The Burden of Disease of Medication Errors: Analysis in ADR Caused Emergency Visits

1. Introduction

Medication errors (MEs) are among the most preventable causes of adverse drug reactions (ADRs) and remain a critical challenge for patient safety. Globally, ADRs contribute substantially to emergency hospital admissions, morbidity, mortality, and healthcare costs, yet the specific role of MEs is often underexplored [1,2,3,4].

In Germany, most available information stems from spontaneous reporting systems, pharmacovigilance centers, or critical incident reporting systems (CIRS) in hospitals. While these sources are valuable for case detection and signal generation, they are inherently limited: they do not allow reliable estimates of incidence or prevalence, nor do they provide systematic insight into how ME-related ADRs compare with those arising independently of errors [5,6,7].

The scale of the problem becomes clearer when considering routine emergency care. A single maximum-care hospital emergency department (ED) treats approximately 30,000 patients annually. Even with a conservative ADR rate of 10%, this would correspond to around 3,000 ADR-related admissions per hospital each year [8,9,10]. Previous Drug therapy safety projects have suggested that about one quarter of such ADRs are attributable to MEs [11,12,13,14,15]. Extrapolated nationally, this translates into tens of thousands of potentially preventable admissions annually—a burden that cannot be ignored.

Despite these figures, prospective data on the prevalence, characteristics, and outcomes of ME-related ADRs in German EDs remain scarce. In particular, it remains unclear whether ME-related ADRs differ systematically from non-ME ADRs in terms of presentation, severity, and prognosis, or whether specific patient and treatment characteristics predispose individuals to such errors [16,17,18,19]. Without evidence, preventive strategies remain difficult to target effectively.

The ADRED (Adverse Drug Reactions in Emergency Departments) study was established to address this gap. Conducted prospectively across six German university hospitals, it systematically documented ADR-related ED admissions, assessed causality, and evaluated preventability using standardised methods [16,17,20,21,22,23,24,25,26].

2. Results

This section may be divided by subheadings. It should provide a concise and precise description of the experimental results, their interpretation, as well as the experimental conclusions that can be drawn.

2.1. Objectives

This analysis aimed to:

• Quantify the proportion of ADR-related ED admissions attributable to MEs,
• Compare the clinical presentation, severity, and outcomes of ADRs with and without ME involvement, and
• Identify patient- and treatment-related factors associated with the occurrence of MEs in the ED setting.

2.2. Study Population

Across the six EDs, 7,967 cases were documented. Of these, 6,364 (79.9%) were classified as ADRs without ME involvement and 1,603 (20.1%) as ADRs caused by MEs. Thus, one in five ADR-related ED admissions was potentially preventable.

Table 1 summarizes patient characteristics. Patients with MEs were approximately of the same age (mean 69.8 vs. 68.6 years) and less frequently male (49.3% vs. 51.1%). Both groups carried a heavy multimorbidity burden (median six chronic conditions) and were on extensive medication (median seven drugs). The number of suspected drugs was higher in ME cases (2.17 vs. 1.8). Figure 1 illustrates the distribution of polypharmacy across age groups in the study population. As expected, both the number of medications and the number of individuals with polypharmacy increased with age.

2.3. Clinical Presentation and Severity

Triage severity was broadly comparable between groups . Immediate triage level (resuscitation) was uncommon overall but slightly more frequent among ME-related cases (4.1% vs. 3.3%). Most patients were categorised as urgent or emergency, indicating similar clinical severity at presentation regardless of ME involvement.

Overall, ME-related ADRs initially presented with clinical features largely comparable to ADRs without ME involvement. Observed differences were mainly related to outcome, with permanent damage more frequently in the ME group. With regard to patient characteristics, MEs were more often observed among patients who reported current smoking or alcohol consumption (Table 1). Documentation of lifestyle factors and outcome parameters was incomplete across cases, which may limit the interpretability of these findings.

