Understanding and Analyzing MEPS Household Component Medical Condition Data


by Steven Machlin, Anita Soni, and Zhengyi Fang

Introduction

MEPS Household Component (MEPS-HC) respondents are asked open-ended questions about their own medical conditions as well as those of other family members’ in various sections of the computer assisted personal interview (CAPI) questionnaire. A particular condition for an individual may be reported in all, some, or none of these sections. This document describes the reporting of medical conditions in MEPS and uses 2008 data to quantify the extent to which sample persons’ conditions are reported in different sections of the survey. A particular focus is on the extent to which the series of questions related to medical events account for condition reporting in the MEPS-HC.

^top

Condition Data Collection

Household respondents are asked open-ended questions about MEPS sample persons’ medical conditions in the following sections of the questionnaire: 1) Condition Enumeration (CE), 2) Medical Events (ME) and 3) Disability Days (DD). The CE section is the first among these sections and asks respondents to identify any specific physical or mental health problems for the person during the interview reference period (regardless of whether there was associated medical care or disability days). In most ME sections (medical provider office visits (MV), emergency rooms (ER), outpatient departments (OP), hospital inpatient stays (HS), prescribed medicine purchases (PM) and home health providers (HH)), household respondents are asked to identify the conditions that are associated with care/events they reported (conditions are not asked in the dental (DN) or other medical (OM) event sections)). Finally, conditions that caused sample persons to miss school or work, or spend more than half a day in bed, are ascertained in the DD section.  

A particular condition for a sample person may be reported in all, some, or none of the sections described above. Interview responses that identify conditions are recorded as verbatim text by interviewers and then professionally coded for MEPS analytic files to fully-specified ICD-9-CM codes. Specific questions for each section with condition probes are provided in the Appendix. 

In addition to open-ended reporting of conditions in the CE, ME, and DD sections, the priority condition enumeration section (PE) contains a series of “yes/no” questions on whether the person has ever been diagnosed as having each of several specific conditions that are generally chronic in nature. These conditions, which include high blood pressure, heart disease, stroke, emphysema, chronic bronchitis, high cholesterol, cancer, diabetes, joint pain, arthritis, asthma, and attention deficit disorder/attention deficit hyperactivity disorder, were deemed by the Agency for Healthcare Research and Quality (AHRQ) as “priority” for the survey due to their relatively high prevalence and generally accepted criteria for assessing appropriate clinical care. While variables based on responses to the PE questions are included in the MEPS Full Year Consolidated files, this report focuses only on the reporting of conditions in the CE, ME, and DD sections (i.e., not PE) because these sections have the potential to capture any condition and comprise the data in the MEPS Medical Conditions public use files. However, because the PE questions are asked prior to the CE, ME, and DD sections, reporting of the specific PE conditions may be greater in these subsequent sections than would have occurred in the absence of a PE section.

^top

MEPS Medical Conditions Files

Each record in a MEPS annual medical condition file represents a “current” condition reported as existing for a MEPS sample person at any time during the specific data year (i.e., those identified in the CE, ME, or DD sections of the questionnaire). To preserve confidentiality, diagnosis condition codes provided on the Medical Conditions file are collapsed from fully-specified 5-digit ICD-9-CM codes to 3-digit ICD-9-CM code categories.

The following table provides SAS code that can be used to identify conditions in a Medical Condition file according to which section of the questionnaire they were reported. This code reflects the methodology used to develop table 1 for this report (see “Methods” section below).

CAPI section SAS code
Medical Events (ME) If HHNUM>0 or IPNUM>0 or OPNUM>0 or OBNUM>0 or ERNUM>0 or RXNUM>0 then Section=ME;
Disability Days (DD) but not ME Else if MISSWORK=1 or MISSSCHL=1 or INBEDFLG=1 then Section=DD;
Condition Enumeration (CE) only Else Section=CE_only;

While it is reasonable to presume that many conditions reported in the ME and/or DD sections were also reported in the CE section, only those conditions reported exclusively in the CE can be identified based on information provided in the Medical Conditions file (see SAS code in table above).

