Effectiveness and Safety of Dementia Care Management in Primary Care: A Randomized Clinical Trial

Key Points

Question  What is the effect of dementia care management, a model of collaborative care, on the treatment and care of people with dementia and their caregivers in primary care?

Findings  In this randomized clinical trial of 634 people with dementia, dementia care management significantly reduced neuropsychiatric symptoms and caregiver burden and increased use of antidementia drugs compared with care as usual. Dementia care management was found to be a safe intervention.

Meaning  Dementia care management may significantly improve the outcomes of treatment and care among people with dementia and caregiver burden and should be incorporated into routine care.

Importance  Dementia care management (DCM) can increase the quality of care for people with dementia. Methodologically rigorous clinical trials on DCM are lacking.

Objective  To test the effectiveness and safety of DCM in the treatment and care of people with dementia living at home and caregiver burden (when available).

Design, Setting, and Participants  This pragmatic, general practitioner–based, cluster-randomized intervention trial compared the intervention with care as usual at baseline and at 12-month follow-up. Simple 1:1 randomization of general practices in Germany was used. Analyses were intent to treat and per protocol. In total, 6838 patients were screened for dementia (eligibility: 70 years and older and living at home) from January 1, 2012, to March 31, 2016. Overall, 1167 (17.1%) were diagnosed as having dementia, and 634 (9.3%) provided written informed consent to participate.

Interventions  Dementia care management was provided for 6 months at the homes of patients with dementia. Dementia care management is a model of collaborative care, defined as a complex intervention aiming to provide optimal treatment and care for patients with dementia and support caregivers using a computer-assisted assessment determining a personalized array of intervention modules and subsequent success monitoring. Dementia care management was targeted at the individual patient level and was conducted by 6 study nurses with dementia care–specific qualifications.

Main Outcomes and Measures  Quality of life, caregiver burden, behavioral and psychological symptoms of dementia, pharmacotherapy with antidementia drugs, and use of potentially inappropriate medication.

Results  The mean age of 634 patients was 80 years. A total of 407 patients received the intended treatment and were available for primary outcome measurement. Of these patients, 248 (60.9%) were women, and 204 (50.1%) lived alone. Dementia care management significantly decreased behavioral and psychological symptoms of dementia (b = −7.45; 95% CI, −11.08 to −3.81; P < .001) and caregiver burden (b = −0.50; 95% CI, −1.09 to 0.08; P = .045) compared with care as usual. Patients with dementia receiving DCM had an increased chance of receiving antidementia drug treatment (DCM, 114 of 291 [39.2%] vs care as usual, 31 of 116 [26.7%]) after 12 months (odds ratio, 1.97; 95% CI, 0.99 to 3.94; P = .03). Dementia care management significantly increased quality of life (b = 0.08; 95% CI, 0 to 0.17; P = .03) for patients not living alone but did not increase quality of life overall. There was no effect on potentially inappropriate medication (odds ratio, 1.86; 95% CI, 0.62 to 3.62; P = .97).

Conclusions and Relevance  Dementia care management provided by specifically trained nurses is an effective collaborative care model that improves relevant patient- and caregiver-related outcomes in dementia. Implementing DCM in different health care systems should become an active area of research.

Trial Registration  clinicaltrials.gov Identifier: NCT01401582

Dementia is a public health priority that affects 47.5 million people worldwide.¹ The rapidly growing number of people with dementia presents a challenge to the health care systems. People with dementia need comprehensive medical, nursing, psychological, and social support to delay the progression of disease and sustain autonomy and social inclusion. Primary care has been identified as the first point of contact for people with dementia and is thus a promising setting for identification, comprehensive needs assessment, and initiating dementia-specific treatment and care.² However, primary care systems worldwide are insufficiently prepared for these tasks.^3-6

