Discrepancies Between Patient Self-Reported and Electronic Health Record Documentation of Medication Allergies and Adverse Reactions in the Acute Care Setting: Room for Improvement

Abstract

Background: Allergy information is commonly transcribed into an electronic health record (EHR) for all patients admitted to acute care hospital units by a licensed health care professional. The allergy history is utilized each time a new inpatient medication is prescribed to identify the patient’s risk of having an allergic reaction and/or anaphylaxis. There is potential for negative consequences in cases where the allergy history is incorrectly documented. Objective: The objective of this study was to assess the discordance between documented allergy information in the EHR and verbally reported allergy information from a patient interview. Methods: Prospective, observational, nonrandomized study performed within a 2-month period during the Spring of 2016. The study was performed at a teaching community hospital in Chicago, Illinois. A total of 270 patients were interviewed on the general medicine (n = 216) and headache (n = 54) units regarding their medication allergies and reactions. The outcomes were discordance among EHR-documented and verbally stated medication allergies and reactions. Results: The agreement across all medications and reactions between the EHR and patient self-reported interview was 80.9%. There were 31 reactions (6.7%) that were verbally reported by patients but were not documented in the EHR (omissions) and 57 reactions (12.4%) that were verbally reported but were incorrectly documented in the EHR (incorrect documentations). Only 20 out of the 264 verbally reported reactions (7.5%) met the study definition of anaphylaxis. The highest rate of incorrect documentations occurred with opiate agonists, and the highest rate of omissions occurred with anticonvulsants. EHR documentation was more likely to be incorrect among patients who reported gastrointestinal reactions and was more likely to be correct among patients who reported cutaneous reactions. Conclusion: There was a high rate of discordance amid EHR-documented and verbally stated medication allergies and reactions. Errors among opiate agonists, anticonvulsants, and sulfa drugs were most prevalent.

Introduction

Medication allergies are ubiquitous within the health care system, presenting clinicians with the perennial challenge of understanding and managing their impact. While medication allergies are difficult to characterize, the Gell-Coombs’ classification system (Type 1-4) is frequently utilized to classify the likely mechanism of hypersensitivity.¹ Anaphylaxis (eg, Type 1, immunoglobulin E-mediated reaction) is a rare but serious and acute immunologically mediated systemic illness, often necessitating hospitalization and whose incidence appears to be increasing.^2,3 While anaphylaxis may arise in response to re-exposure to various medications, patients frequently experience other types of medication hypersensitivity reactions (MHR), as well as medication adverse events (MAE). Allergies, MAE, and MHR are major problems that pose barriers to optimal medication management and result in increased health care costs.^4,5 Given the pervasive nature of medication allergy labels, the difficulty discerning allergies and adverse effects by patients, and providers’ reluctance to challenge a history of labeled allergy by other providers, the prevalence of labeled medication allergies will likely increase in the foreseeable future.

The incorrect classification of allergy and the allergy label itself has negative consequences for patients.⁶ Patients carrying a drug allergy label are more likely to receive more toxic or less active alternative agents (eg, most infection guidelines prefer β-lactams as first-line therapy, with second-line therapy reserved for patients with allergies to β-lactams).⁶ Thus, accurately identifying historical information with regard to each reported allergy and placing it in the context of the reported syndrome is critical. Information regarding when the reaction occurred, the time between medication administration and reaction occurrence, and the reaction that occurred should all be carefully vetted when documenting patient medication allergies.⁵

The accuracy of medical record documentation is complicated in practice by a variety of factors, including underdiagnosis (most often due to patient underreporting) and overdiagnosis (most often due to the overuse of the term “allergy”). In addition, a reported allergy to one medication may lead to the distortion that the patient is allergic to the entire medication class. Furthermore, literature that can be useful for assessing a medication allergy or adverse effect in clinical practice is limited. A number of regional, national, and international documents exist but they differ in terms of scope and methodology.⁶ While electronic health records (EHRs) are expected to improve documentation, they can also serve as a reservoir of incorrect information if not actively managed.^7,8 In many EHR systems, anaphylaxis, MHR, and MAE are poorly differentiated. Improper EHR documentation of allergies can lead to poor patient outcomes.⁹ The purpose of this study was to evaluate the concordance between verbally expressed and EHR-documented medication allergies and reactions.

