Using the ‘presenting problem’ field in emergency department data improves the enumeration of intentional self-harm in NSW hospital settings

To the Editor

Hospital data are a key source of timely information on intentional self-harm and the only routinely available source for the surveillance of a national public health priority. However, the extent to which accurate information on intentional self-harm is captured, or coded appropriately, in these data sources may be variable. This is important because, in the absence of a formal register or surveillance system of intentional self-harm, these data sources remain primary sources to evaluate the effectiveness of investments in suicide prevention.

The Emergency Department Data Collection (EDDC) was implemented in New South Wales (NSW) in 1994, covering more than 80% of ED presentations from 2007 (O’Connell et al., 2014). A single diagnosis field is used to code self-harm, using the International Classification of Diseases (ICD-10-AM) codes X60-X84 or corresponding SNOMED-CT codes. Additional information is comprised in the ‘Presenting problem’ 50-character free text field that provides a short case description, but which is not routinely coded.

Using data linked to the 45 and Up Study (45 and Up Study Collaborators, 2008) (n = 267,153) in NSW, we identified likely cases of intentional self-harm that appeared in the ‘presenting problem’ field but were not coded as such in the ‘diagnosis field’. We searched the presenting problem field of 1,026,276 records linked to the 45 and Up Study (over the period 2006–2018) looking for words and expressions related to self-harm, with manual checks conducted by two authors.

There were 1118 cases of intentional self-harm (87% of cases were subsequently admitted to hospital), of which 662 cases (59%) were identified from the presenting problem field alone (Table 1). Of these, 10% presented a clear reference to a suicide attempt (expressions such as ‘suicide attempt’, ‘attempted suicide’, ‘attempted hanging’). Another 29% presented clear descriptions of ingestion of a large quantity of pills, sometimes combined with alcohol. 11% of cases were identified in both diagnosis and presenting problem fields.

Table 1.

Intentional self-harm cases that appear in the NSW Emergency Department Data Collection ‘presenting problem’ field (but not coded in the ‘diagnosis field’), grouped by method of identification.

Group	ICD or SNOMED codes	Presenting problem text	Present in both codes and text
1. Clear description of a suicide attempt	0 (0.0)	65 (97.0)	2 (3.0)
2. Intentional overdose, polypharmacy overdose, drugs plus alcohol and others	2 (0.9)	192 (91.0)	17 (8.1)
3. Other self-harm behaviour a	76 (13.0)	405 (69.2)	104 (17.8)
4. Undetermined intent or unrelated to self-harm b	255 (100.0)	0 (0.0)	0 (0.0)
Total	333 (29.8)	662 (59.2)	123 (11.0)

Values represent absolute (relative) frequency.

a

Example text includes ‘attempted self harm’, ‘biba [brought in by ambulance] with self harm, laceration to r [right] arm’, ‘injury unspecified self harm superficial lac [laceration]’, ‘suicidal ideation. cuts lt [left] wrist. biba’.

b

Categorised with ICD or SNOMED code as ‘undetermined’ or ‘unrelated’, but where presenting problem text was suggestive of intentional self-harm (e.g. ‘poisoning’, ‘overdose’, ‘MH-anxious’).

In this older age cohort, it appears that a significant number of intentional self-harm cases are not captured by routine coding, even when there is a description of suicidal intent. Strategies to incorporate information from additional diagnosis fields and case descriptions, perhaps via the implementation of a mandatory intentional self-harm flag or the development of machine learning algorithms using text fields to automate coding, would more accurately capture the incidence of intentional self-harm in the hospital setting. Such strategies would result in an improved resource for the evaluation of suicide prevention activity.