Many states are developing statewide data bases of emergency department records that can potentially provide useful information about occupational injuries. In September, 2001 MDPH received funding under NIOSH Grant R01-OH-004262 to evaluate the feasibility of using electronic medical records from the Massachusetts Emergency Department Injury Surveillance System (EDISS), a stratified sample of Massachusetts hospital emergency departments, to characterize the magnitude and distribution of occupational injuries treated in EDs in Massachusetts and to identify potential gaps in the traditional occupational injury data systems. The study aims were: 1) To estimate the sensitivity, specificity, and predictive value of two indicators of work-related injury in EDISS, independently and in combination, 2) To assess the extent to which information about employer, industry and occupation were included in the medical charts and/or the electronic systems of participating hospitals, 3) To estimate the accuracy of E-codes for work-related injuries, 4) To evaluate the feasibility of using two alternative coding protocols (Short ICECI and NEISS) for work-related injuries, and 5) To compare summary data on work-related injuries from EDISS and findings from the Massachusetts Survey of Occupational Injuries and Illnesses to identify potential biases in findings based on these alternative data sources. EDISS cases with either Workers' Compensation (WC) designated as expected payer or an Injury-at-Work (IAW) value of "yes", or both, were considered probable work-related (PWR). All other cases were considered probable non-work-related (PNWR). Medical charts were reviewed for a stratified study sample of 1002 PWR and 250 PNWR cases drawn from injuries to persons aged 14 through 75 years treated in hospital EDs during the one year period March 1, 1999 through February, 2000. Chart abstractors recorded information on patient employment, injury circumstances and codes, and payment. Abstractors also assessed the "work-relatedness" of the injury based on narrative information in the chart, independent of the payer source and IAW field. E-codes were assigned by an expert E-coder based on the narrative injury descriptors. Findings from EDISS for the period October - September 2000 were compared with findings from the Massachusetts Survey of Occupational Injuries and Illness for the calendar year 2000. The indicators of work relatedness (WC and IAW) had sensitivity, specificity and predictive value positive (PVP) of 83%, 98% and 85%, respectively, using the chart review assessment of work-relatedness as the gold-standard. However, the medical charts were an imperfect measure of work-relatedness because they capture only the information both queried and recorded by emergency department staff. Insufficient information (other than IAW or WC) to determine work relatedness, rather than contradictory information, largely accounting for the relatively low PVP. The use of the IAW field was infrequent and inconsistent across hospitals making it difficult to assess the value of this field in capturing work-related injuries to workers not covered or not applying for WC. Employer name was present in 89% of the charts of work related cases reviewed, less so for non-work-related cases, and was available as a data field in all medical registration systems surveyed, suggesting this information is readily available for industry specific analyses and special studies. External cause of injury E-codes were found to be accurate to four digits for 65% of work-related cases (59% overall), similar to or higher than previous studies. Using broad cause of injury categories typically used for injury surveillance, E-codes were accurate in 85% of cases (work related and overall). The comparison of EDISS data and the BLS SOII was problematic. The two systems covered different segments of the working population, used different broad case definitions, and used different nature and cause of injury coding systems that were difficult to reconcile. These systems should be seen as complimentary. The comparison did raise concerns about under reporting and sampling error in SOII at the state level that need to be further explored. EDISS has the capability to capture a large and representative sample of injuries, both work related and non-work related, across a large and representative sample of the state population into a data set with considerable analytical flexibility, providing a powerful tool for conducting surveillance of workplace injuries. Improvements in the attribution of work-relatedness through more consistent use of the IAW field, and in the accuracy of e-codes, both of which could be addressed through training, would further enhance the value of EDISS. The feasibility of including of an activity field in electronic medical records should be explored to enhance the value of ED data in general and occupational surveillance. Practical methods for routinely using employer information should also be explored to enhance the ability to target surveillance and intervention activities.
Occupational Health Surveillance Program, Massachusetts Department of Public Health, 250 Washington Street, Boston, MA 02108