Abstract
Background:
The accuracy of data on suicide-related presentations to Emergency Departments (EDs) has implications for the provision of care and policy development, yet research on its validity is scarce.
Objective:
To test the reliability of allocation of ICD-10 codes assigned to suicide and self-related presentations to EDs in Queensland, Australia.
Method:
All presentations due to suicide attempts, non-suicidal self-injury (NSSI) and suicidal ideation between 1 July 2017 and 31 December 2017 were reviewed. The number of presentations identified through relevant ICD-10-AM codes and presenting complaints in the Emergency Department Information System were compared to those identified through an application of an evolutionary algorithm and medical record review (gold standard).
Results:
A total of 2540 relevant presentations were identified through the gold standard methodology. Great heterogeneity of ICD-10-AM codes and presenting complaints was observed for suicide attempts (40 diagnostic codes and 27 presenting complaints), NSSI (27 and 16, respectively) and suicidal ideation (38 and 34, respectively). Relevant ICD codes applied as primary or secondary diagnosis had very low sensitivity in detecting cases of suicide attempts (18.7%), NSSI (38.5%) and suicidal ideation (42.3%). A combination of ICD-10-AM code and a relevant presenting complaint increased specificity, however substantially reduced specificity and positive predictive values for all types of presentations. ED data showed bias in detecting higher percentages of suicide attempts by Indigenous persons (10.1% vs. 6.9%) or by cutting (28.1% vs. 10.3%), and NSSI by female presenters (76.4% vs. 67.4%).
Conclusion:
Suicidal and self-harm presentations are grossly under-enumerated in ED datasets and should be used with caution until a more standardised approach to their formulation and recording is implemented.
Keywords
Get full access to this article
View all access options for this article.