^top

Methods

Among persons with one or more records in the 2008 Medical Conditions file, table 1 shows the distributions by MEPS questionnaire section that the condition was reported for the 50 most common condition categories while table 2 provides two different types of population prevalence estimates for the 50 conditions. Persons are classified in table 1 into 3 mutually exclusive categories: 1) those with one or more medical events reported as associated with the condition, 2) those with the condition reported in DD section but no associated medical events reported, or 3) those with the condition reported in the CE section only. Table 2 shows “total” population prevalence estimates based on all persons and “treated” population estimates based on only persons with associated medical event(s) reported. The condition categories were defined using the Clinical Classification Software (CCS) which is a tool developed by AHRQ for clustering diagnoses into a manageable number of clinically meaningful policy-relevant categories (see http://meps.ahrq.gov/mepsweb/data_stats/download_data/pufs/h120/h120doc.pdf for crosswalk between CCS condition categories and component ICD-9-CM codes). The number of sample persons range across the 50 condition categories from 177 to 6,469. All estimates were weighted by the MEPS person-level weight (PERWT08F) to produce national estimates and standard errors shown in table 2 were computed using a Taylor Series estimation approach.

^top

Results


Table 1 shows that, in general, a high proportion of persons with a condition reported in any section had the condition identified in the medical events sections. For 32 of the 50 conditions shown, 80 percent or more of persons had medical events associated with the condition.  Moreover, 95 percent or more of persons with thyroid disease (CCS=48), diabetes (CCS=49–50), epilepsy/convulsions (CCS=83) or male genital disorders (CCS=164–166) in the Medical Conditions file had medical events associated with the condition.  In contrast, only 30 percent of persons with influenza (CCS=123) and 23 percent of those with an intestinal infection (CCS=135) reported in any section had a medical event reported for those conditions. These two conditions were more likely to be reported in the DD section (55 percent of those with influenza and 69 percent of those with intestinal infections reported).  Other conditions with a notable proportion of persons identified in the DD section (but not ME) include acute bronchitis/upper respiratory infection (URI) (36 percent, CCS=125–126), and headache (33 percent, CCS=84).  Most conditions (42 of the 50) had less than 10 percent reported in DD with no medical events reported for the same condition. Finally, for all but three of the conditions, less than 20 percent of persons were identified as having the condition in the CE section only. The three conditions with more than 20 percent identified in the CE section only were trauma-related disorders (21 percent, CCS=225–236 etc.), disorders of teeth or jaws (25 percent, CCS=136), and osteoarthritis (33 percent, CCS=201–204).

Table 1. Percentage Distribution by Section in which Condition was Reported for Selected Condition Categories, 2008

