Reducing inappropriate psychotropic drug use in nursing home residents with dementia: protocol for participatory action research in a stepped-wedge cluster randomized trial

We will use a two-armed cluster RCT with a stepped-wedge design to allow each NH to participate in the intervention phase and to increase the study’s power (Fig. 1). We expected that recruitment to a classic RCT design would be problematic because half do not receive the intervention, which can be off-putting. A waiting list procedure would also require including more NHs, and we want to avoid this because of the resource-intensive design in terms of time, dedication, and money. In addition to allowing for smaller sample sizes, other advantages of stepped-wedge designs are the possibility to compare both between- and within-cluster effects, as well as the ability to model effects over time [44].

The study duration will be 16 months, split into two 8-month periods. Based on a pilot study, we estimated that the problem analysis for PAR-RCT would take about 2 months. We plan to allow 6 months to implement the chosen interventions because a longer period may lead to a loss of enthusiasm among NH staff, and because we anticipate high turnover rates among residents [5] and staff [32, 39]. The coaching will also be time and cost intensive, and the amount of data collected will require a huge investment from research assistants and NH staff.

Although we believe that 8 months will be optimal for these reasons, we do recognize that it might be insufficient to bring about a change in NH practice. Therefore, we plan to use the stepped-wedge design to compare the 8 month intervention with a 16 month extended intervention that will allow us to study possible long-term effects. Using a computer program, 16 NHs will be randomized by fixed-block randomization into an intervention group or a deferred intervention group (8 NHs each). In Period 1, the intervention group (Fig. 1, green blocks) will start by implementing the PAR method, while the deferred intervention group will start with care as usual. After 8 months, Period 2 will start, and the deferred intervention group will start the intervention and the original intervention group will complete a second intervention phase.

The primary outcome will be the reduction of inappropriate psychotropic drug prescribing based on data collected at baseline, 8 months, and 16 months. The statistician will be blinded to the cluster randomization process, and the NH staff and researchers will be aware of their participation group by design. Given that we aim to intervene at the DSCU level and that residents are not directly subject to an intervention, we have no discontinuing criteria, trial stopping rules, or modifying allocations.

NHs will be recruited by organizing a national kick-off and using media channels to gain attention. Any interested NHs in the Netherlands can apply to participate and will be considered eligible if the board of directors and the client council agree to participate and are prepared to invest the requisite time. We will exclude NHs in which major organizational changes are expected during the study period or if other projects related to psychotropic drug use are currently running.

The study population will comprise residents with dementia who reside in psychogeriatric units (e.g., DSCUs). NHs can have several DSCUs, and several DSCUs from one NH can participate. However, DSCUs designed to deliver care for residents with Korsakov syndrome, acquired brain injury, Down syndrome, or young-onset dementia will be excluded. Although interventions will be aimed at the DSCU level, we will still need to gather data on residents to assess outcomes. Residents will be eligible to participate if they have a diagnosis of dementia according to the Diagnostic and Statistical Manual of Mental Disorders (fifth edition), have a life expectancy of at least 3 months, as judged by a physician, and provide written informed consent. We will also include users and nonusers of psychotropic drugs. It is anticipated that most residents will be at an advanced stage of dementia and may be mentally incompetent. The physician of the relevant DSCU will therefore be asked to assess a resident’s mental competence to provide informed consent. If they are not deemed competent, an employer of the registry office at each NH will send the informed consent form to a residents’ legal representative. In the absence of a response, a reminder request to return the informed consent form will be made once. If no response will be obtained, residents will not be included.

The PAR-RCT element will involve a cyclic approach of planning, acting, observing, and reflecting (Fig. 2) [42]. Half of the NHs will complete two cycles (i.e., those in the starting intervention group) and the other half of the NHs will complete one cycle (i.e., the deferred intervention group). Each NH will initially form a multidisciplinary project team (MPT) that will include, as a minimum, an internal project leader, a nursing staff representative, a psychologist, and a physician. An external coach will also participate in the MPT to facilitate the whole process.

The action research cycle will start with the researchers carrying out a problem assessment. We will start with the observation phase because effective interventions should meet the needs identified by staff [45]. This assessment will focus on current daily practice and difficulties managing NPS, including inappropriate psychotropic drug use, and the results will be presented to the MPT. Next, the MPT and the coach will reflect on these results. In this reflection phase, opportunities for improvement will be identified before the MPT moves on to formulate goals to tackle the identified problems. Two expected outcomes are predicted. First, a problem analysis may reveal that the evaluation of psychotropic drugs is inappropriate. In this instance, NHs may opt to pursue a clear work policy between physicians, to clarify the evaluation of psychotropic drugs, or to do medication reviews more frequently. Second, early detection of NPS may be identified as a problem. In this instance, they may recommend e-learning on NPS or implementing the routine use of the Neuropsychiatric Inventory-Nursing Home version (NPI-NH) by nurses for monitoring. How an NH intervenes will be decided by the coach and the MPT, with only indirect input from the researchers. The identified goals will be converted into an intervention and implementation plan. The coach will support this process and the researchers will assess whether the plan aims to reduce inappropriate psychotropic drug use. In summary, the information gained during the problem assessment will be used to formulate and operationalize the goals for the intervention and implementation plan (planning phase), after which NHs will implement the interventions (action phase).

