The economic evaluation was conducted from a societal perspective alongside a cluster-randomized controlled trial that compared the Namaste Care Family program with usual care over a time horizon of 12 months. The protocol was evaluated by the Medical Ethics Review Committee of the VU University Medical Center in Amsterdam, the Netherlands (protocol number: 2016.399) and is registered in the Netherlands Trial Register (http://www.trialregister.nl/trialreg/index.asp, identifier: NL5570). The committee declared that no further formal review was needed, since the study was declared exempt from the scope of the Medical Research Involving Human Subjects Act. Detailed information on the study design and the Namaste Care Family program trial is provided elsewhere . Funding sources were not involved in the design and execution of the study or writing of the study results.
Recruitment and setting
Nineteen Dutch nursing homes with a psychogeriatric ward were recruited. Dutch nursing homes offer long-term care, some in a domestic-small-scale living style environment. It includes 24-h functional support and (medical and nursing) care for people that are in need of assistance with activities of daily living.
Randomization and sample size
Participating nursing homes were matched on several characteristics that potentially affect the outcomes. These characteristics were collected by means of a questionnaire sent out to the managers of the nursing homes and included questions on the number of beds, whether the nursing home offered small-scale living arrangements for persons with dementia yes or no, whether it was situated in a rural or urban area, and the number of offered psychosocial programs such as “Snoozelen”. Furthermore, the religious orientation of the nursing home was also used as a matching criterion when it was felt it affected end-of-life decision-making care practice. Possible matches were judged by three researchers (HJAS, KJJ and JTvdS). Subsequently, matched pairs of nursing homes were randomized by an independent statistician. Due to the nature of the intervention, the group allocation could not be masked. Sample size calculations indicated that with 192 participants, power sufficed to detect a relevant difference in participants’ quality of life .
Recruitment of participants
Within each nursing home, nursing staff was asked to indicate potentially eligible residents. Residents were eligible if they had advanced dementia and were unable to participate in regular activity programs, or if they had moderate dementia with behavioral symptoms of dementia such as agitation, aberrant motor behavior, aggression or apathy. If the nursing staff was of the opinion that the resident or family caregiver could benefit from the program, for example because the resident was responsive to touch, the resident was considered eligible. Further, eligible residents had family caregivers with sufficient proficiency in Dutch who were willing and able to fill in questionnaires. Written informed consent was provided by family caregivers. No financial incentive to participate was provided.
Ultimately, nineteen nursing homes consented to participate and were allocated to the Namaste Care Family program (n = 10) or usual care (n = 9).
Namaste Care is a multidimensional care program with psychosocial, sensory and spiritual components that incorporates tailored and personalized care until death for people with advanced dementia. Experiences and outcomes of family caregivers may be improved when their involvement in the program is increased. Hence, we expanded the program by also inviting family caregivers to training sessions and involve them in Namaste Care, which is referred to as the Namaste Care Family program . Ideally, this Namaste Care Family program was given 7-days-a-week in two 2-h sessions . The sessions took place in a calm room with a ‘home-like’, relaxed setting, the so-called ‘Namaste-room’. In this ‘Namaste-room’, pleasant scents were used and nature sounds or soft music were played. Furthermore, attempts were made to avoid external distractions or interruptions. Each session started with personally welcoming each participant when entering the Namaste-room. Each participant was comfortably seated and screened for signs of pain. Appetizing, nutritious foods and drinks were offered frequently. The main aim was to establish a meaningful connection between participants and family caregivers. Extra personal care (e.g. massages, grooming, nail care and washing the face, hands and feet) was offered during the sessions to aid in an experience of a gentle, caring touch. Each session ended with nursing staff personally thanking each participant for attending the session. A more detailed description of the intervention can be found elsewhere . In the control group, participants only received usual care, which was not restricted in any way.
Data were collected between May 2016 and December 2018. Primary clinical outcomes were the Quality of Life in Late-Stage Dementia (QUALID) , the three-level version of the EuroQol (EQ-5D-3L) , and the Gain in Alzheimer Care Instrument (GAIN) , which were assessed at baseline, and 1, 3, 6 and 12 months after baseline. The QUALID is a proxy-rated quality of life instrument for people with dementia and consists of 11 items [18, 21]. Each items has 5 response options. Summed scores range from 11 to 55 with lower scores indicating better quality of life. The QUALID has good psychometric properties in people with advanced dementia [18, 22, 23]. The EQ-5D-3L contains five dimensions (mobility, self-care, activities of daily living, pain/discomfort and depression/anxiety) with 3 answer levels (no problems to severe problems). Both the EQ-5D-3L and QUALID were filled in by nursing staff. The EQ-5D-3L health states were converted to utility scores using the Dutch tariff, where 0 is anchored to death and 1 to full health (range − 0.30 to 1, where negative utilities indicate that a health state is valued as worse than death) . Using the area under the curve method, QALYs were calculated by multiplying the amount of time a participant spent in a specific health state with the utility score associated with that health state. Transitions between health states were linearly interpolated. The GAIN measures family caregivers’ gains in dementia caregiving . The scale has 10 items that are scored on a Likert scale from 0 (disagree a lot) to 4 (agree a lot). Summed scores can range from 0 to 40, with higher scores indicating higher gains. The GAIN has good psychometric properties .
