Development of a standardized chart review method to identify drug‐related hospital admissions in older people

Abstract

Aims

We aimed to develop a standardized chart review method to identify drug‐related hospital admissions (DRA) in older people caused by non‐preventable adverse drug reactions and preventable medication errors including overuse, underuse and misuse of medications: the DRA adjudication guide.

Methods

The DRA adjudication guide was developed based on design and test iterations with international and multidisciplinary input in four subsequent steps: literature review; evaluation of content validity using a Delphi consensus technique; a pilot test; and a reliability study.

Results

The DRA adjudication guide provides definitions, examples and step‐by‐step instructions to measure DRA. A three‐step standardized chart review method was elaborated including: (i) data abstraction; (ii) explicit screening with a newly developed trigger tool for DRA in older people; and (iii) consensus adjudication for causality by a pharmacist and a physician using the World Health Organization‐Uppsala Monitoring Centre and Hallas criteria. A 15‐member international Delphi panel reached consensus agreement on 26 triggers for DRA in older people. The DRA adjudication guide showed good feasibility of use and achieved moderate inter‐rater reliability for the evaluation of 16 cases by four European adjudication pairs (71% agreement, κ = 0.41). Disagreements arose mainly for cases with potential underuse.

Conclusions

The DRA adjudication guide is the first standardized chart review method to identify DRA in older persons. Content validity, feasibility of use and inter‐rater reliability were found to be satisfactory. The method can be used as an outcome measure for interventions targeted at improving quality and safety of medication use in older people.

What is Already Known about this Subject

• Drug‐related hospital admissions (DRA) represent a growing patient safety threat in older people.

• Identifying DRA in older people is complex and there is lack of a standardized approach to identify DRA.

What this Study Adds

• We developed a standardized chart review method to measure DRA in older persons.

• Content validity, feasibility of use and inter‐rater reliability were found to be satisfactory.

• The method can be used as an outcome measure for interventions targeted at improving quality and safety of medication use in older people.

Introduction

Adverse drug events (ADEs) are a leading cause of iatrogenic harm globally 1, 2. A significant proportion of ADEs results in hospitalization and these so‐called drug‐related hospital admissions (DRA) have serious clinical and economic consequences 3, 4, 5, 6. DRA can result from non‐preventable adverse drug reactions (ADR) or from preventable medication errors.

Older adults have almost a seven‐fold increased risk of experiencing a DRA compared to younger persons due to several risk factors such as multi‐morbidity and polypharmacy 7. Around 70% of DRA in older people are caused by potentially preventable ADEs mainly resulting from poor medication adherence and inappropriate prescribing 8, 9, 10, 11, 12, 13. The latter includes the prescription or use of more drugs than are clinically needed (overuse), the incorrect prescription or use of drugs that are needed (misuse) and the failure to prescribe or use drugs that are needed (underuse) 14. Identifying DRA in older people is challenging because ADEs often present as common geriatric problems such as falls, confusion or renal impairment, which might be due to the ageing process, underlying diseases or medications 13, 15.

No standardized and validated method to identify DRA in older people exists in the literature. Yet measuring DRA is potentially an important issue in the light of the World Health Organization's (WHO's) Global Patient Safety challenge on medication‐related harm 2. Studies have reported DRA prevalence rates ranging from 6% to 50% of all admissions in older adults 16, 17, 18, 19, 20. The wide variance in prevalence rates is associated with the considerable heterogeneity in definitions and methods used to identify DRA, the study population and the setting 20, 21. DRA identification often relies on a highly subjective and variable process and few attempts have been made to measure DRA resulting from underuse of medications 12, 19, 22, 23.

We aimed to develop a standardized chart review method to identify DRAs resulting from ADR, overuse, misuse and underuse of medications, specific to older people: the DRA adjudication guide. In this paper we present the developmental pathway of the DRA adjudication guide and the evaluation of its content validity, feasibility of use and reliability, which are defined as desirable attributes of a quality measure by the Agency for Healthcare Research and Quality 24.

