Changing response rates from Māori and non-Māori in national sleep health surveys
View Article PDF
The Kaupapa M ori principle of equal explanatory power promotes the inclusion of equal numbers of M ori and non-M ori participants in health research. This concept has been developed to ensure M ori health needs are effectively addressed and understood at a comparable level to those of non-M ori.Several national sleep health surveys in New Zealand have integrated equal explanatory power into their study design.1-8 Incorporating this principle allows separate analysis of data for M ori and non-M ori with the same level of power, while also allowing M ori and non-M ori comparisons to be made.This programme of work, undertaken in partnership by the Sleep/Wake Research Centre and Te R 014dp 016b Rangahau Hauora a Eru P 014dmare, aims to produce accurate population prevalence estimates for sleep disorders and risk factors among M ori and non-M ori adults, and to examine predictive factors for sleep problems for both groups.Whilst the first national sleep survey in 1999 successfully achieved high response rates from both M ori and non-M ori (Table 1), the response rates for both samples have steadily decreased in subsequent surveys, despite very similar methods being used.In this paper we aim to identify the contextual changes and specific aspects of the questionnaires and research design that may be affecting this process. Table 1. Response rates for M ori descent and non-M ori samples in sleep population surveys 1999-2008 Survey topic Year Response rates M ori descent (%) non-M ori (%) Total (%) Obstructive sleep apnoea syndrome Insomnia Morningness/eveningness Circadian rhythm sleep disorders 1999 2001 2003 2008 70.6 58.2 53.4 48.1 78.9 76.2 58.2 61.6 74.4 72.5 55.7 54.3 Note: The denominator for calculating total response rates excluded: return to senders, people who had moved outside the target region and those who were identified as deceased. In addition, the 2003 survey excluded from the denominator those unable to answer the questionnaire due to physical or mental illness and in the 2008 survey, incorrect telephone numbers. Survey method In all surveys, the electoral roll was used as the sampling frame. Since it includes information on M ori descent and year of birth for all electors, it is possible to take a random sample of the population stratified by M ori descent and age. The electoral roll does not collect information on sex, so samples cannot be stratified by this variable, although it is presumed that samples from the electoral roll will contain approximately equal numbers of males and females. The survey method was developed in the first national sleep survey in 1999, which investigated the prevalence of symptoms and risk factors for Obstructive Sleep Apnoea Syndrome (OSAS) among M ori and non-M ori aged 30-59 years.5,8 It built on a regional pilot study that sampled 300 adults of M ori descent and 300 non-M ori which achieved an overall RR of 75% (M ori descent response rate=70%, non-M ori response rate= 80%, with the denominator excluding, people identified as deceased or no longer living in the target region). In the pilot study, the number of responders that identified as being in the M ori ethnic group as a proportion of responders of M ori descent was 91%. Since ethnicity using the M ori ethnic group was to be used to classify M ori and non-M ori during data analysis, the national OSAS survey included an age stratified random sample of 5500 electors of M ori descent and 4,500 non-M ori, in order to achieve approximately equal numbers of participants by ethnicity. The key elements of the first survey that facilitated high response rates included the use of a short questionnaire (two A4 sides), questions requiring primarily tick-box or single value answers, providing the opportunity to enter a draw to win an incentive prize, and the use of information sheets that were tailored to M ori and non-M ori samples (i.e. for the M ori sample appropriate salutations in te reo M ori, the use of Te R 014dp 016b Rangahau Hauora a Eru P 014dmare Centre letterhead, and identification of senior M ori researchers in the signature). The survey method also included intensive follow-up. After the initial mail-out of study packages, at approximately 2-weekly intervals a reminder postcard was sent to non-responders, and then a new study pack. Telematching was provided by a search service using name and address information provided on the electoral roll. Telephone follow-up was then undertaken where telephone numbers were available, with M ori interviewers available for M ori participants. Table 2. Differences between New Zealand sleep population surveys 1999-2008 Survey topic Obstructive sleep apnoea syndrome Insomnia Morningness eveningness Circadian rhythm sleep disorders Year of survey April 1999 April 2001 August 2003 October 2008 Sample size M ori descent 5500 2100 2674 5000 Non-M ori 4500 1900 2326 4100 Total 10,000 4000 5000 9100 Age range (10-year age groups) 30-59 years 20-59 years 30-49 years 20-59 years Location Nationwide Nationwide Wellington region Nationwide Questionnaire length (A4 pages) 1 00d7 double-sided 1 00d7 double-sided 2 00d7 double-sided 1.5 00d7 double-sided Question style Majority tick boxes Majority tick boxes Majority tick boxes ~50% tick boxes, ~50% written answers Collaboration with M ori Research Centre Yes. Led by M ori researcher. Questionnaire sent from TRRHaEPa Yes. Led by M ori researcher. Questionnaire sent from SWRC b No, but led by M ori researcher. Questionnaire sent from SWRC b No, but led by M ori researcher. Questionnaire sent from SWRC b Incentive prize Mystery holiday weekend Sleepyhead product Rimu mirror 5 00d7 iPod shuffles Telematches of published telephone numbers M ori descent 39.0% 31.5% 45.8%c 24.8% Non-M ori 50.0% 45.5% 60.8%c 36.9% Non-responder follow-up Postcard Postcard Postcard Postcard New study pack New study pack New study pack New study pack Telephone call Telephone call Telephone call Telephone call New study pack to M ori New study pack to sample with no telephone number M ori telephone interviewers Yes Yes Yes No Notes: a TRRHaEP = Te R 014dp 016b Rangahau Hauora a Eru P 014dmare; b SWRC = Sleep/Wake Research Centre; c The Wellington region is socioeconomically less deprived than other regions in New Zealand.10 Both the M ori descent and non-M ori samples in this study were less deprived than the national population surveyed in 1999. This may help to explain the high rate of telematching success for this survey. Table 2 describes the major differences in the four population-based sleep surveys including the sample sizes and age ranges, questionnaire style and length, increasing requirement for further