Patients and Methods

Study Design

We conducted a prospective, observational cohort study of patients that was approved by the institutional review board at our institution. The study was performed at a 313-bed teaching community hospital during a 2-month period in the Spring of 2016. As a standard of care, the nursing staff and medical residents at our facility were responsible for entering and updating patient allergies. Patients admitted weekdays during daytime hours to the postoperative/general medicine unit and headache unit underwent a comprehensive allergy history evaluation in a structured patient interview as part of a standard-of-care clinical pharmacist service for this study’s protocol. The interviewers were pharmacists attached to the respective wards.

Patient Assessment and Inclusion/Exclusion Criteria

All patients older than 18 years of age who were admitted to the inpatient postoperative/general medicine unit and headache unit were eligible to participate in the study. Any patient with documented altered mental status was not eligible to be interviewed. Informed written consent was collected for all patients who agreed to participate in the study. Eligible and consented patients were prospectively interviewed and underwent allergy assessment. The structured patient interview utilized a standard questionnaire (Supplemental Table S1, available online). After the interview, investigators recorded the verbally stated medication allergy(ies) using a standardized de-identified case report form. Documentation of medication allergy(ies) and reaction(s) within the EHR (Epic 2017, Verona, WI) were then extracted and recorded in the standardized case report form. At our institution, reaction descriptors for the medication allergies were not required. When a chart was accessed, no automatic EHR notification or provider alerts prompted any update of the allergy profile.

Data Elements and Definitions

Patient demographic data (patient’s age, gender, and assigned hospital unit) were recorded during the structured interviews and recorded on the de-identified case report forms. Adverse medication reactions were classified as meeting criteria for anaphylaxis or not. The criteria of Sampson et al (Appendix, available online) were used to determine if the patient had an anaphylactic reaction.⁵ Offending medications were classified according to the American Hospital Formulary Service drug classes.¹⁰ Adverse hypersensitivity reactions were classified as immediate if they occurred within 6 hours of medication administration and delayed if they occurred greater than 6 hours after administration.

Outcomes and Statistical Analyses

The primary outcome was the concordance between the verbally stated and documented medication allergy or allergies in the EHR expressed as a percent essential agreement. In order to meet the primary outcome (eg, for there to be a match between the verbally stated and documented medication allergy), all of the stated medications had to match the documented medications in the EHR. For example, if the patient had 3 documented medication allergies but only stated that he or she has 2 medication allergies during the interview, then the primary outcome of concordance was not achieved. If a patient was unaware of the exact reaction, could not recall the nature of the reaction, or could not verify details of the reaction, then the verbally expressed and documented reactions were classified as discordant. The patient needed to verbally state all reactions that were in the EHR in order to meet the primary outcome. The secondary outcomes were the discordance rates between the verbally stated and documented reaction(s) in the EHR expressed as omission (eg, false-negative EHR documentation; the patient verbally stated the reaction he or she experienced but it was omitted or no reaction was documented for the medication in the EHR) or incorrect documentation (ie, false-positive EHR documentation; the patient verbally stated the reaction he or she experienced but it was incorrectly documented as a different type of reaction for the same medication in the EHR).

For the primary and secondary outcomes, descriptive statistics were calculated for concordance. Demographic differences between patients with correct and incorrect EHR documentation status were compared using Student’s t tests for continuous variables and χ² or Fishers exact test for categorical variables, as appropriate. The primary and secondary endpoints were evaluated using confusion matrices, with sensitivity and specificity of EHR documentation calculated for each major medication class. Statistical significance was set at P < .05, and all statistical tests were performed using Intercooled Stata version 14.2 (StataCorp LLC, College Station, TX).

Results

There were 292 patients who were asked to participate in the study and 270 patients gave consent and agreed to participate, yielding a 92% participation rate. Overall, included patients were predominantly female (63.3%) with a mean ± SD age of 53.7 ± 17.6 years. Patients were recruited primarily from the general medicine ward (80%). Two thirds of patients interviewed had no documented reactions in the EHR; similarly, 59.3% of patients included reported no known history of adverse medication or hypersensitivity reactions.

Across all documented and reported medication adverse reactions, a total of 18 patients were incorrectly classified by the EHR (ie, misclassified with or without an adverse reaction label), and 50% of these represented a false omission. Demographic characteristics of the study population stratified by incorrect EHR classification are summarized in Table 1. Patient demographics (eg, age and sex) did not significantly differ by incorrect EHR classification. Patients who were incorrectly classified by the EHR were more likely to have no stated history of medication reactions at the time of the interview compared with those who were correctly classified (88.9% [n = 16/18] vs 57.1% [n = 144/252]; P = .008).