The DRA adjudication guide will be used in four European centres to measure the primary outcome DRA in the OPERAM trial (http://operam-2020.eu) that will assess the impact of a pharmacotherapy optimization intervention in 2000 multi‐morbid older people.

Methods

Design

The DRA adjudication guide was developed in four steps: (i) the first draft of the guide was developed based on literature review; (ii) this version was subsequently refined based on evaluation of content validity by an expert panel; (iii) user feedback in a pilot test; and (iv) a reliability study (Figure 1).

Figure 1.

DRA adjudication guide development process. ADE, adverse drug event; DRA, drug‐related hospital admission

Literature review

Two literature searches were performed in PubMed by the first author for articles published between 1 January 1990 and 1 August 2015. Screening of titles and abstracts and data extraction was performed by the first author.

A first exploratory search aimed to review existing structured ADE or DRA identification approaches to inform the development of the overall DRA identification strategy. The search included the following medical subject headings (MeSH): ‘Patient admission’, ‘Drug‐related side effects and adverse reactions’, ‘Quality assurance, Health Care’, ‘Patient outcome assessment’. Studies published in English, French or Dutch that focused on defining, identifying, and/or characterizing ADE or DRA in the adult in‐hospital setting were included.

A second literature search aimed to review common causes for DRA in older people to inform the development of a trigger tool for DRA in older people for inclusion in the DRA adjudication guide. To improve efficiency and to standardize identification of ADEs, trigger‐based chart review has been advocated as the premier ADE identification approach 25, 26, 27. Triggers are defined as occurrences, prompts or flags found upon chart review that trigger further investigation to determine the presence or absence of an adverse event 28. Trigger tools have been designed for a variety of clinical settings but, to our knowledge, no trigger tool for identifying DRA in older people exists. To compile a preliminary trigger tool, the second literature search aimed to identify common causes for DRA in older people and to review previously developed adverse event triggers tools designed for other settings. PubMed was searched using the following search terms and/or combinations: ‘Aged’[MeSH], ‘Drug‐Related Side Effects and Adverse Reactions’[MeSH], ‘Hospitalization’[MeSH], ‘Trigger’[All fields], ‘Adverse drug events trigger tool’[All fields], ‘Pharmaceutical preparations’[MeSH], ‘Underuse’[All fields], ‘Prescribing omission’[All fields]. Studies on hospitalizations in people aged ≥65 years resulting from preventable ADEs and non‐preventable ADRs were included. Studies on the development or evaluation of adverse event trigger tools designed for other settings were also included. Studies on DRA in patients younger than 65 years were excluded. Trigger tool studies focusing on specific patient groups such as surgical patients were also excluded.

A data extraction form was developed to document study characteristics including study aims, population, design, setting, methods used to detect ADE or DRA, causality algorithms used, professionals involved in ADE or DRA assessment, most frequent causes of DRA, most frequent medications involved or omitted in DRA, triggers and their positive predictive value.

Evaluation of content validity

Content validity refers to the relationship between an instrument's content and the construct it is intended to measure 29. In the absence of a gold standard to measure DRA, content validity of the DRA adjudication guide was assessed by an expert panel.

First, the overall DRA identification method suggested by the guide was agreed on a consensus basis through face‐to‐face discussions by three physicians (B.B., J.B.B., J.D.) and two clinical pharmacists (A.S., O.D.) with expertise in geriatric pharmacotherapy and medication safety.

Secondly, a two‐round online modified Delphi survey using LimeSurvey software was conducted to validate the triggers derived from the literature review. The Delphi method is a consensus technique that is widely used for questions addressing medication safety in older adults 30. A modified online two‐round Delphi survey was selected in this study as a way to combine scientific rigor and pragmatism to obtain consensus from a geographically diverse expert panel. Experts were selected based on their recognized academic or clinical expertise on the subject of drug‐related morbidity in older patients or were personal contacts. Of the 29 experts invited, respectively 15 and 14 experts from eight different countries took part in the first and second Delphi round (Table 1).

Table 1.

