nsch.Rmd

# National Survey of Children's Health (NSCH) {-}

[![License: GPL v3](https://img.shields.io/badge/License-GPLv3-blue.svg)](https://www.gnu.org/licenses/gpl-3.0) <a href="https://github.com/asdfree/nsch/actions"><img src="https://github.com/asdfree/nsch/actions/workflows/r.yml/badge.svg" alt="Github Actions Badge"></a> 

Estimates of children's health care and family environments to inform reports like [Healthy People 2030](https://www.healthypeople.gov/).

* 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](http://www.mchb.hrsa.gov/).

---

## Recommended Reading {-}

Four Example Strengths & Limitations:

✔️ [Detailed information about child and family health measures](https://www.childhealthdata.org/docs/default-source/nsch-docs/2023-nsch-content-map_child-and-family-health-measures_cahmi.pdf)

✔️ [The redesigned 2016 NSCH converted to an annual survey](https://pmc.ncbi.nlm.nih.gov/articles/PMC6372340/#S13)

❌ [The redesigned 2016 NSCH broke trend with earlier surveys](https://www.childhealthdata.org/learn-about-the-nsch/survey-revisions)

❌ [Longitudinal rather than cross-sectional data may be more accurate at assessing childhood trends](https://doi.org/10.3390/pediatric16040084)

<br>

Three Example Findings:

1. [Across 2016-2022, nonsupine sleep averaged 12%, 13%, 19%, and 23% at months 4, 6, 9, and 12](http://doi.org/10.1001/jamanetworkopen.2024.50277).

2. [A minimum wage increase through childhood is associated with significant child health improvement](http://doi.org/10.3386/w26691).

3. [Publicly insured kids in 2018 were more likely to have dental sealants in the past year than others](http://doi.org/10.22514/jocpd.2024.138).

<br>

Two Methodology Documents:

> [2022 National Survey of Children's Health Methodology Report](https://www2.census.gov/programs-surveys/nsch/technical-documentation/methodology/2022-NSCH-Methodology-Report.pdf)

> [2022 National Survey of Children's Health Data Users Frequently Asked Questions (FAQs)](https://www2.census.gov/programs-surveys/nsch/technical-documentation/methodology/2022-NSCH-FAQs.pdf)

<br>

One Haiku:

```{r}
# "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:

```{r eval = FALSE , results = "hide" }
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:

```{r eval = FALSE , results = "hide" }
nsch_screener_url <-
	"https://www2.census.gov/programs-surveys/nsch/datasets/2022/nsch_2022_screener_Stata.zip"

nsch_topical_url <-
	"https://www2.census.gov/programs-surveys/nsch/datasets/2022/nsch_2022_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:

```{r eval = FALSE , results = "hide" }
# nsch_fn <- file.path( path.expand( "~" ) , "NSCH" , "this_file.rds" )
# saveRDS( nsch_df , file = nsch_fn , compress = FALSE )
```

Load the same object:

```{r eval = FALSE , results = "hide" }
# 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:
```{r eval = FALSE , results = "hide" }
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:
```{r eval = FALSE , results = "hide" }
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:
```{r eval = FALSE , results = "hide" }
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:
```{r eval = FALSE , results = "hide" }
nsch_list <- split( nsch_long_df , nsch_long_df[ , 'implicate' ] )
```

Define the design:
```{r eval = FALSE , results = "hide" }
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:
```{r eval = FALSE , results = "hide" }
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:
```{r eval = FALSE , results = "hide" }
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:
```{r eval = FALSE , results = "hide" }
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:
```{r eval = FALSE , results = "hide" }
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:
```{r eval = FALSE , results = "hide" }
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:
```{r eval = FALSE , results = "hide" }
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:
```{r eval = FALSE , results = "hide" }
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:
```{r eval = FALSE , results = "hide" }
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:
```{r eval = FALSE , results = "hide" }
MIcombine( with( nsch_design ,
	svyratio( numerator = ~ liveusa_yr , denominator = ~ sc_age_years , na.rm = TRUE )
) )
```

### Subsetting {-}

Restrict the survey design to only children:
```{r eval = FALSE , results = "hide" }
sub_nsch_design <- subset( nsch_design , agepos4 == 1 )
```
Calculate the mean (average) of this subset:
```{r eval = FALSE , results = "hide" }
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:
```{r eval = FALSE , results = "hide" }
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:
```{r eval = FALSE , results = "hide" }
degf( nsch_design$designs[[1]] )
```

Calculate the complex sample survey-adjusted variance of any statistic:
```{r eval = FALSE , results = "hide" }
MIcombine( with( nsch_design , svyvar( ~ sc_age_years ) ) )
```

Include the complex sample design effect in the result for a specific statistic:
```{r eval = FALSE , results = "hide" }
# 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:
```{r eval = FALSE , results = "hide" }
# MIsvyciprop( ~ under_six_ever_breastfed , nsch_design ,
# 	method = "likelihood" )
```

### Regression Models and Tests of Association {-}

Perform a design-based t-test:
```{r eval = FALSE , results = "hide" }
# MIsvyttest( sc_age_years ~ under_six_ever_breastfed , nsch_design )
```

Perform a chi-squared test of association for survey data:
```{r eval = FALSE , results = "hide" }
# MIsvychisq( ~ under_six_ever_breastfed + poverty_categories , nsch_design )
```

Perform a survey-weighted generalized linear model:
```{r eval = FALSE , results = "hide" }
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](https://www2.census.gov/programs-surveys/nsch/technical-documentation/methodology/NSCH-Analysis-with-Imputed-Data-Guide.pdf) recommend a method matching the `MIcombine` syntax shown above.** 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?](https://www.childhealthdata.org/browse/survey/results?q=10033&r=1&g=1081):

```{r eval = FALSE , results = "hide" }
results <-
	svyby( 
		~ as.numeric( overall_health == 'excellent or very good' ) ,
		~ poverty_categories , 
		nsch_design$designs[[1]] , 
		svymean , 
		na.rm = TRUE 
	)

published_proportions <- c( 0.817 , 0.862 , 0.922 , 0.947 )

published_lb <- c( 0.797 , 0.843 , 0.914 , 0.941 )

published_ub <- c( 0.835 , 0.879 , 0.930 , 0.952 )

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 ) )
```