Allergy Safety Events in Health Care: Development and Application of a Classification Schema Based on Retrospective Review

Abstract

BACKGROUND:

Allergy safety requires understanding the operational processes that expose patients to their known allergens, including how and when such processes fail.

OBJECTIVE:

To improve health care safety for patients with allergies, we developed and assessed an allergy safety event classification schema to describe failures resulting in allergy-related safety events.

METHODS:

Using keyword searches followed by expert manual review of 299,031 voluntarily-filed safety event reports at 2 large academic medical centers, we identified and classified allergy-related safety events from 5 years of safety reports. We used driver diagrams to elucidate root causes for commonly observed allergy safety events in health care settings.

RESULTS:

From 299,031 safety reports, 1922 (0.6%) were extracted with keywords and 744 (0.2%) were manually confirmed as allergy-related safety events. Safety failures were due to incomplete/inaccurate electronic health record documentation (n = 375, 50.4%), human factors (n = 175, 23.5%), allergy alert limitation and/or malfunction (n = 127, 17.1%), data exchange and interoperability failures (n = 92, 12.4%), and electronic health record system default options (n = 30, 4.0%). Safety failures resulted in known allergen exposures to drugs (n = 537), including heparin (n = 27) and topical anesthetics such as lidocaine (n = 8); latex (n = 114); food allergens (n = 73); and adhesive (n = 23).

CONCLUSIONS:

We identified 744 allergy-related safety events to inform a novel safety failure classification schema as an important step toward a safer health care environment for patients with allergies. Improved systems are required to address safety issues with certain food and drug allergens.

A survey-based study estimates that at least 10% of American adults have a food allergy;¹ adverse drug reactions (ADR) occur in 10% to 15% of hospital admissions.² With over 36 million annual hospitalizations in the United States,³ it is critical to optimize the safety of the over 7 million annual interactions that patients with allergies have with the health care system. Allergic reactions, one type of adverse drug event, carry high potential for patient morbidity and mortality; however, they are difficult to detect, and health care systems lack comprehensive mechanisms to track and prevent such reactions.^4,5 Previous studies focused on ADR identification; few have included reactions to nondrug allergens, such as foods, latex, and adhesive, which are also important culprits for allergic reactions.^6,7 These studies additionally exclude allergy-related medical errors that expose patients to their known allergens without resulting in patient harm (“near misses”).

Multiple operational processes occur from a patient’s initial health care presentation to their exposure to a known allergen. To improve allergy safety in health care settings, a detailed understanding of each of these steps and the interactions between them is essential. Many potential safety failures contribute to allergy safety events: system contributors such as clinical data exchange and electronic health record (EHR) usability, functionality, and configuration; human contributors include performance contributors such as distraction, fatigue, knowledge, skills, and ability and behavior contributors such as human response to rules, policies, and procedures.^8,9 Human- and system-based processes are definitively intertwined, and suboptimal interactions between humans and the EHR (ie, the human-computer interface—both the EHR and human user perform well on their own, but neither is designed nor trained to perform optimally when working with the other) can also lead to allergy safety events. For example, although a clinician independently documents a patient’s allergies correctly, and the EHR independently provides appropriate clinical decision support (CDS), if the clinician documents the allergies in the incorrect section of the EHR, CDS will fail to activate, resulting in patient harm, particularly for clinicians who rely on CDS to catch such errors.¹⁰

The patient safety field includes safety event frameworks that facilitate event monitoring and root cause analyses. One example is the “Swiss cheese model,” in which each slice of cheese represents a safety checkpoint designed to prevent an adverse event, and each hole in a slice of Swiss cheese represents a safety failure that could allow the safety event to occur.¹ Combining multiple slices of Swiss cheese makes it less likely that an adverse event will occur, but even if multiple safety barriers are set up to prevent an event, the holes in the Swiss cheese might properly align, allowing a preventable adverse event to occur.¹¹ Despite the importance of allergy safety in health care settings, particularly given increasing food and drug allergy prevalence,¹² there is no schema to facilitate tracking and promotion of allergy safety.¹³ Given that improvement of allergy safety relies on tools to understand allergy safety events, we aimed to identify, measure, and classify the safety failures that led to allergy-related safety events at 2 academic medical centers over a 5-year period.

