Correspondence to: Sarah Fox, Centre for Eye Research Australia, Locked Bag 8, East Melbourne, Victoria 8002. Fax: (03) 9662 3859; e-mail: firstname.lastname@example.org
Objective: To review the process of sample selection and highlight the methodological difficulties encountered during a nationwide survey of Indigenous Australians, to determine the prevalence and causes of vision impairment and evaluate access to and utilisation of eye care services.
Methods: Using a multi-stage, random cluster sampling methodology, 30 geographic areas stratified by remoteness, were selected to provide a representative population of approximately 3,000 Indigenous Australians aged 5–15 and 40 years and older, and a small non-Indigenous sample in selected remote areas. Recruitment was adapted to local conditions. The rapid assessment methodology included a questionnaire, tests of visual acuity, trachoma grading, frequency doubling perimetry and non-mydriatic fundus photography.
Results: The number of people examined was 2883/3662 (78.7%) Indigenous and 136 (83.4%) non-Indigenous. The percentage of the expected population who were enumerated during the survey varied; discrepancies were largest in urban areas (34.5%) compared to very remote areas (97.1%).
Conclusions: The unexpected variation in predicted population numbers and participation rates could be explained in part by local circumstances, degree of urbanisation, interpretation of the definition of ‘Indigenous’ and time constraints.
Implications: For successful recruitment, a community-specific approach is essential, including collaboration with local organisations and liaison with health workers of each gender.
Eye-related problems were the most frequently reported long-term health condition in Indigenous Australians (30%) in the 2004/05 National Aboriginal and Torres Strait Islander Health Survey.1 Diabetes and diabetic eye disease are increasingly common in Indigenous Australians, and high rates of refractive error and ocular trauma, with pockets of trachoma and blinding cataract, are still reported.1–2 The last comprehensive nationwide data on the prevalence and causes of vision impairment in Indigenous Australian communities were collected in the 1970s during the National Trachoma and Eye Health Program.3 Since then, intermittent reports have investigated specific problems such as trachoma. A current, comprehensive evaluation of eye health in Indigenous Australians is needed4 to plan and prioritise effective eye care services.
The National Indigenous Eye Health Survey (NIEHS) aimed to determine the prevalence and causes of vision impairment in Aboriginal and Torres Strait Islander peoples; to evaluate the access to and utilisation of eye care services in Indigenous communities; and to provide the evidence base for the planning of effective eye care programs.
This paper reviews the process of sample selection and highlights the methodological difficulties encountered.
The NIEHS fieldwork was conducted between January and December 2008.
The survey was designed to collect a representative sample of approximately 3,000 Indigenous Australians aged 5–15 and 40 years and older. A pilot study validated the survey protocol and established the target age ranges, finding the majority of vision impairment in adults aged over 40, and that examination of school-age children would be representative of the eye health status of Indigenous children.5
The June 2006 Census conducted by the Australian Bureau of Statistics (ABS) estimated the total residential Indigenous population to be 517,043.6 The NIEHS site selection was based on geographic and population data from the 2006 Census and the Accessibility and Remoteness Index of Australia (ARIA), which measures remoteness based on the physical distance to the nearest urban centre.7 Using a multi-stage, random cluster sampling methodology, 30 geographic areas containing approximately 300 Indigenous people, stratified by remoteness, were selected.
Sample size calculation
The sample size was determined after considering the prevalence of vision impairment, design effect, likely response rate, and the relative proportion of children and adults in communities.
The sample size was calculated to detect a doubling of the prevalence of presenting vision impairment compared to that reported in other Australians (a=0.05, b=0.20) after adjusting for the design effect (1.5). In mainstream Australia, rates of vision impairment have been reported as 4.2% in the better eye of adults,8 and as 3.1% and 5.0% in the worse eye of children aged 6 and 12 years, respectively.9,10 This gave a sample size of 1,631 adults and 1,281 children.
In Indigenous communities, the proportions of children and adults are approximately equal. A target cluster sample of 300 (range 200–400) was selected to contain 82 children aged 5 to 15 years (27%) and 67 adults aged 40 years and older (22%).6 This gave a prospective sample of 2,460 children and 2,010 adults. Allowing for 75% compliance, the final predicted sample sizes were 1,845 children and 1,507 adults. The minimum cluster sample was set at 200, containing 54 children (27%), which accorded with the WHO recommended sample size for the assessment of trachoma.11
The ABS organises the country into Australian Indigenous Geographic Classification areas, comprising Indigenous Areas (IARE) and smaller Census Collection Districts (CCD). IARE were stratified into five ARIA Remoteness Areas: Major City, Inner Regional, Outer Regional, Remote and Very Remote. To give better geographic representation by over-sampling, the Very Remote stratum was further divided into Very Remote Inland and Very Remote Coastal. Very Remote Inland was defined as a distance ≥100 km from the nearest coastline, and Very Remote Coastal as <100 km from the nearest coastline.
