The nature and impact of caring for family members with a disability in Australia

Research Report No. 16 – June 2008

3. Method and sample

Ben Edwards

This chapter describes the process by which the sampling and fieldwork for the Families Caring for a Person with a Disability Study (FCPDS) were conducted. The major steps involved in the process of designing the interview schedule and the recruitment and interviewing of the participants are outlined. Response rates are presented and the demographic characteristics of the sample are then compared with the population.

Overview of the sampling procedure and fieldwork

The sample was obtained through random selection from the Centrelink administrative database of carers over 18 years of age who were at the time receiving the Australian Government Carer Allowance or Carer Payment. As discussed in Chapter 2, those receiving Carer Payment may also be eligible for Carer Allowance. The interviews were conducted using computer-assisted telephone interviewing.

The sample selection and recruitment involved the following steps:

  1. A random sample of 5,000 people receiving Carer Payment and/or Carer Allowance in June 2006 was selected from the Centrelink administrative database.12
  2. Selected Centrelink customers were sent a brochure advising them that they had been selected to be potentially contacted to participate in the survey and that if they did not want their contact details released to the fieldwork company, they needed to contact FaHCSIA to opt out of the study. The brochure also detailed the nature of the study and its aims (see Appendix B). Sixty-five carers opted out of the survey at this point, leaving a total of 4,935 potential respondents available for the survey.
  3. Contact details and the type of payment received for the potential respondents were then provided to the fieldwork agency.13
  4. The fieldwork agency contacted potential respondents by telephone. A total of 10,856 calls were placed to 2,126 sample people to achieve 1,00314 completed interviews. This equates to one completed interview for every 10.8 calls. The most commonly occurring call outcomes were: no answer (26.8%), "soft" appointment15 (26.5%) and answering machine (15.4%).

The survey collected information from primary carers about how their families cared for a person with a disability and the impact of this on their family.16 The average interview length was 34 minutes. The final sample size achieved was 1,002 carers.17

As with all surveys, there was a level of non-response. Almost three-quarters (72.7%) of the carers with whom contact was made and who were eligible to be interviewed agreed to participate and completed the interview. This is a relatively high response rate for a telephone survey, particularly given that the interview was quite long.

Figure 3.1 provides detailed information about the CATI interview process. Of the 4,935 names provided to the fieldwork company, only 2,126 were used. In order to reduce the possibility of participation bias, batches of new numbers were progressively released to the telephone interviewers only after maximum effort had been put into attempting to contact the original set of carers. Of these, 261 had unusable telephone numbers, most commonly because the number had been disconnected (n = 196). Other reasons for numbers being unusable included them being wrong numbers, being a fax or modem number, not being a residential number, or the selected person not being known at the number. A further 201 carers had a valid telephone number, but it was not possible to contact the carer on that number. This occurred for a variety of reasons, including that the number was engaged at all attempts, there was an answering machine and calls were not returned, there was no answer, or because the carer was contacted and an appointment was made to conduct the interview at a later date, but that appointment was not kept because the period allocated for conducting the CATI interviews had expired. Of the 1,664 people who were contacted, 284 were out of scope, mainly because they said they were not carers or they were away for the duration of the interviewing period. Of the 1,380 in-scope carers with whom contact was made, 1,003 (72.7%) completed an interview. The response rate was not significantly different between those receiving only Carer Allowance (72.3%) and those receiving Carer Payment and/or Carer Allowance (74.0%).18

Figure 3.1 The FCPDS computer-assisted telephone (CATI) interview process and results

Figure 3.1 The FCPDS computer-assisted telephone (CATI) interview process and results, described in text.

Of the remaining 377 carers who were contacted, 240 personally refused to participate, while a further 39 household members refused on the carers' behalf. There were a further 98 instances where it was not clear who refused - the carer or another household member.

Carers who declined to participate in the survey were asked their reasons for not wanting to participate. While not all of those who declined to participate gave reasons, many did. For carers, the most common reasons for refusal to participate were: not interested (39%), too busy (22%), or the survey was too long (8%) (Table 3.1). In some instances, the carer made no comment or just hung up (9%); this was the second most prevalent response from another household member who answered the phone (41%). A further 46% of members of the remaining carers' households said that the carer was not interested.