METHODS

Data source, allergy documentation, and allergy-related safety event identification

We searched electronic voluntary safety reports (rL, Cambridge, Mass) filed at 2 large academic medical centers (AMC) in Boston, Mass, from May 5, 2014, to January 1, 2019. Any employee with an EHR login can file a safety report, and there are approximately 32,000 safety reports filed annually across both institutions. Safety reports can be filed in less than 5 minutes, and reporters can request feedback on the outcome of a safety report. Employees are encouraged to file reports for any safety issue, no matter how minor; there is no financial incentive to file a report (Figure E1, available in this article’s Online Repository at www.jaci-inpractice.org).

Allergies can be documented in the EHR in several ways. The preferred method is use of the “Allergy Module,” in which clinicians begin typing an allergy and then choose the most appropriate allergen from a drop-down pick list. This pick list includes medications, foods, diagnostic contrast agents, latex, and other common allergens. If the appropriate allergen is not available, clinicians can enter the allergy as a “free text” entry, which is not detected by the clinician decision support system. Clinicians might also enter allergies elsewhere in the EHR, such as in a clinical note, but they are trained and encouraged to enter allergies using the pick list in the allergy module. The allergy module then interacts with other electronic health systems at the institution and within the enterprise.

Because coded entries in the allergy module undergo analysis through CDS, clinicians enter not only true, IgE-mediated, type 1 hypersensitivity reactions into the module, but also intolerances and contraindications that would make it unsafe for a patient to be exposed to a certain allergen (eg, a patient with G6PD deficiency may be listed as “allergic” to chloroquine, or a patient with a history of spontaneous subdural hematoma may be listed as “allergic” to aspirin). By doing so, the entering clinician ensures that an alert is triggered for future prescribing clinicians, thus adding an additional layer of safety protection.

We used a keyword search algorithm on safety report narrative free text entered within “event description” to identify potential allergy safety events. We expanded a previously developed lexicon of allergist- and pharmacist-devised terms related to allergic reactions³ to include terms with high likelihood of indicating both an allergy and a patient safety issue (Figure E2, available in this article’s Online Repository at www.jaci-inpractice.org). Keyword searches applied to 299,031 safety reports identified 1922 reports describing potential allergy-related safety events.

Creation of allergy safety event classification schema

The allergy safety event classification schema was developed using qualitative methodology in which data are reviewed until thematic saturation is reached, meaning no additional understanding of the breakdown of allergy safety events is derived by reviewing additional events.¹⁴ A random sample of 300 safety reports meeting the extraction criteria were reviewed and discussed by a coding group of allergists/immunologists (NAP, PW, KB), biomedical informaticians (LW, LZ), and quality and safety experts (NAP, PW, EM, DWB, CS, KB). We developed a preliminary classification schema (ie, coding frame with hazard categories) to categorize allergy safety failures by content analyses. To ensure that this classification schema would still be applicable to data not directly used to create the classification schema, we then reviewed 5 additional random samples of 10 to 20 events matching extraction criteria. Given that no additional schema modifications were needed based on these additional reviews, we confirmed that inter-reviewer agreement in classifying safety events using this schema was high and proceeded to classify all safety events matching extraction criteria using this schema. Guided by limited terminologies in other safety-related domains, our coding frame included allergy safety categories, sub-categories, and safety failures attributable to 3 underlying root causes established by expert consensus based on a review of the literature: system contributors, human performance contributors, and human behavior contributors.^8,9,15,16

Allergy safety event coding

Each potential allergy safety event was reviewed and coded by 2 annotators, all with expertise in both quality and safety and allergy/immunology (NAP, PW, KB). Discrepancies were discussed until there was agreement by all. If there were multiple potential contributing factors, the one felt to be most responsible was chosen; if 2 or more factors were equally responsible, they were all marked as contributing factors for that event. For events lacking sufficient information related to CDS/allergy alert firing, we queried our institution’s enterprise data warehouse to determine the exact allergy alert(s) that did or did not appear to the EHR user as well as how that user responded to these alerts; this information was used to appropriately code uncertain events. Both institutions used Longitudinal Medical Record (Boston, Mass), a home-grown EHR system in 2014 and had transitioned to Epic (Verona, Wisc) by 2019; our study includes data from both EHR systems.

