Impact of Antipsychotic Review and Nonpharmacological Intervention on Antipsychotic Use, Neuropsychiatric Symptoms, and Mortality in People With Dementia Living in Nursing Homes: A Factorial Cluster-Randomized Controlled Trial by the Well-Being and Health for People With Dementia (WHELD) Program

There are 35 million people with dementia worldwide, many of whom reside in nursing homes. In the United Kingdom, one-third of people with dementia live in care homes (1), and in the United States, 64% of people receiving Medicare in nursing homes have dementia (2). The majority of these individuals have moderate or severe dementia and have highly complex care needs resulting from a combination of cognitive, functional, and communication impairments, neuropsychiatric symptoms, and medical comorbidity. Neuropsychiatric symptoms such as aggression, agitation, and psychosis affect 90% of people at some point during the course of their condition (3). They cause significant distress and can place the individual and others at risk. Furthermore, they present a substantial challenge for health and care professionals as there are limited treatment options. As a result, many people are prescribed antipsychotic medications.

There is evidence to support modest benefits of antipsychotic treatment, particularly risperidone, olanzapine, and aripiprazole, for some neuropsychiatric symptoms for the short-term management of severe aggression, but benefits with longer-term treatment are less clear (4–7). Moreover, antipsychotics are associated with severe safety concerns, including increased cognitive decline, stroke, and death, particularly when used in the long term (5, 7–9). Best practice guidance emphasizes the importance of frequent monitoring and judicious prescribing not only to reduce these risks, but also to ensure identification of cases where antipsychotic use is warranted (10).

There is a growing evidence base to support the value of person-centered care and nonpharmacological interventions for the management of neuropsychiatric symptoms in nursing homes (11–16). A recent meta-analysis particularly highlighted the benefit conferred by social interaction and pleasant activities on both neuropsychiatric symptoms and antipsychotic use (17), and a positive effect of physical activity, through personalized exercise, on mood has also been demonstrated (18). This suggests that augmentation of person-centered care with these elements would provide an effective approach to care.

Up to 2008, cohort studies and audits in the United States and Europe reported that 40% of people with dementia in nursing homes were receiving an antipsychotic (19, 20). In recent years there has been a concerted effort to reduce unnecessary prescribing of antipsychotics in people with dementia, which has led to a shift in the landscape of their use, with audits demonstrating 15%−50% reductions in prescriptions across the United States and Europe (21, 22). With this reduction in unnecessary prescriptions, use of antipsychotics has likely been focused to a greater extent on people with more severe neuropsychiatric symptoms and may have altered the benefit-to-harm ratio. While randomized discontinuation studies have reported benefits following review and withdrawal of antipsychotics (6), we know of no randomized trials evaluating the impact of rigorous antipsychotic review.

This raises key questions regarding the potential to build an effective, feasible real-world intervention to manage neuropsychiatric symptoms and antipsychotic use in the complex landscape of nursing homes. It will be important to establish whether routine implementation of antipsychotic review and evidence-based nonpharmacological interventions can contribute to improved outcomes for people with dementia. This cluster-randomized controlled trial evaluated an intervention to rigorously implement best practice guidelines for the prescribing and review of antipsychotics in people with dementia living in nursing homes alongside nonpharmacological approaches, delivered through primary care physician and nursing home education. The primary hypotheses were that antipsychotic review would reduce antipsychotic use, that social interaction would reduce agitation, and that exercise would improve depression.

Method

Study Design

The study was part of the Well-Being and Health for People with Dementia (WHELD) program. We used a cluster-randomized 2×2×2 factorial design with two replications in 16 nursing homes in South London, North London, Oxfordshire, and Buckinghamshire. The unit of randomization was the nursing home. Each home (cluster) received a randomly allocated intervention, with most homes randomly assigned to more than one of the three interventions, for 9 months (see Figure S1 in the data supplement accompanying the online version of this article). The study received ethical approval from South-Central Oxford Research Ethics Committee (REC) C (REC number 11/SC0066). The trial is registered as a clinical trial (ISRCTN number 40313497), and the protocol is available online at http://www.kcl.ac.uk/ioppn/depts/wolfson/about/people/staff/ballardclive.aspx.