Evidence-based interventions alleviate the burden of disease, as no curative treatment is currently available. Involving caregivers in intervention is important because they provide the largest proportion of care for people with dementia. The burden of informal care^7-9 is the main determinant of nursing home admissions of people with dementia,^10,11 and informal care contributes the most to total care costs.^12,13 General challenges in the management of dementia include providing antidementia drug treatment, addressing neuropsychiatric symptoms and behavioral problems, reducing inappropriate psychoactive medication use, and managing caregiver burden.¹⁴ Collaborative care programs address these challenges. There is some evidence that programs for general practitioner (GP)–based dementia care can be successfully implemented into health systems.¹⁵ However, presently the scientific evidence does not match the enthusiasm for these programs.¹⁶ There is a need to test the effectiveness of care management before implementation in primary care.¹⁷

A Cochrane review¹⁸ from 2015 analyzing 13 randomized clinical trials revealed beneficial effects of care management, specifically in reducing patients’ behavior disturbance, and caregivers’ burden and depression as well as in improving caregivers’ well-being and social support. However, there is heterogeneity in interventions, study designs, sample size, and outcomes measured. Thus, the review concluded that studies that are rigorous in design and intervention delivery are needed.¹⁸ Intervention modules and a standard set of outcome measures should furthermore be clearly defined to improve comparability.^18-20

The present randomized clinical trial describes the effectiveness of dementia care management (DCM) on relevant patient- and caregiver-oriented outcomes, including (1) quality of life, (2) caregiver burden, (3) behavioral and psychological symptoms of dementia, (4) pharmacotherapy with antidementia drugs, and (5) use of potentially inappropriate medication (PIM). Dementia care management uses a well-defined, computer-supported,²¹ and model-based intervention²² implemented by specifically trained nurses.²³

The Dementia: Life- and Person-Centered Help in Mecklenburg-Western Pomerania (DelpHi) trial was a pragmatic, GP-based, cluster-randomized intervention study with 2 arms, an intervention group and a care as usual (CAU) group. The study protocol was approved by the Ethical Committee of the Chamber of Physicians of Mecklenburg-Western Pomerania, Germany (registry number BB 20/11). The reporting of this study follows the CONSORT statement²⁴ and its extensions regarding cluster-randomized,²⁵ pragmatic trials²⁶ with nonpharmacologic treatments.²⁷ The design, eligibility and inclusion criteria, intervention, and baseline characteristics of the trial have been described in detail elsewhere.^21-23,28 The full trial protocol is available in Supplement 1.

General practices were the unit of randomization and determined the patients’ group status. A total of 854 GPs in 5 municipalities of Mecklenburg-Western Pomerania were invited to participate by mail. General practitioners expressing an interest in the study were visited by the investigators to convey additional detailed information about the study. Finally, 136 GPs (15.9%) gave written informed consent to participate and agreed to adhere to the DelpHi trial protocol. There were no restrictions regarding the GPs’ treatment of patients.

General practitioners assessed the eligibility of patients (≥70 years, living at home) and systematically screened patients who met the inclusion criteria using the DemTect procedure.²⁹ This interview-based instrument is widely used for dementia screening in GP practices in Germany and is more sensitive than the Mini-Mental State Examination for detecting milder forms of cognitive impairment (DemTect, 98% vs Mini-Mental State Examination, 46%).^30,31 Thus, it is possible that some patients screened positive for dementia by the DemTect procedure are not considered cognitively impaired according to the Mini-Mental State Examination (score 27 to 30 of 30).

Study enrollment started January 1, 2012, and ended December 31, 2014. The follow-up period ended on March 31, 2016. Patients who screened positive for dementia were informed about the study by their GP, invited to participate, and asked to provide written informed consent. If the patients listed a caregiver, he or she was asked to participate as well. When patients were unable to provide written informed consent, their legal representative was asked to sign the consent form on their behalf. General practitioners received allowances for screening (€10 [US $11.15] per patient) and study enrollment (€100 [US $111.54] per patient).

