Cost-effectiveness of the FindMyApps eHealth intervention vs. digital care as usual: results from a randomised controlled trial

Abstract

Objectives

Despite growing interest, the cost-effectiveness of eHealth interventions for supporting quality of life of people with dementia and their caregivers remains unclear. This study evaluated the cost-effectiveness of the FindMyApps intervention, compared to digital care-as-usual. FindMyApps aims to help people with dementia and their caregivers find and learn to use tablet apps that may support social participation and self-management of people with dementia and sense of competence of caregivers.

Method

A randomised controlled trial (Netherlands Trial Register NL8157) was conducted, including people with mild cognitive impairment (MCI) or mild dementia and their informal caregivers (FindMyApps n = 76, digital care-as-usual n = 74). Outcomes for people with MCI/dementia were Quality-Adjusted Life-Years (QALYs), calculated from EQ-5D-5L data and the Dutch tariff for utility scores, social participation (Maastricht Social Participation Profile) and quality of life (Adult Social Care Outcomes Toolkit), and for caregivers, QALYs and sense of competence (Short Sense of Competence Questionnaire). Societal costs were calculated using data collected with the RUD-lite instrument and the Dutch costing guideline. Multiple imputation was employed to fill in missing cost and effect data. Bootstrapped multilevel models were used to estimate incremental total societal costs and incremental effects between groups which were then used to calculate Incremental Cost-Effectiveness Ratios (ICERs). Cost-effectiveness acceptability curves were estimated.

Results

In the FindMyApps group, caregiver SSCQ scores were significantly higher compared to care-as-usual, n = 150, mean difference = 0.75, 95% CI [0.14, 1.38]. Other outcomes did not significantly differ between groups. Total societal costs for people with dementia were not significantly different, n = 150, mean difference = €-774, 95%CI [–2.643, .,079]. Total societal costs for caregivers were significantly lower in the FindMyApps group compared to care-as-usual, n = 150, mean difference = € −392, 95% CI [−1.254, −26], largely due to lower supportive care costs, mean difference = €−252, 95% CI [−1.009, 42]. For all outcomes, the probability that FindMyApps was cost-effective at a willingness-to-pay threshold of €0 per point of improvement was 0.72 for people with dementia and 0.93 for caregivers.

Conclusion

FindMyApps is a cost-effective intervention for supporting caregivers’ sense of competence. Further implementation of FindMyApps is warranted.

Keywords:

• Dementia
• quality of life
• psychosocial interventions
• eHealth
• randomised controlled trial
• cost-analysis
• cost-effectiveness

Introduction

The number of people living with dementia globally is expected to rise from 57.4 million in 2019 to 152.8 million in 2050 (GBD 2019 Dementia Forecasting Collaborators, 2022). People living with dementia experience cognitive and functional impairments and express unmet needs, including a lack of meaningful daytime activities and social contact (van der Roest et al., 2009). In Europe, the estimated cost of dementia care was €400 billion in 2019, which is expected to increase with the rising prevalence (World Health Organization, 2021). Care provided by informal caregivers (family members and friends) accounts for 40–90% of the economic burden of dementia (Meijer et al., 2022; Nay et al., 2015). Informal caregivers often experience high stress and health problems (Gauthier et al., 2022; Ma et al., 2018; Nay et al., 2015), and adverse outcomes for caregivers are associated with adverse outcomes for people with dementia. Therefore, cost-effective psychosocial interventions are needed to support physical, mental and social health of people with dementia and caregivers, to maintain quality of life (Dröes et al., 2017; Moniz-Cook et al., 2011; WHO, 2021).

Available cost-effectiveness evaluations of psychosocial interventions in dementia have in many cases demonstrated modest benefits at increased cost (Henderson et al., 2021; Huo et al., 2021; Mountain et al., 2022; Nickel et al., 2018). The Journeying Through Dementia intervention in the UK had no effect on quality of life, but a small benefit on self-reported flourishing, despite incremental costs of £608 (€713) over usual care (Mountain et al., 2022). An international study evaluating the Dutch Meeting Centre model for people with dementia found a 50% probability of cost-effectiveness at a willingness-to-pay (WTP) of €5,000 per point of increase in dementia-specific quality of life, but the probability of cost-effectiveness for quality-adjusted life years (QALY) was 0 at willingness-to-pay thresholds up to €350,000 (Henderson et al., 2021).

Low-cost eHealth interventions, such as smartphone or tablet apps and other interventions incorporating computer software, are considered promising psychosocial interventions for meeting care needs, improving well-being and reducing burden among people with dementia and caregivers. However, few digital psychosocial interventions have been evaluated in randomised controlled trials (RCTs) (Ghani et al., 2022; Knapp et al., 2022; van Santen et al., 2022) and limitations of the few existing studies include incomplete cost information or lack of outcome data. Thus, there remains little evidence as to whether digital interventions are cost-effective for supporting quality of life or social health in dementia. FindMyApps is a digital intervention targeted at community-dwelling people with mild cognitive impairment (MCI) or mild dementia, and their informal caregiver (Kerkhof et al., 2019). The goal of FindMyApps is to support self-management and social participation for people with MCI/mild dementia and improve caregivers’ sense of competence to provide care, by helping people with dementia learn to use a tablet computer, and find easy-to-use apps that meet their personal needs and interests. This paper evaluates the cost-effectiveness of FindMyApps compared to digital care-as-usual alongside a randomised controlled trial, from a societal perspective, with respect to quality of life of people with dementia and caregivers, social participation of people with MCI/mild dementia and sense of competence of their informal caregivers.