IARE were sorted by strata, and those with an Indigenous population of <200 were combined with contiguous IARE. Thirty sites were required to obtain the sample size, and we selected five primary sites plus 1 alternate site per stratum (i.e. six sites per stratum). The total Indigenous population per stratum was summated and divided by six to give the sampling fraction. A random number between 0 and 1 was multiplied by the sampling fraction to give a seed value (n) (Table 1). Sites that contained the nth, n2th, …n6th person were selected (Table 2, Figure 1), totalling 30 primary and six alternate sites. This method ensured random selection and controlled for probability proportional to population size.
Table 1. Sampling fraction for the NIEHS site selection.
Remoteness area description
Very Remote Inland
Very Remote Coastal
Table 2. NIEHS Sites: selected Indigenous areas and sub-selected census collection districts.
Using Global Mapper Software v9.03 (Parker, CO USA, 2001) and ABS digital boundary data, each IARE was sub-sampled at the level of Statistical Local Area (SLA) or CCD. These were grouped into contiguous CCD or SLA to give clusters of approximately 300 people. A cluster was randomly selected for the survey site and mapped using CCD or SLA boundaries; the specific survey areas can be found at http://www.cera.org.au/niehs.
In some remote areas, a single CCD contained more than 300 people. After consulting with health centres and community councils, the CCD was subdivided into clusters of 300 people, and one was randomly selected for the survey.
Following data collection in the first three sites (Moe, Tamworth, Doonside), the observed sample was considerably less than expected. To ensure an adequate sample size, contiguous CCDs were added to the remaining urban and regional survey areas to bring the sample size to 400 (range 350 to 450). Very Remote and some Remote strata did not require sample size adjustments.
Data exist for urban and rural non-Indigenous Australians,6,12 so a small sub-sample was defined in Remote and Very Remote areas based on the non-Indigenous population. If the original IARE did not contain a sufficient number of non-Indigenous people, the site was omitted from this aspect of the study (Table 2).
Recognising the great diversity of Indigenous Australian communities, we adapted recruitment strategies to local conditions for a community-specific approach. Each community was consulted in advance to gain permission and establish support for the survey, facilitate relationships and guide the recruitment process.
Local health services were contacted, including Aboriginal Community Controlled Health Services (ACCHS), community health services, medical centres and hospitals. Community councils, local schools, aged care facilities and other Aboriginal and Torres Strait Islander organisations were also consulted. Aboriginal Health Workers (AHW), community elders, Eye Health Coordinators and Aboriginal Liaison Officers (ALO) were engaged to facilitate participation, community acceptance, provide translation and ensure cultural appropriateness. They were offered casual employment and training opportunities, and community vehicles with local drivers were employed to transport participants. Local or visiting optometrists and ophthalmologists were consulted to assist with recruitment and referrals, which fostered relationships with community members and ensured continuum of care.
Children and adults who resided in the community during the week of fieldwork were considered eligible. When possible, lists were obtained from health services, schools and community councils. Organisations were understandably reluctant to share lists for privacy reasons, but researchers indirectly accessed lists through AHWs, ALOs and Eye Health Coordinators. We performed an informal census by combining these lists, door knocking and utilising community networks. A complete door-to-door private census was beyond the means of this program and was not undertaken.
Multiple recruitment methods were employed with varying success. ACCHS and health centres sent flyers and made phone calls to invite all eligible persons to participate. Eligibility was based on age, address and Indigenous status. This method was subject to the accuracy of the health centre's database. Media releases were sent to local newspapers, radio and television stations, and during the survey week, health workers door-knocked. Family networks and word of mouth were the most effective recruitment methods. Participation typically increased towards the end of a survey week. In each state, we obtained permission from the Department of Education Regional Offices and Catholic Education Boards to distribute information to parents, and either examine children at school or transport them from schools to the survey venue. A letter, information sheet, flyer, consent form and survey form were sent home and returned to teachers or ALOs.
Written, informed consent was obtained from each participant prior to examination. Everyone completed a questionnaire followed by a standardised rapid-assessment eye examination (Table 3, Figure 2). If any abnormality was detected during the screening process, participants were referred to the GP, optometrist or ophthalmologist. Results were explained to each participant, and community reports were compiled at the end of data collection.
