Only four of the fourteen-predictor variables on univariate analysis made unique statistically significant contributions to a logistic regression model. The odds ratios suggest that crisis nurses were twice as likely to refer patients who were not in any relationship or those who were known to psychiatric services compared to those who were in a relationship or not known to psychiatric services respectively; nineteen and six times respectively more likely to refer patients presenting with psychotic symptoms or a drug or alcohol problem compared to someone presenting with a medical problem and twenty and seven times respectively less likely to refer patients who were given a probable primary diagnosis of ‘drug/alcohol use disorder’ or ‘personality disorders’ compared to patients who were given the probable primary diagnosis of a ‘Depressive disorder’, controlling for other factors in the model.

In a task-shifting model of care, protocols need to be developed with clear guidelines around patients who need to be referred to a psychiatrist for further management.