Hospital admissions due to adverse drug reactions and adverse drug events in older adults: a systematic review

Abstract

Background

Adverse drug reactions (ADRs) and adverse drug events (ADEs) are consistently reported to cause up to 30% of hospital admissions in older adults, resulting in significant morbidity, mortality and an added health economic burden.

This systematic review aimed to establish the frequency of ADRs and ADEs as a cause of hospitalisation in older adults.

Methods

Standard databases and citations were searched from 2015 to 2025. Studies specifically assessing ADR and ADE prevalence and risk factors in older adults were included. The review was registered in PROSPERO (CRD42024613426) and quality was assessed using the Joanna Briggs Institute criteria with bias determined via ROBINS-E.

Results

Nine studies underwent full evaluation. The reported prevalence of ADRs ranged from 3.3% to 23.1% and ADEs ranged from 11.75% to 18% as causes of hospitalisation. The median age of those included, in the respective studies, was between 77 and 86 years. Falls (19.4%–20.9%), delirium (7.3%–12.9%) and bleeding (8%–30.2%) were the most frequently encountered ADR/ADEs causing hospitalisation. Anti-thrombotics (11.5%–30.2%) diuretics (14.7%–30.2%) and renin-angiotensin-aldosterone system (RAAS) inhibitors (7.5%–8.9%) accounted for the highest proportion of ADR/ADE causative agents.

Conclusion

This review has limitations stemming from the heterogeneity of the included studies and the exclusion of the grey literature. However, ADRs and ADEs remain a significant cause of hospital admissions in older patients. The possible preventability of these and their possible hospital admission avoidance highlights the critical need for further research into ADR/ADE risk prediction methods in addition to co-morbidity and medication optimisation for older adults.

Key Points

• Adverse drug events (ADEs) and adverse drug reactions (ADRs) are consistently reported to be the cause of up to 30% of hospital.

• Falls, delirium and bleeding are the most frequently encountered ADR/ADEs in older adults.

• Future research into methods for risk assessment and prevention of ADEs and ADRs is urgently needed.

Introduction

The widely accepted definition of an adverse drug reaction (ADR) is that of an ‘appreciably harmful or unpleasant reaction, resulting from an intervention relating to the use of a medicinal product’ and ‘occurs at doses normally used in man for the prophylaxis, diagnosis, or therapy of disease or for the modification of physiologic function’ [1–3], which differs from that of an adverse drug event (ADE), ‘an unwanted occurrence after exposure to a drug that is not necessarily caused by the drug’ [3, 4]. Therefore, ADRs can be thought of as ADEs with a causal link to a medicine (Figure 1). Both ADRs and ADEs are common and mostly preventable. Their prevalence is reported to represent 5%–30% of hospital admissions with the variability in prevalence possibly due to heterogeneity in methodologies, definitions used and whether the ADR/ADE is deemed causative or contributory to the admission [5]. The Adverse Drug Reactions in an Ageing Population (ADAPT) study conducted in Ireland reported the prevalence of ADRs as at least contributing to 10% of hospital admissions, with over 70% of these being deemed to be preventable [6]. Not only are ADRs a significant cause of hospitalisation but they can have further important economic impacts. A hospital admission secondary to an ADR can prolong length of hospital stay by up to 12 days [7]. The cost of ADRs and their subsequent effects are significant in the US (30 billion dollars annually), the UK (700 million euro annually) and Germany (435 million euro annually) with a single ADR costing 2250 to 9538 euro per event [8–11].

Figure 1.

Relationship between adverse drug reactions (ADRs), adverse Drug Events (ADEs) and medication errors.

ADEs and ADRs are a common cause of significant morbidity and mortality. One study reported that over a quarter of patients who experience an ADE have residual disability as a direct result of the ADE but it can be higher depending on the presenting complaint, e.g. 56% of patients presenting with bleeding have residual disability [12]. With the ageing process there are changes in pharmacokinetics and pharmacodynamics which increase the risk of ADRs and ADEs in addition to rising polypharmacy [13, 14]. In this population ADRs and ADEs also pose a diagnostic challenge as they can resemble symptoms of geriatric syndromes, e.g. falls, incontinence and cognitive decline [15]. With an ageing population worldwide, it is critical to understand the extent of ADRs and ADEs in older persons in order to develop an effective prevention strategy.

Aims

The primary aim of this study is to systematically review the literature to establish the frequency of ADRs and ADEs as a cause of hospitalisation in older adults with a secondary aim of determining the implicated drugs and risk factors.

