ALSWH aims

Sample

Survey 1 (baseline) 1996

More than 40,000 women responded to the baseline surveys for the main cohorts in 1996. Because of uncertainties about the accuracy of the Medicare database (which was used as the sampling frame for the stratified random samples), response rates cannot be exactly specified. It is estimated that 41%-42%, 53-56%, and 37-40% of the 1973-78 cohort, the 1946-51 cohort and the 1921-26 cohort, respectively, responded to the initial invitation to participate. Confidentiality restrictions meant that the names of the selected women were unknown to researchers. Usual methods of encouraging participation such as by telephone could not be used. The response rates were pleasing given that the invitation included a request for women to participate in the longitudinal study for up to 20 years.

In light of these response rates, it is important to assess any response bias so that the generalizability of the study findings can be determined. A comparison of the demographic characteristics of respondents and non-respondents was not possible because privacy guidelines prevented the researchers from having any information about women who were selected to receive an invitation but did not respond. We were able, however, to obtain aggregate data for non-respondents’ use of health services (from the Australian Medicare database). These data suggest that there are small differences in use of health services among respondents and non-respondents, with non-respondents less likely, for example, to have visited a medical specialist in the last 2 years (1946-51 cohort, 49% versus 54%; 1921-26 cohort, 65% versus 72%). There was not a significant difference in health service use between respondents and non-respondents from the 1973-78 cohort.

A proportion of this difference may be explained by the fact that some women who were selected may no longer be living in Australia or may have died, as the Medicare database is not routinely linked to emigration records or the National Death Index in Australia.

Although we were not able to ascertain reasons for non-response (because we were not allowed to know any details about the selected women), we were able, through comparison with the 1996 census data, to confirm that the participants in each of the cohorts are reasonably representative of the general population of women of the same age in Australia (Table 1). There is some response bias in terms of overrepresentation of women with tertiary education and underrepresentation of some groups of immigrant women.

 Table 1 Socio-demographic characteristics

The above table and information are taken from:
Brown, W. J., Dobson, A. J., Bryson, L., & Byles, J. E. (1999). The Australian Longitudinal Study on Women’s Health: on the progress of the main cohort studies. Journal of Women's Health & Gender-Based Medicine, 8(5), 681-688.

Sample for the longitudinal study of the 1973-78, 1946-51 and 1921-26 cohorts

Some participants completed Survey 1 in 1996 and did not provide any contact details (532 women from the 1973-78 cohort, 383 women from the 1946-52 cohort and 508 women from the 1921-26 cohort). Hence the numbers of women actually enrolled in the Australian Longitudinal Study on Women's Health (ALSWH) were 14,247 women in the 1973-78 cohort, 13,716 women in the 1946-51 cohort and 12,432 women in the 1921-26 cohort (Lee, C., Dobson, A. J., Brown, W. J., Bryson, L., Byles, J., Warner-Smith, P., & Young, A. F. (2005). Cohort Profile: the Australian Longitudinal Study on Women's Health. International Journal of Epidemiology, 34(5), 987-991).

1973-78 cohort

Among the 1973-78 cohort, 69% responded to Survey 2 in 2000, 65% to Survey 3 in 2003, 67% to Survey 4 in 2006, 61% to Survey 5 in 2009 and 61% to Survey 6 in 2012 (See Table 2). This retention compares well with other surveys of this highly mobile age group. The major reason for non-response among the 1973-78 cohort was that the research team was unable to contact the women (between 21% and 28% of the cohort at subsequent surveys), despite using all possible methods of maintaining contact. Women in their twenties are characterised by high levels of mobility, change of surnames on marriage, often not having telephone listings and not being registered to vote and making extended trips outside Australia for work, education or recreation. Despite these losses, modelling has shown that there is no serious bias in estimates of associations between risk factors and health outcomes in longitudinal models (Powers, J., & Loxton, D. (2010). The Impact of Attrition in an 11-Year Prospective Longitudinal Study of Younger Women. Annals of Epidemiology, 20(4), 318-321).

