gweissman

# Libraries
library(gmish) # devtools::install_github('gweissman/gmish')

# Size
N <- 1000

# Softmax helper function
sm <- function(x) exp(x)/sum(exp(x))

# Make some predictions

gweissman

---
title: 'Aim 2: Simulated prelim data for Roybal Proposal'
author: "Gary E. Weissman, MD, MSHP"
date: "8/29/2018"
output: html_document
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
library(igraph)

gweissman

# A helper function to one-hot encode a _single_ variable
# Assumes data.table column or vector input
# Returns a data.table with x columns where x is number of levels
ohev <- function(vars, drop_ref = TRUE) {
  vname <- deparse(substitute(vars))
  lvls <- unlist(unique(vars))
  tmp_list <- list()
  for (lev in lvls) {
    tmp_list[[make.names(paste0(vname,'_',lev))]] <- as.numeric(vars == lev)
  }

gweissman

# highlight communities and clusters in an igraph plot

library(igraph)

g <- erdos.renyi.game(50, .02)

# plot regular

plot(g)

gweissman

# extract the ICD10 Cardiac procedure codes from PDF 
# Appendix B of AHRQ PQI v6.0
require(pdftools)
download.file(url = 'http://www.qualityindicators.ahrq.gov/Downloads/Modules/PQI/V60/TechSpecs/PQI_Appendix_B.pdf',
              destfile = 'PQI_Appendix_B.pdf')

raw_text_pages <- pdf_text('PQI_Appendix_B.pdf')
raw_text <- paste(raw_text_list, collapse = '')
tokens <- unlist(strsplit(raw_text, split = "\\s"))
tokens_filt <- tokens[ ! tokens %in% c('APPENDIX', '(ACSCARP)')]

gweissman

# Convert ICD 9 to ICD 10 codes based on the CMS General Equivalency Mappings 2016
# https://www.cms.gov/Medicare/Coding/ICD10/2016-ICD-10-CM-and-GEMs.html

library(data.table)
library(icd)

gem <- fread('~/Desktop/SCIP - Study of Critical Illness Pathways/2018-redo/final_code/2016-General-Equivalence-Mappings/2016_I9gem.txt',
            colClasses = rep('character',3))
setnames(gem, c('icd9', 'icd10', 'flags_composite'))

gweissman

# this script finds ambulatory sensitive conditions (ACSCs) in ICD-9 codes
# the data are based on the appendices from
# http://www.ahrq.gov/downloads/pub/ahrqqi/pqiguide.pdf (Appendix A)
# usage: acsc(list_of_icd9_dx_codes, list_of_drgcodes, list_of_icd9_proc_codes) 
# returns: TRUE or FALSE
# or set getlist = TRUE to get the list of which ACSCs were found

# NB. that some of the definitions for these PQIs exclude hospital transfers
# paripartum admissions, or certain age groups, etc. 
# This script ***only*** assess ACSC status by ICD-9 code.

gweissman

# sample calculating mixing matrix and assortativity coefficient with igraph

require(igraph)
set.seed(12)

# create a random graph
g <- random.graph.game(1000,0.15)

# assign some characteristics
V(g)$color <- c('red','white','blue','orange','green')

gweissman

# calculate the assortativity coefficient for a mixing matrix of a graph
# ref: MEJ Newman, 'Mixing patterns in networks', Phys Rev E 67, 026126 (2003)
#
# define assortativity coefficient as
#        trace (m) - sum (m^2)
# ac = -------------------------
#           1 - sum (m^2)
# 
# where m is the mixing matrix of a graph

gweissman

# calculate the mixing matrix of in igraph graph object 'mygraph', by some vertex attribute 'attrib'
# can change the default use.density=FALSE to return a matrix with raw number of edges rather than density

mixmat <- function(mygraph, attrib, use.density=TRUE) {
  
  require(igraph)
  
  # get unique list of characteristics of the attribute
  attlist <- sort(unique(get.vertex.attribute(mygraph,attrib)))