What this Notebook Contains

Assuming access to the available source data files, the Notebook contains the complete statistical code for replicating:

  • “Table 1. Characteristics of patients in overall sample”

Step 1. Import “Duran_Schillinger_KNPC_restricted_demographics.csv.”

See README in repository for instructions on how to access the .csv file.

Please note: Kaiser Permanante (owners of data) will not allow any individual-level data (even a small subset) to be shown in a public repository. A version of this HTML exists with the restricted .csv files that includes a subset of individual-level data to aid in tutorial interpretation.

library(tidyverse)
library(pander)

df.demo = read.csv("Duran_Schillinger_KNPC_restricted_demographics.csv")
pander(names(df.demo))

PAT_ID, Age, sex, race, educ_p, CHARLSON_INDEX and predictions_LP

  • PAT_ID: Randomly generated numeric code associated with each unique patient
  • Age: Age of patient
  • sex: Sex of patient; 0 = female, 1 = male
  • race: Race of patient; 1 = White, 2 = Black, 3 = Hispanic, 4 = Asian, 5 = Other
  • educ_p: Education level of patient; 1 = No degree, 2 = GED/high school, 3 = Some college or more
  • CHARLSON_INDEX: Co-morbidity index, interval
  • predictions_LP: Patient’s assigned score of “high” or “low” health literacy (otherwise known as “LP Score”) based on previously developed linguistic model; see: S. A. Crossley, R. Balyan, J. Liu, A. J. Karter, D. McNamara, D. Schillinger, Developing and testing automatic models of patient communicative health literacy using linguistic features: Findings from the ECLIPPSE study, Health Commun. (2020)

Step 2: Reassign variable levels for readability

df.demo.prep = df.demo %>%
  mutate(sex = case_when(
    sex=="0" ~ "Women", 
    sex=="1" ~ "Men",
    TRUE ~ NA_character_)) %>%
  mutate(race = case_when(
    race=="1" ~ "White", 
    race=="2" ~ "Black",
    race=="3" ~ "Hispanic", 
    race=="4" ~ "Asian",
    race=="5" ~ "Other", 
    TRUE ~ NA_character_)) %>%
  mutate(educ_p = case_when(
    educ_p=="1" ~ "No degree", 
    educ_p=="2" ~ "GED/high school",
    educ_p=="3" ~ "Some college or more",    
    TRUE ~ NA_character_))

Step 3: For each variable, generate summary statistics based on whether high or low health literacy; Start with education level as example

## Get stratified counts and percentage of each education category
demo_ed = df.demo.prep %>% group_by(educ_p) %>%
  summarize(count=n(), perce = (n()/nrow(df.demo.prep))*100) 
demo_edlow = filter(df.demo.prep, predictions_LP=="Low") %>% group_by(educ_p) %>%
  summarize(count=n(), perce = (n()/nrow(filter(., predictions_LP=="Low")))*100)  
demo_edhigh = filter(df.demo.prep, predictions_LP=="High") %>% group_by(educ_p) %>%
  summarize(count=n(), perce = (n()/nrow(filter(., predictions_LP=="High")))*100)  

## Assemble into readable matrix
demoEduc = matrix(c(
  "No degree ", demo_ed$count[2], round(demo_ed$perce[2],1) , demo_edlow$count[2], round(demo_edlow$perce[2],1), demo_edhigh$count[2], round(demo_edhigh$perce[2],1),  
  "GED/high school", demo_ed$count[1], round(demo_ed$perce[1],1) , demo_edlow$count[1], round(demo_edlow$perce[1],1), demo_edhigh$count[1], round(demo_edhigh$perce[1],1),  
  "Some college or more", demo_ed$count[3], round(demo_ed$perce[3],1) , demo_edlow$count[3], round(demo_edlow$perce[3],1), demo_edhigh$count[3], round(demo_edhigh$perce[3],1) ),  
  ncol=7,byrow=TRUE)
colnames(demoEduc) = c("Levels", "Total", "Total-%", "LowHL", "LowHL-%", "HighHL", "HighHL-%")

pander(head(demoEduc))
Levels Total Total-% LowHL LowHL-% HighHL HighHL-%
No degree 402 9.3 188 12.1 214 7.7
GED/high school 995 23 413 26.5 582 21
Some college or more 2934 67.7 959 61.5 1975 71.3

Step 4: Follow same procedure as above for remaining variables

## Get stratified counts and percentage of each sex category
demo_sex = df.demo.prep %>% group_by(sex) %>%
  summarize(count=n(), perce = (n()/nrow(df.demo.prep))*100)  
demo_sexlow = filter(df.demo.prep, predictions_LP=="Low") %>% group_by(sex) %>%
  summarize(count=n(), perce = (n()/nrow(filter(., predictions_LP=="Low")))*100)  
demo_sexhigh = filter(df.demo.prep, predictions_LP=="High") %>% group_by(sex) %>%
  summarize(count=n(), perce = (n()/nrow(filter(., predictions_LP=="High")))*100)  

