track the same conditions across the two coding schemes.

Figure 1.2.1. Inpatient Stays With a Diagnosis of Adverse Effects of Opioids by Age, 2015 Q1 Through 2016 Q3

Back to Top

2. Events

The event record on IPUMS MEPS is composed of seven separate files that are produced by AHRQ: (1) office-based medical provider visits, (2) outpatient visits, (3) emergency room visits, (4) inpatient hospitalization stays, (5) dental visits, (6) other medical expenses, and (7) home health. IPUMS combines these files together, allowing users to add all event types that selected variables are available in at once, or to filter their extracts to include only certain event types. Which file an event record was derived from can be determined with the variable EVENTYPE. While the variables offered differ between files, they share several commonalities. Each record represents a single medical event, which is defined as a visit for most event files, as a stay for inpatient hospitalizations, and a purchase for other medical expenses. The home health event type is slightly different from the other event types. Please see the Home Health section below for an explanation of the home health event type record.

Users should be aware that from 1996-2004, purchases of insulin and diabetic supplies were included on the other medical expenses and prescribed medicines files. Beginning in 2005, purchases of insulin and diabetic supplies were excluded from the other medical expenses file and only included on the prescribed medicines files.

2.1. Home Health

Home health event records have a slightly different meaning from the other event types where each record represents a discrete event (i.e., visit to a medical provider, purchase of medical equipment, etc.). The records for home health events represent one month of similar home health services by a provider. This means that if a person has a nurse that comes every day throughout the year to provide care, there will be 12 home health event records for that person: one record for every month in which they received those home health services. If the person has multiple providers providing home health services during the same month, each provider would be represented on a different record.

Home health services are eligible to be reported in MEPS if they are provided by one of three kinds of providers: (1) formal (paid) home health agency providers, (2) paid independent providers (self-employed), and (3) informal providers who do not reside in the same household as the MEPS sampled person. Care from informal providers who live in the same household as the sampled person are not reported in MEPS.

3. Prescribed Medicines

3.1. General Information

The MEPS prescribed medicines data provide detailed information about household-reported prescribed medicines and are well-suited for studying prescribed medicines utilization and expenditures over time. Information about prescribed medicines is obtained at each interview round and is collected during both the household and the medical provider components of the MEPS. For detailed information about how these data were collected and the editing and imputation procedures used by AHRQ, please see the prescribed medicines user note. The prescribed medicines data are available under the RX medicines drop-down menu on the IPUMS MEPS website and are referred to as medication-round fills in the IPUMS MEPS documentation. Each record in the prescribed medicines data represents a single acquisition, or fill, of a prescribed medication. While the MEPS provides summary level utilization (RXPRMEDSNO) and expenditure (RXEXPTOT, RXEXPSRC) variables for prescribed medicines on the person record (available under the "Annual" drop down menu), the prescribed medicines data provides more detailed characteristics about each medication-round fill, such as the National Drug Code (RXNDC), therapeutic class (MULTC1), FDA pregnancy category (PREGCAT), and expenditure amount by payment source (RXFEXPSRC) for each medication-round fill.

The prescribed medicines data include all survey participants for whom the family respondent reported either the purchase or receipt at least one prescribed medicine during the calendar year. Over the counter medications are not included in the prescribed medicines data. Some household respondents may have more than one medication fill, which means they will have more than one record in the prescribed medicines data. Other household respondents may not have reported any medication fills, in which case they will not have any records in the prescribed medicines data. Both original purchases and refills are included; however, users should note that original purchases and refills are not distinguished from one another in the data. It may be the case that all acquisitions in a given round are refills of an original purchase made in a prior round or before the MEPS panel began.

Each record in the prescribed medicines data includes a unique record identifier; linking keys to merge prescribed medicines data with other record types (see Linking Prescribed Medicine and Event Records and Linking Prescribed Medicine and Condition Records, for more information on how to link prescribed medicines data); total expenditure and sources of payment information; medication attributes (e.g., National Drug Code (NDC)); number of days supplied; when medication was started; and Multum Lexicon classification variables. We illustrate how the prescribed medicines data are offered by IPUMS MEPS in the next section.