CCS code(s) Condition category Number of persons unweighted) Percentage distribution of condition report by section, ME1 Percentage distribution of condition report by section, DD but not ME2 Percentage distribution of condition report by section, CE only
1–9 Infectious diseases 1,993 75.2 12.5 12.4
10, 254–258 Other care and screening 1,866 84.6 2.7 12.7
11–45 Cancer 1,321 89.0 1.2 9.9
46, 47 Non-malignant neoplasm 381 83.6 1.7 14.7
48 Thyroid disease 1,285 97.3 0.3 2.4
49, 50 Diabetes mellitus 2,263 95.4 0.8 3.8
51, 52, 54–58 Other endocrine, nutritional, & immune disorder 1,180 87.8 2.6 9.6
53 Hyperlipidemia 4,488 87.2 0.1 12.6
59 Anemia and other deficiencies 401 84.8 1.1 14.1
79–81 Hereditary, degenerative, and other nervous system disorders 248 90.5 0.5 9.0
83 Epilepsy and convulsions 242 95.2 0.8 4.1
84 Headache 1,394 51.6 33.2 15.2
86 Cataract 423 86.0 0.1 13.9
87, 89–91 Other eye disorders 1,823 78.7 2.9 18.4
88 Glaucoma 352 92.1 0.0 7.9
92 Otitis media 1,164 81.6 6.8 11.6
93–95 Other central nervous system (CNS) disorders 1,727 76.8 4.6 18.7
96, 97, 100–108 Heart conditions 2,343 86.7 2.1 11.2
98, 99 Hypertension 5,607 93.0 0.6 6.4
109–113 Cerebrovascular disease 402 79.9 6.1 14.0
114–121 Other circulatory conditions arteries, veins, and lymphatics 578 81.5 3.6 14.9
122 Pneumonia 432 78.9 12.7 8.5
123 Influenza 1,097 30.0 55.4 14.6
125, 126 Acute bronchitis and URI 6,469 44.7 36.0 19.3
127–134 COPD, asthma 6,105 74.2 8.6 17.3
135 Intestinal infection 3,957 23.3 68.8 7.9
136 Disorders of teeth and jaws 1,132 63.9 11.0 25.1
138–141 Disorders of the upper gastrointestinal (GI) 2,602 81.0 12.0 6.9
143 Hernias 240 80.7 5.7 13.6
144–148 Other stomach and intestinal disorders 194 86.8 4.8 8.5
149–152 Gallbladder, pancreatic, and liver disease 339 81.3 5.6 13.1
153–155 Other GI 1,041 77.3 8.8 13.9
156–158, 160, 161 Kidney disease 431 89.9 3.7 6.4
159 Urinary tract infections 849 83.2 3.7 13.1
162, 163 Other urinary 495 89.2 1.6 9.2
164–166 Male genital disorders 410 95.0 0.5 4.6
167 Non-malignant breast disease 195 84.4 0.7 14.9
168–176 Female genital disorders, and contraception 1,398 86.3 6.7 7.0
177–195 Complications of pregnancy and birth 192 78.1 9.3 12.6
196, 218 Normal birth/live born 865 80.0 6.2 13.8
197–200 Skin disorders 2,310 85.2 1.7 13.1
201–204 Osteoarthritis and other non-traumatic joint disorders 5,014 62.7 4.6 32.8
205 Back problems 2,076 75.8 7.2 17.0
206–209, 212 Other bone and musculoskeletal disease 789 89.3 1.4 9.4
210–211 Systemic lupus and connective tissues disorders 1,657 77.5 4.1 18.4
213–217 Congenital anomalies 177 82.1 0.9 17.0
225–236, 239, 240, 244 Trauma-related disorders 4,058 72.7 6.7 20.5
241–243 Poisoning by medical and non-medical substances 242 74.7 9.4 15.9
253 Allergic reactions 785 80.5 3.2 16.3
650–670 Mental disorders 4,289 79.0 6.1 15.0

1 Condition may also have been reported in DD or CE sections.
2 Condition may also have been reported in CE section.

^top

Table 2 compares “total population prevalence” estimates for the 50 condition categories based on the entire Medical Conditions file to “treated population prevalence” estimates based only on condition records associated with medical events.  It is apparent from the table that the rank orders of the two sets of population estimates are highly correlated (r=.98).  However, acute bronchitis/URI (CCS=125–126), intestinal infections (CCS=135), and influenza (CCS=123) rank notably higher in terms of total reported conditions than in terms of treated prevalence. The ratio of “treated prevalence” to “total reported” population estimates[1] ranges from 23 percent for intestinal infections (9,725,098/41,676,660) to 97 percent for thyroid disease (CCS=48) (15,434,555/15,869,425). In other words, the 2008 MEPS treated prevalence population estimate for intestinal infections is substantially lower than the total estimate (9.7 versus 41.7 million) while treated prevalence and total prevalence estimates are similar for thyroid disease (15.4 versus 15.9 million).

Table 2. Population Prevalence Estimates Based on Full Condition File Data versus only Persons w/ Medical Events, Selected Condition Categories, 2008

