A Method for Recruiting Participants from Isolated Islands of Small Island Developing States (SIDS) for Survey Research

Abstract

Representing isolated small island communities through social survey research continues to be challenging. We examine a locally developed method to reach and recruit older people (65+ years) for a survey on well-being in the small island developing state of Maldives. The use of messengers to recruit participants is examined in the context of these isolated islands. The approach allowed for the participation of the isolated island communities in the survey yielding a high response rate. We consider the aspects of this approach for further application to survey research in isolated communities.

Keywords

survey research SIDS recruitment response rate older people

Isolated island communities are often underrepresented in social survey research. While we acknowledge that anthropologists continue to study isolated hard-to-reach communities (Averill 2006), the methods adopted by many such researchers are qualitative (Cloos et al. 2010; Connell 2010). Isolated communities are underrepresented in survey research at both country and international levels (Palloni and McEniry 2007; United Nations Children’s Fund 2012). This underrepresentation poses a problem for comparative analysis of population groups with different geospatial and sociocultural characteristics. It also has implications for establishing the relevance of theories and concepts, as researchers may overlook important correlates that they may not have observed otherwise (Yancey et al. 2006).

Research using inclusive research methods indicates that guaranteeing representation of isolated small population groups is challenging (Ejiogu et al. 2011). One issue relates to the study design, where low representation occurs when direct random selection is used without area sampling or household screening (Pavlik et al. 1996). In Asia-Pacific island countries, a cluster sampling method is adopted to obtain representation of isolated communities (Fua et al. 2007; United Nations Children’s Fund 2012). A second issue is the high cost of doing survey research in isolated communities (Stineman et al. 2011), where most of the resources are spent on developing contacts with communities (Ory et al. 2002). The appropriateness of the recruitment approach is a third reason for the underrepresentation of isolated communities. Passive approaches of recruitment that produce lower participation rates from isolated groups include activities such as mailing letters, sending out flyers, and placing announcements or stories in local media (Schoenfeld et al. 2000).

In researching the indigenous population in Australia, it was observed that the recruitment approaches are often not culturally congruent, and participants are unable to establish any links between the research and their needs (Maar et al. 2011). Use of proactive approaches yields increased participation rates from isolated communities. Proactive approaches that involve the researcher establishing face-to-face contact and building trusting partnerships with local community by spending adequate time to inform them about the research yield increased response rates (Mendez-Luck et al. 2011; Panapasa et al. 2012). In the context of small island developing states (SIDS), establishing contact with local institutions and leaders is seen as necessary for survey research (Aga Khan University 2010; Fua et al. 2007), while a multipronged flexible approach is adopted for recruiting participants. However, the practical constraints of financial resources often limit the extent of applying these approaches. This article sets out a locally appropriate, low-cost method for obtaining participation of isolated island communities of Maldives for social survey research.

Background

Maldives is a SIDS of 1,190 islands scattered over the Indian Ocean, of which only 196 are inhabited. The population was 298,968 in the 2006 census (Ministry of Planning and National Development 2008); 70% lived on islands with a population less than 10,000 (Department of National Planning 2010). The population is socioculturally homogeneous, monolingual (speaking the language Dhivehi), practices one religion (Islam), and lives in an extended family arrangement. Although a middle-income country (The World Bank 2013), some basic services such as public transport and postal services are not established in the smaller islands. The exception is the mobile telephone that reaches even the most isolated island. For transport purposes, the island communities usually rely on fishing boats owned by islanders when the need arises to travel to other islands. Locals in these islands also depend on the boat crews’ generosity when they need to deliver goods to their family and friends in other islands, as there is no regular postal service.

We describe the approaches we used to reach and recruit participants from the islands of Maldives for a cross-sectional survey of people aged 65+ years in 2012, using structured telephone interviews.

Methods

The methods and approaches we used were not only required to achieve sufficient participation of the target population from different island contexts but also to maintain societal interest, cultural authenticity, and ethical robustness. We present the six stages of the sampling and recruitment (summarized in Table 1) process, which comprised of two stages for sampling the islands and four stages for sampling and recruitment of the participants. We illustrate the messenger approach, a locally appropriate method we developed for reaching the participants and its impact on the response rate.

Table 1.

Sampling of 190 Islands, and Sampling and Recruitment of Participants (65+ years) for the Survey of Well-being in Maldives.

