Health-related quality of life in elderly, multimorbid individuals with and without depression and/or mild cognitive impairment using a telemonitoring application

The overall study “Autonomy despite multimorbidity in Saxony through patient empowerment, holistic care for older people with networking of all regional institutions and service providers” (Acronym: ATMoSPHAERE) contributed to the care of elderly, multimorbid people. A significant challenge is the feasible identification of risk groups within the cohort of participants for whom it could be difficult to use telemonitoring measures. These include individuals with MCI or depression—a patient cohort that was excluded or underrepresented in previous studies [43], although the afflicted suffer from the most common diseases of old age. Supported by professionals in the fields of primary care, geriatric medicine, information technology, social services and economics, a regional ecosystem was created that combined both medical and non-medical services on the basis of a medical technology platform. Cross-sectoral networking enabled e.g. the coordination, optimization and rapid adjustment of care plans and the recording of patient needs. Thus, a comprehensive care concept for elderly, multimorbid individuals was developed, which included aspects of prevention, diagnosis, therapy and aftercare/rehabilitation. Its aim was to enable people to lead a largely self-determined life by controlling their own disease-related restrictions. The increase in HRQoL of the study participants (in the following referred to as “participants”), measured with the SF-12, was defined as a secondary outcome in the study [44, 45]. Participants were provided with a TMA, which enabled them to measure vital signs, transfer these data to a care coordination center, and to request home-based assistance and services from regional providers.

The ATMoSPHAERE study was designed as a longitudinal intervention and feasibility study and was approved by the ethics committee at the Technische Universität Dresden (approval number EK 1012016). The study was performed in the two largest cities in Saxony (Dresden, Leipzig). GP practices from a network of accredited academic teaching practices were included to facilitate recruitment of elderly, multimorbid individuals.

Study participants Individuals who met inclusion criteria (Table 1) were informed about the study by their GP. Eligibility of individuals was assessed by applying validated instruments measuring cognition (Mini-Mental State Examination, MMSE [46], and clock-drawing test [47]), mobility (Timed “Up & Go” test, TUG) [48] and their independence in everyday life activities (Instrumental Activities of Daily Living assessment) [49]. If eligible individuals decided to participate in the study, they signed a declaration of consent and were informed about the possibility to withdraw their participation at any time.

After inclusion, participants were assessed with regard to possible depression using the validated measuring instrument Geriatric Depression Scale (GDS) [50]. According to the results of the MMSE and the GDS, participants were assigned to risk groups (RGs), as shown in Table 2. RGs were defined as such if mental disorders could make it more difficult for the elderly participant to operate the TMA. The study focused on two of the most common mental syndromes in old age: depression and cognitive impairment [51]. The following specific inclusion criteria were applied:

The TMA consisted of the telemonitoring hardware ASUS ZenPad 7.0, Samsung Tab 4, sphygmomanometer, pulse oximeter, body weight scale, and the telemonitoring software Motiva and was provided by the technical project partner Philips Medical Systems GmbH.

Sociodemographic and health data of new participants were collected by study assistants and then transferred to a Care Coordination Centre (CCC). Then a technician from the CCC instructed participants in their homes on the use of the TMA and they had to carry out the first complete measurement and transfer process of their vital data in presence of the technician. After confirmation of having understood the use of the TMA, the care process started. More detailed description of the telemedicine care process for participants has already been described in a previously published paper [52].

According to treatment regimens of participating GPs, participants measured their blood pressure (BP), heart frequency (HF), blood oxygen saturation (SpO₂), and body weight (BWT) via measuring devices. The frequency of these vital sign measurements was individually determined by each participant’s GP (once weekly, 2–6 times per week, and daily). Vital sign data were transferred from measuring devices to the tablet and to the CCC, which constantly monitored and critically assessed the data for intervention necessity. If any limit values were exceeded, control questionnaires were sent to the participant and a care coordinator contacted the participant by telephone. If intervention was required, the GP was provided with relevant information immediately.

For each participant and vital sign, each complete calendar week may be evaluated with respect to the relationship between actual and recommended frequencies of measurement (see supplement (Table S1)). We classified a participant as exactly adherent in a certain calendar week if both the number of days with measurements and the total number of measurements were in line with GP recommendations. The calendar week was classified as over-measured if the number of days with measurements and/or the total number of measurements was higher than recommended by the GP. In contrast, the calendar week was classified as under-measured if the number of days with measurements and/or the total number of measurements was lower than recommended by the GP.

