Medical Expenditure Panel Survey (MEPS)

The Household Component captures person-level spending across service categories, coverage types.

• The consolidated file contains one row per individual within each sampled household, other tables contain one record per event (like prescription fills, hospitalizations), per job, per insurance policy.

• A complex sample survey designed to generalize to the U.S. civilian non-institutionalized population.

• Released annually since 1996.

• Administered by the Agency for Healthcare Research and Quality.

---

Recommended Reading

Four Example Strengths & Limitations:

✔️ Detailed information about individual medical events

✔️ Detailed sources of health expenditures

❌ Methodological changes could make a notable impact on trend analyses for conditions

❌ Household-reported medical events may be undercounted

Three Example Findings:

1. In 2021, the top 1 percent of the population ranked by their healthcare expenditures accounted for 24.0 percent of total healthcare expenditures among the U.S. civilian noninstitutionalized population.

2. Compared to those losing a job during the three prior years, nonelderly adults who lost a job during 2014 thru 2016 had a 6% net reduction in loss of health insurance coverage.

3. Americans lose ~$1,500 per year (in 2013 USD) over their life-course due to bad health realizations.

Two Methodology Documents:

MEPS HC-224 2020 Full Year Consolidated Data File

Wikipedia Entry

One Haiku:

# king dumpty's horsemen
# ahrq stitches payors, bills, claims
# fractured health system

---

Function Definitions

Define a function to download, unzip, and import each sas file:

library(haven)

meps_sas_import <-
    function( this_url ){
        
        this_tf <- tempfile()
        
        download.file( this_url , this_tf , mode = 'wb' )
        
        this_tbl <- read_sas( this_tf )

        this_df <- data.frame( this_tbl )
        
        names( this_df ) <- tolower( names( this_df ) )
        
        this_df
    }

---

Download, Import, Preparation

Download and import the 2022 consolidated file and the replicate weights file:

meps_cons_df <-
    meps_sas_import( "https://meps.ahrq.gov/mepsweb/data_files/pufs/h243/h243v9.zip" )

meps_brr_df <-
    meps_sas_import( "https://meps.ahrq.gov/mepsweb/data_files/pufs/h036brr/h36brr22v9.zip" )

Merge the consolidated file with the replicate weights:

meps_df <- merge( meps_cons_df , meps_brr_df )

stopifnot( nrow( meps_df ) == nrow( meps_cons_df ) )

meps_df[ , 'one' ] <- 1

Save Locally  

Save the object at any point:

# meps_fn <- file.path( path.expand( "~" ) , "MEPS" , "this_file.rds" )
# saveRDS( meps_df , file = meps_fn , compress = FALSE )

Load the same object:

# meps_df <- readRDS( meps_fn )

Survey Design Definition

Construct a complex sample survey design:

library(survey)

meps_design <-
    svrepdesign(
        data = meps_df ,
        weights = ~ perwt22f ,
        type = "BRR" ,
        combined.weights = FALSE ,
        repweights = "brr[1-9]+" ,
        mse = TRUE
    )

Variable Recoding

Add new columns to the data set:

meps_design <- 
    update( 
        meps_design , 
        
        one = 1 ,
        
        insured_december_31st = ifelse( ins22x %in% 1:2 , as.numeric( ins22x == 1 ) , NA )
        
    )

---

Analysis Examples with the survey library  

Unweighted Counts

Count the unweighted number of records in the survey sample, overall and by groups:

sum( weights( meps_design , "sampling" ) != 0 )

svyby( ~ one , ~ region22 , meps_design , unwtd.count )

Weighted Counts

Count the weighted size of the generalizable population, overall and by groups:

svytotal( ~ one , meps_design )

svyby( ~ one , ~ region22 , meps_design , svytotal )

Descriptive Statistics

Calculate the mean (average) of a linear variable, overall and by groups:

svymean( ~ totexp22 , meps_design )

svyby( ~ totexp22 , ~ region22 , meps_design , svymean )

Calculate the distribution of a categorical variable, overall and by groups:

svymean( ~ sex , meps_design )

svyby( ~ sex , ~ region22 , meps_design , svymean )

Calculate the sum of a linear variable, overall and by groups:

svytotal( ~ totexp22 , meps_design )

svyby( ~ totexp22 , ~ region22 , meps_design , svytotal )