Table 3.1 Reasons for refusal to participate in the study
  Respondent (%) Other member of the household (%) Total (%)
Not interested
39
46
40
Too busy
22
5
20
No comment or just hung up
9
41
14
25 minutes is too long
8
3
7
Too personal or intrusive
3
0
3
Don't believe surveys are confidential or privacy concerns
3
0
3
Never do surveys
4
0
3
Don't like subject matter
2
0
1
Don't trust surveys or government
0
3
1
Get too many calls for surveys or telemarketing
1
0
1
Asked to be taken off list (add to Do Not Call Register)
1
0
1
Other reason
8
3
7
Total
100
100
100
Number of observations
240
39
279

Note: Percentages may not total 100% due to rounding.

Source: FCPDS 2006

Representativeness of the sample

One of the advantages of obtaining the sample from an administrative database is that it is possible to compare the characteristics of those who responded to the survey with those of the population from which the sample was obtained (that is, the total number of payment recipients in Australia). Table 3.2 presents data on the characteristics of the respondents and the population. The characteristics that can be examined are obviously limited to those contained in the administrative database. As can be seen from the percentages in Table 3.2, the characteristics of the respondents to the survey were very similar to the total population of carers who receive government assistance.19 Statistical tests also indicated that there were no significant differences for gender, state/territory and payment type considered; however, there were differences for the age of the carer. There were more carers aged 35 to 44 and 65 or more, and fewer carers aged 18 to 24 in the sample than in the population of payment recipients. In summary, with the exception of the age distribution of carers, the sample is representative of carers who receive Carer Allowance or Carer Payment for the characteristics examined.

Table 3.2 Characteristics of survey respondents and total population receiving Carer Allowance or Carer Payment

 

FCPDS sample (%) Population of carers (%)
Gender
Male
22.5
26.2
Female
77.5
73.8
Age
18-24 years
1.0
2.1
25-34 years
8.2
9.9
35-44 years
25.3
21.9
45-54 years
20.9
22.1
55-64 years
20.8
23.6
65 years or more
23.7
20.4
State
 
 
ACT
1.2
0.9
NSW
32.4
34.3
NT
0.5
0.5
QLD
21.5
19.5
SA
9.2
8.2
TAS
3.7
3.0
VIC
24.0
26.0
WA
7.6
7.7
Payment type
 
 
Only Carer Allowance
74.8
77.8
Carer Payment (including those on both payment types)
25.2
22.2
Number of observations
1,002
471,815

Notes: Statistical tests showed that carers' age was different (χ2 (5) = 23.01, p < .01), but not the proportion of male and female carers (χ2 (1) = 0.09, p > .01), their state/territory of origin (χ2 (7) = 8.70, p > .01), or the proportion in each payment type (χ2 (1) = 5.12, p > .01). Percentages may not total 100% due to rounding.

Sources: FCPDS 2006; population figures provided by Centrelink

The interview schedule

The survey20 comprised several modules focusing on a number of themes, including:

  • Disabling condition and caring role;
  • Care of person with disability;
  • Household composition and demographic characteristics;
  • Service use;
  • Support networks;
  • Family life events;
  • Family relationships;
  • Caregiver's general health and wellbeing;
  • About your social life;
  • About education;
  • About work; and
  • Financial issues.

Wherever possible, the questions were sourced from valid and reliable measures or from widely used surveys, such as the ABS 2003 Survey of Disability, Ageing and Carers or the Household Income and Labour Dynamics in Australia (HILDA) survey. The use of standardised measures ensured that the information gathered was valid and reliable. In addition, these measures have also been used in other large-scale surveys of the general population, such as the HILDA survey, thus enabling data collected in the FCPDS to be compared to the general population who do not have caring responsibilities. There are several instances in the empirical chapters of this report where data from carers in the FCPDS are compared with data from the general population from the HILDA survey.

This section provides a summary of the standardised measures used in the survey.

Mental health

The five-item Mental Health Inventory from the Medical Outcomes Study Short Form 36 (SF-36), a widely used measure of health-related quality of life (Ware, Kosinski, & Gandek, 2002), was used to assess the mental health status of carers. Scores range from 0 to 100, with higher scores indicating better mental health status. People who score less than 52 have been found to satisfy the clinical diagnostic criteria for depression and related disorders. The sensitivity and specificity of this cut-off has been found to be high (Berwick et al., 1991). We used this cut-off as an indicator of carers with clinical levels of depression and a continuous score as an indicator of the carers' mental health. The internal consistency of the scale was very good (α = .84).21

Carers were asked whether they or other household members had been depressed or sad (even if they felt OK sometimes) for 6 months or more. The question was adapted from one of the items from the Kemper three-item screening measure for depression, which has been found to be a valid indicator of clinical levels of depression in nationally representative samples (Kemper & Babonis, 1992; Lanzi, Pascoe, Keltner, & Ramey, 1999). Carers were also asked in what year the depressive episode occurred and for how long (one year, a period of consecutive years, or multiple non-consecutive years). The format of these questions enabled the timing of depressive episodes since the family started caring to be estimated.22