Dementia care management aims to provide optimal care by integrating multiprofessional and multimodal strategies for improving patient- and caregiver-related outcomes within the framework of the established health care and social service system. It was developed according to current guidelines,^32,33 targeted at the individual participant level, and delivered at patients’ homes by 6 nurses with dementia-specific qualifications supported by a computer-based intervention-management system (IMS) to improve systematic identification of patients’ and caregivers’ unmet needs. The nurses conducted an in-depth assessment. Based on these data, the IMS generated an individual preliminary intervention task list, and the nurses discussed and finalized the task list in a weekly interdisciplinary case conference with a nursing scientist, a neurologist/psychiatrist, a psychologist, and a pharmacist. Afterwards, the list of intervention tasks was summarized in a semistandardized GP information letter. This letter was then discussed between the GP and nurse to establish an individual treatment plan. During the first 6 months of the intervention period, the nurse conducted 6 home visits with an average duration of 1 hour, carrying out his or her standard intervention tasks in close cooperation with the caregiver, the GP, and health care and social service professionals. During the subsequent 6 months, the study nurse monitored the completion of all intervention tasks. In line with the Pacala scale³⁴ for intensive case managements, each study nurse delivered intervention to, on average, 60 patients with dementia. Training, intervention, and the IMS are described in more detail elsewhere^28,35 and in the eAppendix in Supplement 2.

The primary outcomes were assessed within identical, standardized, computer-assisted face-to-face interviews at the patients’ homes by specifically qualified nurses over an average of 3 separate visits at baseline and 12 months after baseline and pertain to the individual patients’ (1) quality of life, measured by the Quality of Life in Alzheimer Disease instrument,³⁶ which assesses physical health, mental health, and social and financial domains; (2) caregiver burden, measured by the Berlin Inventory of Caregivers’ Burden with Dementia Patients, an inventory with 88 items in 20 different dimensions, which assesses subjective and objective burden³⁷; (3) behavioral and psychological symptoms, measured by the Neuropsychiatric Inventory, an interview by proxy on 12 dimensions of neuropsychiatric behaviors³⁸; (4) use of pharmacotherapy with antidementia drugs, which included the following substances recommended by relevant guidelines: donepezil, galantamine, rivastigmine, and memantine; and (5) use of PIM, which is defined as a drug for which the risk of an adverse drug effect outweighs the clinical benefit and evaluated using the PRISCUS criteria.^39-41 The predefined secondary outcomes included cognitive status, measured using the Mini-Mental State Examination⁴²; activities of daily living, measured using the Bayer-Activities of Daily Living Scale; and health care resource use, especially institutionalization.^43,44

No previous data on the main outcome measures were available for sample size calculation. Therefore, sample size was estimated based on theoretical assumptions. In the design, the minimally important difference for determining the effectiveness was considered to be of at least a small effect, defined by Cohen d (Cohen d = 0.2).⁴⁵ Comparing 2 groups at a significance level of α = .05, assuming a statistical power of 80% and an intraclass correlation with clustering by GP practice of 0, a sample size of 310 persons per group would have been sufficient.⁴⁵ Considering the longitudinal design, we accounted for a loss over time of 35% (eg, death or withdrawal of informed consent) and determined that 477 persons per group with complete data sets would have been needed to be included in the study. We estimated that GPs would identify 1000 patients over the course of 2 years. Recruitment turned out to be slower than expected. Thus, recruitment was prolonged from 2 to 3 years. The achieved sample size allows to detect a medium effect size (Cohen d = 0.5).⁴⁵

We used simple 1:1 randomization without stratification or matching. This procedure was sufficient because of the high number of expected clusters in our study.⁴⁶ General practitioners were not informed of their randomization status. However, because of the type of intervention, GPs became aware of their status throughout the course of the study. Patients were recruited and enrolled by participating GPs but allocated to the study group by study center. Because baseline assessment, primary outcome assessment, and delivery of intervention needed to be performed by the same nurses, blinding was not possible.

