Who's Missing: Identifying Work-Relatedness in Emergency Department Data Using ICD-9-CM Codes

Tuesday, June 16, 2015: 5:00 PM
104, Hynes Convention Center
Laurel Harduar Morano , University of North Carolina, Chapel Hill, NC
Scott Proescholdbell , North Carolina Department of Health and Human Services, Raleigh, NC
David B. Richardson , University of North Carolina, Chapel Hill, NC

BACKGROUND: Hospital discharge and emergency department (ED) data are an important resource for population-based occupational surveillance.   Identification of work-related injuries is difficult in these data systems.  Identification of work-related cases based upon worker’s compensation (WC) as payer misses 33-47% of all work-related injuries/illnesses.  We explored the use of an additional method for identifying work-related injuries in ED data based on ICD-9-CM codes.

METHODS:   Two sets of potential work-related ICD-9-CM codes were identified: probable = E000.0,1 (civilian/military activity for pay) and possible = work-related transportation Ecodes, location Ecodes (farm, mine/quarry, industrial), and work-related Vcodes. All 2010-2013 North Carolina (NC) ED records where age ≥16 with an injury code and payer = WC or one of the potential work-related codes were obtained from NC Department of Health and Human Services NC DETECT system. Descriptive statistics were calculated. A random sample of 800 records with available triage notes were selected and manual review of triage notes was conducted to verify occupational injury.

RESULTS:   131,742 injuries were identified: 68.9% (90,793) by WC, 18.3% (24,167) by probable work-related codes, and 12.7% (16,782) by possible work-related codes. The male to female ratio was higher among the probable and possible work-related groups than among the WC group (OR = 1.3 and 1.2, respectively). The age distribution was different among the three groups. WC method identified 81.2% of all amputations but only 60.7% of all sprains/strains. Probable work-related codes identified 10% of amputations and 22.8% of sprains/strains. 100,482 records (76.3%) had a cause-of-injury Ecode of which 17.4% and 13.6% of all falls were identified by probable and possible work-related codes, respectively. Verification by triage notes found: 47% (188) of the 400 records with WC were confirmed work-related, 5.5% (22) were probable work-related, 1 record was not work-related; 33.5% (67) of the 200 records with probable work-related codes were confirmed work-related, 13% (26) were probable work-related, and 1.5% (3) records were not work-related; 22% (44) of the 200 records with possible work-related codes were confirmed work-related, 9% (18) were probable work-related, and 11% (22) records were not work-related.

CONCLUSIONS:   The use of the ICD-9-CM codes greatly increased the number of work-related injuries ascertained. The demographics of cases ascertained by ICD-9-CM codes differ from those ascertained by WC.   Taking triage notes as a gold standard, use of ICD-9-CM codes improved sensitivity but reduced specificity in work-related case ascertainment. Users of these data should be aware of the implications of coding practices in their state.