In kamclean/finalpsm: Propensity-score matching with tidyverse and finalfit

knitr::opts_chunk$set(echo = F)
library(dplyr); library(tibble); library(survival)

data <- tibble::as_tibble(survival::colon) %>%

  dplyr::filter(etype==2) %>% # Outcome of interest is death
  dplyr::filter(rx!="Obs") %>%  # rx will be our binary treatment variable
  dplyr::select(-etype,-study, -status) %>% # Remove superfluous variables

  # Convert into numeric and factor variables
  dplyr::mutate_at(vars(obstruct, perfor, adhere, node4), function(x){factor(x, levels=c(0,1), labels = c("No", "Yes"))}) %>%
  dplyr::mutate(rx = factor(rx),
                mort365 = cut(time, breaks = c(-Inf, 365, Inf), labels = c("Yes", "No")),
                sex = factor(sex, levels=c(0,1), labels = c("Female", "Male")),
                differ = factor(differ, levels = c(1,2,3), labels = c("Well", "Moderate", "Poor")),
                extent = factor(extent, levels = c(1,2,3, 4), labels = c("Submucosa", "Muscle", "Serosa", "Contiguous Structures")),
                surg = factor(surg, levels = c(0,1), labels = c("Short", "Long")))

Overview

finalpsm is designed not just to make the process of propensity-score matching itself easier, but also the process of modeling with the matched dataset. This aims to act as an extention to the finalfit package to create final results tables

Risk adjustment was subsequently undertaken using multivariable binary logistic regression models (based on the variables used to generate the propensity score) to allow doubly robust estimation

Funk MJ, Westreich D, Wiesen C, Stürmer T, Brookhart MA, Davidian M. Doubly robust estimation of causal effects. Am J Epidemiol 2011; 173: 761–767.

This does not include mixed-effects models given the propensity-score matching would be assumed to trump ...

Each has the same output structure: 1. "balance" = 2. "fit" The 3. "table" = The finalfit table for the psm analysis 4. "metric" The finalfit metrics for the psm analysis. 5. ? plot??????????????????

All the usual assumptions should still be investigated as appropriate.

Binary Logistic Regression

data %>%
  finalpsm::match(strata = "rx",
                    explanatory = c("age","sex", "obstruct", "differ", "surg"),
                    id = "id",
                    dependent = c("mort365", "time", "status"),
                    method = "full") %>%
  finalpsm::finalpsm(dependent = "mort365")

Cox-Proportional Hazards

data %>%
  finalpsm::match(strata = "rx",
                    explanatory = c("age","sex", "obstruct", "differ", "surg"),
                    id = "id",
                    dependent = c("mort365", "time", "status"),
                    method = "full") %>%
  finalpsm::finalpsm(dependent = "survival::Surv(time, status)")

Linear Regression

data %>%
  finalpsm::match(strata = "rx",
                    explanatory = c("age","sex", "obstruct", "differ", "surg"),
                    id = "id",
                    dependent = c("mort365", "time", "status"),
                    method = "full") %>%
  finalpsm::finalpsm(dependent = "survival::Surv(time, status)")