Tackling inpatient penicillin allergies: Assessing tools for antimicrobial stewardship

Abstract

Background

Reported penicillin allergy rarely reflects penicillin intolerance. Failure to address inpatient penicillin allergies results in more broad-spectrum antibiotic use, treatment failures, and adverse drug events.

Objective

We aimed to determine the optimal approach to penicillin allergies among medical inpatients.

Methods

We evaluated internal medicine inpatients reporting penicillin allergy in three periods: (1) standard of care (SOC), (2) penicillin skin testing (ST), and (3) computerized guideline application with decision support (APP). The primary outcome was use of a penicillin or cephalosporin, comparing interventions to SOC using multivariable logistic regression.

Results

There were 625 patients: SOC 148, ST 278, and APP 199. Of 278 ST patients, 179 (64%) were skin test eligible; 43 (24%) received testing and none were allergic. In the APP period, there were 292 unique website views; 112 users (38%) completed clinical decision support. While ST period patients did not have an increased odds of penicillin or cephalosporin use overall (aOR 1.3 [95% CI 0.8, 2.0]), we observed a significant increased odds of penicillin or cephalosporin use overall in the APP period (aOR 1.8 [95% CI 1.1, 2.9]), and in a per protocol analysis of the skin tested subset (aOR 5.7 [95% CI 2.6, 12.5]).

Conclusions

Both the computerized guideline with decision support and penicillin skin testing – when completed – increased use of penicillin and cephalosporin antibiotics among inpatients reporting penicillin allergy. While the skin tested subset showed an almost 6-fold impact, the computerized guideline significantly increased penicillin or cephalosporin use overall nearly 2-fold and was readily implemented.

INTRODUCTION

Penicillin allergy is reported in up to 15% of inpatients and is associated with increased use of alternative antibiotics, including vancomycin, clindamycin, aminoglycosides, and aztreonam.^1–4 Compared to beta-lactam antibiotics, these drugs are less effective in some clinical circumstances,^5–8 more toxic,^4,9 more costly,^10,11 and generally cover a broader antimicrobial spectrum. When a beta-lactam antibiotic is the preferred inpatient antibiotic, but not administered due to alleged allergy, patients experience more treatment failures and adverse events.^4,8 Patients reporting penicillin allergy have increased odds of antibiotic resistant organisms, such as methicillin-resistant Staphylococcus aureus and vancomycin-resistant Enterococcus.¹²

Despite a reported penicillin allergy, more than 95% of patients evaluated for such allergy are found penicillin and cephalosporin tolerant.^10,12–16 Therefore, active attention to clarifying old and inaccurate penicillin allergies is supported by various U.S. guidelines and agencies as an important feature of antimicrobial stewardship.^17–20 Because the optimal approach to the evaluation and management of inpatient penicillin allergy is unknown, yet impacts a substantial number of patients per year, we implemented and assessed two healthcare delivery system innovations to improve antibiotic choice among medical inpatients reporting a history of penicillin allergy.

METHODS

Design Overview

We conducted a quasi-experimental study evaluating prospectively identified cohorts of internal medicine inpatients at the Brigham and Women’s Hospital (BWH). We sequentially evaluated three strategies: (1) BWH standard of care (SOC); (2) history-appropriate penicillin skin testing (ST), a process-based innovation; and (3) a computerized guideline application with clinical decision support, a technology-based innovation (APP). We compared antibiotic use in the intervention periods to the SOC period. This study was approved by the Partners Human Research Committee.

Study Population

A daily electronic tracker identified medical inpatients with a history of penicillin allergy prescribed one or more doses of any antimicrobial in all periods (Fig 1). Patient readmissions and patients not meriting treatment of a presumed infection were excluded. The latter exclusion comprised patients who did not receive therapeutic antibiotics in the first seven days of hospitalization, and those who received < 48 hours of antibiotic therapy, accounting for both discharge antibiotics and amended dosing frequency associated with renal dosing.²¹

Fig 1.