CCS code(s) Condition category Total (full conditions file), population Total (full conditions file), SE Total (full conditions file), rank Associated w/ medical
event(s),
population
Associated w/ medical
event(s),
SE
Associated w/ medical
event(s),
rank
1–9 Infectious diseases 21,989,219 788,465 14 16,526,955 625,421 14
10, 254–258 Other care and screening 20,347,107 931,518 16 17,219,615 808,617 13
11–45 Cancer 17,400,124 736,081 19 15,479,928 709,324 17
46, 47 Non-malignant neoplasm 4,708,789 327,107 37 3,936,964 304,736 36
48 Thyroid disease 15,869,425 705,056 20 15,434,555 687,100 18
49, 50 Diabetes mellitus 22,170,464 723,833 13 21,157,779 717,549 11
51, 52, 54–58 Other endocrine, nutritional, and immune disorder 13,226,874 531,229 23 11,608,845 512,025 21
53 Hyperlipidemia 50,898,959 1,459,715 5 44,401,137 1,365,597 3
59 Anemia and other deficiencies 3,619,403 234,119 41 3,069,478 219,283 41
79–81 Hereditary, degenerative and other nervous system disorders 3,072,974 220,610 43 2,780,439 212,717 43
83 Epilepsy and convulsions 2,603,482 212,087 46 2,477,566 207,632 44
84 Headache 14,129,062 568,284 22 7,297,193 367,256 28
86 Cataract 5,247,714 362,558 34 4,514,756 328,839 34
87, 89–91 Other eye disorders 20,479,939 827,815 15 16,115,728 675,354 15
88 Glaucoma 3,970,976 302,947 40 3,656,679 293,172 38
92 Otitis media 11,783,337 520,238 24 9,619,883 449,185 23
93–95 Other CNS disorders 20,196,297 707,327 17 15,503,906 634,403 16
96, 97, 100–108 Heart conditions 26,788,637 936,846 10 23,234,187 879,698 8
98, 99 Hypertension 59,334,619 1,556,908 3 55,197,052 1,490,396 1
109–113 Cerebrovascular disease 4,610,873 320,606 38 3,684,310 290,958 37
114–121 Other circulatory conditions arteries, veins, and lymphatics 6,768,348 338,658 32 5,517,244 289,264 31
122 Pneumonia 4,527,763 305,919 39 3,571,130 285,979 39
123 Influenza 10,657,800 683,241 27 3,201,693 262,339 40
125, 126 Acute bronchitis and URI 68,738,432 1,895,626 1 30,735,580 928,828 7
127–134 COPD, asthma 62,697,942 1,856,151 2 46,492,922 1,469,667 2
135 Intestinal infection 41,676,660 1,423,233 8 9,725,098 514,380 22
136 Disorders of teeth and jaws 11,477,850 540,660 25 7,331,240 383,290 27
138–141 Disorders of the upper GI 27,790,982 857,891 9 22,521,248 714,505 9
143 Hernias 2,787,080 211,583 45 2,248,293 186,716 46
144–148 Other stomach and intestinal disorders 2,597,150 252,798 47 2,253,680 231,991 45
149–152 Gallbladder, pancreatic, and liver disease 3,571,157 243,743 42 2,904,198 207,600 42
153–155 Other GI 11,147,354 535,300 26 8,612,365 468,067 25
156–158, 160, 161 Kidney disease 4,719,110 321,565 36 4,242,635 318,244 35
159 Urinary tract infections 9,454,716 435,691 29 7,869,037 386,379 26
162, 163 Other urinary 5,473,173 327,429 33 4,883,913 317,867 32
164–166 Male genital disorders 5,097,348 338,889 35 4,840,044 325,804 33
167 Non-malignant breast disease 2,164,857 197,391 48 1,827,690 174,594 48
168–176 Female genital disorders, and contraception 15,812,862 637,006 21 13,644,918 617,463 20
177–195 Complications of pregnancy and birth 1,751,045 186,002 50 1,367,201 152,547 50
196, 218 Normal birth/live born 7,557,688 439,270 31 6,045,920 356,662 30
197-200 Skin disorders 25,801,812 825,649 11 21,976,284 773,751 10
201–204 Osteoarthritis and other non-traumatic joint disorders 55,944,585 1,719,769 4 35,053,634 1,133,487 5
205 Back problems 23,606,471 802,807 12 17,895,697 671,758 12
206–209, 212 Other bone and musculoskeletal disease 9,667,717 438,216 28 8,630,869 432,237 24
210–211 Systemic lupus and connective tissues disorders 18,869,533 695,706 18 14,625,583 592,138 19
213–217 Congenital anomalies 2,017,420 174,932 49 1,656,071 155,477 49
225–236, 239, 240, 244 Trauma-related disorders 45,575,137 1,297,478 7 33,140,004 1,075,738 6
241–243 Poisoning by medical and non-medical substances 2,914,645 222,170 44 2,177,129 186,994 47
253 Allergic reactions 8,230,755 427,283 30 6,624,015 365,856 29
650–670 Mental disorders 47,879,960 1,426,609 6 37,799,317 1,225,226 4