NHs will be provided with a toolkit that contains several multidisciplinary care programs, such as GRIP or PROPER, as well as person-centered psychosocial interventions. The toolkit is a bundle of existing evidence and practices from which NHs can draw. They will be free to implement any intervention, either from the toolkit or elsewhere, provided the selected interventions match the problems identified in managing NPS and psychotropic drug use. Table 1 provides an overview of the toolkit’s content. In contrast to traditional PAR, the primary outcome (reduction of inappropriate prescribing of psychotropic drugs) is fixed in our study.

During the implementation phase, the MPT will have regular meetings with the coach to monitor progress (observation phase). At 8 months, NHs that start in the intervention group will receive feedback on the level of inappropriate psychotropic drug prescribing. The results of this interim analysis should allow them to reflect on changes in inappropriate prescribing (reflection phase), making it possible to adapt the intervention and implementation plan (planning phase), and act accordingly (action phase). When there is little or no reduction in the rate of inappropriate prescribing of psychotropic drugs, the MPT will be guided to consider choosing different interventions, changing their implementation strategy, or increasing the effort put into implementation.

Coaches will be responsible for the following aspects: (1) advising and supervising the internal project leader and the MPT; (2) being present at meetings of the MPT; (3) advising on the logistical aspects of the research and improvement process in the organization, including problem analysis, planning, multidisciplinary embedding, and sustainability; (4) offering substantive knowledge about psychotropic drugs, inappropriate prescribing, and reducing psychotropic drug use; (5) supporting the organization to get set up; (6) providing access to tools, methodologies, and learning networks; and (7) advising on quality assurance and dissemination of the results. We plan to recruit eight coaches through the Vilans Center of Expertise for Long-term Care, who in turn, will source them internally or through cooperating partners. The coaches must be knowledgeable about dementia and have previous consultation expertise in nursing home organizations.

Each coach will support two NHs. They may spend an average of 3 h per week with an NH over each 8-month intervention period. The 8 NHs that start in the intervention group and have an extended intervention duration will only receive coaching in the first 8 months, which will end after discussing the interim results. It is expected that time-intensive weeks will be compensated for by weeks in which little to no input is needed. Coaches will not be required to use all allocated hours, and the actual time spent with each NH is to be determined in consultation with the relevant MPT. The coaches will be asked to keep a logbook in which they will be asked to write down the number of hours spent on coaching, any agreements made with the NH, and other findings they consider relevant. They will also participate in monthly supervision sessions, led by a trainer at Utrecht University. The aim of these sessions is to discuss any difficulties experienced, to find suitable solutions, and for coaches to exchange tips. The extent to which a coach succeed will be addressed in the process evaluation.

It is essential for credibility that researchers acquire information about the quality of an intervention and its implementation in a given study [46, 47]. Leontjevas et al. (2012) proposed a process evaluation model based on first- and second-order data, which we will adopt in this study. The first-order data provide information about the internal and external validity, comprising of the sample quality (e.g., recruitment, randomization, and reach) and the intervention quality (e.g., relevance, feasibility, and extent to which an intervention was performed). The second-order data are then examined (implementation knowledge), such as implementation components that were delivered and received and the barriers to, and facilitators of, implementation. The internal project leader and coach will each receive a digital questionnaire to assess these aspects. We will follow this up with a semi-structured telephone interview to address ambiguities and to seek additional comments.

An overview of the outcomes and measurement instruments is provided in Table 2. Demographic data will be extracted from medical files by the researchers, including residents’ age, sex, type of dementia, pro re nata psychotropic drug use, date of last medication review, and date of admission to the DSCU. Cognitive abilities will be examined using the Cognitive Performance Scale (CPS), a reliable and valid scale that rates cognition, communication, activities of daily living and consciousness from 0 (intact) to 6 (very severe impairment) [48]. The instrument has a sensitivity and specificity of 0.94 and has been validated against the Mini Mental State Examination [49]. The CPS will be used by nursing staff, on paper, in the presence of a researcher.

