Antibiotic allergy labels in hospitalized and critically ill adults: A review of current impacts of inaccurate labelling

Abstract

Antibiotic allergy labels (AALs) are reported by approximately 20% of hospitalized patients, yet over 85% will be negative on formal allergy testing. Hospitalized patients with an AAL have inferior patient outcomes, increased colonization with multidrug‐resistant organisms and frequently receive inappropriate antimicrobials. Hospitalized populations have been well studied but, to date, the impact of AALs on patients with critical illness remains less well defined. We review the prevalence and impact of AALs on hospitalized patients, including those in in critical care.

Introduction

Antibiotic allergy labels (AALs) are patient‐reported antibiotic allergies, that may represent an unpredictable immune mediated adverse drug reaction (ADR; e.g. anaphylaxis) or predictable nonimmune mediated (pharmacological) ADR (e.g. gastrointestinal upset). AALs are frequently reported in hospitalized patients 1, 2. Irrespective of whether these labels correctly apply to a true immune mediated reaction or incorrectly to a predictable nonimmune reaction, an AAL has the potential to persist in the medical record 3. AALs lead to the prescription of antibiotics that diverge from guidelines with increased risk of antimicrobial resistance, mortality and morbidity in hospitalized patients 4, 5, 6. AALs increase rates of Clostridium difficile infection, emergence of methicillin‐resistant Staphylococcus aureus, vancomycin‐resistant Enterococcus faecium and multidrug resistant Gram negatives – organisms of particular importance and pathogenicity for the critically ill subgroup of hospitalized patients 7, 8, 9, 10. It has been reported that upwards of 10–20% of hospitalized patients have an AAL, most commonly to a penicillin or β‐lactam, highlighting the enormous burden these AALs present 1, 6, 11, 12, 13.

A focus of the global effort to reduce antimicrobial resistance include the optimization of antimicrobial prescribing through stewardship programmes. Whilst novel antibiotic allergy testing (AAT) programmes have been successfully employed in hospital antimicrobial stewardship (AMS) practices, with subsequent reductions in restricted antibiotic usage noted, the impact of AALs in hospitalized patients, especially in critical care and the intensive care unit (ICU), is unknown 14, 15. We sought to understand the current burden and impact of AALs on the hospitalized and critically ill patient.

Methods

We performed a narrative review, examining literature via the Medline (Ovid) platform (September 2017 and February 2018), utilizing search terms: antibiotic allergy label, antibiotic allergy, antimicrobial allergy, penicillin allergy, β‐lactam allergy, antibiotic hypersensitivity, acute care, critical care, ICU, intensive care, acute medical, high dependency, antimicrobial stewardship, hypersensitivity delayed, hypersensitivity immediate, anti‐infective agents, antibacterial agents, antimicrobial resistance and bacterial resistance. We found 702 articles after limitation to the English language and articles published after 1990. In total, 150 articles were read in full after studies were excluded (case reports, case discussions, exclusively paediatric population or allergy studies that did not focus on the inpatient impacts of AALs). The bibliographies of identified publications were also checked for relevant studies.

Antibiotic allergy: mechanisms and definitions

An AAL can reflect an immunological or non‐immunological ADR and can be categorized into Type A or B according to their aetiology 16. Type A ADRs are predictable, dose‐dependent, pharmacological (non‐immune) effects of a drug, such as nausea and diarrhoea, and account for between 16–80% of AALs 1, 6, 11, 17, 18. Type B ADRs (or true antibiotic allergies) are unpredictable, non‐dose‐dependent and are immunologically mediated. Type B ADRs are subdivided into four categories (I–IV) depending on their immunological mechanism (Table 1) 19. There have been proposed revisions to these classical definitions, as some Type A reactions have been found to be dependent both on dose and genetic factors and the Type B reactions demonstrating predictable dose‐dependency, with some suggesting that a more simplified classification might refer to those reactions that are either on target (Type A) or off target (Type B) 20.

Table 1.

