9  Step 6: Propensity

9.1 hdPS model

C = investigator-specified covariates and EC = hdPS covariates (Schneeweiss et al. 2009)

Then the hdPS can be used as matching, weighting, stratifying variables, or as covariates (usually in deciles) in outcome model.

(Wyss et al. 2022)

9.1.1 Create propensity score formula

hdps.data$exposure <- as.numeric(I(hdps.data$obese=='Yes'))
hdps.data$outcome <- as.numeric(I(hdps.data$diabetes=='Yes'))
proxy.list.sel <- names(out3$autoselected_covariate_df[,-1])
proxyform <- paste0(proxy.list.sel, collapse = "+")
covform <- paste0(investigator.specified.covariates, collapse = "+")

rhsformula <- paste0(c(covform, proxyform), collapse = "+")
ps.formula <- as.formula(paste0("exposure", "~", rhsformula))
ps.formula
#> exposure ~ age.cat + sex + education + race + marital + income + 
#>     born + year + diabetes.family.history + medical.access + 
#>     smoking + diet.healthy + physical.activity + sleep + uric.acid + 
#>     protein.total + bilirubin.total + phosphorus + sodium + potassium + 
#>     globulin + calcium.total + systolicBP + diastolicBP + high.cholesterol + 
#>     rec_dx_I10_once + rec_dx_R73_once + rec_dx_I10_frequent + 
#>     rec_dx_R60_once + rec_dx_E78_once + rec_dx_M79_once + rec_dx_E87_once + 
#>     rec_dx_I51_once + rec_dx_I50_once + rec_dx_D75_once + rec_dx_K21_once + 
#>     rec_dx_Z79_once + rec_dx_N28_once + rec_dx_F41_once + rec_dx_E79_once + 
#>     rec_dx_M10_once + rec_dx_M1A_once + rec_dx_L70_once + rec_dx_F32_once + 
#>     rec_dx_I80_once + rec_dx_F90_once + rec_dx_F43_once + rec_dx_I50_frequent + 
#>     rec_dx_B00_once + rec_dx_R12_once + rec_dx_N19_once + rec_dx_B20_once + 
#>     rec_dx_E28_frequent + rec_dx_R07_once + rec_dx_M19_once + 
#>     rec_dx_I48_once + rec_dx_G47_once + rec_dx_T14_once + rec_dx_I70_once + 
#>     rec_dx_M06_once + rec_dx_N40_once + rec_dx_R51_once + rec_dx_R21_once + 
#>     rec_dx_J45_once + rec_dx_H40_once + rec_dx_K31_once + rec_dx_I63_once + 
#>     rec_dx_H10_once + rec_dx_R52_once + rec_dx_R19_once + rec_dx_N95_once + 
#>     rec_dx_K92_once + rec_dx_J20_once + rec_dx_E28_once + rec_dx_R32_once + 
#>     rec_dx_M13_once + rec_dx_G30_once + rec_dx_J98_once + rec_dx_N52_once + 
#>     rec_dx_I49_once + rec_dx_N30_once + rec_dx_N42_once + rec_dx_R39_once + 
#>     rec_dx_N92_once + rec_dx_B96_once + rec_dx_H40_frequent + 
#>     rec_dx_M54_once + rec_dx_K59_once + rec_dx_G25_once + rec_dx_F31_once + 
#>     rec_dx_T78_once + rec_dx_R41_once + rec_dx_G31_once + rec_dx_I99_once + 
#>     rec_dx_G30_frequent + rec_dx_C50_once + rec_dx_F39_once + 
#>     rec_dx_R06_once + rec_dx_F17_once + rec_dx_K30_once + rec_dx_R45_once + 
#>     rec_dx_K25_once + rec_dx_B37_once + rec_dx_L08_once + rec_dx_T88_once + 
#>     rec_dx_J42_once + rec_dx_R10_once + rec_dx_M06_frequent + 
#>     rec_dx_G20_once + rec_dx_E07_once + rec_dx_K58_once + rec_dx_R35_once + 
#>     rec_dx_K04_once + rec_dx_R05_once + rec_dx_K08_once + rec_dx_I20_once + 
#>     rec_dx_H57_once + rec_dx_R25_once + rec_dx_H66_once + rec_dx_R42_once + 
#>     rec_dx_B20_frequent + rec_dx_I48_frequent + rec_dx_M62_once + 
#>     rec_dx_G43_once + rec_dx_F29_once

This is an overly simplistic scenario where we are adding only the main effects in the non-transformed form.

9.1.2 Fit PS model

require(WeightIt)
W.out <- weightit(ps.formula, 
                    data = hdps.data, 
                    estimand = "ATE",
                    method = "ps")

9.1.3 Obtain PS

hdps.data$ps <- W.out$ps

Always a good idea to check propensity score overlap in both exposure groups

9.1.4 Obtain Weights

hdps.data$w <- W.out$weights

Always a good idea to check the summary statistics of the weights to assess whether there are extreme weights

9.1.5 Assessing balance

Always a good idea to assess balance. Here we are measuring against SMD 0.1. Use love.plot function from the cobalt package. See more descriptions of balanced diagnostics elsewhere for a propensity score context.

Schneeweiss, Sebastian, Jeremy A Rassen, Robert J Glynn, Jerry Avorn, Helen Mogun, and M Alan Brookhart. 2009. “High-Dimensional Propensity Score Adjustment in Studies of Treatment Effects Using Health Care Claims Data.” Epidemiology (Cambridge, Mass.) 20 (4): 512.

Wyss, Richard, Chen Yanover, Tal El-Hay, Dimitri Bennett, Robert W Platt, Andrew R Zullo, Grammati Sari, et al. 2022. “Machine Learning for Improving High-Dimensional Proxy Confounder Adjustment in Healthcare Database Studies: An Overview of the Current Literature.” Pharmacoepidemiology and Drug Safety 31 (9): 932–43.