To describe the sample, metric variables were summarized as means and SD, and nominal variables were presented as proportions. Baseline and follow-up values were compared using paired t tests or McNemar tests, as applicable.

The primary analyses (intention to treat [ITT] and per protocol) included generalized regression models, with a model specification corresponding to the scale level of the outcome variable. The ITT analysis was performed as modified ITT for all cases with valid baseline data,⁴⁷ and the per-protocol analysis included complete cases only. Missing data in the follow-up variables were imputed by multiple imputations via chained equations. The outcome variable at follow-up was the dependent variable, and study group was the predictor of interest. To account for the stochastic dependency of patients treated by the same GP, GPs were included as random effects. The baseline value of the outcome variable was included as a covariate to reduce residual variance and to account for interindividual variance at baseline. Furthermore, age, sex, and living situation (alone vs not alone) were included as covariates. A positive intervention effect was defined as a significant regression coefficient of the study group variable. Sensitivity analyses were performed by introducing random slopes for the difference of DCM vs CAU and the baseline variable of the outcome and recalculating the P values and 95% CIs by bootstrapping (2000 replications). All P values for the primary analyses are 1-sided. Data analysis and management were conducted using Stata version 13.1 (StataCorp). Details of the statistical analyses are provided in Supplement 1.

Because the intervention targeted the patients’ entire social system, exploratory prior analyses, prespecified in the analysis plan, were conducted by stratifying the models by patients’ living situation, identifying whether the intervention would show stronger effects in persons living alone or not living alone.

The CONSORT statement is illustrated in the Figure. Overall, 634 patients provided written informed consent, and a total of 407 (64.2%) received the intended treatment (DCM, 291 [71.5%]; CAU, 116 [28.5%]). Whereas all 407 participants were included in the per-protocol analyses, in the ITT analyses, all patients with valid baseline variables were included (Table 1). In total, 227 patients were lost to follow-up. Most of the patients dropped out before starting the baseline assessment at home (118 of 634 [18.6%]), which took place on average 138 days after initial screening by the GP because of the study procedure. The dropout rate between completion of baseline and follow-up was lower (94 of 516 [18.2%]) and more frequent in the control group (DCM, 46 of 348 [13.2%] vs CAU, 48 of 168 [28.6%]). There were no statistical differences between patients assessed at follow-up (n = 407) and those who dropped out before follow-up (n = 227) in age, sex, and DemTect score (eTable 1 and eTable 2 in Supplement 2). The intervention was safe, as no dropout was reported because of GPs’ advice or problems with the intervention reported by the patients with dementia or the caregiver. There was no significant effect of the study group on mortality.

Participant characteristics at baseline and follow-up are summarized in Table 2. Primary outcome measures for baseline and follow-up are given by group in Table 3. The groups did not differ significantly according to primary outcomes and sociodemographic variables (eTable 3 in Supplement 2) in the ITT analyses data set. In the per-protocol analyses set, the CAU group reported a significantly higher quality of life.

In the primary ITT analyses, a significant decrease in patients’ behavioral and psychological symptoms of dementia (b = −7.45; 95% CI, −11.08 to −3.81; P < .001) and caregiver burden (b = −0.50; 95% CI, −1.09 to 0.08; P = .05) was observed in the intervention group compared with CAU. Patients with dementia receiving DCM had an increased chance of receiving antidementia drug treatment (DCM, 114 of 291 [39.2%] vs CAU, 31 of 116 [26.7%]) after 12 months (odds ratio, 1.97; 95% CI, 0.99 to 3.94; P = .03). There was no effect on quality of life (b = 0.02; 95% CI, −0.09 to 0.05; P = .26) or on PIMs (DCM, 77 [26.5%] vs CAU, 19 [16.4%]; odds ratio, 1.86; 95% CI, 0.62 to 3.62; P = .97) after 12 months. The secondary analyses indicated a significant effect on quality of life in the intervention group (b = 0.08; 95% CI, 0 to 0.17; P = .03) for patients not living alone (Table 4).

