Research techniques and use of data in research on ageing

1
Research techniques and use of data in research
on ageing

• Yvonne Wells
• Lincoln Centre for Research on Ageing

2
Where this workshop is going

• What do you need?
• Focus on quantitative data
• Dont under-rate qualitative research
• Cross-sectional vs. longitudinal research
• Measures of change
• ABS and administrative by-product data sets
• Content and examples of use
• Discussion

3
What goes without saying

• Never generalize from younger populations to
  older ones
• Never assume that all older people are alike

4
Old people are pretty much alike

• False
• On a task measuring speed and accuracy
  performance (Stroop Task), older adults displayed
  as much individual variation as other populations
  (Rush, Panek, Russell, 1990).
• Individual variation in mental abilities
  increases as participants age (Rabbitt et al.,
  2001)
• Older adults show large individual differences in
  terms of subjective well-being and quality of
  social relationships (Smith Baltes, 1993).

5
Research on ageing is increasing
Number of international PsycInfo journal
publications addressing ageing, childhood and
adolescence 1993-2002.
Wells, Y. (2005). Research and practice with
older adults The picture in Australia.
Australian Psychologist, 40(1), 2-7.
6
especially in Australia
Growth curves comparing Australian with
international research on ageing, 1993-2002
7
Time in ageing research

• Age years since birth
• Wave number of data collection points since T0
• Analyses can centre on an event of interest
• e.g., time to death, time before/after a dementia
  diagnosis
• This technique can help separate normative from
  non-normative changes with ageing

8
Hazards of research with older people

• Participation rates
• Survivor effects
• Physical health and disability issues
• vision, deafness, fatigue, mental status, balance
• methods of gathering data do have consequences.
• Response bias may increase with age
• Issues of meaning and acceptability
• Ethical issues and privacy
• Cross-sectional vs. longitudinal studies
• In longitudinal studies, selective attrition

9
Longitudinal vs. cross-sectional

• Cross-sectional studies can be misleading because
• Cohort effects
• Period effects
• Selective attrition
• Longitudinal research permits estimates of
  individual change over time
• Easier to retain individuals with failing health
  to an ongoing study than recruit them to a new
  one
• Longitudinal research is expensive
• Cross-sequential analysis is a good compromise

10
Response bias and ageing

• McAvay, G. J. et al. (2005). Symptoms of
  depression in older home-care patients Patient
  and informant reports. Psychology and Aging, 20,
  3, 507-518.
• Older, medically ill population reports of
  depressive symptoms vary systematically between
  patients and family member informants.
• Agreement on somatic symptoms was poor, even
  though reported most often by patients and
  informants
• Patients were more likely to report sleep
  problems whereas informants were more likely to
  report fatigue
• Informants were more likely to report
  indecisiveness and inability to concentrate
• Patients reported more suicidal thoughts than
  informants
• Younger informants reported more cognitive
  symptoms than patients, whereas older informants
  reported fewer cognitive symptoms that patients

11
Response rate and ageing

• Kaldenberg, D. O., Koenig, H. E., Becker, B. W.
  (1994). Mail survey response rate patterns in a
  population of the elderly Does response
  deteriorate with age? Public Opinion Quarterly,
  58, 68-76.
• Response rate 58 for 60-62 year-olds
• Fell by half a percentage point for each age
  group
• Influence of age on data quality varied between
  items types
• No age difference in missing data on open-ended
  questions
• See graph

12
Response rates by age and type of question (from
Kaldenberg et al., 1994)
13
Attrition is considerable

• Figure is from the Health Status of Older People
  study

14
Attrition data quality in HSOP 1994-2005
15
Loss rate in HSOP
Loss average of 12 per year, accelerating.
16
Attrition is not random
Figure 2
Sliwinski, M., Buschke, H. (1999).
Cross-sectional and longitudinal relationships
among age, cognition, and processing speed.
Psychology and Ageing, 14, 18-33.
17
Retest effects are considerable
Figure 1

• Salthouse, T., Schroeder, D. H., Ferrer, E.
  (2004). Estimating retest effects in longitudinal
  assessments of cognitive functioning in adults
  between 18 and 60 years of age. Developmental
  Psychology, 40, 813-822.
• Is improvement with retest error?

