The DAVOS study has been approved by the Ethics Committee of the Goethe University
of Frankfurt am Main, Germany (reference 129/18 and conforms to the Declaration of
Helsinki (Version Fortaleza 2012). Trial registration is DRKS00015686 (www.drks.de).
DAVOS is a controlled cluster-randomized trial that uses the stepped-wedge design.
This type of waiting control group design employs repeated assessments, whereby each
of the clusters (= nursing home) of the mentioned cluster group begins in the control
phase. Each cluster group in DAVOS includes residents from 3-4 different nursing homes.
After a certain delay, the intervention is gradually introduced in each cluster group
(see Figure 1). During the waiting control phases, patients receive “usual care”. The point at
which clusters pass from the control to the intervention phase is fixed. As soon as
a new cluster group passes into the intervention phase, relevant data is collected
from all clusters. In this way, data is gathered throughout the duration of the intervention
(T1-T4), irrespective of whether the intervention has already been introduced or not.
The T4 assessment can be classified as post-data collection. Furthermore, baseline
(T0) and follow-up (T5) assessments are conducted four months before and four months
after the intervention, respectively. Thus, depending on time point of recruitment
and inclusion in to the study, residents will be repeatedly measured up to five times.
Overall, 10 nursing homes (about 1,250 care places) are included and randomized to
three cluster groups. We aim to recruit a study population of N = 380. As a result
of newly admitted nursing home residents, the recruitment process will be continuous
(open cohort), enabling the size of the study population included at baseline to be
kept more or less constant throughout the study. In this way, the drop-outs that are
a natural part of the process, perhaps due to severe illness or death, can be compensated
for, and the number of assessments in each survey period can kept almost constant.
The stepped-wedge design allows comparisons to be made within and between the clusters,
receiving and not receiving the intervention. Furthermore, insights can be obtained
into whether and from when the complex intervention is effective (the separate modules
of the intervention have already been demonstrated to be effective in previous randomized
The nursing homes are the units of randomization (unrestricted randomization). We
randomly assigned the 10 nursing homes (cluster) to the three cluster groups with
computer-generated random numbers. We used online software to generate a randomization
using a list randomizer (http://www.random.org). Finally, each cluster group includes residents from 3-4 different nursing homes.
There are certain time points at which the cluster groups pass from the control to
the intervention phase. All nursing homes and the included residents are informed
about these time points.
Setting and study population
The recruitment of participants and implementation of the intervention will take place
at 10 outpatient nursing home facilities with a total of more than 1,250 care places.
This will be possible as a result of the cooperation with the two social services
organizations (German: Frankfurter Verband für Alten- und Behindertenhilfe e.V.; Agaplesion Markus Diakonie
gGmbH). Based on nursing home records (e.g., on known diagnoses and health status), the
nursing home residents that are, in principle, suitable for participation in the study,
will be approached by the previously appointed case managers. At this point, residents
with a clinical diagnosis of dementia (ca. 50%), are unable to provide consent, or
who have a known alcohol or substance related disorder will be excluded from participation.
It can be assumed from previous research that about one third of the overall population
(ca. 36%) will be eligible for participation in the study and will provide their consent
(Figure 2). The data of residents that agree to participate in the study will be collected
at baseline and checked against the inclusion and exclusion criteria.
Residents over 60 years of age and without obvious signs of dementia, an addictive
disorder, or another severe mental illness will be included in the study. Inclusion
criteria for participation in the interventional modules as described below are the
presence of a subsyndromal depressive disorder or clinical depression. It makes sense
to include subsyndromal symptomatic depression because it is common in old age and
defines an important target population for the use of secondary preventive measures.
In any case diagnosis of subsyndromal or clinically manifest depression is based on
ICD 10 criteria following the judgement of a psychotherapist who is licenced by a
state board (in Germany: Approbierter Psychotherapeut). The establishment of a proper clinical diagnosis is already part of the intervention
as described below.
The case management program as well as the interventional modules are shown in Figure 3. The depression case managers play an important integrative role in the intervention,
as they are at the interface between residents, nurses, physicians, and psychotherapists.
The case managers are nurses selected by the management of the respective nursing
homes, and who have been trained for this role before the intervention begins. In
every participating nursing home two depression case managers are selected. Altogether
20 depression case managers are trained during the study. The case manager’s tasks
include the identification of suitable study participants, prompt presentation during
the psychotherapeutic consultation hour (see below) in the case of positive screening
of depression, and coordinate the treatment modules for the participants.
The trainings for case managers will include the following four elements: 1) communication
of basic medical-psychological information on late-life depression, 2) use of the
screening instrument, 3) information on how to deal with residents with depression
and 4) the organization of project-related requirements.