Characteristics of Delphi panellists

	Experts invited n (%)	Participation Round 1 n (%)	Participation Round 2 n (%)
Total	29 (100)	15 (52)	14 (48)
Profession, area of expertise
Physician, geriatric medicine	10 (34)	6 (40)	6 (43)
Physician, internal medicine	8 (28)	2 (13)	2 (14)
Physician, primary care	1 (3)	–	–
Pharmacist, geriatric medicine	5 (17)	4 (27)	3 (21)
Pharmacist, medication safety	5 (17)	3 (20)	3 (21)
Country
Belgium	5 (17)	5 (33)	4 (29)
Canada	1 (3)	1 (7)	1 (7)
Italy	1 (3)	–	–
Ireland	2 (7)	1 (7)	1 (7)
France	2 (7)	1 (7)	1 (7)
Switzerland	4 (14)	2 (13)	2 (14)
The Netherlands	6 (21)	3 (20)	3 (21)
UK	2 (7)	1 (7)	1 (7)
USA	6 (21)	1 (7)	1 (7)
Sex
Female	15 (52)	9 (60)	8 (57)
Male	14 (48)	6 (40)	6 (43)

The Delphi panel was asked to assess the content validity of the preliminary trigger tool, to develop consensus on the most relevant triggers and to identify additional triggers. Furthermore, the panel was asked to assess two screening questions for non‐triggered, spontaneously detected events. In the first Delphi round participants were asked to rate for each of the 29 triggers derived from the literature and for the two screening questions the relevance to screen for a DRA in older people on a five‐point Likert scale (ranging from absolutely irrelevant to absolutely relevant; relevance was defined as the degree to which the item comprehensively includes the full scope of the outcome it intends to measure). A free‐text field was provided for each item, allowing comments to improve the trigger design or to suggest new triggers.

For each item, consensus measurement was based on the median Likert response and the interquartile range. The following cut‐off values of consensus were defined before data analysis: consensus that a trigger should be retained if the median score on the five‐point Likert scale was ≥4 and the 25th percentile ≥4 (i.e. ≥75% of the experts considered the trigger as relevant or absolutely relevant); consensus that a trigger should be excluded if the median score was <3 and the 75th percentile <3 (i.e. ≥75% of the experts considered the trigger as irrelevant or absolutely irrelevant); no consensus for triggers that failed to meet either of the latter cut‐off values.

Triggers that were accepted or rejected unanimously after the first round were not presented in the second round. In the second Delphi round, participants were asked to rate the triggers for which revisions were suggested in the first round. Furthermore, participants were asked to re‐evaluate the equivocal triggers on the five‐point Likert scale, taking into account the groups' responses. Participants were provided with a reminder of their own responses from round 1, the median group rating and interquartile range and a summary of the comments made by participants. Equivocal triggers that were rated equivocal again, were not included in the final trigger tool (Supporting Information S1).

Pilot test

A pilot test was performed aimed at ensuring that the newly developed DRA adjudication guide was a workable instrument and to identify points for improvement. For this purpose, the DRA adjudication guide was piloted independently by a geriatrician and a pharmacist from one centre (J.B.B., S.T.). For the pilot test, 15 cases from a medical record database of frail older patients admitted to a teaching hospital were randomly selected by using a random number generator. The reviewers' suggestions for improvement were discussed within the OPERAM research team and modifications were subsequently implemented in the DRA adjudication guide.

Reliability study

A reliability study was conducted to assess whether the DRA adjudication guide yields reproducible results when applied by different raters. Raters were OPERAM research team members with clinical and/or research experience in geriatric medicine. Pairs of raters in three centres (Brussels, Cork and Utrecht) consisted of a pharmacist and physician (S.M., F.V., I.W., A.V., S.C., D.O.M.) whereas in one centre (Bern) the pair was composed of physicians only (C.F., C.S.). The raters had no prior experience in using the DRA adjudication guide and were provided with a video training tutorial (https://www.youtube.com/watch?v=fadmO-WcCHM).

