Retrospective analysis of adverse drug reactions leading to short-term emergency hospital readmission

of all readmissions within 30 days. The most frequent ADRs were gastrointestinal disorders (26%), infections and infestations (19%), and nervous system disorders (10%). The most frequent drug classes leading to ADRs were antineoplastic/immunomod-ulating (35%) and antithrombotic agents (25%). Only 8 (3%) of the 270 cases were reported to the Swiss ADR reporting system. CONCLUSION: ADR-related readmissions constituted a considerable part of short-term emergency readmissions. Despite being a relevant cause for rehospitalisation, only a minority of the ADRs were reported to the regulatory authorities. Strategies to prevent ADR-related readmissions and to improve reporting rates are needed.


Introduction
Adverse drug reactions (ADRs) are unintended noxious responses to medicinal products and can present a major burden on health care [1,2].Approximately 3-5% of hospital admissions are estimated to be related to ADRs [2][3][4], with even higher rates in geriatric populations [5].Patients hospitalised owing to an ADR have a significantly prolonged length of hospital stay and an almost 2-fold increased risk of death compared with other hospitalised patients [6].Therefore, efforts to decrease ADRs are essential to reduce patient harm and healthcare costs.
Hospital readmissions are increasingly used as a measure of healthcare quality [7].According to a recent systematic review including 19 studies, the median prevalence rate of drug-related hospital readmissions was 21%, with an estimated preventability of 69% [7].Hospital readmissions shortly after hospital discharge represent a subgroup of great interest in terms of preventive measures and quality improvement.Although short-term hospital readmissions can be associated with non-drug related causes such as premature discharge due to pressure on beds, poor community support services and medical complications [8], in a previous study from the United States nearly one-fourth of the cases with hospital readmission within 30 days had a contributing ADR [9].In a German study, ADRs led to hospitalisation in 6.2% of first admissions and in 4.2% of readmissions [10].In some cases, a combination of the above-mentioned reasons may lead to a short-term readmission; for example, an ADR caused by a new drug therapy started during hospitalisation might not be detected in time owing to premature discharge in an effort to reduce costs, with short-term readmission as a possible consequence [10].Importantly, approximately half of the ADRs leading to hospital admission have been found to be preventable [8,10], which highlights the importance of ADR monitoring in clinical practice to optimise patient care and public health.
Spontaneous ADR reports transmitted from health professionals to drug regulatory authorities play an important role in providing postmarketing pharmacovigilance data.In Switzerland, ADR reports are processed by regional pharmacovigilance centres and Swissmedic's national pharmacovigilance centre, which collaborates with the international centre for drug safety run by the World Health Organization (WHO) [11].In accordance with the new Law on Therapeutic Products [12], all serious adverse reactions must be reported.ADRs are considered serious if they result in death, are life-threatening, lead to or prolong hospitalisation, involve a persistent disability or incapacity, or are otherwise to be considered medically significant (e.g., when a timely medical intervention was needed to prevent one of the above-mentioned outcomes).Spontaneous reports can contribute to drug safety by generating signals of possible ADRs that can then be followed more closely.
Investigation of ADR-related readmissions can contribute to the identification of vulnerable groups and high-risk drugs and to public health by offering guidance regarding preventive measures.Currently, insufficient data are available regarding the frequency and characteristics of ADRrelated emergency readmissions in Switzerland.The main aim of this retrospective study was to characterise ADRs leading to short-term emergency readmissions in a large Swiss University Hospital.Further, we aimed to assess the reporting frequency of such ADRs to the Swiss national pharmacovigilance centre.