Data analysis

We combined allergy-related safety events across both institutions. We described overall allergy safety failure counts as well as counts for the level of harm associated with these events; the most commonly reported, potentially preventable allergy safety events from common allergens were displayed using driver diagrams.

We report descriptive characteristics using numbers with frequencies. All analyses were performed in SAS (version 9.4; Cary, NC). This study was approved by the Mass General Brigham Institutional Review Board before data collection with waiver of informed consent from study participants.

RESULTS

From 299,031 safety reports, 1922 (0.6%) reports were identified by the keyword search algorithm to potentially contain an allergy safety event; 744 (0.2%) reports were confirmed as allergy safety events by manual review (Figure E2, available in this article’s Online Repository at www.jaci-inpractice.org). Thematic saturation was reached,¹⁴ and allergy safety events were coded using the newly developed classification schema, which included 5 primary categories with 22 subcategories (Table 1, Figure 1): (1) incomplete/inaccurate EHR documentation (n = 375, 50.4%), including entering/updating allergies in the allergy module (n = 219, 29.4%); inadequately obtaining, confirming, and recording allergy history (n = 81, 10.9%); administration of medications without EHR orders (n = 35, 4.7%); free text allergy entries (n = 21, 2.8%); and failure to adequately confirm patient identity during the patient care process (n = 19, 2.6%); (2) human factors (n = 175, 23.5%), including inadequate intra- and interteam communication (n = 121, 16.3%), overridden allergy alerts (n = 49, 6.6%), and inadequate personnel training (n = 5, 0.7%); (3) allergy alert limitation and/or malfunction (n = 127, 17.1%), including the absence of allergy alerts for kit contents and devices (n = 86, 11.6%), absence of allergy alerts for minor ingredients (n = 14, 1.9%), alerts that the receiving user could not act upon (n = 12, 1.6%), inadequate alert/problem list synchronization (n = 5, 0.7%), absence of alerts for orders that were active before the allergy was entered (n = 5, 0.7%), inappropriate allergy alerts firing (n = 3, 0.4%), and appropriate allergy alerts not firing (n = 2, 0.3%); (4) data exchange and interoperability failures (n = 91, 12.2%), including inadequate interdepartmental EHR-to-EHR communication within a hospital (n = 60, 8.1%), inadequate care pathway EHR-to-EHR communication (n = 4, 0.5%), inadequate EHR-to-EHR communication at the time of transition between EHR systems (n = 15, 2.0%), and inadequate EHR communication between EHR systems at different hospitals within a health care system (n = 12, 1.6%); and (5) EHR system default options (n = 30, 4.0%), including inappropriate alert filtering based on user settings (n = 13, 1.7%), inadequate allergy dictionary defaults (n = 13, 1.7%), and inadequate medication dictionary defaults (n = 4, 0.5%).

TABLE 1.