Back to Top

3.2. An Illustration of the Prescribed Medicines Data

IPUMS MEPS offers data derived from the prescribed medicines file as two different types of records: one corresponding to the medication-round and another corresponding to the medication-round fill. As illustrated in Table 3.2.1, the medication-round, with a RECTYPE value of "M," corresponds to a unique prescribed medicine obtained in a given MEPS interview round, assigned a unique identifier captured in the variable (MEPSLINKIDM). When the same medication is obtained in a different interview round, it is assigned a different value of MEPSLINKIDM. The medication-round fill, with a rectype value of "F," corresponds to unique fills or refills of a particular prescribed medicine in a given round. Medication-round fill records are uniquely identified with the variable MEPSRXRECID.

Back to Top

Table 3.2.1 contains a subset of prescribed medicines data available for a MEPS participant in Panel 21, from 2017. The first row in the table contains data for medication-round MEPSLINKDM value = "2110002101047103". This medication-round record refers to a series of prescribed medicine acquisitions, or fills, for TAMSULSIN (found in the variable RXNAME), a drug used to treat prostate conditions. There were two fills of TAMSULSIN during the third MEPS interview round (indexed in purchase round (PURCHRD). Each prescribed medication fill lists the total expenditure amount incurred in the variable RXFEXPTOT. Although the same drug was acquired four times during MEPS interview round 4 (rows 5 through 8), note that the corresponding medication-round unique identifier has changed to a new value in row 4 (MEPSLINKIDM value = "2110002101068103"). This is because MEPSLINKIDM is unique to each round. For a medication identifier that is consistent across rounds, see MEPSDRUGID.

Table 3.2.1.: 2017 Prescribed Medicine Data for the Person with MEPSID = "2110002101" for records associated with the drug name "TAMSULOSIN"

Back to Top

Section 3.3. Changes in the Prescribed Medicines Data Over Time

AHRQ has made several changes to the collection and delivery of the Prescribed Medicines data over time, which we describe below.

3.3.1 Inclusion and treatment of diabetic supplies

In 1996-2017 samples (excluding Panel 21 Round 5 and Panel 22 Round 3), diabetic supplies, such as syringes and insulin, were first reported in the Other Medical Expenses (OM) section of the MEPS-HC questionnaire. If diabetic supplies were mentioned, additional information about these items was then collected in the Prescribed Medicines section of the questionnaire. Charge and payment information was asked for these events up until 2014. Beginning in Panel 21 Round 5 (2017) and Panel 22 Round 3 (2017-2018), diabetic supplies were only collected in the Prescribed medicines section questionnaire.

Back to Top

3.2.2. Changes in editing and imputation procedures

Insurance coverage and sources of payment

In 1999-2018 samples, inconsistencies between household-reported insurance coverage and pharmacy-reported sources of payment were edited to allow for more consistency in sources of payment across MEPS event files. These inconsistencies were edited by moving the positive amount from a source of payment that was inconsistent with the reported insurance coverage to one or both of the "other" source categories: Other Private (RXFEXPOPR) and Other Public (RXFEXPOPU). Beginning in 2019, the "other" source category variables were discontinued, and the inconsistencies between household-reported insurance coverage and pharmacy-reported sources of payment were no longer edited for consistency.

In 2009, if the third party payer was a public payer and the third party amount was missing, then pharmacy reports of zero out-of-pocket amounts were preserved rather than imputed.

Back to Top

Handling of outliers

In 2007, the rules used to identify outlier prices for prescriptions in the pharmacy component became less strict than in prior years. These changes allowed for greater variation in prices for brand name and generic drugs than was previously accepted. These new outlier thresholds were established based on the distribution of the ratio of retail unit prices relative to the AWUP in the 2007 MarketScan Outpatient Pharmaceutical Claims database.