Immunological differentiation of Type B adverse drug reaction (ADRs)

Type B ADR	Timing	Immunological Involvement	Antibiotic Examples	Clinical presentation
Class I	30–60 min (immediate) or 6–48 h (accelerated)	IgE and mast cell	Beta‐lactams, sulfa antimicrobials, macrolides, fluoroquinolones	Pruritis, urticaria, angioedema, bronchospasm, anaphylaxis
Class II	5–72 h (accelerated or delayed)	IgG or IgM, cytotoxic cells (phagocytes, natural killer cells)	Beta‐lactams, sulfa antimicrobials, rifampicin, dapsone, vancomycin	Haemolytic anaemia, neutropenia, thrombocytopenia
Class III	3–72 h (accelerated or delayed)	IgG, antigen–antibody complex	Beta‐lactams (especially amoxicillin and cefaclor), sulfa antimicrobials, minocycline 57	Serum sickness, acute interstitial nephritis, hypersensitivity vasculitis,
Class IV	24–72 h (delayed)	T‐cell mediated IVa (macrophage) IVb (eosinophil) IVc (T Cell) IVd (neutrophil)	Beta‐lactams, sulfa antimicrobials, fluoroquinolones, tetracyclines, macrolides, dapsone, vancomycin, anti‐tuberculosis drugs	Severe cutaneous adverse reactions, erythema multiforme, morbilliform eruptions, SJS, TENS, DRESS

Adapted from Trubiano et al. (with permission) 25, 58, Gell and Coombs 16, Pichler 19. Comparative table of the different type B adverse drug reactions and their immunological components, clinical presentations and time frames. SJS, Stevens–Johnsons syndrome; TEN, toxic epidermal necrosis; DRESS, drug reaction with eosinophilia and systemic symptoms

The impact of AALs on hospitalized patients

North American and Australian studies have estimated that somewhere between 15 and 25% of all hospitalized patients have an AAL, with even higher rates reported in the immunocompromised 21, 22. It has been estimated that 30–40% of hospitalized patients with an AAL do not receive first‐line (preferred) antibiotics 12. Further, that the presence of a penicillin allergy also significantly delays time to first antibiotic dose (236.1 vs. 186.6 min; P = 0.03) 23. MacFadden et al. demonstrated that β‐lactam allergies in hospitalized patients were associated with a number of inferior patient outcomes (increased readmission rate, C. difficile infection, drug reaction or acute kidney injury) when patients were unable to receive therapy with a first‐line β‐lactam 24. Multiple reports have found that AALs are associated with a prescriber divergence from antibiotic guidelines, lower use of β‐lactams, greater use of broad spectrum antibiotics, prolonged use of intravenous compared to oral antibiotics, inferior patient outcomes and greater risk of antimicrobial resistance emergence (Table 2) 1, 4, 5, 7, 11, 13, 18, 25, 26. A review by Wu et al. found that seven of eight studies of hospitalised patients had various negative clinical, prescription and antibiotic utilization outcomes associated with AALs 2.

Table 2.

Comparison of studies looking at the impact of antibiotic allergy labels (AALs) on patient outcomes

