One hundred and nine closed files from an undisclosed state within Australia were selected by searching the National Coronial Information System (NCIS) for cases of deaths from 'unnatural' causes. The final sample included cases where the final determination was suicide (n=29), accident (n=36), homicide (n=22), and undetermined (n=22). Each file was reviewed to identify the presence or absence of pre-determined demographic and evidentiary variables, based on those factors known to be associated with manner of death determinations. Binary logistic regression and classification and regression trees were used to identify significant predictor variables, and formulate decision making models for each manner of death outcome type. Based on the optimal classification tree models for each manner of death determination type, individual cases where coroners' determinations did not match the statistically predicted outcome were identified as representing evidence/determination discrepancies. It was hypothesised that an active cause of death (e.g., asphyxia), adult deaths (>18 years), having a formal history of mental illness, and the presence of a suicide note will predict suicide manner of death determinations. A passive cause of death (e.g., substance related), child-deaths (<18 years), absence of a history of mental illness and no suicide note will significantly predict accidental' death and undetermined manner of death determinations. It was also expected that the police hypothesis of the manner of death will significantly predict each respective manner of death outcome (e.g., a police hypothesis of suicide would significantly predict a coronial determination of suicide). Secondly, it was hypothesised that there would be a less than 100% match between coroners' predicted determinations in cases of equivocal death and the observed determination the classification tree model). Due to their relatively unequivocal nature, no discrepancies between predicted and observed outcomes were expected in cases of homicide. The results found that asphyxia related deaths, presence of a suicide note, a negative mood state (or stressors) prior to death, and a police hypothesis of suicide, were predictors of suicide determinations. The predictors of accident determinations included the absence of a negative mood state (or stressors) prior to death, and the absence of a police hypothesis of homicide. The logistic regression analysis also found that substance related deaths significantly predicted accident determinations, but this was not replicated by the classification tree model. A police hypothesis of homicide also significantly predicted determinations of homicide. Against what was hypothesised, age and a history of mental illness were not significant predictors for any manner of death determination type. However, as expected, there were overall discrepancies between the observed manner of death determinations and what was statistically predicted. It was concluded that a more standardised approach is necessary to reduce coroner's susceptibility to making inconsistent decisions.
History
Publication status
Unpublished
Rights statement
Copyright 2010 the author - The University is continuing to endeavour to trace the copyright owner(s) and in the meantime this item has been reproduced here in good faith. We would be pleased to hear from the copyright owner(s). Thesis (PhD)--University of Tasmania, 2010. Includes bibliographical references