Methods

Search strategy and study selection

A systematic search for published material was performed from January 2015 to May 2025 (developed by a librarian), using standardised databases, namely Cochrane, OVID Medline and Embase, to identify relevant studies. The search combined keywords relating to medication harm, older adults and hospitalisation. The full search strategy is provided in Appendix 1 in the Supplementary Data section and was registered on PROSPERO CRD42024613426 [16]. All titles identified through the computerised search were screened independently by two reviewers (FB and IF) and unrelated studies were discarded. The remainder of the studies’ abstracts were then assessed independently by two reviewers (NC and FB) for their inclusion. A thorough search of reference lists of all reviewed articles and of primary studies was performed to find studies not identified in the computerised searches. A PICOT statement was developed to clearly define the inclusion criteria [17].

• Population: Older adults (defined as ≥65 years)

• Intervention/exposure: ADR or ADE

• Outcome: Hospitalisation

• Time: January 2015 to May 2025

Exclusion criteria applied in this systematic review included studies which focused on poisonings, drug abuse and studies which assessed interventions to either identify or ameliorate ADRs/ADEs.

Data extraction

All relevant information was independently extracted by both reviewers (FB and NC) using a common data extraction template. Data extraction related to the following:

• (i) Study characteristics—Sample size, population, study setting and study design

• (ii) ADR/ADE characteristics—definitions and criteria for defining and establishing causality, severity, preventability.

• (iii) Medications: implicated drugs and presenting feature of ADR/ADE.

Data analysis

ADR and ADE percentages were calculated for each study if not specifically reported. The systematic review was undertaken according to the Preferred Reporting Items for Systematic Reviews and Meta-analysis (PRISMA) checklist (Figure 2: Prisma Flow Diagram) (the full details can be viewed in Appendix 2 and Appendix 3 [18]). To accommodate the expected heterogeneity in the studies, a narrative synthesis approach was adopted to amalgamate and interpret the results across the studies. Any instances of missing or unclear information were reported in their original form, without making assumptions about the nature or content of the absent data unless explicitly stated otherwise.

Figure 2.

PRISMA flow diagram.

Quality assessment

The Joanna Briggs Institute (JBI) checklist for analytical cross sectional studies was also used to assess quality [19]. This was performed independently by two reviewers (N.C. and F.B.). Quality assessment of the reporting was undertaken using ‘the strengthening the reporting of observational studies in epidemiology’ (STROBE) guidelines and checklists [20]. This process was undertaken independently by two reviewers (S.L. and I.F.) and all discrepancies were discussed with a third reviewer (N.C.).

Risk of bias assessment

The risk of bias (ROB) in non-randomised studies—of exposure (ROBINS-E) was applied to each of the included studies independently by two reviewers (N.C. and F.B.). This tool evaluates the ROB across seven domains: (i) confounders, (ii) participants, (iii) exposure measurement, (iv) post-exposure interventions, (v) missing data, (vi) outcome measurement and (vii) reporting [21].

Results

Study selection

From the literature search, 15 460 titles were initially identified, of which 10 999 titles were screened. 23 studies associated with the primary outcome were retrieved for further evaluation. Of these, nine studies met the inclusion criteria and underwent a complete evaluation [6, 12, 22–28].

Characteristics of included studies

Table 1 outlines the study characteristics. The specific objectives of the included studies were the prevalence of hospitalisation due to ADRs in five publications [6, 12, 25–27], ADEs in three publications [22, 24, 28] and one study assessed both ADRs and ADEs causing hospitalisation [23]. All nine studies included participants aged ≥65 years and two studies included only those who were admitted from a long-term care facility [22, 27]. The studies were conducted in eight different countries in Europe and Australia (Australia [22, 25], Finland [23], France [27, 28] Spain [26], Czech Republic [24], Italy [12] and Ireland [6]). The number of patients included in each study ranged from 263 to 60 263 and the study duration ranged from 4 months to 7 years. Six studies [12, 22, 23, 26, 27, 28] were conducted with a retrospective design based on medical records or database and three were prospective studies conducted on inpatient medical wards [6, 24, 25].

Table 1.

Characteristics of included studies.