follow-up, change in the leadership of studies from Te R 014dp 016b Rangahau Hauora a Eru P 014dmare to M ori health researchers at the Sleep/Wake Research Centre, and the lack of M ori interviewers during telephone follow-up for the final survey. Response rates in sleep surveys The response rates achieved for each survey are shown in Table 1. The response rate for M ori participants in each survey is consistently lower than that for non-M ori ( 03c72, p<0.0001) and has declined with each successive survey. M ori were also over-represented in the most deprived NZDep9deciles in all surveys, and response rates decreased with increasing socioeconomic deprivation (Cochrane-Armitage test for trend, p<0.0001) in each survey. Figure 1 demonstrates that the gradients of socioeconomic deprivation have not changed over time for either the M ori descent or non-M ori sample. Additionally, in each survey there was a significant trend for increasing response rates with each additional decade of age (Cochrane-Armitage test for trend, p<0.0001). Figure 1. Response rate in each NZDep decile for the M ori descent and non-M ori sample in 1999 and 2008, showing a trend of decreasing response with increasing socioeconomic deprivation The changes in response rate are most apparent between the initial survey in 1999 and the most recent survey in 2008. Table 3 compares the response rate at each stage of data collection for these two surveys. While telephone follow-up achieved a higher percentage of responses in the 2008 survey, the first mail-out elicited only about half as many responses as in the 1999 survey. Furthermore, the percentages of the sample who were ineligible or unable to be contacted had more than doubled between the two surveys. One factor contributing to this was the inclusion of 20-29 year olds in the 2008 survey, but not the 1999 survey. Figure 2 shows the breakdown of response rate by descent and age for the two surveys. In both the M ori and non-M ori samples, the percentage of respondents in each age group was lower in 2008 than in 1999. Table 3. Comparison of responses at each stage of data collection between the M ori descent and non-M ori samples (30-59 year age groups) in 1999 and 2008 in addition to the complete sample (20-59 year age groups) in 2008 Variables M ori descent sample non-M ori sample 1999 2008 1999 2008 Age 30-59 Age 30-59 Age 20-59 Age 30-59 Age 30-59 Age 20-59 % of 5,500 % of 3,750 % of 5,000 % of 4,500 % of 3,075 % of 4,100 Mail-out 1 30.4 16.1 14.1 43.2 23.6 21.3 Mail-out 2 14.0 11.6 10.4 14.5 16.0 14.5 Mail-out 3 15.9 6.7 6.1 13.2 10.5 9.6 Phone follow-up 6.1 10.7 10.0
Summary
Abstract
Aim
To understand declining response rates in New Zealand sleep health surveys by examining contextual changes and specific aspects of the questionnaires and research design that may have contributed.
Method
From 1999-2008, four population surveys were undertaken, seeking to recruit equal numbers of M ori and non-M ori, consistent with the Kaupapa M ori principle of equal explanatory power; using the electoral roll as a sampling frame and including extensive follow-up.
Results
In successive surveys, there were fewer respondents in all age groups. Response rates from M ori were lower in all surveys and the percentage decline was greater than for non-Maori. Between 1999 and 2008, the response rates from the initial mail-out decreased by 50% and the proportion of the sample that were uncontactable increased by 50%. Identified societal trends included decreased currency of electoral roll address information, declining use of listed landline telephone numbers, and possibly declining willingness to participate from increasing respondent burden. Contributing study design features may have included changes in M ori leadership, increasing complexity of questions and saliency of the research topic to potential participants.
Conclusion
The declining response rate in sleep population surveys is likely to be due to a number of factors. The pros and cons of using the electoral roll as a sampling frame in mail surveys should be carefully considered.
Author Information
Acknowledgements
Correspondence
Correspondence Email
Competing Interests
Robson B. Mana Whakam rama - Equal Explanatory Power: M ori and non-M ori sample size in national health surveys. Wellington (NZ): Te R 014dp 016b Rangahau Hauora a Eru P 014dmare, Wellington School of Medicine and Health Sciences, University of Otago for Public Health Intelligence, Ministry of Health, New Zealand; 2002 November.Paine SJ, Gander PH, Harris R, Reid P. Who reports insomnia? Relationships with age, sex, ethnicity, and socioeconomic deprivation. Sleep. 2004;27:1163-9.Paine SJ, Gander PH, Harris RB, Reid P. Prevalence and consequences of insomnia in New Zealand: disparities between M ori and non-M ori. Aust N Z J Public Health. 2005;29:22-8.Paine SJ, Gander PH, Travier N. The epidemiology of morningness/eveningness: influence of age, gender, ethnicity, and socioeconomic factors in adults (30-49 years). J Biol Rhythms. 2006;21:68-76.Mihaere KM, Harris R, Gander PH, et al. Obstructive sleep apnea in New Zealand adults: prevalence and risk factors among M ori and non-M ori. Sleep. 2009;32:949-56.Gander PH, Marshall NS, Harris R, Reid P. The Epworth Sleepiness Scale: influence of age, ethnicity, and socioeconomic deprivation. Epworth Sleepiness scores of adults in New Zealand. Sleep. 2005;28:249-53.Gander PH, Marshall NS, Harris RB, Reid P. Sleep, sleepiness and motor vehicle accidents: a national survey. Aust N Z J Public Health. 2005;29:16-21.Harris R. Obstructive sleep apnoea syndrome: symptoms and risk factors among M ori and non-M ori adults in Aotearoa. [Master of Public Health] Wellington (NZ): University of Otago, 2003.Salmond C, Crampton P, Atkinson J. NZDep2006 Index of Deprivation [report on the Internet]. Wellington (NZ): Department of Public Health, University of Otago; 2007 August.http://www.uow.otago.ac.nz/academic/dph/research/NZDep/NZDep2006%20research%20report%2004%20September%202007.pdfStatistics New Zealand. New Zealand: An Urban/Rural Profile [report on the Internet]. Wellington (NZ); 2004.http://www.stats.govt.nz/~/media/Statistics/Publications/Urban-Rural%20Profile/nz-urban-rural-profile-report.ashxElectoral Enrolment Centre [homepage on the Internet]. Wellington (NZ): 2008 [cited 10 August 2009]. Enrolment Statistics: Comparison of Estimated Eligible Voting Population to Enrolled Electors for the whole of New Zealand.http://www.elections.org.nz/ages/electorate_all_0809.htmlStatistics New Zealand. Household Economic Survey: Year ended 30 June 2007 [report on the Internet]. Wellington (NZ); 2007.http://www.stats.govt.nz/browse_for_stats/people_and_communities/Households/HouseholdEconomicSurvey_HOTPYeJun07/Commentary.aspxGray A, Sunteralingham R. Feasibility of using samples of telephone numbers for Tier 1 Official Statistical Household Surveys. Official Statistics Research Series. 