For the purpose of the reliability study, each centre provided four cases of multi‐morbid older patients including the discharge and/or admission letter, laboratory values and medication lists. Translation of the cases was performed by OPERAM research team members from their mother tongue (Dutch, French, Swiss‐German) to English. No formal back‐translation process was undertaken.

Raters were asked to first assess the cases individually and subsequently to come to a consensus result on the case within the pair. The time needed to adjudicate a case was recorded. A dichotomous outcome variable (DRA identified yes/no) was defined and inter‐rater reliability (IRR) was determined by calculating percentage agreement and agreement corrected for chance between pairs of raters from four European centres (Fleiss' κ) as well as within each pair (Cohen's κ) for the dichotomous outcome variable. Kappa values were interpreted as slight agreement if <0.20, fair agreement if 0.21–0.40, moderate agreement if 0.41–0.60, substantial agreement if 0.61–0.8 and almost perfect agreement if 0.81–1.00 31. Next, adjudication results and discrepancies were shared among all raters, who were asked for feedback. The primary goal was to determine whether discrepancies were due to difficulties in using the adjudication method, missed information or case interpretation.

Ethics approval

The ethics committee from the Cliniques universitaires Saint‐Luc (Brussels, Belgium) provided approval for anonymous use of the medical record database (reference number B40201111806).

Results

Literature review and development of the DRA adjudication guide

Development of the overall DRA identification strategy

Twenty‐five studies on ADE or DRA identification were reviewed 3, 7, 12, 26, 27, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51. Chart review by two or more reviewers has been considered as a gold standard in many patient safety studies because of its high ADE yield and high specificity 32. To evaluate the relationship between drug treatment and the occurrence of an adverse event, several causality assessment methods have been developed. No causality assessment method is universally accepted but expert judgement is the most widely used 47. Chart review is however often conducted in an implicit and unstructured way, resulting in low IRR 32. Our method selected to adjudicate DRA therefore involved a structured chart review with the aid of a trigger tool to improve efficiency and standardization in ADE detection 25. Previous research has demonstrated that by restricting ADE detection to trigger tools only, whole classes of ADE can be missed 32, 52, 53. Therefore, two screening questions for non‐triggered, spontaneously detected events were also compiled.

A three‐step approach for DRA identification based on chart review was elaborated (Figure 2). The three steps are: (i) abstraction of a standardized list of data from the medical record into an electronic case report form, the main source documents including the admission and discharge letter, laboratory values and medication lists; (ii) explicit screening for ADE(s) that are potential DRA with the DRA trigger tool and screening questions for non‐triggered events; and (iii) adjudication: consensus judgement in terms of ADE causality and ADE contribution to hospital admission with the WHO–Uppsala Monitoring Centre (UMC) and Hallas criteria respectively 36, 54. Steps 2 and 3 are performed by an adjudication pair composed of a pharmacist and a physician given their complementary knowledge and experience 55, 56. Definitions, step‐by‐step instructions for use and examples are contained in the DRA adjudication guide (Supporting Information S2).

Figure 2.

Three‐step approach for identifying DRA. ADE, adverse drug event; DRA, drug‐related hospital admission

Development of the trigger tool

Twenty‐three studies on common causes of DRA in older people 3, 7, 8, 9, 10, 12, 16, 23, 38, 51, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69 and 12 trigger tools studies were reviewed 30, 52, 53, 70, 71, 72, 73, 74, 75, 76, 77, 78. Based on the information from the literature and their own clinical expertise, the research team compiled a preliminary list of 29 triggers and two screening questions for non‐triggered events related to ADR, overuse, underuse or misuse of medications. Key considerations for selecting the triggers were the reported positive predictive value of the triggers, severity (i.e. the trigger should be severe enough to result in hospital admission) and ease of detection. The triggers were divided into three categories: diagnoses, abnormal laboratory values and other triggers (e.g. antidote use). Each trigger was elaborated with potential causative drugs or potential causes for drug underuse based on the STOPP/START criteria version 2 and by consulting pharmacology and pharmacotherapy references 79. Consequently, each trigger consists of a diagnosis or abnormal laboratory value and a corresponding list of potential causative drugs or causes for drug underuse allowing explicit chart screening for DRA.

