New function: derive_params_growth_age - for BY AGE growth parameters

[rossfarrugia]

New function:

derive_params_growth_age(
  dataset, 
  sex, 
  age, 
  age_unit, 
  meta_criteria, 
  parameter, 
  set_values_to_sds, 
  set_values_to_pctl)

Arguments:

• dataset: input dataset
• sex, age and age_unit would define the variables to use from input dataset: e.g. sex = SEX, age = AAGECUR, age_unit = AAGECURU
• meta_criteria: metadata as defined in step 4 of MS Teams doc
• parameter: the input dataset PARAMCD for which we want growth derivations: e.g. parameter = exprs(BMI = VSTESTCD == "BMI"). Options: “BMI”, “HEIGHT”, “WEIGHT”, “HEAD CIRCUMFERENCE”
• set_values_to_sds: variables to be set for zscore (SDS) parameter: e.g. set_values_to_sds(PARAMCD = “BMIASDS”, PARAM = “BMI-for-age z-score”). If left blank then parameter not derived in output dataset.
• set_values_to_pctl: variables to be set for percentile parameter: e.g. set_values_to_pctl(PARAMCD = “BMIAPCTL”, PARAM = “BMI-for-age Percentile”). If left blank then parameter not derived in output dataset.

Checks:

• dataset
  • The variables for sex, age and age_unit are in input dataset – age then first needs to get converted to DAYS using admiral::derive_vars_duration()
  • The PARAMCD in parameter is available in input dataset
  • the unit associated with each parameter is as expected (Height/Head CIRCUMFERENCE in cm, weight in kg)
• metadata
  • has variables for SEX, AGE, L, M, S

Derivations:

Example function call:

advs_wgt_age <- advs %>% 
  derive_params_growth_age(
    sex = SEX, 
    age = AAGECUR, 
    age_unit = AAGECURU, 
    meta_criteria = mymetadata,
    parameter = WGHT,
    set_values_to_sds(
      PARAMCD = “WGTASDS”,
      PARAM = “Weight-for-age Z-Score”
    ),
    set_values_to_pctl(
      PARAMCD = “WGTAPCTL”,
      PARAM = “Weight-for-age Percentile”
    )
  )

NOTE from David for BMI: CDC BMI percentiles >95 are calculated differently than this <95th percentile. Here are the 2 places that 'introduce' the new (2023) CDC calculations for BMIs > 95th percentile

1. https://www.cdc.gov/nccdphp/dnpao/growthcharts/resources/sas.htm - (This a page that I largely wrote). You can see the big notification at the top of the screen and there's a section on 'Extended z-scores and percentiles' about halfway down the long webpage
2. And here's a page written by CDC people in Washington (Natl Center for Health Statistics): https://stacks.cdc.gov/view/cdc/122933

We called them extended BMIz and extended BMI percentiles, but the SAS and R programs name these calculations as 'BMIz' and 'BMIp'. If someone wanted to use the old visions of the calculations to check their work done in years before 2023, the variables are named something like original_BMIz and original_BMIp

[zdz2101]

@kabis-ops here is a rough work-flow of how I've started to get the appropriate metadata/join, "function-ing" it as we speak

library(admiral)
library(admiralpeds)
library(dplyr)
library(lubridate)
advs <- dm_peds %>%
  select(USUBJID, BRTHDTC, SEX) %>%
  right_join(., vs_peds, by = "USUBJID") %>%
  filter(USUBJID != "PEDS-1010") %>%
  mutate(VSDT = ymd(VSDTC),
         BRTHDT = ymd(BRTHDTC)) %>%
  derive_vars_duration(
    new_var = AGECUR_D,
    new_var_unit = CURU_D,
    start_date = BRTHDT,
    end_date = VSDT,
    out_unit = "days",
    trunc_out = FALSE
  ) %>%
  derive_vars_duration(
    new_var = AGECUR_M,
    new_var_unit = CURU_M,
    start_date = BRTHDT,
    end_date = VSDT,
    out_unit = "months",
    trunc_out = FALSE
  ) %>%
  mutate(AGECUR = ifelse(AGECUR_D >= 365.25*2, AGECUR_M, AGECUR_D),
         AGECURU = ifelse(AGECUR_D >= 365.25*2, CURU_M, CURU_D))

# metadata is in months
cdc_meta_criteria <- admiralpeds::cdc_wtage %>%
  mutate(age_unit = "months")

# metadata is in days
who_meta_criteria <- bind_rows(
  (admiralpeds::who_wt_for_age_boys %>%
     mutate(SEX = 1,
            age_unit = "days")),
  (admiralpeds::who_wt_for_age_girls %>%
     mutate(SEX = 2,
            age_unit = "days"))
) %>%
  rename(AGE = Day)

meta_criteria <- bind_rows(
  cdc_meta_criteria,
  who_meta_criteria
) %>%
  arrange(SEX, age_unit, AGE) %>%
  group_by(SEX, age_unit) %>%
  mutate(next_age = lead(AGE),
         SEX = ifelse(SEX == 1, "M", "F"))

# process metadata and join
advs_meta <- advs  %>%
  filter(VSTESTCD == "WEIGHT") %>%  # swap in appropriate PARAM statement
  left_join(., meta_criteria, by = c("SEX", "AGECURU" = "age_unit"), relationship = "many-to-many") %>%
  filter(AGE <= AGECUR & AGECUR < next_age) %>%
  mutate(SDS = ((VSSTRESN/M)^L - 1)/(L*S),
         perc = pnorm(SDS)*100) %>%
  select(-c(L, M, S, next_age)) %>%
  pivot_longer(., cols = SDS:perc, names_to = temp, values_to = "AVAL")

[rossfarrugia]

@zdz2101 reminder here to see the doc David sent around unit testing tips for this function so we cover a good range of boundary cases. Also around his tip on deriving for very high BMI values.