2007;1. http://www.stats.govt.nz/sitecore/content/statisphere/Home/official-statistics-research/series.aspxKuusela V, Callegaro M, Vehovar V. The Influence of Mobile Telephones on Telephone Surveys. In: Lepkowski JM, Tucker C, Brick JM, et al (eds). Advances in Telephone Survey Methodology. Miami: John Wiley & Sons, pp. 87-112, 2008.Blumberg SJ, Luke JV. Wireless substitution: Early release of estimates from the National Health Interview Survey, July-December 2008 [report on the Internet]. Hyattsville (MD): National Centre for Health Statistics; 2009 May.http://www.cdc.gov/nchs/data/nhis/earlyrelease/wireless200905.pdfThe New Zealand Herald [homepage on the Internet]. Auckland (NZ): 2006 October 19 [cited 30 September 2009]. NZ Post introduces fees for redirecting mail. http://www.nzherald.co.nz/nz/news/article.cfm?c_id=1&objectid=10406669Howell B, Corbett L, Mishra V, Ryan L. Information and Communications Technologies in New Zealand: Nine Case Studies, Case Study 2: New Zealand Post Electoral Enrolment Centre [report on the Internet]. Wellington (NZ): New Zealand Institute for the Study of Competition and Regulation and Victoria Management School, Victoria University of Wellington for the Ministry of Economic Development; 2004 February 10.http://www.med.govt.nz/upload/2483/02-electoral-enrolment-centre.pdfElectoral Enrolment Centre [homepage on the Internet]. Wellington (NZ) [cited 30 September 2009]. Enrolment Update Campaign - FAQ.http://www.elections.org.nz/enrolment/enrol-faqs/update-faq.htmlGroves RM. Survey errors and survey costs. Hoboken (NJ): John Wiley & Sons, 2004.Rogers A, Murtaugh MA, Edwards S, Slattery ML. Contacting controls: Are we working harder for similar response rates, and does it make a difference? Am J Epidemiol. 2004;160:85-90.Groves RM, Couper MP. Nonresponse in household interview surveys. New York: John Wiley & Sons, 1998.Galea S, Tracy M. Participation rates in epidemiologic studies. Ann Epidemiol. 2007;17:643-53.Finn A, Gendall P, Hoek J. Two attempts to increase the response to a mail survey. Marketing Bulletin. 2004;15. http://marketing-bulletin.massey.ac.nz/V15/MB_V15_N1_Finn.pdfBrennan M. Survey participation and attitudes towards surveys in New Zealand. N Z J Business. 1991;13:72-95.Statistics New Zealand. Respondent load strategy for Statistics New Zealand: Strategies and initiatives for reducing respondent load [report on the Internet]. Wellington (NZ); 2008. http://www.stats.govt.nz/about_us/policies-and-guidelines/respondent-load-strategy.aspxPaine SJ. Towards a balanced and ethically responsible approach to understanding differences in sleep timing. [PhD thesis] Wellington (NZ): Massey University, 2006.Salmond C, Crampton P, Hales S, et al. Asthma prevalence and deprivation: a small area analysis. J Epidemiol Commun H. 1999;53:476-480.Devlin N, Hansen P, Herbison P. Variations in self-reported health status: results from a New Zealand survey. N Z Med J. 2000;113:517-20.North FM, Sharples K. Changes in the use of hormone replacement therapy in New Zealand from 1991-1997. N Z Med J. 2001;114:250-253.Grace VM, Zondervan KT. Chronic pelvic pain in New Zealand: prevalence, pain severity, diagnoses and use of the health services. Aust N Z J Public Health. 2004;28:369-75.Hoek J, Maubach N. Consumers' knowledge, perceptions, and responsiveness to direct-to-consumer advertising of prescription medicines. N Z Med J. 2007;120. http://www.nzmj.com/journal/120-1249/2425/content.pdfMetcalf P, Scott A. Using multiple frames in health surveys. Stat Med. 2009;28:1512-1523.Hox JJ, de Leeuw ED. A Comparison of nonresponse in mail, telephone, and face-to-face surveys: Applying multilevel modeling to meta-analysis. Qual Quant. 1994;28:329-344.
For the PDF of this article,
contact nzmj@nzma.org.nz
Changing response rates from Māori and non-Māori in national sleep health surveys
View Article PDF
The Kaupapa M ori principle of equal explanatory power promotes the inclusion of equal numbers of M ori and non-M ori participants in health research. This concept has been developed to ensure M ori health needs are effectively addressed and understood at a comparable level to those of non-M ori.Several national sleep health surveys in New Zealand have integrated equal explanatory power into their study design.1-8 Incorporating this principle allows separate analysis of data for M ori and non-M ori with the same level of power, while also allowing M ori and non-M ori comparisons to be made.This programme of work, undertaken in partnership by the Sleep/Wake Research Centre and Te R 014dp 016b Rangahau Hauora a Eru P 014dmare, aims to produce accurate population prevalence estimates for sleep disorders and risk factors among M ori and non-M ori adults, and to examine predictive factors for sleep problems for both groups.Whilst the first national sleep survey in 1999 successfully achieved high response rates from both M ori and non-M ori (Table 1), the response rates for both samples have steadily decreased in subsequent surveys, despite very similar methods being used.In this paper we aim to identify the contextual changes and specific aspects of the questionnaires and research design that may be affecting this process. Table 1. Response rates for M ori descent and non-M ori samples in sleep population surveys 1999-2008 Survey topic Year Response rates M ori descent (%) non-M ori (%) Total (%) Obstructive sleep apnoea syndrome Insomnia Morningness/eveningness Circadian rhythm sleep disorders 1999 2001 2003 2008 70.6 58.2 53.4 48.1 78.9 76.2 58.2 61.6 74.4 72.5 55.7 54.3 Note: The denominator for calculating total response rates excluded: return to senders, people who had moved outside the target region and those who were identified as deceased. In addition, the 2003 survey excluded from the denominator those unable to answer the questionnaire due to physical or mental illness and in the 2008 survey, incorrect telephone numbers. Survey method In all surveys, the electoral roll was used as the sampling frame. Since it includes information on M ori descent and year of birth for all electors, it is possible to take a random sample of the population stratified by M ori descent and age. The electoral roll does not collect information on sex, so samples cannot be stratified by this variable, although it is presumed that samples from the electoral roll will contain approximately equal numbers of males and females. The survey method was developed in the first national sleep survey in 1999, which investigated the prevalence of symptoms and risk factors for Obstructive Sleep Apnoea Syndrome (OSAS) among M ori and