Participants

The participants were people with dementia as defined by the Clinical Dementia Rating (23) (score of 1 or greater) and the Functional Assessment Staging (24) (stage 4 or greater). Nursing homes were identified from those rated “adequate” or better in the Care Quality Commission register in 2013. Eight homes were selected from a convenience sample, and another eight were randomly selected. Homes were excluded if fewer than 60% of the residents had dementia or they were in receipt of local authority special support. All residents meeting the eligibility criteria within the homes were invited to participate. Baseline and follow-up data were collected on all residents for whom consent was given and who met the inclusion criteria at each participating nursing home.

Consent for nursing home involvement was obtained from the management of the homes. If residents lacked capacity, informed consent was obtained through the involvement of a nominated or personal consultee who represented the residents’ interests and wishes in accordance with the Mental Capacity Act. Research assistants carried out baseline assessments prior to randomization.

Interventions

All 16 homes received an intervention to implement person-centered care. Eight nursing homes were randomly assigned to antipsychotic review, eight to an intervention to increase social interaction, and eight to an exercise intervention, as shown in Figure S1 in the data supplement accompanying the online version of this article. The interventions were delivered by a therapist who had attended an intensive 10-day training program and who coordinated the delivery of the intervention into all homes assigned to that intervention. In each home a minimum of two lead staff members (“champions”) were trained to implement the intervention.

Person-centered care.

The person-centered care intervention primarily used tools developed for the Focussed Intervention for Training of Staff (FITS) program, which demonstrated efficacy in a robust randomized controlled trial (12). Supplementary materials were drawn from the best available training manuals (13), and they were augmented by leadership training based on the principles identified in a systematic qualitative review of the elements of effective implementation (25) and input from an expert therapy development group. The intervention had five focal points: 1) embedding an understanding of dementia and person-centered care, 2) assessing how each home personalizes care in terms of plans and provision of opportunities for individuals’ person-centered care, 3) developing the staff’s understanding of the relationship between an individual resident’s experience, behavior, and well-being through the use of his or her life story and the principles of functional analysis to understand challenging behavior, 4) recognizing the impact of staff–resident interactions on the care experience by using cognitive-behavioral principles, and 5) implementing person-centered care planning based on these principles. This training package was delivered to all staff in the participating homes.

Antipsychotic review.

Antipsychotic review focused specifically on the review of antipsychotic prescriptions by primary care physicians or psychiatry specialists, based on the National Institute for Health and Clinical Excellence dementia guidelines (25) and facilitated by antipsychotics guidance developed by the Alzheimer’s Society in partnership with the U.K. Department of Health. The core of the guidelines and the related educational materials was consistent with best practice advice internationally, with educational materials describing the modest benefits and potential harmful effects of antipsychotic medications in people with dementia. The guidelines emphasized careful medical assessment of underlying causes of neuropsychiatric symptoms such as pain and factors leading to delirium, the use of monitoring and/or nonpharmacological interventions as a first-line approach before considering pharmacotherapy (unless symptoms were severe or causing risk to the person or others), regular review of antipsychotic prescriptions in people for whom these treatments were already prescribed, and the recommendation to constrain treatment periods with newly commenced antipsychotics to a maximum duration of 12 weeks when possible. A trial discontinuation was recommended as the preferred practice for patients who had had antipsychotic prescriptions for more than 3 months, but on the basis of evidence from a previous randomized controlled trial, caution was recommended in people with baseline Neuropsychiatric Inventory scores above 14 (26). Physicians were invited to an interactive seminar and/or practice meeting, provided with a toolkit or best practice guide, and given an opportunity for detailed discussion, including scenarios with individual patients. Seminars were conducted for care staff regarding safe antipsychotic prescribing, monitoring, and review. WHELD therapists worked with the champions and other staff to develop processes to prompt physician review according to best practice guidelines. Therapists also worked with physicians and staff to augment person-centered care during antipsychotic withdrawal. The goal was to promote informed review. Prescribing decisions were still made entirely by the participants’ own physician. In the majority of cases this was the primary care physician.