In 2009, stricter outlier thresholds were adopted for fills with partial payment data and less strict thresholds were adopted for fills with complete payment data.

Beginning with the 2020 data, the process for identifying outlier prices for prescription medications was updated in response to newer price benchmarks and analyses. As a result, the 2020 data has lower prices paid for generic drugs when compared to 2019 data, and fewer generic fills have third party payments.

Back to Top

Medicare/Medicaid

In 2008, improvements to the data editing procedures led to changes in the distribution of payments by source. As a result, there was more spending on Medicare beneficiaries by private insurance, rather than Medicare, and there were less out-of-pocket payments and more Medicaid payments among Medicaid enrollees. Beginning with the 2009 data, another change affected the data for Medicare beneficiaries with both Part D and Medicaid coverage: reported Medicaid and other state and local program payments were no longer edited to be Medicare payments.

Drug Pricing

Beginning with the 2017 data, higher imputed prices were allowed to account for the rising prices of specialty drugs. As a result, specialty drugs with imputed prices for missing payment data resulted in higher total expenditures.

Back to Top

4. Condition-Event Linking

In addition to the condition and event records, IPUMS MEPS also offers the condition-event link record (RECTYPE = "J") which enables users to link condition and event records together (When you add a condition variable to your data cart, all variables from the condition-event link record are automatically added to your data cart as well). To link these records together, IPUMS MEPS staff have created MEPSCONDID on the condition record, MEPSEVNTID on the event record, and MEPSCONDIDJ and MEPSEVNTIDJ on the condition-event link record. IPUMS MEPS staff advise additionally using YEAR when linking multiple years of condition and event records because records reported in Round 3 are allowed to be present in both years of data collection for that person (e.g., a condition reported in Round 3 by a member of Panel 21 could be present in the condition data in 2016 and 2017).

The general process for linking condition and event records can be described as follows (there are other processes that can be used to link the condition and event records; this is just one example):

  1. Link the condition and condition-event link record using MEPSCONDID to MEPSCONDIDJ, and YEAR. Save the result. This will be a one-to-many link going from condition to condition-event link records. All records on the condition-event link record will have a match on the condition record, but not all condition records will have a match on the condition-event link record prior to 2018. Beginning in 2018 all condition records will have a match on the condition-event link record because conditions reported by the survey respondent are only included in the data if the condition links to an event or prescribed medicine purchase during the calendar year.

  2. Link the event and condition-event link record using MEPSEVNTID to MEPSEVNTIDJ, and YEAR. Save the result. This will be a one-to-many link going from event to condition-event link records. All records on the condition-event link record with an EVENTYPEJ value not equal to 8 (Prescribed Medicines) will have a match on the event record, but not all event records will have a match on the condition-event link record because, for example, a person can visit a medical provider for a general checkup that is not related to a specific condition.

  3. Link the result from steps 1 and 2 using MEPSCLNKID and YEAR. This will be a one-to-one link and all records (except those with EVENTYPEJ equal to 8) will have a match. The result will be a file with records that represent an event-condition pair, where one event can be associated with multiple conditions and one condition can be associated with multiple events.

To help users properly link condition and event records, IPUMS MEPS staff have provided linking syntax files. These linking files will work with multiple years of IPUMS MEPS data.

Additionally, IPUMS MEPS offers the original condition and event identifiers created by AHRQ, CONDID and EVNTID, which can be used to link to the original AHRQ files to incorporate variables that are not yet offered through IPUMS MEPS into analyses. Event and condition records can also be linked to prescribed medicine records, see section 6 for guidance on how to link prescribed medicine and event data and section 7 for guidance on how to link prescribed medicine and condition data.

Figure 4.1. Diagram of linking condition and event records

5. Linking to Person and Round Variables

Condition, event, and prescribed medication-round records can also be linked to the Person and Round records using the variables MEPSID and YEAR. Linking to the person and/or round records allows users to add demographic, employment, health insurance, and other information that can enrich analyses of the condition and event records.