Classification of allergy-related safety events in health care settings

Safety failure category	Safety failure subcategory	n = 744 (%)*	Safety failure root cause	Definition	Detail†
Incomplete/inaccurate EHR documentation		375 (50.4)
	Entering/updating allergies in allergy module	219 (29.4)	Human behavior contributor	Pre-existing allergies incorrectly documented	Patient was documented as allergic to oxycodone and Percocet, but his actual allergy was to acetaminophen, a component of Percocet
				New-onset (during hospitalization) allergies not updated at the time of occurrence	Patient developed an ampicillin allergy during his hospitalization, but because this was not entered into the EHR, he received and reacted to another penicillin antibiotic during that same hospitalization
	Inadequately obtaining, confirming, and recording allergy history	81 (10.9)	Human performance contributor	Clinicians did not correctly and/or thoroughly obtain, confirm, or record an allergy history in the appropriate section of the EHR	Only some of a patient’s many food allergies were documented in one example. In another, a patient’s latex allergy was not documented in the EHR
	Administration of medications without EHR orders	35 (4.7)	Human behavior contributor	Verbal orders without associated EHR order	A physician gave a verbal order for morphine without an EHR order, and the patient received morphine despite a morphine allergy listed accurately in the EHR
				Medications without EHR orders	A nurse flushed a patient’s intravenous lines with heparin despite the patient’s having a heparin allergy because the flush happened before (or without) an order, and therefore, no EHR alert fired
	Free text allergy entries	21 (2.8)	Human behavior contributor	Inadequate allergy pick lists; if a specific allergy was absent or difficult to find, clinicians entered free text allergens	A patient was listed as allergic to “*oxycodone” (free text entry with an asterisk) instead of “oxycodone” (coded entry), so clinical decision support did not fire an alert when another narcotic was ordered A patient’s EHR included a free text allergy entry of “sulfa” instead of a coded entry of “sulfonamide” or “sulfamethoxazole-trimethoprim”; as a result, clinical decision support could not cross-check with the patient’s allergy list, and he was prescribed Bactrim (sulfamethoxazole-trimethoprim, a sulfonamide antibiotic)
	Failure to adequately confirm patient identity	19 (2.6)	Human behavior contributor	Clinicians did not confirm patient identity before updating EHR	One patient’s chart was noted to list another patient’s allergies; this was only recognized when the first patient denied being exposed to the multiple medications listed in his allergy list
Human factors		175 (23.5)
	Inadequate intra- and interteam communication	121 (16.3)	Human behavior contributor	Allergy information appropriately documented within the EHR was not adequately communicated to team members who interfaced with the patient without EHR interaction	A latex-allergic patient was ordered for intravenous catheter placement; the team who places these catheters does not always interact with the EHR before making patient contact. As a result, they used latex gloves during the catheter placement
	Overridden allergy alerts	49 (6.6)	Human performance contributor	Alerts overridden due to either inadequate clinical knowledge or alert fatigue	A cefepime-allergic patient had been desensitized to cefepime on a prior admission, inducing a temporary state of tolerance to the drug that lasts only as long as the patient continues to take the drug. The patient completed the antibiotic course and stopped taking the drug but required cefepime on a different admission. The ordering clinician overrode the allergy alert stating that the patient had previously been desensitized to cefepime but did not understand that the desensitized state no longer applied Patients on a ketogenic diet are frequently listed as being allergic to dextrose to help prevent inadvertent administration of the sugar, which would subsequently worsen their seizures; clinicians sometimes override this alert, not understanding the medical basis behind it and thus compromising patient safety
	Inadequate personnel training	5 (0.7)	Human behavior contributor	Staff unfamiliarity with the new EHR system	A patient’s penicillin allergy was appropriately listed in the EHR, but not acknowledged by clinical staff due to unfamiliarity with how to navigate the system; as a result, the patient inadvertently received penicillin
Alert limitation and/or malfunction		127 (17.1)
	Absence of allergy alerts for kit contents and devices	86 (11.6)	System contributor		Patients were exposed to their known allergens through either devices that are not typically scanned into the EHR before use (eg, materials in the supply closets such as catheters, tubing, lines, electrodes, adhesives, etc) or prepackaged procedure kits that contain medications that are not individually scanned before use (eg, central line, lumbar puncture, foley catheter kits may contain lidocaine, chlorhexidine, and/or iodine); despite appropriately documented allergies, clinical decision support was not activated as the allergen-containing kit or device was not scanned before administration. For example, a patient with latex allergy was catheterized with a latex-containing foley catheter. A patient with lidocaine allergy received lidocaine as part of a lumbar puncture. A patient with iodine allergy was cleansed with iodine before foley catheter placement