Stage I	Stage II	Stage III	Stage IV	Stage V	Stage VI
Developing sampling frame of islands (classifying island clusters)	Sampling of islands	Developing sampling frame of potential participants	Sampling of participants	Recruiting participants	Administering interviews
Dense cluster (population 100,000 +) 1 island	1 island	List of people 65+ years compiled by NSPA	Participants selected randomly from the NSPA list by researchers	Participants contacted by phone by researchers	Telephone interview by researchers
Moderate cluster (population 1,000–99,000) 63 islands	Random selection of 6% of islands (4 islands)	NSPA list updated by door-to-door screening by messenger	Participants selected randomly from the NSPA list by researchers	Participants contacted by messenger visit	Telephone interview by researchers
Sparse cluster (population below 1,000) 127 islands	Random selection of 5% of islands (6 islands)	NSPA list updated by door-to-door screening by messenger	Total population 65+ selected from the NSPA list by researchers	Participants contacted by messenger visit	Telephone interview by researchers

Note: NSPA = National Social Protection Agency.

Sampling of the Islands

The first stage was to develop the sampling frame of islands based on the list of islands in the local governance areas and classify them into three clusters according to the population density. The three clusters were labeled “sparse” (population below 1,000), “moderate” (population 1,000–99,999), and “dense” (population 100,000 and above). This resulted in a sampling frame of 190 islands with one island in the dense cluster, 63 islands in the moderate cluster, and 127 islands in the sparse cluster. It must be noted that despite the use of this classification, none of the islands in the moderate cluster had a population of more than 10,000.

At the second stage, sampling of the islands, we randomly selected 5% of the islands in the moderate and sparse cluster, but since the dense cluster included only one island this was not necessary. The criterion of at least 5% of the islands was chosen, as it provides the estimated minimum number of islands that would yield an adequate sample size from the sparse cluster, which had the least number of people per island. However, in consultation with local stakeholders, we increased the proportion of islands in the moderate cluster to 6% and selected a fourth island randomly. This adjustment was necessary to address the concerns of the local stakeholders that the varying island contexts in this cluster were not adequately represented when the 5% criteria is used. Hence, the final sample consisted of 11 islands in the three clusters (1 island in the dense, 4 in the moderate, and 6 in the sparse cluster).

Sampling and Recruitment of the Participants

Developing the sampling frame of target population in the 11 islands was the third stage. A list of people aged 65+ years in Maldives was obtained from the National Social Protection Agency (NSPA), the statutory body that provides social health insurance to the population, including people aged 65+ years. The NSPA list included names and contact details of people aged 65+ years in all the islands.

We started with the NSPA list as the sampling frame since it was a random list in that surnames or first names of Maldives residents are not listed in alphabetical order. However, as the NSPA list was last updated in 2009, it included people who had migrated to other islands and some who had passed away. Also the telephone numbers in the NSPA list were complete only for the island in the dense cluster, with very few contact numbers of people in the islands of the moderate and sparse clusters. We had to identify a local way to reach and obtain the contact numbers of those in the islands of moderate and sparse clusters. We held discussions with the local stakeholders and proposed the idea of using dhathuru-verin (messengers) for reaching and recruiting the participants from these clusters. This is an approach commonly used by the island communities for trade and delivery purposes and is a socioculturally appropriate method to approach potential participants in these clusters. This idea was confirmed by the stakeholders and is further discussed subsequently. We used messengers to update the NSPA list with contact numbers by conducting door-to-door screening in the 10 islands in the moderate and sparse clusters.

At the fourth stage, sampling of participants, we randomly selected participants from the NSPA list in the dense cluster and moderate cluster. In a case of noncontact in the dense cluster, we replaced the selection using the same random process. This was not required in the moderate cluster, as we were able to reach all the selected participants. In the sparse cluster, however, we selected the total population aged 65+ years in each island due to the small number of target population in these islands and in the interest of maintaining harmony in the community.

In the fifth stage, where participants were recruited, the messenger in each island of the moderate and sparse clusters visited the selected participants, provided information about the survey, and obtained their consent. In fact, in the sparse cluster, stages 3 to 5 were combined as the total target population was selected. During the recruitment, the participants were informed by the messengers that the telephone interview would be conducted by the researchers and that the messengers did not have access to the information the participants provided in the survey interview. Once the messengers recruited the participants, they brought the consent forms to the researchers. In the dense cluster, recruitment was by direct dialing of the selected participants, providing information, and obtaining consent. Where possible, this stage was combined with administration of the interview, the sixth stage. In the moderate and sparse clusters, the telephone interviews were administered by the researchers as a separate stage. The interview included 60 structured questions, with the responses coded on scales of 3 or 5.