Psychotropic drugs will be categorized according to the Anatomical Therapeutic Chemical (ATC) classification [50]. The primary outcome is the appropriateness of psychotropic drug use, and this will be measured with the Appropriateness of Psychotropic drug use In Dementia (APID) index [23]. This index rates the prescription of regular psychotropic drugs for NPS in people with dementia, including antipsychotic, anxiolytic, hypnotic, antidepressant, anticonvulsant, and anti-dementia drugs. Psychotropic drugs administered for dementia, sleep disturbance, and delirium will be scored with the APID index, but those prescribed for psychiatric disorders will not be scored. Treatment appropriateness will be measured on seven domains: indication, evaluation, dosage, drug-drug interactions, drug-disease interactions, duplications, and therapy duration. For each domain, a score between zero (appropriate) and two (inappropriate) can be given, allowing an overall appropriateness score to be calculated (weighted sum score). The APID has been validated in residents of DSCUs in the Netherlands (intraclass correlation coefficient, 0.577–1.000) [23]. The researchers will extract information on psychotropic drug prescribing from medical records.

The secondary outcomes will be the frequency of psychotropic drug use, the frequency of NPS, and the QoL. The frequency of psychotropic drug use will be extracted by the researchers from medical files, and NPS and QoL will be assessed by nursing staff on paper, in the presence of a researcher. Proxy measures of NPS and QoL (nurse assessments) will be used on the assumption that residents lack capacity. NPS will be measured with the NPI-NH and the Cohen–Mansfield Agitation Inventory (CMAI). QoL will be measured using the visual analog scale (VAS) of the EQ. 5D (i.e., the EQ-VAS).

The NPI-NH measures the prevalence, frequency, severity, and associated caregiver distress of 12 neuropsychiatric symptoms. All symptoms are rated on Likert-type scales, with four-point scales used for frequency, three-point scales used for severity, and six-point scales used for caregiver distress. When a symptom is not present, the frequency, severity, and caregiver distress are not scored [51]. We plan to use the Dutch version of the NPI-NH, which has demonstrated high inter-rater agreement and validity as a rating scale [52].

The CMAI is the most used tool for determining the frequency of agitation and aggression [53]. The CMAI consists of 29 items, subdivided into three subscales: physical aggression, physically nonaggressive behavior, and verbally agitated behavior. Scores are rated on seven-point Likert-type scales, rating symptoms over the preceding 2 weeks from “never” to “several times an hour” [54]. The translated and validated Dutch version by De Jonghe and Kat (1996) will be used, which has established reliability (Cronbach’s α = 0.82) and construct validity (Inter-rater agreement = 0.89) [55‐57].

The EQ-VAS records a respondent’s self-rated health on a vertical scale ranging from “best imaginable health state” to “worst imaginable health state” [58]. In our study, we will use the two proxy-versions of the EQ-VAS reported in a previous Dutch study [59]. Nursing staff and family members will be asked to rate the residents’ QoL and health status, both from their own perspective and from the perspective of the resident. In addition, the social engagement of residents will be measured as an important proxy of QoL, using the Revised Index of Social Engagement (RISE) [60]. The RISE is part of the inter-RAI Long-Term Care Facilities Assessment System and consists of six dichotomous items related to social behavior [61‐63]. Its reported reliability and validity are considered sufficient [60].

We will gather information on inappropriate psychotropic drug prescribing, frequency of psychotropic drug use, NPS frequency, and QoL. Also, we will analyze current difficulties in managing NPS using a self-designed, digital questionnaire for NH staff (i.e., physicians, psychologists, and nurses) to assess the following: (1) the detection, analysis, treatment, and evaluation of NPS, and (2), efforts to prevent NPS, any views about NPS, and the presence of multidisciplinary cooperation. The questionnaire will not be used to measure an outcome; instead, it will only be used to examine problems that NH staff perceive when dealing with NPS and inappropriate psychotropic drug prescribing. Scores will be rated on four-point Likert-type scales ranging from “never/rarely” to “at all times” for the frequency measure, and from “not satisfied” to “very satisfied” for the extent of satisfaction. The questionnaire has been piloted in two NHs and adjusted based on user experiences.

To gain a more in-depth insight into the processes that play a role in managing NPS and the prescription of psychotropic drugs, researchers (CGK, CvT) will conduct several semi-structured interviews with members of the MPT (e.g., physicians, psychologists, nursing staff, and managers). Several interviews were trialed in a pilot study at two NHs, and the interview formats have been adjusted according to user experience.

