Patient safety, cost-effectiveness, and quality of life: reduction of delirium risk and postoperative cognitive dysfunction after elective procedures in older adults—study protocol for a stepped-wedge cluster randomized trial (PAWEL Study)

Eligible patients are aged 70 years or older and scheduled for elective surgery (cardiac, thorax, vessels, proximal large joints or spine, genitourinary, abdominal, or general elective surgery procedures) with a planned duration of surgery of at least 60 min (cut-to-suture time) under general, spinal, or regional anesthesia. As delirium is a high risk in dementia and frailty, we include patients with dementia or frailty who can consent to the trial or whose substitute decision-makers provide consent.

Patients undergoing emergency surgery procedures, patients unable to consent due to insufficient mastery of the German language or with newly discovered severe dementia (red flag: Mini Mental State Examination (MMSE) < 15, Montreal Cognitive Assessment (MoCA) < 8) without a substitute decision-maker, patients with a poor clinical prognosis (expected survival of less than 15 months), and patients who have a long driving distance to the study site (> 120 km) are to be excluded from the trial.

The recruitment procedure is described in Fig. 2. Patients who are not able to consent may also be recruited if the legal guardian consents to the patient’s study participation, given that such patients are especially at risk of developing delirium and POCD after surgery [10], and they could eminently benefit from the intervention.

The intervention implements a cross-sectoral all-encompassing multimodal delirium prevention and management approach, and will be carried out in each study center after the center is randomized to take part in the intervention phase of the trial:

Given the patient-centered care research implementation goal of this study, the outcomes of patients undergoing the intervention will be compared to the outcomes of patients receiving treatment as usual (TAU) provided by the centers before the randomized start of the intervention.

Patient group involvement and staff involvement has been implemented in an earlier version of this intervention approach adapting the HELP (Hospital Elderly Life Program) structure (see [18]). In addition, we will implement focus groups at the general practitioner level post hospital stay to further evaluate patients’ perception of the intervention itself.

The training modules are expected to play a fundamental role in improving adherence to the intervention protocols. They will be supplemented by additional talks, case discussions, and a web-based knowledge base providing webinars, training videos, and so forth. Moreover, a support system for the intervention teams will be established comprising an email service and a telephone hotline, as well as a data pool on a project server. Modules prescribed by the intervention team will be monitored, and the time span of the intervention will be daily documented. Furthermore, adherence to the manuals and “prescription checklists” will be tested by unannounced visits to every site.

Adverse events and serious adverse events (SAEs) will be recorded and documented. Falls, strokes, infections, and other severe perioperative complications (death, reoperation, pneumonia, sepsis) are to be expected in this patient group independently of the intervention, while SAEs related to the intervention are expected to be very rare. An example of a study-related SAE could be, for instance, a patient’s fall during early mobilization by the delirium companion during the active intervention phase. The feasibility of the project and the occurrence of SAEs will be assessed every 3 months by the REDMB, who will also carry out audits every year. In the case of substantial differences between the SAEs in the different groups, this will be discussed by the REDBM.

The criteria for discontinuing the intervention for a participant will be death; or study withdrawal requested by the patient, guardian, or authorized relative; or (re)operation of more than 1 h (cut-to-suture time) during the first perioperative week, given that in this case a delirium or POCD cannot be precisely assigned to the first operation; or malignancy surgery with prognosis under 15 months, neoadjuvant chemotherapy, brain radiotherapy, or primary metastasis surgery for pancreas or bronchial cancer.

The assessment of all the outcomes will always be performed by trained assessors, who will be blinded for the intervention. Specifically, delirium raters will be told that the data will be used for validating a delirium risk score. Staff will be instructed not to reveal the nature of the intervention to these assessors.

The primary outcome will be delirium prevalence. It will be measured by daily delirium screening (I-Confusion Assessment Method-based scoring system for delirium severity (I-CAM)/CAM-S)) [28, 29] over 7 days after surgery and after 2 and 6 months, the Nursing Delirium Screening Scale (NuDESC) [30] (days 2 and 6 after surgery), a chart review at discharge applying the DSM-V delirium criteria as a reference standard [31], and the clinical evaluation.

