ESRA 2017 Programme
|ESRA Conference App|
Friday 21st July, 09:00 - 10:30 Room: F2 107
Surveying elderly and people with age-associated cognitive impairment - barriers, challenges and opportunities
|Chair||Mr Patrick Kutschar (Institute of Nursing Science, Paracelsus Medical University )|
|Coordinator 1||Professor Martin Weichbold (University of Salzburg)|
Session DetailsTaking a survey basically assumes respondents to be able to understand the questions and to be aware of their answers. But what if these assumptions have to be doubted? In this session we want to discuss the challenges of surveying old people with or without cognitive impairments. Ageing is not causally linked but often accompanied by multifaceted changes and declines in cognitive functions. Against this background, we want to ask for the consequences and specifities which should be taken into account when dealing with such special populations in survey research.
Methodological research on the quality of data obtained by standardized surveys in such vulnerable, multi-morbid older populations especially with restrictions in cognitive function is rather scarce. Implementing fixed age-limits as it is common in some general population survey programmes and national traditions seems not to be an appropriate solution, especially in times of societal ageing. On the contrary, the demand for information about these people is rapidly growing and standardized surveys are applied in gerontological, medical, health care, or social science studies, for instance to examine quality of life or to evaluate satisfaction with perceived care services.
We encourage researchers to share their experiences and suggestions for surveying cognitively impaired and/or old people. Presentations should cover - not exclusively, but predominantly - the following topics:
- Special sampling strategies for these population groups
- Screening instruments or strategies to measure cognitive impairment and to determine the ability to survey participation
- Experiences with the implementation of thresholds in terms of cognitive impairment to obtain vaild and reliable data
- Findings about data quality in surveys with elderly people or people with restrictions in cognitive functions
- Experiences with the linkage of additional measures (e.g. biomarkers, proxy interviewing or non-reactive process-produced data) to the primary survey data
- Alternative (quantitative) data collection strategies apart from the strict standardized approach
Paper Details1. Balanced sampling to survey elderly people living in healthcare institutions
Mr Laurent Costa (INSEE)
Mr Xavier Besnard (SSM Santé - DREES)
The CARE-Institutions French survey of elderly people living in institutions aims to complete the CARE french survey (Capacities, Aids and REsources) of elderly people living in ordinary households and pursues the same objectives: to follow the evolution of their handicap and to measure the involvement of their entourage. Its is directed by the French ministry of health. This population is hard to reach because of its singularity and because there is no census or centralized administrative data describing it : the population registers of every healthcare institution is only available at a local level. Therefore, we need to select only a few establishment in which we survey elderly people.
In order to obtain a representative sample of this population, we use a a three-phase sampling where we first select 30 departments (French NUTS2), then 1,000 institutions and then 6 of the elderly people (over 60 years old) living there. All the departments are not equivalent : they have more or less healthcare institution or residents. In order to make a typology of the departments, we regrouped them with a hierarchical cluster analysis in three groups of homogeneous departments using the type and capacity of the healthcare institutions they contained. To account the differences, we then selected them in the three strata formed by the groups with an unequal probability sampling according to their total number of residents.
In the second phase, we want that the distribution of seniors in the sample will be identical to the distribution of all residents in healthcare institutions. We considered only the institutions within all the departments previously selected in this phase. In order to allow a representative selection of the institutions, a balanced sampling was carried out using the CUBE algorithm to get the same means in the population and the sample for some auxiliary variables such as their type, their legal status and their capacity. We also considered the probability of inclusion of the institution in the sample in the balance and wanted to respect the constraint given of a size fixed sampling (same number of institutions by department in the sample).
The selection of the 6 seniors is random : their ability to answer the questions is judged by the pollster only and not by the healthcare institution to avoid the creation of a selection bias.
