This report looks at routinely collected national data on injury cases who attended a public hospital emergency department (ED) in Australia in 2013–14, which are collected in the National Non-admitted Patient Emergency Department Care Database (NAPEDCD). It describes and illustrates possible applications of these data for injury surveillance.
Description of the data
Diagnosis is not coded according to the same system for all records in the NAPEDCD. Two very different systems have been applied: ICD and SNOMED-CT. In addition, several editions of ICD-10-AM were used as was an earlier edition of ICD, ICD-9-CM.
Of the nearly 7.2 million records in the database, 68% included a principal diagnosis coded according to ICD-10-AM, 26% were reported as SNOMED-CT (all in New South Wales), 0.6% were coded according to ICD-9-CM and 5% had a missing diagnosis value.
Use of different coding systems made identification of injury cases difficult in some instances. Conversion of SNOMED-CT terms to equivalent ICD-10-AM codes is not straightforward for certain types of injury.
The data did not include a field for external cause of injury (such as fall or traffic crash). The lack of this item markedly reduced the value of the data for injury surveillance.
Exploration and application
More than 1.3 million ED presentations in 2013–14 were assigned at least 1 ICD-10-AM diagnosis code in the injury and poisoning range S00–T98 (27% of all presentations that were assigned an ICD-10-AM diagnosis code). Of these presentations, almost 93% (1.2 million) were assigned a principal diagnosis in the injury range S00–T75 or T79 (community injury), the usual scope of AIHW injury reports.
For presentations assigned a principal diagnosis of S00–T75 or T79, 22% were assigned a principal diagnosis for a soft-tissue injury, and 21% were assigned a principal diagnosis for a fracture. Overall, almost 18% of ED injury cases were admitted to hospital. Admission proportions varied widely by nature of injury. Among frequently occurring injuries, the proportion admitted ranged from 95% for fracture of the femur to 1.7% for dislocations, sprains, and strains of joints and ligaments at the wrist or hand. ED admission proportions can supplement and add value to reports based on admitted patient data.
Comparisons were made between data from EDs at hospitals with and without a level 1 trauma centre. Differences in patterns of cases generally aligned with expectations, providing some evidence of the face validity of the ED data. For example, the percentage of injury ED cases admitted to the same hospital where a person presented was much higher in level 1 trauma service hospitals (26%) than other hospitals (13%), and high urgency cases were concentrated at the EDs of hospitals with trauma centres.
While linked data studies are needed to provide a more complete assessment of ED injury data, the results presented in this report suggest that the source has value for injury surveillance. The analysis of ED injury data in this report will enable the NAPEDCD to be used for injury surveillance based on more recently available annual ED data sets. The utility of the data source would be improved by including external cause data in the NAPEDCD, and by specifying the coding better, which varied markedly among states and territories.