Accuracy of Suicidal Behaviors in Administrative Data as Measured by International Classification of Diseases, Tenth...

Suicidal behaviors present public health challenges worldwide. Surveillance and research aimed at preventing suicidal behaviors often rely on administrative data. Existing systematic reviews examine the validity of methods for identifying suicidal outcomes in administrative datasets but do not include codes based on the International Classification of Diseases, Tenth Revision (ICD-10), despite their widespread use for >2 decades. This rapid review evaluates methods for identifying suicidal behaviors using ICD-10 codes.

We searched PubMed and PsycINFO to identify relevant studies worldwide. Studies were included if they measured suicidal behaviors, used ICD-10 codes, validated the behaviors against a gold standard, and provided measures of accuracy (eg, specificity, sensitivity). We extracted data on study populations, data sources, ICD-10 codes used, gold-standard comparators, and accuracy measures from included studies.

Of 2246 studies identified, 9 met our inclusion criteria. For methods identifying suicide attempts, sensitivity ranged from 19% to 45%, specificity from 47% to 99%, positive predictive value from 21% to 82%, and negative predictive value from 65% to 92%. For methods identifying self-harm, sensitivity ranged from 12% to 85%, specificity from 98% to 100%, positive predictive value from 64% to 100%, and negative predictive value from 86% to 87%. Only sensitivity was reported for suicide death (range, 78%-97%).

Findings indicate that sensitivity and positive predictive value for identifying suicide attempts using ICD-10 codes are low. Studies and surveillance methods relying on these codes might substantially underestimate rates of suicidal behaviors. ICD-10 codes may provide higher accuracy for identifying self-harm and other nonfatal suicidal behaviors than codes from the International Classification of Diseases, Ninth Revision.