To examine changes of participants´ HRQoL and symptoms of depression between t₀ (baseline, installation of TMA), t₁ (after six months with TMA), and t₂ (after 12 months with TMA), participants were interviewed by a research scientist via telephone using the validated questionnaires SF-12 [53] and the GDS [50]. The SF-12 was used in the present study to measure HRQoL. The questionnaire covers eight domains of health that are summarized in two summary scores, the PCS and the MCS, both including six items. The scores of the HRQoL domains range from 0 to 100, with higher scores referring to higher quality of life [54]. Several studies have reported the good psychometric properties of the SF-12 in different age groups and specifically in elderly persons [55]. The GDS consists of 15 questions whose sum score differentiates depressive and non-depressive individuals. Research on screening performance in the elderly showed that the use of the GDS allows an appropriate screening of the very old individuals and a reliable detection of a major depression [56].

Data from BP and BWT measurements as recommended by GPs served as the basis for evaluating participant engagement. Due to limited participant and time coverage, two measurement types (heart frequency and blood oxygen saturation) were not included in the analysis.

Baseline characteristics of study participants were analyzed descriptively. Data on measurements/transfer of vital data were also analyzed descriptively regarding frequencies of engagement with measurements, over-measurement, and under-measurement over time (classification according to supplement (Table S1)). Relationships between HRQoL and risk group membership were visualized by box plots. Statistical significance of differences between risk groups regarding HRQoL was tested using Wilcoxon rank-sum tests for unpaired samples. Bivariate associations between participant engagement with blood pressure and body weight measurements, sociodemographic characteristics and PCS / MCS scores at baseline were investigated using the Spearman’s rank correlation test and the Kruskal–Wallis test, respectively. Relationships between HRQoL and participant characteristics were investigated using panel data on the PCS and the MCS gathered at baseline (t₀), at the six-month follow-up (t₁), and the 12-month follow-up (t₂). To explore associations between HRQoL and covariates, linear multilevel regressions were conducted using the PCS and the MCS as dependent variables and participant characteristics as covariates. In addition to the covariates, the right-hand side of these models contains a random intercept at the participant level. This random intercept captures correlation of the participants’ outcomes over time. To explore potential relationships between changes in HRQoL and participants engagement with measurements, linear multilevel regression models for changes in the PCS scores and the MCS scores between t₀ and t₁, and t₁ and t₂, respectively, were specified. The share of calendar weeks between the PCS and the MCS measurements in which a participant either followed the recommendations of the physician exactly or measured more often than recommended (over-measured) was included as a covariate capturing active study participation. The rationale behind pooling engagement with exact measurement and over-measurement was that both might be expected to be associated with a higher HRQoL relative to a limited study participation as reflected by under-measurement. Hence, exact and over-measurement were combined to generate a single binary indicator capturing active study participation. Sensitivity analyses were conducted by treating the shares of calendar weeks classified as exact or over-measured as separate variables (see supplement (Table S2)). Associations were estimated separately for study participants measuring BP and BWT, respectively. All estimates were adjusted for participant characteristics. Significance levels of 5% and 1% were reported for all regression analyses. Statistical analysis was performed using the statistical software IBM SPSS Statistics for Windows v25.0 (IBM Corp., Armonk, NY) and R v3.6.1.

Descriptive analyses of the changes in the physical and mental HRQoL of the participants showed differences between the risk groups (Fig. 2). Statistically significant differences could be determined using Wilcoxon rank-sum tests (see supplement (Table S3)).

The results indicate that the median PCS score of study participants with no depression or MCI (RG 1) remained stable over time and was consistently higher than in study participants with mental disorders (RG 2, RG 3, RG 4). Study participants with MCI only (RG 3) differed only slightly in their median PCS score from RG 1, although this group of study participants showed fluctuations in the median PCS scores over time. Overall, study participants with depression had lower median scores for the PCS than RG 1 and RG 3, but fluctuated only slightly over time. Study participants with both depression and MCI (RG 4) had the lowest median PCS scores at baseline. Their median PCS scores fluctuated over time, but were similar to the median scores of the study participants in RG 2. Statistically significant differences regarding physical HRQoL of participants were found at t₀ (baseline) between RG 1 and RG 2 (p = 0.0013) and between RG 1 and RG 4 (p = 0.003). After six months of using the TMA (t₁), statistically significant differences in RG 1 compared to RG 2 (p = 0.001), RG 3 (p = 0.024) and RG 4 (p = 0.017) were found. At t₂ there were statistically significant differences in RG 1 compared to RG 2 (p < 0.001) and to RG 4 (p = 0.005) and between RG 2 and RG 3 (p = 0.048).