^top

Analytic Considerations

Quality of household reported information

MEPS condition information is reported by a household respondent and consequently subject to limitations in detail and accuracy. In general, conditions that are salient, painful, require hospitalization, require ongoing treatment, have specific recognizable treatment, alter lifestyle and/or affect daily life tend to be more accurately reported by household respondents (Machlin et. al, 2009). Moreover, research has suggested that the agreement between conditions reported by household respondents and medical providers is weak for very specific conditions, but improves as condition categories are expanded to include broader ranges of conditions (Cox and Cohen, 1985). Consequently, the Clinical Classification Codes (varname=CCCODEX) generated by CCS, which aggregate somewhat similar ICD-9-CM conditions into broad but clinically meaningful categories, are provided for analysts on the Conditions file.

Prevalence estimation limitations

As described earlier, medical conditions for MEPS sample persons are identified if reported as a general problem, associated with medical care events, or as the reason for missing school or work, or for spending days in bed during the year. When analyzing MEPS condition data, it is important to consider that persons may have conditions that are not captured in any of these sections so MEPS condition data cannot be regarded as completely exhaustive. Moreover, the household respondent reporting for other household members may not be fully aware of problematic conditions for those persons (particularly those not associated with medical events) and also may not be aware of some of their own conditions. Consequently, estimates of condition prevalence in the population during the year based on MEPS should be undertaken with caution and the quality of such estimates likely varies substantially by type of condition.

To avoid some of the problems associated with estimating condition prevalence, an option for MEPS analysts is to focus on “treated prevalence” in which conditions are ascribed to persons only if they are reported as associated with medical events in the survey. This approach has the advantages of tying the analysis directly to MEPS utilization and expenditure data, which are key components of the survey, and reducing the risk that the estimates or analysis will be misconstrued as reflecting “true” prevalence. Nonetheless, “treated prevalence” estimates are also subject to underestimation to the extent that medical events and associated conditions are underreported by respondents. It should also be considered that treated prevalence estimates from MEPS are likely to better reflect true prevalence for conditions that are salient and require ongoing treatment. For example, MEPS is commonly used to produce estimates of treated prevalence and expenditures for chronic conditions. The AHRQ Healthcare Utilization Project Chronic Condition Indicator (CCI) (http://www.hcup-us.ahrq.gov/toolssoftware/chronic/chronic.jsp) tool provides a method for identifying chronic conditions in MEPS.

^top

Summary

This document uses 2008 MEPS data to describe the types, sources, and limitations of condition information available from the survey. MEPS condition information is household reported and consequently subject to limitations in detail and accuracy. Household respondents are asked open-ended questions about MEPS sample persons’ medical conditions in the following sections of the questionnaire: 1) Condition Enumeration (CE), 2) Medical Events (MV, ER, OP, HS, PM, and HH) and 3) Disability Days (DD) sections. A particular condition for an individual may be reported in all, some, or none of these sections. 

Estimates of condition prevalence in the population based on MEPS should be undertaken with caution and the quality of such estimates likely varies substantially by type of condition. Most conditions are identified in MEPS as a result of their association with medical events reported by the household respondent. To avoid some of the problems in estimating total population condition prevalence, MEPS analysts may choose to focus on “treated prevalence” by ascribing conditions to persons only when reported as associated with medical events in the survey. This approach ties the analysis directly to MEPS utilization and expenditure data which are key components of the survey. 

^top

References

Cox B., Cohen S. A Comparison of Household and Medical Provider Reports of Medical Conditions. In: Cox B., Cohen S., eds. Methodological Issues for Health Care Surveys. New York, NY: Marcel Dekker Inc.; 1985: 150–189.