Case managers will be supervised throughout the study.
The intervention is initiated by a screening applied to the participating residents
using a modified version of the Depression Monitoring List (DeMoL) with integrated
PHQ-D assessment . The screening is performed by the depression case manager or
by other members of the nursing staff under the supervision of the case manager. In
case of positive screening the participant is referred to a psychotherapeutic consultation hour (in German: Psychotherapeutische Sprechstunde) in accordance with §92, paragraph 6a, German Social Code (in German: Sozialgesetzbuch) during which a board licenced psychological psychotherapists will provide a diagnostic
assessment (according to ICD 10 criteria). As part of DAVOS, the psychotherapeutic
consultation hour will be implemented as an “in house” service in the nursing home
which is an innovative approach compared to the usual practice in the German health
care system. The assessment in the psychotherapeutic consultation hour will conclude
with recommendations for several interventions that are elaborated in accordance with
the German S3 guideline, and the National Disease Management Guideline on Unipolar
Depression , and are part of three interventional modules. Ranging from “watchful
waiting”, participation in basic intervention (module 1) and a recommendation for
psychotherapy (module 3) to the involvement of the general practitioner or a specialist
physician, e.g. psychiatrist (module 2), the measures will cover a wide spectrum of
Module 1 (= basic intervention) consists of participation in group sessions that are offered
to all participants with and without any symptoms of depression (including persons
suffering from subsyndromal depressive disorders). Key components of this module are
supportive and psychoeducative approaches (on the subjects of “successful ageing”,
“mindfulness” etc.). By preparing a weekly plan, for example, the residents are to
be animated to participate in measures involving physical activity, as well as other
social and leisure pursuits. The case managers initially assist to carry out this
module but later take on full responsibility for it. The aim is to establish this
basic intervention as part of the daily nursing routine and to encourage everyday
companions and other nurses to use it after the intervention is over. The case managers
are thus to adopt the role of multipliers.
Module 2 contains aspects of treatment that require the therapeutic involvement of the general
practitioner in charge of the resident and/or a specialist physicians (such as exclusion
or treatment of somatic causes of depression, drug therapy / antidepressants, interactions
with other drugs, polypharmacy, hospital admissions etc.). The role of the case manager
here is to prompt and coordinate appointments with the doctors in charge following
recommendations derived from the psychotherapeutic consultation hour.
Module 3 covers participation in psychotherapeutic groups and, where applicable, individual
psychotherapy conducted by psychologists. The employed interventions include elements
of cognitive behavioral therapy (e.g., planning pleasant daily activities, problem
solving, mindfulness-based meditation and cognitive restructuring [6, 11–15]). Individual
psychotherapy sessions will be provided if required, to residents with major depression,
dysthymia, and adjustment disorders. Psychotherapy will partly be delivered by psychologists
in clinical training from the outpatient clinic of the Department of Clinical Psychology
and Psychotherapy at Goethe University Frankfurt. The psychologists will receive additional
training in CBT for late-life depression  and mindfulness-based cognitive therapy
The psychologists’ tasks are: 1) participation in regular meetings at the nursing
home facilities with case managers to ensure exchange of information about cases and
treatments, 2) conducting psychotherapeutic consultations in order to motivate the
patient for psychotherapy and select a suitable setting, 3) conducting either individual
or group CBT including elements of mindfulness-based meditation [6, 11–15]).
After two months, individuals who meet the inclusion criteria for the intervention
but do not wish to participate will be contacted again and informed about the provided
Questionnaires and psychometric instruments that have been validated in clinical and
gerontology research will be used to collect data face-to-face. For this purpose,
instruments have been selected that are well-established and time-efficient, but that
simultaneously cover a wide range of outcome-relevant variables. Data collection will
be mainly quantitative but also supplemented qualitative methods (e.g., interviews
with case managers, focus groups). These qualitative data will be analyzed by the
sequential analysis in dependence on Rosenthal  and the Grounded Theory .
Subsequently, paper-pencil data will be digitalized, checked and subjected to missing
values analysis. Questionnaires with more than 30% of values missing will not be included
in the subsequent analysis. An estimate of isolated missing values will be made, using,
for example, the FIML (full information maximum likelihood algorithm).
The raters responsible for data collection are part of the study team but not involved
in the intervention (e.g. psychotherapy). They will receive intensive training in
using the deployed instruments (test methods, questionnaires). The exclusive use of
standardized instruments will ensure the influence of individual raters is negligible
and that the data is valid. During the trainings, the trustworthiness of the data
collection process will also be controlled by calculating interrater reliability.
The prevalence of depressive disorders and the severity of depression symptoms (or
any change in them) among nursing home residents are the primary outcomes of DAVOS.