Table 3. NIEHS examination materials and methods.
Demographic, medical and eye health history
Adult NIEHS Questionnaire
Self-administered or interview.
Questionnaires were adapted from the Vision Initiative.5,13–15
Child NIEHS Questionnaire
Presenting Visual Acuity (VA)
WHO Low Vision Kit16
Three metre simplified E chart.
Vision impairment was defined as VA <6/12 in the better eye.13–16
Pinhole and best-corrected VA
WHO Low Vision Kit
VA assessed with pinhole, and if indicated, refraction and correction.
Participants with VA <6/12 were considered as having vision impairment.15
Righton Retinomax 3 Hand-held Auto Refractor (Tokyo, Japan)
Magnon Trial Lenses – 134M f/a (Japan)
Near Visual Acuity
WHO Low Vision Kit
Presenting near vision measured; if indicated free reading glasses were given to correct to N8.
Both tarsal plates were everted and graded clinically; the left was photographed and graded retrospectively.
Clinical grading and photograding according to the WHO Simplified Grading System.17–18
Nikon D40X digital camera with AF Nikkor 85mm lens (Tokyo, Japan)
Welch Allyn/ Humphrey Zeiss Frequency Doubling Technology (San Leandro, CA USA)
C-20–1 screening mode.
A diagnosis of glaucoma was made on the basis of abnormal fields (2 points missed on FDT)19–21 and an abnormal cup-to-disc ratio on fundal photographs.15,22
Canon CR-DGi Non-mydriatic retinal camera (Tokyo, Japan)
Two fundus photographs of each eye.
Images were evaluated by a blinded clinician15 for the evidence of glaucoma,22 age-related macular degeneration23 and diabetic retinopathy.20
Canon CR-DGi Non-mydriatic fundus camera (Tokyo, Japan)
Anteriorly focused lens photographs were taken in retro-illumination.
Images were evaluated by a blinded clinician15 for the presence of cataract.25
Impact of Vision Impairment (IVI) on activities of daily life
Self-administered or interview.
The IVI assesses the rehabilitation needs of people with low vision.26 This version is currently being developed for use with Indigenous Australians.
The ethical process for the NIEHS has been described in detail elsewhere.27,28 Primary ethical approval was obtained from the Royal Victorian Eye and Ear Hospital. The survey protocol was formally approved by the following Human Research Ethics Committees: Aboriginal Health and Medical Research Council of NSW, Aboriginal Health Council of South Australia, Menzies School of Health Research, Central Australia HREC, Western Australian Aboriginal Health Information and Ethics Committee, ACT Health, Queensland Aboriginal and Islander Health Council, Tasmania Health and Medical HREC and Tasmanian Scientific Research Advisory Committee. This research was conducted in accordance with the tenets of the Declaration of Helsinki as revised in 2000.
The number of eligible people examined was 2,883 (78.7%) Indigenous and 136 (83.4%) non-Indigenous (Figure 3). Child participation rates were generally higher overall (84.4%) compared to adult rates (71.8%) due to assistance from the schools. The ratio of male to female participation was 466 to 722 (0.65) for Indigenous adults, 849 to 830 (1.02) for children, and 52 to 84 (0.62) for non-Indigenous adults. Participation rates increased as the proportion of Indigenous people per total population increased (Table 4).
Table 4. Comparison of the mean participation rate with the mean proportion of the population who were the target NIEHS sample.
Mean percentage of Indigenous population examined
Mean percentage of total population who were Indigenous
Mean percentage of Non-Indigenous population examined
Mean percentage of total population who were target Non-Indigenous population
Very Remote Inland
Very Remote Coastal
Barriers to recruitment
Barriers to recruitment were identified through observation and feedback from community members.
A high turnover of AHWs, nurses and schoolteachers was observed nationwide. Seventeen sites (57%) had established Aboriginal Eye Health Coordinators who assisted with the survey. Community residents were employed as health workers in 11 locations (37%) where there were no AHWs or insufficient clinic staff to support the survey. Three communities had new AHWs (<6 months experience), which hindered recruitment. In seven sites (23%), our primary contact person either left the community just prior to the survey week or failed to attend the survey site.
In 12 sites (40%), AHWs and the community questioned the random sampling methodology, thus many ineligible people were examined. We were frequently advised that the majority of the Indigenous community lived outside the survey area. In five urban sites (17%), this resulted from increased land prices and government housing relocation.