The CAM [32], with its four-step diagnostic algorithm, is a widely used screening test for assessing delirium. Originally developed from the DSM-III-R, it is now predominantly used for delirium screening and research according to the DSM-IV and DSM-V criteria. It has been operationalized and translated into German [33], and then revealed a high sensitivity of 0.77 in a cohort of geriatric patients with a high prevalence dementia, and a specificity of 0.96–1.00 with excellent inter-rater reliability (Cohen’s κ = 0.95 (CI 0.74–1.0) for the algorithm, single item’s κ values varied between 0.5 and 1).

The I-CAM (I for ICD-10) [28] extends the German version of the original CAM adding abnormal psychomotor activity, to allow taking also the ICD-10 delirium diagnosis as a reference standard and assessing the motor delirium subtypes as well. The CAM-S [29] is a CAM-based scoring system for assessing delirium severity and was operationalized for use with the German version of the I-CAM. As the CAM might be confounded by the fluctuating nature of delirium, we use a chart-based review [31] filled out by trained medical staff at discharge to evaluate for fluctuations in sleep–wake rhythm or psychomotor activity indicating delirium.

The NuDESC is a five-item scale based on nurses’ observations assessing disorientation, inappropriate behavior and communication, hallucinations, and psychomotor retardation over a 24-h period. For the German version of the NuDESC, in a sample of patients after elective surgery, a sensitivity of 0.98, a specificity of 0.92, and an inter-rater reliability of 0.83 were observed [34].

The secondary outcomes will be: delirium duration as described in the primary outcome assessment; prevalence of POCD 2 and 6 months after surgery; and persistence of POCD after 12 months. The prevalence of POCD will be measured by the following neuropsychological test battery: the Montreal Cognitive Assessment (MoCA) [35], the digit span backwards [36], the Trail Making Test A and B (TMT A and B) [37], and cognitive performance measured with the continuous nonstandardized test values of these scales. A cognitive deficit is defined as the presence of a test value of ≤ 0.5 standard deviations, normalized for age, gender, and education, in one of these test procedures.

The MoCA is a brief cognitive screening test for assessing cognitive impairment among older people. The test assesses multiple cognitive domains including visuospatial ability, executive functions, memory, attention, language, abstraction, and orientation. The MoCA has high sensitivity (0.90) and specificity (0.87) to detect individuals with mild cognitive impairment and distinguish them from cognitively intact older people [35] and is available in three parallel versions.

The digit span backwards is commonly used to assess working memory capacity. Participants are required to recall a sequence of spoken digits in reverse order. For people older than 70 years, the digit span backwards had a test–retest reliability > 0.60 and an internal consistency of 0.882 [38].

The TMT is a widely used instrument in neuropsychological assessment that measures the speed of scanning and visuomotor tracking, divided attention, and cognitive flexibility [36]. The test consists of two parts, A and B. TMT A requires an individual to draw lines sequentially connecting consecutive numbers from 1 to 25. TMT B involves drawing a similar line, connecting an ascending sequence of numbers and letters in an alternating manner. In a sample of healthy older adults, Part A had a test–retest reliability of 0.78 and Part B of 0.73, and Part A an inter-rater reliability of 0.99 and Part B of 0.93 [39]. In a sample of elderly volunteers, Part B had a sensitivity of 0.63 for cognitive dysfunction, 0.72 for dementia, and 0.77 for AD, and a specificity of 0.89 [40].