Back to Top

6. Linking Prescribed Medicine and Event Records

TAKE NOTE: Users should note that beginning in 2021, AHRQ discontinued the public use version of the RXLK file, which contains all K rectype variables needed for linking prescribed medicines and event records. AHRQ determined that the data in the RXLK file are not useful for studies of prescriber behavior because so few drugs link to prescribers. IPUMS MEPS continues to offer the RXLK variables for the 1996-2020 samples, allowing users to link prescribed medicines and event record data in these years. Users interested in the RXLK variables for the 2021-forward samples can refer to AHRQ's webpage on procedures to access restricted variables in a data center.

IPUMS MEPS offers the event-medication-round link (K) record, which enables users to link medication-round (M) and medication-round fill (F) records to event (E) records. Linking event and medication-round or medication-round fill records together creates a dataset that is well-suited for studying research questions about the types of medical providers who prescribe medications, the frequency in which medications are prescribed, and the costs of prescription medications associated with particular types of medical events. When you add a medication-round fill record variable to your data cart, all variables from the medication-round record are automatically added to your data cart. Similarly, when you add an event record variable to your data cart, all variables from the event-medication-round link record are automatically added to your data cart. To link these records together, IPUMS MEPS staff have created MEPSLINKIDF on the medication-round fill record, MEPSLINKIDM on the medication-round record, MEPSLINKIDK on the event-medication-round link record, and MEPSEVNTID on the event record. IPUMS MEPS staff advise additionally using YEAR when linking multiple years together because prescribed medicine fills reported in Round 3 may be split across two years of data collection for that person (e.g., a medication reported in Round 3 by a member of Panel 21 could have fills present in the 2016 and 2017 public-use files).

Back to Top

The general process for linking medication-round and medication-round fill records to event records can be described as follows.

  1. IPUMS MEPS staff encourage users to first either 1) summarize medication-round fill records to the medication-round level or 2) reshape medication-round fill records from long to wide so that each prescribed medicine fill associated with a given medication-round record (MEPSLINKIDM) is on the same record line. This is because multiple medication-round fill records can be associated with multiple event records, resulting in a many-to-many merge when linking these records together.

  2. Link the summarized medication-round or the reshaped medication-round fill records and event-medication-round link records using MEPSLINKIDM or MEPSLINKIDF and MEPSLINKIDK and YEAR. Save the result. This will be a one-to-many link going from medication-round or medication-round fill to event-medication-round link records. All records on the event-medication-round link record will have a match on the medication-round or reshaped medication-round fill record, but not all medication-round or medication-round fill records will have a match on the event-medication-round link record.

  3. Link the event and event-medication-round link records using MEPSEVNTIDK and MEPSEVNTID and YEAR. Save the result. This will be a one-to-many link going from event to event-medication-round link records. All records on the event-medication-round link record will have a match on the event record, but not all event records will have a match on the event-medication-round link record because, for example, a person can visit a medical provider for a general checkup that does not result in a prescribed medicine fill.

  4. Link the results from steps 2 and 3 using MEPSRXLKIDK and YEAR. This will be a one-to-one link and all records will have a match. The result will be a file with records that represent a medication-round-event pair, where one event can be associated with multiple medications and one medication can be associated with multiple events.

Back to Top

IPUMS MEPS also offers the original prescribed medicines and event identifiers created by AHRQ, LINKIDX (LINKIDF/LINKIDM/LINKIDK) and EVNTIDX (EVNTID/EVNTIDK), which can be used to link to the original AHRQ files to incorporate variables that are not yet offered through IPUMS MEPS into analyses.

