Completeness and accuracy of adverse drug reaction documentation in electronic medical records at a tertiary care hospital in Australia

Introduction

Adverse drug reactions (ADRs) are common. Studies have reported that 20–35% of the hospital inpatient and outpatient population had an ADR documented in their electronic medical record. (Blumenthal et al., 2017; Zhou et al., 2016; Foreman et al., 2020) A minority of these ADRs, estimated at 5–10%, are truly immune-mediated allergies (Zhou et al., 2016). While many non-allergy ADRs are important to document, inappropriate labelling of a reaction as an allergy can lead to patient harm (Inglis et al., 2017). Recent studies of penicillin allergy labelling have shown that an overwhelming majority of patients who are labelled as allergic to penicillins are able to tolerate them, (Inglis et al., 2017; Chua et al., 2020) and inaccurate allergy labelling contributes to inappropriate antimicrobial prescribing and antimicrobial resistance (Trubiano et al., 2016)

Electronic medical records (EMRs) should facilitate accurate and complete documentation of previous ADRs, so that these details are available for healthcare workers during future episodes of care. A recent study of ADR documentation before and after implementation of EMR in the South Australian hospitals showed significantly higher rates of ADR documentation following EMR introduction, theorised to be due to persistent electronic prompts when this field was left incomplete. (Al-Sarawi et al., 2019).

ADR documentation processes in EMRs should be designed to reduce the possibility of accidental omission of information. However, the current design of EMRs does not usually make recording of all key details mandatory, nor allow for meaningful differentiation of mild or moderate adverse drug reactions or intolerances from more serious reactions mediated by allergic mechanisms (Foreman et al., 2020). Clinicians should be accurately informed of previous reactions to enable appropriate and safe treatment decisions to be made. For example, the use of erythromycin may be avoided if an EMR alert fires in a patient who had previously experienced nausea and vomiting to the medication, however, such a reaction should not rule out its use if it is the most suitable choice for the patient. Accurate and complete documentation of previous adverse drug reactions informs this clinical decision making.

At the hospital where this study was undertaken, when healthcare workers enter ADRs into the EMR (Cerner Millenium), the mandatory fields are: drug name and ADR type (allergy or intolerance/sensitivity). ADR type defaults to allergy unless the clinician changes it to intolerance or sensitivity. Reaction and severity fields must be completed but can be recorded as ‘unknown’. Non-mandatory fields include information on when the reaction occurred and the source of the information. The selection of reaction is not tied to the reaction classification. For example, ‘drowsiness’ can be entered as a reaction even if the classification is selected as allergy.

The aim of this study was to determine the completeness of ADR documentation (drug name, reaction and severity) and the accuracy of ADR classification (allergy vs intolerance) in a sample of EMRs at a major metropolitan hospital.

Method

We performed a cross-sectional audit of EMRs of patients on general medicine, surgical, aged care and haematology/oncology wards at a metropolitan tertiary referral hospital (780 + beds) in Melbourne, Australia, on three separate days, spread over 3 weeks in May 2019. Patients on the wards on more than one of the study days were only included once and excluded on subsequent days. The ADR section of the EMR for every patient on these wards on the study days was reviewed and the data collated by one investigator (AB) in Microsoft Excel, in which the analyses were performed using Excel functions to determine proportions, means, medians and interquartile ranges.

Results

Two hundred and sixty-four patients were evaluated: 133 (50.4%) were male; mean age 70 years, median of 9.0 regular medications). Of these, 102 patients (38.6%) had a total of 210 ADRs (83 to antimicrobials, 59 to pain medicines and 68 to other medicines) documented in their EMR (median 1.0 per patient, interquartile range 1–8.5). Twenty-two (10.5%) ADRs were categorised as a sensitivity or intolerance, with the remainder classified as allergies.

Documentation was considered complete for 105/210 (50.0%) ADRs. No ADRs were missing drug name (as this was a mandatory field); 63/210 (30.0%) were missing a reaction description and 88/210 (41.9%) were missing severity information. Documentation was most complete for antimicrobial ADRs (48/83, 57.8%), then pain medicine ADRs (29/59, 50.0%) and lowest for other ADRs (28/68, 41.2%).

When reaction information was documented (n = 147 ADRs), 97 (66.0%) were classified correctly as either allergy or intolerance/sensitivity. The remaining 50 (34.0%) were non-allergic reactions (e.g. nausea, diarrhoea) that were incorrectly classified as an allergy.