Materials and methods
This retrospective study included all ADR-related readmissions presenting to the emergency department of the University Hospital Bern within 30 days after hospital discharge between 1 January and 31 December 2012.The emergency department of the University Hospital Bern is both a primary care facility (walk-in patients) and tertiary referral centre for hospitals in the greater Bern area (patients ≥16 years of age), with about 48,000 emergency admissions a year (2018).The division of Clinical Pharmacology and Toxicology of the hospital also hosts the local regional pharmacovigilance centre, which receives and processes ADR reports and forwards them to the national pharmacovigilance centre (Swissmedic).The study was reviewed by the local ethics committee (Cantonal Ethics Committee Bern).
Cases were identified by reviewing the electronic records of all patients discharged from the University Hospital Bern within the 12-month period with emergency readmission within 30 calendar days after hospital discharge.The follow-up period of 30 days has been commonly used in previous studies investigating drug-related hospital readmissions [7], and hospital readmission within 30 days of discharge has also been described as a standard measurement of hospitalisation quality [13].Case inclusion required a known ADR (listed in the official Swiss [14] or US drug information [15]) and, in line with the definition of ADRs [1], a temporal relationship between the ADR and drug intake.Cases were included if the reason for the readmission was an ADR (causality could be possible, probable or certain).The assessment was based on the reason of admission as stated in the emergency department report and information on patient history (medication history).In some, but not all, cases the drug cause was mentioned in the admission diagnosis section.Cases with intentional overdosing, evident lack of compliance, insufficient documentation, decreasing symptoms despite continuation of the suspected drug(s), or readmitted for non ADR-related signs and symptoms (e.g., in the context of the patient's primary disease, cases of violent assaults) were excluded.Patient records were reviewed independently by two experienced medical professionals and unclear cases (divergent opinions of the two primary assessors) were additionally reviewed together with a senior physician with experience in this field.Identified cases were searched in the Swiss ADR reporting system to assess the reporting rate.A flowchart of the procedures is shown in figure 1.
An ADR was defined as "an appreciably harmful or unpleasant reaction, resulting from an intervention related to the use of a medicinal product, which predicts hazard from future administration and warrants prevention or specific treatment, or alteration of the dosage regimen, or withdrawal of the product" [1].Drugs were classified using the WHO classification system based on the Anatomical Therapeutic Chemical (ATC) code, a unique code assigned to a medicine according to the organ or system it acts on and how it works [16].For the evaluation of drug interactions, the drug interaction screening programme Pharmavista was used [17].For the description of ADRs, the WHO Adverse Reaction Terminology (WHO-ART) Lowest Level Terms (LLTs) were used to provide maximum specificity [18].The causality assessment was based on the Swiss ADR reporting system criteria (table 1) [19], which are based on the WHO Uppsala Monitoring Centre (UMC) causality assessment system [20].
For the investigation of differences between the ADR-related readmissions (study population) and non ADR-related emergency readmissions during the study period, for which data were collected on age, sex, days between first hospitalisation and readmission, and duration of hospitalisation after readmission, comparisons were tested using the chi-square test for categorical variables, the t-test for normally distributed continuous variables, and the Mann-Whitney test for nonparametric variables.Values of p <0.05 were considered statistically significant.Statistical analyses were conducted using SPSS statistical software (IBM SPSS Statistics 25.0).