non-M ori aged 30-59 years.5,8 It built on a regional pilot study that sampled 300 adults of M ori descent and 300 non-M ori which achieved an overall RR of 75% (M ori descent response rate=70%, non-M ori response rate= 80%, with the denominator excluding, people identified as deceased or no longer living in the target region). In the pilot study, the number of responders that identified as being in the M ori ethnic group as a proportion of responders of M ori descent was 91%. Since ethnicity using the M ori ethnic group was to be used to classify M ori and non-M ori during data analysis, the national OSAS survey included an age stratified random sample of 5500 electors of M ori descent and 4,500 non-M ori, in order to achieve approximately equal numbers of participants by ethnicity. The key elements of the first survey that facilitated high response rates included the use of a short questionnaire (two A4 sides), questions requiring primarily tick-box or single value answers, providing the opportunity to enter a draw to win an incentive prize, and the use of information sheets that were tailored to M ori and non-M ori samples (i.e. for the M ori sample appropriate salutations in te reo M ori, the use of Te R 014dp 016b Rangahau Hauora a Eru P 014dmare Centre letterhead, and identification of senior M ori researchers in the signature). The survey method also included intensive follow-up. After the initial mail-out of study packages, at approximately 2-weekly intervals a reminder postcard was sent to non-responders, and then a new study pack. Telematching was provided by a search service using name and address information provided on the electoral roll. Telephone follow-up was then undertaken where telephone numbers were available, with M ori interviewers available for M ori participants. Table 2. Differences between New Zealand sleep population surveys 1999-2008 Survey topic Obstructive sleep apnoea syndrome Insomnia Morningness eveningness Circadian rhythm sleep disorders Year of survey April 1999 April 2001 August 2003 October 2008 Sample size M ori descent 5500 2100 2674 5000 Non-M ori 4500 1900 2326 4100 Total 10,000 4000 5000 9100 Age range (10-year age groups) 30-59 years 20-59 years 30-49 years 20-59 years Location Nationwide Nationwide Wellington region Nationwide Questionnaire length (A4 pages) 1 00d7 double-sided 1 00d7 double-sided 2 00d7 double-sided 1.5 00d7 double-sided Question style Majority tick boxes Majority tick boxes Majority tick boxes ~50% tick boxes, ~50% written answers Collaboration with M ori Research Centre Yes. Led by M ori researcher. Questionnaire sent from TRRHaEPa Yes. Led by M ori researcher. Questionnaire sent from SWRC b No, but led by M ori researcher. Questionnaire sent from SWRC b No, but led by M ori researcher. Questionnaire sent from SWRC b Incentive prize Mystery holiday weekend Sleepyhead product Rimu mirror 5 00d7 iPod shuffles Telematches of published telephone numbers M ori descent 39.0% 31.5% 45.8%c 24.8% Non-M ori 50.0% 45.5% 60.8%c 36.9% Non-responder follow-up Postcard Postcard Postcard Postcard New study pack New study pack New study pack New study pack Telephone call Telephone call Telephone call Telephone call New study pack to M ori New study pack to sample with no telephone number M ori telephone interviewers Yes Yes Yes No Notes: a TRRHaEP = Te R 014dp 016b Rangahau Hauora a Eru P 014dmare; b SWRC = Sleep/Wake Research Centre; c The Wellington region is socioeconomically less deprived than other regions in New Zealand.10 Both the M ori descent and non-M ori samples in this study were less deprived than the national population surveyed in 1999. This may help to explain the high rate of telematching success for this survey. Table 2 describes the major differences in the four population-based sleep surveys including the sample sizes and age ranges, questionnaire style and length, increasing requirement for further follow-up, change in the leadership of studies from Te R 014dp 016b Rangahau Hauora a Eru P 014dmare to M ori health researchers at the Sleep/Wake Research Centre, and the lack of M ori interviewers during telephone follow-up for the final survey. Response rates in sleep surveys The response rates achieved for each survey are shown in Table 1. The response rate for M ori participants in each survey is consistently lower than that for non-M ori ( 03c72, p<0.0001) and has declined with each successive survey. M ori were also over-represented in the most deprived NZDep9deciles in all surveys, and response rates decreased with increasing socioeconomic deprivation (Cochrane-Armitage test for trend, p<0.0001) in each survey. Figure 1 demonstrates that the gradients of socioeconomic deprivation have not changed over time for either the M ori descent or non-M ori sample. Additionally, in each survey there was a significant trend for increasing response rates with each additional decade of age (Cochrane-Armitage test for trend, p<0.0001). Figure 1. Response rate in each NZDep decile for the M ori descent and non-M ori sample in 1999 and 2008, showing a trend of decreasing response with increasing socioeconomic deprivation The changes in response rate are most apparent between the initial survey in 1999 and the most recent survey in 2008. Table 3 compares the response rate at each stage of data collection for these two surveys. While telephone follow-up achieved a higher percentage of responses in the 2008 survey, the first mail-out elicited only about half as many responses as in the 1999 survey. Furthermore, the percentages of the sample who were ineligible or unable to be contacted had more than doubled between the two surveys. One factor contributing to this was the inclusion of 20-29 year olds in the 2008 survey, but not the 1999 survey. Figure 2 shows the breakdown of response rate by descent and age for the two surveys. In both the M ori and non-M ori samples, the percentage of respondents in each age group was lower in 2008 than in 1999. Table 3. Comparison of responses at each stage of data collection between the M ori descent and non-M ori samples (30-59 year age groups) in 1999 and 2008 in addition to the complete sample (20-59 year age groups) in 2008 Variables M ori descent sample non-M ori sample 1999 2008 1999 2008 Age 30-59 Age 30-59 Age 20-59 Age 30-59 Age 30-59 Age 20-59 % of 5,500 % of 3,750 % of 5,000 % of 4,500 % of 3,075 % of 4,100 Mail-out 1 30.4 16.1 14.1 43.2 23.6 21.3 Mail-out 2 14.0 11.6 10.4 14.5 16.0 14.5 Mail-out 3 15.9 6.7 6.1 13.2 10.5 9.6 Phone follow-up 6.1 10.7 10.0
Summary
Abstract
Aim
To understand declining response rates in New Zealand sleep health surveys by examining contextual changes and specific aspects of the questionnaires and research design that may have contributed.