Table 2: Participation and retention of the 1973-78 cohort

 

Survey 2

Survey 3

Survey 4

Survey 5

Survey 6

Age in years

22-27

25-30

28-33

31-36

34-39

Eligible at previous survey

14247

14116

13886

13557

13,337

Ineligible

         

Deceased between surveys

22

10

15

8

17

Frailty (e.g. intellectual disability)

3

6

4

3

1

Withdrawn before mailout survey date

106

213

311

209

188

Total ineligible

131

229

330

220

206

Eligible at current survey

14116

13887

13557

13337

13,131

Non-respondents

         

Withdrawn from the project

124

200

171

113

132

Contacted but did not return survey

1332

653

1371

1994

1454

Unable to contact participant

2972

3953

2870

3030

3535

Total non-respondents

4428

4806

4412

5137

5121

Respondents

         

Completed survey

9688

9081

9145

8200

8010

Retention rate as % eligible

68.6%

65.4%

67.5%

61.5%

61.0%


1946-51 cohort

Retention has been much higher among the 1946-51 cohort; 91% responded to Survey 2 in 1998, 84% to Survey 3 in 2001, 84% to Survey 4 in 2004, 84% to Survey 5 in 2007 and 82% responded to Survey 6 in 2010 (See Table 3). The major reasons for non-response among the 1946-51 cohort was that the research team was unable to contact the women (6% to 8% of eligible women between Survey 2 and Survey 6), and non-return of questionnaires by women who could be contacted (2% at Survey 2 and 7% to 9% of eligible women at subsequent surveys) Women in the 1946-51 cohort typically lead busy lives often working, as well as caring for parents and their children. The women who could not be contacted were more likely to be separated, divorced or widowed.

Table 3: Participation and retention of the 1946-51 cohort

 

Survey 2

Survey 3

Survey 4

Survey 5

Survey 6

Age in years

47-52

50-55

53-58

56-61

59-64

Eligible at previous survey

13715

13605

13310

12979

12694

Ineligible

         

Deceased between surveys

50

65

88

99

119

Frailty (e.g. dementia, stroke)

7

14

14

19

28

Withdrawn before mailout survey date

53

216

229

167

277

Total ineligible

110

295

331

285

424

Eligible at current survey

13605

13310

12978

12694

12270

Non-respondents

         

Withdrawn from the project

155

155

136

226

201

Contacted but did not return survey

254

998

886

995

1153

Unable to contact participant

858

931

1052

935

905

Total non-respondents

1268

2084

2074

2056

2259

Respondents

         

Completed survey

12338

11226

10905

10638

10011

Retention rate as % eligible

90.7%

84.3%

84.0%

83.8%

81.6%


1921-26 cohort

Of women from the 1921-26 cohort, 91% responded to Survey 2 in 1999, 85% to Survey 3 in 2002, 84% to Survey 4 in 2005, 79% to Survey 5 in 2008 and 77% to Survey 6 in 2011 (see Table 4). Among the 1921-26 cohort the major reason for non-response was non-return of the questionnaire (4% to 9% of eligible women at Surveys 2 to 5 and 17% at Survey 6), although up to 9% of participants could not be contacted. Non-respondent women tended to report poorer self-rated health at Survey 1 than respondents. The effects of these losses were evaluated in terms of losses due to death and non-death (Brilleman, S. L., Pachana, N. A., & Dobson, A. J. (2010). The impact of attrition on the representativeness of cohort studies of older people. BMC Medical Research Methodology, 10. doi: 7110.1186/1471-2288-10-71). Brilleman et al. concluded that non-death losses were potentially a greater source of bias than the effects of death.

Table 4: Participation and retention of the 1921-26 cohort

 

Survey 2

Survey 3

Survey 4

Survey 5

Survey 6

Age in years

73-78

76-81

79-84

82-87

85-90

Eligible at previous survey

12432

11537

10186

8531

7002

Ineligible

         

Deceased between surveys

529

569

770

869

914

Frailty (e.g. dementia, stroke)

106

265

379

322

563

Withdrawn before mailout survey date

260

517

506

338

263

Total ineligible

895

1351

1655

1529

1740

Eligible at current survey

11537

10186

8531

7001

5262

Non-respondents

         

Withdrawn from the project

313

384

269

159

228

Contacted but did not return survey

481

860

592

640

883

Unable to contact participant

309

294

511

642

96

Total non-respondents

 1103

1539

1373

1442

1207

Respondents

         

Completed survey

10434

8647

7158

5560

4055

Retention rate as % eligible

90.5%

84.9%

83.9%

79.4%

77.1%

Six-month follow up surveys of the 1921-26 cohort

From November 2011, shorter surveys containing a set of core questions were mailed to the 1921-26 cohort every six months after the return of the previous survey. Table 5 shows the numbers of eligible participants and respondents at end of each six month period. Response rates have been consistently above 60%.