2.4. Drug Involvement

Drug-level analyses provided further insights into differences between ADRs with and without MEs (Table 2). Several commonly prescribed agents were disproportionately implicated in ME cases. Pantoprazole was more than twice as likely to be suspected in ME-related ADRs compared to non-MEs (5.5% vs. 2.4%, p < 0.001). Torasemide (39.1% vs. 26.1%, p < 0.001), metoprolol (30.1% vs. 19.3%, p < 0.001), ramipril (25.4% vs. 13.9%, p < 0.001), and metformin (32.9% vs. 9.8%, p < 0.001) showed similar patterns. Conversely, acetylsalicylic acid was less frequently suspected in ME cases (22.7% vs. 45.4%, p < 0.001), as was apixaban (28.2% vs. 51.2%, p < 0.001). Prednisolone and phenprocoumon, while major contributors to ADRs overall, were also significantly more often implicated in ME cases (both p < 0.01). Additional signals included bisoprolol, levothyroxine sodium, candesartan, ibuprofen, and allopurinol, all of which showed higher suspected proportions in ME cases.

Taken together, these results show that ME-related ADRs are not evenly distributed across all drugs but cluster around a subset of widely used chronic medications

2.5. Preventability Factors

Among 1,602 ME cases with Schumock assessment, the most frequent factors were inappropriate dosing (criterion 2; 481/1,602, 30.0%), lack of adherence (criterion 7; 454/1,602, 28.3%), contraindication (criterion 1; 410/1,602, 25.6%), and missing monitoring/laboratory tests (criterion 3; 316/1,602, 19.7%). Less common were drug interactions (criterion 5; 198/1,602, 12.4%), known allergy or intolerance (criterion 4; 35/1,602, 2.2%), and toxic serum levels (criterion 6; 29/1,602, 1.8%) (Figure 2).

Most patients had only one preventability factor (1,330/1,602, 83.0%). Two factors co-occurred in 230/1,602 (14.4%), three factors in 37/1,602 (2.3%), and more than three in only 5 cases (<0.5%). The most frequent double combination was inappropriate dosing + missing monitoring (criteria 2+3), observed in 51/1,602 (3.2%). The most frequent triple combination was contraindication + inappropriate dosing + missing monitoring (1+2+3), present in 13/1,602 (0.8%) (Figure 3, Table A1).

Taken together, the Schumock analysis highlights that preventable ADRs in the ED were predominantly driven by routine processes such as dosing, monitoring, and adherence, while complex mechanisms (interactions, allergies, toxic levels) played only a minor role.

2.6. Symptom Burden and Clusters

Analysis of symptom frequencies revealed several statistically significant differences between ADRs with and without MEs (Table 3). General deterioration of physical health was reported less frequent in ME-related cases (4.8% vs. 7.1%, p < 0.001). Dyspnea occurred more often in ME cases (5.4% vs. 4.3%, p < 0.001), while hematochezia (1.9% vs. 3.7%, p < 0.001) and anemia (1.5% vs. 3.4%, p < 0.001) were less commonly observed. Pyrexia was also less frequent among ME patients (0.9% vs. 2.5%, p < 0.001).

For the remaining symptoms—dizziness, nausea, vomiting, and abdominal pain—no significant differences were observed. Fatigue was slightly less common in ME cases (1.6% vs. 2.1%), reaching statistical significance (p = 0.01), although the absolute difference was small.

When examining the overall symptom burden, patients without MEs presented with a marginally higher number of Medical Dictionary for Regulatory Activities (MedDRA)-coded symptoms per case (mean 3.57, range 1–25) compared to those with MEs (mean 3.29, range 1–16; Mann–Whitney U, p < 0.001). Although statistically significant, the difference in symptom count between groups was small.

Overall, although several individual symptoms differed statistically between ADRs with and without MEs, the absolute effect sizes were small (data not shown). The general clinical picture remained broadly comparable, reinforcing that ME-related ADRs cannot be reliably distinguished from non-ME ADRs based on presenting symptoms alone.

To explore whether medication errors result in distinct clinical presentations, we compared drug-specific symptom complexes for the five most frequently implicated agents—metoprolol, Torasemide, ramipril, phenprocoumon and ibuprofen—across ADRs with and without MEs.

2.6.1. Metoprolol

Symptom complexes for metoprolol were dominated by cardiovascular effects, especially bradycardia, hypotension, syncope, dizziness, and falls (Figure 4a). In both ADR and ME cases, these features accounted for the majority of presentations, with respiratory symptoms and general deterioration contributing smaller proportions. No clinically meaningful differences were observed between ME and non-ME cases.

2.6.2. Torasemide

Torasemide-related ADRs were characterised by electrolyte disturbances, dehydration, dizziness, syncope, and confusion, reflecting the expected spectrum of diuretic toxicity. Both ADR and ME cases showed highly overlapping clusters, with renal impairment and weight fluctuations appearing in smaller numbers.