Results
During the study period there were 4792 readmissions (14% of all admissions) within 30 days after discharge and 1294 (27% of all readmissions) of these were emergency readmissions.We identified 270 cases of ADR-related emergency readmissions, corresponding to 21% of emergency readmissions and 6% of all readmissions within 30 days.Nine hundred and sixty-one cases were not ADRrelated and in 63 cases an adequate evaluation was not pos-sible because of insufficient or missing documentation (fig.2).
Among the 270 cases of ADR-related readmissions, 78% were readmitted from home and 22% from a medical institution (hospital or rehabilitation facility).Most patients were elderly (59% ≥65 years old) and were male (63%).The median number of drugs on readmission was 8 (range 0-22; causative drugs already discontinued before presen-tation at the emergency department in 2 cases and no information on the number of drugs available in 32 cases).The median number of (active) main diagnoses was 6 (range 1-18).In 125 cases (46%), the ADR was associated with a drug that was newly started or changed during the index hospitalisation, in 136 cases (50%) the associated drug was either unchanged or started after the index hospitalisation, and in 9 cases (3%) an evaluation was not possible because of insufficient information.ADR-related cases were  In 231 cases (86%), the ADR concerned a reaction to the drug itself, in four cases (2%) an ADR was caused by a drug-drug interaction (DDI), and in 35 cases (13%) both.The four cases in which a DDI led to an ADR included hyponatraemia under the combination of oxcarbazepine, hydrochlorothiazide and citalopram, hyperammonaemic encephalopathy under the combination of valproic acid and topiramate, drug-induced delirium under ritonavir and midazolam, and a fatal case of acute renal failure under perindopril, indapamide, torasemide and lercanidipine.Further details about cases with fatal outcome can be found in table 4.
The further 31 cases with ADRs caused not only by the drugs themselves but also by a DDI included cases with increased risk of bleeding (drugs involved: acetylsalicylic acid, clopidogrel, nadroparin, enoxaparin, heparin, phenprocoumon, acenocoumarol, escitalopram, ibuprofen, ciprofloxacin, dexamethasone), one case of increased toxicity of fluoropyrimidines (drugs involved: fluorouracil, calcium folinate), one case of increased risk of extended respiratory depression and sedation (drugs involved: morphine, flunitrazepam), one case of a skin reaction in a patient treated with lamotrigine and valproic acid, and one case with two interactions, blood pressure decrease (candesartan, hydrochlorothiazide) and hyponatraemia and ventricular arrhythmias (hydrochlorothiazide, trimipramine).
In accordance with the new Law on Therapeutic Products [12], all ADRs of the study were classified as "serious" as they led to (re-)hospitalisation; in 228 of the cases (84%), this was the only criterion for "seriousness", 34 cases (13%) required admission to the intensive care unit and were thus considered to be life-threatening (table S3), and 8 cases (3%) were fatal (table 4).Despite fulfilled criteria for seriousness in all of the included cases, only 8 (3%) of the 270 cases and none of the fatal cases were reported to the Swiss ADR reporting system (table 5).

