The 131st Annual Meeting (November 15-19, 2003) of APHA

Abstract #72247

Bruce A. Lawrence, PhD¹, Ted R. Miller, PhD¹, and Harold B. Weiss, MPH, PhD². (1) Pacific Institute for Research and Evaluation, 11710 Beltsville Drive, Suite 300, Calverton, MD 20705-3102, 301/755-2731, lawrence@pire.org, (2) Center for Injury Research and Control, University of Pittsburgh, 200 Lothrop St., Suite B400, Pittsburgh, PA 15213

In injury research, analysts are often concerned with the intent underlying an injury episode – i.e., whether the injury was unintentional, self-inflicted, or intentionally inflicted by another person. Classifying the intent of injuries is especially important in selecting cases for research on suicide and interpersonal violence. For this study, the authors examined all poisoning cases from pooled hospital inpatient data containing all 1997 discharges from 19 states, which represented just over half the U.S. population. Of 124,101 cases, 55.1% were E-coded as self-inflicted, 34.9% as unintentional, 5.4% as undetermined intent, and 0.2% as other. Another 4.4% lacked any E code. Therefore, the intent of nearly 10% of poisonings was unknown. This percentage varied widely, from 4.3% in California to 19.1% in Maine. Given that most suicide attempts involve poisoning, this has important implications for estimating suicide counts and for interstate comparisons of suicide rates. In order to determine which of the unknown-intent cases were probably self-inflicted, the authors estimated a logistic regression on the cases E-coded as unintentional or self-inflicted, and then applied this model to the cases with unknown intent. Independent variables in the regression included patient demographics, admission and discharge information, and diagnosis information including specific poison substances and conditions for which medicine is commonly taken. The resulting model correctly identified the intent of 90% of known self-inflicted poisonings. The model estimated that 53% of the poisoning cases of unknown intent were self-inflicted.

Learning Objectives:

• The participant will learn a method for examining and classifying poisoning cases whose intent cannot be inferred from an E code on the data record.

Keywords: Suicide, Data/Surveillance

Presenting author's disclosure statement:
I do not have any significant financial interest/arrangement or affiliation with any organization/institution whose products or services are being discussed in this session.