	Absence of allergy alerts for minor ingredients	14 (1.9)	System contributor	Alerts failed to fire if a medication contained a minimal quantity of a potential allergen	A patient with a known anaphylactic latex allergy was ordered for IV furosemide; even though furosemide vials are known to contain latex, an alert did not fire to notify the ordering clinician or administering nurse of this potential safety issue
	Alerts that the receiving user could not act upon	12 (1.6)	System contributor	Alert fired when a clinician ordered a medication for a patient that matched an active patient allergy, but the clinician was unable to act on the alert based on that clinician’s EHR configuration	A clinician ordered oxycodone for a patient with a documented oxycodone allergy; despite an alert appropriately firing, the EHR build did not prompt the clinician to return to the order entry screen to cancel the order. As a result, the patient received oxycodone and developed hives
	Inadequate synchronization between problem list and alerts	5 (0.7)	System contributor	Medical problem-drug alerts did not fire when a medication was contraindicated	A pregnant woman was ordered for isotretinoin despite a known risk of birth defects A patient with G6PD deficiency was ordered for chloroquine despite a known risk of hemolysis A patient on fluoxetine was ordered for dextromethorphan despite a known risk of serotonin syndrome
	Absence of alerts for orders that were active before the allergy was entered	5 (0.7)	System contributor	Clinical decision support did not cross-check with previously placed, still active medication orders when a new allergy was entered	A patient who had been taking meropenem for several days developed an allergy to the antibiotic; although the clinician caring for the patient appropriately updated the patient’s allergies, the EHR did not fire an alert to inform the clinician that the patient still had an active order for meropenem. As a result, the EHR prompted the nurse to administer the antibiotic as it was scheduled even though it was a known allergen
	Inappropriate allergy alerts firing	2 (0.3)	System contributor	EHR alerted the ordering clinician to a potential medication reaction for an allergy that did not exist	For a patient allergic only to penicillin and iodinated contrast, an alert for “ibuprofen” allergy fired when “acetaminophen” was ordered despite the fact that the patient had never received either medication before
	Appropriate allergy alerts not firing	2 (0.3)	System contributor	Alerts failed to fire when a patient was ordered for a drug to which s/he was known to be allergic	A patient with documented aspirin allergy was ordered for aspirin; the system auto-verified the order, meaning that it was not transmitted to pharmacy or nursing for review after placement of the order. Auto-verification thus bypassed the traditional safety steps that might have otherwise led the reviewing pharmacist or administering nurse to catch the error that the ordering clinician missed
Data exchange and interoperability failures		92 (12.4)
	Inadequate interdepartmental EHR-to-EHR communication within a hospital	60 (8.1)	System contributor	Orders/allergies in the EHR did not communicate with specific health information technology systems used by other institutional departments	A patient with history of angioedema to shellfish received a salad containing shrimp from nutrition. A patient on a ketogenic diet with a documented dextrose allergy received medications mixed in dextrose from pharmacy
	Inadequate care pathway EHR-to-EHR communication	4 (0.5)	System contributor	Clinical order entry system did not communicate with the medication administration system	A patient with a penicillin allergy was ordered for ampicillin; the nurse scanned the medication into the electronic medication administration record, but the allergy alert did not fire
	Inadequate EHR-to-EHR communication at the time of transition between EHR systems	15 (2.0)	System contributor	Institutions transitioned from EHR systems, but allergies were not appropriately transferred	An intravenous contrast allergy for a patient was visible in a legacy viewer of the old EHR system but not in Epic (the new EHR system)
	Inadequate EHR communication between EHR systems at different hospitals within a health care system	12 (1.6)	System contributor	Allergies at one institution were not visible at other institutions even within the same health care system with a shared EHR	A patient’s ceftriaxone allergy, documented at hospital A, was not visible at hospital B despite both institutions sharing an EHR; the patient received ceftriaxone at hospital B and developed hives
EHR system default options		30 (4.0)
	Inappropriate alert filtering based on user settings	13 (1.7)	System contributor	EHR system build prevented users from viewing clinically relevant alerts	A patient with known contrast allergy was ordered for contrast; the ordering user’s EHR configuration filtered the alert, preventing it from firing to notify the ordering user of the safety concern
	Inadequate allergy dictionary defaults	13 (1.7)	System contributor	EHR allergen pick list not comprehensive	A patient with a cocoa allergy presented to care; the EHR did not include “cocoa” (a food) in the allergy list and instead only included “cocoa butter” (a topical emollient) A clinician tried to enter “milk” as the allergen, but only found “milk of magnesia”
	Inadequate medication dictionary defaults	4 (0.5)	System contributor	EHR medication pick list listed incorrect default options	Desensitization protocols defaulted to an intravenous route, although they can be by oral or subcutaneous routes as well When intramuscular or subcutaneous epinephrine was ordered, the default listed dose was 1 mg, not 0.3 mg, the latter of which is the recommended adult dose