Discussion
Our data show that ADR-related readmissions constitute a considerable part of short-term emergency readmissions.The most frequent ADRs associated with emergency readmissions within 30 days after hospital discharge were gastrointestinal disorders (approximately one-fourth of the cases, including cases of gastrointestinal bleeding), as well as infections and infestation, (approximately one-fifth of the cases).In line with this, the most frequent drug classes involved were antineoplastic/immunomodulating and antithrombotic agents, and most (five out of eight) fatal cases were bleeding related.Despite fulfilling the criteria for seriousness, only a minority of the ADRs leading to emergency readmissions was reported to the regulatory authorities.In our study, ADR-related emergency readmissions corresponded to 21% of emergency readmissions and 6% of all readmissions within 30 days after discharge.According to a recent systematic review [7], rates of drug-related readmissions in previous studies were 3-64% (median 21%).
The follow-up time between first admission and readmission in these studies varied from 28 days to more than 4 years, but readmission within 30 days was the most commonly used measure [7].Besides data on the rates and causes of readmissions, other aspects such as the patients' emotional costs, loss of quality of life and economic burden should also be considered for the estimation of the global clinical and economic consequences related to hospital readmissions.Although these were not assessed in the current study, previous studies from the United States report approximately 20% rate of rehospitalisations of Medicare patients within 30 days after discharge with an estimated annual cost of unplanned rehospitalisations of US$17 billion [21].
In a previous study investigating ADR-related emergency department visits leading to hospitalisation among adults ≥65 years of age [22], warfarin / oral antiplatelet agents and insulins / oral hypoglycaemic agents were the implicated drugs / drug classes in two-third of the cases, whereas high-risk drugs, as defined by the Healthcare Effectiveness Data and Information Set (HEDIS) measure for "Use of high-risk medications in the elderly" [23], were involved in only a minority of the cases.In another study from the same group [24], investigating emergency department visits for ADRs involving medications identified as potentially inappropriate based on the Beers criteria (a consensus-based and repeatedly updated list of medications considered potentially inappropriate for use in patients ≥65 years of age, mostly owing to a high risk for adverse events [25]), three drugs (warfarin, insulin, digoxin) were implicated in one-third of the cases, whereas Beers criteria medications caused lower numbers of emergency department visits.Similar findings have been reported in studies in geriatric patients (≥80 years of age) with ADR-related readmission within 30 days, in which anticoagulants / antiplatelet agents and bleeding were the most common drug classes and adverse event [9]; prescription of nervous system drugs (third most frequent drug class in our study) was identified as a risk factor for ADR-related readmissions within 30 days in a previous study with elderly patients [13].Although our study also included younger patients, antithrombotic agents and bleeding complications (e.g., gastrointestinal, epistaxis, haematuria, haematoma) were among the most commonly reported drugs and disorders, and bleeding was the underlying ADR in five of the eight fatal cases.These findings have important clinical implications, since such reactions (also known as Type A or pharmacological ADRs) are largely dose dependent with known mechanisms and therefore preventable [26].This is different from idiosyncratic reactions (also known as Type B or hypersensitivity ADRs), such as many cases of druginduced liver injury or allergic skin reactions which are less influenced by dosage and often are immunologically mediated [26].Therefore, future strategies to prevent ADRs and ADR-related readmissions should focus not only on available lists of potentially inappropriate medications for specific age groups as listed in the Beers criteria [25] or the German PRISCUS list [27], but also on other considerations such as comorbidities and DDIs that can lead to Type A reactions, and adequate follow-up for a timely check for preventable ADRs, especially in patients treated with anticoagulants and/or antiplatelet agents.Preventive measures to decrease Type A ADRs could include automated red flags and DDI alerts in electronic medical records based on patient profiles and laboratory values to provide reminders for, e.g., a dose reduction based on the renal function or a pharmacodynamic interaction with increased risk of bleeding in the case of a combination of, for example, an anticoagulant agent and a nonsteroidal anti-inflammatory drug (NSAID).Furthermore, pharmacist-or clinical pharmacologist-led medication reconciliation interventions could further contribute to the reduction of medication discrepancies and ADRs [28,29].In the case of Type B reactions, some could be prevented with validated pharmacogenetic testing (e.g., human leucocyte antigen (HLA)-B*5701 and associated increased risk for hypersensitivity reactions to abacavir [30]).
Most of the patients in our study belonged to the elderly group and ADR-related cases were significantly older than non-ADR-related cases, most probably as a result of factors such as polypharmacy, impaired renal function or other comorbidities, which are common among older patients.The complexity of medication regimens, which is calculated on the basis of number of prescribed drugs, dosage form and frequency, and additional instructions has also been shown to be predictive for unplanned hospital readmissions within 30 days in previous studies [31].Not all of those factors were investigated in our study (and a high medication regimen complexity score might also lead to readmission due to lower adherence and not due to an ADR) [31].However, a median of 8 drugs, extending up to 22 drugs on readmission, can be taken as an indicator of a rather high complexity of medication regimens in the ADR-related cases in our study.The large majority of the cases were readmitted from home, whereas only one-fifth of the patients were readmitted from another medical institution.Returning home after discharge was identified as a risk factor for emergency readmission within 30 days also in another matched case-control study with elderly patients, after adjustment for sex and age [13].These findings also highlight the importance of regular follow-up as a strategy to prevent ADR-related readmissions, since it can be assumed that patients in medical institutions receive more regular and thorough medical supervision than patients at home.In a previous prospective cohort study investigating preventability of ADRs among outpatients [32], 63% of the ameliorable events were attributed to the physician's failure to respond to drug-related symptoms and 37% to the patient's failure to report the symptoms to the physician.Regular follow-up with enough time available to check the patient's medication list regarding indication, correct dosage and DDIs, and also to ask the patient about any potential drug-related symptoms can thus contribute to the prevention of ADRs.In contrast to previous studies [6,10], we did not find a significant difference in the duration of hospitalisation of the ADR-related and the non-ADR-related cases, which might be in part attributable to different hospital discharge policies among countries.In our study, there was no significant difference in the days between first hospitalisation for ADR-vs non-ADR-related readmission; in a previous study no signif-