Study (Year)	Study Setting	Study design	Included ICU patients (Y/N)	Sample Size (n = patients)	AAL Prevalence (%)	Increased duration of ABx	Increased number of ABx	Reduced β‐lactam use	Increased restricted ABx use	Increased vancomycin use	Increased Quinolones use	Increased non‐oral ABx administration	Increased length of stay	Increased Mortality	Increased readmissions	Increased ICU admission	Increased Cost	Reduced ABx appropriateness
Lee et al. (2000) 32	Tertiary hospital (USA)	Retrospective single‐centre	Yes	1893	25	a	a	✓	✓	✓	✓	a	a	a	a	a	a	a
Charneski et al. (2011) 6	Medical wards (USA)	Retrospective single‐centre, cohort	Yes	11 872	11.2	a	✓	a	a	a	a	a	✓	✓	✗	✓	a	a
Picard et al. (2013) 46	ICU, CCU, medical wards (CAN)	Retrospective single‐centre, cross‐sectional	Yes	1738	9.9	a	a	a	✓	✓	✓	a	a	a	a	a	✓	a
Macy et al. (2014) 7	Californian hospitals (USA)	Retrospective multicentre, matched cohort	Yes	51 582	11.2	a	✓	a	✓	✓	✓	a	✓	✗	a	a	✓	a
Trubiano et al. (2015) 18	Tertiary hospital (AUS)	Retrospective single‐centre point prevalence	Yes	509	25	✓	a	✓	✗	a	a	✓	a	a	a	a	a	✗
Trubiano et al. (2016) 11	Medical wards (AUS)	Retrospective multicentre, cross‐sectional	No	453	24	a	a	✓	✓	a	✓	a	a	a	a	a	a	a
Trubiano et al. (2015) 17	Oncology wards (AUS)	Retrospective single‐centre	No	198	23	✓	✓	a	✓	✗	✓	a	✗	✗	✓	a	a	✓
Trubiano et al. (2016) 1	Tertiary hospital (AUS)	Retrospective multicentre point prevalence	Yes5	21 031	18	a	✓	✓	✓	✓	✓	✗	a	a	a	a	a	✓
Knezevic et al. (2016) 59	Tertiary hospital (AUS)	Retrospective single‐centre, cross‐sectional	No	687	18	a	a	✓	a	a	a	a	✗	a	✓	✗	✗	a
Lutomski et al. (2008) 60	Tertiary hospital (USA)	Retrospective single‐centre	No	300	NA	a	a	a	✓	✓	✓	a	a	a	a	a	a	a
Van Dijk et al. (2016) 61	Tertiary hospital (NET)	Prospective single‐centre, matched cohort	No	17 959	5.6b	a	✓	a	✓	a	✓	a	✗	✗	✓	a	a	a
Conway et al. (2017) 23	Veterans hospital ED (USA)	Retrospective single centre, study & interview	Yesd	403c	14.1	✗b	✗	a	✓	a	✓	a	✗	a	✗	a	a	a
Huang et al. (2018) 22	Oncology wards (USA)	Retrospective cohort	No	4671	35.1	✓	✓	✓	a	a	a	a	✓	✓	✓	✗	✓	a
Torda et al. (2018) 62	Tertiary hospital (AUS)	Prospective single‐centre, cross‐sectional	No	3855	14.4	a	a	a	a	a	a	a	a	a	a	a	a	a

Direct comparison of observational articles providing insight of AAL impact on patient outcomes between various studied hospitalized and non‐hospitalized populations. Outcomes include effect on length of stay, mortality, readmission rate, type of antibiotic use and compliance to guidelines. No studies directly looked at AAL prevalence or impacts on ICU however Trubiano et al. 1 found an AAL prevalence of 24% for acute care specifically. Note that Huang et al. 22 found 35.1% of oncology patients had allergy to a non‐monobactam β‐lactam but classified patients with β‐lactam plus other antibiotic allergies as unexposed for analysis of outcomes. AUS, Australia; CAN, Canada; NET, The Netherlands; ICU, intensive care unit; ABx, antibiotic. ✓ Study looked for this finding and found a statistically significant result (P < 0.05). ✗ Study looked for this finding and did not find a statistically significant result (P > 0.05)

^aStudy did not look for this finding

^bStudy found increased time to first antibiotic administration

^cStudy population included only those who presented with an infection requiring antibiotics

^dThe ICU population was included and analysed separately

The burden of AALs on the critically ill

Recent work by Trubiano et al. identified that 24% of patients in acute care units, including ICU, had an AAL 1, while more specifically, up to 8–12% of patients in the emergency department and 8% in the ICU have a penicillin AAL 1, 12, 27. The reason for this high prevalence of AALs is probably multifactorial and includes those accumulated in childhood during a concurrent viral illness (e.g. viral exanthems or urticaria) 28, mislabelling of predictable antibiotic side effects as allergies (e.g. nausea, diarrhoea) and following high‐antibiotic usage in the acute care setting 3, 23, 29. Further compounding confusion surrounding AALs is that allergies and ADRs are often recorded within the same area of a medical record chart without allowance for differentiation of reaction types 30 and up to 36% of electronic records lacking an adequate description of the AAL 1, 6, 31, 32. These challenges are of particular importance in critically ill hospitalized patients where treatment is often time critical and clinicians may not have the opportunity to take a detailed history from the critically unwell patient or review a patient file to clarify the history of the ADR. This has been reflected in a 50‐min time‐delay in antibiotic administration seen in those reporting a penicillin allergy 23.