This table describes the safety failure categories and subcategories of allergy-related safety events detected through the keyword-search algorithm followed by a manual review. For each subcategory, we describe the number of relevant events and the safety failure root cause (human behavior contributor, human performance contributor, and system contributor); we also provide additional details on that event type including relevant examples.

EHR, Electronic health record.

^*Percentages add to greater than 100% because some events were felt to have multiple equally contributing factors (799 total contributing factors in 744 events).

^†Actual free text examples were modified to preserve peer-protected safety reporting data.

FIGURE 1.

Allergy safety hazards. This “Swiss cheese” schematic provides an overview of the clinical allergy pathway at the system level; each step is a safety hazard that, if not carried out correctly, could expose patients to a known allergen. The timeline demonstrates both the events and the interactions that occur at each step of the process. EHR, Electronic health record.

Drugs were the most common (n = 537, 72.1%) culprit for allergy-related safety events; among them, antibiotics (n = 164, 22.0%) were the most common followed by diagnostic contrast agents (n = 89, 12.0%) and unknown drug culprits (n = 54, 7.3%). Commonly reported and potentially preventable allergy safety events also occurred when allergy checking required communication outside of the primary EHR, such as with latex allergy and procedures involving latex gloves and catheters (n = 114, 15.3%), food allergy communication with nutrition and pharmacy (n = 73, 9.8%), heparin allergy and flushing intravenous lines (n = 27, 3.6%), and prepackaged procedure kits with adhesives (n = 23, 3.1%) or a topical anesthetic such as lidocaine (n = 8, 1.1%) (Figure 2, A-E). A full list of these culprits is included in Table EI, available in this article’s Online Repository at www.jaci-inpractice.org.

FIGURE 2.

Allergen exposure driver diagrams. These prototypical diagrams demonstrate the allergy safety hazards perfectly aligned in a real-world setting to allow the occurrence of a safety event due to (A) latex, (B) food allergens, (C) heparin, (D) adhesive, and (E) lidocaine. Steps that are duplicative within and across figures may warrant the most attention to reduce allergy safety events. EHR, Electronic health record.

The 1922 events were further classified based on the level of harm to the patient as follows: 496 (25.8%) events did not reach the patient and hence caused no harm; 647 (33.7%) events reached the patient, but did not cause harm; 723 (37.6%) events reached the patient and caused temporary or minor harm; 7 (0.4%) events reached the patient and caused permanent or major harm; 3 (0.2%) events resulted in death; and 46 (2.4%) events were not classified by harm level.

DISCUSSION

By applying qualitative research and safety science methods to 5 years of safety report data across 2 major US academic medical centers, we identified 744 allergy-related safety events that informed a novel allergy safety classification schema that can be used to track and reduce patient exposure to known allergens in health care settings, thus helping organizations to identify allergy safety hazards and develop improvement efforts to reduce allergy-related patient safety events. The most common allergy safety failure was attributable to incomplete or inaccurate EHR documentation (>50% of allergy safety events); other human factors such as inadequate communication, purposeful allergy alert overrides, and inadequate training comprised 24% of identified events. Although the EHR is intended to improve allergy safety, we identified that allergy alert limitations or malfunction resulted in 17% of allergy safety events, data exchange and interoperability failures resulted in 12% of allergy safety events, and inadequate EHR system defaults resulted in 4% of allergy safety events. We additionally found that 72% of allergy-related safety events were due to drug culprits, but nondrug culprits such as foods, latex, and adhesive were also responsible for many events. This work is an important step in allergy safety optimization including allergy safety assessment, reporting, and prevention.

