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

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.

Immediate reaction to ibrutinib amenable to oral desensitization.

INTRODUCTION: Although up to half of patients receiving chemotherapeutic agents develop hypersensitivity reactions to the same, desensitization protocols can induce temporary tolerance to allow patients to continue to receive first-line treatment. Approximately 25% of patients develop cutaneous hypersensitivity reactions to ibrutinib, but there are no published management guidelines. CASE REPORT: We describe the case of a 71-year-old woman with chronic lymphocytic leukemia who developed a delayed maculopapular rash with lip tingling and swelling following ibrutinib therapy. MANAGEMENT AND OUTCOME: We performed a novel 11-step desensitization procedure to ibrutinib allowing us to successfully induce tolerance against IgE-mediated symptoms in this patient. DISCUSSION: As indications for ibrutinib use expand and more patients present with IgE-mediated symptoms, we expect that this protocol will provide benefit for many such patients.

Allergic Reactions Captured by Voluntary Reporting.

BACKGROUND: The epidemiology of hospital adverse reactions (ARs), particularly allergic reactions, or hypersensitivity reactions (HSRs), is poorly defined. To determine priorities for allergy safety in healthcare, we identified and described safety reports of allergic reactions. METHODS: We searched the safety report database of a large academic medical center from April 2006 to March 2016 using 101 complete, truncated, and/or misspelled key words related to allergic symptoms, treatments, and culprits (e.g., medications, foods). Patient and event data were summarized for ARs and two types of ARs, HSRs and side effects/toxicities. RESULTS: Among 9111 key word search-identified events, 876 (10%) were ARs, of which 436 (5%) were HSRs and the remaining 440 (5%) were side effect reactions or toxicities. Whereas the most common HSRs were simple cutaneous reactions (83%), the following severe immediate HSRs were also identified: shortness of breath (16%), anaphylaxis (14%), and angioedema (12%). Most HSRs were caused by drugs (81%), with antibiotics (26%), particularly ß-lactams (11%), and vancomycin (8%), commonly implicated. Other causes of drug HSRs included contrast agents (24%), chemotherapeutics (7%), and opioids (6%). Nondrug HSRs were from blood products (8%), latex (3%), and devices (3%). Food reactions were rarely identified (1%). CONCLUSIONS: We identified ARs, HSRs, and side effects/toxicities, contained in a decade of safety reports at an academic medical center. Allergy safety in the healthcare setting should target approaches to common and severe reactions, with a focus on the safe administration of ß-lactams, vancomycin, contrast agents, chemotherapeutics, and opioids. Priority nondrug HSR culprits include blood products, latex, and devices.

Penicillin Allergy Assessment in Pregnancy: Safety and Impact on Antibiotic Use.

BACKGROUND: Penicillin and other beta-lactam antibiotics are recommended for group B Streptococcus and cesarean section prophylaxis, but approximately 10% of pregnant patients report a penicillin allergy. OBJECTIVE: To assess the safety and impact of penicillin allergy evaluation in pregnant patients. METHODS: In this retrospective study of obstetrician-ordered Allergy/Immunology (AI) electronic consultations (e-consults) from September 20, 2017 through December 31, 2019, we reviewed the electronic health record for e-consult recommendation; patient demographic, obstetric, and allergy histories; and peripartum antibiotic utilization with indication. For patients whose electronic consultation recommended an in-person AI evaluation, testing outcomes were determined, and multivariable logistic regression models were used to compare antibiotic use between patients who did and did not receive an in-person AI evaluation. RESULTS: Of 389 obstetrician-ordered e-consults, 363 (93%) recommended an in-person AI evaluation; of these, 222 (61%) patients received an in-person AI evaluation. Of 220 (99%) patients skin tested, 209 (95%) had their penicillin allergy label safely removed. Compared with patients who did not receive an in-person AI evaluation despite it being recommended (n = 141), patients with in-person AI evaluation (n = 222) had reduced peripartum vancomycin (adjusted odds ratio [aOR], 0.07; 95% CI, 0.01-0.33), clindamycin (aOR, 0.17; 95% CI, 0.08-0.34), and gentamicin (aOR, 0.39; 95% CI, 0.19-0.78) use and increased penicillin (aOR, 18.0; 95% CI, 6.30-51.2) use. The fully AI evaluated patients had increased first-line antibiotic prophylaxis for group B Streptococcus (aOR, 26.9; 95% CI, 6.32-114) and cesarean section (aOR, 1.94; 95% CI, 1.06-3.52). CONCLUSIONS: In a sample of 220 pregnant patients with penicillin allergy histories and in-person AI evaluation, penicillin allergy testing was safe and associated with significantly reduced broad-spectrum antibiotic use and increased first-line beta-lactam antibiotic use.

Conducting Safety Research Safely: A Policy-Based Approach for Conducting Research with Peer Review Protected Material.

A multidisciplinary team developed a policy-based approach that provides guidance for using peer review protected information for safety research while maintaining peer review privilege. The approach includes project approval by an ad hoc review committee, signed confidentiality agreements by investigators and study staff, early removal of case identification numbers, standards for maintaining data security, and publication of aggregate data without data set sharing. By describing this procedure and embedding into an institutional policy on Data for Performance Improvement, the team encourages other institutions to develop similar policies consistent with their state regulations.

Development and Validation of a Deep Learning Model for Detection of Allergic Reactions Using Safety Event Reports Across Hospitals.