Table 1.

Baseline Characteristics of Patients Interviewed.

Characteristic	Overall	EHR Correct	EHR Incorrect	P
Unique patients (N = 270), n (%)	270 (100)	252 (93.3)	18 (6.7)
Age, mean ± SD	53.7 ± 17.6	54.0 ± 17.8	49.8 ± 15.3	.34
Male, n (%)	99 (36.7)	94 (37.3)	5 (27.8)	.42
General medicine unit, n (%)	216 (80)	203 (80.6)	13 (72.2)	.39
Documented reactions, n (%)	90 (33.3)	81 (32.1)	9 (50)	.12
No documented reactions, n (%)	180 (66.7)	171 (67.9)	9 (50)
Stated reactions, n (%)	110 (40.7)	108 (42.9)	2 (11.1)	.008
No stated reactions, n (%)	160 (59.2)	144 (57.1)	16 (88.9)

Abbreviation: EHR, electronic health record.

Among the 270 patients assessed, 200 reported at least 1 adverse reaction. A total of 460 reactions (or lack thereof) to various medications were recorded across all patients. Only 20 out of the 264 verbally reported reactions (7.5%) met the study definition of anaphylaxis. The proportion of all adverse and hypersensitivity reactions, as recounted in the interview, that were correctly classified by the EHR is summarized in Table 2. The essential agreement across all agents and reactions (or lack of reactions) between the EHR and patient self-reported history was 80.9%. Notably, the patient interview uncovered an additional 31 (6.7%) adverse or hypersensitivity reactions omitted by the EHR. The structured interview also revealed 57 (12.4%) adverse or hypersensitivity reactions that were incorrectly documented. Across the 4 most common medication classes associated with adverse or hypersensitivity reactions, essential agreement ranged from 70% to 100%, as shown in Table 2. The highest rates of omissions were associated with anticonvulsants (30%) and opiate agonists (16.7%) within the 4 medication classes most frequently reported. On the other hand, the highest rates of incorrect documentations were associated with opiate agonists (18.9%) followed by sulfa medications (8.7%).

Table 2.

Discordance Rates Between All EHR-Documented and Patient-Reported Allergies or Adverse Events According to Documented Drug Class.

Drug class	Agreementa, EHR = History; n/N (%)	Incorrect Documentationb, EHR + | History −; n/N (%)	Omissiona, EHR − | History +; n/N (%)
Anticonvulsant	7/10 (70%)	0 (0%)	3 (30%)
Opiate agonist	30/36 (81.1%)	7/37 (18.9%)	6 (16.7%)
Penicillin	40/40 (100%)	1/41 (2.4%)	0 (0%)
Sulfa drug	21/23 (91.3%)	2/23 (8.7%)	2/23 (8.7%)
Overall reactions	372/460 (80.9%)	57/460 (12.4%)	31/460 (6.7%)

Abbreviation: EHR, electronic health record.

^aDenominator is patient-stated reactions.

^bDenominator is EHR-documented reactions.

Among the 200 patients evaluated with a stated reaction, 214 documented reactions and 264 stated reactions were recorded. A wide range of adverse medication and hypersensitivity reactions were documented and reported by included patients. The correct classification of adverse medication and hypersensitivity symptoms within the EHR documentation is summarized in Table 3, whereas the corresponding correct classification of symptoms according to patient self-report is summarized in Table 4. By far, the most common adverse or hypersensitivity reactions documented within the EHR were cutaneous reactions (40.7%; n = 87/214) followed by central nervous system (CNS) adverse reactions (19.2%) and gastrointestinal (GI) adverse effects (14.0%), as shown in Table 3. Correct classification of allergy or adverse event symptoms according to documented reaction occurred 93% of the time with cutaneous reactions, 80% of the time with CNS reactions, and 73% of the time with GI reactions. Similar patterns were noted within the analysis of patient self-reported reactions (Table 4). EHR documentation was numerically more likely to be incorrect among patients with documented GI reactions (24.2% vs 12.2%; P = .066), whereas EHR documentation was significantly more likely to be correct in patients with documented cutaneous reactions (44.8% vs 18.2%; P = .006), as shown in Table 3. Similar trends were observed for EHR classification of reactions according to patient self-report with the exception that reported GI reactions were significantly more likely to be incorrectly documented in the EHR (22.2% vs 8.8%; P = .015), as shown in Table 4. Correct classification of allergy or adverse event symptoms according to stated reaction occurred 93% of the time with CNS reactions, 89% of the time with cutaneous reactions, and 71% of the time with GI reactions.