A study in hospitalized oncology patients, often critically ill, found that there was an increase in total antibiotic usage per admission, increased restricted antibiotic use, longer antibiotic duration, higher readmission rates and poorer concordance with antibiotic guidelines 17. Charneski et al. studied over 11 000 hospitalized patients and found that AALs were associated with increased length of stay (additional 1.2 days), higher mortality (adjusted odds ratio 1.6) and increased rate of ICU admission (adjusted odds ratio 1.4; Table 2) 6. Penicillin AALs are also associated with an increased rate of surgical site infections, relevant to surgical ICUs 33. Most studies looking at the prevalence of AALs and their associated outcomes are retrospective (Table 2) and literature describing the impacts of AAL on hospital cohorts with a high proportion of critically ill, such as ICU patients, is limited 34.

From a health economic perspective, AALs are associated with increased treatment cost – up to 63% – probably due to receipt of restricted antimicrobials and extended hospital length of stay 7, 35, 36. Costs are also increased in ICU patients who acquire nosocomial infections during their inpatient stay 10. Chagoya et al. demonstrated that inpatient testing of penicillin allergy with skin testing in the ICU facilitated more appropriate antibiotic selection without an increase in hospitalization cost 37.

Pathways to AAL de‐labelling in hospitalized patients: choice of testing and safety

In‐vivo confirmation of allergy status

Skin prick testing (SPT), intradermal testing (IDT) and oral provocation (i.e. oral rechallenge) are the most widely accepted methods of establishing an allergy phenotype 38. Whilst frequently undertaken in the outpatient setting, there is increased interest in employing these methods for hospitalized patients 5, 14, 15, 39. A 2017 systematic review of inpatient and ICU populations by Sacco et al. found that inpatient allergy testing was safe, efficacious and associated with subsequent improved antibiotic utilisation 40. Sacco et al. found that the population‐weighted mean for a negative SPT was 95.1%, similar to the studies listed in Table 3 40. There is evidence that critically ill patients or those taking systemic glucocorticoids have the high rates of negative SPT/IDT or a negative histamine result, raising concern of false negative results and highlighting the need for further validation in ICU cohorts 41, 42, 43. Whilst SPT/IDT is being increasingly performed by non‐allergists, the role of the intensivist, high‐dependency clinician or generalist is ill defined 14, 31, 43. A summary of the use of β‐lactam allergy testing and utilization in a variety of hospitalized patients, including the critically ill, is provided in Table 3.

Table 3.

Summary of studies comparing skin prick testing for penicillin adverse drug reactions

Study (year)	Study setting	Sample Size (n = patients)	Staff Performing SPT/IDT	Included ICU / ED / acutely unwell patients	Negative SPT/IDT (%)	Indeterminant SPT/IDT (%)	Included OP	Negative rate after OP (%)	Systemic adverse reactions	Recommendation to use PCN	Analysis of antibiotic use after SPT/IDT	Patients who changed antibiotics	Increased narrow spectrum antibiotics	Decreased restricted antibiotics
Harris et al. (1999) 63	Hospital‐wide & presurgical	44	Not provided	Unknown	86.0	7.0	No	NA	0	86%	Yes	95%	✓b	✓a
Arroliga et al. (2000) 12	Intensive care unit	24	Allergy service	Yes ICU, d	95.0	4.8	No	NA	0	48%	No	48%	✓b	✓b
Arroliga et al. (2003) 26	Intensive care unit	96	Allergist supervision	Yes ICU, d	88.5	10.4	No	NA	0	12–81% f	Yes	39%	✓a	c
Nadarajah et al. (2005) 44	Hospital‐wide	101	Not provided	Yese	91.1	3.9	No	NA	0	49%	Yes	49%	✓b	✓b
Macy (2006) 64	Pregnant inpatients and allergy clinic outpatients	47	Not provided	Unknownd	93.6	0.0	No	NA	6	89%	Yes	89%	✓b	c
De Real et al. (2007) 42	Hospital‐wide	596	Allergy Service	Yes ICU, e	88.4	3.4	Yes	NA	5	49%	Yes	55%	✓b	✓b
Raja et al. (2009) 36	Emergency department	150	ED physicians	Yes ED, d	91.3	0.0	No	NA	0	NA	No	c	c	c
Rimawi and Mazer (2014) 47	Intensive care unit	100	Various staff	Yes ICU, e	100.0	0.0	Yes	100	NA	100%	Yes	100%	✓b	✓b
Heil et al. (2016) 43	Inpatient allergy clinic	90	ID fellow	Yes ICU, e	96.0	12.0	Yes	NA	0	80%	Yes	84%	✓b	✓a
Marwood et al. (2017) 27	Emergency department	100	ED physicians	Yes ED, g	81.0	NA	Yes	NA	0	81%	Yes	c	✓a	c