Ten sites (33%) had no existing ACCHS within or near the survey area. Alternate, culturally appropriate venues were sought, including schools, football clubs, churches and community centres. In seven locations (23%), the Indigenous community was divided, and in eight sites (27%), the primary health centre supporting the survey was not utilised by the entire community, so multiple venues were required. It was not possible to work past 5 pm in nine sites (30%), resulting in low participation by working people; this mainly affected urban areas and mining towns.
Examination in schools was not possible due to holidays in two sites, and long processing times for permission from central Department of Education offices (7 sites). In remote areas, many older children attended boarding school elsewhere, resulting in lower participation by children aged 13–15 years (9 sites) (Figure 4).
The transient nature of Indigenous peoples influenced sample sizes, particularly in the remote areas. Community numbers were low in 10 sites (33%) due to concurrent funerals or land rights meetings. Substance abuse occasionally affected participation rates (6 sites). The Australian Government Intervention in the Northern Territory challenged recruitment, due to recent eye examinations and the sentiment that there had been too many visitors to the community and too many examinations.
In 14 sites (47%), the survey team either could not access a formalised community list or none existed. Seven sites encompassed a large area (range 1,346–115,884 km2). In 10 urban sites, the community was highly dispersed, meaning that door knocking was not a feasible method of identifying participants. In seven of these 10 areas, no formalised community network existed.
Sixteen communities (53%) advised that participation would have increased if the site coordinator had arrived the week before the survey to build relationships and compile the sample list. This would have been especially helpful in sites where community members used a number of different primary health care providers, as we were not aware of this until arrival. A survey team member was able to visit 14 sites prior to the survey, and in nine sites we had indirect contact with the community via visits by the ophthalmologists or Eye Health Coordinators.
Practical and logistical difficulties are commonly reported in Indigenous research, arising from differing cultural attitudes and priorities between researchers and community members.29 Despite carefully planned sampling methodology and community consultation, the NIEHS met unexpected local challenges.
Due to local circumstances, data collection on the non-Indigenous population in Mount Isa, Esperance and Ashburton was not possible, and participation rates in the Outer Regional stratum were low. The Port Augusta ACCHS had a well-established visiting optometrist and ophthalmologist program. As a result of recent eye examinations, half of the adult population was unwilling to participate in the survey. Geographically, the Upper Murray site was large (11,739 km2), and participation was subject to time constraints. The survey team was scheduled to visit a different location within the site each day, so people with scheduling conflicts or who cancelled appointments did not participate. In Cherbourg, the selected CCD was sub-sampled to contain approximately 300 Indigenous people. The community as a whole responded enthusiastically to the survey, though most adults who self-presented came from outside of the sample area. A revisit was attempted but was not possible due to the restructuring of the ACCHS.
The survey venue type and location directly impacted participation, and guidance from multiple community members regarding culturally appropriate venues was important for maximum recruitment. In a study of young, Indigenous peoples’ health with the Victorian Aboriginal Health Service (VAHS), Holmes et al (2002) found that collaboration with a community controlled organisation caused the community to be more eager to participate. In the NIEHS, collaboration with local organisations elicited a similar response, but collaboration with a single organisation sometimes excluded some of the community. Multiple venues were required in most of the urban areas and some of the remote areas. In locations where the health centre was outside of the survey area, it was critical to find a convenient location within the survey area, as participants were unwilling to travel long distances (30 minutes) even if transportation was provided. The importance of a central examination site was also illustrated in the Melbourne Visual Impairment Project,30 TVI14 and the Visual Acuity Impairment Survey in the United States.31
Participation generally snowballed towards the end of the survey week as word spread and family and friends were encouraged to participate. Extensive family networks assisted in spreading the word, and it often required a physical presence to engage with them. In some areas, recruitment efforts could not begin until the start of the survey week for local reasons. Accurate sample lists required a physical presence, as every participant was asked about family and friends. Many communities advised that recruitment would have improved if the site coordinator had arrived early to engage with the community and build trust. It was not logistically possible to spend an extra week in each site, as the fieldwork in the 30 sites was completed in 42 weeks.
The benefit of early arrival was demonstrated in the Swan survey site, where the Indigenous community was widely dispersed. During the initial visit, equipment was lost in transit, so only children were examined. We learned that the community utilised multiple health services, as is the norm in urban settings, thus multiple venues were required. We used the initial visit to establish relationships with the extensive community networks and create a sample list. During a second visit, the remaining sample was examined quickly and effectively.
