Data sources
We used the Healthcare Cost and Utilization Project (HCUP) Nationwide Inpatient Sample (NIS) from the Agency for Healthcare Research and Quality for 2005 to 2015.9 We chose these years as NIS coding practices were different prior to 2005, and in 2016 NIS moved to the International Classification of Diseases, 10th revision coding system. With 8 million discharges a year, NIS is the largest inpatient care database in the USA and contains a stratified sample of non-federal, short-term, and general specialty hospitals; it also provides sample weights that allow national estimates to be derived. However, while NIS provides geographic region, it does not provide state-level details, thus all results could be reported only at the regional level—Northeast, South, Midwest, West.
We identified admissions caused by firearm-related injuries using the International Classification of Diseases, ninth Revision, Clinical Modification (ICD-9-CM)) codes.10 We included patients if they had an ICD-9-CM diagnosis code of E922.0–0.3, 0.8, 0.9, E955.0–0.4, E965.0–4, E979.4, E985.0–0.4, or E970. Because E-codes distinguish emergency conditions, we were able to identify injured patients.
We did not include patients treated and released from an emergency department (ie, not admitted) or subsequent inpatient experiences not associated with an E-code (ie, readmission). We derived injury severity score using ICD Programs for Injury Characteristic, a Stata module that translates diagnosis codes into standard injury categories and scores.11 We reported race/ethnicity according to NIS classification: White, Black, Hispanic, Asian/Pacific Islander, Native American, and other. Asian/Pacific Islanders, Native Americans, and other comprised small numbers in our sample and were combined as ‘other’.
Primary outcome of interest was cost associated with hospitalizations for firearm-related injuries and payer of record. NIS contains information regarding the total charges billed for services and cost to charge ratios, allowing us to estimate costs. We inflation-adjusted costs to 2015 dollars using Consumer Price Index rates.
We used the Brady Gun Law Score card as a measure of firearm regulation restrictiveness by state. We converted letter grades for each state into a condensed continuous variable (all A grades given a score of 5, all B grades given a score of 4, down to F given a score of 0) for each region. T-tests and ANOVA tests were used to determine significance between firearm restrictiveness and firearm hospitalizations by region.
We performed unadjusted and adjusted analyses. We used the Student t-test for normally distributed continuous data. We used Χ2 analysis and analysis of variance to compare categorical variables. We considered p<0.05 to be significant. We used linear regression to determine adjusted costs on the basis of payer status. Regression variables included demographics, length of stay, injury severity scores, and hospital region; we controlled for center.
We used Stata SE version 14.1 (StataCorp LP, College Station, TX) for analyses. NIS contains survey strata using US Census division, location, teaching status, ownership, and bed size. We applied survey weights according to HCUP recommendations to create national estimates for the entire US population.