Costs were measured from a societal perspective (secondary care costs, medication costs, family costs and Namaste Care Family program costs) using questionnaires primarily based on a Dutch standardized data collection tool for older nursing home residents, the TOPIC-MDS  at 1, 3, 6, and 12 months after baseline. Additional questions on the costs of the Namaste Care Family program were developed specifically for this study. Medication use was assessed using medication data from the nursing homes.
Healthcare utilization was valued using standard costs from the Dutch costing guideline . Medication costs were valued using prices from the Royal Dutch Pharmacists Association . Family costs included time spent with the participant, administrative tasks for the participant, travel time and distance to visit the participant, finding replacement for daily activities when visiting, and lost productivity due to family caregivers’ absenteeism from work. The shadow price of these time investments is assumed to be equal to the tariff for cleaning work. Lost productivity costs due to absenteeism from work were calculated using gender-specific income values of the Dutch population.
The Namaste Care Family program costs were estimated using a bottom-up micro-costing approach, which included costs of supplies for the intervention, any change (increase or decrease) in nursing staff time as well as hiring extra nursing staff, and family and volunteer time investments. Costs related to supplies and other investments, as well as actual costs of donated items, were collected by asking the participating nursing homes to estimate the monetary value of supplies and donations they received. Extra staff costs were estimated using their hourly wage. All costs were expressed in Euros for the year 2018 using consumer price indices . Discounting was not necessary, because follow-up was restricted to 12 months.
The cost-effectiveness analyses were conducted according to the intention-to-treat principle. Missing data were replaced using Multiple Imputation with Chained Equations (MICE) . Cost and effect data were assumed to be missing at random, which means that missing observations are explained by observed variables . The imputation model included outcome variables and predictor variables that either differed at baseline, were related to missing data or were associated with the outcome (see Table 2 for variables included in imputation model). To account for the skewed distribution of cost data, predictive mean matching was used in MICE . The number of imputed datasets was increased until the loss of efficiency was less than 5%, resulting in 10 imputed datasets . Each of the imputed datasets was analyzed separately as described below. Results from the multiple datasets were pooled using Rubin’s rules .
Multilevel regression models were used to estimate incremental costs and effects between the treatment groups, while accounting for the clustering of the data by allowing the intercepts to vary across clusters (i.e. random intercepts model). For costs and QALYs, a two-level structure was used where nursing homes and participants represented the first and second level, respectively. QALYs were adjusted for baseline utility. For the difference in QUALID and GAIN, an additional level accounted for repeated observations within persons (i.e. scores at different time points). The differences in QUALID and GAIN were additionally adjusted for confounders (see Table 3 for list of confounders). Incremental Cost-Effectiveness Ratios (ICERs) were calculated by dividing the incremental costs by incremental effects. Bias-corrected bootstrapping was used to estimate statistical uncertainty (2000 replications). Statistical uncertainty surrounding ICERs was illustrated by plotting the bootstrapped cost-effect pairs on a cost-effectiveness plane (CE plane). Cost-effectiveness acceptability curves (CEACs) were also estimated, which demonstrate the probability that the Namaste Care Family program is cost-effective compared to usual care for a range of different ceiling ratios (i.e. the willingness-to-pay threshold for one point effect extra) . In the Netherlands, the generally used willingness-to-pay threshold for healthcare interventions ranges between 10,000 and 80,000 € per QALY gained . For outcome measures such as the QUALID and GAIN, no willingness-to-pay thresholds have been determined. Analyses were performed in IBM SPSS Statistics 24® (IBM Corp., Armonk, NY, US), StataSE 14® (StataCorp LP, CollegeStation, TX, US) and MLwiN® (University of Bristol, Bristol, UK)  from within StataSE 14® . To check the robustness of the results, four sensitivity analyses were performed. First, the economic evaluation was performed without adjustment for confounders (SA1). Second, it was performed from the healthcare perspective (SA2), which included secondary care costs, medication costs, and Namaste Care Family program costs. Third, clustering was ignored in the estimation of incremental costs and effects (SA3). Finally, the economic evaluation was performed using observed data, i.e. missing data was handled using complete-case analysis (SA4).