In UMCSTaR/medicareR: Standardize Medicare Data and Create Analytic File

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

These are the commonly used covaraites in medicare outcomes models.

Fixed effects:

'surgeon_years_experience'
'sex_male',
'age_at_admit_std',
'ahrq_score_std',
'race_white',
'ses_top3quintiles',
'emergent_admission',
'year', # facility claim year
"had_assist_surg",
'hospital_urban', # from flg_hosp_urban_cbsa
'hospital_beds_gt_350',
'hospital_icu',
'hospital_rn2bed_ratio_std',
'hospital_mcday2inptday_ratio_std'

Note: Do not use flg_hosp_urban to define hospital_urban.

flg_hosp_urban values in 2007's medicare data were all missing. Since we only use complete data in the analysis, we would lose all 2007's data in the models.

Random effects:

'procedure', # ECS label or project-specific procedure group
'cpt_cd',
'id_surgeon',
'id_hospital'

Code to name the varaibles in medicare

The code to rename all original medicare variable to more readable names is in function medicareAnalytics::prep_data_for_model.

Below is the renaming code from the function above.

data = data %>%
    mutate(sex_male = ifelse(flg_male == 1, 1, 0),
           hospital_beds_gt_350 = ifelse( e_hosp_beds_4grp %in% c(3,4), 1, 0)) %>% 
    rename(
      id_surgeon = id_physician_npi,
      id_hospital = facility_prvnumgrp,
      procedure = e_proc_grp_lbl,
      year = facility_clm_yr_from_year0,
      surgeon_years_experience = val_yr_practice,
      hospital_urban = flg_hosp_urban_cbsa,
      hospital_icu = flg_hosp_ICU_hosp,
      hospital_rn2bed_ratio_std = val_hosp_rn2bed_ratio_std,
      hospital_mcday2inptday_ratio_std = val_hosp_mcday2inptday_ratio_std,
      hospital_rn2inptday_ratio_std = val_hosp_rn2inptday_ratio_std,
      death_30d = flg_death_30d,
      severe_complication = flg_cmp_po_severe_not_poa,
      readmission_30d = flg_readmit_30d,
      reoperation_30d = flg_util_reop,
      any_complication = flg_cmp_po_any_not_poa,
    )