This flow chart demonstrates patient cohort identification. The care redesign team was notified prospectively if patients met initial criteria, including (1) internal medicine inpatient, (2) reported penicillin allergy, and (3) prescribed one or more antibiotic doses. From this cohort, we excluded readmissions and patients whose admission was not for treatment of an infection. Of 625 patients meeting these criteria, 148 were admitted in the standard of care period, 278 were admitted in the penicillin skin testing period, and 199 were admitted in the computerized guideline period.

Study Periods

BWH Standard of Care (SOC)

SOC was the comparison period when no active intervention was performed. SOC patient data were collected from June 9, 2014 through November 5, 2014. As an academic, tertiary care facility, BWH has an antibiotic stewardship program that restricts some broad-spectrum and costly antibiotics (e.g., linezolid, daptomycin). BWH also has a drug allergy program with inpatient Allergy/Immunology consultation and 24-hour on-call services. During SOC, all penicillin skin testing and test dose challenge²² procedures were performed only when referred by the primary team and deemed appropriate after Allergy/Immunology consultation (eMethods 1, Fig E1a).

Penicillin Skin Testing (ST)

The ST period began November 12, 2014, and continued through June 30, 2015. During the ST period, all tracker-identified patients were screened by the care redesign team for skin test eligibility. Patients ineligible for skin testing included patients with penicillin intolerances (e.g., gastrointestinal upset), patients taking medications that interfered with skin testing (e.g., antihistamines), and patients with multiple beta-lactam allergies, penicillin anaphylaxis in the last five years, or a type II–IV hypersensitivity reaction²³ to penicillin. Skin testing was routinely intended for all skin test eligible patients, but required permission from the primary team, coordination of skin testing using a moonlighting pool of allergy trainees and nurses, and patient consent (eMethods 1, Fig E1b). Patients with both negative penicillin skin testing and tolerance of an oral amoxicillin test dose were deemed not allergic. The primary medical team and the patients were updated regarding changes in allergy status.

Computerized Guideline Application with Clinical Decision Support (APP)

After a five month study break due to a hospital-wide electronic health record conversion,²⁴ the APP period ran from November 20, 2015 through June 13, 2016 (Fig E1c). A clinical pathway that guided beta-lactam antibiotic use in patients with listed penicillin allergy was previously developed, implemented, and assessed at an academic hospital affiliate (eMethods 2).^5,25–27 The guideline empowered inpatient providers to group allergic reactions into hypersensitivity type, then recommended if and how specific beta-lactam antibiotics be used (i.e., very low risk: full doses; low risk: test doses; medium to high risk: Allergy/Immunology consultation; serious type II–IV hypersensitivity reactions: avoidance).

The previously studied pathway was adapted into a computerized guideline,^28,29 a mobile-friendly website with optional clinical decision support, functionally similar to a smartphone application (eMethods 3). Due to the coincident electronic health record conversion at BWH, the computerized guideline was not integrated into the electronic health record, but remained a distinct clinical workflow. The guideline was accessible at any BWH desktop computer or mobile device on the secure intranet. Providers could access the pathway figures directly from the website and/or login to use clinical decision support. After decision support computed the patient’s likely allergic reaction type, it stratified the reaction into a risk category and displayed recommendations for further action (Fig E2). The website housed additional educational information and provider videos.

Data Collection

All patient data were collected from the electronic health record, with duplicative entry, initially by research assistants (NP, AEN), followed by internal medicine housestaff. Data were entered and maintained using Research Electronic Data Capture hosted at Partners HealthCare.³⁰

Demographic Characteristics

Collected patient data included age, sex, race, admission date, discharge date, admission diagnosis, allergy history, intensive care unit stay and duration time, Infectious Diseases consultation, Allergy/Immunology consultation, history of colonization or infection with methicillin-resistant Staphylococcus aureus or vancomycin-resistant Enterococcus, renal disease, and overall length of stay. Two board-certified internists and allergists/immunologists (KGB, PGW) determined which admission diagnoses were related to an infection as well as which penicillin allergies were intolerances.

Intervention Uptake

In the ST period, we determined the frequency with which eligible patients completed skin testing. In the APP period, we tracked usage through reports from Google Analytics Solutions (website views) and clinical decision support responses.

Outcomes

The primary outcome was use of formulary unrestricted penicillins or cephalosporins (Table E2). Penicillin and cephalosporin use was identified through inpatient antibiotic administrations. Secondary outcomes included the proportion of patients discharged on a penicillin or cephalosporin antibiotic, inpatient use of alternative antibiotics, and resultant adverse drug reactions.