The table compares articles that confirmed allergy status of patients with an antibiotic allergy label using penicillin SPT/IDT ± OP. Negative rates varied between 88.4–95% of patients tested, similar to recent reviews. Studies that occurred in acute care settings (ED or ICU) are bolded for ease of reading. Studies used different combinations of penicillin derivatives for SPT/IDT as noted here (where available): Arroliga, Rimawi and Heil used benzylpenicilloyl polylysine (BPP) and penicillin G ‐ Heil used amoxycillin as oral provocation. Macy used BPP, penicillin, amoxicillin, penilloate, and penicilloate. Raja used penicillin major and minor determinants. Bourke used BPP, minor determinant mix (MDM) (sodium benzylpenicillin, benzylpenicilloic acid, sodium benzylpenicilloate), benzylpenicillin, amoxicillin, cefazolin and ceftriaxone plus/minus other penicillins, cephalosporins, or carbapenems was performed if clinically indicated. Marwood used BPP, MDM and amoxycillin. Harris used BPP and MDM (sodium penicilloate, sodium penilloate, and sodium benzylpenicillin). SPT, skin prick testing; IDT, intradermal testing; OP, oral provocation; ED, emergency department; ID, infectious disease physician; PCN, penicillin; NA, not available. ✓ Study looked for this outcome;

^astudy found a statistically significant result (P < 0.05);

^bNo statistical analysis conducted however clinically relevant results were found;

^cStudy did not look for this outcome;

^dFurther patient demographics unknown;

^estudy included possibly acutely unwell patients (e.g. endocarditis, sepsis, bacteraemia);

^fstudy found 12% reduction for prophylactic and 18% reduction for therapeutic indications for antibiotic use;

^gstudy excluded severely unwell patients and those taking corticosteroid medications

Pathways to AAL de‐labelling: impact on AAL burden and antibiotic prescribing

The studies listed in Table 3 show a negative SPT/IDT result in 88–95% of patients tested, suggesting that only 5–11% of patients have a positive immunological reaction to penicillin derivatives 5, 12, 26, 36, 42, 44. Therefore only 1–2% of hospitalized patients are truly antibiotic allergic compared with the 10–20% of patients that report an allergy during hospital admission.

The SPT/IDT studies promoted an antibiotic change in approximately 70% of those tested hospitalized patients, with greatest numbers of antibiotic changes in orthopaedic or critically ill patients (Table 3) 4, 40, 42, 45. This is key, as Picard et al. 46 found that most patients with an indication for β‐lactam antibiotics were the patients in ICU and De Real et al. 42 found that 70% of their critically ill patients with an incorrect penicillin AAL had their antibiotic changed after SPT/IDT. SPT/IDT allows for the deployment of more appropriate antibiotics, reduced restricted antibiotic usage and lower risk of acquiring a drug resistant infection 12, 26, 36, 40, 44. Arroliga et al. 12, 26 and Rimawi and Mazer 47 demonstrated that SPT/IDT in the ICU allowed for increased guideline preferred antibiotic prescribing, crucial when considering the treatment of sepsis in the hospitalized and critically ill 48. Arroliga et al. also found that the rate of adverse reaction following SPT/IDT was not greater than the general population 26. None of the studies investigating penicillin allergies using SPT/IDT reported any serious adverse reactions with subsequent exposure to β‐lactams – with just five of 596 patients noting nonsevere dermatological type reactions (hives, nonspecific rashes, flushing, urticaria) 42. Whilst the potential benefit to critically ill patients following AAT is significant, larger studies looking at feasibility and impact on patient outcomes are required. A proposed algorithm for the assessment and management of patients with antibiotic allergies in the ICU, extrapolated from available research and available AAT literature is provided in Figure 1. This proposed algorithm requires future validation in the critical care setting.

