Published on in Vol 9, No 1 (2017):

Improving Detection of Call Clusters through  Surveillance of Poison Center Data

Improving Detection of Call Clusters through Surveillance of Poison Center Data

Improving Detection of Call Clusters through Surveillance of Poison Center Data

Authors of this article:

Royal K. Law1 ;   Howard Burkom2 ;   Josh Schier1
The full text of this article is available as a PDF download by clicking here.

ObjectiveOur objective was to compare the effectiveness of applying thehistorical limits method (HLM) to poison center (PC) call volumeswith vs without stratifying by exposure type.IntroductionThe Centers for Disease Control and Prevention (CDC) uses theNational Poison Data System (NPDS) to conduct surveillance ofcalls to United States PCs. PCs provide triage and treatment advicefor hazardous exposures through a free national hotline. Informationon demographics, health effects, implicated substance(s), medicaloutcome of the patient, and other variables are collected.CDC uses automated algorithms to identify anomalies in both purecall volume and specific clinical effect volume, and to identify callsreporting exposure to high priority agents. Pure and clinical effectvolume anomalies are identified when an hourly call count exceeds athreshold based on historical data using HLM.1Clinical toxicologistsand epidemiologists at the American Association of Poison ControlCenters and CDC apply standardized criteria to determine if theanomaly identifies a potential incident of public health significance(IPHS) and to notify the respective health departments and localPCs as needed. Discussions with NPDS users and analysis of IPHSshowed that alerting based on pure call volume yielded excessivefalse positives. A study using a 5-year NPDS call dataset assessed thepositive predictive value (PPV) of the call volume-based approach.This study showed that less than 4% of anomalies were IPHS.2A low PPV can cause unnecessary waste of staff time and resourcesanalyzing false positive anomalies.As an alternative to pure call volume-based detection where allcalls to each PC are aggregated for anomaly detection, we consideredseparating calls by toxicologically-relevant exposure categories formore targeted anomaly detection. We hypothesized that this stratifiedapproach would reduce the number of false positives.MethodsWe derived our exposure categories based on the criteria that thecategories must: 1) relate to hazardous exposures of public healthimportance, 2) reflect categories based on clinical effects andtreatment modalities, 3) avoid high priority exposures that may betriggered by single calls, 4) be compatible with exposure substanceidentification codes currently used by PCs and NPDS, and 5) includeenough calls for meaningful tracking. We queried all calls reportingexposures to the proposed categories between January 1, 2009and July 31, 2015 for ten PCs. We applied the HLM method afterstratifying by exposure category and tabulated the number of alertstriggered for each category during the study period. We then appliedthe HLM method for the ten PCs on all combined exposure calls torepresent the traditional non-stratified approach. We compared thecombined alert burden generated by stratifying by exposure categorywith the alert burden for the non-stratified approach for varying timewindows (1-, 2-, 4-, 8- and 24-hours). We conducted analysis in R.ResultsWe derived a total of 20 exposure categories, including chemicals(n=4), drugs of abuse (n=6), pesticides (n=3), gas/fume/vapors (n=2),contaminated food/water (n=1), and others (n=4). Call counts during2015 for these categories ranged from approximately 5,000 to 90,000.Table 1 shows the total number of alerts triggered for each methodby time windows. There was a marked reduction of alert burdenwhen first stratifying by exposure category for time windows shorterthan eight hours compared to the alert burden for the non-stratifiedapproach.ConclusionsStratification of call volume by exposure category and timewindow suggests potential improvement over traditional non-stratified approach by having a lower alert burden. Further workshould focus on refining the exposure categories, refining the timewindow for surveillance, and assessing other detection performancemetrics, such as sensitivity.Table 1: Alert burden comparison for the non-stratified vs stratified approach