Method
From 1999-2008, four population surveys were undertaken, seeking to recruit equal numbers of M ori and non-M ori, consistent with the Kaupapa M ori principle of equal explanatory power; using the electoral roll as a sampling frame and including extensive follow-up.
Results
In successive surveys, there were fewer respondents in all age groups. Response rates from M ori were lower in all surveys and the percentage decline was greater than for non-Maori. Between 1999 and 2008, the response rates from the initial mail-out decreased by 50% and the proportion of the sample that were uncontactable increased by 50%. Identified societal trends included decreased currency of electoral roll address information, declining use of listed landline telephone numbers, and possibly declining willingness to participate from increasing respondent burden. Contributing study design features may have included changes in M ori leadership, increasing complexity of questions and saliency of the research topic to potential participants.
Conclusion
The declining response rate in sleep population surveys is likely to be due to a number of factors. The pros and cons of using the electoral roll as a sampling frame in mail surveys should be carefully considered.
Author Information
Acknowledgements
Correspondence
Correspondence Email
Competing Interests
Robson B. Mana Whakam rama - Equal Explanatory Power: M ori and non-M ori sample size in national health surveys. Wellington (NZ): Te R 014dp 016b Rangahau Hauora a Eru P 014dmare, Wellington School of Medicine and Health Sciences, University of Otago for Public Health Intelligence, Ministry of Health, New Zealand; 2002 November.Paine SJ, Gander PH, Harris R, Reid P. Who reports insomnia? Relationships with age, sex, ethnicity, and socioeconomic deprivation. Sleep. 2004;27:1163-9.Paine SJ, Gander PH, Harris RB, Reid P. Prevalence and consequences of insomnia in New Zealand: disparities between M ori and non-M ori. Aust N Z J Public Health. 2005;29:22-8.Paine SJ, Gander PH, Travier N. The epidemiology of morningness/eveningness: influence of age, gender, ethnicity, and socioeconomic factors in adults (30-49 years). J Biol Rhythms. 2006;21:68-76.Mihaere KM, Harris R, Gander PH, et al. Obstructive sleep apnea in New Zealand adults: prevalence and risk factors among M ori and non-M ori. Sleep. 2009;32:949-56.Gander PH, Marshall NS, Harris R, Reid P. The Epworth Sleepiness Scale: influence of age, ethnicity, and socioeconomic deprivation. Epworth Sleepiness scores of adults in New Zealand. Sleep. 2005;28:249-53.Gander PH, Marshall NS, Harris RB, Reid P. Sleep, sleepiness and motor vehicle accidents: a national survey. Aust N Z J Public Health. 2005;29:16-21.Harris R. Obstructive sleep apnoea syndrome: symptoms and risk factors among M ori and non-M ori adults in Aotearoa. [Master of Public Health] Wellington (NZ): University of Otago, 2003.Salmond C, Crampton P, Atkinson J. NZDep2006 Index of Deprivation [report on the Internet]. Wellington (NZ): Department of Public Health, University of Otago; 2007 August.http://www.uow.otago.ac.nz/academic/dph/research/NZDep/NZDep2006%20research%20report%2004%20September%202007.pdfStatistics New Zealand. New Zealand: An Urban/Rural Profile [report on the Internet]. Wellington (NZ); 2004.http://www.stats.govt.nz/~/media/Statistics/Publications/Urban-Rural%20Profile/nz-urban-rural-profile-report.ashxElectoral Enrolment Centre [homepage on the Internet]. Wellington (NZ): 2008 [cited 10 August 2009]. Enrolment Statistics: Comparison of Estimated Eligible Voting Population to Enrolled Electors for the whole of New Zealand.http://www.elections.org.nz/ages/electorate_all_0809.htmlStatistics New Zealand. Household Economic Survey: Year ended 30 June 2007 [report on the Internet]. Wellington (NZ); 2007.http://www.stats.govt.nz/browse_for_stats/people_and_communities/Households/HouseholdEconomicSurvey_HOTPYeJun07/Commentary.aspxGray A, Sunteralingham R. Feasibility of using samples of telephone numbers for Tier 1 Official Statistical Household Surveys. Official Statistics Research Series. 2007;1. http://www.stats.govt.nz/sitecore/content/statisphere/Home/official-statistics-research/series.aspxKuusela V, Callegaro M, Vehovar V. The Influence of Mobile Telephones on Telephone Surveys. In: Lepkowski JM, Tucker C, Brick JM, et al (eds). Advances in Telephone Survey Methodology. Miami: John Wiley & Sons, pp. 87-112, 2008.Blumberg SJ, Luke JV. Wireless substitution: Early release of estimates from the National Health Interview Survey, July-December 2008 [report on the Internet]. Hyattsville (MD): National Centre for Health Statistics; 2009 May.http://www.cdc.gov/nchs/data/nhis/earlyrelease/wireless200905.pdfThe New Zealand Herald [homepage on the Internet]. Auckland (NZ): 2006 October 19 [cited 30 September 2009]. NZ Post introduces fees for redirecting mail. http://www.nzherald.co.nz/nz/news/article.cfm?c_id=1&objectid=10406669Howell B, Corbett L, Mishra V, Ryan L. Information and Communications Technologies in New Zealand: Nine Case Studies, Case Study 2: New Zealand Post Electoral Enrolment Centre [report on the Internet]. Wellington (NZ): New Zealand Institute for the Study of Competition and Regulation and Victoria Management School, Victoria University of Wellington for the Ministry of Economic Development; 2004 February 10.http://www.med.govt.nz/upload/2483/02-electoral-enrolment-centre.pdfElectoral Enrolment Centre [homepage on the Internet]. Wellington (NZ) [cited 30 September 2009]. Enrolment Update Campaign - FAQ.http://www.elections.org.nz/enrolment/enrol-faqs/update-faq.htmlGroves RM. Survey errors and survey costs. Hoboken (NJ): John Wiley & Sons, 2004.Rogers A, Murtaugh MA, Edwards S, Slattery ML. Contacting controls: Are we working harder for similar response rates, and does it make a difference? Am J Epidemiol. 2004;160:85-90.Groves RM, Couper MP. Nonresponse in household interview surveys. New York: John Wiley & Sons, 1998.Galea S, Tracy M. Participation rates in epidemiologic studies. Ann Epidemiol. 2007;17:643-53.Finn A, Gendall P, Hoek J. Two attempts to increase the response to a mail survey. Marketing Bulletin. 