2. "We do not want to lose you"! Panel attrition in a longitudinal study of older people in Italy
Dr Emanuela Sala (Dipartimento di Sociologia e ricerca sociale)
Dr Daniele Zaccaria (Fondazione Golgi Cenci)
Monitoring older people’s physical and cognitive conditions is a key issue for sociologists, health researchers and policy makers. Longitudinal surveys of older people are very powerful research resources that can be used to address these topics. However, longitudinal surveys pose specific methodological challenges that may hamper the quality of the data collected and undermine the validity of the research findings. Together with wave one non-response, attrition (i. e., wave on wave non response), is one of the main challanges to data quality in longitudinal surveys. There is an established body of knowledge on attrition in household panel surveys; however, little is known on wave on wave non response in longitudinal surveys of older people. Our starting point is that the processes that lead the general population to continue partecipating in a panel may be different from those of the older people; older people are more socially and psycologically frigile and their propensity to survey response may be more dependent on external circumstances (i. e., the death of a partner). The aim of this paper is to a identify the drivers of wave-on-wave response and identify response patterns; the focus of the analysis, in particular, is to understand the role played by the partner in influencing the decision to continue partecipating. We use data from the Brain Aging in Abbiategrasso study linked to the administrative data from the population registry files provided by the Municipality of Abbiategrasso. The Brain Aging in Abbiategrasso study (also known as InveCe.Ab) is a biannual cohort study of older people born between 1935 and 1939 (aged 70-75 at the first wave) and living in Abbiategrasso, a town near Milan, in northern Italy. InveCe.Ab is a registry-based population study that aims to assess older people’s physical conditions and identify factors associated with the risk of developing dementia and cognitive impairment. We use panel analysis regression models.
3. Factorial Survey among Older Respondents: Age-associated Consistency Effects in Vignette Judgements
Dr Christiane Gross (Institute of Sociology, Leibniz University Hannover, Germany)
Dr Andrea Teti (Epidemiology and Health Monitoring, Robert Koch Institute, Berlin, Germany)
Background: The factorial survey (FS) method is increasingly used in the social sciences. It is particularly suitable for studying events or decision situations that are difficult to assess empirically. Despite the growing popularity of the method, empirical knowledge of the methodological implications of its use remains scarce. Especially the feasibility of the FS method among older respondents has been investigated up to now insufficiently. This contribution address three main questions: (1) Whether the FS method is feasible for surveying decision situations in the elderly, (2) Does response consistency in vignette judgements decrease with age? and (3) Does the number of vignette dimensions that have been accounted for decrease with higher age?
Methods: This paper draws on data from the HOME study, which investigated residential mobility in old age. Specifically, n = 103 respondents aged between 55 and 90 years were asked to make hypothetical relocation decisions. The consistency of these responses was assessed as a function of respondents’ age, gender, immigration background, education, household income, employment status, and intention to move.
Results: No significant differences in response consistency were attributable to age, gender, education, or immigration background. Response consistency increased with income level, although employed participants gave less consistent responses than did participants who were not in paid employment (lack of time effect). The personal relevance of the survey topic was also positively related to response consistency. We find no association of age and number of vignettes that have been accounted for.
Conclusion: The main finding is that the FS method can, under certain methodological conditions, be used among older adults. Very old adults, who are otherwise often excluded from complex assessments, can thus indeed be surveyed by means of the FS method. Moreover, the number of evaluated vignette dimensions does not decrease with increasing age.
4. Combining negative and positive items in scales for older respondents: a split-ballot experiment on data quality
Dr Wander van der Vaart (University of Humanistic Studies, Utrecht)
Dr Tina Glasner (University of Humanistic Studies, Utrecht, Netherlands)
This paper examines the impact of combining negatively and positively worded items
in one scale on response behaviour of older respondents. It is a common rule in questionnaire design that a scale for one construct consists of a mix of positive and negative items. One reason for this is that combining both types of wordings provides a more balanced representation of the issues being measured. This is equivalent to the rule that in questionnaires the orientation of one independent question should be balanced and not directed at a positive or negative answer only. A second, related reason is that combining items of both directionalities is needed to avoid response bias. Combining positive and negative items urges respondents to read more carefully and it reduces acquiescence effects and satisficing behaviour.
While a more balanced representation of items may enhance data quality, research also indicates that combining positive and negative items may lower reliability and construct validity. Positively and negatively worded items often form two factors in factor analysis, even when the content of these items is consistent. Studies usually suggest that negative items are harder to process by respondents and also report that positive items result in higher mean scores. In particular in case of respondents with reduced cognitive or motivational abilities, combining negative and positive items may hamper data quality instead of enhancing it. In this line, the current study focuses on a population of older respondents: they may experience specific difficulty in processing negatively worded items and find alternation of item directionality more burdensome. In addition older respondents may experience more physical adversities diminishing their response motivation.