Study name	Study design	Patient characteristics	Study outcomes and definitions used	ADR/ADE assessment	Common ADEs/ADRs and implicated drugs	Identified risk factors
Ali et al. 2024 [22]	Australia Retrospective cross-sectional study using digital medical records in 4 Tasmanian public hospital July 2018 to June 2021 ≥ 65 years long-term care residents with unplanned admission to hospital	N = 500 Age: 83 years (median; IQR 76–89 years) Gender: Female 60.4% (n = 302) Co-morbidities: Hypertension 54.2% (n = 271) Rheumatologic disease 35.4% (n = 177) Cognitive Impairment 31% (n = 155) Medication Related: Potentially inappropriate medicines (Beers criteria) median 1, range 1–2 Anticholinergic burden score median 3, range 2–5 Previous ADR 11.4% (n = 57)	ADEs: WHO definition ADE 18% (n = 91) Causality: Naranjo criteria: 64% possible ADRs Preventability: Schumock and Thornton: 34% ADRs definitely or probably avoidable Severity: Hartwig and Siegel; 99% ADRs moderate severity	Chart review by Pharmacist and ADEs assessed by pharmacy researchers and Geriatrician	Fall 20.9% n = 19 Diuretics Hypnotics Sedatives Opioids RAAS Hypotension 17.6% (n = 16) Diuretics Antihypertensives Confusion/delirium 11% (n = 10) Opioids Hypoglycaemia (8%) Rectal or gastrointestinal bleeding (8%)	Logistic regression Rheumatologic disease (OR 1.89, CI 1.09–3.30, P = .024) Previous ADR (OR 12.91, CI 8.84–24.37, P□.01) Number of diagnoses (OR1.00, CI 0.87–1.15, P = > .99) Potentially inappropriate medicines (OR 1.22, CI 0.92–1.96, P = .18) Anticholinergic burden score (OR1.01, CI 0.90–1.15, P = .82) GerontoNet Risk Score (OR 1.12, CI 0.99–1.27, P = .08)
Bres et al. (2023) Poster abstract [27]	France Retrospective, cross-sectional study 2019 Elderly patients in nursing homes admitted to hospital (age not specified)	N = 263 Age: 87.1 years (mean) Co-morbidities: 52.1% with three co-morbidities but details not reported Polypharmacy: 85.6%	Warnings and Precautions section of the drugs summary of product characteristics ADR: 17.8% Predictability: 80% predictable	NR	NR	NR
Cahir et al. 2023 [6]	Ireland Prospective cross-sectional study at tertiary hospital November 2016 to June 2017 Patients age: ≥65 years admitted to an acute hospital	N = 3760 screened (N = 814 for analysis; ADR 377, non-ADR 437) Age: Mean 79.9 (SD7.3) in ADR; 81.6 years (SD7.7) in non-ADR admissions Gender: Female 51% (n = 184) ADR group; 53.1% (n = 233) non-ADR admissions Co-morbidities: Chronic lung disease 19.1% (n = 33) Heart failure 14.1% (n = 51) Cerebrovascular disease 13.3% (n = 48) Medication Related: Polypharmacy 87.2% (n = 315)	ADR: Edwards and Aronson ADR: 10% (n = 377) Causality: WHO (possible 72.9%, probable 20.5%, definite 6.1%), Naranjo (possible 77.4%, probable 20.5%, definite 2.1%) and Liverpool Algorithm (possible 66.3%, probable 23.4%, certain 10.4%) Preventability: Hallas; 11.4% definitely preventable, 59.7% possibly preventable, 28.9% unavoidable Severity: Hartwig Scale; 92.8% moderate severity, 7.2% severe	Multifaceted review of hospital admission by Consultant Geriatrician, two pharmacists and a research nurse using a previously validated screening process	Bleeding disorders 30.8% n = 116 Aspirin Warfarin Gastrointestinal disorders 23.3% n = 88 Respiratory disorders 16.7% n = 63 Falls and Syncope 13.3% n = 50 Renal and urinary disorder 11.1% n = 42	Multivariable analysis Immobility (AOR 2.33, CI 1.01–5.38) Previous fall (AOR 0.25, CI 0.10–0.64) Frailty (TRS) (AOR 2.51, CI 1.39–4.53) Delirium (DSM-4) (AOR 1.63, CI 1.06–2.50) Chronic lung disease (AOR 0.51, CI 0.28–0.94) Anticoagulants (AOR 2.00, CI 1.35–2.97) Antiplatelets (AOR 1.64, CI 1.13–2.38)
Dauny et al. 2025 [28]	France Retrospective cohort study in a geriatric medicine unit 2023 Patient age criteria not reported	N = 483 Age: Mean 86.7 years (SD 6.15) Gender: Female 59% Co-morbidities: Charlson Co-morbidity Index score median 6 (IQR 5–8) Hypertension 72% (n = 349) Diabetes mellitus 25% (n = 119) Medication related: Median number of drugs taken 7 (IQR 4–9)	ADE: ‘any adverse incident resulting from medication usage that causes harm or injury to the patient including adverse drug reactions and medication errors, such as those associated with overuse, underuse and misuse of drug therapy’ ADE: 41% (n = 198) by trigger tool DRA: 43% (n = 207) by trigger tool DRA: n = 113 by geriatrician review Causality: WHO Drug-related admission: main or contributory reason Preventability determined using STOPP/START criteria: 70% preventable	Initial review by one Geriatrician to determine if admission was drug-related Retrospective chart review independently by two Geriatricians using AT-HARM trigger tool	Falls/Fractures 33% Bleeding 23% Delirium 14% Diuretics 21% Renin-angiotensin system inhibitors 20% Direct oral anticoagulants 15%	NR