2004;15. http://marketing-bulletin.massey.ac.nz/V15/MB_V15_N1_Finn.pdfBrennan M. Survey participation and attitudes towards surveys in New Zealand. N Z J Business. 1991;13:72-95.Statistics New Zealand. Respondent load strategy for Statistics New Zealand: Strategies and initiatives for reducing respondent load [report on the Internet]. Wellington (NZ); 2008. http://www.stats.govt.nz/about_us/policies-and-guidelines/respondent-load-strategy.aspxPaine SJ. Towards a balanced and ethically responsible approach to understanding differences in sleep timing. [PhD thesis] Wellington (NZ): Massey University, 2006.Salmond C, Crampton P, Hales S, et al. Asthma prevalence and deprivation: a small area analysis. J Epidemiol Commun H. 1999;53:476-480.Devlin N, Hansen P, Herbison P. Variations in self-reported health status: results from a New Zealand survey. N Z Med J. 2000;113:517-20.North FM, Sharples K. Changes in the use of hormone replacement therapy in New Zealand from 1991-1997. N Z Med J. 2001;114:250-253.Grace VM, Zondervan KT. Chronic pelvic pain in New Zealand: prevalence, pain severity, diagnoses and use of the health services. Aust N Z J Public Health. 2004;28:369-75.Hoek J, Maubach N. Consumers' knowledge, perceptions, and responsiveness to direct-to-consumer advertising of prescription medicines. N Z Med J. 2007;120. http://www.nzmj.com/journal/120-1249/2425/content.pdfMetcalf P, Scott A. Using multiple frames in health surveys. Stat Med. 2009;28:1512-1523.Hox JJ, de Leeuw ED. A Comparison of nonresponse in mail, telephone, and face-to-face surveys: Applying multilevel modeling to meta-analysis. Qual Quant. 1994;28:329-344.
For the PDF of this article,
contact nzmj@nzma.org.nz
Changing response rates from Māori and non-Māori in national sleep health surveys
View Article PDF
The Kaupapa M ori principle of equal explanatory power promotes the inclusion of equal numbers of M ori and non-M ori participants in health research. This concept has been developed to ensure M ori health needs are effectively addressed and understood at a comparable level to those of non-M ori.Several national sleep health surveys in New Zealand have integrated equal explanatory power into their study design.1-8 Incorporating this principle allows separate analysis of data for M ori and non-M ori with the same level of power, while also allowing M ori and non-M ori comparisons to be made.This programme of work, undertaken in partnership by the Sleep/Wake Research Centre and Te R 014dp 016b Rangahau Hauora a Eru P 014dmare, aims to produce accurate population prevalence estimates for sleep disorders and risk factors among M ori and non-M ori adults, and to examine predictive factors for sleep problems for both groups.Whilst the first national sleep survey in 1999 successfully achieved high response rates from both M ori and non-M ori (Table 1), the response rates for both samples have steadily decreased in subsequent surveys, despite very similar methods being used.In this paper we aim to identify the contextual changes and specific aspects of the questionnaires and research design that may be affecting this process. Table 1. Response rates for M ori descent and non-M ori samples in sleep population surveys 1999-2008 Survey topic Year Response rates M ori descent (%) non-M ori (%) Total (%) Obstructive sleep apnoea syndrome Insomnia Morningness/eveningness Circadian rhythm sleep disorders 1999 2001 2003 2008 70.6 58.2 53.4 48.1 78.9 76.2 58.2 61.6 74.4 72.5 55.7 54.3 Note: The denominator for calculating total response rates excluded: return to senders, people who had moved outside the target region and those who were identified as deceased. In addition, the 2003 survey excluded from the denominator those unable to answer the questionnaire due to physical or mental illness and in the 2008 survey, incorrect telephone numbers. Survey method In all surveys, the electoral roll was used as the sampling frame. Since it includes information on M ori descent and year of birth for all electors, it is possible to take a random sample of the population stratified by M ori descent and age. The electoral roll does not collect information on sex, so samples cannot be stratified by this variable, although it is presumed that samples from the electoral roll will contain approximately equal numbers of males and females. The survey method was developed in the first national sleep survey in 1999, which investigated the prevalence of symptoms and risk factors for Obstructive Sleep Apnoea Syndrome (OSAS) among M ori and non-M ori aged 30-59 years.5,8 It built on a regional pilot study that sampled 300 adults of M ori descent and 300 non-M ori which achieved an overall RR of 75% (M ori descent response rate=70%, non-M ori response rate= 80%, with the denominator excluding, people identified as deceased or no longer living in the target region). In the pilot study, the number of responders that identified as being in the M ori ethnic group as a proportion of responders of M ori descent was 91%. Since ethnicity using the M ori ethnic group was to be used to classify M ori and non-M ori during data analysis, the national OSAS survey included an age stratified random sample of 5500 electors of M ori descent and 4,500 non-M ori, in order to achieve approximately equal numbers of participants by ethnicity. The key elements of the first survey that facilitated high response rates included the use of a short questionnaire (two A4 sides), questions requiring primarily tick-box or single value answers, providing the opportunity to enter a draw to win an incentive prize, and the use of information sheets that were tailored to M ori and non-M ori samples (i.e. for the M ori sample appropriate salutations in te reo M ori, the use of Te R 014dp 016b Rangahau Hauora a Eru P 014dmare Centre letterhead, and identification of senior M ori researchers in the signature). The survey method also included intensive follow-up. After the initial mail-out of study packages, at approximately 2-weekly intervals a reminder postcard was sent to non-responders, and then a new study pack. Telematching was provided by a search service using name and address information provided on the electoral roll. Telephone follow-up was then undertaken where telephone numbers were available, with M ori interviewers