Social interaction with pleasant activities.

A manual for the social interaction intervention was developed to operationalize the way social activities were selected, with the aim of enhancing resident interactions with staff, family, and volunteers and increasing the amount of time residents spent in meaningful activity. The objectives were to provide positive planned social interactions for each resident delivered through individual or group sessions, with at least three sessions per week. The activities were based on three evidence-based approaches to promote social interaction and specific communication skills training to enhance staff–resident interactions. These were the published Positive Events Schedule (27), a social interaction intervention (28), and the Needs, Environment, Stimulation, and Techniques (NEST) program (29). Individualized care plans were developed by taking into account life history information and interests to ensure that the activities and interactions were individually tailored. The aim was for residents to have at least 1 hour a week of social interactions or to increase social interactions by 20% by the end of the study.

Exercise.

The exercise intervention aimed to promote exercise through enjoyable physical activities based on the Seattle protocols (27) and NEST manual (29). Assessment of the resident’s interests informed a personalized exercise plan developed by the therapist and champion, accounting for health and fitness levels. Walking was encouraged as a routine activity, where appropriate, alongside other individual and group activities, such as dancing, exercising to music, or chair volleyball. The aim was for residents to engage in at least 1 hour a week of exercise or to increase exercise by 20% by the end of the study.

Outcome Measures

Antipsychotic and other psychotropic drugs were classified according to the British National Formulary. Depression was evaluated by using the Cornell Scale for Depression in Dementia, an informant- and patient-report scale validated for dementia (30). Agitation was evaluated by informant interview using the 29-item Cohen-Mansfield Agitation Inventory (31). Secondary outcomes included neuropsychiatric symptoms measured through the 10 domains of the Neuropsychiatric Inventory, nursing home version (32) (which also includes a domain for depression), mortality, and dementia severity as measured by the Clinical Dementia Rating (23) and Functional Assessment Staging instrument (24).

Assessments were carried out at baseline and 9 months by research assistants blind to intervention allocation. The factorial design made it more straightforward to maintain blinding than is usually the case in trials of a nonpharmacological intervention.

Randomization

Randomization was performed as a constrained complete list randomization stratified on the four participating sites. All homes had been recruited before randomization. The constraint ensured an approximately equal distribution of the number of interventions to each geographic location. The randomization system was held at the North Wales Organisation for Randomised Trials in Health, and it has been coded and validated in R (statistical package) (33). Selection bias was reduced by inclusion of all participants identified as eligible and with consent. Homes were approached in the order of appearance on the randomized list.

Sample Size

The primary outcome measure was reduction of antipsychotics. It was estimated that about 30% of the residents with dementia living in nursing homes were receiving antipsychotics, on the basis of the best available data. We estimated that this proportion would be reduced to 10% for residents receiving antipsychotic review.

Based on PASS 11 power analysis and sample size software (NCSS, Kaysville, UT), a sample size of 96 in each group gives 82% power to detect a difference between the two proportions of −0.20. The test statistic used is the two-sided Z test to compare two independent proportions for a cluster-randomized trial where the intracluster correlation is assumed to be 0.05. The significance level of the test is 0.05. After adjustment for a dropout rate of 25%, the sample size required was 128 per arm, or about 16 participants per home. Based on the effect size (>0.50) seen in the Caring for Aged Dementia Care Resident Study (11), the study was designed to detect an effect size of 0.5 for the other outcomes. A total of 128 patients for each of the group comparisons gives 80% power to detect a treatment difference at a two-sided 0.05 significance level, if the true effect size is 0.5. Cluster randomization reduces efficiency and leads to loss of power but was essential, as the intervention has to be implemented throughout individual nursing homes. The design effect (DE), otherwise known as the inflation factor, is defined as the ratio of the total number of patients required in cluster randomization to the number required in individual randomization. Statistical theory leads to the formula DE = 1 + [(m − 1) × r1], where r1 = s2b / (s2b + s2w), called the intracluster correlation coefficient, where s2b is the between-cluster variance and s2w is the within-cluster variance. Intracluster correlation coefficients for authentic resident outcome measures (rather than process outcomes) rarely exceed 0.03. An estimated average of 16 eligible participants per cluster leads to an inflation factor of 1.45. Therefore, 186 participants were required to give this level of power for each outcome. Given the frailty of the population and the estimated mortality rate, a total sample size of 240 was stipulated to allow for mortality and dropouts. This sample size does not give power to correct for multiple testing with respect to the three primary hypotheses.