Materials and methods

Design

The economic evaluation was conducted alongside a two-arm, non-blinded randomised controlled trial, comparing the FindMyApps intervention with digital care-as-usual. The RCT protocol has been described in full elsewhere (Neal et al., 2021). Participants were randomly allocated to one of the two trial arms using block randomization, with target ratio 1:1, stratified by prior tablet use (yes/no), diagnosis (MCI/mild dementia) and whether the person with dementia and their caregiver cohabited (yes/no). Outcomes relating to people with MCI/dementia and their caregiver were measured at baseline and at three-month follow-up. This economic evaluation is reported according to the Consolidated Health Economic Evaluating Reporting Standards statement (CHEERS) (Husereau et al., 2022).

Ethics and consent

The trial was registered in the Netherlands Trial Register (NL8157) and was approved by the Medical Ethics Committee of VUmc (2019.605). The study was declared exempt from the Medical Research Involving Human Subjects Act. All participants provided verbal and written informed consent to participate before commencing, and consent was verbally reaffirmed during every interaction.

Sample and setting

Community-dwelling people with an established diagnosis of MCI or mild dementia (MMSE 18-25 or Brief Cognitive Rating Scale 17–32), who had an internet connection and contact with an informal caregiver at least twice per week were eligible to participate. Exclusion criteria were blindness or severe visual impairment, a diagnosis of primary progressive aphasia, insufficient proficiency in the Dutch language to provide informed consent, or concurrent participation in another intervention study. Participants were recruited from all regions of the Netherlands, between January 2020 and July 2022, via the Amsterdam and Groningen Alzheimer Centres, the Centre of Geriatrics Amsterdam at Amsterdam University Medical Centre, and national networks of meeting centres and case managers. Participants also self-referred via the project website.

Intervention

The FindMyApps intervention comprises a tablet, training in tablet use, and the FindMyApps app, which allows users to search for and download apps from a database of apps that have been assessed as easy to use by people with dementia, which match their personal user-profile, and may contribute to their social health (Neal et al., 2021). By making it easy to find apps that are relevant and easy to use, FindMyApps is expected to support people with MCI/mild dementia to independently manage daily life and have sufficient social contact, and therefore reduce caregiver burden (increase sense of competence to provide care). The FindMyApps intervention was implemented as described in the study protocol (Neal et al., 2021), based on Small-scale pilot studies which found FindMyApps to be feasible to implement and positively experienced (Beentjes et al., 2020; 2023; Kerkhof et al., 2022). A large RCT and process evaluation comparing FindMyApps to digital care-as-usual found that people receiving FindMyApps downloaded more apps, reported the tablet as easier to learn to) use (Neal et al. (2023a) and that informal caregivers experienced a significantly greater sense of competence with FindMyApps (Neal et al., 2023b). Participants used the FindMyApps app installed on their own tablet, or on a tablet (Apple iPad or Lenovo Table 3, Table 4 or Table M10) borrowed from Amsterdam UMC for the research. Caregivers each received a single 30-minute training session by video call on the use of the tablet and how best to support their counterpart with dementia during the study, to (learn to) use the tablet and search for, download and use apps. People with dementia received a 30-minute training session by video call on the use of the tablet and the FindMyApps app, with the caregiver also present. Participants received paper and digital copies of a handbook, containing information about the tablet and the FindMyApps app, and had access to training films that also covered the same information. Participating dyads were advised to practise together with the tablet at least twice a week for the first four weeks.

As described in the study protocol (Neal et al., 2021), participants randomised to receive digital care-as-usual were not provided the FindMyApps app. They used their own tablet or a tablet of the same models from Amsterdam UMC, and they received training sessions, handouts and access to training films, but these materials did not cover the use of the FindMyApps app. The handouts instead included ia list of websites, on which they could find apps recommended for people with dementia. They were given the same advice regarding practice with the tablet. For any questions or problems, participants in both arms were able to contact a help desk run by investigators, by email or telephone, at any time during the study.

Effect measures

The primary outcome measure for this cost-effectiveness evaluation was health-related quality of life, assessed by the EQ-5D-5L for both persons with dementia as well as for caregivers. The EQ-5D-5L contains five dimensions of quality of life (mobility, self-care, activities of daily living, pain/discomfort and depression/anxiety) with five response levels (no problems to extreme problems) (EuroQol Research Foundation, 2019). The combination of all five dimensions presents a total of 3125 different health states. The Dutch tariff was used to convert participant’s EQ-5D-5L health states to utility scores (range: −0.446 to 1) (Versteegh et al., 2016). To calculate QALYs, utility scores were multiplied by the amount of time a participant spent in a specific health state (area under the curve method). Transitions between health states were linearly interpolated.