Machlin S., Cohen J., Elixhauser A., et. al. Sensitivity of Household Reported Medical Conditions in the Medical Expenditure Panel Survey. Medical Care 2009; 47: 618–625.

^top

Appendix. MEPS CAPI Questions Ascertaining Medical Conditions

I.                    Condition Enumeration Section

CE04

====

Between (START DATE) and (END DATE), did (PERSON) have any physical or mental health problems, accidents, or injuries?} [Please include all of (PERSON)’s conditions, accidents, or injuries regardless of whether (PERSON) saw a medical provider, received treatment, or took medications {since (START DATE)/between (START DATE) and (END DATE)}. {Also include health problems that may have been mentioned during a previous interview, but have also bothered (PERSON) {since (START DATE)/between (START DATE) and (END DATE)}.}]

CE05

====

What did (PERSON) have?

PROBE: Did (PERSON) have any other health problems, accidents, or injuries?

II.                  Medical Events Sections

Medical Provider Visits

MV08

=====

Was this visit related to any specific health condition or were any conditions discovered during this visit?

MV09

=====

What conditions were discovered or led (PERSON) to make this {visit/telephone call}?

PROBE: Any other condition?

Emergency Room

ER03

====

Was this visit related to any specific health condition or were any conditions discovered during this visit?

ER04

====

What conditions were discovered or led (PERSON) to make this visit?

PROBE: Any other condition?

Home Health

HH05

=====

What health condition led (PERSON) to receive home health care services from {someone from} (PROVIDER) during (VISIT MONTH)?

PROBE: Any other health condition?

Hospital Stay

HS03

====

Was this hospital stay related to any specific health condition or were any conditions discovered during this hospital stay?

HS04

====

What conditions were discovered or led (PERSON) to enter the hospital?

PROBE: Any other condition?

Outpatient Department

OP08

====

Was this {visit/telephone call} related to any specific health condition or were any conditions discovered during this {visit/telephone call}?

OP09

====

What conditions were discovered or led (PERSON) to make this {visit/telephone call}?

PROBE: Any other condition?

Prescribed Medicines

PMO8

=====

Is (MEDICINE) used for a specific health problem?

PMO9

=====

What health problem is (MEDICINE) prescribed for?

PROBE: Any other health problems?

III. Disability Days Section

The next questions ask about time when (PERSON) may have missed a half day or more from work or school or spent a half day or more in bed {since (START DATE)/between (START DATE) and (END DATE)}. In answering these questions, please include any time when this occurred because of (PERSON)’s physical illness or injury, or a mental or emotional problem such as stress or depression.

DD02

=====

Let's start with work. {Including the time (PERSON) (were/was) in {the hospital} {and} {the long-term care facility}, how/How} many days did (PERSON) miss a half day or more from work {since (START DATE)/between (START DATE) and (END DATE)}? Please do not include work around the house.

PROBE: Include any time when a half day or more was missed because of a physical illness or injury, or a mental or emotional problem.

DD03

=====

What are the health problems that caused (PERSON) to miss work on those days?

PROBE: Any other health problems?

DD04

=====

Of those days, how many did (PERSON) stay in bed for a half day or more?

DD05

=====

Let's talk about school (and day care). {Including the time (PERSON) (were/was) in {the hospital} {and} {the long-term care facility}, how/How} many days did (PERSON) miss a half day or more of school (or day care) {since (START DATE)/between (START DATE) and (END DATE)}?

PROBE: Include any time when a half day or more of school (or day care) was missed because of a physical illness or injury, or a mental or emotional problem.

DD06

=====

What are the health problems that caused (PERSON) to miss school on those days?

PROBE: Any other health problems?

DD07

=====

Of those days, how many did (PERSON) stay in bed a half day or more?

DD08

=====

{Besides the days in bed you just told me about, how/How} many {additional} days did (PERSON) spend a half day or more in bed {since (START DATE)/between (START DATE) and (END DATE)} because of a physical illness or injury, or mental or emotional problem?

DD09

=====

What are the health problems that caused (PERSON) to spend a half day or more in bed on those days?

PROBE: Any other health problems?



[1] This ratio is equivalent to the percent of persons with the condition identified in the events section shown in table 1.