Statistical Analysis

The primary hypotheses were that in comparison to person-centered care alone, antipsychotic review would lead to a greater reduction in antipsychotic use, the social interaction intervention would improve agitation, and the exercise intervention would reduce depression at the individual resident level. Analyses accounted for intervention, baseline demographic and clinical characteristics, and site as covariates and exposure variables. All continuous outcome measures were scored by using a 20% missing rule. The analysis used multiple linear regression models for continuous outcome measures and logistic regression models for binary outcome measures. Robust standard errors were used throughout to account for clustering effects (34–36).

The primary outcome analysis was an intention-to-treat analysis. For the main analysis, age, gender, and severity of dementia were included as covariates in the modeling of the five outcome measures. Site was also included as a stratification variable. For antipsychotic drug use and Neuropsychiatric Inventory total score, the corresponding baseline measures were also covariates. For each outcome, a model was fitted that consisted of the baseline and all three interventions simultaneously to reflect the nature of a factorial design. When significant interaction effects were identified, these were included in linear models. Throughout, scores on the Functional Assessment Staging and Clinical Dementia Rating scales were modeled as linear effects, as they are naturally ordered. This reduced the degree of freedom and increased the statistical power. A p value of 0.05 was adopted. Analyses were conducted by using Stata version 13 (StataCorp, College Station, Tex).

The primary analyses were treatment as allocated for all individuals with outcome data. In the sensitivity analyses, we did a reanalysis allocating the one nursing home that withdrew and did not receive any intervention as if the participants were assigned to the group receiving person-centered care only. For the main analysis, only participants with follow-up data were included. Two further sensitivity analyses were undertaken; these imputed data for participants who had died or withdrawn by using best- and worst-case scenario assumptions for missing data. In the worst-case scenario, the total score on the Cohen-Mansfield Agitation Inventory or the Neuropsychiatric Inventory for all residents who died was imputed as the maximum score in the corresponding nursing homes. For all those lost to follow-up or those who completed the follow-up but with the corresponding outcome measures missing, the score was imputed as the mean score in the corresponding homes. For antipsychotic use, all participants missing data for the follow-up drug status were classified as taking drugs. In the best-case scenario, the Cohen-Mansfield Agitation Inventory or Neuropsychiatric Inventory total score for participants who died was imputed as the mean score in the corresponding nursing homes. For all those lost to follow-up or those who completed the follow-up but for whom the corresponding outcome measures were missing, these were imputed as the minimum score in the corresponding nursing homes. For antipsychotic use, all participants missing data for follow-up drug status were classified as not receiving antipsychotics.

Results

Cohort Characteristics

Sixteen nursing homes were recruited and randomized between August and December 2011, and 277 participants were randomly assigned to the interventions. Of these 277, 195 (70%) completed the study. One home withdrew after randomization but before commencement of the intervention. Outcome measures for 12 of the 21 participants from this home were collected at 9 months. The flow of participants through the study is summarized in Figure S2 in the online data supplement.

The participants had a mean age of 85.26 (SD=7.02), and 74% were female. According to the Clinical Dementia Rating, 12% had mild dementia, 40% moderate, and 47% severe. According to the Functional Assessment Staging scale, 11% had mild dementia, 6% moderate, 64% moderately severe, and 19% severe. Of the 277 participants, 49 (18%) were taking antipsychotics, with no significant difference between the groups that did and did not receive antipsychotic review. Baseline characteristics are described fully in Table 1. The intrahome correlation coefficient for Neuropsychiatric Inventory total score at baseline was 0.05. This is modest, suggesting that the clustering effect should not be overly influential in relation to the outcomes.