We performed a split-ballot experiment on item directionality in a study on inhabitants of senior residences in the Netherlands. Accommodations in these residences are open to persons aged 50 years and older who live an independent life (without specific care facilities). Ten residences were selected from medium sized cities as to cover different regions in the Netherlands, include small and large residences (ranging from 36 to 335 inhabitants), and include younger and older populations (mean age ranges from 69 to 83). All 1259 inhabitants of the ten residences were send a questionnaire, resulting in a sample of N=397 respondents (31%).
The split-ballot experiment was performed on a ‘meaning in life scale’ (“MiL”, 7 items) and a ‘self-reliance scale’ (“SL”, 8 items). Two versions of the questionnaire were composed: in one version all SL-items were positively formulated while the MiL-items were alternately negatively and positively worded; the opposite was done in the second questionnaire version. Each respondent randomly obtained one version of the questionnaire.
Analysis focuses on reliability and construct validity of the two scale-versions, and specific indicators of data quality, such as straight lining and number of item non-response. The outcomes are related to cognitive and motivational mechanism that may underlie response behaviour of older respondents when dealing with scales of positive and negative items. Possible impact of item directionality on the substantive outcomes
5. Question and response order effects in a survey of nursing home residents with different grades of cognitive decline
Mr Patrick Kutschar (Institute of Nursing Science and Practice, Paracelsus Medical University Salzburg)
Little is known about the specifics of cognitively impaired elderly in the question-answer-process and the possible consequences for the quality of survey data. At the same time, standardized surveys of old and oldest old respondents are broadly used, for instance in Gerontology, Nursing or Medical Science. Here, it is usually assumed that self-report instruments can be applied in elderly with up to moderate cognitive impairment. Mainly due to age-associated declines in cognitive function, it can be hypothesized that the quality of survey data obtained in nursing home resident (NHR) populations is threatened and defective. One possible source of error may be varying answers as of order effects.
This talk will present very recent findings of question and response order effects in NHR and focus on the moderating effects of different grades of cognitive decline.
Data comes from the initial data evaluation (February-April 2017) of the large-scale research project ‘PIASMA’ [Project to implement an adequate pain management in nursing homes]. The PIASMA study aims to improve the multiprofessional pain management and the pain situation of nursing home residents by means of a structured non-pharmacological intervention. In this cluster-randomized controlled trial, more than 1,000 residents with up to moderate cognitive impairment from 20 randomly selected nursing homes in Bavaria, Germany, will be interviewed face-to-face about pain and depression by research assistants using standardized self-report instruments. The Mini-Mental-State-Examination (MMSE) is used to assess the cognitive function of NHR. Residents with MMSE-scores between 18-30 points are stratified as residents with no or mild cognitive impairment, those with MMSE 10-17 as residents with moderate cognitive impairment.
Instruments and Split-ballots.
Pain is measured by the BPI - Brief Pain Inventory (four NRS-10 items for pain intensity, seven NRS-10 items for pain interference); depression is measured by the GDS-SF15 - Geriatric Depression Scale-Short Form (15 dichotomous ‘yes/no’ items). Split-ballot #1 (coin-flip assignment): To test question order effects, the seven items of BPI’ pain interference scale are arranged in reversed order in the experimental version of the instrument. Split-ballot #2 (coin-flip assignment): To test response order effects, the GDS-SF15 dichotomous response categories are switched to ‘no/yes’ and compared to the originally validated version.
In general, it is hypothesized that no or very little question order effects but considerable response order effects can be observed. This is constituted on the assumption that elderly exhibit poor working memory capacities. As a consequence, the context (question order) in which questions are presented would not have an essential impact on the further’ questions ratings, while those answer categories presented last are easier on call than those presented first (response order). The effects are expected to be moderated by the cognitive function: residents with no or mild cognitive impairment are assumed to exhibit more question order effects but less response order effects than those with more progressed declines in cognitive function.