18
How?

• Prospective vs. retrospective
• Ways of collecting data (interview vs.
  self-complete questionnaire)

19
Retest vs self-ratings of change

• Data from Healthy Retirement Project
• Prospective panel study
• n gt 400
• Next table provides data on change in first 12
  months of retirement

20
Retest vs self-ratings of change
21
What predicts discrepancy?

• Floor and ceiling effects
• Self-image
• Not
• Recency
• Social comparison

22
Self-rated health in HSOP
23
Use of existing datasets

• Australian Bureau of Statistics
• Census data
• National Health Survey
• Survey Disability, Ageing and Carers
• Administrative by-product data sets for
• Home and Community Care program
• Aged Care Assessment Program
• Death Registry
• Hospital Data
• ACCMIS

24
Examples

• Focus on CALD in health data
• Interesting analyses from ACAP and HACC data

25
Using the ACAP and HACC MDSs
26
ACAP MDS

• 200,000 assessments per year
• 1.2 involve a psychologist
• 27.5 of clients have a diagnosis of dementia
• Increases to 34.5 of assessments that involve a
  psychologist

27
ACAP MDS domains

• Registration
• Accommodation and living arrangements
• Dates
• Carer data
• Activity Limitations
• Current assistance and recommended assistance
• Current government service use and recommended
  use
• Health conditions
• Recommended accommodation
• Assessor profession
• Approvals
• Care coordination

28
Reporting

• Numbers
• Timeliness
• Access to key groups
• Recommendations for key groups
• Ad hoc data analyses

29
Example 1 Assessment numbers
Total equivalent assessment numbers by location
of assessment, ACAP, Australia 1994-1995 to
2005-2006
30
Example 2 Change in client age
Figure 16 Change in client age over time,
1995-1996 to 2005-2006 ( of client group)
31
Example 3 Dependency

• How do dependency and service use at assessment
  influence recommendations?

Mean dependency score by location at assessment,
service use, and outcome
32
Example 4 ACAT Variability
40-80 of variance in recommendations explained
33
HACC and ACAP MDSs

• Advantages of using both datasets together
• HACC has much better data on service use
• ACAP has much better data quality and information
  on outcomes

34
Are small amounts of service protective? 1
recommended to community by assessment setting,
hours of service and dependency
35
Are small amounts of service protective? 2
recommended to community by assessment setting,
cost of service and dependency
A. By costs of services used
36
Using the National Health Survey
37
CALD groups aged 60 in NHS 2001
38
Mental health of older Italians

• Comparison with
• Australian-born
• migrants from English-speaking countries
• migrants from other CALD countries

39
Italian-born group much more likely to have high
scores indicating low well-being
Mental Health (K10)
40
with High/Very high scores by Sex
Italian-born women much more likely to have high
scores indicating low well-being
41
Mean scores on feeling depressed
Italian-born group much more likely to have high
scores indicating low well-being
42
Used medication for mental wellbeing
43
Medication use and mental health
44
Risk factors for poor mental health

• Physical health problems
• Health behaviours (inactivity, alcohol, smoking)
• Social isolation
• Financial insecurity

45
Self-rated healthItalian-born group much more
likely to rate health as fair or poor
46
Arthritis Italian-born group much more likely to
currently have arthritis
47
Diabetes Italian-born group more likely to
currently have diabetes than Australian-born or
ESB groups
48
Sedentary Italian-born group was more likely to
be sedentary
49
Obesity
Italian-born group was more likely to be obese
than other groups
50
Italian-born group (especially women) much more
likely to be in the lowest income group
Income quintiles
51
Italian-born group much more likely to be rated
speaking it not well or not at all
English language capacity
52
Multivariate analysis

• In a logistic regression analysis controlling for
  sex
• Italians twice as likely to have high/very high
  scores as the Australian-born
• Other CALD groups 1.5 times as likely to have
  high/very high scores as the Australian-born

53
Mediators What explains group differences?

• Proficiency in English
• Health status
• And what doesnt
• Exercise level (mediates sex difference only)
• Arthritis (mediates sex difference only)
• Income

54
What are your research issues?
55

• Australian Institute for Primary Care
• La Trobe University
• Victoria 3086
• 61 3 9479 3700
• aipc_at_latrobe.edu.au