Table 5: Participation in 6-month follow up surveys of the 1921-26 cohort

 

May 2012

Nov 2012

May 2013

Nov 2013

May2014

 

Deceased

4909

5281

5430

6061

6531

 

Withdrawn

2516 

2603

2854

2625

2467

 

Total ineligible

7425

7884

8284

8686

8998

 

Respondents

3431 

3299

2858

2486

2107

 

Total eligible

5007

4548

4148

3746

3434

 

Response rate

68.5%

72.5%

68.9%

66.4%

61.4%

 


Comparison of the cohorts with Australian Censuses in 2001, 2006 and 2011

Demographic characteristics (Indigenous status, country of birth, marital status, lone person household, education, and employment) of respondents at each of the surveys have been compared with those of women of the same age in the Australian population, using data from the Australian Census conducted closest to the survey (see (1) Powers, J. (2004) Comparison of the Australian Longitudinal Study on Women’s Health cohorts with women of the same age in the 2001 Census. Technical Report. Newcastle: ALSWH.  At: http://www.alswh.org.au/InfoData/data%20technical%20reports/WHAvsCensus2001Feb2004.pdf; (2) Australian Longitudinal Study on Women’s Health (2009). Comparison of the ALSWH cohorts with women of the same ages in the 2006 Census and the 2004/2005 National Health Survey. In: Technical Report No. 32. 111-142; (3) Australian Longitudinal Study on Women’s Health (2013). Comparison of the ALSWH cohorts with women of the same ages in the 2011 Census. In: Technical Report No. 36. 107-135. As was found previously, there was some over-representation of Australian-born women and of women with tertiary education. In addition, ALSWH participants were more likely to be married, be employed and work longer hours than women of the same age in the general population. Some of these differences will be due to differences in the questions asked and the sampling frame i.e. to be selected for ALSWH women must have a Medicare card.

1989-95 cohort

In 2012 and 2013, 17070 18-23 year old women were enrolled in the 1989-95 cohort. Women were mainly recruited using the internet and social media platforms. Consistent with the other cohorts, women were required to have a Medicare card. Women completed the online survey and provided consent to linkage of survey data with administrative databases such as Medicare.

Due to this different method of recruiting it was not possible to provide a response rate. Therefore it was necessary to compare the representativeness of the sample with women of the same age in the closest Australian Census (see Table 6). The 1989-95 cohort were broadly representative in terms of area of residence, State and Territory distribution, marital status and age distribution. Women with tertiary education were over-represented.

Table 6: Comparison of sociodemographic characteristics of the 1989-95 cohort of women aged 18 to 23 in 2013 with women of the same age range in the 2011 Australian Census

 

1989-95 cohort 2013 (n=17,010)

%

2011 Census (N=847,042)

%

Age

   
 

18

15.2

16.0

 

19

17.5

16.2

 

20

17.2

16.8

 

21

16.5

17.1

 

22

16.9

17.0

 

23

16.7

17.0

State/Territories

   
 

New South Wales

27.8

31.0

 

Victoria

24.0

25.4

 

Queensland

22.3

20.6

 

Western Australia

11.0

10.6

 

South Australia

7.6

7.3

 

Australian Capital Territory

3.3

2.0

 

Tasmania

2.9

2.1

 

Northern Territory

0.8

1.0

Area of residence

   
 

Major city

75.3

74.5

 

Inner regional

16.6

16.0

 

Outer regional

6.7

7.2

 

Remote

0.8

1.1

 

Very remote

Migratory/no usual address

0.3

N/A

0.9

0.3

 

Missing

0.3

-

Highest qualification

   
 

Less than Year 12

7.4

14.9

 

Year 12 or equivalent

43.0

46.1

 

Certificate / Diploma

25.9

21.7

 

University degree

22.5

9.4

 

Missing/not stated/inadequately described

1.2

7.8

Aboriginal or Torres Strait Islander origin*

   
 

No

97.3

91.9

 

Yes

2.7

3.4

 

Missing

0.0

4.7

Marital Status

   
 

Never married

95.5

94.5

 

Married

3.0

4.9

 

Separated/divorced/widowed

0.4

0.6

 

Missing

1.2

-

*n=14,909 as this question was only asked in a later version of the survey