Laatikainen et al. 2016 [23]	Finland Retrospective cross-sectional study at Oulu University Hospital Finland using emergency department (ED) patient records 2014 Patients age: >65 years	N = 290 Age: 77 years (mean), range 65–95 years Gender: NR Co-morbidities: NR Medication related: Polypharmacy 71% (n = 206)	ADR: WHO definition ADE: Edwards and Aronson ADRs 23.1% (n = 67) Causality: Customised Naranjo: 38 probably, 29 possibly Preventability: NR Severity: NR	ED patient record review by MDT team of pharmacist, clinical pharmacologist and health science researcher	Falling, vertigo, fractures (19.4%) Opioids, antipsychotics and centrally acting agents Bleeding (12%) Disorientation, delirium (8.9%) Opioids, benzodiazepines and centrally acting agents Constipation (8.9%) Opioids and anti-cholinergic agents
Marikova et al. 2020 [24]	Czech Republic Prospective cross-sectional observational study using electronic hospital records April to December 2017 Unplanned hospital admission to a geriatric medicine ward (age limit >78 years)	N = 336 Age: 86 years (median) Gender: Female 65.6% Co-morbidities: NR Medication related: Polypharmacy 78.4% (n = 287)	ADE: Nebeker definition and includes ADR and medication errors ADEs: 11.75% (n = 43) Causality: WHO-UMC criteria: 76.7% probable, 23.3% possible Preventability: Modified Schumock and Thornton: 53.5% potentially preventable Severity: NR	Geriatric Ward inpatients on a weekly basis Causality assessment performed by clinical pharmacist	Gastrointestinal haemorrhage and ulcerations 30.2% (n = 13) Antithrombotics Hyponatraemia 23.3% (n = 10) Diuretics Digoxin intoxication 11.6% (n = 5)
Ognibene et al. 2018 [12]	Italy Retrospective observational study using electronic medical records of internal medicine unit inpatients July 2015 to June 2016 Patients age: >65 years	N = 1750 Age: 83.5 years (median; IQR 78–87 years) Gender: Female 55.7% (n = 53) Co-morbidities: only reported for those with an ADR Hypertension 65.1% (n = 69) Cardiovascular diseases 60.4% (n = 64) Dementia 50% (n = 53) Medication related: Polypharmacy 56.6%	ADR: WHO definition ADRs: 6.1% (n = 106) Causality: Naranjo: 16.5% (n = 28) probable, 65.1% (n = 69) possible and 8.5% (n = 9) definite Preventability: NR Severity: Karch and Lasagna; 50% moderate, 27.4% severe, 10.4% lethal	Internal medicine unit inpatient chart review	Electrolyte disorders 16.5% (n = 28) Diuretics, psychoactive drugs Dehydration 15.9% (n = 27) Diuretics Bleeding 14.7% (n = 25) Vitamin K antagonists, antiplatelets, DOACs Altered state of consciousness 12.9% (n = 22) Benzodiazepines, antidepressants Acute Kidney injury Diuretics, ACE-inhibitors Syncope 8.8% (n = 15) Antihypertensives, diuretics
Parameswaran et al. 2017 [25]	Australia Prospective cross-sectional study using medical records and patient interviews in two acute hospital medical wards March 2014 to March 2015 in one hospital September 2015 to December 2015 in one hospital Patients age: >65 years	N = 1008 Age: 81 years (median; IQR 74–86) Gender: Female 53.4% Co-morbidities: Cardiovascular disease 89.1% Renal failure 52.7% Dementia 6.8% Medication-related: Number of medications 10 (median; IQR 7–14) Previous ADR 58.3%	ADR: WHO definition ADR: 18.9% Causality: Naranjo: 24.1% possible, 70.1% probable, 5.8% definite Preventability: Modified Schumock and Thornton 88.5% Severity: Hartwig et al.; 97.5% moderately severe, 2.1% lethal	Chart review and patient questionnaire/interview performed by pharmacist	Hypotension, orthostatic hypotension, syncope 33.5% (n = 64) Diuretics, RAAS inhibitors, beta-blockers, calcium channel blockers Renal impairment 25% (n = 49) Diuretics, RAAS inhibitors Dizziness 23% (n = 44) Diuretics, RAAS inhibitors, beta-blockers, calcium channel blockers Haemorrhage 11.5% (n = 22) Antiplatelets, anticoagulants Confusion 7.3% (n = 14) Benzodiazepines, opioids, anticonvulsants, antidepressants
Pedros et al. 2015 [26]	Spain Retrospective cross-sectional study using institutional database January 2008 to December 2014 Patients age: >65 years	N = 60,263 Age: NR Gender: NR Co-morbidities: NR Medication related characteristics: NR	ADR definition: European Commission ADR: 3.3% (n = 1976) Causality: NR Preventability: NR Severity: Modified Karch and Lasagna; 10.2% fatal	ADRs identified through systematic daily review of admission diagnosis and reviewed by clinical pharmacologist	Acute renal failure 22.9% (n = 453) RAAS inhibitors, diuretics Upper gastrointestinal bleeding 16.6% (n = 328) Antiplatelet drugs, vitamin K antagonist, NSAIDs Lower gastrointestinal bleeding 11.2% (n = 221) Vitamin K antagonists Intracranial bleeding 9.3% (n = 184) Vitamin K antagonists Digitalis intoxication 3.7% (n = 74) Digoxin