There was no significant difference in the total Neuropsychiatric Inventory score or the Cohen-Mansfield Agitation Inventory score at baseline between noncompleters and completers among either the people receiving antipsychotic review (neuropsychiatric symptoms: −1.82 difference for noncompleters, 95% CI −7.79 to 4.15, p=0.53; agitation: −3.37, 95% CI −8.54 to 1.80, p=0.18) or not receiving antipsychotic review (neuropsychiatric symptoms: +0.59, 95% CI −6.20 to 7.38, p=0.86; agitation: −6.46, 95% CI −14.24 to 1.31, p=0.10).

Primary Outcomes

Twenty of 118 (17%) people were receiving antipsychotics at baseline in the clusters receiving review. Twenty of 99 people (20%) were receiving antipsychotics at baseline in clusters not receiving antipsychotic review. Ten of the 20 (50%) people taking antipsychotics in the reviewed group stopped receiving antipsychotics before follow-up, but none of the individuals taking antipsychotics stopped treatment in the group not receiving antipsychotic review. Three people began receiving antipsychotics in each group (<4%). Overall there was a significant reduction in antipsychotic use in the review group compared with the non-review group (odds ratio 0.17, 95% CI 0.05 to 0.60, p=0.006) (Table 2). In addition, there were no prescriptions of typical antipsychotics at follow-up in the homes that received antipsychotic review. The doses used were similar at baseline and follow-up in both groups. The details of individual antipsychotic use at baseline and follow-up in the homes receiving antipsychotic review are shown in more detail in Table 3. All of the participants for whom antipsychotics were discontinued in the review group had been receiving antipsychotics for at least 3 months at baseline, and therefore all individuals were eligible for discontinuation according to the recommendations in the educational package. Additional caution was recommended for individuals with baseline Neuropsychiatric Inventory scores above 14, on the basis of evidence from a previous randomized controlled trial (26). Only three of the patients for whom antipsychotics were discontinued had scores over this threshold, although it is noteworthy that they had a mean deterioration of more than 20 points on the Neuropsychiatric Inventory after discontinuation.

Secondary Outcome: Mortality

After adjustment for baseline covariates, the proportion of people dying in the group receiving neither antipsychotic review nor the social interaction intervention was 35%. This was reduced to 28% in the group receiving antipsychotic review but not the social intervention and 19% in the group receiving both antipsychotic review and the social intervention. The adjusted main analysis for mortality showed a significant effect of the interaction between antipsychotic review and social interaction (odds ratio 2.06, 95% CI 1.06–4.01, p=0.04). In this main analysis model, which included the baseline covariates, antipsychotic review, social interaction, and exercise and the interaction between antipsychotic review and social interaction, antipsychotic review conferred a nonsignificant reduction in mortality (odds ratio 0.67, 95% CI 0.39–1.14, p=0.15). The variable predominantly affecting reduced mortality in this model was social interaction (odds ratio 0.26, 95% CI 0.13 to 0.51, p<0.001). Exercise did not significantly contribute to mortality (odds ratio 1.18, 95% CI 0.78 to 1.79, p=0.44) (Table 4). A further analysis focusing specifically on the contrast between participants receiving both antipsychotic review and social interaction and those receiving neither demonstrated that the group receiving both of these interventions had significantly reduced mortality compared with the group receiving neither (odds ratio=0.26, 95% CI 0.13 to 0.51, p<0.001).

Secondary Outcome: Neuropsychiatric Symptoms

There was a significant interaction between antipsychotic review and social interaction (score difference −7.81, 95% CI −14.74 to −0.88) and after interactions were accounted for, the group receiving antipsychotic review had a 7.37-point (95% CI 1.53 to 13.22, p=0.02) disadvantage compared with the group not receiving antipsychotic review. Importantly, the disadvantage of antipsychotic review in regard to the Neuropsychiatric Inventory score disappeared for the group receiving both antipsychotic review and the social intervention in comparison to patients not receiving antipsychotic review (−0.44, 95% CI −4.39 to 3.52, p=0.82). Exercise significantly improved neuropsychiatric symptoms (−3.59, 95% CI −7.08 to −0.09, p<0.05) but did not confer any significant benefit with respect to the score on the Cornell Scale for Depression in Dementia (mean difference −1.21, 95% CI −4.35 to 1.93, p=0.43). For the Cohen-Mansfield Agitation Inventory, there was no significant difference between the groups that did and did not receive antipsychotic review (score difference 4.60, 95% CI −1.43 to 10.63, p=0.13) (Table 4). Results of the full analysis are described in Table 4.

