Addressing health disparities using multiply imputed injury surveillance data

Table 2 Comparison of the different missing data strategies in the simulation study: estimating crude rate of nonfatal assault injuries treated in EDs (Emergency Departments) per 100,000 population among non-Hispanic white individuals

	Weighted Number of Injuries	Crude Rate per 100,000 Population ^b	95% CI	Absolute Deviation ^c \(\|\frac{\left({\varvec{R}}{\varvec{a}}{\varvec{t}}{\varvec{e}}-{\varvec{S}}{\varvec{T}}{\varvec{D}}\right)}{{\varvec{S}}{\varvec{T}}{\varvec{D}}}\|\)
CCA	217,080	244.5	(175.3, 313.7)	32.3%
JM MI	275,609	310.4	(223.3, 397.5)	14.1%
FCS MI	321,759	362.4	(255.0, 469.8)	0.3%
Standard Control^a	320,629	361.2	(257.9, 464.5)	-

The simulation study built on real non-missing data from National Electronic Injury Surveillance System-All Injury Program (NEISS-AIP) 2018 was conducted to identify the appropriate MI model for handling missing data
Abbreviations: NH non-Hispanic, CCA complete case analysis, JM joint modeling, MI multiple imputation, FCS fully conditional specification, EDs Emergency Departments
^aNon-missing data was used as standard control data
^b Excludes sexual assault cases
^cAbsolute deviation was calculated by using the standard control rate as the reference

ISSN: 1475-9276