Messenger Approach

We adopted the messenger approach for reaching and recruiting participants from the moderate and sparse clusters for many reasons. First, the only way to obtain phone numbers of potential participants was to approach the households directly. The island communities mostly use mobile phones, a common practice in other rural settings (Mammen and Sano 2012), and some older people use the phones of their children or friends. A second reason was the cost savings made by the use of messengers to reach the target population rather than the researchers traveling to the islands. Third, this is a locally accepted way of reaching people in the isolated islands of Maldives, thereby making the approach culturally appropriate and logistically simple. The fourth reason relates to the connectedness of the messengers to their individual islands. Although one of the researchers is familiar with the culture and the local language, we do not have in-depth knowledge of the practices of individual island communities. The messengers, as local residents, have current, in-depth knowledge of the households and the community and are also familiar with their way of doing things. Hence, there is greater acceptance of the messenger by the island community compared to the researchers.

We identified a set of messenger tasks and characteristics based on the research goals. The specific tasks included household screening and updating the NSPA list with the contact numbers, informing participants about the research, and obtaining consent of the target population to participate in the survey.

The messenger characteristics were the following: (1) a local resident of the island with similar sociocultural practices to the island community; (2) a person who regularly traveled back and forth from the island to the capital city, Malé, with the responsibility to undertake informal tasks for the community; (3) a friendly person known by the community; (4) a person engaged in community social activities; and (5) a person able to read and write the local language fluently.

The messengers were identified with the assistance of local stakeholders who shortlisted two to three possible messengers for each island. We selected one after a telephone interview to assess the willingness and interest of the messenger to engage in this research. A face-to-face meeting was held with each messenger to provide information about the research. A decision was then made as to the communication capabilities of the messenger in conveying information to the participants. We provided the messengers with written instructions, tasks they were required to undertake, and materials for the tasks.

Finally, we agreed on a token payment for the messenger services based on the understanding that it would be a small token as the researchers were self-financing and that the messengers were doing it voluntarily as a contribution to the community. The token payment was provided at the end of their work and varied from US$50 to US$150, depending on the time and effort they had to put in as the households to visit varied from 20 to more than 200, depending on the island. These figures must be taken with caution as they are just token appreciation and specific to the sociocultural context of the islands.

Ten messengers were selected, one messenger for each island in the moderate and sparse clusters. During the time the messengers spent recruiting the participants, one of the researchers was in daily contact with them by telephone to provide any further clarification.

Results

In the dense cluster where direct dialing were used, the recruitment rate was 93% (see Table 2). In the two clusters where the messenger approach was used, the average recruitment rate was 90% (91% in moderate and 88% in sparse clusters). However, this difference in recruitment rate by the method of reaching the participants is not statistically significant (p = .380).

Table 2.

Recruitment and Response Rates of Maldives Residents Aged 65+ Years by the Method of Recruitment and Island Cluster.

Method of recruitment	Island cluster		Participants contacted	Recruitment rate (participants recruited)	Response rate (participants completed survey)
Direct dialing by researcher	Densely populated	N	127	118	85
Direct dialing by researcher	Densely populated	Percentage within island cluster		93	72
Visit by messenger	Moderately populated	N	244	223	191
Visit by messenger	Moderately populated	Percentage within island cluster		91	86
Visit by messenger	Sparsely populated	N	139	123	117
Visit by messenger	Sparsely populated	Percentage within island cluster		88	95
Total		N	510	464	393
Total		Percentage of total	100%	91	85

Note: NPSA = National Social Protection Agency.

In all three clusters, we obtained a high response rate (Table 1) as measured by those who completed the interviews from those who agreed to participate. The response rate is significantly higher (p < .01) in the two clusters where the messenger approach was used to recruit participants (average of 89%) than those recruited using direct dialing (72%).

Discussion

The results of this study show that the messenger approach is an effective method for reaching and obtaining participation of isolated communities in SIDS in survey research. We were able to recruit a high proportion of the target population from the moderate and sparse clusters, which included very isolated small communities that are often underrepresented in survey research. Although it appears that the use of direct dialing yielded a better recruitment outcome compared to the messenger approach, the difference observed is not statistically significant.