The MCS score showed a similar picture in the distribution between the risk groups, although study participants’ mental HRQoL scores were higher than the physical HRQoL scores. Participants from RG 1 also showed the highest MCS overall scores as well as the highest MCS scores over time, with the highest median MCS score of all risk groups at time point t₂. Participants from RG 3 differed only slightly from the median MCS scores of the RG 1 study participants and remained nearly stable over time. Study participants with depression (RG 2) had slightly lower median scores, although their scores also changed only slightly over time. As with regard to the PCS score, the study participants from RG 4 had the lowest median MCS score at baseline; however, this increased over time and became similar to the MCS scores of the participants from RG 2. With regard to the MCS at t₀, RG 4 showed statistically significant differences to RG 1 (p = 0.006) and to RG 3 (p = 0.022). In addition, statistically significant differences between RG 1 and RG 2 (p < 0.001) after six months of TMA usage (t₁) were found. At t₂ there were statistically significant differences between RG 1 and RG 2 (p = 0.001), RG 1 and RG 4 (p = 0.023), as well as between RG 2 and RG 3 (p = 0.015).

The descriptive results withstand an adjustment for other potential covariates of PCS and MCS. Linear mixed effects regression results for the PCS and the MCS levels as dependent variables showed significant differences between risk groups (Table 4). Relative to study participants with no mental disorders (RG 1), study participants with depression (RG 2) and with both depression and MCI (RG 4) had significantly lower PCS and MCS scores. Compared to RG 1, no significant differences were found for participants with only MCI (RG 3). Higher age was associated with lower PCS and MCS scores. Female participants had lower MCS scores than male participants but did not differ significantly from the latter with respect to the PCS scores. The estimation results did not provide evidence for associations between school education and marital status. While the PCS did not differ between study dates (t₀, t₁, t₂), there was evidence for an increase of the MCS over time as reflected by the positive and statistically significant coefficients of the indicator variables for t₁ and t₂.

The linear mixed effects regressions for changes in the PCS scores and the MCS scores over time provided evidence for different relationships between the PCS, the MCS, and TMA usage (Table 5). The share of calendar weeks in which a study participant was exact or measured more often than recommended was not associated with changes in the PCS. This finding holds for both BP and BWT. On the contrary, a higher share of exact or over-measured calendar weeks was significantly related to an increase in the MCS over time. Relative to a study participant that measured BP less than recommended in all calendar weeks, a participant that measured as recommended or more often in all calendar weeks was estimated to experience an MCS increase of approximately 3 points. A similar positive and significant association was found for changes in the MCS and BWT measurement with an estimated effect size of 5.8 points. With respect to the other covariates, there was little evidence for significant relationships with changes in the PCS or the MCS.

We found no statistically significant associations between engagement with blood pressure or body weight measurements and risk group assignment, sociodemographic characteristics or HRQoL at baseline (see supplement (Table S4)).

Our study sample is a non-representative convenience sample. Eligible individuals were easily accessible and recruitable via the participating GPs. Selection bias could exist, as the GPs and the individuals voluntarily participated in the study. We conducted telephone interviews with the participants to assess their HRQoL and possible depression over time using the SF-12 questionnaire and the GDS, respectively. The SF-36 and the GDS have been validated for interviews [54, 70]. A comparison of the survey types with the SF-36 showed that the results for the MCS score in telephone interviews were 2.4 scale points higher than in the postal survey [54]. This may be associated with a tendency towards social desirability among respondents, which could result in a response bias [54]. Participants may have behaved differently because they knew that they were participating in a study and were under observation (Hawthorne effect). By implementing a randomized controlled study design, statements about the effectiveness of the telemonitoring intervention used would have been possible. However, that design was not used in the present study.

We were able to include a vulnerable target group relevant for telemedicine research in our study: multimorbid individuals aged 65 years and over with mental comorbidities. Even if the risk groups examined are not large in number, the results provide important information about the changes in physical and mental HRQoL of these target groups. The results obtained complement the research area of telemedicine applications in elderly people.

Our feasibility study showed that depression and the coexistence of depression and MCI are negatively associated with HRQoL of elderly, multimorbid people using a TMA. Engagement of individuals with vital data measurements via telemonitoring applications which are in line with the treatment regimen of the GP may increase the mental HRQoL, but not the physical HRQoL of individuals. An increased HRQoL of individuals can be a decisive factor in their constant engagement with telemonitoring measures. In order to examine the effectiveness of future telemonitoring interventions, randomized controlled study designs must be applied. Future research should focus on further determinants that can influence the usability of telemedicine applications for mentally impaired people.