For baseline assessment, the following variables will be evaluated. Basic sociodemographic patient information to be collected includes age, gender, weight, height, dominant hand, marital status, immigrant background, educational level, occupation, living arrangements, nicotine consumption, alcohol consumption, falls, and statutory level of care dependency. The self-reported subjective memory impairment (SMI) will be assessed, for which subjects will be asked “Do you feel that your memory is getting worse?” (no; yes; I don’t know). If the patient answers yes, the patient will then be asked whether he or she is worried about this (no; yes, that worries me; yes, that worries me very much; I don’t know, no answer) [41]. The personal medical history, including comorbidities, is to be quantified by the expanded version of the Charlson’s comorbidity Index (CCI) [42]. The patient’s preoperative physical status will be assessed by the American Society of Anesthesiologists Physical Status classification (ASA) [43]. The risk of stroke will be estimated with the CHA₂DS₂-VASc score (Congestive heart failure/left ventricular dysfunction, Hypertension, Age > 75, Diabetes, Stroke/transient ischemic attack/thromboembolism, Vascular disease, Age 65–74, Sex category) [44]. Laboratory results (hemoglobin, sodium, creatinine, total amount of protein, C-reactive protein), the history of delirium in the past, and a neurological examination will be recorded. The grip strength as a frailty marker will be measured with the Jamar® Hydraulic Hand Dynamometer. Hearing and visual integrity will be tested by the whisper and visual acuity tests [45]. Functional mobility and risk of falls will be investigated by the Timed Up and Go Test (TUG) [46]. Anxiety and depression will be assessed by the Patient Health Questionnaire (PHQ-4) [47]. The health-related quality of life will be measured by the EuroQol five dimensions questionnaire (EQ-5D-5 L) [48, 49] (this version includes five levels of severity in each of the existing five EQ-5D dimensions) and by the 12-Item Short Form Survey (SF-12) [50, 51]. Nutritional status will be assessed by the Mini Nutritional Assessment Short Form (MNA-SF) [52, 53]. Functional status will be evaluated with the Hamburg Classification Manual [54] version of the Barthel Index [55]. Frailty will be analyzed with the Clinical Frailty Scale of the Canadian Study of Health and Aging (CSHA Clinical Frailty Scale) [56]. Pain will be measured by the Numerical Rating Scale of Pain (NRS Pain) [57]. Sleeping behavior will be investigated by the Pittsburgh Sleep Quality Index—Basic (PSQI-Basic), a four-item version of the PSQI [58]. Sleep apnea will be screened for by the STOP-BANG questionnaire (Stop-BANG) [59]. Olfactory function will be scrutinized by the Sniffin’ Sticks 12 version [60]. Cognitive decline will be assessed by the very short version (seven items) of the Informant Questionnaire on Cognitive Decline in the Elderly (IQCODE) [61]. The subjective burden of the family caregivers will be measured by the German Zarit Burden interview (G-ZBI) [62, 63]. Classification as a geriatric patient, will be assessed with the Geriatric-Check using the geriatric concept of Baden-Württemberg (GC) [45]. Finally, the patient’s medications will be recorded, including the type, dose, and frequency of administration.

For planned analysis, the following variables are currently anticipated: anxiety and depression will be evaluated with the PHQ-4 questionnaire, the health-related quality of life with the EQ-5D-5 L questionnaire and the SF-12 survey, the nutrition status with the MNA-SF, the functional status with the Barthel Index, and the frailty with the CSHA Clinical Frailty Scale and hand grip. The duration and extent of use of physical restraints and patient care attendants will also be assessed. Moreover, the relevant aspects of behavior in daily life will be assessed by the Nurses’ Observation Scale for Geriatric Patients (NOSGER II) [64, 65], the presence of cognitive decline will be assessed by IQCODE evidenced by relatives, and the subjective burden of family caregivers by G-ZBI. The timelines are summarized in Fig. 3.

For planned validity assessment of postoperative delirium, the validity of the I-CAM-S and NuDESC vs the Richmond Agitation-Sedation Scale (RASS) [66] and vs chart review will be evaluated. The RASS is a simple observational scale that quantifies the level of consciousness, assessing both sedation and agitation.

The study started in November 2017, and the recruitment phase will last until April 2019.

There will be three follow-up examinations after 2, 6, and 12 months, with the last planned for April 2020. We plan to complete the evaluations and cost analyses by December 2020.

From the patient’s perspective, the study will start with a preadmission screening (T0, at most 3 weeks before admission) (see Fig. 3). Admission (T1) will be followed by surgery and then daily delirium screening for 7 days (T2–T8) during hospitalization. A final discharge assessment (T9) will take place 1–8 days after surgery. If, at this point, the patient is considered to suffer from delirium, then delirium screening and monitoring will continue for another 2 weeks, so long as the patient remains hospitalized in the center. The discharge examination (T9) will take place, at the latest, 3 weeks after the elective surgery, with the participation of a physician who will note any postoperative complications. Follow-up assessments will be performed after 2 months (T10), 6 months (T11), and 12 months (T12). Moreover, the patient’s relatives will be asked at T12 about their care burden using the G-ZBI, and the cognitive state of the patient will be assessed using the IQCODE instrument. If a patient is unable or unwilling to go to the regional center for any follow-up assessment, the local PAWEL team will offer the assessment either at some hospital closer to the patient or at the patient’s general physician’s practice. If the patient is not able to go to any of these places, a team member will assess the patient at the patient’s home. Finally, if the patient refuses this option, he or she will be asked for a telephone interview.