For people with MCI/mild dementia, aspects of quality of life broader than health alone were measured using the Adult Social Care Outcomes Toolkit questionnaire (ASCOT) (van Leeuwen et al., 2015a). The ASCOT is a 9-item scale with four response levels of need (ideal state to high-level need). The range of outcome values (-0.17 to 1) represents the sum of weighted scores per item, assessing the participant’s quality of life. The ASCOT has been shown to be a reliable and valid measure (van Leeuwen et al., 2015b). Social participation was measured using the Maastricht Social Participation Profile (MSPP) (Mars et al., 2009). The MSPP is a 26-item instrument containing four indices: consumptive participation, formal social participation, informal social participation-acquaintances and informal social participation-family. Diversity scores and frequency scores for each index (the total and mean score of the items in each index, respectively) can be calculated. Total diversity scores were calculated by summing the diversity scores (number of items in the index on which a respondent scored at least one) per section. The total frequency score was calculated as the mean score of the items in the indices. For informal caregivers, the Short Sense of Competence Questionnaire (SSCQ) was used to assess caregivers’ sense of competence. This 7-item scale of five response levels (strongly agree to strongly disagree) has been demonstrated to have strong reliability and validity (Vernooij-Dassen et al., 1999). The total SSCQ score (0-7) was the sum of items with which participants either disagreed or strongly disagreed.

Costs

Healthcare utilisation data were collected at baseline (as a potential source of confounding), and at one, two and three months follow-up from a societal perspective using the RUD-Lite instrument (Dutch version) (Wimo et al., 2013), for people with dementia and caregivers separately. Healthcare utilisation was valued using standard costs from the Dutch costing guideline (Kanters et al., 2017) and grouped into primary care (general practitioner, physiotherapist, occupational therapy and manual therapy visits), secondary care (hospitalisation, intensive care admission, emergency, psychiatric and outpatient care), supportive care (district nurse, home care, day care centre), informal care and other costs (meals and transportation). Further details are provided in Appendix B (Costing Table). The incremental intervention costs of FindMyApps were calculated using a bottom-up micro-costing approach, comprising prices of fixed expenses needed to provide and maintain the intervention (i.e. FindMyApps app database hosting, website hosting, app updates, analytics analysis, certification, compliance, insurance, marketing, sales) and customer support costs. We estimated that the target population in the Netherlands, of community dwelling people with diagnosed dementia (Vektis Factsheet dem, 2022), who have access to the internet at home (Centraal Bureau voor Statistiek, 2020) is 110,560 people (see Appendix A). From trial data, the average rate of uptake of the intervention was 55%. Scaled up, this would result in an expected number of 60,808 users of FindMyApps. To estimate costs for one user of the FindMyApps intervention, the fixed costs of FindMyApps were divided by 60,808 users and added to the estimated cost of customer support, per user. This resulted in intervention costs of €18 per dyad. Details are provided in Appendix A. Costs of informal care, based on the total time that caregivers spent on taking care of their counterpart with dementia and performing household tasks, were added to health care costs to estimate costs from the societal perspective. Lost productivity costs were not assessed, because it was assumed that the majority of persons with dementia and many caregivers had already reached the age of retirement. All costs were expressed in Euros for the year 2021. Prices were converted to 2021 using consumer price indices if necessary. Discounting was not necessary as the follow-up period was shorter than one year.

Statistical analysis

Baseline differences between the FindMyApps and digital care-as-usual groups were tested using t-tests for continuous variables and Chi-square tests for categorical variables. The main analysis was performed from the societal perspective according to the intention-to-treat principle. Missing observations were imputed using Multiple Imputation with Chained Equations (MICE), a technique which preserves statistical power whilst minimising the risk of introducing bias (van Buuren & Groothuis-Oudshoorn, 2011). It was assumed that data was missing at random. During the imputation procedure, predictive mean matching (PMM) was used to account for the skewed distribution of cost data. The number of imputed datasets was increased until there was a loss of efficiency less than 5%, resulting in 10 imputed datasets. Imputed datasets were analysed separately, after which results were pooled using Rubin’s rules. For estimating the differences in costs and effects between the intervention and control group, bivariate regression was used while adjusting for education, which was identified as a possible confounder during the main effect analyses (Neal et al., 2023b). Incremental cost-effectiveness ratios (ICERs) were calculated by dividing the difference in total costs between groups by the difference in effects. Bias-corrected bootstrapping with 5000 replications was used to estimate 95% confidence intervals around the mean cost and effect differences. Decision uncertainty surrounding ICERs was shown by plotting the bootstrapped cost-effect pairs on a cost-effectiveness plane. Cost-effectiveness acceptability curves were also plotted, to demonstrate the probability of the intervention being cost-effective based on different WTP thresholds (range of willingness-to-pay thresholds for one unit of effect gained). The WTP thresholds for EQ-5D-5L and ASCOT were set based on the generally accepted thresholds for QALYs, ranging from €0 through €20,000 up to €80,000 per QALY gained. For secondary health outcomes (MSPP, SSCQ) there are no WTP thresholds available and we therefore used arbitrarily chosen WTP thresholds of €0, €750 and €3000 per point of improvement. All analyses were conducted using R (version 4.2.2).

Sample description

In total, from 1st January 2020—30th June 2022, 76 dyads were randomised to the FindMyApps group and 74 dyads to digital care-as-usual. No statistically significant differences were observed between groups at baseline (see Table 1).

Table 1. Descriptive statistics of demographic and experimental outcomes of the FindMyApps group and digital care-as-usual group at baseline. Note: N, count; SD, standard deviation.