Penicillins and cephalosporins on BWH formulary were included; cephalosporins were grouped by generation for analysis. Because of the intent to improve antibiotic choice, we excluded penicillins and cephalosporins historically restricted by BWH’s antibiotic stewardship program, including piperacillin-tazobactam, ceftaroline, ceftolozane-tazobactam, and ceftazidime-avibactam. Alternative antibiotics included drugs that are potentially more toxic, less effective, more costly, and/or more broad-spectrum: vancomycin, clindamycin, daptomycin, linezolid, carbapenems, aztreonam, and aminoglycosides.

Balance Measure

We analyzed macrolide antibiotic utilization as a balance measure, given its use would not be expected to change as a result of interventions related to penicillin allergy evaluation.

Statistical Analysis

Study period duration and sample size were estimated in advance (eMethods 4). Pairwise study group comparisons were specified a priori. While the primary analysis between periods was an intention-to-treat analysis, we performed an additional secondary (per protocol) analysis for the ST period (ST-PP) to estimate the potential impact of this strategy if it were fully implemented. Demographic data were reported as medians with quartiles or as frequencies with percentages. Fisher’s exact tests were used to compare study groups for binary variables, and Wilcoxon rank sum tests were used for continuous variables. Racial differences among study groups were tested using the Freeman-Halton test. Logistic regression was used to test group differences, with unadjusted and adjusted odds ratios (ORs) and 95% confidence intervals (CI) presented. Selection of variables to include in multivariable models involved prior knowledge of association with outcome, and imbalance across the study duration. Nominal p-values were reported in this context of three a priori non-independent pair-wise group comparisons, considering 5% two-sided alpha. Statistical analyses were conducted using SAS 9.4 (SAS Institute Inc., Cary, NC, USA).

RESULTS

Cohort Characteristics

Of 1,000 medicine inpatients with a reported penicillin allergy prescribed one or more antibiotic dose, there were 780 unique patients and 625 unique patients admitted with a presumed infection across three periods: SOC 148, ST 278, and APP 199 (Fig 1).

Patient characteristics were largely balanced across periods, with a few exceptions (Table I). Compared to the SOC period, there were fewer females in the subset of patients skin tested (i.e., skin test per protocol analysis [ST-PP]), more patients with an admission diagnosis of infection in the ST and APP periods, more frequent Allergy/Immunology consultation in the ST period and in the ST-PP analysis, more frequent methicillin-resistant Staphylococcus aureus colonization in the ST and APP periods, and more reported cephalosporin allergies in the ST and APP periods.

Table I.

Cohort characteristics of medicine patients with reported penicillin allergy, requiring antibiotics over three periods (n= 625).