According to secondary outcomes, we found no significant effect on patient’s cognitive status, daily living activities, or institutionalization. Overall, 24 of 407 patients (5.9%) were institutionalized 1 year after baseline (DCM, 16 [5.5%] vs CAU, 8 [6.9%]).

Sensitivity analyses confirmed the results of the ITT analyses. As expected, the clinical characteristics showed serious clustering, and sociodemographic variables, such as sex, age, and living status, were not GP-dependent. The per-protocol analyses, sensitivity analyses, and the intraclass correlations for the main outcomes are reported in eTable 4 in Supplement 2.

In our study, DCM was beneficial for optimizing treatment and care in patients with dementia. We found medium to large effects of DCM for community-dwelling patients with dementia in primary care on behavioral and psychological symptoms, caregiver burden, and pharmacologic treatment with antidementia drugs. Referring to neuropsychiatric symptoms measured by the Neuropsychiatric Inventory, a decrease in 4 points would be regarded as clinically meaningful.³⁸ In our analysis, DCM reduced neuropsychiatric symptoms by 8 points, with a larger effect size compared with previous studies included in the Cochrane review by Reilly et al⁴⁸ (standardized mean difference, −0.20; 95% CI of difference, −0.41 to 0.01; n = 368; I² = 83%; P = .06). Referring to caregiver burden, the effect size of the DCM was medium but larger when compared with other studies (−0.18 vs −0.07).⁴⁸ Thus, our results indicate meaningful clinical relevance. The study methods were in line with the demand to use standardized sets of outcome measures²⁰ and well-defined interventions¹⁹ to improve comparability across studies, and our results contribute empirical evidence to currently inconclusive research¹⁸ on DCM approaches in primary care.

The results suggest that DCM increased the quality of dementia care. Improvements included a higher use of antidementia drugs. Although this is a simple proxy for good medical dementia care, the data do not indicate whether drug treatment conformed to guidelines. To our knowledge, there is no benchmark for the percentage of people who should be treated with antidementia drugs in primary care that we could have used for comparison. Our data suggested the use of a careful approach with antidementia drug treatment, with a prevalence of 39% in the intervention group. This proportion is comparable with other studies.^49,50 Use of antidementia drugs is recommended in many international guidelines.⁵¹ However, assessing adherence to guidelines in their full complexity could not be done in this study because of limited sample size.

Neuropsychiatric symptoms and caregiver burden are among the most important risk factors for institutionalization of people with dementia.^10,11 To our knowledge, our study is the first to show positive effects on both factors.^52,53 Another risk factor is functional inabilities.¹⁰ Because the use of antidementia drugs can have a positive impact on functional abilities,³² all 3 effects of DCM are likely synergistical to delay institutionalization. This could save long-term costs.

A small effect on quality of life was restricted to patients not living alone. This result should not be overestimated because validity and reliability of quality of life measures in people with dementia are limited. However, this finding implies that further analyses could identify target groups with an increased benefit. We speculate that the effectiveness of DCM could be associated with socioeconomic status, functional ability, or severity of dementia.

There was no effect of DCM on the frequency of PIMs. This is unexpected because comprehensive medication management was part of the intervention. We speculate that the intensity was probably too low in this trial because recommendations to the GP regarding pharmacologic treatment were provided only once. An effective reduction of PIMs may require a higher intensity of care management and follow-up reviews.⁵⁴

Screening and recruitment were part of routine care so that selection bias cannot be ruled out. Any systematic control mechanisms would have interfered with GPs’ routine, causing adverse effects, including dropout of GPs. However, all participating GPs agreed to recruit systematically while adhering to the requirements of the study design.