Original article
icant difference regarding the delay between hospitalisations was found in patients with only one ADR-related hospitalisation and patients hospitalised twice or more owing to ADRs [33].
Only a minority (8 of the 270 cases) of the ADRs in our study were reported to the regulatory authorities, although all cases identified led to (re-) hospitalisation and thus formally fulfilled the criteria for seriousness.This highlights one of the major limitations of pharmacovigilance data, which are plagued by high underreporting rates [34].Reasons for underreporting include lack of time or unawareness of reporting requirements; for example, the medical personnel may not be aware that expected or only suspected ADRs also could or must (in the case of serious ADRs) be reported [12,34].Based on the reported cases in our study, it seems that the decision to report a case may have been based more on the clinical presentation rather than the formal criteria for seriousness of the adverse reaction.Another possible factor favouring reporting could be knowledge of the possible adverse reactions of the drugs, since many of the reported cases referred to well-described ADRs of the specific agents (e.g., agranulocytosis associated with metamizole, liver injury associated with paracetamol, exanthemas associated with antibiotics, heparin-induced thrombocytopenia).However, none of the fatal cases had been reported, which might be due to lack of time (especially in an emergency setting), unawareness regarding reporting requirements [34] or fear of possible legal consequences.Although currently not the case in Switzerland, policies such as the Hospital Readmission Reduction Program [35] have been introduced in the United States to reduce readmissions, by, for example, imposing payment penalties on hospitals with excessive readmissions for specific diagnoses, and similar developments are seen in some European countries [7,36].Since spontaneous reports are a useful drug safety evaluation tool and can generate signals, which can then be followed more closely, it is important to raise awareness regarding the importance of pharmacovigilance among medical personnel and also to clarify relevant aspects such as the anonymity of the reports and that proof of causality is not required.In addition, organisation of ADR monitoring systems by clinical pharmacologists and/or pharmacists within hospitals could also significantly contribute to the timely recognition and reporting of ADRs to the regulatory authorities.
Limitations of our study include the retrospective design, with some missing information in some cases, and data from only one emergency department, which may not be representative for the whole country or other health systems.Furthermore, most cases were assessed as "possible" and only few cases as "probable/likely" or "certain" based on the formal causality criteria, and our data represent prescription patterns that reflect clinical practice during the observation period of the study, and new drug categories (e.g., direct oral anticoagulants) have been introduced into the Swiss market since then.It is also possible that newly introduced ADR screenings by clinical pharmacologists and pharmacists on hospital wards have contributed to increased ADR reporting in the recent years.We investigated emergency department readmissions, and thus the total number of ADR-related readmissions is most likely higher, since cases admitted directly to a hospital ward (>70% of the total readmissions) were not included in the analysis.
The strengths of the study include the sensitive search, the individual review of the cases and the investigation of the reporting frequency to drug regulatory authorities.To our knowledge, this is the first study to investigate the frequency and characteristics of ADR-related emergency readmissions in a large Swiss University Hospital.It could thus contribute to public health by offering guidance regarding ADR preventive measures and also raise awareness regarding the importance of ADR reporting as a drug safety tool.
In conclusion, ADR-related readmissions constituted a considerable part of short-term emergency readmissions, with potentially preventable ADRs (e.g., bleeding, which might have been prevented by more regular measurement of the international normalised ratio (INR), prescription of a proton pump inhibitor together with NSAID to prevent gastroduodenal toxicity, dental hygiene and regular dental visits in case of bisphosphonates) involved in many of the cases.Despite being a relevant cause for rehospitalisation, only a minority of the ARDs were reported to the regulatory authorities.Strategies to prevent ADR-related readmissions and to improve reporting rates are needed.

Figure 1 :
Figure 1: Flowchart of the procedures used for the identification of adverse drug reaction (ADR)-related emergency readmissions.(* = Listed in the official Swiss or US drug information.)

Figure 2 :Table 2 :
Figure 2: Number of cases of non-emergency readmissions and adverse drug reaction (ADR)-and non-ADR-related emergency readmissions.

Table 4 :
Adverse drug reaction (ADR)-related fatal cases (death in possible relation to ADR and not cases of patients who died during hospitalisation for other reasons; n = 8).

Table 5 :
Cases reported to the Swiss national pharmacovigilance centre (n = 8).
ADR: adverse drug reaction * No ADR after normalisation of tacrolimus concentration Original article Swiss Med Wkly.2021;151:w20400 Swiss Medical Weekly • PDF of the online version • www.smw.chPublished under the copyright license "Attribution -Non-Commercial -No Derivatives 4.0".No commercial reuse without permission -https://smw.ch/permissions

Table S1 :
Adverse reactions and drugs involved by Anatomical Therapeutic Chemical (ATC) code (>1 drug involved in some cases, n = 477 drugs).