## Assemble into readable matrix
demoSex = matrix(c(
  "Women", demo_sex$count[2], round(demo_sex$perce[2],1) , demo_sexlow$count[2], round(demo_sexlow$perce[2],1), demo_sexhigh$count[2], round(demo_sexhigh$perce[2],1) ),
  ncol=7,byrow=TRUE)
colnames(demoSex) = c("Levels", "Total", "Total-%", "LowHL", "LowHL-%", "HighHL", "HighHL-%")

pander(demoSex)
Levels Total Total-% LowHL LowHL-% HighHL HighHL-%
Women 1930 44.6 688 44.1 1242 44.8 
## Get stratified counts and percentage of each race category
demo_race = df.demo.prep %>% group_by(race) %>%
  summarize(count=n(), perce = (n()/nrow(df.demo.prep))*100) 
demo_racelow = filter(df.demo.prep, predictions_LP=="Low") %>% group_by(race) %>%
  summarize(count=n(), perce = (n()/nrow(filter(., predictions_LP=="Low")))*100)  
demo_racehigh = filter(df.demo.prep, predictions_LP=="High") %>% group_by(race) %>%
  summarize(count=n(), perce = (n()/nrow(filter(., predictions_LP=="High")))*100)  

## Assemble into readable matrix
demoRace = matrix(c(
  "White", demo_race$count[5], round(demo_race$perce[5],1) , demo_racelow$count[5], round(demo_racelow$perce[5],1), demo_racehigh$count[5], round(demo_racehigh$perce[5],1),   
  "Black", demo_race$count[2], round(demo_race$perce[2],1) , demo_racelow$count[2], round(demo_racelow$perce[2],1), demo_racehigh$count[2], round(demo_racehigh$perce[2],1),  
  "Hispanic", demo_race$count[3], round(demo_race$perce[3],1) , demo_racelow$count[3], round(demo_racelow$perce[3],1), demo_racehigh$count[3], round(demo_racehigh$perce[3],1),  
  "Asian", demo_race$count[1], round(demo_race$perce[1],1) , demo_racelow$count[1], round(demo_racelow$perce[1],1), demo_racehigh$count[1], round(demo_racehigh$perce[1],1),  
  "Other", demo_race$count[4], round(demo_race$perce[4],1) , demo_racelow$count[4], round(demo_racelow$perce[4],1), demo_racehigh$count[4], round(demo_racehigh$perce[4],1) ),
ncol=7,byrow=TRUE)
colnames(demoRace) = c("Levels", "Total", "Total-%", "LowHL", "LowHL-%", "HighHL", "HighHL-%")

pander(demoRace)
Levels Total Total-% LowHL LowHL-% HighHL HighHL-%
White 1407 32.5 431 27.6 976 35.2
Black 582 13.4 237 15.2 345 12.5
Hispanic 572 13.2 262 16.8 310 11.2
Asian 1348 31.1 484 31 864 31.2
Other 422 9.7 146 9.4 276 10 
## Get mean and SD for continuous variable of age; Assemble into readable matrix
demo_low = filter(df.demo.prep, predictions_LP=="Low") 
demo_high = filter(df.demo.prep, predictions_LP=="High") 

demoAge = matrix(c(
  "Age", round(mean(df.demo.prep$Age),1), round(sd(df.demo.prep$Age),1), round(mean(demo_low$Age),1), round(sd(demo_low$Age),1), round(mean(demo_high$Age),1), round(sd(demo_high$Age),1) ), 
  ncol=7,byrow=TRUE)
colnames(demoAge) = c("Levels", "Total", "Total-%", "LowHL", "LowHL-%", "HighHL", "HighHL-%")

pander(demoAge)
Levels Total Total-% LowHL LowHL-% HighHL HighHL-%
Age 57.2 10 56.6 10.2 57.6 9.8 
## Get mean and SD for interval variable of comorbidity score; Assemble into readable matrix
democoMorbid = matrix(c(
  "Comorbidity Score", round(mean(df.demo.prep$CHARLSON_INDEX),1), round(sd(df.demo.prep$CHARLSON_INDEX),1), round(mean(demo_low$CHARLSON_INDEX),1), round(sd(demo_low$CHARLSON_INDEX),1), round(mean(demo_high$CHARLSON_INDEX),1), round(sd(demo_high$CHARLSON_INDEX),1) ), 
  ncol=7,byrow=TRUE)
colnames(democoMorbid) = c("Levels", "Total", "Total-%", "LowHL", "LowHL-%", "HighHL", "HighHL-%")

pander(democoMorbid)
Levels Total Total-% LowHL LowHL-% HighHL HighHL-%
Comorbidity Score 2.2 1.5 2.2 1.6 2.1 1.5

Step 5: Simple tests to compare counts between high and low health literacy values

library(compareGroups)

df.demo2 = df.demo

df.demo2$sex = as.factor(df.demo2$sex)
df.demo2$race = as.factor(df.demo2$race)
df.demo2$educ_p = as.factor(df.demo2$educ_p)

demo.test <- compareGroups(predictions_LP ~ Age + sex + race + educ_p + CHARLSON_INDEX,
    data = df.demo2)
print(demo.test)
## 
## 
## -------- Summary of results by groups of 'predictions_LP'---------
## 
## 
##   var            N    p.value  method            selection
## 1 Age            4331 0.001**  continuous normal ALL      
## 2 sex            4331 0.671    categorical       ALL      
## 3 race           4331 <0.001** categorical       ALL      
## 4 educ_p         4331 <0.001** categorical       ALL      
## 5 CHARLSON_INDEX 4331 0.067*   continuous normal ALL      
## -----
## Signif. codes:  0 '**' 0.05 '*' 0.1 ' ' 1