National Survey of Children’s Health (NSCH)

Estimates of children’s health care and family environments to inform reports like Healthy People 2030.

• One screener table with one row per eligible child (1+ rows per household), one topical table with the sampled child (only one row per household) from three stacked age-specific questionnaires.

• A complex sample survey designed to generalize to non-institutionalized U.S. children under 18.

• Released every four or five years since 2003, annually since 2016.

• Sponsored by the Maternal and Child Health Bureau, Health Resources and Services Administration.

---

Please skim before you begin:

1. 2021 National Survey of Children’s Health Methodology Report

2. 2021 National Survey of Children’s Health Data Users Frequently Asked Questions (FAQs)

3. A haiku regarding this microdata:

# "age but a number"
# lied babe from crib. "your nose grows"
# cried gramps changing bib

---

Function Definitions

Define a function to download, unzip, and import each comma-separated value file:

library(haven)

nsch_stata_import <-
    function( this_url ){
        
        this_tf <- tempfile()
        
        download.file( this_url , this_tf , mode = 'wb' )
        
        unzipped_files <- unzip( this_tf , exdir = tempdir() )
        
        this_stata <- grep( '\\.dta$' , unzipped_files , value = TRUE )
        
        this_tbl <- read_stata( this_stata )
        
        this_df <- data.frame( this_tbl )
        
        file.remove( c( this_tf , unzipped_files ) )
        
        names( this_df ) <- tolower( names( this_df ) )
        
        this_df
    }

---

Download, Import, Preparation

Download and import the sample adult interview and imputed income files:

nsch_screener_url <-
    "https://www2.census.gov/programs-surveys/nsch/datasets/2021/nsch_2021_screener_Stata.zip"

nsch_topical_url <-
    "https://www2.census.gov/programs-surveys/nsch/datasets/2021/nsch_2021_topical_Stata.zip" 

nsch_screener_df <- nsch_stata_import( nsch_screener_url )

nsch_df <- nsch_stata_import( nsch_topical_url )

Save Locally  

Save the object at any point:

# nsch_fn <- file.path( path.expand( "~" ) , "NSCH" , "this_file.rds" )
# saveRDS( nsch_df , file = nsch_fn , compress = FALSE )

Load the same object:

# nsch_df <- readRDS( nsch_fn )

Survey Design Definition

Construct a multiply-imputed, complex sample survey design:

Remove the fpl columns from the main data.frame:

fpl_columns <- grep( '^fpl_i[0-9]' , names( nsch_df ) , value = TRUE )

fpl_wide_df <- nsch_df[ c( 'hhid' , fpl_columns ) ]

nsch_df[ fpl_columns ] <- NULL

Reshape the fpl columns from wide to long:

fpl_long_df <- 
    reshape( 
        fpl_wide_df , 
        varying = list( fpl_columns ) , 
        direction = 'long' , 
        timevar = 'implicate' , 
        idvar = 'hhid' 
    )
    
names( fpl_long_df )[ ncol( fpl_long_df ) ] <- 'fpl'

Merge the fpl table with multiple records per child onto the main table:

nsch_long_df <- merge( nsch_df , fpl_long_df )

stopifnot( nrow( nsch_long_df ) == nrow( fpl_long_df ) )

stopifnot( nrow( nsch_long_df ) / length( fpl_columns ) == nrow( nsch_df ) )

Reshape the imputed income data.frame into a list based on the implicate number:

nsch_list <- split( nsch_long_df , nsch_long_df[ , 'implicate' ] )

Define the design:

library(survey)
library(mitools)

nsch_design <- 
    svydesign( 
        id = ~ 1 , 
        strata = ~ fipsst + stratum , 
        weights = ~ fwc , 
        data = imputationList( nsch_list ) ,
        nest = TRUE
    )

Variable Recoding

Add new columns to the data set:

nsch_design <-
    update(
        nsch_design ,
        
        one = 1 ,
        
        state_name =
            factor(
                fipsst ,
                levels = 
                    c(1L, 2L, 4L, 5L, 6L, 8L, 9L, 10L, 
                    11L, 12L, 13L, 15L, 16L, 17L, 18L, 
                    19L, 20L, 21L, 22L, 23L, 24L, 25L, 
                    26L, 27L, 28L, 29L, 30L, 31L, 32L, 
                    33L, 34L, 35L, 36L, 37L, 38L, 39L, 
                    40L, 41L, 42L, 44L, 45L, 46L, 47L, 
                    48L, 49L, 50L, 51L, 53L, 54L, 55L, 
                    56L) ,
                labels =
                    c("Alabama", "Alaska", "Arizona", "Arkansas", "California", 
                    "Colorado", "Connecticut", "Delaware", "District of Columbia", 
                    "Florida", "Georgia", "Hawaii", "Idaho", "Illinois", "Indiana", 
                    "Iowa", "Kansas", "Kentucky", "Louisiana", "Maine", "Maryland", 
                    "Massachusetts", "Michigan", "Minnesota", "Mississippi", "Missouri", 
                    "Montana", "Nebraska", "Nevada", "New Hampshire", "New Jersey", 
                    "New Mexico", "New York", "North Carolina", "North Dakota", "Ohio", 
                    "Oklahoma", "Oregon", "Pennsylvania", "Rhode Island", "South Carolina", 
                    "South Dakota", "Tennessee", "Texas", "Utah", "Vermont", "Virginia", 
                    "Washington", "West Virginia", "Wisconsin", "Wyoming")
            ) ,

        
        overall_health =
            factor( 
                c( 1 , 1 , 2 , 3 , 3 )[ k2q01 ] , 
                levels = 1:3 , 
                labels = c( 'excellent or very good' , 'good' , 'fair or poor' ) 
            ) ,
                    
        poverty_categories = 
            factor( 
                1 + findInterval( fpl , c( 100 , 200 , 400 ) ) ,
                labels = 
                    c( "below poverty" , "100-199% fpl" , "200-399% fpl" , "400%+ fpl" )
            ) ,
        
        under_six_ever_breastfed =
            as.numeric( k6q40 == 1 ) ,

        sc_sex =
            factor( ifelse( sc_sex %in% 1:2 , sc_sex , NA ) , labels = c( "male" , "female" ) )
        