	FindMyApps	Digital care-as-usual	Difference test	p-value
People with MCI/dementia	N = 76	N = 74
Age (years), mean (SD)	73.2 (9.48)	72.42 (8.77)	t = 0.522	0.602
Gender, N (% female)	34 (44.74%)	29 (39.19%)	Chi2 = 0.273	0.601
Educational level			Chi2 = 3.298	0.192
Low, N (%)	10 (13.16%)	8 (10.81%)
Middle, N (%)	23 (30.26%)	33 (44.59%)
High, N (%)	43 (56.58%)	33 (44.59%)
EQ-5D-5L, mean (SD)	0.91 (0.08)	0.91 (0.08)	t = −0.348	0.728
ASCOT, mean (SD)	0.82 (0.13)	0.82 (0.13)	t = −0.048	0.962
MSPP total diversity score, mean (SD)	12.1 (4.68)	12.34 (4.72)	t = −0.274	0.784
MSPP total frequency score, mean (SD)	2.46 (1.24)	2.35 (1.17)	t = 0.487	0.627
Caregivers	N = 76	N = 74
Age (years), mean (SD)	65.37 (11.45)	62.5 (14.31)	t = 1.354	0.178
Gender (% female)	54 (71.05%)	56 (75.68%)	Chi2 = 0.207	0.649
Educational level			Chi2 = 2.07	0.355
Low, N (%)	7 (9.21%)	3 (4.05%)
Middle, N (%)	31 (40.79%)	36 (48.65%)
High, N (%)	38 (50.99%)	35 (47.30%)
EQ-5D-5L, mean (SD)	0.93 (0.06)	0.92 (0.08)	t = 1.270	0.206
SSCQ, mean (SD)	4.95 (1.92)	5.05 (1.71)	t= −0.291	0.771

Complete follow-up data for people with MCI/dementia across all variables were available for 65% (n = 97/15i0). At the follow-up of 3 months, 16% of EQ-5D-5L data was missing (n = 24, 13 in intervention group and 11 in control group), and 19% of ASCOT and MSPP scores (n = 29, 15 in intervention group and 14 in control group). For total societal costs, 23% of the data was missing (n = 35, 20 in the intervention group and 15 in the control group). For caregivers, complete follow-up data across all variables were available for 77% (n = 115/150). At the follow-up of 3 months, 16% of EQ-5D-5L data and SSCQ scores were missing (n = 24, 13 in intervention group and 11 in control group). For total societal costs 23% of data were missing (n = 35, 20 in intervention group and 15 in control group).

Sensitivity analysis

To assess the robustness of the results, five sensitivity analyses (SA) were performed. In SA1, regression models were not adjusted for confounders. SA2 was performed from the healthcare perspective meaning that only healthcare costs were included. In SA3, costs and QALYs for people living with dementia and their caregivers were summed. In SA4 (dropout analysis), a complete case analysis was performed and in SA5 outliers for total costs (observations >1.5 times the interquartile range from the median) were excluded.

Results

Effects

There was a statistically significant difference in SSCQ scores between caregivers in the FindMyApps group and the digital care-as-usual group (0.75, 95%CI: 0.14 to 1.38, Table 2). Differences in QALYs (people with dementia and caregivers), ASCOT and MSPP scores (people with dementia) were not statistically significant (Table 2). No harms were identified.

Table 2. Multiply imputed effects and costs for the FindMyApps group and digital care-as-usual group after 3 months of follow-up. Note: MSPP-d: MSPP total diversity score; MSPP-f: MSPP total frequency score; SE: standard error; 95% CI: 95% confidence interval. All costs are presented in €.

	FindMyApps		Digital care-as-usual
	Mean (SE)	Range (min; max)	Mean (SE)	Range (min; max)	Mean difference (95% CI)
People with MCI/dementia
EQ-5D-5L	0.91 (0.003)	(0.62; 1.00)	0.91 (0.003)	(0.66; 1.00)	0 (−0.025; 0.023)
ASCOT	0.82 (0.01)	(0.39; 1.00)	0.82 (0.01)	(0.46; 1.00)	−0.003 (−0.051; 0.044)
MSPP-d.	13.52 (0.17)	(4; 25)	12.99 (0.19)	(4; 26)	0.528 (−1.026; 2.058)
MSPP-f.	2.64 (0.04)	(0.55; 5.22)	2.54 (0.05)	(0.6; 6.35)	0.109 (−0.279; 0.488)
Costs
Primary care	276 (14)	(0; 2.623)	280 (16)	(0; 3.024)	−4 (−147; 119)
Secondary care	250 (20)	(0; 3.997)	166 (14)	(0; 2.965)	84 (-46; 262)
Supportive care	2,047 (119)	(0; 17,387)	2.605 (125)	(0; 16,472)	−558 (−1,619; 530)
Informal care	2,166 (111)	(0; 14,942)	2.413 (130)	(0; 16,527)	−247 (−1,328; 787)
Other costs	115 (9)	(0; 1.514)	182 (13)	(0; 2,046)	−68 (−173; 20)
Total health care costs	2592 (126)	(18; 17,405)	3.051 (132)	(0; 16,705)	−459 (−1585; 687)
Total societal costs	4.873 (200)	(18; 22,178)	5.647 (225)	(0; 28,757)	−774 (−2643; 079)
Caregivers
EQ-5D-5L	0.9 (0.001)	(0.8; 0.95)	0.92 (0.002)	(0.63; 0.96)	0.013 (−0.002; 0.032)
SSCQ	5.16 (0.06)	(0; 7)	4.41 (0.08)	(0; 7)	0.754 (0.139; 1.382)
Costs
Primary care	116 (5)	(0; 661)	190 (12)	(0; 1.548)	−74 (−170; −6)
Secondary care	98 (6)	(0; 908)	184 (12)	(0; 2.106)	−86 (−190; −15)
Supportive care	195 (27)	(0; 4.478)	447 (65)	(0; 16,330)	−252 −1.009; 42)
Other costs	28 (4)	(0; 700)	26 (6)	(0; 1.280)	2 (−60; 31)
Total health care costs	427 (30)	(18; 4.914)	820 (74)	(0; 17,597)	−393 (−1.179; −45)
Total societal costs	455 (32)	(18; 5.256)	846 (79)	(0; 18,877)	−392 (−1.254; −26)