The number of participants was imbalanced between the intervention and control groups. Fewer GPs were randomized to the control group. Furthermore, there was the tendency that GPs in the control group included less patients. We expect that during the trial, GPs noticed their assignment, which resulted in a loss of motivation for an inclusion of further patients. However, there were no significant group differences according to primary outcomes or sociodemographic variables.

The DelpHi trial was not a diagnostic trial. The identification of patients with dementia was based on a screening instrument. A state-of-the-art diagnostic procedure was not required. However, the DemTect was designed for this specific purpose and is widely used in routine care.^30,55,56 Furthermore, the main analysis was ITT, and any false positives would have caused an underestimation of the intervention effect.

The study was incorporated into routine care as closely as possible so that the external validity of the results is high. However, because of the rigorous design in the context of this trial, there were restrictions in time, length, and content of DCM activities. In routine care, nurses have more freedom to decide what, when, and how activities are performed. Additionally, generalizability might be limited because of the region and health care system being studied. It is possible that differences in access and availability of health care resources in other health care systems may affect the effectiveness. However, challenges of dementia care are mainly triggered by the disease itself and require similar resources that are available in different regions and health systems.

Dementia care management provided by specially trained nurses and supported by a computer-based IMS is an effective and safe collaborative care model that has clinically relevant patient- and caregiver-related effects on treatment and care. Therefore, implementation in routine care could be beneficial for people with dementia and their relative caregivers. Further analyses should identify specific subgroups of people with dementia with higher effectiveness of DCM and should evaluate cost-effectiveness to adapt DCM to other settings and health care systems.

Corresponding Author: Jochen René Thyrian, PhD, German Center for Neurodegenerative Diseases (DZNE), site Rostock/Greifswald, Ellernholzstr. 1/2, 17489 Greifswald, Germany (rene.thyrian@dzne.de).

Accepted for Publication: June 1, 2017.

Published Online: July 26, 2017. doi:10.1001/jamapsychiatry.2017.2124

Author Contributions: Drs Hoffmann and Thyrian had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Thyrian, Dreier-Wolfgramm, Teipel, Hoffmann.

Acquisition, analysis, or interpretation of data: Thyrian, Hertel, Wucherer, Eichler, Michalowsky, Dreier-Wolfgramm, Zwingmann, Kilimann, Hoffmann.

Drafting of the manuscript: Thyrian, Hertel, Michalowsky, Zwingmann, Teipel.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Thyrian, Hertel, Teipel.

Obtained funding: Hoffmann.

Administrative, technical, or material support: Thyrian, Eichler, Zwingmann, Kilimann.

Study supervision: Thyrian, Wucherer, Dreier-Wolfgramm, Eichler, Teipel, Hoffmann.

Conflict of Interest Disclosures: None reported.

Funding/Support: The study was performed in cooperation with and funded by the German Center of Neurodegenerative Diseases and the University Medicine of Greifswald.

Role of the Funder/Sponsor: Both funders participated in and funded the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review and approval of the manuscript; and decision to submit the manuscript for publication.

Additional Contributions: We acknowledge Ines Abraham, RN; Ulrike Kempe, RN; Sabine Schmidt, RN; Vaska Böhmann, RN; Kathleen Dittmer, RN; and Saskia Moll, RN (Greifswald Medical School, University of Greifswald, Greifswald, Germany); for data collection and intervention delivery; Daniel Fredrich, Dipl-Inf (Greifswald Medical School, University of Greifswald, Greifswald, Germany), and Henriette Rau, MSc (German Center for Neurodegenerative Diseases [DZNE], Rostock/Greifswald, Germany), for information technology development and support in conducting the trial; Kerstin Albuerne, MedDok, and Andrea Pooch, BSc (DZNE, Rostock/ Greifswald, Germany), for data collection, data quality assurance, and data provision; and Viktoria Kim-Böse, MSc, and Kerstin Wernecke, PhD (DZNE Rostock/ Greifswald, Germany), for writing and editing assistance as well as administrative assistance in conducting the trial. All persons mentioned were compensated for their contributions as part of their employment.