Overview

This repository contains the analytic dataset, R code, and supporting classification files used to examine associations between Psychosocial Safety Climate (PSC) and occupational health outcomes (work-related stress and injury) among lone and non-lone workers in the United States, using data from the 2022 General Social Survey (GSS).

This is a secondary analysis of de-identified, publicly available survey data. No raw GSS microdata are redistributed in this repository. Only the derived analytic dataset created by the authors, and two small occupation/industry classification crosswalks, are shared.

Data Source

Original data: 2022 General Social Survey (GSS), NORC at the University of Chicago, public-use cross-sectional file.
GSS is publicly available at: https://gss.norc.org/

Researchers wishing to reproduce the full processing pipeline from raw GSS data must download the original 2022 GSS data directly from NORC's official GSS Data Explorer / repository and comply with GSS's data use terms. File paths to the raw GSS file in the cleaning script are placeholders and must be updated.

Human Subjects / De-identification Statement

This dataset is derived from the 2022 General Social Survey (GSS), a public-use, de-identified dataset released by NORC at the University of Chicago, which applies its own statistical disclosure limitation procedures prior to public release (e.g., no exact geography below the national/regional level, top-coded/bracketed income and age where applicable, no names, addresses, or dates of birth).

The dataset shared here (PSC_LWs.csv) contains no direct identifiers (no names, addresses, exact birth dates, or exact interview/exam dates). Demographic and health-related variables retained (age in years, sex, race, marital status, education category, and self-rated general health) are unmodified or recoded/binned versions of variables already present in NORC's public-use GSS file and are required as covariates in the study's adjusted statistical models. Total family income (income) was recoded during cleaning but is not used in any reported analysis and has been removed from the shared file to minimize retained indirect identifiers. No variable identifies geography more granular than the national GSS sample, no rare medical condition or diagnosis is recorded (only a 5-category self-rated general health scale), and no religion variable is included. The survey design variables vstrat and vpsu are NORC-assigned variance strata/primary sampling units used for design-based variance estimation and are part of the standard public-use GSS release; they do not indicate geographic location.

Files Included

Data files

Software needed: Any spreadsheet program (Excel, Google Sheets, etc.) or a statistical package (R, Python, Stata) can open these CSV files directly. Reproducing the cleaning and analysis scripts requires R with the following packages: readxl, dplyr, janitor, stringr (cleaning), and tidyr, survey, srvyr, gtsummary, flextable, officer, MASS (analysis).

Code files

Data Dictionary — PSC_LWs.csv

N = 2,034 rows (one per eligible respondent). Missing values are coded as blank/NA.

Note on the PSC composite score: The standardized PSC score (psc_score_std) used as the main exposure in the regression models is not a pre-computed column in PSC_LWs.csv. It is calculated in PSC_LWs_analysis.R by summing strmgtsup_fact, psysamephys_fact, and allorglevel_fact (each coded 1–5, retaining only respondents with non-missing data on all three items) and then standardizing (z-scoring) the sum. Users who want the pre-computed score can either run the relevant lines of PSC_LWs_analysis.R or contact the authors.

Note on binary recodes used only at analysis time: PSC_LWs_analysis.R also creates Lonewrkr_binary (0 = Non-Lone Worker, 1 = Lone Worker; "Partial Lone Worker" excluded) and sex_binary (0 = Male, 1 = Female) from lnwkr_cat and sex_cat respectively. These are not separate columns in the CSV.

Lone Worker Classification (lnwkr_cat)

Lone worker status was determined using a combination of:

1. Work arrangement — respondent works mostly alone (partteam = "No, I work mostly on my own") and rarely or never works from home (wrkhome = "Never" or "A few times a year"); and
2. Occupation — whether the respondent's detailed occupation code is on a predefined list of occupations considered eligible for lone work (provided in lw_occupation_codes.csv, a subset of the codes in occupation_classification_codes.csv).

Respondents meeting both criteria 1 and 2 are classified Lone Worker; those meeting criterion 1 only are classified Partial Lone Worker; all others are Non-Lone Worker. Respondents missing wrkhome, partteam, or occupation code are coded NA.

Note on income_group

Data_Cleaning_PSC_LWs.R constructs an income_group variable (family income collapsed into ordered brackets) from the raw GSS income item, but it is not used in Table 1 or in any regression model reported for Aim 1. It has been dropped from PSC_LWs.csv to avoid sharing an unnecessary indirect identifier; the cleaning script still shows how it was derived for transparency, but the final select() step in that script no longer includes it in the exported file.

Eligibility / Exclusions

Starting from the 2022 GSS respondent pool, the analytic sample excludes respondents who are retired, students, unemployed/looking for work, keeping house, "other" work status, on active duty military, or in a military-specific occupation. Remaining missing-value codes (e.g., "Don't know," "Refused," "Inapplicable," "Skipped on Web") from the raw GSS file are recoded to NA throughout.

Reproducibility

All processing and analysis steps are documented in the accompanying R scripts. File paths to raw GSS data and the Excel-based classification files in Data_Cleaning_PSC_LWs.R are placeholders and must be updated by users after they obtain the original public 2022 GSS dataset directly from NORC.

Ethics Statement

This study was determined to be exempt from Institutional Review Board (IRB) review. All data shared here are de-identified and derived from the publicly available 2022 General Social Survey. See the Human Subjects / De-identification Statement above for details on identifier handling.

Contact

Corresponding author: Stefani Florez-Acevedo, spflorez@uw.edu, ORCID: https://orcid.org/0000-0001-8148-5799]

Dataset Citation

The data supporting the findings of this study are publicly available in Dryad Digital Repository at: https://doi.org/10.5061/dryad.05qfttfhg

Human subjects data

This dataset is derived from the 2022 General Social Survey (GSS), a public-use, de-identified dataset released by NORC at the University of Chicago, which applies its own statistical disclosure limitation procedures prior to release. The shared file contains no direct identifiers (no names, addresses, exact birth dates, or exam/interview dates). Demographic variables retained (age, sex, race, marital status, education category, self-rated health) are recoded/binned versions of variables in NORC's public-use file, required as covariates in the study's adjusted statistical models. A recoded income bracket variable was removed from the shared file as it was not used in any reported analysis. No variable indicates geography more granular than the national GSS sample, no rare medical condition is recorded, and no religion variable is included.