Identifying Patient Social Risk from Administrative Health Care Data.

socialrisk

The goal of socialrisk is to create an efficient way to identify social risk from administrative health care data using ICD-10 diagnosis codes.

Installation

You can install socialrisk from CRAN with:

install.packages("socialrisk")

The development version of socialrisk can be downloaded from GitHub with:

# install.packages("devtools")
devtools::install_github("WYATTBENSKEN/socialrisk")

Background on Social Risk and Z-Codes

With the introduction of the 10th Revision of the International Classification of Disease (ICD-10) in 2015 came a new class of codes, which start with the letter Z and thus are referred to as Z-codes which aimed to allow for care providers to capture social factors of their patients. These codes include factors which influence health and health outcomes such as homelessness or housing insecurity, unemployment, low education, and economic concerns. As these were ICD-10 codes they followed a standard format as all other diseases and start with the letter Z and thus are referred to as Z-codes. Since their implementation they’ve attracted substantial attention from policy makers and health care leaders as a potential mechanism to document and report a practice or system’s patient needs, as well as a mechanism to adjust purchasing or payment programs.

Z-codes have been referred to as both “social determinants of health” as well “health-related social needs.” While oftentimes used interchangeably, these are distinct constructs and the lexicon around this has been of high attention recent. From this work, we first note that social determinants of health (SDoH) operate at the community level. These are the characteristics of the collective community, neither positive nor negative, which can affect health.15,16 When SDoH are adverse they engender social risk factors for individuals within the community. However, not every patient will identify or prioritize a social risk as a need that must be tended to. When a patient identifies a social risk factor as a priority for which assistance is needed, that risk becomes a need termed a health-related social needs (HRSN).

There is notable divergence in the literature around which Z codes, and their conceptual domains, are part of social risk/social need analyses. First, CMS in their work has included Z55 through Z65 as the relevant Z codes to capture HRSN. This range of codes was expanded on by the Missouri Hospital Association to create 5 domains distinct domains (socioeconomic status, employment, housing, family and psychosocial), which have subsequently been used in additional studies. Finally, the third source which has provided a list and domains for Z codes is the Social Interventions Research and Evaluation Network (SIREN) at the University of California San Francisco. In the SIREN Compendium, which includes both ICD codes as well as HRSN codes for other taxonomies, there is a much wider range of ICD Z-codes, beyond Z55-Z65 that are included.

This package allows one to implement any three of these approaches and taxonomies in their data using ICD-10 Z-codes to allow for flexibility and efficiency in identifying social risk and social needs from administrative claims data.

Package and Function Options

Data Cleaning

This package includes a clean_data() function which allows us to obtain a cleaned dataset by specifying the dataset name, patient ID variable, whether the data is wide (multiple diagnosis columns) or long (single diagnosis column), as well as the prefix of the diagnoses columns. It’s recommended that this is run on all datasets to prepare them for then running the socialrisk() function.

clean_data(dat = i10_wide, id = patient_id, style = "wide", prefix_dx = "dx")

Social Risk Codes

There are three different taxonomies built into socialrisk() which are called with the taxonomy argument as such:

socialrisk(dat = data, id = id, dx = dx, taxonomy = "cms")
socialrisk(dat = data, id = id, dx = dx, taxonomy = "mha")
socialrisk(dat = data, id = id, dx = dx, taxonomy = "siren")

Please see the vignette for greater details on how to implement these functions.