Sensitivity Analysis

An additional analysis was performed in which the nursing home that did not implement any of the four interventions was reallocated to the group that received person-centered care only. The effect estimates became 0.25 (95% CI 0.07 to 0.91, p=0.04) for the impact of antipsychotic review on reduction of antipsychotic drugs. The effects of the interaction between antipsychotic review and the social interaction intervention became 1.83 (95% CI 0.93 to 3.57, p=0.08) for mortality and −7.57 (95% CI −14.31 to −0.83, p=0.03) for Neuropsychiatric Inventory score.

The two additional sensitivity analyses imputing data on the basis of the best- and worst-case scenarios produced results that were consistent with the main analysis both numerically and in terms of statistical significance (see Table S2 in the online data supplement).

Discussion

This real-world intervention focused on training of primary care physicians and nursing home staff with support tools to reinforce best practice guidelines. The intervention significantly reduced antipsychotic use in people with dementia by 50%, even in a population with a baseline antipsychotic use below 20%. Descriptive data also indicated that no patients were receiving typical antipsychotics at follow-up in the group assigned to antipsychotic review. Exercise conferred significant benefits in overall neuropsychiatric symptoms but not in depression, and social interaction did not improve either agitation or total neuropsychiatric symptoms. In addition, the group receiving antipsychotic review in combination with social interaction had a significant reduction in mortality. However, compared with the group of residents not assigned to antipsychotic review, those in the antipsychotic review group experienced a significantly worse outcome on overall neuropsychiatric symptoms. Importantly, this impact was mitigated by the concurrent delivery of social interaction, and the group receiving both antipsychotic review and social interaction had no deterioration in their Neuropsychiatric Inventory score. These results strongly indicate that while substantial reductions in antipsychotics can be achieved by using this real-world approach, current best practice guidelines may not be achieving the best outcomes for people with dementia unless effective nonpharmacological interventions are implemented alongside review of antipsychotics. Importantly, combining antipsychotic review with social interaction significantly reduced antipsychotic use and mortality without a worsening of neuropsychiatric symptoms, and exercise improved neuropsychiatric symptoms.

The detrimental impact of antipsychotic review on neuropsychiatric symptoms is an important finding and can likely be explained by the changed landscape of antipsychotic prescribing that is evident in recent international studies and in this cohort (21, 22, 37, 38). The antipsychotic review intervention was based on guidance created before the substantial reductions in antipsychotic use that have occurred over the last 5 years. While this has achieved significant benefits, it has meant that the severity of neuropsychiatric symptoms in people who are now receiving antipsychotics is likely to be much higher compared with previously. The finding indicates the urgent need for revision of guidelines and, in particular, the need for a greater emphasis on providing evidence-based nonpharmacological interventions in conjunction with antipsychotic review to achieve overall benefits.

Interestingly, the negative impact of antipsychotic review on neuropsychiatric symptoms was not mitigated by person-centered care but was mitigated by the addition of social interaction. This is difficult to interpret as the social interaction intervention on its own was actually associated with a worsening of neuropsychiatric symptoms, but it is probably explained by the numerical reduction of antipsychotics in the nursing homes receiving social interaction. It is also possible that the mitigating benefits of social interaction were related to the specific use of social interaction as a therapeutic approach during antipsychotic withdrawal. The experience of therapists was that although nursing home staff were able to gather life story information, they found it difficult to develop and maintain tailored plans or interventions resulting from this information based on person-centered care alone. The person-centered care intervention was based largely on understanding its underlying principles and the life story of individuals without further specified methods of applying this to improve communication, care planning, and management of neuropsychiatric symptoms. While nursing homes receiving only the person-centered care intervention had support from the therapist to devise plans, the addition of the more structured social interaction or exercise intervention provided this framework, which appears to have made implementation more straightforward.