available for M ori participants. Table 2. Differences between New Zealand sleep population surveys 1999-2008 Survey topic Obstructive sleep apnoea syndrome Insomnia Morningness eveningness Circadian rhythm sleep disorders Year of survey April 1999 April 2001 August 2003 October 2008 Sample size M ori descent 5500 2100 2674 5000 Non-M ori 4500 1900 2326 4100 Total 10,000 4000 5000 9100 Age range (10-year age groups) 30-59 years 20-59 years 30-49 years 20-59 years Location Nationwide Nationwide Wellington region Nationwide Questionnaire length (A4 pages) 1 00d7 double-sided 1 00d7 double-sided 2 00d7 double-sided 1.5 00d7 double-sided Question style Majority tick boxes Majority tick boxes Majority tick boxes ~50% tick boxes, ~50% written answers Collaboration with M ori Research Centre Yes. Led by M ori researcher. Questionnaire sent from TRRHaEPa Yes. Led by M ori researcher. Questionnaire sent from SWRC b No, but led by M ori researcher. Questionnaire sent from SWRC b No, but led by M ori researcher. Questionnaire sent from SWRC b Incentive prize Mystery holiday weekend Sleepyhead product Rimu mirror 5 00d7 iPod shuffles Telematches of published telephone numbers M ori descent 39.0% 31.5% 45.8%c 24.8% Non-M ori 50.0% 45.5% 60.8%c 36.9% Non-responder follow-up Postcard Postcard Postcard Postcard New study pack New study pack New study pack New study pack Telephone call Telephone call Telephone call Telephone call New study pack to M ori New study pack to sample with no telephone number M ori telephone interviewers Yes Yes Yes No Notes: a TRRHaEP = Te R 014dp 016b Rangahau Hauora a Eru P 014dmare; b SWRC = Sleep/Wake Research Centre; c The Wellington region is socioeconomically less deprived than other regions in New Zealand.10 Both the M ori descent and non-M ori samples in this study were less deprived than the national population surveyed in 1999. This may help to explain the high rate of telematching success for this survey. Table 2 describes the major differences in the four population-based sleep surveys including the sample sizes and age ranges, questionnaire style and length, increasing requirement for further follow-up, change in the leadership of studies from Te R 014dp 016b Rangahau Hauora a Eru P 014dmare to M ori health researchers at the Sleep/Wake Research Centre, and the lack of M ori interviewers during telephone follow-up for the final survey. Response rates in sleep surveys The response rates achieved for each survey are shown in Table 1. The response rate for M ori participants in each survey is consistently lower than that for non-M ori ( 03c72, p<0.0001) and has declined with each successive survey. M ori were also over-represented in the most deprived NZDep9deciles in all surveys, and response rates decreased with increasing socioeconomic deprivation (Cochrane-Armitage test for trend, p<0.0001) in each survey. Figure 1 demonstrates that the gradients of socioeconomic deprivation have not changed over time for either the M ori descent or non-M ori sample. Additionally, in each survey there was a significant trend for increasing response rates with each additional decade of age (Cochrane-Armitage test for trend, p<0.0001). Figure 1. Response rate in each NZDep decile for the M ori descent and non-M ori sample in 1999 and 2008, showing a trend of decreasing response with increasing socioeconomic deprivation The changes in response rate are most apparent between the initial survey in 1999 and the most recent survey in 2008. Table 3 compares the response rate at each stage of data collection for these two surveys. While telephone follow-up achieved a higher percentage of responses in the 2008 survey, the first mail-out elicited only about half as many responses as in the 1999 survey. Furthermore, the percentages of the sample who were ineligible or unable to be contacted had more than doubled between the two surveys. One factor contributing to this was the inclusion of 20-29 year olds in the 2008 survey, but not the 1999 survey. Figure 2 shows the breakdown of response rate by descent and age for the two surveys. In both the M ori and non-M ori samples, the percentage of respondents in each age group was lower in 2008 than in 1999. Table 3. Comparison of responses at each stage of data collection between the M ori descent and non-M ori samples (30-59 year age groups) in 1999 and 2008 in addition to the complete sample (20-59 year age groups) in 2008 Variables M ori descent sample non-M ori sample 1999 2008 1999 2008 Age 30-59 Age 30-59 Age 20-59 Age 30-59 Age 30-59 Age 20-59 % of 5,500 % of 3,750 % of 5,000 % of 4,500 % of 3,075 % of 4,100 Mail-out 1 30.4 16.1 14.1 43.2 23.6 21.3 Mail-out 2 14.0 11.6 10.4 14.5 16.0 14.5 Mail-out 3 15.9 6.7 6.1 13.2 10.5 9.6 Phone follow-up 6.1 10.7 10.0
Summary
Abstract
Aim
To understand declining response rates in New Zealand sleep health surveys by examining contextual changes and specific aspects of the questionnaires and research design that may have contributed.
Method
From 1999-2008, four population surveys were undertaken, seeking to recruit equal numbers of M ori and non-M ori, consistent with the Kaupapa M ori principle of equal explanatory power; using the electoral roll as a sampling frame and including extensive follow-up.
Results
In successive surveys, there were fewer respondents in all age groups. Response rates from M ori were lower in all surveys and the percentage decline was greater than for non-Maori. Between 1999 and 2008, the response rates from the initial mail-out decreased by 50% and the proportion of the sample that were uncontactable increased by 50%. Identified societal trends included decreased currency of electoral roll address information, declining use of listed landline telephone numbers, and possibly declining willingness to participate from increasing respondent burden. Contributing study design features may have included changes in M ori leadership, increasing complexity of questions and saliency of the research topic to potential participants.
Conclusion
The declining response rate in sleep population surveys is likely to be due to a number of factors. The pros and cons of using the electoral roll as a sampling frame in mail surveys should be carefully considered.
