Enhancement of the National Poison Data System for Detection of Incidents of Public Health Significance
Project Name: Enhancement of the National Poison Data System for Detection of Incidents of Public Health Significance
Project Status: Proposed
Point of Contact: Royal Law
Keywords: Surveillance, Poison Center, Analysis
Project Description: Since 2001, the Health Studies Branch (HSB) has collaborated with the American Association of Poison Control Centers (AAPCC) in the maintenance and enhancement of the National Poison Data System (NPDS), a surveillance system using poison center (PC) data to identify incidents of public health significance. On average every 10 minutes, all regional PCs that cover the US and US territories upload data collected from calls made to their organizations to the national poison center reporting database known as NPDS. NPDS is owned and operated by AAPCC. HSB and AAPCC have developed methods to use NPDS data for near real-time surveillance of exposures to hazardous substances of potential public health significance. Incidents of public health significance are defined in NPDS as illnesses or exposures called to poison centers that may impact the public health and may need further follow up by appropriate state and local public health officials. Examples of incidents of public health incidents include carbon monoxide exposures causing a school evacuation or a group of marine toxin exposures that may indicate a commercial seafood outbreak. HSB uses NPDS to: – improve national surveillance for chemical, environmental, drug, foodborne, biological and radiological exposures and illnesses of potential public health significance – identify early markers of chemical, environmental, drug, foodborne, biological and radiological events in order to provide an effective and rapid public heath response – identify and track exposures and cases of illness during an emerging or known public health threat The methods used to detect data anomalies have been tested and refined over the years, and are based on the methodologies used by the National Notifiable Disease Surveillance System and CDC Influenza Surveillance. For one main method, an anomaly is identified when the number of calls to a poison center exceeds a statistical threshold within a given time frame per PC. Automated data anomalies are jointly reviewed by medical toxicologists and epidemiologists in HSB and AAPCC using standardized criteria for public health significance. For any anomaly that rises to the categorization of an incident of public health significance, a national notification protocol is initiated. This protocol is vetted by CDC, the Council of State and Territorial Epidemiologists, and AAPCC, details the content of the communication message and designates points of contact for PCs and departments of health. The appropriate PC and department of health contacts are designated based on the location of the incident, and a standardized message with details of the incident are sent simultaneously to the PC and public health contact. This information about the identified incident can be used for situational awareness, collaboration opportunity, and initiation of a public health response to the incident as necessary. One aspect of automated data anomaly detection is that the system receives call record information from individual PCs in periodic batches, rather than continuously. In order for a call record to be uploaded by the local PC servers and be accepted by NPDS, a minimal number of fields must be completed, such as demographic and exposure information. Because of this step in the data transmission process, some issues arise with NPDS automated data anomaly detection. Sometimes, during emergencies when PCs are receiving a large number of calls each hour, PC specialists cannot fill out all fields and upload call record data to NPDS in a timely manner. Thus, case records are not uploaded to NPDS in time for the automated algorithms to identify the anomaly. For example, a PC receives a lot of calls due to a lot of exposures related to a hazardous substance spill at 4:00PM. Because the PC is overwhelmed with handling the medical information and performing triage for the calls, PC staff cannot upload the case records into NPDS until 11:00PM that night. By then, the automated anomaly detection algorithm for the surveillance period of 4:00-5:00PM has passed. The public health emergency posed by the hazardous spill was missed by the surveillance system and the relevant public health officials were not notified about the incident.
The HSB and AAPCC technical team for NPDS has identified this lag in reporting of call records to NPDS as a critical fault in the ability of NPDS to detect public health events. One proposed solution is to have NPDS rerun the same anomaly detection algorithms 24 hours after the surveillance period and filter out anomalies that were identified in the previous day. The added function will only identify new anomalies that were otherwise missed during the first run of the surveillance algorithms. In the above example, the automated algorithms will rerun at 4:00PM the next day, and will identify the anomaly that was missed the first time around. This addition will enhance the ability of the NPDS anomaly surveillance system to identify incidents even when the PC staff are overwhelmed with calls and are forced to enter call record data after the fact. The 24-hour window was chosen because this is the longest reported time lag between initiation of a surge and eventual upload of call record data into NPDS. Having NPDS rerun the anomaly detection algorithms later than 24 hours may reduce the timeliness of notifications to public health officials and reduce the public health impact of NPDS surveillance activities. The validation phase of the enhancement will determine whether the 24-hour window is sufficient in alleviating the critical fault, or whether a 48-hour or 72-hour window is more appropriate.