Costs

Although none of the observed cost differences between groups among people living with dementia were statistically significant, total societal costs were somewhat lower in the FindMyApps group (€–774, 95%CI: −2643 to 1079, Table 2) with supportive care (€-558 (–1.619; 530)) and informal care costs (€–247 (–1.328; 787)) being the main contributors to this difference. Mean total societal costs for caregivers were significantly lower in the FindMyApps group compared to the digital care-as-usual group (€–392, 95%CI: −1.254 to −26, Table 2), largely due to lower supportive care costs (€–252, 95%CI: −1009 to 42, Table 2).

Cost-effectiveness analyses

Tables 3 and 4 and Figure 1 show the outcomes of the cost-effectiveness analyses. The ICER per QALY gained among people with dementia from a societal perspective was €541,503 with 42% of the bootstrapped cost-effect pairs in the southwest quadrant (FindMyApps was less expensive and less effective compared with digital care-as-usual). At WTP thresholds of €0, 20,000 and 80,000 per QALY gained, the probability of cost-effectiveness was 0.72, 0.71 and 0.67, respectively. The ICER per QALY gained among caregivers was estimated at €-115,649 with 89% of the bootstrapped cost-effect pairs in the southeast quadrant (FindMyApps was more effective and less expensive compared with digital care-as-usual, i.e. dominant over digital care-as-usual). The probability that the FindMyApps program was cost-effective compared to digital care-as-usual at WTP thresholds of €0, €20,000 and €80,000 per QALY gained for caregivers were 0.93, 0.95 and 0.98, respectively. A similar trend could be observed for the ASCOT with an ICER of €–69,256 per one point of improvement and 45% of the bootstrapped pairs in the southeast quadrant (pCE: 0.72, 0.73 and 0.68 at WTP thresholds of €0, €20,000 and €80,000, respectively) and MSPP-diversity with an ICER of €-818 per point gained and 56% bootstrapped pairs in the southeast quadrant (pCE: 0.72, 0.82, 0.83 at WTP thresholds of €0, €750 and €3000, respectively). In both cases, FindMyApps was associated with lower costs, while being more effective (i.e. dominant) than digital care-as-usual. For MSPP-frequency, the ICER was €84,588 per point gained (35% bootstrapped cost-effect pairs located in the southeast and 36% in the southwest quadrant) and the probabilities that the FindMyApps program is cost-effective compared to digital care-as-usual were 0.72, 0.71 and 0.65 at WTP thresholds of €0, €750 and €3000, respectively. The ICER per point gained on the SSCQ was €-505 per point of improvement with 94% of the bootstrapped pairs in the southeast quadrant (FindMyApps was dominant over digital care-as-usual). At WTP thresholds of €0, €750 and €3,000 per point of improvement on the SSCQ, the probabilities that the FindMyApps program is cost-effective compared to digital care-as-usual were 0.93, 1 and 1, respectively.

Figure 1. Interventions. A) Cost-effectiveness plane for difference in QALY - People living with dementia. B) Cost-effectiveness acceptability curve for difference in QALY - People living with dementia. C) Cost-effectiveness plane for difference in ASCOT. D) Cost-effectiveness acceptability curve for difference in ASCOT. E) Cost-effectiveness plane for difference in MSPP – diversity. F) Cost-effectiveness acceptability curve for difference in MSPP – diversity. G) Cost-effectiveness plane for difference in MSPP – frequency. H) Cost-effectiveness acceptability curve for difference in MSPP – frequency. I) Cost-effectiveness plane for difference in QALY – Caregivers. J) Cost-effectiveness acceptability curve for difference in QALY – Caregivers. K) Cost-effectiveness plane for difference in SSCQ. L) Cost-effectiveness acceptability curve for difference in SSCQ

Table 3. Results of the cost-effectiveness analysis, cost-utility analysis and sensitivity analyses with respect to primary outcomes (QALYs). Note: N: number of observations in the analysis; ΔC: difference in costs; ΔE: difference in effects; 95% CI: 95% confidence interval; ICER: Incremental cost-effectiveness ration; NE: northeast; SE, southwest; SW: southwest; NW: northwest; pCE: probability that the intervention is cost-effective as compared to usual care; QALY: quality-adjusted life-year;*adjusted for education. All costs are presented in €.