	Standard of Care (n=148)	Penicillin Skin Testing				Computerized Guideline/APP (n=199)
		All (n=278)		Per Protocol (n=43)
	No (%)	No (%)	p value*	No (%)	p value*	No (%)	p value*
Age (Med, IQR)	64 [48,79]	68 [54, 80]	0.09	71 [52, 77]	0.66	65 [52, 76]	0.96
Female sex	97 (66)	173 (62)	0.53	21 (49)	0.05	128 (64)	0.82
Race			0.14		0.01		0.72
White	109 (77)	183(69)		25 (61)		143 (73)
Black	23 (16)	46 (17)		6 (15)		36 (18)
Hispanic	6 (4)	26 (10)		7 (17)		10 (5)
Other	3 (2)	11 (4)		3 (7)		8 (4)
Admission diagnosis of infection	72 (49)	163 (59)	0.05	25 (58)	0.30	124 (62)	0.01
Reported penicillin allergy history†
Rash or hives	79 (53)	149 (54)	>0.99	28 (65)	0.22	108 (54)	0.91
Angioedema‡	17 (11)	27 (10)	0.62	11 (26)	0.03	14 (7)	0.18
Anaphylaxis	12 (8)	22 (8)	>0.99	4 (9)	0.76	17 (9)	>0.99
Shortness of breath|	2 (1)	7 (3)	0.73	2 (5)	0.22	6 (3)	0.49
Unknown history	26 (18)	65 (23)	0.17	6 (14)	0.65	35 (18)	>0.99
Intolerance only	11 (7)	15 (5)	0.40	1 (2)	0.30	16 (8)	>0.99
ICU Stay > 24 hours	23 (16)	37 (13)	0.56	3 (7)	0.21	27 (14)	0.64
Infectious Diseases consultation	43 (29)	78 (28)	0.82	11 (26)	0.71	55 (28)	0.81
Allergy/Immunology consultation	6 (4)	21 (8)	0.21	6 (14)	0.03	8 (4)	>0.99
Penicillin skin test	2 (1)	43 (16)	<0.001	43 (100)	<0.001	5 (3)	0.70
Resistant organism history
MRSA colonization	9 (6)	36 (13)	0.03	1 (2)	0.46	32 (16)	0.004
VRE colonization	9 (6)	25 (9)	0.35	1 (2)	0.46	15 (8)	0.67
Renal disease	45 (30)	96 (35)	0.45	13 (30)	>0.99	74 (37)	0.21
Other reported antibiotic allergies
Cephalosporins	1 (0.7)	42 (15)	<0.001	3 (7)	0.04	35 (18)	<0.001
Sulfonamides	30 (20)	70 (25)	0.28	8 (19)	>0.99	54 (27)	0.16
Fluoroquinolones	21 (14)	28 (10)	0.21	3 (7)	0.30	31 (16)	0.76
Vancomycin	7 (5)	14 (5)	>0.99	0 (0)	0.35	12 (6)	0.64
Macrolides	18 (12)	26 (9)	0.40	3 (7)	0.42	20 (10)	0.60
Length of stay	5 [3,8]	5 [3,8]	0.61	5 [4,8]	0.91	5 [3,9]	0.61

^*Fisher’s exact test or Wilcoxon Rank Sum test between innovations and standard of care.

^†Records may contain more than one description or reaction of clinical feature of drug allergy history.

^‡Includes swelling.

^|Includes bronchospasm.

^**Median time from patient admission to penicillin skin test was 2.0 days [IQR 1.0, 3.0 days] in the Skin Testing period and 3.0 days [IQR 2.0, 4.0 days] in the Computerized Guideline/APP period. Neither median times were significantly different from the median time in the Standard of Care period (2.0 days [IQR 1.0, 3.0 days], p=0.80 and p=0.48 respectively).

Abbreviations: ICU, intensive care unit; MRSA, methicillin-resistant Staphylococcus aureus; VRE, vancomycin-resistant Enterococcus

Intervention Uptake

Among 278 patients in the ST period, 179 (64%) were skin test eligible. Of 43 (24%) patients skin tested, none were allergic (Fig 2). There were 2,802 webpage views in the APP period, with 292 unique page views lasting 26 seconds on average (Table E1). There were 112 unique users who completed decision support.

Fig 2.

This flow chart includes all penicillin skin testing period patients. Of 278 patients in the skin testing period, 179 patients were eligible for skin testing. Of the 179 patients for whom skin testing was intended, 43 patients completed penicillin skin testing, and none were allergic.

Primary Outcome, Inpatient Penicillin or Cephalosporin Use

Of 148 patients in the SOC period, 56 (38%) were treated with a penicillin or cephalosporin (Table II). Of 278 patients in ST period, 116 (42%) were treated with a penicillin or cephalosporin antibiotic. Of 43 ST-PP patients, 31 (72%) of patients were treated with a penicillin or cephalosporin antibiotic. Of 199 patients in the APP period, 99 (50%) were treated with a penicillin or cephalosporin antibiotic; treated patients received the penicillins and/or cephalosporins after negative penicillin skin testing (n=5, 5%) or directly with a test dose or full dose challenge (n=94, 95%).

Table II.

The impact of interventions on antibiotic use over three periods.