Maintaining allergy safety throughout the patient care continuum relies on multiple safeguards (Figure 1); at a systems level, safety events occur when workflows circumvent these safeguards. Clinicians must obtain, confirm, document, and communicate a complete allergy history using an intuitive, user-friendly EHR that is configured to accept relevant data entry and provide an appropriate level of CDS that neither misses important potential safety issues nor induces alert fatigue in the user.^17-19 Ideally, this action is performed in real time to prevent inadvertent re-exposures, particularly to newly discovered allergens and those such as chlorhexidine, which may not be as common as drug allergens. In addition, clinical information and orders must communicate with all clinical systems within and across institutions. Lastly, all potential allergic exposures need to be part of this closed system, such that all foods, drugs, and other potential allergens are electronically and manually checked against the EHR allergy list and allergies documented elsewhere in the EHR (clinical notes, clinical flowsheets, and clinician orders).

Health information technology can mitigate allergy safety hazards. Computerized physician order entry and CDS, including drug-allergy interaction alerts, decrease allergy-related medication errors.^20-23 However, with nearly 80% of drug allergy alerts overridden, this decision support alone is insufficient.²⁴ Although some overridden allergy alerts are clinically appropriate, approximately 6% of allergy alert overrides result in adverse drug events, of which nearly half are serious.²⁵ Our current system additionally falls short with significant limitations to allergy documentation (eg, 13 allergy safety events resulted from inaccurate EHR system defaults for allergens such as peanut and cocoa), allergy communication (eg, 91 allergy safety events resulted from lack of allergy communication between EHR systems), and allergy alerting (eg, 127 allergy safety events resulted from alert limitations or malfunctions). Even keyword search on safety reporting free text inadequately captures all allergy safety events reported.²⁶ Prior studies have made specific recommendations to decrease alert overrides.²⁵ Epic has made substantial progress in displaying allergies across institutions, but much work still remains. Given increasing food and drug allergy prevalence,^1,2,12 the high potential for allergy-related morbidity and mortality, and the medicolegal concerns surrounding allergy safety events,²⁷ allergy specialist guidance is imperative to optimizing patient safety; however, even Epic EHR, the dominant EHR system in the United States, which has a steering board of clinicians on a variety of specialty topics, does not have an allergy steering board.

Drugs were the most common culprit for allergy-related safety events. This is not surprising given that a prior study of our hospital system showed that 35.5% of patients seen here reported at least 1 drug allergy; among these, penicillins were the most common, which is similar to our findings here.²⁸ Other EHR systems failures included access to allergens that did not require scanning and interaction with CDS, such as foods, medications (eg, heparin flush, medications contained in procedure kits), and other health care products (eg, latex gloves, adhesive on electrocardiogram leads). Although forcing all items containing potential allergens to be scanned before use would have avoided many of these safety events (eg, events related to medications without orders [n = 35] or prepackaged kits and devices [n = 86]), the added burden of scanning each heparin flush and chlorhexidine swab used may not be operationally feasible in a large health care setting. Understanding how to optimally add barriers that reduce the risk of future allergy safety events without increasing health care team burden is an important area of future study.

EHR systems played a substantial contributing role in the occurrence of allergy-related safety events. Three categories (alert limitation or malfunction, data exchange/interoperability, and EHR system default options) and 4 additional subcategories (overridden alerts, free text allergy entries, administration of medications without EHR orders, and entering/updating allergies in allergy module) comprised 573 (77%) safety failures that could directly be attributed to an EHR deficiency; several others could indirectly be attributed to EHR usability issues. For example, a clinician recorded allergies in the clinical note instead of the EHR allergy module because the module was less easily accessible and user-friendly than free text documentation. In addition, if a health care system’s alerts are not properly configured, and alerts fire inappropriately (eg, alert firing in all egg-allergic patients ordered for the influenza vaccine²⁹), the excess alerts may result in alert fatigue, blinding clinicians to allergy alerts that are clinically important. Lastly, if the EHR interface into which a clinician enters clinical information/orders does not adequately communicate with the EHR interface at which such orders are carried out (eg, pharmacy, nutrition, nursing, radiology), no checking occurs on the back end to inform clinicians about potential safety risks. These issues must be brought to the attention of EHR vendors in the United States, and future studies might evaluate allergy EHR modifications to facilitate clinician allergy and order entry.³⁰