Outcome	N	ΔC (95% CI)	ΔE (95% CI)	ICER	NE	SE	SW	NW	pCE	pCE	pCE
									0	20,000	80,000
Main analysis*
People with MCI/dementia
QALY	150	−542 (−2.386; 1.301)	−0.001 (–0.007; 0.005)	541,503	7	29	42	22	0.72	0.71	0.67
Caregivers
QALY	150	−401 (−946; 143)	0.003 (–0.001; 0.008)	−115,649	5	89	5	1	0.93	0.95	0.98
SA1: Unadjusted analysis
People with MCI/dementia
QALY	150	−774 (–2.683; 1.136)	−0.001 (-0.006; 0.005)	1,278,509	5	36	42	16	0.79	0.78	0.75
Caregivers
QALY	150	−392 (–928; 145)	0.003 (−0.001; 0.008)	−116,072	5	89	6	1	0.92	0.95	0.97
SA2: Healthcare perspective*
People with MCI/dementia
QALY	150	−459 (–1.627; 708)	−0.001 (–0.006; 0.005)	759,043	5	37	42	17	0.78	0.77	0.72
Caregivers
QALY	150	−402 (–912; 107)	0.003 (–0.001; 0.008)	−115,961	4	91	5	0	0.94	0.96	0.98
SA3: Costs merged together*
QALY	150	−1353 (–3.630; 924)	0.002 (–0.005; 0.009)	−728,592	7	62	25	6	0.88	0.88	0.88
SA4: Complete-case analysis*
People with MCI/dementia
QALY	114	301 (−2.380; 2.983)	−0.001 (–0.009; 0.007)	−432,448	20	22	18	39	0.41	0.41	0.41
Caregivers
QALY	115	−77 (–422; 268)	0.004 (–0.002; 0.009)	−19,767	28	64	4	4	0.67	0.78	0.88
SA5: Excluding outliers*
People with MCI/dementia
QALY	136	−1.258 (–2.956; 440)	−0.001 (–0.007; 0.005)	1470,365	1	37	55	6	0.93	0.92	0.89
Caregivers
QALY	140	−1.524 (−2.649; −400)	0.003 (–0.002; 0.007)	−586,584	0	88	12	0	1	1	1

Table 4. Results of the cost-effectiveness analysis, cost-utility analysis and sensitivity analyses with respect to secondary outcome. Note: N: number of observations in the analysis; ΔC: difference in costs; ΔE: difference in effects; 95% CI, 95% confidence interval; ICER: Incremental cost-effectiveness ratio; NE: northeast; SE: southwest; SW: southwest; NW, northwest; pCE: probability that the intervention is cost-effective as compared to usual care at willingness to pay thresholds listed in bld (€per point of improvement); QALY, quality-adjusted life-year; MSPP-d: MSPP total diversity score; MSPP-f: MSPP total frequency score;*adjusted for education. All costs are presented in €.

Outcome	N	ΔC (95% CI)	ΔE (95% CI)	ICER	NE	SE	SW	NW	pCE	pCE	pCE
Main analysis: Societal perspective*
People with MCI/dementia
									0	20,000	80,000
ASCOT	150	−542 (−2.392; 1307)	0.008 (−0.046; 0.062)	−69,256	18	45	26	11	0.72	0.73	0.68
									0	750	3,000
MSPP-d.	150	−542 (−2.383; 1.298)	0.663 (−0.963; 2.289)	−818	23	56	15	6	0.72	0.82	0.83
MSPP-f.	150	−542 (−2396; 1.311)	−0.006 (−0.441; 0.429)	84,588	13	35	36	15	0.72	0.71	0.65
Caregivers
SSCQ	150	−401 (−940; 138)	0.795 (0.145; 1.446)	−505	5	94	0	0	0.93	1	1
SA1: Unadjusted analysis
People with MCI/dementia
									0	20,000	80,000
ASCOT	150	−774 (−2.676; 1.129)	0.006 (−0.048; 0.059)	−139,445	12	47	32	9	0.79	0.78	0.69
									0	750	3,000
MSPP-d.	150	−774 (−2.680; 1.132)	0.681 (−0.959; 2.322)	−1,135	17	63	16	4	0.79	0.87	0.84
MSPP-f.	150	−774 (−2.668; 1121)	0.003 (−0.426; 0.432)	−288,932	10	40	39	11	0.79	0.78	0.74
Caregivers
SSCQ	150	−392 (−927; 143)	0.792 (0.141; 1.442)	−495	5	94	0	0	0.92	1	1
SA2: Healthcare perspective*
People with MCI/dementia
									0	20,000	80,000
ASCOT	150	−459 (−1.624; 705)	0.006 (−0.048; 0.059)	−82,788	12	47	31	9	0.78	0.75	0.65
									0	750	3,000
MSPP-d.	150	−459 (−1.624; 705)	0.681 (−0.959; 2.322)	−674	18	62	16	4	0.78	0.87	0.84
MSPP-f.	150	−459 (-−1.623; 705)	0.003 (−0.425; 0.43)	−171,537	11	39	39	11	0.78	0.77	0.77
Caregivers
SSCQ	150	−402 (−911; 107)	0.795 (0.146; 1.445)	−506	4	95	0	0	0.94	1	1
SA4: Complete-case analysis*
People with MCI/dementia
									0	20,000	80,000
ASCOT	108	187 (−2.540; 2,914)	0.009 (−0.045; 0.062)	21,750	36	27	17	20	0.45	0.50	0.58
									0	750	3,000
MSPP-d.	103	959 (−1.804; 3,721)	0.127 (−1.841; 2.094)	7,575	41	14	11	35	0.25	0.3	0.43
MSPP-f.	103	959 (−1.804; 3,721)	−0.104 (−0.596; 0.387)	−9,198	25	9	16	51	0.25	0.24	0.22
Caregivers
SSCQ	115	−77 (−422; 268)	0.885 (0.218; 1.552)	−87	32	67	0	0	0.67	0.99	1
SA5: Excluding outliers*
People with MCI/dementia
									0	20,000	80,000
ASCOT	136	−1,258 (−2.963; 448)	0.009 (−0.048; 0.066)	−140,489	4	57	36	3	0.92	0.91	0.78
									0	750	3,000
MSPP-d.	136	−1,258 (−2.964; 449)	0.804 (−0.898; 2.505)	−1.565	6	77	16	1	0.93	0.96	0.91
MSPP-f.	136	−1,258 (−2.957; 441)	0.033 (−0.409; 0.474)	−38,259	4	52	41	3	0.92	0.91	0.86
Caregivers
SSCQ	140	−1,510 (−2/.655; −366)	0.757 (0.101; 1.413)	−1,995	0	99	1	0	1	1	1