Complete ADR information was most common on the haematology/oncology ward (25/34; 73.5%), and lowest on the general medicine ward (26/69; 37.7%). Accuracy of classification was highest on the general medicine ward (52/69, 75.0%) and lowest on the surgical ward (27/48, 56.0%) (Figure 1).

OPEN IN VIEWER

Figure 1.

Completeness of adverse drug reaction (ADR) documentation and accuracy of ADR classification by ward.

Discussion

Consistent with previous studies, ADRs were common in our patient population, with 39% of patients having an ADR documented in the EMR. This is a slightly higher ADR rate than previous studies which may be due to the population studied (inpatients on a general medicine, oncology/haematology, surgical and aged care wards). Previous studies have shown increased documentation of ADR fields in EMR compared to paper charts. (Al-Sarawi et al., 2019), yet few studies have explored the completeness and accuracy of ADR documentation. Our audit found that half of the ADRs documented in patients’ EMRs lacked the minimum basic information required to inform clinicians’ future treatment decisions, and one-third misclassified the ADR type.

Incorrect classification of ADR type may occur due to inadequate clinician knowledge and training, and a lack of awareness of the importance of selecting the right ADR category. Another important factor is that ‘allergy’ is the default category within the EMR used at the study hospital, so if the reaction is an intolerance, the clinician needs to actively change it. Having the ADR category default to a particular ADR type within an EMR creates risk of incorrect information that may compromise patient care, and should be avoided by EMR developers.

Some of the disparity in completeness of ADR documentation noted between wards in our study may have been due to a concurrent project on the haematology/oncology ward utilising a risk stratification tool to analyse the appropriateness of documented penicillin allergies (Devchand et al., 2019). The tool encouraged nurses to consider the nature and severity of the reaction, and the time elapsed since reaction to determine which patients would be suitable for an oral rechallenge. This may have led to better documentation of reactions and severity for penicillins as well as other medicines by encouraging detailed ADR history taking, an essential component of a proper documentation. Further research is required to investigate the reasons behind the differences in degree of accuracy and completion between wards.

A limitation of our study is that our reported number of incorrect classifications of reactions is likely to be an underestimation due to the conservative nature of our pre-determined list of reactions indicating potential allergy. For example, not all urticaria, rashes or angioedema develop as a result of allergic mechanism (e.g. opioid release of histamine causing a rash), and yet according to our audit criteria they were always considered allergic in nature. Our percentage of incorrectly classified reactions is smaller than that described in a South Australian study, using a different EMR, where the incidence of incorrect classification as allergy rather than intolerance was 54.0%. (Foreman et al. 2020). This likely relates to the differences in EMR with automatic linking of particular reactions with allergy or intolerance classification in their study. Our study was performed at a single site, with a single EMR design, so cannot be extrapolated to all settings. However, many hospitals use the same EMR system and may have the same structure for documentation of allergies.

One of the advantages of EMRs is endurance of ADR information. Unlike paper charts, ADR information does not need to be documented repeatedly during an admission or on further admissions. However, if the EMR information is not accurate and complete, this advantage is lost and incorrect information can be perpetuated. A forcing function or alert for clinicians to review ADR information at the start of each admission may reduce this inaccuracy, but may contribute to ‘alert fatigue’.

Another advantage of EMRs is that documentation of ADRs in the EMR allows alerts to fire when that medication, or ones considered a cross-reaction risk, are prescribed. The usefulness of these alerts is diluted when ADR data are incomplete or if there is no differentiation between allergy or intolerance reactions. An example is that penicillins would not need to be avoided if a documented reaction is diarrhoea, yet the alert will fire anyway. In an ideal world, the alerts about cross-reactivity would only fire when they are appropriate, such as when allergic mechanisms are involved or when different agents within a medication class may cross react (for example, statin induced hepatotoxicity). A decrease in alerts may reduce alert fatigue and reduce the number of times allergy alerts are overridden without due consideration, which has been observed at rates greater than 80% (Topaz et al. 2016).

Our audit may enable us to identify key successes and areas for improvement in awareness and history taking skills to improve allergy documentation by healthcare workers at our institution. It also led to an ongoing review of the categories available within the EMR for documentation of ADRs to further simplify reaction categorisation.

Conclusion

In conclusion, this study highlights deficiencies in completeness and accuracy of ADR documentation within the EMR at the study hospital. EMR developers need to enhance the design of ADR documentation functionality through design and testing with end users. It is also imperative to educate healthcare workers about the importance of accurate and complete documentation of ADRs to protect our patients.