To our knowledge, there is not presently a classification schema for how to understand and classify allergy-related events that occur in the health care setting. This study was the first to create such a classification schema and apply it to understand where to focus safety efforts. As drugs are the primary culprit allergen, prevention efforts should be focused here; however, nondrug culprits such as foods, allergens such as latex and adhesives that are in health care products but not directly administered to the patient, and allergens such as topical anesthetics that are in prepackaged kits must also be an area of focus. In addition, the patient care continuum is a complex series of interrelated steps; some steps are particularly prone to errors that result in allergy-related safety events and should serve as a starting point for safety efforts. For health care organizations across the country, we recommend beginning efforts at these steps with a focus on drug-related culprits. This classification schema, however, is only the first step; institutions also need to devote health care resources and create appropriate infrastructure, to ensure patient safety for allergic individuals.

Our study had several limitations. We relied on voluntarily-reported safety events to extract allergy-related safety failures, but most safety events are not reported, and safety reporting culture varies between institutions and over time; safety reports themselves are 1-sided accounts of events that may omit key information. Because much of the event was described in a nonstructured free text field, keyword searches were necessary to identify events, which may have limited the sensitivity and scope of events identified. Our classification schema was limited to allergy-related safety events identified through keyword searching and may have missed other allergy safety issues such as those related to order sets (eg, incorrect default medication dosage) or reference information (eg, incorrect reference range). In addition, although our safety failure classification schema incorporated a comprehensive overview of system, human performance, and human behavior contributors, we may have missed other contributors such as knowledge gaps from inadequate education or organizational training. Given the absence of tools to classify and count allergy safety events, creating this classification schema required a retrospective observational study design applied to 5 years of multisite data rather than a prospective application of the same. Lastly, our data collection was restricted to 2 urban academic medical centers, and results may not be generalizable to smaller hospitals, nonurban hospitals, or even other similar-sized urban academic medical centers, particularly if they do not have a similarly advanced EHR or safety reporting system in place.

We identified a large number of spontaneously reported allergy safety events where patients were exposed to allergens to which they were known to be allergic. The safety failure categorization we developed to classify events is the initial step toward reducing allergy safety events in health care settings and would ideally form the foundation of an allergy safety failure taxonomy approved by international consensus. Drugs should be the primary target of prevention efforts. Latex, food allergens, heparin, adhesive, and lidocaine warrant special attention because many of these exposures in health care settings bypassed current safety checks. Both human and EHR factors allowed these events to occur; although further study is needed to elucidate the details of specific deficiencies in both areas, prevention efforts cannot exclude either as doing so would be the equivalent of adding another slice of Swiss cheese that is more hole than cheese. Future directions include prospectively validating this classification schema across sites while simultaneously engaging EHR and health information technology leaders to improve allergy documentation and checking systems. Clinically, focusing attention and efforts on these key steps may prevent many potential allergy-related safety events as research moves forward.

What is already known about this topic?

Allergy safety requires complex synchronization of many steps in the clinical care pathway; failures in these processes can cause patients to be exposed to known allergens, resulting in morbidity and mortality.

What does this article add to our knowledge?

This article uses safety reports to demonstrate the allergy safety pathway with key steps that protect patients from allergy-related safety events. It additionally identifies drugs and several other allergens as key safety culprits.

How does this study impact current management guidelines?

This study demonstrates key allergy safety hazards and culprit allergens, thus helping to guide leadership in understanding where in the clinical care pathway to focus safety efforts.

Abbreviations used

ADR

Adverse drug reaction

AMC

Academic medical centers

CDS

Clinical decision support

EHR

Electronic health record