2.6.3. Ramipril

Symptom clusters for ramipril highlighted dizziness, syncope, and hypotension, alongside gastrointestinal complaints and occasional bleeding events. Pain-related complaints formed a substantial fraction in both groups. The distribution of symptoms was broadly similar between ME and non-ME cases, with no distinctive “error-specific” pattern emerging.

2.6.4. Phenprocoumon

Phenprocoumon clusters were dominated by bleeding manifestations, including upper and lower gastrointestinal bleeding, respiratory and urogenital tract bleeding, and anaemia-related sequelae. The relative distribution of bleeding sites varied slightly, but the overall profile was nearly identical in ME and non-ME cases, reflecting the pharmacological mechanism of vitamin K antagonism.

2.6.5. Ibuprofen

Ibuprofen was primarily associated with gastrointestinal complaints (abdominal pain, nausea, vomiting, diarrhea), as well as bleeding and general health deterioration. Pain-related complaints and neurological symptoms such as dizziness were also represented. Again, the relative proportions of these symptom complexes were consistent between ADRs with and without MEs.

2.7. Predictors of Medication Errors

In multivariable logistic regression, multimorbidity emerged as the only significant predictor of medication errors. In the binary model, multimorbidity was associated with a modestly reduced risk of MEs (OR = 0.84; 95% CI 0.71–1.00; p < 0.05). Neither sex, polypharmacy (≥5 medications), nor age group (<65 years as reference) showed significant associations.

The continuous model, which included age in years, number of diagnoses, and number of medications, likewise identified no significant predictors. Odds ratios for these variables all approximated unity, with confidence intervals crossing 1.0. Thus, across both modeling approaches, multimorbidity was the only factor linked to ME occurrence, and the association was protective rather than predisposing (Figure A1).

Analyses of LOS, adjusted and unadjusted, revealed no systematic difference between ADRs and MEs (Figure A2). Once an ADR occurred, its trajectory was essentially the same, whether or not an error had triggered it.

3. Discussion

This prospective multicenter study provides one of the most comprehensive assessments of MEs in German emergency care. Among nearly 8,000 ADR-related admissions, 20.1% were attributable to preventable medication errors, underscoring their substantial contribution to the burden of drug-related morbidity and highlighting a relevant and avoidable fraction of harm within routine pharmacotherapy.

A central finding of this study is that ADRs caused by MEs were not clinically milder or distinct from ADRs without error. Symptom burden, discharge outcomes, and triage severity overlapped almost completely, with only minor differences in selected MedDRA-coded symptoms. Given the substantial burden of disease associated with ADRs, the high proportion of preventable MEs represents a major opportunity to reduce both patient harm and healthcare costs attributable to drug therapy. For frontline clinicians, these findings indicate that an underlying ME cannot be inferred from clinical presentation alone. Instead, the ED is the site where consequences of errors manifest, not where their origins can be identified—highlighting that prevention must occur upstream.

Medication errors clustered predominantly around high-volume, everyday therapies rather than rare or classical “high-alert” drugs. Proton pump inhibitors, diuretics, beta-blockers, ACE inhibitors, and anticoagulants accounted for a substantial share of ME-related ADRs. When drugs were suspected of contributing to an ADR, the likelihood of a ME was often significantly higher, particularly for agents such as torasemide, metoprolol, ramipril, phenprocoumon, and ibuprofen. This emphasises that prevention strategies must not be restricted to exceptional or high-risk substances, but instead target routine prescribing and monitoring of the most commonly used chronic medications.

The Schumock analysis revealed that most preventable ADRs were linked to basic process failures. Inappropriate dosing (30%), lack of adherence (28%), overlooked contraindications (26%), and missing monitoring (20%) were the dominant preventability criteria, while drug interactions, allergies, and toxic serum levels occurred less frequently. Importantly, although most cases involved a single preventability criterion, a relevant minority showed multifactorial origins, most frequently the combination of dosing plus missing monitoring (3.2%) or the triple of contraindication, dosing, and monitoring (0.8%). These findings reinforce the need to strengthen routine safeguards—dose adjustment by renal function and age, structured laboratory monitoring, and adherence support—rather than focusing narrowly on rare or exotic error types.