Back to Top

7. Linking Prescribed Medicine and Condition Records

The condition-event link (J) record that enables users to link condition and event records together also enables users to link medication-round (M) and medication-round fill (F) records to condition (X) records. When you add a medication-round fill record variable to your data cart, all variables from the medication-round record are automatically added to your data cart. Similarly, when you add a condition variable to your data cart, all variables from the condition-event link record are automatically added to your data cart. To link these records together, IPUMS MEPS staff have created MEPSCONDID on the condition record, MEPSLINKIDM on the medication-round record, MEPSLINKIDF on the medication-round fill record, and MEPSCONDIDJ and MEPSEVNTIDJ on the condition-event link record. IPUMS MEPS staff advise additionally using YEAR when linking multiple years together because prescribed medicine fills reported in Round 3 may be split across two years of data collection for that person (e.g., a medication reported in Round 3 by a member of Panel 21 could have fills present in the 2016 and 2017 public-use files).

Back to Top

The general process for linking medication-round, medication-round fill, and condition records can be described as follows.

  1. IPUMS MEPS staff encourage users to first either 1) summarize medication-round fill records to the medication-round level or 2) reshape medication-round fill records from long to wide so that each prescribed medicine fill associated with a given medication-round record (MEPSLINKIDM) is on the same record line. This is because multiple medication-round fill records can be associated with multiple condition records, resulting in a many-to-many merge when linking these records together.

  2. Link the summarized or reshaped medication-round records and condition-event link records using MEPSLINKIDM and MEPSEVNTIDJ and YEAR. Save the result. This will be a one-to-many link going from medication-round to condition-event link records. All records on the condition-event link record with EVENTYPEJ == 8 - Prescribed medicine will have at least one match on the medication-round record, but not all medication-round records will have a match on the condition-event link record. In other words, all conditions associated with a prescribed medicine will have a condition-event link record with EVENTYPEJ==8 and will be associated with at least one prescribed medicine record, but not all prescribed medicine records will be associated with a household-reported medical condition.

  3. Link the condition and condition-event link record using MEPSCONDID to MEPSCONDIDJ and YEAR. Save the result. This will be a one-to-many link going from condition to condition-event link records. All records on the condition-event link record will have a match on the condition record, but not all condition records will have a match on the condition-event link record prior to 2018. Beginning in 2018 all condition records will have a match on the condition-event link record because conditions reported by the survey respondent are only included in the data if the condition links to an event or prescribed medicine purchase during the calendar year; i.e., beginning in 2018, all condition records are treated conditions.

  4. Link the results from steps 2 and 3 using MEPSCLNKID and YEAR. This will be a one-to-one link and all records will have a match. The result will be a file with records that represent a medication-round-event pair, where one event can be associated with multiple medications and one medication can be associated with multiple events.

Back to Top

IPUMS MEPS also offers the original prescribed medicines and event identifiers created by AHRQ, LINKIDX (LINKIDF/LINKIDM/LINKIDK) and CONDID (CONDID/CONDIDJ), which can be used to link to the original AHRQ files to incorporate variables that are not yet offered through IPUMS MEPS into analyses.

Back to Top

7.1 Analytic Guidance

The prescribed medicines data may be used to study person-level research questions, such as the proportion of people in the U.S. with prescribed medicines or the health status of those with versus without prescribed medicines. When examining person-level research questions, users should be careful to accurately define their research question and estimate their unit of analysis up to the person-level, meaning each unit in their analysis represents persons rather than medication fills. The prescribed medicines file alone is not suitable for calculating estimates for the entire population because not everyone in the MEPS is represented in the prescribed medicines data. In this case, users must use the person-level records as the denominator to accurately calculate estimates for the population such as those mentioned above.

Back to Top

8. Additional Resources

  1. AHRQ Publication from 2008: Understanding and Analyzing MEPS Household Component Medical Condition Data
  2. 2018 Medical Conditions Survey Documentation (MEPS HC-207)
  3. 2012 Medical Conditions Survey Documentation (MEPS HC-154)
  4. 2018 Office-based Medical Provider Visits Survey Documentation (MEPS HC-206G)
  5. 2017 Home Health Survey Documentation (MEPS HC-197H)
  6. 2005 Other Medical Expenses Survey Documentation (MEPS HC-094C)

Back to Top