The power analysis for the primary outcome of delirium prevalence assumes a reduction in the outcome from 25% to 15% [20] as a result of the intervention. A conventional analysis to detect differences in proportions (delirium rate) between the intervention and control groups, using Fisher’s exact test, results in a total of 514 patients with a 1:1 randomization, given a power of 1 – β = 0.80 and an α error of 5%. Using the adjustment form proposed by Woertman et al. [67] with five crosspoints in a stepped-wedge design, with a maximum of 50 patients per cluster per period and an intracluster correlation of 0.01, leads to a correction factor for the stepped-wedge design of KF = 2.63, and therefore a number of 514 × 2.63 = 1351 patients. Assuming a dropout rate of 15%, this leads to a total requirement for about 750 patients per arm. The minimum number of clusters is given by the ratio of the total number of patients to the product between the number of crossing points and the patients per cluster per period, and is 15.

For the secondary outcomes, a reduction of the persistent cognitive deficit from 20% to 10% results in 428 patients using the conventional analysis with Fisher’s exact test and a 1:1 randomization, a power of 1 – β = 0.80, and an α error of 5%, which after the design correction leads to a requirement of 1079 patients.

Thus, the planned sample size is of at most 1500 patients (see Fig. 1 with the numbers per period). Furthermore, relatives of the patients will be asked about the burden of caring and the cognitive status of the patients.

The study takes place in Germany in the state of Baden-Württemberg. The study will be performed in university hospitals in Tübingen, Freiburg, and Ulm, and in tertiary care hospitals (one in Stuttgart and two in Karlsruhe). In total, five medical centers (one per city) with 12 surgical departments (three in Stuttgart and Tübingen, and two at the other centers) are included. To balance the surgical subspecialties per center, recruitment is restricted to a maximum of 2/3 from one of the following surgical subspecialties: orthopedics, vessels and cardiac, or abdominal. We intend to maintain this ratio for every center and for every 12-week interval of the study.

Patient recruitment will be carried out by a medical specialist. Study staff, including doctors and scientists, will provide the necessary information and obtain the written informed consent from the patients, and whenever possible also from one relative (to assess the caregiver burden and the cognitive status of the patient).

The medical directors and the managers of anesthesia and nursing at the hospitals included in the study have provided a letter of intent expressing their interest in introducing and evaluating preventive procedures, and their willingness to recruit the necessary number of patients. Hence, the recruiting objective is realistic. However, if a department fails to meet its recruitment target (17 patients for the departments in Stuttgart and Tübingen, 25 patients for the others) during one period, concrete and binding measures will be taken to reach the planned number of patients. In this case, the center will commit either to increase the recruitment in its departments or to include another surgical department. If several departments do not reach their target, then the five recruitment blocks, each lasting 3 months, will be extended to 4 months, adding 33% more time to recruit. In this case, for the patients recruited during the last period of the project, the POCD assessments would be completed after 6 months instead of after 12 months.

The collection of the clinical data during all stages of the study (preadmission, admission and postoperatively) will be performed exclusively by authorized staff of the study sites, using electronic Case Report Forms (eCRFs) via the web-based electronic data capture (EDC) system secuTrial® from interActive Systems GmbH. The access to the secuTrial® platform is protected by an authentication process, and the transmission of data between the study centers and the secuTrial® server is protected by a secure connection using Secure Sockets Layer (SSL) encryption. All collected data will be checked for reliability and validity every 3 months. Data access will be granted initially to project and data management and consortium leaders. All the data from scales and neuropsychological tests, as well as the clinical data and the data for the health-economic analysis, will be stored in a pseudonymized fashion. After the end of the study, all paper and electronic documents will be stored for at least 10 years more.

For the primary outcome, logistic regression analyses with cluster adjustment are planned.

For the secondary outcomes, the changes in the medication and the care burden of the patients’ relatives, we plan analyses using mixed linear and logistic regression models, with fixed and random factors for estimating heterogeneity over clusters [68].

Subgroup analyses for specific patient groups (e.g., patients with cardiac surgery regarding cognitive status or patients with brief hospital stays/brief surgeries) are planned.