    )

---

Analysis Examples with the survey library  

Unweighted Counts

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

MIcombine( with( nsch_design , svyby( ~ one , ~ one , unwtd.count ) ) )

MIcombine( with( nsch_design , svyby( ~ one , ~ state_name , unwtd.count ) ) )

Weighted Counts

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

MIcombine( with( nsch_design , svytotal( ~ one ) ) )

MIcombine( with( nsch_design ,
    svyby( ~ one , ~ state_name , svytotal )
) )

Descriptive Statistics

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

MIcombine( with( nsch_design , svymean( ~ sc_age_years ) ) )

MIcombine( with( nsch_design ,
    svyby( ~ sc_age_years , ~ state_name , svymean )
) )

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

MIcombine( with( nsch_design , svymean( ~ poverty_categories ) ) )

MIcombine( with( nsch_design ,
    svyby( ~ poverty_categories , ~ state_name , svymean )
) )

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

MIcombine( with( nsch_design , svytotal( ~ sc_age_years ) ) )

MIcombine( with( nsch_design ,
    svyby( ~ sc_age_years , ~ state_name , svytotal )
) )

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

MIcombine( with( nsch_design , svytotal( ~ poverty_categories ) ) )

MIcombine( with( nsch_design ,
    svyby( ~ poverty_categories , ~ state_name , svytotal )
) )

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

MIcombine( with( nsch_design ,
    svyquantile(
        ~ sc_age_years ,
        0.5 , se = TRUE 
) ) )

MIcombine( with( nsch_design ,
    svyby(
        ~ sc_age_years , ~ state_name , svyquantile ,
        0.5 , se = TRUE ,
        ci = TRUE 
) ) )

Estimate a ratio:

MIcombine( with( nsch_design ,
    svyratio( numerator = ~ liveusa_yr , denominator = ~ sc_age_years , na.rm = TRUE )
) )

Subsetting

Restrict the survey design to only children:

sub_nsch_design <- subset( nsch_design , agepos4 == 1 )

Calculate the mean (average) of this subset:

MIcombine( with( sub_nsch_design , svymean( ~ sc_age_years ) ) )

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 <-
    MIcombine( with( nsch_design ,
        svymean( ~ sc_age_years )
    ) )

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

grouped_result <-
    MIcombine( with( nsch_design ,
        svyby( ~ sc_age_years , ~ state_name , svymean )
    ) )

coef( grouped_result )
SE( grouped_result )
confint( grouped_result )
cv( grouped_result )

Calculate the degrees of freedom of any survey design object:

degf( nsch_design$designs[[1]] )

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

MIcombine( with( nsch_design , svyvar( ~ sc_age_years ) ) )

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

# SRS without replacement
MIcombine( with( nsch_design ,
    svymean( ~ sc_age_years , deff = TRUE )
) )

# SRS with replacement
MIcombine( with( nsch_design ,
    svymean( ~ sc_age_years , deff = "replace" )
) )

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

# MIsvyciprop( ~ under_six_ever_breastfed , nsch_design ,
#   method = "likelihood" )

Regression Models and Tests of Association

Perform a design-based t-test:

# MIsvyttest( sc_age_years ~ under_six_ever_breastfed , nsch_design )

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

# MIsvychisq( ~ under_six_ever_breastfed + poverty_categories , nsch_design )

Perform a survey-weighted generalized linear model:

glm_result <- 
    MIcombine( with( nsch_design ,
        svyglm( sc_age_years ~ under_six_ever_breastfed + poverty_categories )
    ) )
    
summary( glm_result )

---

Replication Example

As noted in the bold red footnotes of their published table, this technique is not correct and should not be used. The technical documents recommend a method matching the MIcombine syntax shown above. Nonetheless, this code matches statistics and confidence intervals within 0.5% from the Excellent or very good column of Indicator 1.1: In general, how would you describe this child’s health?:

results <-
    svyby( 
        ~ as.numeric( overall_health == 'excellent or very good' ) ,
        ~ poverty_categories , 
        nsch_design$designs[[1]] , 
        svymean , 
        na.rm = TRUE 
    )

published_proportions <- c( 0.833 , 0.859 , 0.907 , 0.955 )

published_lb <- c( 0.810 , 0.838 , 0.894 , 0.949 )

published_ub <- c( 0.854 , 0.878 , 0.919 , 0.961 )

stopifnot( all( abs( round( coef( results ) , 3 ) - published_proportions ) < 0.005 ) )

( ci_results <- confint( results ) )

stopifnot( all( abs( ci_results[ , 1 ] - published_lb ) < 0.005 ) )

stopifnot( all( abs( ci_results[ , 2 ] - published_ub ) < 0.005 ) )