Factors associated with health-related quality of life among home-dwelling older adults aged 75 or older in Switzerland: a cross-sectional study – Health and Quality of Life Outcomes

Study design and setting

This is a cross-sectional study conducted in 2019 in Canton BL, in northwestern Switzerland [37]. Canton BL is a German-speaking region, inhabited by around 290,000 citizens and has the second-highest proportion of population aged 65 or above (22.4%) and aged 80 or above (6.7%) in Switzerland [2].

Study participants and data collection

The participants were recruited via postal mail, with no sampling method necessary as we included all those eligible, namely all home-dwelling older adults living in Canton BL who were aged 75 and above. The INSPIRE Population Survey is embedded within the larger INSPIRE project (https://inspire-bl.unibas.ch/), in which an important component of the care model is screening for frailty. As frailty increases with age [39], the age cut-off of 75 years was chosen as an age when we consider older adults are more likely to be at risk of frailty and can thus benefit the most from the integrated care intervention.

A survey package containing the questionnaire along with instructions for filling it out, an information sheet, a personalized cover letter, a prepaid return envelope and the informed consent form was mailed to the home address of all community-dwelling persons aged 75 years or older in Canton BL, which we received from the Cantonal Statistical Office. Thus, the filled-out questionnaires were also returned by postal mail. All the questionnaires were pseudonymized prior to being delivered, with the intent to allow potential follow-up in the future. However, due to concerns of the general public on data security and based on several stakeholder recommendations, we anonymized the questionnaires after having sent them and destroyed all documents containing identifiable information.

The survey was successfully delivered to 28,791 older adults living at home in Canton BL and a total of 8,846 questionnaires were returned (Response Rate = 30.7%). During the validation process, 60 questionnaires were excluded from the analysis (i.e., based on ineligible ZIP codes, respondent’s age, or residents in a long-term care institution), resulting in a final sample of 8786 participants. We consider the response rate to be representative, as it is much higher than what is reported in literature for postal surveys [40]. Furthermore, we found that the prevalence of frailty among community-dwelling older adults as measured by the GFI in a comparable study population to be in line with our observed results [41].

A detailed description on the development, dissemination and characteristics of the population survey have been reported elsewhere [37].

Variables and measurements

As the current study is part of an implementation science project, the survey was designed with the input of various stakeholders. The list of stakeholders includes but is not limited to a group of older adults, representatives of local policymakers, community care providers and representatives of nursing homes. The survey items are henceforth a combination of validated tools and investigator-developed items. Detailed information on the development of the survey and overall participants’ characteristics have been reported elsewhere [37].

Outcome variable

HRQoL was assessed using the EQ-5D-5L instrument [16], a generic standardized instrument comprising of a short descriptive questionnaire and a visual analogue scale (EQ-VAS). The descriptive questionnaire includes the following dimensions of health: mobility, self-care, usual activities, pain/discomfort and anxiety/depression. Each dimension has a five-level response of severity, ranging from 1—no problems, 2—slight problems, 3—moderate problems, 4—severe problems to 5—unable to/extreme problems, which correspond to potential health states [42]. These health states are then converted into a single EQ-5D-index, by applying a country-specific valuation algorithm [42]. In lack of a Swiss value set, we used the German value set algorithm by Ludwig et al. [43]. Along with the descriptive questionnaire, this instrument also includes a visual analogue scale (EQ-VAS). This scale is similar to a thermometer, where the endpoints are labelled ‘The worst health you can imagine (0)’ and ‘The best health you can imagine (100)’. The construct validity of the EQ-5D-5L instrument is well examined in use among older adults, such as in the Bhaduri et al. study, who computed the Spearman’s rho between each of the EQ-5D items and the Barthel Index (Spearman coefficients 0.42) [44].

Micro level variables

Year of birth was used to calculate the age of the participants at the time of the survey completion and was recorded as a continuous variable. Gender information was collected as “Male” or “Female”. The original answers categories for the level of education question were regrouped into four categories: “Tertiary” (“University” and “University of Applied Sciences”); “Secondary” (“Gymnasium” and “Apprenticeship”); “Elementary or None” (“Elementary School” and “No degree”) and “Other”. Income, which was originally collected as a monthly household income in Swiss Francs (CHF), was converted to individual income by dividing the household income by the number of people living in the household, following the guideline of the Swiss Centre of Expertise in the Social Sciences on how to measure income in surveys [45].

Living situation of the participants was assessed through an investigator-developed item asking who they currently lived with. For the purpose of the analysis, the answer choices were dichotomized into: living alone vs living with others (a spouse/partner, an adult child, other adults, siblings or a professional caretaker).

The health status of the participants was assessed by asking whether they experienced vision, hearing or memory problems in their daily life, or if they had unintentionally lost weight in the past 6 months. Polypharmacy, defined by the Groningen Frailty Index (GFI) tool [46] as taking four or more medications at once, was also recorded. Variables pertaining to health status and polypharmacy had dichotomized “Yes” or “No” answer choices. The criterion validity of the GFI tool among older adults has been examined (r − 0.62) [47].

Socioemotional well-being of the participants was assessed using three questions from the GFI tool [46] which ask the participants whether they feel empty, miss the company of others or feel abandoned. The answer choices for these questions included: “Yes”, “Sometimes” or “No”, which for the purpose of the analysis were dichotomized into “Yes / Sometimes” and “No”.

