Clustering Analysis of a Colombian Toxicological Database

Abstract

ABSTRACT The database analysis allows the return of experience needed to support decision-making processes in risk management. A Colombian toxicological database (TED) developed and maintained by the Center of Safety Information and Chemical Products (CISPROQUIM by its Spanish abbreviation in) is analyzed here using a demographic clustering technique. Data-quality processes were performed on the raw data (more than 170 variables) and as a result the database was reduced to 20 meaningful variables. The variables characterized by values with categories were selected for clustering analysis: gender, age, type of emergency, emergency location, means of poisoning, product use, and physical state of the toxic substance. Clustering analysis showed that there are three profiles that are prevalent in the TED database: Young Adult Suicidal Woman, Unsupervised Child, and Man at Work. These profiles could not be identified using traditional statistical analyses performed on the data collected by CISPROQUIM or defined a priori from the categorical variables. The identification of vulnerable populations and the cause of the toxicological events are critical in order to develop national prevention programs and policies. The analysis described provided a methodology for a critical analysis of toxicological databases that can be applied to other databases such as security databases.