Table 3.

Correct Classification of Allergy or Adverse Event Symptoms According to Documented Reaction.

Documented Reaction(s)	Total		EHR Correct		EHR Incorrect
	n	%a	n	%a	n	%a
Anaphylaxis (n = 8)	8	3.7	6	3.3	2	6.1
CNS (n = 41)	41	19.2	33	18.2	8	24.2
Cardiovascular (n = 8)	8	3.7	4	2.2	4	12.1
Cutaneousb (n = 87)	87	40.7	81	44.8	6	18.2
Edema (n = 1)	1	0.5	1	0.6	0	0
Gastrointestinalc (n = 30)	30	14.0	22	12.2	8	24.2
Hypothermia (n = 1)	1	0.5	0	0	1	3
Pain (n = 3)	3	1.4	2	1.1	1	3
Paralysis (n = 1)	1	0.5	1	0.6	0	0
Respiratory (n = 13)	13	6.1	12	6.6	1	3
Sick (n = 1)	1	0.5	1	0.6	0	0
Slurred speech (n = 2)	2	0.9	2	1.1	0	0
Stroke (n = 2)	2	0.9	0	0	2	6.1
Suicidal (n = 3)	3	1.4	3	1.7	0	0
Swelling (n = 10)	10	4.7	10	5.5	0	0
Thrombocytopenia (n = 1)	1	0.5	1	0.6	0	0
Transaminitis (n = 1)	1	0.5	1	0.6	0	0
Tremor (n = 1)	1	0.5	1	0.6	0	0
Total (N = 214)	214	100	181	100	33	100

Abbreviations: CNS, central nervous system; EHR, electronic health record.

^aColumn percent.

^bCutaneous reactions: P = .006.

^cGastrointestinal reactions: P = .066.

Table 4.

Correct Classification of Allergy or Adverse Event Symptoms According to Stated Reaction.

Documented Reaction(s)	Total		EHR Correct		EHR Incorrect
	n	%a	n	%a	n	%a
Anaphylaxis (n = 5)	5	1.9	4	1.8	1	2.8
Anorexia (n = 1)	1	0.4	1	0.4	0	0
CNS (n = 42)	42	15.9	39	17.1	3	8.3
Cardiovascular (n = 9)	9	3.4	7	3.1	2	5.6
Cutaneousb (n = 112)	112	42.4	100	43.9	12	33.3
Edema (n = 1)	1	0.4	1	0.4	0	0
Eye spasm (n = 1)	1	0.4	1	0.4	0	0
Eye swelling (n = 1)	1	0.4	1	0.4	0	0
Fever (n = 1)	1	0.4	1	0.4	0	0
Gastrointestinalc (n = 28)	28	10.6	20	8.8	8	22.2
Indeterminate (n = 23)	23	8.7	19	8.3	4	11.1
Infection (n = 1)	1	0.4	1	0.4	0	0
Neuropathy (n = 1)	1	0.4	1	0.4	0	0
Pain (n = 1)	1	0.4	1	0.4	0	0
Paralysis (n = 1)	1	0.4	1	0.4	0	0
Paresthesia (n = 1)	1	0.4	1	0.4	0	0
Respiratory (n = 22)	22	8.3	19	8.3	3	8.3
SLE (n = 1)	1	0.4	0	0	1	2.8
Slurred speech (n = 2)	2	0.8	2	0.9	0	0
Suicidal (n = 3)	3	1.1	3	1.3	0	0
Swelling (n = 2)	2	0.8	2	0.9	0	0
Thrombocytopenia (n = 1)	1	0.4	1	0.4	0	0
Transaminitis (n = 1)	1	0.4	1	0.4	0	0
Tremor (n = 3)	3	1.1	1	0.4	2	5.6
Total (N = 264)	264	100	228	100	36	100

Abbreviations: CNS, central nervous system; EHR, electronic health record; SLE, systemic lupus erythematosus.

^aColumn percent.

^bCutaneous reactions: P = .24.

^cGastrointestinal reactions: P = .015.