Although medication-related harm is highly heterogeneous and context-dependent, the drug-specific symptom complexes identified in this study do not support the notion of attenuated effects, but instead point to clinical manifestations of comparable severity between MEs and ADRs. For example, phenprocoumon-related cases were dominated by bleeding regardless of whether an error was present, and torasemide clustered around electrolyte and volume disturbances in both ADR and ME cases, with potentially severe consequences for the patients. Across metoprolol, ramipril, and ibuprofen, the patterns mirrored expected pharmacological adverse effects, largely unaffected by ME status. This finding aligns with the broader observation that clinical presentation is driven by the drug’s inherent risk profile, which underlines the potential benefit of preventability of ME.

Regression analyses identified multimorbidity as a modest protective factor against MEs (OR 0.84, 95% CI 0.71–1.00). In contrast, polypharmacy, age, sex, number of diagnoses, and number of medications were not significant predictors. This paradoxical association may reflect that multimorbid patients are managed more cautiously, with closer clinical supervision and structured follow-up, thereby reducing error risk despite their higher baseline vulnerability. The finding challenges the assumption that complexity automatically translates into higher ME rates, and instead suggests that quality of management is more decisive than sheer disease or drug counts.

However, this observation may also be influenced by detection and selection biases. Firstly, MEs may be more readily identified in patients with fewer admission diagnoses, where clinical constellations are less complex and deviations from standard therapy are easier to detect. Second, as the study design involved explicit screening for MEs, patients with simpler clinical profiles may have been more likely to be flagged and included, potentially enriching the cohort with less complex cases.

Smoking and alcohol consumption were more frequently observed among ME cases in univariate analyses; however, due to substantial missing data and the lack of consistent associations in multivariable models, these findings should be interpreted cautiously.

Regarding outcomes, ME-related ADRs were not associated with a more favorable course. While most discharge categories did not differ markedly between groups, permanent damage was documented more frequently in ME-related cases. This is particularly concerning, as it suggests that preventable errors may contribute disproportionately to irreversible harm.

Length of hospital stay (LOS) analyses showed only modest effects. While unadjusted median LOS did not differ significantly between ME and non-ME cases. Adjusted models indicate a small and statistically significant prolongation of hospital stay among ME-related cases. with a 4–5% longer hospital stay, corresponding to an average increase of 0.37 days. Although the absolute difference is small, at a population level this translates into a meaningful cumulative burden.

To our knowledge, this study is the first analysis comparing ADRs caused by medication errors with the same suspected drugs in the absence of documented errors. The strengths of this study include its prospective multicenter design, large sample size, and systematic assessment of ADR causality and preventability using standardized criteria. Limitations include exclusion of omission errors from the main analysis, potential under-recognition of subtle MEs, incomplete documentation of lifestyle factors, and restriction to German EDs, which may limit generalizability. Moreover, while the Schumock assessment provides structured insights into preventability, it relies on documented information and cannot capture all contextual clinical decision-making.

Taken together, our findings demonstrate that medication errors drive a significant share of ADR-related ED admissions, but are not less severe and even not clinically recognizable once harm has occurred. Errors arise mainly from routine prescribing and monitoring processes in common drugs, not from rare or exotic scenarios. This calls for system-level interventions upstream in care—electronic prescribing support, structured medication reviews, pharmacist involvement, monitoring protocols, and adherence support. By focusing on everyday therapies and process safeguards, a considerable proportion of drug-related hospitalizations could be prevented.

4. Materials and Methods

4.1. Study Design and Setting

The ADRED study is a prospective, multicenter cohort that was conducted in six central EDs of German maximum-care hospitals between 2015-2021. Each center handles ~between 10,000-90,000 ED admissions annually, providing a robust base for studying drug-related presentations.

4.2. Patient Population

Adult patients presenting unplanned in the ED were screened for causality analysis of their symptoms with medication. Inclusion required at least “possible” causality by established pharmacovigilance criteria (WHO-UMC causality assessment criteria [14]).

4.3. Data Collection

Patients presenting to the ED were screened by trained study personnel (e.g., hospital pharmacists, medical doctors) for drug-related presentations. Cases were included if the ED visit was attributable to an ADR based on a structured causality assessment. All cases with a WHO-UMC causality score of ≥4 (possible, probable, certain) were reviewed for preventability using the Schumock and Thornton criteria [27]. The following factors were assessed:

• Drug inappropriate for the patient’s condition (contraindication)
• Dose, mode of administration, or frequency inappropriate for the patient’s age, weight, or health condition
• Missing monitoring or missing laboratory tests
• Known allergy or intolerance to the drug
• Drug interaction
• Toxic serum level documented
• Lack of adherence or compliance

Each case could be assigned more than one criterion. In addition to the prevalence of single factors, we also analyzed co-occurrence patterns of criteria, reporting the most frequent combinations.