IQR: interquartile range, RAAS: renin-angiotensin-aldosterone system agents, OR: Odds Ratio, CI: Confidence Interval, AOR: adjusted odds ratio, TRS: Triage risk Stratification tool, DMS: diagnostic and statistical manual of mental disorders, NR: not reported, ACE: angiotensin converting enzyme, NSAIDs: non-steroidal anti-inflammatory drugs.

Patient characteristics

Participant age was reported as a mean in four studies with it ranging from 77 to 87.1 years in the respective studies [6, 23, 27, 28] and the median age ranged from 81 to 86 years in studies using this statistic [12, 22, 24, 25]. One of the studies was conducted on a geriatric medicine ward where the age limit for admission is ≥78 years. In the six studies reporting gender, females constituted a higher proportion in all (51%–65.8%) [6, 12, 22, 24, 25, 27].

Patient co-morbidities were reported fully in four studies [6, 22, 25, 28]. The most common co-morbidities related to the cardiovascular system (hypertension 54.2% [22] to72% [28], heart failure 14.1% [6]), neurological system (cognitive impairment/dementia 6.8% [25] to 51% [28]) and others including rheumatological diseases (35.4% [22]) and cerebrovascular diseases (25% [28]). Co-morbidities were reported only in those with an ADR in one study [12].

Medication-related factors reported in the studies included polypharmacy [6, 12, 23, 24, 27], potentially inappropriate medications as per Beer’s criteria [22], and previous ADR [22, 25]. The prevalence of reported polypharmacy ranged from 56.6% [12] to 87.2% [6]. Previous ADRs were identified in 11.4% [22] to 58.3% [25] of cases.

ADR/ADE definitions and results

The definitions of ADR and ADE varied with the most common being the WHO definition followed by Edwards and Aronson with one study using their own definition. The prevalence of ADRs was reported to be 3.3% [26] to 23.1% [23] and the reported prevalence of ADE ranged from 11.75% [24] to 41% [28]. To assess causality, the most frequently used algorithms were Naranjo (four [12, 22, 23, 25]) and WHO-UMC (two studies [24, 28]) with one study using three different assessment algorithms [6]. Causality was not reported in one study [26].

Preventability was assessed using Schumock and Thornton in three studies and one using Hallas criteria [6], with 11.4%–88.5% reported as possibly preventable to definitely preventable [6, 22, 24, 25]. The study by Dauny et al. used the STOPP/START criteria as their preventability assessment [28]. Severity was assessed in five studies with two using Karch and Lasagna [12, 26] and three studies used a method developed by Hartwig et al. [6, 22, 25]. Severity was at least of moderate severity in 50%–99% of cases.