Figure 1.

Proposed algorithm for assessment and management of patients with antibiotic allergy label in ICU. A proposed algorithm to assist with the management of penicillin antibiotic allergy labels that are admitted to the intensive care unit who require antibiotics 13, 41, 65. ICU, intensive care unit; MPE, maculopapular exanthem; ADR, adverse drug reaction; IV, intravenous; SJS, Stevens–Johnston syndrome; TEN, toxic epidermal necrosis; AAT, antibiotic allergy testing. ^Cautious of drugs with antihistamine properties, including antidepressants and cyclizine, *Refer to allergy testing guidelines for drug selection, doses and administration technique

Hospital antimicrobial stewardship and antibiotic allergies

Inappropriate antibiotic use has been identified as a strong mediator of antimicrobial resistance 49. Furthermore, a penicillin allergy is strongly associated with the acquisition of C. difficile infection, presumably related to the use of more broad‐spectrum antibiotics that disrupt the normal enteric flora 50. Infections with resistant organisms, particularly from multidrug resistant Gram negatives, are associated with excess morbidity and mortality 9. Recently, the assessment and testing of patients with AALs has been recognized as a novel antimicrobial stewardship tool to add to the fight against antimicrobial resistance 51, 52. In an Australian context, Trubiano and colleagues 15 have extensively studied the impact of AALs on antibiotic appropriateness and the role of AAT and multidisciplinary de‐labelling programmes in AMS (AAT‐AMS), to increase the use of appropriate and narrow spectrum antibiotic treatments. However, the incorporation of AAT‐AMS into hospitalized settings with a predominance of critically ill patients is yet to be formally evaluated 15.

Vitrat et al. recently reviewed the impact of AMS programmes on antibiotic utilization in the ICU and antimicrobial resistance and found that AMS was effective at increasing the appropriateness of antibiotic use and improving patient outcomes through improved guideline adherence, reduction in mortality, reduced length of stay, reduced drug resistant infections and reduced cost 53. To the authors knowledge, there have not been any published studies on the use of AAT by AMS services in the ICU setting but there is reason to hope that additional benefit may be gained from incorporating such AAT strategies into existing AMS programmes. The variety of AMS programmes already proven to be effective in the ICU population include antibiotic restriction, preauthorization, education and de‐escalation; however, an as yet unresearched avenue is the effect of de‐labelling inaccurate AALs 53, 54.

Conclusion

AALs are associated with increased hospital readmission rates, hospital costs, ICU admission and increased ICU length of stay. Furthermore, AALs negatively impact clinician antibiotic choices and are associated with poorer outcomes for hospitalized patients. The existing literature demonstrates the extent of the burden of AALs in hospitalized patients, particularly those critically unwell and in the ICU. Further work must be done to define the extent of AAL phenotypes in the setting of critical illness as part of a multifaceted approach to reducing antimicrobial resistance and optimizing patient outcomes.

Data demonstrating the utility of SPT/IDT to define AAL in the critically ill continue to mount. While numbers are small, they show that AAT may lead to significant changes in appropriate and preferred antibiotic usage. Concerns still exist regarding the potential for a false negative result in the critically ill and those receiving vasopressor support or immunosuppression, patients who are over‐represented in the ICU. While further work is necessary to determine the efficacy and safety of SPT/IDT in a range of illness settings in hospitalized, there is reason to hope that these techniques will evolve into a key component of antibiotic stewardship programmes in the ICU and lead to improved short‐ and long‐term patient outcomes in the future 55, 56.

Competing Interests

There are no competing interests to declare.

J.A.T. is supported by postgraduate scholarships from the National Health and Medical Research Council and The National Centre for Infections in Cancer, Peter MacCallum Cancer Centre, Melbourne, Australia.

Moran, R. , Devchand, M. , Smibert, O. , and Trubiano, J. A. (2019) Antibiotic allergy labels in hospitalized and critically ill adults: A review of current impacts of inaccurate labelling. Br J Clin Pharmacol, 85: 492–500. 10.1111/bcp.13830.