	Standard of Care (n=148)	Penicillin Skin Testing				Computerized Guideline/APP (n=199)
		All (n=278)		Per Protocol (n=43)
	No (%)	No (%)	p value*	No (%)	p value*	No (%)	p value*
Penicillin and cephalosporin antibiotics†
Inpatient use
Penicillins or cephalosporins	56 (38)‡	116 (42)§	0.47	31 (72)|	<0.001	99 (50)¶	0.03
Penicillins	7 (5)	14 (5)	>0.99	6 (14)	0.08	11 (6)	0.81
Cephalosporins (1st or 2nd)	6 (4)	12 (4)	>0.99	5(12)	0.07	13 (7)	0.35
Cephalosporins (3rd or 4th)	47 (32)	102 (37)	0.34	27 (63)	<0.001	86 (43)	0.03
Discharge Use
Penicillins or cephalosporins	24 (16)	33 (12)	0.23	11 (26)	0.18	34 (17)	0.88
Penicillins	5 (3)	9 (3)	>0.99	5 (12)	0.05	7 (4)	>0.99
Cephalosporins (1st or 2nd)	5 (3)	8 (3)	0.77	3 (7)	0.38	9 (5)	0.78
Cephalosporins (3rd or 4th)	15 (10)	16 (6)	0.12	3 (7)	0.77	18 (9)	0.85
Alternative antibiotics**
Inpatient use	74 (50)	148 (53)	0.54	26 (60)	0.30	109 (55)	0.39

^*Fisher’s exact test p values compare the intervention periods to the standard of care period.

^†Excludes piperacillin-tazobactam, ceftaroline, ceftolozane-tazobactam and ceftazidime-avibactam (see methods).

^‡Penicillin or cephalosporin-treated patients received a median of 5.5 [IQR 3.0, 10.5] doses.

^§Penicillin or cephalosporin-treated patients received a median of 5.0 [IQR 2.0, 7.0] doses.

^|Penicillin or cephalosporin-treated patients received a median of 5.0 [IQR 2.0, 11.0] doses.

^¶Penicillin or cephalosporin-treated patients received a median of 6.0 [IQR 2.0, 8.0] doses.

^**Includes vancomycin, clindamycin, aztreonam, daptomycin, linezolid, carbapenems, and aminoglycosides (see methods).

In univariable analyses compared to SOC, penicillin or cephalosporin use was similar in the ST period overall (p=0.44), but greater in the ST-PP analysis (p<0.001) and overall in the APP period (p=0.03). In the multivariable logistic regression model, patients in the ST period did not have a significant increased odds of receiving a penicillin or cephalosporin (aOR 1.3 [95% CI 0.8, 2.0], Table III). ST-PP patients had increased odds of receiving a penicillin or cephalosporin, with an adjusted OR 5.7 [95% CI 2.6, 12.5]. APP period patients had increased odds of receiving a penicillin or cephalosporin, with an adjusted OR 1.8 [95% CI 1.1, 2.9].

Table III.

The impact of innovations on use of penicillin or cephalosporin antibiotics in unadjusted and adjusted analyses.

	Unadjusted Analysis*			Adjusted Analysis*
	Skin Testing	Skin Testing Per Protocol	ComputerizedGu ideline/APP	Skin Testing	Skin Testing Per Protocol‡	ComputerizedGuideline/APP†
Inpatient Penicillin or Cephalosporin	1.2 [0.8, 1.8]	4.2 [2.0, 8.9]|	1.6 [1.1, 2.5]§	1.3 [0.8, 2.0]	5.7 [2.6, 12.5]|	1.8 [1.1, 2.9]¶
Discharge Use Penicillin or Cephalosporin	0.7 [0.4, 1.2]	1.8 [0.8, 4.0]	1.1 [0.6, 1.9]	0.7 [0.4, 1.3]	2.5 [1.04, 6.2]**	1.0 [0.6, 1.9]
Inpatient Use of Alternative Antibiotics	1.1 [0.8, 1.7]	1.5 [0.8, 3.1]	1.2 [0.8, 1.9]	1.1 [0.7, 1.7]	1.8 [0.9, 3.8]	1.0 [0.7, 1.7]

^*Data presented as OR [95% CI]

^†Adjusts for frequency of methicillin-resistant Staphylococcus aureus or vancomycin-resistant Enterococcus colonization, an infection as the admission diagnosis, length of stay, concomitant cephalosporin allergy, penicillin intolerance, and Allergy/Immunology consultation.