We also plan to address the attitudes of health care staff toward the use of new interventions or treatments. Values and beliefs on these issues influence the degree to which innovations are initiated and implemented in clinical practice [64‐66], and considering the attitudes of professionals toward adopting new interventions can facilitate implementation [67]. Therefore, we will administer the Evidence-Based Practice Attitude Scale (EBPAS) developed by Aarons in 2004. This scale consists of four subscales for 15 items that are measured on five-point Likert scales, ranging from 0 (not at all) to 4 (to a very great extent). The EBPAS assesses the extent to which a professional (1) finds the intervention intuitively appealing, (2) would adopt an intervention if required by a supervisor, (3) has a general openness to trying new interventions, and (4) perceives interventions as being of limited clinical value and less important than clinical experience. We will use the Dutch version, for which the factor structure and reliability are comparable to the original version [68].

Furthermore, the organizational culture of the DSCUs will be assessed using a Competing Values Framework (CVF) scale [69]. This involves a questionnaire that consists of 24 items on four-point Likert-type scales, ranging from 1 (not characteristic) to 4 (very characteristic). It aims to provide insight into the cooperation between staff members and the working conditions and characteristics of the DSCU, and it suggests either a dominant culture type, a market type, an adhocracy type, or a hierarchy culture type. We will use a Dutch version of the questionnaire [70].

Both the EBPAS and the CVF scale will also be administered in the process evaluation, and they will be sent digitally to NH staff. Finally, a self-designed paper and pencil questionnaire will be used to assess the opinion of family members regarding the delivered care and degree of communication received with respect to NPS and/or psychotropic drug use.

For our primary outcome, we aim to detect a reduction in inappropriate psychotropic drug prescribing of 5 points minimum (standard deviation = 15) on the APID index from the baseline to the final measurement (16 months) [23, 71]. We expect that APID values will be nested within NHs, the main level of randomization in our design. Assuming an average size of 25 residents per NH, a power of 0.80, a significance level (alpha) of 0.05, we will need 284 psychotropic drug users. Given the multilevel design with two measurements after baseline, we anticipate that we will then need to increase this to a sample size of 364 (15 clusters) for an intraclass correlation coefficient of 0.1 [44] and a calculated design factor of 1.28 (the factor at which the original N has to multiplied). To allow for a cluster dropout of about 10% [72], and to obtain an even number of clusters, we plan to include 16 clusters with 364 residents in total. Given that approximately 60% of residents with dementia will be prescribed psychotropic drugs [20], this number will need to be further increased to 607 residents. Previous studies have shown that we can also anticipate a loss to follow-up of about 30% per year [5], which would amount to 40% in the 16 months in our study. However, in the event a resident dies or moves away from the unit, we will enroll the newly admitted resident, precluding the need to account further for attrition. Consequently, the case mix during our study will vary, and information from dropouts will need to be included in an intention to treat analysis.

Primarily, we will examine results between both arms; the intervention group and the control group. Secondarily, we will examine results between the short intervention duration (8 months) and the long intervention duration (16 months). All data will be entered in a secured digital data management program. After collection, the data will be checked for outliers and extracted into a statistical software package. Descriptive statistics will be used to compare the baseline data between groups. A multilevel model will then be used to study the effects of a tailored intervention and implementation plan on reducing inappropriate psychotropic drug prescribing, based on the methods described by Twisk [73]. This model will be used for both the primary and secondary outcomes to account for the dependency of information due to the repeated measurements and cluster randomization. A restricted iterative generalized least squares algorithm will be used to estimate the regression coefficients, and the Wald test will be used to obtain a P value for each regression coefficient. Models will include a fixed and a random intercept.

Analyses will be adjusted for baseline outcome values (e.g., inappropriate psychotropic drug prescribing and NPS), time, and confounders (e.g., cognitive abilities, as measured with the CPS, type of dementia, distress in nurses due to NPS). We will adjust for factors that could explain why an intervention did not succeed at the team level (e.g., staff attitudes toward the use of new interventions or treatments, as measured by the EBPAS, and the cooperation between staff members, the working conditions and characteristics of the DSCU, as measured by the CVF scale). Possible interaction effects of the intervention with time and with an NH will also be investigated. Sensitivity analyses will be used to examine differences between NHs with respect to extent of performance, considering whether degree of implementation (such as more coaching) is associated with a greater reduction of inappropriate prescribing or whether attitude serves as an effect modifier. When indicated, subgroup analyses will be used to consider NHs that implement their plans unsuccessfully and NHs that implement their plans successfully. We will also conduct a missing value analysis to evaluate whether missing data are likely to be missing at random, and we will consider replacement if appropriate. Normal probability plots and plots of standardized residuals versus predicted values will be inspected to assess whether the assumptions of normality and homogeneity of variance are met. In the event of noncompliance, data transformation will be considered. The double ratings of QoL and health status, as perceived by nursing staff and family members, will be analyzed separately. Data from the process evaluation interviews will be examined in the content analysis and any barriers and facilitators will be analyzed according to the Consolidated Framework For Implementation Research [74].