In HunterRatliff1/hcup: Implementation of HCUP Software Tools in R

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.path = "man/figures/README-",
  out.width = "100%"
)

hcup- Implementation of HCUP Software Tools in R

Adapts the software provided by the Healthcare Cost and Utilization Project (HCUP) for analyses of administrative database using International Classification of Diseases (ICD) codes within the R statistical environment.

Installation

# Hide the warning messages 
library(hcup)
library(dplyr)
library(purrr)
library(tidyr)

You can install the development version of hcup from GitHub with:

# install.packages("devtools")
devtools::install_github("HunterRatliff1/hcup")

Classify diagnosis codes by chronicity (CCI)

This works for ICD-9 or ICD-10 diagnosis codes, although the categories differ across ICD versions

library(hcup)
classify_chronic(icd_dx = "G4730", icd_version = "dx10")
classify_chronic(icd_dx = "56081", icd_version = "dx9")


# Vectorized version
icd_codes <- c("G4730", "L563", "M25151")
classify_chronic(icd_codes, icd_version = "dx10")

Classify procedures by class

Identifies procedures by Procedure Classes, which can be [1] diagnostic or therapeutic and [2] whether they major (expected to be performed in an operating room) or minor. This works with both ICD-9-CM procedure codes and ICD-10-PCS procedure codes.

classify_pr("0PS443Z") # Spine surgery
classify_pr("009U3ZX") # Lumbar puncture
classify_pr("0015")    # Infusion of IL-2

Classify ICD-9 into CCS

This function categorizes ICD codes for both diagnoses and procedures into mutually exclusive Clinical Classifications Software (CCS) categories. This was designed for use with ICD-9 codes, but the HCUP used to provide a beta version that works with ICD-10

library(dplyr)
## Look up ICD-9 DX codes
classify_ccs("8442", code_type = "dx9", level = "single")

# Vectorized
tibble(ICD = c("8442", "1403", "9682",  "36463", "3595")) %>%
  mutate(CCS      = classify_ccs(ICD, code_type = "dx9", "single"),
         CCS_expl = explain_ccs(CCS))

## Works the same for ICD-9 PR codes
classify_ccs("066", code_type = "pr9")

As mentioned above, classify_ccs() will also work with ICD-10 using a beta version that is no longer updated, but the CCSR (below) is preferred. However, if you're working with a mix of ICD-9 and ICD-10 codes, you can use classify_ccs() understanding the limitations

# Fake patient data
tibble::tribble(
  ~patient, ~pr_vers,     ~code,
       "A",       10, "0TY10Z1",
       "B",       10, "0RJM4ZZ",
       "C",       10, "0210088",
       "D",        9,    "3352",
       "E",        9,    "3323",
       "F",        9,    "3639"
  ) %>%

  mutate(CCS = case_when(
    pr_vers==9  ~ classify_ccs(code, code_type = "pr9"),
    pr_vers==10 ~ classify_ccs(code, code_type = "pr10"))) %>%
  mutate(expln = explain_ccs(CCS))

Classify ICD-10 into CCSR

The Clinical Classifications Software Refined (CCSR) software for ICD-10 works similar to CCS for procedures, but capitalizes on ICD-10's specificity

classify_ccsr_pr("0016070")

library(dplyr)
tibble(icd_10 = c("04L04DZ", "009Y40Z", "0F7C7DZ", "GZ2ZZZZ")) %>%
  mutate(CCSR      = classify_ccsr_pr(icd_10),
         explained = explain_ccsr(CCSR))

For ICD-10 diagnostic codes, the specificity actually creates a bit of a problem as CCSR doesn't categorize diagnoses into mutually exclusive categories (see vignette("ccsr-dx")).

ccsr_dx("I110") # Hypertensive heart disease with heart failure
ccsr_dx("I110") %>% explain_ccsr()

This is wonderful in some senses, because it captures more information, but makes it difficult to answer some questions (e.g. do admissions for heart failure cost more than admissions for hypertension?). To address this, the principal diagnosis can still be categorized into a single CCSR category.

tibble::tribble(
  ~Pt_id, ~DX_NUM,     ~ICD,
     "A",  "DX_1",   "I110",
     "A",  "DX_2", "F17210",
     "B",  "DX_1",   "I150",
     "B",  "DX_2",   "Z370",
     "B",  "DX_3",   "E876"
  ) %>%

  # Limit to the principal diagnosis
  filter(DX_NUM == "DX_1") %>%
  mutate(prin_CCSR      = classify_ccsr_dx1(ICD, setting="ip"),
         principal_CCSR = explain_ccsr(prin_CCSR))