2.1 Study Design and Setting

A single-centre cohort study at LMU University Hospital Munich, Germany was conducted from March 2022 to October 2022 with the primary aim of developing a drug-based risk score for POD [23]. The study was approved by the Ethics Committee of LMU University Hospital Munich (no. 23-0041). This is a secondary sub-analysis that included inpatients over 65 years of age who underwent surgical intervention in orthopaedics or trauma surgery and who received pharmacist-led medication reconciliation at hospital admission. In the primary study [23], patients with preoperative delirium and delirium due to alcohol withdrawal, as indicated in patient records, were excluded. This sub-analysis further excluded patients with missing information on drug dose.

Patients from three orthopaedic and trauma surgery wards were included, which all participated in the project ‘gertrud – age-appropriate proactive health care’ with a focus on reducing postoperative complications in older adult patients, specifically delirium [29]. Therefore, ward staff were especially trained for delirium awareness, and trained nurses assessed delirium two to three times a day using the four A’s test (4AT) [30]. The 4AT considers alertness, attention assessment (through the month backwards test), the four-item Abbreviated Mental Test (AMT4: age, date of birth, current place and year), and evidence for acute change or fluctuating course. For patients with a 4AT ≥ 4, physicians checked for the presence of delirium and, if present, documented a diagnosis code according to the International Classification of Diseases, 10th Revision (ICD-10).

Pharmacist-led medication reconciliation is routinely performed at admission for all surgical patients from Monday to Friday. This results in a detailed medication history of drugs (prescribed, over-the-counter, and phytopharmaceuticals), including long-term and on-demand medication, which is saved as admission medication in the electronic medication record Meona® (Mesalvo GmbH Freiburg, Germany).

2.2 Data Collection

Drugs and dosages of the admission medications were retrieved from Meona®. Sociodemographic and laboratory data as well as disease-related information of the patients (diagnoses coded according to ICD-10, 4AT scores, chart entries, and data from the preoperative anaesthesia assessment) were collected from the electronic patient information system (i.s.h.med®, Cerner Corporation, North Kansas City, USA). Dementia status was recorded according to ICD-10 codes (F00.-*, F01, F02.-*, F03, F05.1), chart review (keyword: dementia), or the use of anti-dementia drugs. The estimated glomerular filtration rate (eGFR) was calculated using the chronic kidney disease epidemiology collaboration (CKD-EPI) equation [ml/min/1.73 m2] [31].

2.3 Retrospective Assessment of Postoperative Delirium

For each inpatient stay, POD was assessed up to 7 days post-surgery according to the documented ICD-10 codes (F05.0, F05.1, F05.8, and F05.9). Additionally, as validated in previous studies [32, 33], a subsequent chart review was performed (keywords: delirious, confused, disoriented, disturbed attention, hallucination, restless, and agitated). POD was considered to be present if either an appropriate ICD-10 code was documented or the chart review clearly indicated the development of POD. The assessment occurred independently of the knowledge of AC burden scores. After the initial assessment by a pharmacist, a physician confirmed the final POD rating.

2.4 Extension of the MARANTE Scale

The authors of the original MARANTE scale published dose ranges for 41 AC drugs and suggested the completion of additional country-specific or newly developed AC drugs [16]. Thus, we defined the dose values (minimal effective value, main dose, and maximal effective value) for all the remaining AC drugs determined using the GerACB score in our patient cohort. To adapt the potency values, we assigned ACB 1 as ‘low potency’ (potency value of 1) and ACB 2 and 3 as ‘high potency’ (potency value of 2).

Following the methodological approach of the original MARANTE scale, we retrieved dosage information from multiple international sources and invited an expert panel to rate dosage concepts. We determined the main indications according to the World Health Organization (WHO) Collaboration Centre for Drug Statistics Methodology [34]. As international reference sources for dosage information for the main indication, we consulted UpToDate® [35], the British National Formulary [36], and the Geriatric Dosage Handbook [37]. Second, we retrieved information from the German Summary of Product Characteristics (SmPC) [38].