^‡Adjusts for frequency of methicillin-resistant Staphylococcus aureus or vancomycin-resistant Enterococcus colonization, an infection as the admission diagnosis, length of stay, concomitant cephalosporin allergy, and penicillin intolerance.

^|p<0.001;

^§p=0.03;

^¶p=0.02;

^**p=0.04

Secondary Outcomes

Although a similar proportion of patients in the ST period used a penicillin or cephalosporin on discharge compared to the SOC period, a greater frequency of ST-PP patients received a penicillin or cephalosporin on discharge compared to SOC (26% vs 16%, p=0.18, Table II). In the multivariable model, there was an increased odds of a penicillin or cephalosporin used for discharge treatment for ST-PP patients, compared to SOC (OR 2.5 [95% CI 1.04, 6.2], Table III). In both univariable and multivariable analyses, the frequency of discharge use of a penicillin or cephalosporin was similar in the APP period, compared to SOC (Tables II and III).

Overall frequency of alternative antibiotic use was largely similar across periods (Tables II and III, Bottom).

There were no adverse drug reactions identified in SOC or ST periods. One patient in the APP period with history of penicillin allergy developed erythema and pruritus to amoxicillin after a negative penicillin skin test.

Balance Measure

The frequency of inpatient macrolide antibiotic use was unchanged by the interventions, compared to SOC (10%), for all comparisons (ST 10%, p =0.89; ST-PP 2%, p=0.10; APP 12%, p=0.67).

DISCUSSION

In this study of two antibiotic stewardship innovations designed to safely increase use of favorable beta-lactam antibiotics in medical inpatients reporting penicillin allergy, we found that the allergy history-driven computerized guideline as well as completed penicillin skin testing significantly increased penicillin or cephalosporin antibiotic use. While skin testing period patients did not have increased odds of penicillin or cephalosporin use, patients in the computerized guideline period had a significant almost 2-fold increased odds of receiving a penicillin or cephalosporin antibiotic. Completed skin testing, in the per protocol analysis, had an almost 6-fold increased odds of receiving a penicillin or cephalosporin antibiotic.

Among inpatients treated for infections, 5–25% report an allergy to penicillin.^1,4,7,16,31 Although up to 3 in 4 patients with alleged penicillin allergy warrant inpatient treatment with a beta-lactam antibiotic,⁴ only half were receiving these antibiotics in the baseline SOC period. We implemented interventions that were associated with increased use of favorable penicillins and cephalosporins, reflecting improved antibiotic choice and stewardship. Since the impact of over-reporting penicillin allergy is felt beyond antibiotic utilization to resultant readmissions,³¹ treatment failures,⁸ and adverse events,⁴ safely increasing the use of penicillin and cephalosporin antibiotics in this patient population is a crucial first step towards improved quality of care and reduced cost.³²

We did not observe a significant impact of the interventions on the frequency of discharge use of penicillins or cephalosporins in the primary analyses, or inpatient use of alternative antibiotics. The former may be explained by providers’ choice for outpatient therapy placing more emphasis on medication convenience. Yet, we observed that patients who completed skin testing in the per protocol analysis had an increased odds of penicillin or cephalosporin use at discharge, which may indicate that skin test negative patients have improved antibiotic use beyond their hospitalization. It was not surprising that neither intervention had a significant impact on use of alternative antibiotics, since medical inpatients with infections are often given empiric treatment that includes the alternative antibiotics.

Prior studies addressing over-reported penicillin allergies have demonstrated the utility of penicillin skin testing in emergency, intensive care, and perioperative patients.^10,13,33 Fewer skin testing studies are reported among highly selected inpatient subsets.^16,34–36 Alternative approaches to address inpatient penicillin allergies have included consultation by allergy specialists^34,37 and guidelines/clinical pathways.^26,38 By implementing and assessing both a skin testing approach and guideline approach in sequential cohorts at the same institution, we were able to determine the overall hospital policy level impact of these interventions.