There are several issues that need to be taken into account when interpreting data collected from this set of questions. First, although the questions used can be seen as being indicative of prolonged depressive symptoms, it was not a standardised diagnostic interview, which is the gold standard in terms of measuring clinical levels of depression. There is only one Australian national epidemiological study that we are aware of that has employed standardised diagnostic interviews - the 1997 National Survey of Mental Health and Wellbeing (Andrews, Hall, Teesson, & Henderson, 1999) - largely because of the time and costs associated with conducting such interviews. Consequently, for the purposes of this study, carers or family members who have reported experiencing depression for 6 months or more should be considered to have been at risk of having (i.e., highly likely to have) suffered clinical levels of depression over this time period.

Second, inaccuracies can be introduced when people are asked to retrospectively recall their mental health status. However, studies that have examined recall bias indicate that it is more likely that people fail to recall previous depressive episodes rather than provide overestimates (Patten, 2003).

Third, carers were asked to indicate whether other family members had experienced a prolonged period of depression. Research examining the validity of proxy reports23 of the mental health of people with a disability, and more generally, have found that mental health problems tend to be under-reported by proxies because feelings are not directly observable (Ball, Russell, Seymour, Primrose, & Garrett, 2000; McPherson & Addington-Hall, 2003; Todorov & Kirchner, 2000). It is less likely for there to be under-reporting of depression in the current study, as family members' depressive symptoms needed to occur for 6 months or more for carers to report them. Previous research examining proxy reports of mental health problems has typically examined mental health problems in the last 4 weeks, so it may be that mental health problems that occur for a longer period of time would be less likely to go unnoticed by the proxy reporter. Nevertheless, there is still a possibility that carers may have under-reported family members' experiences. Consequently, the estimates of depression of family members in this report should be considered to be lower-bound estimates.

Vitality

The vitality or energy levels of carers was assessed using the Vitality subscale of the Medical Outcomes Study Short Form 36 (SF-36). Scores range from 0 to 100, with higher scores indicating more vitality or energy. A four-item measure, the Vitality subscale has been found to be a valid and reliable measure in the general population as well as in other populations (Ware et al., 2002). The internal consistency was .83 in the current sample.

General health

General health was measured by the question: "In general, would you say your health is excellent, very good, good, fair or poor?" The question was drawn from the Medical Outcomes Study Short Form 36 (SF-36). Self-rated health has been found to be highly predictive of subsequent morbidity and mortality, independent of other factors (e.g., Jenkinson & McGee, 1998). Research examining this measure has found that this question reflects the presence of acute and chronic physical health problems (Idler & Benyamini, 1997).

Relationship satisfaction

Relationship satisfaction of the carer with other members of their family was also assessed, using a series of items that have been used in the HILDA survey. The carers' satisfaction with their relationship with their partner, the person (or persons) with a disability, their children (not including the person with a disability), and their parents; the carers' partner's relationship with their children; and how the children got along with each other were all assessed. The relationship items have been widely used and have established validity. For example, low levels of relationship satisfaction with partners were associated with later marital separation (Heady, Warren, & Harding, 2006).

Increased arguments with partner

Carers were also asked: "Since you began caring for [name of person with disability], was there a period or periods in your life when the number of arguments with your spouse or partner increased significantly?" This item has been frequently used in standardised measures of stressful life events (Lin, Dean, & Ensel, 1986).

Separation

Carers were asked whether they had separated from their spouse or long-term partner since caring began.

Family functioning

Family functioning was assessed using a five-item scale. Three of these items were derived from the Family Relationship Inventory. Each of the three items used were developed to assess different dimensions of family functioning: family cohesion (the commitment, help and support family members provide for one another), expressiveness (acting openly and expressing feelings directly) and conflict (the open expression of anger and aggression) (Edwards & Clarke, 2005). To assess an additional dimension of family functioning - flexibility (the degree of flexibility within the family in dealing with new challenges) - an item from the Adaptability subscale of the Family Adaptability and Cohesion Scale IV was used (Olson, 1993). We developed a final item - "Our family copes with and recovers from times of hardship" - to assess family resilience. Carers were asked whether they strongly agreed, agreed, disagreed or strongly disagreed with each of the five items. For example, disagreeing with the statement: "There is a feeling of togetherness in our family" suggested low family cohesion, while agreeing with: "There is a lot of tension in our family" indicated family conflict. Previous studies of family functioning have indicated that problems in just one dimension of family functioning can have implications for mental health (Edwards & Clarke, 2005). For this study, we considered "family problems" to be where carers indicated they "disagreed" or "strongly disagreed" with that item (except for the conflict item, which was negatively worded, so we looked at where carers "agreed" or "strongly agreed" in that case). We also summed the number of family problems to indicate on how many of the five different dimensions of family functioning problems occurred.