Recruitment in urban areas
Researching Indigenous peoples in urban areas is hindered by the lack of a sampling frame from which to draw participants.29,32,33 Indigenous communities are often ‘hidden’ within the urban context. They are characterised by “a heterogeneous network of family relations and organisational memberships geographically dispersed and intermixed within a much larger non-Indigenous population, with a constant inflow and outflow between the urban area and other sites, and frequent changes of address for many who stay within the urban area” (Scrimgeour, Scrimgeour 2007, p. 51). These factors make sample size estimates difficult.
NIEHS recruitment was easiest in remote areas with discrete Indigenous communities. We found the ABS data to be increasingly accurate the more remote we travelled, despite reports that ABS data tends to under-report the Indigenous population particularly in remote areas.34 As the proportion of the Indigenous population decreased, accuracy in our informal census declined due to outdated ACCHS contact lists and lack of organised community networks. Similar difficulties were reported in the VAHS study29 and the Western Australia Aboriginal Child Health Survey (WAACHS).35
These challenges were illustrated in particular by the City of Monash NIEHS site. Only 21 Indigenous people were identified as living within the survey area, following consultation with Bunurong Aboriginal Health Service, VAHS, the housing authority, the Regional Aboriginal Education Officer, Monash University, Monash Medical Centre and numerous Aboriginal people who lived in the region but outside of the sample area. Only six of the 21 were confirmed as still living in the survey area.
The definition of Indigenous
Similar to the NIEHS, WAACHS also reported that selected CCDs yielded significantly fewer observed than expected families in urban areas. This was attributed to population mobility, and investigations of public housing stocks showed that 12% to 15% of the target sample had moved to CCDs outside the survey area.35 This raised questions about the use of ABS data for urban sampling in Indigenous research.
The 2006 Census question about Indigenous status specifically addressed descent: “Is the person of Aboriginal or Torres Strait Islander origin?” The Commonwealth's working definition of an Aboriginal or Torres Strait Islander person is, “a person of Aboriginal or Torres Strait Islander descent who identifies as an Aboriginal or Torres Strait Islander and is accepted as such by the community in which he or she lives”.36,37 A similar definition is used by the general public and in community healthcare, and this was the definition used by the NIEHS.
The Census question can be interpreted multiple ways, especially by people with mixed heritage. It specifically refers the biological component of the definition, but does not explicitly address either of the social components: self-identification and community acceptance. Some responders with distant Aboriginal or Torres Strait Islander heritage are likely to respond affirmatively to the question of origin, though they may not self-identify as Indigenous or be so identified by the community. Others may assume self-identification and community acceptance is implicit in the question. Community identification has been shown to contribute significantly to the definition but is difficult to quantify, thus it is not addressed by the Census question.36,37 We suspect that the variation in reported Indigenous origin with respect to self-identification contributed to the difference between the expected and observed numbers encountered in more heavily populated areas.
Successful recruitment strategies
Community awareness, collaboration with schools, stable family networks, and local guidance contributed to high participation rates. Assistance from health workers was the most critical component of recruitment, and was most successful when the AHW had an established relationship with the community. It was most difficult to recruit adult men; participation improved when they were recruited and examined by a male health worker or survey team member. We agree with Holmes et al. (2002) that Aboriginal peoples are more willing to participate in surveys if approached by peers rather than researchers, and that this is a more culturally sensitive method of recruitment.
Despite the methodological challenges encountered during the NIEHS, recruitment was successful. Community engagement and local guidance facilitated the relationships necessary to achieve a high overall participation rate of 78%. The importance of working in close consultation with local Aboriginal and Torres Strait Islander organisations is reiterated, as is the need to engage knowledgeable and connected health workers of each gender. Additional time on the ground helps team members to build relationships and trust and so engage with the community.
The authors gratefully acknowledge the contributions of Ms Mitasha Marolia, Ms Anthea Burnett, Ms Rebecca Still, Mr Ignatios Koukouras, Dr Jing Xie, Dr Tomer Shemesh, the NIEHS Advisory and Steering Committees, the Aboriginal Health Workers, especially Mr Cyril Oliver, Aboriginal Liaison Officers, school principals, Eye Health Coordinators, especially Ms Barbara O'Connor, Ms Desley Culpin and Mr Shaun Tatipata, the collaborating optometrists and ophthalmologists, and the fieldwork volunteers, especially Dr Jan Lovie-Kitchin, Ms Carmel McInally and Ms Nora Ley. We especially thank the thirty communities for their participation in this project.
The NIEHS was supported by the RANZCO Eye Foundation, the Vision Cooperative Research Centre, the International Centre for Eyecare Education, and the Harold Mitchell Foundation.