The lifestyle section included questions on smoking, alcohol intake and physical activity. The participants were asked about their smoking habits, with answer choices being regrouped into “No” (“Not currently, but I was a smoker before” and “No”) and “Yes” (“Yes, daily” and “Yes, not daily”). Additionally, alcohol intake was assessed by asking the number of drinks a participant consumed in a typical day; with a glass of wine, one dosage of beer of 355 ml or a 40 ml spirit alcohol counting as one drink [48]. The answer choices for this question included: “No drink”, “1–2 drinks”, “3–4 drinks” and “5 or more drinks”. The answer choices were dichotomized into “ ≤ 2 drinks/day” or “ > 2 drinks/day”, based on recommendations of the Swiss Federal Commission for Issues Related to Addiction and Prevention [49]. The participants were also asked about how many minutes they engaged in vigorous-intensity, moderate-intensity physical activity and in muscle-strengthening activities in a typical week. The WHO recommends that an older adult should engage in at least 75 min of vigorous-intensity, or in at least 150 min of moderate-intensity physical activities within a typical week [50]. For additional health benefits, the WHO recommends that an older adult engages in muscle strengthening physical activity in at least 2 days per week [50]. Due to potential multicollinearity among these three variables, we computed one variable related to physical activity. If a person scored 1 or above, which indicated they engaged in any of the three activities as recommended, it was recorded as being physically active. The answer choices were thus scored as: “Per WHO recommendations” and “Below WHO recommendations”.

Meso level variables

Informal daily support from individuals was also assessed and answer choices were dichotomized into: currently receive support from another individual (spouse, younger family member, friend or neighbour) or currently do not need such support. Participants were also asked whether they currently received daily support from organizations, through listing common organizations that older adults receive support from in Switzerland. These include home care organizations, social care organizations, humanitarian organizations (e.g., Red Cross) and disease-specific associations (i.e., Diabetes association, Alzheimer`s association and Parkinson’s association). The answer choices for this question were dichotomous “Yes” or “No”.

Availability of social support was assessed through the Brief Social Support Scale (BS6), which has been validated in German [51]. This instrument includes three questions to assess the availability of tangible support (i.e., someone to accompany them to doctor`s appointments, someone to prepare their meals when unable to and someone to help with daily chores when sick) and three others to assess the availability of emotional support (i.e., someone who can give them good advice, someone they can confide in during a crisis and someone who understands their problems) [51]. The responses are scored on a 4-point Likert scale ranging from 1- “never” to 4- “always”. A sum score of the six-items, ranging from 1 to 24, is calculated and then dichotomized into: “Low to moderate support (a score of up to 17)” and “High to very high support (score of 18 and higher)” [51]. Reliability of the subscales has also been proven, as indicated by Cronbach´s alpha: emotional support α = 0.87, tangible support α = 0.86 and overall α = 0.86 [51].

To assess involvement in social activities, the questionnaire included an investigator-developed list of hobbies and activities (e.g., sports, political parties, church gatherings, volunteering, meeting with family and friends) for which participants could indicate whether they were active in or wished to be active in. To provide more granularity in the results, we grouped the participants into three groups: those who were active in more than one of the activities, those who were active in only one, and those who wished to be active in at least one of the listed activities, but were not currently.

Macro level variables

Type of insurance of the participants was assessed by asking them whether they were insured with statutory health insurance alone or with statutory health insurance plus supplementary private insurance. Although health insurance can be considered an individual factors as well, we have included it as a macro-level factor because in Switzerland, basic health insurance is mandatory. The benefit package of the basic insurance is more comprehensive than in most other countries and defined at the national level, where payment mechanisms are largely defined by federal and cantonal regulations.

Information on supplementary government support was captured by asking the participants whether they received this type of support or not. Supplementary government support is a specific type of help in Switzerland, that support individuals financially if their pension or income do not cover minimum living costs.

Statistical analysis

General descriptive statistics were computed for the EQ-5D-5L domains and all independent variables. Categorical variables (e.g., gender, education, etc.) are reported as frequencies and percentages whereas continuous variables (e.g., age and income) are reported as medians and interquartile ranges or means and standard deviations. The EQ-5D-5L descriptive results are presented by recording the number and percentage of patients reporting each severity level of each dimension of the EQ-5D-5L instrument.

To gain an initial understanding of the association of the independent variables with HRQoL (for both the EQ-VAS and the EQ-5D-index), standard univariate tests such as Mann–Whitney U test and the Kruskal–Wallis test were used for categorical variables. The Spearman`s correlation coefficient was used to test the association of the outcome with continuous predictors.

The influence of independent variables at the macro, meso and micro level on both EQ-5D-index and EQ-VAS were tested using multiple linear regression modelling. All covariates of the conceptual model, from all levels, were included in the regression model, irrespective of significance, in order to determine the relationships of each variable with the outcome variable. Because ceiling effects were observed in previous studies using the EQ-5D-5L in general population surveys [52], we used Tobit-regression modelling. This is a variation of multiple regression, which is capable of correct inference in the presence of ceiling effects [53]. We tested if the underlying assumptions of the linear modelling were met and used the Variance Inflation Factor (VIF) to test the presence of multicollinearity among independent variables. The level of significance was set at 0.05.

Data was primarily missing due to item nonresponse, and after the analysis of missing patterns, we considered our data to be missing at random (MAR). In our dataset, we observed two variables with more than 5% of missing data: individual income (5.3%) and availability of social support (26.6%). As our data met the recommendations of Jakobsen et al. [54] for when to use multiple imputation (i.e. missing data is above 5% but below 40%, data was missing not only on the dependent variable, the Missing Completely at Random—MCAR assumption could not be plausible, and data is considered MAR), we applied multiple imputation by chained equations (MICE) to impute missing values [55]. We also ran a sensitivity analysis using the observed data and found no significant differences in results between the observed and the imputed data.

All analyses were performed using R, version 1.3.1093 for Mac OS [56].