Author Information
Acknowledgements
Correspondence
Correspondence Email
Competing Interests
Robson B. Mana Whakam rama - Equal Explanatory Power: M ori and non-M ori sample size in national health surveys. Wellington (NZ): Te R 014dp 016b Rangahau Hauora a Eru P 014dmare, Wellington School of Medicine and Health Sciences, University of Otago for Public Health Intelligence, Ministry of Health, New Zealand; 2002 November.Paine SJ, Gander PH, Harris R, Reid P. Who reports insomnia? Relationships with age, sex, ethnicity, and socioeconomic deprivation. Sleep. 2004;27:1163-9.Paine SJ, Gander PH, Harris RB, Reid P. Prevalence and consequences of insomnia in New Zealand: disparities between M ori and non-M ori. Aust N Z J Public Health. 2005;29:22-8.Paine SJ, Gander PH, Travier N. The epidemiology of morningness/eveningness: influence of age, gender, ethnicity, and socioeconomic factors in adults (30-49 years). J Biol Rhythms. 2006;21:68-76.Mihaere KM, Harris R, Gander PH, et al. Obstructive sleep apnea in New Zealand adults: prevalence and risk factors among M ori and non-M ori. Sleep. 2009;32:949-56.Gander PH, Marshall NS, Harris R, Reid P. The Epworth Sleepiness Scale: influence of age, ethnicity, and socioeconomic deprivation. Epworth Sleepiness scores of adults in New Zealand. Sleep. 2005;28:249-53.Gander PH, Marshall NS, Harris RB, Reid P. Sleep, sleepiness and motor vehicle accidents: a national survey. Aust N Z J Public Health. 2005;29:16-21.Harris R. Obstructive sleep apnoea syndrome: symptoms and risk factors among M ori and non-M ori adults in Aotearoa. [Master of Public Health] Wellington (NZ): University of Otago, 2003.Salmond C, Crampton P, Atkinson J. NZDep2006 Index of Deprivation [report on the Internet]. Wellington (NZ): Department of Public Health, University of Otago; 2007 August.http://www.uow.otago.ac.nz/academic/dph/research/NZDep/NZDep2006%20research%20report%2004%20September%202007.pdfStatistics New Zealand. New Zealand: An Urban/Rural Profile [report on the Internet]. Wellington (NZ); 2004.http://www.stats.govt.nz/~/media/Statistics/Publications/Urban-Rural%20Profile/nz-urban-rural-profile-report.ashxElectoral Enrolment Centre [homepage on the Internet]. Wellington (NZ): 2008 [cited 10 August 2009]. Enrolment Statistics: Comparison of Estimated Eligible Voting Population to Enrolled Electors for the whole of New Zealand.http://www.elections.org.nz/ages/electorate_all_0809.htmlStatistics New Zealand. Household Economic Survey: Year ended 30 June 2007 [report on the Internet]. Wellington (NZ); 2007.http://www.stats.govt.nz/browse_for_stats/people_and_communities/Households/HouseholdEconomicSurvey_HOTPYeJun07/Commentary.aspxGray A, Sunteralingham R. Feasibility of using samples of telephone numbers for Tier 1 Official Statistical Household Surveys. Official Statistics Research Series. 2007;1. http://www.stats.govt.nz/sitecore/content/statisphere/Home/official-statistics-research/series.aspxKuusela V, Callegaro M, Vehovar V. The Influence of Mobile Telephones on Telephone Surveys. In: Lepkowski JM, Tucker C, Brick JM, et al (eds). Advances in Telephone Survey Methodology. Miami: John Wiley & Sons, pp. 87-112, 2008.Blumberg SJ, Luke JV. Wireless substitution: Early release of estimates from the National Health Interview Survey, July-December 2008 [report on the Internet]. Hyattsville (MD): National Centre for Health Statistics; 2009 May.http://www.cdc.gov/nchs/data/nhis/earlyrelease/wireless200905.pdfThe New Zealand Herald [homepage on the Internet]. Auckland (NZ): 2006 October 19 [cited 30 September 2009]. NZ Post introduces fees for redirecting mail. http://www.nzherald.co.nz/nz/news/article.cfm?c_id=1&objectid=10406669Howell B, Corbett L, Mishra V, Ryan L. Information and Communications Technologies in New Zealand: Nine Case Studies, Case Study 2: New Zealand Post Electoral Enrolment Centre [report on the Internet]. Wellington (NZ): New Zealand Institute for the Study of Competition and Regulation and Victoria Management School, Victoria University of Wellington for the Ministry of Economic Development; 2004 February 10.http://www.med.govt.nz/upload/2483/02-electoral-enrolment-centre.pdfElectoral Enrolment Centre [homepage on the Internet]. Wellington (NZ) [cited 30 September 2009]. Enrolment Update Campaign - FAQ.http://www.elections.org.nz/enrolment/enrol-faqs/update-faq.htmlGroves RM. Survey errors and survey costs. Hoboken (NJ): John Wiley & Sons, 2004.Rogers A, Murtaugh MA, Edwards S, Slattery ML. Contacting controls: Are we working harder for similar response rates, and does it make a difference? Am J Epidemiol. 2004;160:85-90.Groves RM, Couper MP. Nonresponse in household interview surveys. New York: John Wiley & Sons, 1998.Galea S, Tracy M. Participation rates in epidemiologic studies. Ann Epidemiol. 2007;17:643-53.Finn A, Gendall P, Hoek J. Two attempts to increase the response to a mail survey. Marketing Bulletin. 2004;15. http://marketing-bulletin.massey.ac.nz/V15/MB_V15_N1_Finn.pdfBrennan M. Survey participation and attitudes towards surveys in New Zealand. N Z J Business. 1991;13:72-95.Statistics New Zealand. Respondent load strategy for Statistics New Zealand: Strategies and initiatives for reducing respondent load [report on the Internet]. Wellington (NZ); 2008. http://www.stats.govt.nz/about_us/policies-and-guidelines/respondent-load-strategy.aspxPaine SJ. Towards a balanced and ethically responsible approach to understanding differences in sleep timing. [PhD thesis] Wellington (NZ): Massey University, 2006.Salmond C, Crampton P, Hales S, et al. Asthma prevalence and deprivation: a small area analysis. J Epidemiol Commun H. 1999;53:476-480.Devlin N, Hansen P, Herbison P. Variations in self-reported health status: results from a New Zealand survey. N Z Med J. 2000;113:517-20.North FM, Sharples K. Changes in the use of hormone replacement therapy in New Zealand from 1991-1997. N Z Med J. 2001;114:250-253.Grace VM, Zondervan KT. Chronic pelvic pain in New Zealand: prevalence, pain severity, diagnoses and use of the health services. Aust N Z J Public Health. 2004;28:369-75.Hoek J, Maubach N. Consumers' knowledge, perceptions, and responsiveness to direct-to-consumer advertising of prescription medicines. N Z Med J. 2007;120. http://www.nzmj.com/journal/120-1249/2425/content.pdfMetcalf P, Scott A. Using multiple frames in health surveys. Stat Med. 2009;28:1512-1523.Hox JJ, de Leeuw ED. A Comparison of nonresponse in mail, telephone, and face-to-face surveys: Applying multilevel modeling to meta-analysis. Qual Quant. 1994;28:329-344.
Contact diana@nzma.org.nz
for the PDF of this article
Changing response rates from Māori and non-Māori in national sleep health surveys
Subscriber Content
The full contents of this pages only available to subscribers.
Login, subscribe or email nzmj@nzma.org.nz to purchase this article.
This Issue