Importance: Although critical to patient safety, health care-related allergic reactions are challenging to identify and monitor. Objective: To develop a deep learning model to identify allergic reactions in the free-text narrative of hospital safety reports and evaluate its generalizability, efficiency, productivity, and interpretability. Design, Setting, and Participants: This cross-sectional study analyzed hospital safety reports filed between May 2004 and January 2019 at Brigham and Women's Hospital and between April 2006 and June 2018 at Massachusetts General Hospital in Boston. Training and validating a deep learning model involved extracting safety reports using 101 expert-curated keywords from Massachusetts General Hospital (data set I). The model was then evaluated on 3 data sets: reports without keywords (data set II), reports from a different time frame (data set III), and reports from a different hospital (Brigham and Women's Hospital; data set IV). Statistical analyses were performed between March 1, 2019, and July 18, 2020. Main Outcomes and Measures: The area under the receiver operating characteristic curve and area under the precision-recall curve were used on data set I. The precision at top-k was used on data sets II to IV. Results: A total of 299 028 safety reports with 172 854 patients were included. Of these patients, 86 544 were women (50.1%) and the median (interquartile range [IQR]) age was 59.7 (43.8-71.6) years. The deep learning model achieved an area under the receiver operating characteristic curve of 0.979 (95% CI, 0.973-0.985) and an area under the precision-recall curve of 0.809 (95% CI, 0.773-0.845). The model achieved precisions at the top 100 model-identified cases of 0.930 in data set II, 0.960 in data set III, and 0.990 in data set IV. Compared with the keyword-search approach, the deep learning model reduced the number of cases for manual review by 63.8% and identified 24.2% more cases of confirmed allergic reactions. The model highlighted important words (eg, rash, hives, and Benadryl) in prediction and extended the list of expert-curated keywords through an attention layer. Conclusions and Relevance: This study showed that a deep learning model can accurately and efficiently identify allergic reactions using free-text narratives written by a variety of health care professionals. This model could be used to improve allergy care, potentially enabling real-time event surveillance and guidance for medical errors and system improvement.

Electronic Consultations in Allergy/Immunology.

BACKGROUND: Allergic condition management more often requires allergist guidance than allergy testing; necessary testing may be unavailable at initial drug allergy consultations. Electronic consultations (e-consults) provide expedited, problem-focused, potentially cost-saving care in other medical specialties, but have not yet been studied in Allergy/Immunology. OBJECTIVE: To describe e-consult use at an academic allergy/immunology practice. METHODS: E-consult data (August 10, 2016 through July 31, 2018) and in-person consult data (August 1, 2014 through July 31, 2018) were reviewed to determine consult volume, outcomes, indications, and timing. Referral reasons and wait times were compared with chi-square tests. RESULTS: E-consults grew from 1% to 10% of all new consults, with concurrent growth in in-person consults. Of 306 completed e-consults, 41 (13.4%) made diagnostic, therapeutic, or alternative referral recommendations, with 30 (73%) recommendations followed; 183 (59.8%) patients required an in-person Allergy/Immunology consult, and only 5 (<2%) patients saw an allergist without an e-consult recommendation to do so. E-consults were used more often than in-person consults for adverse drug reactions (66% vs 9%; P < .001), especially penicillin allergy (132, 61% of all e-consults) and immunodeficiency (15% vs 2%; P < .001). Allergists completed e-consults in a median of 11 minutes, with a median turnaround time of 22 hours. E-consult implementation was associated with a decreased median in-person consult wait time (1.5 fewer calendar days; P < .05). CONCLUSIONS: E-consults were increasingly used, particularly for historical adverse drug reactions and immunodeficiency. Implementation of an e-consult program resulted in decreased in-person wait times despite an increase in overall consult volume, supporting this model's ability to provide expedited, problem-focused care.

Drug Reaction with Eosinophilia and Systemic Symptoms (DRESS) Syndrome Identified in the Electronic Health Record Allergy Module.

BACKGROUND: Drug reaction with eosinophilia and systemic symptoms (DRESS) syndrome is a rare but severe hypersensitivity reaction that remains poorly characterized in the United States. OBJECTIVE: To identify and describe DRESS syndrome cases in an integrated health care system using electronic health record (EHR) allergy module free-text searches. METHODS: We identified DRESS syndrome cases with rash, absolute eosinophil count of 500/L or more, organ involvement, and a European Registry of Severe Cutaneous Adverse Reactions to Drugs and Collection of Biological Samples score of 2 or more by reviewing those patients from 1980 to 2016 whose EHR allergic reaction matched DRESS-related key words. Liver injury required alanine aminotransferase level of 100 U/L or more, and renal injury required creatinine increase by 0.5 mg/dL (or 50%) or more. Patient and DRESS characteristics were described, resource use was determined, and cost was estimated. RESULTS: Among 3,162,562 patients with 3,319,387 million allergy entries, 538 reactions matched key words, and 69 were DRESS syndrome cases (prevalence 2.18 per 100,000). Patients with DRESS had liver (42%) or renal (42%) injury; 11 (16%) had both liver and renal injury. Primary drug culprits were antibiotics (74%) (vancomycin [39%], ß-lactams [23%], fluoroquinolones [4%], tetracyclines [4%], and sulfonamides [3%]) and anticonvulsants (20%). Of 65 (94%) hospitalized patients with DRESS, 43 (66%) were hospitalized for DRESS syndrome management with median length of stay of 9 days (interquartile range, 6-17 days) and cost approximately $17,101 per patient. CONCLUSIONS: Using free-text search of the EHR allergy module identified a large US DRESS syndrome cohort. DRESS prevalence was 2.18 per 100,000 patients. Both liver and kidney injury were frequent, and vancomycin was the most common drug culprit. DRESS cases were morbid and resource-intensive.