ADR/ADE presentation and implicated medications

All studies, bar one [27], detailed the presenting ADR/ADE and implicated drugs. Falls (19.4%–33%), hypotension (7.6%–33.5%), delirium (7.3%–14%) and bleeding (8%–30.2%) were the most commonly encountered ADRs and ADEs. Anti-thrombotics (including antiplatelets and anticoagulants) (11.5%–30.2%), diuretics (14.7%–30.2%) and renin-angiotensin-aldosterone system (RAAS) inhibitors (7.5%–20%) accounted for the highest proportion of ADR/ADE causative agents.

Risk factors

Only two studies performed logistic regression to establish risk factors for ADR/ADE hospitalisations, which were reported as rheumatologic diseases and previous ADR in the study by Ali et al and immobility, previous falls, anticoagulant or antiplatelet therapy in the study by Cahir et al with these being statistically significantly associated with an increased risk of ADRs [6, 22]. Additionally, the study by Ali et al. also applied the GerontoNet Risk score [29] which showed that a higher score was associated with higher risk of having an ADR responsible for hospitalisation [22].

Quality assessment

As per STROBE guidelines (full details can be found in Appendix 4 of the supplementary data section) and JBI critical appraisal (full details can be found in Appendix 5 of the supplementary data section), all studies included relevant details relating to abstract, introduction, discussion and funding. Six studies detailed their variables [6, 12, 21–24] with one providing incomplete information [25]. One study provided a sample size calculation [22]. Three studies lacked data on statistical methodology [24, 25, 27]. The remaining studies provided varying degrees of information regarding their statistical methods [6, 12, 22, 23, 26]. No study provided details on missing data. One study provided adjusted estimates for their variables [6].

Risk of bias

The ROB was assessed using ROBINS-E (Figure 3) [21]. This was variable when assessing each of the domains and the overall ROB. One study was found to have a low ROB [6], two high level of ROB [24, 27] and the remainder medium ROB [12, 22, 23, 25, 26, 28].

Figure 3.

ROB as determined using the ROBINS-E tool.

Discussion

This systematic review identified nine studies meeting prespecified inclusion criteria over the 10 year search period and has found an ADR/ADE rate in older adults (3.3%–43%) similar to previous reviews and meta-analysis [6, 30]. The variability can be explained by the different definitions used as well as the clear difference between an ADR and ADE, particularly the lack of a causal link of an ADE to a drug [1]. Another potential confounding factor in the variable rates of ADR/ADE are the differing study designs. Prospective studies with large sample sizes provide the most accurate reflection of true rates of ADR/ADEs. Three studies used this approach and had sample sizes of 336 [24], 1008 [25] and 3760 [6] with an ADE prevalence of 11.75% [24] and an ADR rate of 10% [6] and 18.9% [25] respectively, which reflects the results from a previous meta-analysis [30] and most likely reflects real world prevalence.

The studies reviewed included patients admitted acutely to hospital but varied in terms of the site of admission, e.g. emergency department, geriatric medicine inpatient ward with one study specifying long term care facility residents as an inclusion criterion. Ali et al. included patients who were resident in care facilities prior to admission and had an ADE prevalence rate of 18% [22]. This may be lower than expected as patients in residential care tend to be frailer, with higher levels of co-morbidity and resulting polypharmacy and it could be envisaged that subsequently they would have higher ADE rates [31–33]. However, the reason for this somewhat lower than expected rate could be explained by the higher incidence of advanced care planning in these patients with a documented instruction not to be transferred to an acute hospital setting in the event of a medical deterioration [34]. It could be also explained by potentially increased rationalisation and deprescribing in this cohort, but further research should be undertaken in this vulnerable cohort to fully elucidate ADR/ADE rates and their resulting morbidity and mortality in the setting of care homes.

Marikova et al. conducted their study on an inpatient geriatric medicine ward with an admission criterion of age > 78 years [24]. They calculated an ADE rate of 11.75% which could also be considered surprising due to the older patient cohort. However, this may be explained by limiting the study to a single geriatric medicine, as the authors identified that their data may exclude possible ADEs related to falls, acute kidney injury or psychiatric conditions as these presentations would be admitted to other specialty wards. Expansion of their study to these wards may result in a more real-world prevalence of ADE-related hospitalisations.

The study by Dauny et al. report a significantly higher rate of ADEs (43% of admissions) [28] compared with those above [22, 24]. The difference in ADE rate may be explained by differences in methodologies used. The former used their own, broader definition of an ADE and a trigger tool to aid identification. They included all ADEs, which caused or contributed substantially to the participants hospital admission. Their systematic and thorough detection methodology that uncovered often-missed drug-related events, alongside a broad definition of what constitutes a drug-related hospital admission likely contributed to the higher rate of ADEs.