Evaluation of content validity

None of the 29 triggers or screening questions were removed at the end of the first round by the 15‐member Delphi panel. Twenty‐five triggers and two screening questions for non‐triggered events were rated relevant or absolutely relevant to screen for DRA in older people. Of the items on which the group agreed, 10 triggers and two screening questions were adopted without alteration in the final tool, whereas 15 triggers were revised according to the participants' suggestions. Revisions included changing cut‐off thresholds of laboratory values, adding or removing medications associated with a trigger or adding more detail to the triggers. Four triggers (theophylline level > 20 μg ml^–1, rash, Clostridium difficile toxin positive stool, neutrophils <1400 μl^–1) were rated equivocal.

After the second round, all 15 triggers with revisions were rated relevant or absolutely relevant. Three out of four equivocal triggers from the first round were rated equivocal again and these were removed from the trigger tool. The trigger neutrophils <1400 μl ^–1 was now rated relevant and was included in the final trigger tool (Supporting Information S1). Following last refinements, the final 26‐item trigger tool was created (Table 2).

Table 2.

Trigger tool for drug‐related hospital admissions in older persons

Pilot test

The two reviewers involved in the pilot considered the trigger tool as a workable instrument for screening for DRA. The same sets of triggers were identified by the two reviewers; however, adjudication of DRA was the part where most discrepancies arose. Based on feedback from the reviewers, the following modifications were made after the pilot:

• The Naranjo algorithm and Therapeutic Failure Questionnaire 63, 80, which were proposed as causality algorithms in the DRA adjudication guide v.1, were replaced by the WHO‐UMC causality criteria because they reflect clinical practice better. The WHO‐UMC criteria were adapted to allow causality assessment due to medication underuse in line with Klopotowska et al. 32

• Discharge medications were added to the list of data to abstract to aid in the detection of potential underuse.

• The DRA identification strategy and instructions for use were adapted to the process that both reviewers considered as most practical.

Reliability study

Table 3 provides the level of agreement on the presence of a DRA between all centres and within each pair per centre for 16 cases. The DRA adjudication guide achieved a moderate IRR score between adjudication pairs from four European centres (71% agreement, Fleiss' κ = 0.41). Agreement within each pair varied from fair to almost perfect agreement (69%–94% agreement, Cohens' κ = 0.33–0.86). The mean time needed to assess a case individually was 23 ± 6 min and the mean time needed for consensus discussion was 13 ± 5 min.

Table 3.

Inter‐rater reliability for drug‐related hospital admission presence between four adjudication pairs and per centre for the evaluation of 16 cases

Raters	% agreement	Kappaa
Four adjudication pairs	71%	0.41
Centre 1 (two physicians)	94%	0.86
Centre 2 (physician + pharmacist)	75%	0.42
Centre 3 (physician + pharmacist)	69%	0.33
Centre 4 (physician + pharmacist)	88%	0.74

^aRespectively Fleiss' and Cohen's κ were calculated to determine the level of agreement between the four adjudication pairs and within each centre

No differences in IRR for DRA identification were observed for triggered and non‐triggered cases. Detailed analysis of the adjudication results showed that in the majority of cases the same triggers and potential ADEs were identified but discrepancies arose mainly on the level of assessment of contribution to hospital admission. Discrepancies arose for eight cases with more subjective assessments including five triggered cases with potential underuse, two triggered cases with contributory reasons for admission (i.e. an ADE that is not the main reason for admission but plays a substantial role in the admission) 36 and one case with a non‐triggered DRA (Supporting Information S3).

Discussion

To our knowledge the DRA adjudication guide is the first standardized instrument to identify DRA in older persons caused by ADR, overuse, underuse and misuse of medications. The DRA adjudication guide provides definitions, examples and step‐by‐step instructions to measure DRA.