While completed penicillin skin testing was effective at improving inpatient antibiotic choice, 36% of patients were ineligible for testing, consistent with prior estimates that deemed 4–59% of inpatients skin test ineligible.^16,39,40 One modifiable reason that patients in the ST period overall may not have experienced a significant increase in penicillin and cephalosporin use was an inability to coordinate testing for 58 skin test eligible patients. However, we skin tested 16% of all patients reporting penicillin allergy on antibiotics in this period, and 24% of those who were skin test eligible, largely comparable to the 19% yield achieved through the use of a designated pharmacist performing skin testing in a prior study.³⁶ Still, more patients would have had testing completed, and overall findings may have differed, had we hired an on-site skin testing clinician.

Despite challenges associated with coincidental innovation at the time of electronic health record conversion,^41–44 there were almost 300 unique webpage views of the computerized guideline, and 112 providers completed decision support. Given the computerized guideline’s overall positive impact on antibiotic use, reliance on new technology over new human capital, and relative ease of implementation, the computerized guideline and decision support tool have great potential to improve the care of inpatients with reported penicillin allergy. This conclusion is further strengthened by our knowledge of the intervention’s limited electronic health record integration and basic user interface during the study period.^45,46 Further, the computerized guideline and decision support tool can provide guidance when hospitals do not have access to allergists or penicillin skin testing, or when patients are otherwise ineligible or unwilling to complete a skin test.

Innovations in healthcare delivery substantially change care, while quality improvement targets incremental change.⁴⁷ The interventions implemented challenged the status quo by introducing a new process (i.e., skin testing) and a new technology (i.e., computerized guideline) to change the care of inpatients labeled penicillin-allergic. Healthcare systems are ill-equipped to address penicillin allergy because general clinicians have poor drug allergy knowledge,²⁵ there are inadequate numbers of allergy specialists to perform these evaluations,^48,49 and less than 15% of U.S. hospitals have the skin testing reagent on formulary.⁵⁰ Thus, healthcare delivery innovations are needed with investments in infrastructure, technology, and/or human resources. These investments may contain elements of the computerized guideline, which was easy to implement and relatively inexpensive. Hospitals will likely need processes for performing penicillin skin testing and test dose challenges, even without access to allergy specialists.

While the hospital policy perspective of this study afforded us a realistic view of the overall impact of our interventions, the analysis underestimates the potential impact because of incomplete uptake. Using a per protocol analysis for the penicillin skin testing period provided additional insight, but is subject to selection bias towards patients who would most clearly benefit from the penicillin skin test. While a per protocol analysis may have been similarly favorable for patients whose providers used the computerized guideline, these data were not available. Although cohorts were identified prospectively, data were gathered retrospectively, potentially resulting in missing or misclassification information; for example, we could not identify mild adverse reactions, if they were not included in the electronic health record. Data were also gathered in two different electronic health record systems, however, variables were collected by medicine-trained chart verifiers comfortable with both systems. We acknowledge that our quasi-experimental design limits our ability to demonstrate causality, but we included the macrolide balance measure to provide reassurance against unmeasured confounders or secular trends. This study assessed antibiotic outcomes within patients’ incident admission only, and did not capture important longitudinal outcomes, such as clarity of allergy status/documentation or types of antibiotics (i.e., beta-lactams or alternatives) used after discharge or in future hospitalizations. Finally, internal medicine patients may not be similar to other types of inpatients (e.g., surgical, obstetric), and results may differ in hospitals whose patients, resources, and/or personnel differ from those of BWH.

In summary, we implemented and assessed two interventions in the care of hospitalized medical patients with reported penicillin allergy. Although barriers to completing skin testing routinely in medical patients made skin testing impractical and without an overall impact on antibiotic use, skin tested patients had almost 6-fold odds of increased penicillin and cephalosporin use. The computerized guideline had an overall positive, almost 2-fold impact, with a feasible implementation and long-term strategy for greater technology integration.

Clinical Implications.

Improved antibiotic choice among medical inpatients reporting prior penicillin allergy was achieved using a computerized guideline, which resulted in an overall significant increased odds of penicillin or cephalosporin use compared to standard of care (aOR 1.8 [95% CI 1.1, 2.9]).

Abbreviations

SOC

standard of care

ST

penicillin skin testing

APP

computerized guideline application with decision support

BWH

Brigham and Women’s Hospital

OR

odds ratio

CI

confidence interval

IQR

interquartile range