Sensitivity analysis

Results of the unadjusted analysis (SA1) and cost-effectiveness analyses from a healthcare perspective (SA2) were comparable to the main analysis from the societal perspective (Tables 3 and 4). When costs and QALYs for persons with dementia and their caregivers were summed (SA3) the ICER per QALY gained was €-728,592 with 62% of the bootstrapped cost-effect pairs in the southeast quadrant (FindMyApps was dominant over digital care-as-usual). The complete-case analysis (SA4) resulted in a reversed mean difference in societal costs among people living with dementia (higher costs in the FindMyApps group). Effect differences were comparable to the main analysis resulting in lower probabilities of cost-effectiveness for all four outcomes among people living with dementia. Among caregivers, the cost differences were smaller compared to the main analysis, but still in favour of the FindMyApps group, resulting in lower ICERs for both QALY and SSCQ points gained. When excluding outliers in total costs, larger differences in costs between groups iwere found compared to the main analysis for both people with dementia and caregivers, resulting in higher probabilities of cost-effectiveness at all relevant WTP thresholds.

Discussion

This study evaluated the cost-effectiveness of FindMyApps compared with digital care-as-usual in people with MCI/mild dementia and their caregivers. In people with dementia, we found no statistically significant differences between groups in health outcomes or costs and a probability of cost-effectiveness of 0.72 at a WTP threshold of €0 per point of improvement on all outcomes. In caregivers, we observed significantly higher total SSCQ scores in the FindMyApps group, while societal costs were lower. Given the 0.93 probability of the intervention being cost-effective at a WTP threshold of €0 per additional unit of effect on the SSCQ, we can conclude that FindMyApps is cost-effective compared to digital care-as-usual for supporting caregiver sense of competence. For the other caregiver outcome (QALY), no significant difference was found between FindMyApps and digital care-as-usual.

The absence of significant differences between groups in QALYs based on the EQ-5D-5L for people with MCI/dementia, could be because FindMyApps impacts aspects of quality of life not measured by this instrument. This is in line with the results of evaluations of other psychosocial interventions in dementia. For example, the combined Meeting Centres Support programme for people with dementia and their informal caregivers was found to improve dementia-specific quality of life measured with the Quality of Life in Alzheimer’s Disease and Dementia Quality of Life (DQoL) instruments, but not health-related quality of life measured by the EQ-5D-5L (Henderson et al., 2021). Limitations of the EQ-5D-5L in the context of dementia may also partly explain a lack of sensitivity of the instrument, as previous studies have questioned the validity of self-reported EQ-5D-5L measurements among people with dementia and showed important ceiling effects (Hounsome et al., 2011). However, limitations of the EQ-5D-5L cannot fully explain these results, as we also found no significant effect of FindMyApps on ASCOT scores between groups, nor on DQoL scores (Neal et al., 2023b). Another reason for the lack of effect of the intervention on quality of life might be that piarticipants (both people with MCI/dementia and caregivers) benefited from the active digital care-as-usual control condition. Many previous economic evaluations of digital psychosocial interventions lacked a control condition altogether (Knapp et al., 2022). With respect to caregivers’quality of life, one of the few digital interventions shown to be effective in an RCT (a blended care self-management programme for caregivers) was compared to a waiting-list rather than active control (Boots et al., 2018). However, the possibility that the apps used by participants during this study simply did not contribute significantly to the users’ quality of life should also be considered.

The significant effect on sense of competence is likely because caregivers found FindMyApps easier to teach to their partner than the regular apps, thereby reducing caregiver burden (Neal et al., 2023a). This finding is important because higher sense of competence is associated with better health outcomes for caregivers (van der Lee et al., 2014), and interventions which improve caregiver sense of competence have been associated with lower societal costs and delayed nursing home admission (Henderson et al., 2021).

Incremental costs of FindMyApps compared to a regular tablet seem to be low compared to incremental costs of other digital interventions for people with dementia and their caregivers. Incremental costs for eHealth interventions have reportedly been as high as 48 US dollars (€45) per dyad per week (Knapp et al., 2022), compared to the estimated €0.35 per dyad per week for FindMyApps. Even factoring in the cost of the tablets used by participants in both arms of this study (€160 to €260 per dyad), FindMyApps would be less expensive than many psychosocial interventions (Huo et al., 2021; Nickel et al., 2018).

Viery few studies have evaluated the cost-effectiveness of digital psychosocial interventions in dementia care (Knapp et al., 2022). An evaluation of the SMART4MD intervention found similar effect sizes for costs and quality of life outcomes among caregivers of people with MCI compared to our evaluation of FindMyApps (Ghani et al., 2022). However, for people with MCI/dementia we additionally found a trend towards lower societal costs which was not found for SMART4MD. FindMyApps was also more likely to be cost-effective than a recently evaluated exergaming intervention (van Santen et al., 2022). Overall, compared to previous studies of digital psychosocial interventions in dementia, we present somewhat stronger evidence of cost-effectiveness for dyads of people with MCI/dementia and caregivers.