Data were documented in a standardised case report form (CRF), which contained patient demographics, comorbidities, current and suspected medications, diagnoses, ED triage category (Manchester-Triage-System), laboratory results, and discharge outcomes. Symptoms were coded using MedDRA at the preferred-term (PT) level.

Several variables required for the present analyses were derived subsequently by the authors from the eCRF data. For example, length of stay (LOS) was calculated from admission and discharge dates, and measures of polypharmacy (according to Anatomical Therapeutic Chemical (ATC)-Code) and multimorbidity (according to International Classification of Diseases (ICD)-10 code) were generated from the documented medication and diagnosis fields.

4.4. Symptom Clustering

For the cluster analysis, only drugs that were suspected of contributing to an ADR were included, defined as cases with a WHO-UMC causality score of 4 (possible), 5 (probable), or 6 (certain).

Reported symptoms were coded at the MedDRA PT level. To capture higher-level patterns, PTs were first grouped by system organ class (SOC). For each drug, these SOC groupings were then used to generate a tailored symptom complex, representing the characteristic spectrum of ADR manifestations plausibly linked to that drug.

The same procedure was applied separately to ADR cases without MEs and to ADR cases with MEs, resulting in parallel drug-specific symptom complexes. This allowed for direct comparison of symptom profiles between ADRs caused by MEs and ADRs not related to errors, while ensuring that the complexes reflected the pharmacological effects of the respective drug.

For the cluster analysis, we selected phenprocoumon, torasemide, metoprolol, ramipril, and ibuprofen. These drugs were chosen because they were among the most frequently implicated in ADRs in the cohort and represent different pharmacological classes with distinct toxicity profiles (anticoagulant, diuretic, beta-blocker, ACE inhibitor, NSAID). This selection allowed us to examine whether symptom clusters might differ between MEs and non-MEs across pharmacologically diverse agents.

4.5. Definitions

• Polypharmacy: Concomitant use of ≥5 drugs.
• Multimorbidity: ≥2 chronic conditions.

4.6. Statistical Analysis

Continuous variables were summarized as mean, median, and range; categorical variables as counts and percentages. Group comparisons were conducted using Chi² tests for categorical variables and Mann–Whitney U tests for continuous variables. Logistic regression models explored predictors of MEs: a binary model included sex, age group (<65, 65-79, >79), polypharmacy, and multimorbidity, while a continuous model used age (years), number of diagnoses, and number of drugs. Combination drugs were counted per active substance, with each ingredient considered as a separate drug. Odds ratios (OR) with 95% confidence intervals (CI) were reported. Hospital length of stay (LOS) was analyzed with unadjusted and adjusted linear regression models, with LOS log-transformed to account for skewness. Symptom clustering was performed by grouping MedDRA PT terms into clinically coherent complexes and further stratified by suspected drug involvement.

All analyses were performed in Python (v3.10). Data management and descriptive statistics were carried out using pandas (v2.2.2) and numpy (v1.26.4). Hypothesis testing used scipy (v1.13.1), while regression analyses were performed with statsmodels (v0.14.2). Visualization was conducted with matplotlib (v3.9.0) and seaborn (v0.13.2), with supplementary exploration supported by plotly (v5.22.0). Figures were exported in publication-ready format using custom Python scripts.

5. Conclusions

Medication errors accounted for nearly one in five ADR-related emergency admissions in this large prospective multicenter cohort. Their clinical presentation was indistinguishable from ADRs without error involvement, and they clustered around common chronic medications rather than rare high-risk drugs. Preventability was mainly linked to dosing errors, contraindications, missing monitoring, and non-adherence, often in combination, highlighting failures of everyday prescribing and follow-up processes.

Regression analyses identified multimorbidity as a modest protective factor, suggesting that closer medical supervision may offset risks, while length-of-stay analyses showed even a modest increase in hospitalization burden for ME cases.

These findings emphasize that prevention must focus upstream in routine care—through structured medication reviews, electronic prescribing support, pharmacist involvement, and adherence monitoring—rather than attempting to detect errors at the point of ED presentation. Strengthening these safeguards in everyday therapies offers the greatest potential to reduce preventable harm and the cumulative burden of drug-related hospitalizations.