The wide prevalence range of ADR/ADE may also be explained by the complexity of this patient cohort. Older adults are frequently frail, multi-morbid and are exposed to polypharmacy with the added complexity of altered pharmacokinetic profiles, making the distinction of an ADR/ADE from disease challenging [15, 35]. Geriatric syndromes include delirium, impaired mobility and falls and incontinence amongst others [36]. A diagnostic challenge arises when a patient presents to the emergency department with any of these as they could also be attributed to an ADR as is evidenced from the studies included in this review. The most common ADR/ADEs are falls (8.8%–33.5%) and delirium (7.3%–14%) in five of the nine studies, which are common geriatric syndromes [12, 22, 23, 25]. Certain study designs are more effective at identifying these non-specific, yet significant, medication-related presentations. Prospective studies with comprehensive reviews and intensive monitoring [6, 25] or those with a multi-professional (physician and pharmacist) review tend to yield higher detection rates [6, 23]. Studies utilizing trigger tools can improve ADE detection by including non-specific presentations as a trigger [28]. Retrospective studies relying on identification of ADR/ADEs based on International Classification of Diseases (ICD-10) codes can underestimate total prevalence of ADR/ADEs although this was not reported to be the method used in the included studies.

Frailty was investigated in one study where it was assessed using two validated tools—the triage risk stratification (TRS) and the Prisma-7 [6]. A higher TRS score was associated with increased risk of ADR occurrence. This is also shown in the study by Cullinan et al. who reported a significant correlation between a patient’s frailty status and a propensity to develop and ADR [15]. Although many frailty measurement tools exist in the literature, including those used in the study by Cahir et al., tools used commonly in clinical practice should be employed to allow for generalizability to real-world populations [37, 38]. The most commonly used tool in clinical practice is the Rockwood Clinical Frailty Scale but this was not a used as a variable in any study [39]. Studies conducted in older adults, particularly in relation to medication-related harm should aim to include frailty markers as a component of baseline characteristics.

Multimorbidity and polypharmacy are consistently reported as being two of the main risk factors for incident ADR/ADEs. The existence of certain medical conditions also increases the risk of ADRs such as renal and liver impairment and these can be subclinical in older adults [40]. Multimorbidity can result in polypharmacy and this can increase the risk of drug–drug interactions and ADRs [41]. The risk of ADRs increases from 13% in a person taking two medicines to 58% when taking five and 82% when taking seven or more [42]. However, polypharmacy should not always be misinterpreted as inappropriate and a more accurate measurement would be of potentially inappropriate prescriptions (PIPs) using validated tools such as Beer’s criteria and Screening Tool of Older Persons’ Prescriptions/Screening Tool to Alert to Right Treatment (STOPP/START) criteria [43, 44]. The study by Dauny et al. used the STOPP/START criteria as their preventability assessment but this may have limitations as it is validated to assess for PIPs and not for preventability assessment [28]. Despite these well-known documented risk factors for ADR/ADEs in older adults, only four studies detail patient co-morbidities and only one study detailed PIP according to Beer’s criteria [22].

Frailty, multi-morbidity and polypharmacy can be considered as a bidirectional potentially causal relationship and it is essential that studies adjust for this relationship [45]. It may be hypothesised that superimposed frailty (on polypharmacy and multi-morbidity) would increase the risk of medication related harm but well-designed studies are lacking in the literature highlighting an area for future research [45].

Preventability was reported in four of the studies with up to 88.5% of ADR/ADEs being at least possibly preventable. This is a key aspect to consider when analysing ADR/ADEs as it highlights possible opportunity cost of increased morbidity, mortality and hospital admission. There is, however, a lack of clinically acceptable methods and tools for assessing ADR/ADE risk and prevention strategies. A recent meta-analysis has shown that current risk prediction models for ADRs are not reliable or validated and have highlighted the need for the development of such a tool [46]. A meta-analysis assessing the impact of pharmacist-led medication review on hospital admissions and deaths showed no effect on these outcomes in older adults [47] but a Cochrane review did reveal a reduction in hospital re-admissions whilst having no impact on mortality or health-related quality of life [48]. Further research is needed in this area to provide clinically beneficial methods of ADR/ADE risk reduction and prevention.