The expert panel included three experts with expertise and experience in drug use in older adult patients (one clinical geriatrician and two clinical pharmacists with long-term experience in drug information). We conducted two rounds. First, the experts filled in the remaining dosage values (minimal effective value, main dose, and maximal effective value) for the remaining AC drugs based on the reference sources, their clinical experience, and the available dosage forms. Once the rated dosage values were collected, they were evaluated for consensus. Consensus was reached when at least two experts rated an identical dosage value. For drugs for which this was not possible, we conducted a second round in which the experts received anonymous ratings from the first round and were asked to revise their ratings. After the second round, all dosage concepts were determined through consensus.

2.5 Assessment of AC Exposure

For each patient’s admission medication, AC exposure was calculated according to the following three scores/equations:

1.

The German Anticholinergic Burden score (GerACB) [39];

2.

The Muscarinic Acetylcholinergic Receptor ANTagonist Exposure scale (MARANTE) [16];

3.

The German Drug Burden Index (GerDBI).

The GerACB assigns values from one to three to drugs based on their AC potency. Both the MARANTE scale and the GerDBI are equations that consider the dosage. The MARANTE scale links dose and potency, while the GerDBI does not consider potency but additionally includes sedative drugs. Fig. 1 shows an overview of the score calculations according to these three scores. The GerDBI is based on the Drug Burden Index by Hilmer et al. [17, 40] and includes drugs available in Germany. It was developed as part of the ‘COFRAIL’ project [41] (funding code 01VSF17053), and details will be published elsewhere.

Fig. 1

Composition of the scores for assessment of the AC exposure. For each listed drug, an individual burden value is calculated depending on the assigned potency (GerACB [39]), potency and dose (MARANTE [16]), or only dose (GerDBI). To determine a patient’s overall burden, individual burden values are summed up. AC anticholinergic, GerACB German Anticholinergic Burden Score, GerDBI German Drug Burden Index, MainD maintenance dose, MARANTE Muscarinic Acetylcholinergic Receptor ANTagonist Exposure Scale, MaxEV maximal effective value, MinEV minimal effective value

For the calculation of dose-related equations, the average daily dosage was needed. On-demand medication was only rated if the intake frequency could be obtained from the pharmacist-led medication reconciliation; otherwise, drugs were not rated. Cumulative AC exposure was reported either as a continuous burden value or as a categorical burden (no burden, low burden, or high burden) to allow comparability of the scores/equations. For each score, established burden classifications were used for no burden, low burden, and high burden: GerACB (0, 1–2, ≥ 3), MARANTE (0, 0.5–1.5, ≥ 2) and GerDBI (0, > 0 < 1, ≥ 1) [16, 39, 40].

2.6 Statistical Analysis

Descriptive statistics are reported as means ± standard deviation (SD), median and interquartile range (IQR), or as frequencies with percentages. Groups were compared using Mann–Whitney U test or chi-squared test. Pairwise or overall agreement of AC burden classifications between all three scores was assessed using kappa statistics, and the agreement classification followed Landis and Koch [42]. Associations of AC burden (estimated through the GerACB, MARANTE, and GerDBI) with POD were determined via multivariable logistic regression analyses. For adjustment of co-variables, significant variables (p < 0.05) from univariable analysis were added to a stepwise forward multivariable logistic regression model. Sex was added as a forced-in variable. Multicollinearity of co-variables was determined through a correlation matrix. The performance of the model was evaluated by the area under the curve (AUC) of receiver operating characteristic (ROC) analysis. To estimate the score performance, the sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were obtained in an unadjusted analysis. All calculations were performed with SPSS Statistics® version 29.0 (IBM Corp., Armonk, NY, USA). Illustrations were created using Adobe Illustrator® version 27.0 (San Jose, CA, USA). P values < 0.05 were considered statistically significant.

The sample size was calculated considering six variables included in the multivariable analysis, ten outcome events per variable [43], and an estimated POD prevalence of 20% [3], for which a minimum of 300 patients were estimated.