DRA identification is based on chart review with the aid of a trigger tool followed by structured consensus judgement, an approach that has been used successfully in previous ADE studies 25. The novelty of our method lies in the development of a trigger tool for DRA, specific to older people and allowing explicit DRA screening. The DRA adjudication guide calls for a rigorous evaluation of DRA including triggered and non‐triggered events as well as non‐preventable ADR and preventable medication errors, which is the desired broader focus of studying DRA 21, 32, 52, 53. Furthermore, an adjudication pair composed of a pharmacist and a physician is a recommended approach for evaluation of ADEs 55, 56.

To improve safety and quality of care, a valid and practical method to measure and understand a problem is a critical approach to any patient safety threat 1, 81, 82. It has been acknowledged that patient safety measures are often based on insufficient evidence and finding a balance between scientific soundness and feasibility is a challenge 81. We addressed these requirements by utilizing a rigorous developmental pathway based on design and test iterations combining evidence from published literature with expert opinion and user‐feedback from international and multidisciplinary sources. Content validity, feasibility of use and IRR were found to be satisfactory.

Despite the development of a standardized procedure, variability in DRA determination remains. IRR between adjudication pairs in four European centres was moderate, which is the most relevant criterion as it is the consensus judgement between the pharmacist and physician that is of importance. Achieving a good IRR score for ADE identification is a challenge inherent to retrospective chart review studies, with previous adverse event studies reporting κ scores varying from –0.077 to 0.66 19, 32, 56, 83, 84, 85. The trigger tool allowed to detect the same triggers, yet discrepancies arose mainly on the level of assessment of contribution to hospital admission. Expert judgement using causality criteria is not devoid of individual subjective judgements 47. Exploring the reasons for discrepancies highlighted the need for further training and standardization of consensus procedures for more subjective adjudications such as underuse. For example, two out of four centres in the present study considered omission of a statin in a 90‐year‐old patient admitted for myocardial infarction as a DRA, whereas there is limited evidence of benefit of statins over the age of 80–85 86.

Our reliability study is the first one evaluating DRA by international adjudication teams, yet rater pairs only came from four European countries. The IRR score can be considered as a satisfactory result taking into account the following considerations: (i) participants were at the beginning of their learning curve when IRR was evaluated; (ii) composition of adjudication teams varied with regards to profession, clinical experience and experience in ADE identification. It has been shown that IRR among different professions is lower, which explains the almost perfect agreement score in the team that was composed of only physicians 56; (iii) cases were collected in four European hospitals and quality of information in source documents such as admission and discharge letters therefore varied. Furthermore, translation of cases into English was needed and was performed by research team members and not by a translation agency, which might have resulted in differences in case quality. Moreover, interpretation of cases and source documents from another country where guidelines and practices might vary, contributes to complexity. However even if the DRA adjudication procedure is applied correctly by all raters, a certain degree of disagreement is to be expected in adjudication of complex multi‐morbidity cases.

The following recommendations to optimize IRR will be implemented in the OPERAM trial: (i) intensification of training and involvement of experienced clinicians in the adjudication teams; (ii) close monitoring of IRR at different time‐points to identify discrepancies; and (iii) prompt feedback and sharing of questions and experiences among teams 84, 87.

The adjudication guide has several limitations. Firstly, data are collected retrospectively and hence are limited to the information available in medical charts. For assessment of underuse in particular, information on patient preferences, life expectancy or adherence are often undocumented in medical charts 81. To obtain an accurate picture, prospective identification of DRA in combination with patient, caregiver and healthcare professional interviews would be desirable 33, 88, 89. Hindsight bias is another limitation of retrospective chart review; knowing the outcome and its severity may influence the adjudication of causation 90. Furthermore, the response rate of the experts invited to the Delphi survey was limited to 48%, nevertheless the Delphi panel represented various disciplines and countries. Moreover, we did not specify an age cut‐off for older people in the Delphi survey, which might have influenced the outcome. However, in the literature review on which the preliminary list of triggers was based, we only included studies of patients aged 65 years and older. We therefore believe that our trigger tool is broad enough to trigger DRA in people aged 65 years and older, which corresponds to the World Health Organization's age cut‐off to define older people. Finally, we did not compare the adjudication results from the four teams with a gold standard such as adjudication by an expert panel.