Secondary outcomes are quality of life, functional status (instrumental activities
of daily living), social participation and the type, frequency and duration of any
hospitalization during the observation period. In addition to collecting data on primary
and secondary outcomes (T0 to T5), relevant personal and sociodemographic data will
be collected at baseline (including family background, socioeconomic status, educational
level, subjective health status , personality characteristics , cognitive
status , current medication and somatic comorbidities). To minimize the stress
of data assessment nursing home documentation can serve as additional data source.
Some of the variables, such as health status and cognitive status, will be measured
repeatedly at the T5 follow-up assessment. All instruments are well introduced, validated
and have frequently been used in previous research. An overview of the instruments
and associated validation references are given in Table 1.
The primary outcomes (prevalence of depressive disorders and severity of depression
symptoms) will be assessed using the structured clinical interview for DSM-IV (SCID-I)
 and the Geriatric Depression Scale (GDS) . SCID-I is a commonly used means
of recording and diagnosing selected mental syndromes and disorders. The 15 items
on the GDS are part of the geriatric assessment for the evaluation of depressive disorders
in old age.
Baseline assessments are conducted simultaneously in all the clusters to determine
prevalence rates of depression at differing levels of severity. We expect a significant
decrease of prevalence rates of depression as well as severity of depression symptoms
in the clusters that receive interventions, as compared to clusters receiving care
as usual (stepped-wedge approach). The screenings conducted by the case managers that
are part of the intervention are independent of the outcome assessments and cannot
therefore result in an increase in prevalence.
Established instruments and scales which are developed especially for old adults will
be used to measure secondary outcomes such as quality of life, functional competence
and social participation. Examples of these are the Philadelphia Geriatric Center
Morale Scale (PGCMS) , the WHO Quality of Life (short form) (WHOQOL-OLD) [25,
26], a questionnaire for the assessment of aspects of quality of life in later life,
the short form of the Late Life Function and Disability Instrument (SF-LLFDI) ,
and the Social and Emotional Loneliness Scale (short form) . Information on the
frequency and duration of hospitalization will be taken from nursing home records.
Data analysis and data quality assurance
The evaluation will be conducted using a stepped-wedge design with open cohorts .
That means that not only at the beginning but also during the course of the trial,
new participants will be recruited into each of the 10 clusters. Each new recruit
will generally undergo all the data assessments (periods) from his inclusion into
the trial until she/he either leaves the nursing home or the project ends. The percentage
(point prevalence) of participants with depression will be assessed at the end of
each period, with some of the trial participants being assessed for the first time
and others not. Assuming the new care model to be effective, the prevalence of depression
is expected to decline significantly. The use of a stepped-wedge design ensures that
results do not reflect a general period effect, meaning that they do not result from
a change in prevalence in all the clusters, independently of whether a cluster is
in the control or intervention phase.
A hierarchical generalized linear model with a logit link function will be used for
the analysis and will account for not only fixed treatment and period effects but
also random effects on both cluster and patient level. In order to analyze the second
primary endpoint (depression severity), we will fit a hierarchical logistic regression
model for ordinal target variables (cumulative odds model). Details will be laid down
in a statistical analysis plan. The statistical models mentioned above are suitable
for use with correlated categorical data for the incomplete courses that the planned
design leads us to expect. Bias can be avoided or at least limited through the use
of randomization and the standardization of data collection methods.
Based on a sample size of 38 participants to be assessed in each cluster (nursing
home) in each period, cluster groups of three, four and three (10 nursing homes),
a type 1 error probability of 0.05 in a two-tailed test, a prevalence (including subsyndromal
symptomatic depression) of 33% under control conditions  and 22% under intervention
conditions, and assuming a between cluster standard deviation of prevalences of 4%,
and assuming independence between repeated observations on the same patients within
clusters across periods, the power of the study will be 0.80. We assumed a time constant
random cluster effect, which induces a correlation structure with an intraclass coefficient
(ICC) of 0.0072 under the null hypothesis and of 0.0092 under alternative hypothesis
within intervention periods, both within periods and between periods. Note, that this
assumption also implies a perfect autocorrelation of random cluster effects across
time. For power calculation, we used the R package swCRTdesign . In fact, patient
level random effects will introduce dependencies within clusters across periods. This
induces a reduction in effective sample size for repeated observation under the same
treatment and increased effective sample size for repeated observation under different
treatments. We expect that the latter effect will dominate because the majority of
patients will be observed under both treatment conditions. At the planning stage of
the project, we did not perform further sensitivity analyses to account for variability
in cluster size, for attrition, and for more realistic correlation structures. For
attrition, we expect a limited impact, because new patients will be recruited while
the study is ongoing accounting for eventual drop out and patients dying.