Development of a prediction model for the diagnosis of suspected acute poisoning

Abstract

The diagnosis of acute poisoning is based on clinical assessment, electrocardiography and laboratory investigations. Diagnosis of acute poisoning in patients with altered mental status is challenging because an accurate history may be unavailable or unreliable and clinicians rely on clinical assessment, whilst the role of screening toxicology tests in such patients is not clear. The aim of the project was to develop a prediction model derived from clinical risk factors that would help in the diagnosis of suspected acute poisoning. In the first phase of the project, reliability of a history of poisoning and characteristics of patients with suspected acute poisoning presenting to Mater Dei Hospital, Malta were investigated. The second phase of the project involved development, validation and testing of a prediction model for the diagnosis of acute poisoning in patients with altered mental status. The reliability of a history of poisoning was measured by a meta-analysis of studies comparing history of poisoning with diagnosis, showing that history was moderately reliable and varied with different drugs. Pooled kappa for the reliability of a history of paracetamol poisoning was 0.67 (95%CI 0.64-0.71) whilst history of drugs of abuse was less reliable with a pooled kappa ranging from 0.35 (95%CI 0.14-0.56) for MDMA to 0.48 (95%CI 0.41-0.54) for opiates. Features of patients with suspected acute poisoning referred for toxicology investigations were evaluated by comparing clinical features with toxicology laboratory results evaluating demographics, differential diagnosis, range of ingested drugs and diagnostic yield. 51.7% (350/677) had confirmed acute poisoning, classified as primary alcohol intoxication and drug overdose in 28.3% (99/350) and 69.7% (244/350)' respectively. Toxicology testing was carried out on 677 patients of which 464/677 (68.5%) were positive. Univariate analysis identified significant diagnostic factors for acute poisoning and although individual parameters had limited prognostic value, composite clinical scores were accurate predictors of poisoning severity. Based on these results, it was determined that a prediction model would be suitable for diagnosis of poisoning in patients with altered mental status. In the second phase of the project, clinical data were gathered on patients presenting with altered mental state and their characteristics were studied. Mean age was 54 years and 55.4% (484/873) were males. Age and gender distribution varied according to diagnosis. Drug overdose and isolated alcohol intoxication was present in 21.4% (187/873) and 14.4% (126/873)

respectively. A history of overdose was present in 17.1% (149/873) whilst 15.8% (138/873) had a history of alcohol ingestion. 45.5% of patients with AMS had a toxicology investigation of which 66% had a detectable drug result. Univariate analysis identified significant diagnostic variables. A prediction model based on significant clinical risk factors was developed for use in stratifying risk of acute poisoning in patients with altered mental status. The model was validated internally and externally and a decision rule for use of screening tests was derived and tested. A simplified risk score was derived from the model and was found to have a negative predictive value of 94.2%. In this project the diagnostic value of clinical factors in suspected acute poisoning was investigated and a prediction model based on selected predictive factors was developed, validated and applied into a decision rule that was simple to calculate and may be of value in a pre-hospital environment or at triage.