Discussion

We found that discrepancies between patient self-report and EHR documentation were common in our population. Omission errors occurred primarily with anticonvulsants. Omission and incorrect documentation errors were common with opiate agonists. Only 7.5% of verbally stated reactions met the definition of anaphylaxis. The most common reactions documented in the EHR were related to the central nervous, cutaneous, and GI systems. The proportion of patients who had an allergy to an anticonvulsant or experienced a CNS adverse effect were likely enriched in a population presenting to the headache unit. In our cohort, β-lactams were the most common class of antibiotics associated with medication allergy. Literature estimates suggest that β-lactams account for at least 10% of self-reported medication allergies.¹¹ Furthermore, in patients who undergo β-lactam skin testing, 71% to 90% of low-risk patients with a β-lactam allergy history are able to tolerate this class.^11-15 Only 1 (2.4%) reaction within our cohort to a penicillin represented an incorrect documentation error. However, reaction descriptors for medication allergies were not required within our EHR. Therefore, the majority of medication allergies did not contain a reaction descriptor.

Our findings are similar to previous studies focusing on the accuracy of documented allergy history. Wyatt found that 15.7% (n = 38/242) of patients interviewed claimed to be allergic to a medication but could not remember the name.¹⁶ An additional 57 patients could not remember what their allergic reaction was. Lyons et al found that only 50% of allergic reactions verbally confirmed during a patient interview matched the reaction already documented in the patient’s chart.¹⁷ Bates et al found that clinicians were less likely to document new allergy information when the reactions occurred in the hospital setting.⁷ In addition, allergy information was only entered into the computer 16% of the time; MHRs were entered less frequently than allergies. Kuperman et al reported that alert fatigue in an EHR can be secondary to the poor quality of allergy reporting.¹⁸ One way to minimize such alert fatigue and avoid suboptimal alternative therapies is to remove inaccurate allergy labels after appropriate testing and challenge. However, Hsieh et al found that appropriate de-labeling after a challenge only happens 17% of the time outside of a formal consultation.¹⁹

Furthermore, Kiechle et al found that 41% of patients admitted to the emergency department had discrepancies between the self-reported allergy or adverse effect and the EHR.²⁰ Shen et al investigated adverse effects as reported in paper charts and the EHR at a tertiary hospital in New Zealand. They reported that 90% of patients had discrepancies in adverse effect information in the EHR and this number significantly rose to 98% of patients within paper charts.²¹ Hui et al performed a retrospective analysis of adverse drug reaction documentation at a tertiary care pediatric hospital. The results indicated that 44% of patients had incorrectly documented adverse drug reaction information in the EHR.²² The results of our study, performed in 2 separate adult inpatient units of a community hospital, are consistent with the results of other studies performed in various settings and further clarify which medication classes are most prone to discordance.

Our study supports the need for vigilance in maintaining correct allergy documentation. The results indicate that quality improvement measures targeting medication allergies are warranted. There is a clearly defined role for the pharmacist in triaging and clarifying medication allergies. The results indicate that clinical pharmacist involvement could enhance the rate at which allergy, MHR, and MAE are correctly documented compared with the standard of care.

There were some limitations to our study. First, this was an observational and nonrandomized study subject to inherent limitations. Second, our sample was somewhat limited but it represents a severely underreported population (eg, patients within community hospitals). In spite of our limited resources, we were still able to recruit and obtain informed consent from nearly 300 patients in a short period of time. Third, gold-standard classification of allergy, MHR, or MAE was assessed based on the patient’s recall of his or her allergies meaning that recall bias may have influenced our findings. We attempted to circumvent this by only interviewing patients who were most likely to correctly recall any reactions by excluding patients with altered mental status.

Our experience supports that most patients are highly invested in their allergy history and label, even if the time course and specific reactions were unclear. Therefore, the false omission rates are conservative and highly concerning. An allergy and adverse effect history should be taken from each patient on each transition of care by a clinical pharmacist or other health care professional. Studies focusing on specific quality improvement interventions targeting medication allergy, MHR, and MAE documentation are warranted.

Conclusions

Discordance between documented and verbally reported allergies and reactions is highly prevalent. Incorrect documentation errors were most prevalent with opiate agonists and omission errors were most prevalent with anticonvulsants and opiate agonists. Future studies are needed to determine the optimal timing of clinical pharmacist involvement to improve the classification and management of adverse medication reactions.