For planned validity assessments of postoperative delirium, the validity of the I-CAM-S and NuDESC vs the RASS will be evaluated as screening tools for postoperative delirium. Diagnostic test performance of the tools will be evaluated by receiver operating characteristics (ROC) analysis.

Missing data (due to nonadherence of a department or a particular patient) will be documented. We will apply mixed-effect models, which handle the missing data without using any imputation method. Mixed-effect models assume that the missing data are missing at random (MAR); that is, when the probability that an outcome is missing is related to some other fully observable variable in the model, but not to the variable with the missing value itself [69].

Regarding the health-economic evaluation, a micro-costing analysis will be carried out using data from the clinical administration of two hospitals (Tübingen and Stuttgart). Moreover, for the patients insured by a specific health insurance (AOK Baden-Württemberg) who consent, the health and care insurance costs of the 12 months before and after surgery will be calculated using a difference-in-difference approach [70]. In this case, the target parameter is the cost per avoided POCD. The cost-effectiveness of the intervention will be calculated comparing the costs of the intervention group and the control group. In addition, quality-adjusted life years (QALYs) will then be calculated using data from the EQ-5D-5 L form about health-related quality of life, and applied for calculating the incremental cost-effectiveness. When possible, subgroup analyses will be carried out, for example, by age or gender, to determine whether cost-effectiveness is different in the corresponding groups. Sensitivity analyses will also be performed. The evaluation will be carried out following the recommendations of the “Good practice secondary data analysis” (GPS).

Publications about the results of the trial will be submitted to international journals after being approved by the project committee. Use of professional writers will not take place, and authorship eligibility criteria are specified in the trial contract and comply with International Committee of Medical Journal Editors (ICMJE) statements [71].

An analysis of the perioperative patient pathway will be developed as an outcome of the project. The goal of this analysis is to adapt the intervention to everyday care routine and to facilitate its subsequent implementation. To this end, doctors and other members of the geriatric centers will meet in 10 focus groups of 10 participants each. They will discuss exploratory questions about the needs and sensitivity to delirium and POCD, the knowledge about delirium, dementia, and depression, and other aspects of the study. To ensure the implementation of the results, they will be disseminated through workshops: at a training course; at the annual update on dementia (in cooperation with the medical association of Baden-Württemberg); and at the day of general medicine of the Institute of General Medicine and Interprofessional Care in Tübingen. A main result of the project will be a care pathway including information tools (e.g., flyers) and training modules for doctors and doctors’ assistants.

This is an intervention study in which existing guidelines and recommendations will be implemented, including the S3 guideline of the German Society for Anesthesiology and Intensive Care Medicine (Deutsche Gesellschaft für Anästhesie und Intensivmedizin (DGAI)), the S3 guideline of the Working Group in Geriatric Traumatology (Arbeitsgemeinschaft Alterstraumatologie (AG Alterstraumatologie)) of the German Association of Trauma Surgery (Deutsche Gesellschaft für Unfallchirurgie (DGU)), the guidelines of the National Institute for Clinical Excellence (NICE), the guidelines of the American Geriatrics Society (AGS), and the guidelines of the Difficult Airway Society (DAS). Ethics approval according to the occupational regulations (Berufsordnung (BO)) has been obtained for all centers. All relevant changes will be communicated to the scientific advisory board, the Regional Ethics and Data Monitoring Board (REDMB), the relevant institutional review boards (IRBs), and the clinical trial registry. Given that there are no extra risks, the usual health insurance or civil liability will cover the risks of the trial, and there will be no additional medical liability insurance for the study. For the additional outpatient assessments, patients and their relatives will be covered by travel insurance.

In the preadmission assessments, it is to be expected that approximately 40% of the examined patients will have slight or severe cognitive impairment [72]. However, this does not necessarily restrict their capacity to consent or their legal capacity. Following Kim and Caine [73], we consider that this only occurs when transitioning to a more severe dementia (MMSE < 19). For those patients, the possible damage would be the self-stigmatization as a dementia-endangered person, but the benefit of the intervention would prevail because of its stipulated reduction of delirium and POCD.

On the other hand, if a previously unrecognized severe cognitive disorder was found, the patient probably could not consent, requiring consent from a substitute decision-maker.