Calculate the weighted sum of a categorical variable, overall and by groups:

svytotal( ~ sex , meps_design )

svyby( ~ sex , ~ region22 , meps_design , svytotal )

Calculate the median (50th percentile) of a linear variable, overall and by groups:

svyquantile( ~ totexp22 , meps_design , 0.5 )

svyby( 
    ~ totexp22 , 
    ~ region22 , 
    meps_design , 
    svyquantile , 
    0.5 ,
    ci = TRUE 
)

Estimate a ratio:

svyratio( 
    numerator = ~ totmcd22 , 
    denominator = ~ totexp22 , 
    meps_design 
)

Subsetting

Restrict the survey design to seniors:

sub_meps_design <- subset( meps_design , agelast >= 65 )

Calculate the mean (average) of this subset:

svymean( ~ totexp22 , sub_meps_design )

Measures of Uncertainty

Extract the coefficient, standard error, confidence interval, and coefficient of variation from any descriptive statistics function result, overall and by groups:

this_result <- svymean( ~ totexp22 , meps_design )

coef( this_result )
SE( this_result )
confint( this_result )
cv( this_result )

grouped_result <-
    svyby( 
        ~ totexp22 , 
        ~ region22 , 
        meps_design , 
        svymean 
    )
    
coef( grouped_result )
SE( grouped_result )
confint( grouped_result )
cv( grouped_result )

Calculate the degrees of freedom of any survey design object:

degf( meps_design )

Calculate the complex sample survey-adjusted variance of any statistic:

svyvar( ~ totexp22 , meps_design )

Include the complex sample design effect in the result for a specific statistic:

# SRS without replacement
svymean( ~ totexp22 , meps_design , deff = TRUE )

# SRS with replacement
svymean( ~ totexp22 , meps_design , deff = "replace" )

Compute confidence intervals for proportions using methods that may be more accurate near 0 and 1. See ?svyciprop for alternatives:

svyciprop( ~ insured_december_31st , meps_design ,
    method = "likelihood" , na.rm = TRUE )

Regression Models and Tests of Association

Perform a design-based t-test:

svyttest( totexp22 ~ insured_december_31st , meps_design )

Perform a chi-squared test of association for survey data:

svychisq( 
    ~ insured_december_31st + sex , 
    meps_design 
)

Perform a survey-weighted generalized linear model:

glm_result <- 
    svyglm( 
        totexp22 ~ insured_december_31st + sex , 
        meps_design 
    )

summary( glm_result )

---

Replication Example

This example matches the statistic and standard error shown under Analysis of the Total Population:

library(foreign)

xport_2002_tf <- tempfile()

xport_2002_url <- "https://meps.ahrq.gov/data_files/pufs/h70ssp.zip"

download.file( xport_2002_url , xport_2002_tf , mode = 'wb' )

unzipped_2002_xport <- unzip( xport_2002_tf , exdir = tempdir() )

meps_2002_df <- read.xport( unzipped_2002_xport )

names( meps_2002_df ) <- tolower( names( meps_2002_df ) )

meps_2002_design <-
    svydesign(
        ~ varpsu ,
        strata = ~ varstr ,
        weights = ~ perwt02f ,
        data = meps_2002_df ,
        nest = TRUE
    )
            
result <- svymean( ~ totexp02 , meps_2002_design )
stopifnot( round( coef( result ) , 2 ) == 2813.24 )
stopifnot( round( SE( result ) , 2 ) == 58.99 )

---

Analysis Examples with srvyr  

The R srvyr library calculates summary statistics from survey data, such as the mean, total or quantile using dplyr-like syntax. srvyr allows for the use of many verbs, such as summarize, group_by, and mutate, the convenience of pipe-able functions, the tidyverse style of non-standard evaluation and more consistent return types than the survey package. This vignette details the available features. As a starting point for MEPS users, this code replicates previously-presented examples:

library(srvyr)
meps_srvyr_design <- as_survey( meps_design )

Calculate the mean (average) of a linear variable, overall and by groups:

meps_srvyr_design %>%
    summarize( mean = survey_mean( totexp22 ) )

meps_srvyr_design %>%
    group_by( region22 ) %>%
    summarize( mean = survey_mean( totexp22 ) )