A strength of the current economic evaluation is the adoption of the societal perspective, meaning that all relevant costs for decision-making by policymakers were included. Several sensitivity analyses were performed, with results in line with the main analysis results, showing the robustness of our results. By choosing broad, function-based inclusion criteria and a community-setting, the trial had a pragmatic design, meaning that the results are more likely to be generalizable to actual practice than a study conducted under more tightly controlled conditions.

The study also has a number of limitations. First, although the study focussed on MCI and mild dementia, the time-horizon of the economic evaluation was still rather short (i.e. 3 months) relative to the time-course over which dementia progresses. Second, a considerable amount of data was missing (16–23% per outcome variable). We tried to mitigate potential bias due to selective dropout through the use of multiple imputation to impute missing observations, which is considered the most appropriate method to deal with missing data in economic evaluations (Faria et al., 2014). Finally, the generalisability of the findings may have been negatively impacted by selection bias because participants could self-refer to the study, and because data were collected during the COVID-19 pandemic. Measures to restrict the spread of COVID-19 also impacted the delivery of the intervention (only training online as opposed to the originally planned hybrid approach) and likely impacted outcomes, particularly social participation.

This study provides as far as we know the first evidence for cost-effectiveness of an eHealth intervention, FindMyApps, for supporting caregivers of people with MCI/dementia compared to an active digital care-as-usual control condition. Effective caregiver support may reduce burden, reduce the risk of adverse health outcomes for caregivers, and therefore allow them to provide better care, for longer (Henderson et al., 2021; van der Lee et al., 2014). Since informal care accounts for 50-90% of dementia care costs, it is not surprising that better caregiver outcomes are associated with benefits to people with MCI/mild dementia (Ma et al., 2018; Meijer et al., 2022; Nay et al., 2015). Whilst the direct effects of FindMyApps on outcomes for the person with MCI/dementia were mostly small and not statistically significant in this study, even small benefits could be important at a population level, if FindMyApps is implemented at a larger scale. The target population of community-dwelling people with MCI or mild dementia constitutes over 100,000 people in the Netherlands alone (Vektis & Factsheet dementie, 2022). The low cost of FindMyApps compared to other psychosocial interventions may make FindMyApps comparatively more feasible to implement at a larger scale. The results of this study provide evidence to policymakers and health care providers, justifying an investment in FindMyApps for supporting caregivers, whilst simultaneously providing people with MCI/dementia a generally positively experienced intervention (Beentjes et al., 2020; Kerkhof et al., 2022; Neal et al., 2023a).

A previous analysis identified significant effect modifiers for quality of life and engagement in pleasurable activities, including diagnosis (MCI vs dementia) and whether or not the person with dementia was experiencing apathy before the start of the study (Neal et al., 2023b). Whilst there were no significant effect modifiers with respect to the outcomes analysed in this economic evaluation, future evaluations with existing digital interventions could nevertheless focus on the subgroups iof participants who may benefit most, such as people with MCI rather than people with dementia. Future studies should also investigate longer term effectiveness and cost-effectiveness of digital psychosocial interventions for promoting social health and quality of life outcomes. Longer studies would allow measurement of relevant outcomes that were beyond the scope of short-term evaluation, such as time to nursing home admission.

Based on the outcomes of this study, FindMyApps may be cost-effective compared to digital care-as-usual with respect to informal caregivers’ sense of competence to provide good care to their counterpart with mild dementia.

Author contributions

D. Neal designed the study, collected the data and wrote the first draft of the manuscript. M. Kucera conducted the statistical analyses and wrote the first draft of the manuscript. B. van Munster provided resources and input in redrafting the article. T. Ettema designed the study, collected the data, and provided input in redrafting the article. K. Dijkstra designed the study, provided supervision and provided input in redrafting the article the article. M. Muller provided supervision and provided input in redrafting the article. R.M. Dröes designed the study, provided resources, provided supervision and provided input in redrafting the article. J. Bosmans designed the study, provided resources, designed and supported the statistical analyses, provided supervision and provided input in redrafting the article.

Disclosure statement

RMD’s Chair in psychosocial care in dementia was partly funded by Foundation to Support of the Association of Christian care for people with mental and nervous diseases (VCVGZ), from September 2014 until August 2022. RMD is Chair of the MEETINGDEM Network, which aims to disseminate the concept of the Meeting Centres Support Programme for people with dementia, and facilitate knowledge exchange, exchange of experiences, and international collaboration in applied research, and since January 2023 of the non-profit JAIN (Joint Artificial Intelligence Network) Foundation, in both roles unremunerated.

Data availability statement

The data that support the findings of this study are available from the corresponding author, DN, upon reasonable request.

Additional information

Funding

This project was carried out within the Marie Skłodowska Curie Actions Innovative Training Network H2020 MSCA ITN, grant agreement number 813196. Additional funding was received from Foundation to Support of the Association of Christian care for people with mental and nervous diseases (VCVGZ), Foundation Codde & van Beresteyn (Foundation C&vB), Bavostichting and Saxion University of Applied Sciences, to cover material costs of the FindMyApps project. No external funders had any role in study design, nor in the collection, analysis, or interpretation of data, writing the report or the decision to submit the paper for publication and H2020 Marie Skłodowska-Curie Actions.