Support needs

Support needs of the carer were assessed by the question: "Now thinking about the support you receive in providing care for [name of person with a disability] overall, do you feel that you need: a lot more support, a little more support or is the support you are getting now about right?" The question was adapted from a valid and reliable measure of social support - the Arizona Social Support Interview Schedule (Barrera, 1981).

Supportive behaviours

Supportive behaviours of family members and friends were assessed in some detail for the support they provided to carers and to the person with a disability. The support questions assessed the main areas of social support recognised in the wider social support literature - informational, emotional and practical assistance to the carer and the person with a disability (Barrera, 1981). With the assistance and comments from experts in social network measurement and analysis, Associate Professor Garry Robins and Dr Dean Lusher, the questions were tailored to be appropriate for measuring the social networks of carers and people with a disability (see also Robins, Pattison, Kalish & Lusher, 2007). Carers were asked whether there was someone they could turn to for informational support (advice about caring for the person with a disability), emotional support (availability to talk to about worries and fears in general and related to caring), practical support (help with everyday duties other than caring for the person with a disability, like going to the shops or doing housework), and financial support (extra money to assist in the caring role). Carers were also asked whether there was anyone other than them who provided the person with a disability with assistance with self-care, mobility or communication and, when appropriate, whether anyone else assisted the person with household management tasks like preparing meals, doing housework and shopping. The emotional support provided to the person with a disability by people other than the carer was also assessed.24

Care needs

Care needs of the person with the disability were assessed through a series of age-specific questions that were used in the ABS Survey of Disability, Ageing and Carers and the 2006 Census. Carers were asked two sets of three questions about the care needs of the person with a disability. The first set of questions asked whether the adult person with the disability ever needed someone to help with or be with them for self-care activities (e.g., eating, showering, dressing or toileting), body movement activities (e.g., getting out of bed, or moving around at home or at places away from home), and communication activities (e.g., understanding or being understood by others). The second set of questions asked whether the adult person with a disability needed someone to help with making friendships, interacting with others or maintaining relationships, coping with feelings or emotions, and making decisions or thinking through problems.

If the carer was caring for a child with a disability under the age of 7, then carers were asked whether the child needed assistance with self-care, body movement and communication activities more than other children his/her own age. For carers of children under the age of 15 (including children with a disability under the age of 6), instead of the second set of questions, carers were asked whether the child needed more care or help than other children his/her own age to interact or play with others, cope with feelings or emotions, and to manage his/her behaviour. Carers were given the following response options to the six care needs questions: "yes, always" (= 2); "yes, sometimes" (= 1); or "no" (= 0). The responses to the six questions were summed to form a scale score, with a possible range from 0 to 12. The internal consistency of the care needs scale was acceptable, with Cronbach's alpha ranging from .63 to .78,25 depending on whether the person cared for was a child aged 7-14 years or a child under the age of 7 years.

For each of the 120 carers who were caring for either two (n = 114) or three or more people (n = 12) with a disability, the data for the person with a disability who had the highest care needs were used in the statistical analyses. To ease interpretation in the statistical analyses, the scores on care needs were divided into three groups: low (a score of 0 to 4), medium (5 to 8) and high (9 to 12).

Type of disability

Type of disability was derived from a question to the carer about the primary medical diagnosis or disability of the person being cared for. Carers' responses to this question were categorised to form the five major categories used by the Australian Institute of Health and Welfare, using version 4 of the National Community Services Data Dictionary (National Community Services Data Committee, 2006). Two additional categories were also created. Consultations with experts in the field of disability indicated that often the person with the disability suffers from two or more medical or disabling conditions. Consequently, these were categorised as "multiple medical conditions". In some instances, the carers' responses did not provide sufficient information to categorise using the ABS definitions; in these cases we created an "unassigned" category. In this category, carers frequently said "old age" was the disabling condition and therefore it was unclear as to which area or areas the person with the disability had the impairment.

Other caregiving characteristics

Other caregiving characteristics included the number of people with a disability (one, two or three) who were cared for, hours of caring per week, years of caring, and caring for other people who were not disabled (e.g., children or ageing parents). As 120 carers were caring for either two (n = 114) or three or more people (n = 12) with a disability, the highest level of hours of caring per week and years of caring reported for any one of the people with a disability were used in the statistical analyses.