The DRA adjudication guide is time‐consuming for use in clinical practice and is designed for research purposes. The method may be used to study the incidence of DRA or drug‐related emergency department visits or as an outcome measure for interventions targeted at improving quality and safety of medication use in older people.

The performance of the trigger tool for detecting DRA has not yet been evaluated. A future study will determine the predictive validity, sensitivity and specificity of the trigger tool to detect DRA in the OPERAM dataset. An electronic trigger tool consisting of drug–disease combinations with adequate specificity could help identify patients at risk of medication‐related harm in electronic patient records 91.

Competing Interests

There are no competing interests to declare.

This work is part of the project ‘OPERAM: OPtimising thERapy to prevent Avoidable hospital admissions in the Multi‐morbid elderly’ supported by the European Union's Horizon 2020 research and innovation programme under the grant agreement No 6342388, and by the Swiss State Secretariat for Education, Research and Innovation (SERI) under contract number 15.0137. The opinions expressed and arguments employed herein are those of the authors and do not necessarily reflect the official views of the EC and the Swiss government.

We gratefully acknowledge the valuable input of the expert panel in the validation of the trigger tool. Participating experts included: Christine Baumgartner, University of Bern; Manuel Blum, University of Bern; Dominique Bonnet‐Zamponi, Assistance Publique Hôpitaux de Paris, Centre of Pharmaco‐epidemiology; Pascale Cornette, Université catholique de Louvain; Paul Jansen, University Medical Centre Utrecht; Louise Mallet, Université de Montréal; Zachary Marcum, University of Washington; Ariane Mouzon, Université catholique de Louvain; Denis O'Mahony, University College Cork; Mirko Petrovic, Ghent University; Sarah Slight, Newcastle University; Annemie Somers, Ghent University; Stephane Steurbaut, Vrije Universiteit Brussel; Patricia van den Bemt, Erasmus University Medical Centre; Tischa van der Cammen, Delft University of Technology.

We thank Stefanie Hossmann and Sven Trelle from the Clinical Trials Unit of the University of Bern, for the work performed to embed the DRA adjudication method in the electronic data collection for the OPERAM trial. We also thank Séverine Henrard from the Université catholique de Louvain for her advice in statistical analysis.

Contributors

S.T. conceptualized and designed the study, performed the literature review and pilot test, performed analysis and interpretation of data resulting from the validation, pilot and reliability studies and drafted the DRA adjudication guide. O.D. and A.S. conceptualized and designed the study, participated in the development and validation of the DRA adjudication guide and performed analysis and interpretation of data resulting from the validation, pilot and reliability studies. J.B.B. participated in the development and validation of the DRA adjudication guide and performed the pilot test. B.B., J.D. and N.R. participated in the development and validation of the DRA adjudication guide. S.M., F.V., I.W., A.V., C.F., C.S., S.C. and D.O.M. participated in the reliability study. S.T. drafted the initial manuscript with contributions from O.D., A.S., J.B.B., B.B., S.M., D.O.M., S.C., J.D., C.F. and I.W. All authors read and approved the final manuscript.

Supporting information

Supporting Information S1 Overview validation trigger tool for drug‐related admissions (DRA) in older persons using a two‐round online modified Delphi technique with an international expert panel

Supporting Information S2 The drug‐related admissions adjudication guide

Supporting Information S3 Summary adjudications of 16 cases by pairs of raters from four European centres

Thevelin, S. , Spinewine, A. , Beuscart, J.‐B. , Boland, B. , Marien, S. , Vaillant, F. , Wilting, I. , Vondeling, A. , Floriani, C. , Schneider, C. , Donzé, J. , Rodondi, N. , Cullinan, S. , O'Mahony, D. , and Dalleur, O. (2018) Development of a standardized chart review method to identify drug‐related hospital admissions in older people. Br J Clin Pharmacol, 84: 2600–2614. 10.1111/bcp.13716.