The quality assessment of this systematic review demonstrates a good overall quality of the studies included in relation to their abstracts, introductions and discussions. However, the quality of the statistical analysis methods and results sections were variable across the studies. A sample size calculation was performed in one study only [22] and therefore the power of the others could potentially be questioned. Missing data and potential sources of bias were poorly reported throughout the studies with only one study detailing bias [23], which can impact on the overall interpretation and generalizability of the study. The use of reporting guidelines such as STROBE and JBI would improve the overall quality of the studies ensuring robustness and reliable interpretation and generalizability.

This systematic review provides an up-to-date review of ADE and ADRs in older adults. A previous meta-analysis in 2017 found that NSAIDs were the medications most related with admissions due to ADR/ADEs followed by beta-blockers, warfarin and digoxin. This review has found that diuretics, RAAS inhibitors are now the most implicated medications. Digoxin was listed as a cause of ADR/ADE in two studies only [24, 26]. This may reflect the changing prescribing practices over time with recent recommendations against its routine use in heart failure [49]. Warfarin has largely been replaced in the management of non-valvular atrial fibrillation, systemic thrombosis and embolism with direct oral anticoagulants, which have more predictable pharmacokinetics and pharmacodynamics [50]. This reflects the need to conduct up to date systematic reviews and meta-analysis in this area due to the dynamic nature of therapeutics and the implementation of new drug regimens reflecting modern prescribing practices.

Future research

According to the Dutch national polypharmacy guideline, patients aged 70 and over using five or more medications should be screened upon emergency department admission to assess if an ADE is the cause for attendance [51]. Accurate detection and reporting of ADR/ADEs may be addressed by the use of a trigger tool such as that used in the study by Dauny et al. [28]. This approach offers a more streamlined and pragmatic approach to identifying ADR/ADEs over the gold standard of chart review with clinical adjudication and have the added benefit of detecting ADR/ADEs masquerading as geriatric syndromes. Many tools have been reported including the Global trigger tool [52], Optimising thERapy to prevent Avoidable hospital readmissions in Multimorbid older people (OPERAM) [53], ADR-tool [51], Quick Assessment of Drug-Related Admissions over Time (QUADRAT) and AT-HARM [54] [55] but further research into their ability to accurately diagnose ADR/ADEs in different populations is needed before they can be recommended for routine use in clinical practice.

Another approach to mitigating ADR/ADE risk could be the use of risk prediction tools. These could identify patients at the highest risk of experiencing an ADR/ADE and trigger their attending physician to undertake a full medication review. A number of these tools have been described in the literature, but none are routinely used in practice due to their limited evidence [46, 56]. Future research is therefore required to prospectively validate these tools in addition to implementation and impact assessments in real-world settings.

Limitations

This review found studies conducted in Europe and Australia. The age limit of 65 years is considered the current operational definition of an older person. However, it has been reported that this may be impacted by differences in life expectancy worldwide, e.g. it has been suggested that the age for older persons would be 50 to 55 years in Africa, 58 years in the Indian sub-continent and 60 years in China and our cut-off may have unintentionally excluded studies based in these areas [57].

A key limitation of this systematic review is the potential for publication bias. It excluded studies published in languages other than English and unpublished data, which could restrict the comprehensiveness of the review.

The heterogeneity amongst the studies included makes direct comparison and conclusions very difficult. This review included both ADE and ADR hospitalisations with differing definitions used for both [1]. The studies also used varying algorithms for causality, preventability and severity and the assessments were also conducted by a variety of professionals, e.g. consultant geriatricians, pharmacists, nurses and researchers, which may impact on the quality of the data and limit its use in comparison studies and meta-analysis. One study used multiple tools and undertook reviews by a number of members of the multidisciplinary team ensuring the study’s robustness and future use in comparison studies [6]. Studies conducted using databases may have an imbedded information bias and future studies should strive to perform prospective reviews using a multidisciplinary review approach.

Conclusion

ADRs and ADEs remain a significant cause of hospital admissions in older adults. The possible preventability of ADRs and ADEs, along with the potential to avoid hospital admission, highlights a critical need for further research. This research should focus on improved ADR/ADE risk prediction methods, as well as on co-morbidity and medication optimization for older adults. It is also important to remain cognisant of the diagnostic challenge posed by differentiating these events from geriatric syndromes in this patient cohort.

Declaration of Conflicts of Interest

None declared.

Declaration of Sources of Funding

This work is independent research funded by the Health Research Board SDAP-2021-020. The first author is an MD candidate whose fees are funded through a Clinical Lecturer post in RCSI University of Medicine and Health Sciences, Ireland.