Going Beyond Chief Complaints to Identify Opioid-Related Emergency Department Visits

Andrew Walsh

Abstract


ObjectiveTo identify heroin- and opioid-related emergency department visitsusing pre-diagnositc data. To demonstrate the value of clinical notesto public health surveillance and situational awareness.IntroductionOverdoses of heroin and prescription opioids are a growingcause of mortality in the United States. Deaths from opioids havecontributed to a rise in the overall mortality rate of middle-aged whitemales during an era when other demographics are experiencing lifeexpectancy gains.1 A successful public health intervention to reversethis mortality trend requires a detailed understanding of whichpopulations are most affected and where those populations live. Whilemortality is the most relevant metric for this emerging challenge,increased burden on laboratory facilities can create significantdelays in obtaining confirmation of which patients died from opioidoverdoses.Emergency department visits for opioid overdoses can provide amore timely proxy measure of overall opioid use. Unfortunately, chiefcomplaints do not always contain an indication of opioid involvement.Overdose patients are not always conscious at registration whichlimits the amount of information they can provide. Menu-drivenregistration systems can lump all overdoses together regardless ofsubstance. A more complete record of the emergency departmentinteraction, such as that provided by triage notes, could provide theinformation necessary to differentiate opioid-related visits from otheroverdoses.MethodsEmergency department registration data was collected fromhospitals via the EpiCenter syndromic surveillance system. Thisdata included chief complaints, triage notes, discharge disposition,and preliminary diagnosis codes. Data elements were linked across agiven visit using patient identifiers and visit numbers as appropriate.Heroin- and opioid-related indicators were identified in chiefcomplaints and triage notes using regular expressions. These wereseparated into three categories: visits with an indication of overdose,visits for withdrawal symptoms, and visits where opioids werementioned in some other context such as history of use. Thesecategories were designed to be mutually exclusive.Regular expression classification results were compared toclassifications based on opioid-related diagnosis codes.ResultsA total of 2,934,610 ED registrations with triage notes and diagnosiscodes were collected from 82 hospitals between January 1, 2015 andAugust 21, 2016. Of these encounters, 24,012 referenced opioid usein some way; 16,718 mentioned heroin specifically; 3,663 mentionedfentanyl specifically; and 5,350 mentioned opioids generically.Table 1 shows the distribution of heroin-related ED visits acrosscategories and source of the indicator. Column totals are not the sumof individual row amounts; they have been adjusted so that a givenregistration is only counted once.Table 2 shows the overlap of heroin-related ED visits betweensources of indicators. Triage notes showed the least overlap with theother two sources, while chief complaints showed the most.ConclusionsWhile it is possible to find indicators of opioid use or overdosein chief complaint data, that field alone does not provide totalinformation about which ED visits are related to opioids. Triagenotes in particular indicate opioid involvement in a large numberof visits not identified by other data sources. While many of theseare simply mentions of opioids, possibly indicating past history ofuse or even in some cases just that questions about opioid use wereasked, a substantial number of visits with overdose indicators werealso detected solely from triage note data. These results suggest thattriage notes can be a valuable additional data source for more complexhealth concerns such as opioid drug use.Table 1: Heroin-Related ED Visits By Indicator Source and CategoryTable 2: Overlap of Heroin-Related ED Visits between Indicators

Full Text:

PDF


DOI: https://doi.org/10.5210/ojphi.v9i1.7617



Online Journal of Public Health Informatics * ISSN 1947-2579 * http://ojphi.org