Financial hardship

Financial hardship was measured by asking participants whether their household had experienced four events. Respondents were asked: "Since January of this year, did any of the following happen to your family because of a shortage of money: could not pay electricity or the telephone bills on time; could not pay the mortgage or rent on time; pawned or sold something; asked for financial help from friends or family?" Originally included in the ABS 1998-1999 Household Expenditure Survey, these events were also included in the HILDA survey, along with three other events (the three other events were: "went without meals; unable to heat home; and asked for help from community organisations"). "Unable to heat home" was excluded as it was felt that this event was not an indicator of financial hardship in some of the warmer climates in Australia. "Asked for help from community organisations" was addressed in other areas of the survey and it was considered repetitive to include this event. "Went without meals" was the event with the lowest incidence of the remaining events in other surveys of income support recipients (after "unable to heat home" and "asked for help from community organisations" were excluded) and due to concerns about the length of the survey, this event was also excluded. Despite the reduction in the number of events, the four-item financial hardship scale had acceptable levels of internal consistency (α = .73).

Other socio-demographic variables

Other socio-demographic variables included the carer's gender, age and labour force status (employed, unemployed, not in the labour force); household income; and whether the individual was partnered. The number of adults and children in the household was taken into account when calculating household income. This is referred to as "equivalised household income", which is a method of adjusting household income by household size to give a more valid measure of material standard of living. We used the OECD equivalence scale, where one adult is given a score of 1.0, then other adults are given a score of .5 and children are given a score of .3. Household income is then divided by the sum of the individual scores. For example, a household of two adults and two children would have an equivalence score of 2.1 and its equivalised income would be total income (e.g., $100,000) divided by 2.1 ($47,619).

Statistical analyses for the report

When comparing groups of individuals (e.g., males and females) on average scores, tests of statistical significance were calculated using a 95% level of confidence (p < .05). The confidence intervals are displayed on the relevant graphs by means of vertical confidence interval bars. Non-overlapping confidence intervals on two columns in a graph suggest that we can be 95% confident that the two values represented in the columns are significantly different.

When comparing percentages across groups of individuals, we employed a statistical test referred to as a chi-square test (χ2). In the context of this report, the chi-square test tells us whether the distribution of frequencies in two or more groups are different. For example, we tested whether there was a difference in the percentage of males and females that had been depressed. In cases when the chi-square test is statistically significant, we indicate the confidence level at which they are significant at the 95% level (p < .05), 99% level (p < .01) or 99.9% level (p < .001).

Footnote

12. The Centrelink administrative list of 5,000 people did not include recipients whose telephone numbers were silent. Otherwise, all carers who were receiving Carers Payment and/or Carer Allowance were included.

13. The interviews were conducted by the Social Research Centre, Melbourne, in October and November 2006. Participation in the survey was voluntary.

14. A total of 1,002 interviews were provided in the data file. One carer fully responded, then later requested that the information provided be withdrawn.

15. A "soft" appointment was when someone indicated they wanted to participate, but advised the interviewer to call back later at an unspecified time.

16. For a copy of the survey protocol and response options, see the Institute's website at: www.aifs.gov.au/ institute/pubs/resreport16/report16pdf/survey.pdf

17. In this study, we focused on primary carers who were adults. However, we are aware that there are unique issues faced by children and young adults who have care responsibilities for adults or other family members (Williams, 1996).

18. χ2 (1) = 0.60, p >.05

19. People who receive Carer Payment may also be entitled to and often receive Carer Allowance. Therefore, the Carer Payment demographic data in Table 3.2 includes some carers who also received Carer Allowance.

20. A copy of the survey, showing the specific questions asked, can be found on the Institute's website at: www.aifs.gov.au/ institute/pubs/resreport16/report16pdf/survey.pdf

21. Cronbach's alpha is the average of all the correlations between items in a measure and, because it increases as the correlation between items increases, it is often used as a measure of the internal consistency of a scale.

22. In some instances where the person with a disability had a gradually deteriorating condition, it may have been difficult for the carer to recall when caring began.

23. Proxy reports occur when one survey respondent provides information about the experiences of another. In this case, the carer is reporting on the mental health of other members of their family.

24. The validity of proxy reports of tangible assistance from family and friends and social interaction with others has been supported by Conner, Conwell, and Duberstein (2001). However, proxies tended to under-report the level of emotional support provided.

25. See footnote 21 for an explanation of Cronbach's alpha.