The approach for reviewing antipsychotics was based on primary care education and on implementing processes within the nursing homes to prompt primary care review. The advantage of this approach is that it did not require direct involvement of researchers or specialists in the review process, and it would therefore be more feasible to implement in routine practice where many individuals are not under specialist care. As a result, the decision-making process that followed antipsychotic review, including further monitoring and ongoing treatment decisions, was not made into a protocol, but was based on the clinical judgment of the physicians informed by the educational input and best practice toolkit. It is therefore possible that some of the clinical decision making was not optimal, and this could have contributed to some of the worsening of neuropsychiatric symptoms associated with antipsychotic review. In addition, as this review was completed as part of clinical practice, there was no routinely documented information regarding the reasons for changes or clinical decisions. It is of note that it was unusual for participants to be started on antipsychotic regimens whether or not the primary care physicians received the educational antipsychotic review intervention, that the antipsychotic treatment of 50% of the people on antipsychotic regimens was discontinued in the antipsychotic review treatment arm, and that no patients remained on typical antipsychotic regimens at follow-up in this group. In addition, it is of note that all patients whose antipsychotic treatment was discontinued in the antipsychotic review group had had antipsychotic prescriptions for more than 3 months at baseline and therefore met the recommended criteria for a trial discontinuation. In addition, only three of the participants whose antipsychotic treatment was discontinued had Neuropsychiatric Inventory scores above 14, although these individuals had a mean deterioration in total score of more than 20 points. These descriptive data indicate that antipsychotic review did follow the evidence-based principles outlined in the educational package. Although anecdotal, the descriptive data do further support the need for caution in discontinuing antipsychotics in people with Neuropsychiatric Inventory scores above 14.

The finding that the exercise intervention did not reduce depression differs from previous results (39). However, the exercise program did significantly improve neuropsychiatric symptoms, consistent with previous literature (40). Both nonpharmacological interventions were based on enhancing positive personalized activities and carry a common theme of person-centered care as a core part of the interventions (14, 18). Both involved just 60 minutes of activity each week, providing feasible approaches to use in practice and to avoid worsening of symptoms during antipsychotic withdrawal.

The mortality figures also have important implications, particularly since mortality risk has been a key driver in the campaign to reduce antipsychotic use (9). Although antipsychotic review alone reduced mortality by more than 30%, this became statistically significant only in combination with social interaction. The reduction in mortality with antipsychotic review alone is valuable since it goes some way to validating the arguments put forward for continuous review in practice. However, this finding indicates the importance of a multifaceted nonpharmacological approach to care within a person-centered care framework, highlighting the need for social engagement of individuals in a tailored way to improve clinical outcomes.

This study has many strengths and represents a robust evaluation of enhanced person-centered care as an intervention for nursing homes. The study also had excellent retention of surviving participants. The intervention design followed best practice guidelines for antipsychotic review and published approaches for social interaction and person-centered care with established benefits in this patient group. To our knowledge, it is the first study to robustly evaluate a practical intervention that can be easily disseminated and implemented in routine clinical practice. The study demonstrated the clear utility of this approach in reducing antipsychotics in people with dementia living in nursing homes. There were also limitations. As the antipsychotic review was based on real-life practice following an educational review, it was not a formal protocol. In addition, although the study was adequately powered to examine each of the primary outcome measures, there was inadequate power to correct for testing three separate primary outcomes, which must be considered in the interpretation of the results.

Overall, the study demonstrates the feasibility of a real-world intervention to reduce antipsychotic use in people with dementia, but it highlights an urgent need to review current best practice guidelines to ensure that review of antipsychotics is in the best interests of people with dementia. Updated guidance will also need to strongly emphasize the importance of evidence-based nonpharmacological interventions. Our study suggests that focused intervention to promote social engagement is an important component of combined interventions to enable effective antipsychotic discontinuation and clinical outcomes.