A model to improve the accuracy of US Poison Center data collection

Over 2 million human exposure calls are reported annually to United States regional poison information centers. All exposures are documented electronically and submitted to the American Association of Poison Control Center's National Poison Data System. This database represents the largest data source available on the epidemiology of pharmaceutical and non-pharmaceutical poisoning exposures. The accuracy of these data is critical; however, research has demonstrated that inconsistencies and inaccuracies exist.

This study outlines the methods and results of a training program that was developed and implemented to enhance the quality of data collection using acetaminophen exposures as a model.

Eleven poison centers were assigned randomly to receive either passive or interactive education to improve medical record documentation. A task force provided recommendations on educational and training strategies and the development of a quality-measurement scorecard to serve as a data collection tool to assess poison center data quality. Poison centers were recruited to participate in the study. Clinical researchers scored the documentation of each exposure record for accuracy. Results. Two thousand two hundred cases were reviewed and assessed for accuracy of data collection. After training, the overall mean quality scores were higher for both the passive (95.3%; + 1.6% change) and interactive intervention groups (95.3%; + 0.9% change). Data collection accuracy improved modestly for the overall accuracy score and significantly for the substance identification component. There was little difference in accuracy measures between the different training methods.

Despite the diversity of poison centers, data accuracy, specifically substance identification data fields, can be improved by developing a standardized, systematic, targeted, and mandatory training process. This process should be considered for training on other important topics, thus enhancing the value of these data in relation to public health safety.