Nonetheless, the use of messengers for recruitment yielded a significantly higher response rate compared to direct dialing of participants (p < .01). In the dense cluster where participants were recruited by direct dialing, we encountered withdrawal from the interview when we made callbacks and during the interview itself. However, we did not encounter such withdrawals from the participants of the two clusters where a messenger was used. The face-to-face interaction with the messenger for recruitment could have contributed to the higher response rate in these clusters. Where direct dialing was used, participant fatigue caused by telephone sales people or previous researchers (Groves 1990) could have contributed to the withdrawals from the interview, although we did not observe this in the recruitment stage.

Another factor contributing the high response rate where a messenger was used is the respect of and commitment to the island community the messengers already had. This facilitated the establishment of a trusting relationship between the researchers and the participants and the legitimacy of the researchers. Once a trusting partnership is established with the local people, the participants become more willing to be part of the research (Emmel et al. 2007; Mendez-Luck et al. 2011). Another reason could be that the people of isolated communities become more trusting when recruitment relies on informal community ties (Maar et al. 2011; Yancey et al. 2006). All these aspects are inherent in the messenger approach and contribute to the effective retention of participants.

The success of the messenger approach also relies on the messengers’ characteristics. The fact that each messenger was a local resident of the island who shared local sociocultural practices and had the trust of the community and good communication skills made the potential participants more willing to be part of the research. Other researchers have observed that a good knowledge of the community and effective communication are essential for recruiting participants from isolated communities (Areán et al. 2003). Qualitative researchers in isolated communities also rely on informal contacts and gatekeepers as intermediaries to establish contact with their research participants (Panapasa et al. 2012). Such intermediaries often have similar characteristics as those identified for messenger characteristics in this research.

Although provision of incentives to participants has been shown to increase response rates (O’Connor 2011), none was provided to participants in this study. Even though we provided a token payment to the messengers who recruited the participants, it is not considered to have contributed to the response rates as the payment was given to the messengers as a token of appreciation and at the completion of their work. Furthermore, the interviews were administered by the researchers at a later stage.

Overall, the benefits of the messenger approach were that it provided a high level of participation from very isolated island contexts for comparative analysis. Second, this approach was not as resource intensive as the researchers traveling to the islands to recruit participants. We estimated that our budget was cut by 34% by adopting the messenger approach to reach and recruit participants. Third, it enabled the establishment of a link with the community, which could be used later for providing feedback to the community.

We encountered some methodological issues in applying this approach in the field. In some islands in the sparse cluster, exclusion of participants based on eligibility criteria for a telephone interview created some dissatisfaction among the potential participants as almost everybody wanted to be included. Resolving this required additional time and effort from the messengers and the support of island leaders to alleviate concerns of the community. This issue did, however, legitimate our decision to select the total population of those 65 and older in the sparse cluster.

Another issue was how the use of messengers could affect the anonymity of the survey participants. In small isolated communities, information trav-els by word of mouth and often there is no privacy (Leipert and Reutter 1998), and even without the use of intermediaries the community is often aware of those who participate in the research. In the islands in the sparse cluster, we anticipated this issue in that everyone knew that all those 65 and older would be asked to participate in the survey.

We observed a significantly higher response rate for the sparse cluster compared to the moderate cluster (p < .01), which could indicate that there are factors in play other than the messenger approach itself. This includes the influence of filial relations within the family and the close community relationships in the small island settings. An older person within the extended family is subject to the influence of family members exerting filial piety and acting as the gatekeepers for older family members’ wider community involvement (Phillips and Cheng 2012). In this instance, the backing of the family gatekeepers was critical to the involvement of the older person in the survey.

The people in these small island communities were tightly knit, and local leaders had kinship and friendship relations with families that enabled them to persuade the participants to participate in the research. In the islands where the island leaders showed a keen interest in the research, they facilitated the recruitment efforts by informing the community and families about the survey. Our experience was that informing family members about the research, alleviating their concerns, and gaining their support were necessary before successful recruitment of their older family members was possible. While we sent written communication to the island leaders informing them about the research, the messengers also had close contact with the island leaders and were able to gain their support. Other researchers have also observed that collaboration and partnership with local institutions is an important factor that facilitates engaging with potential participants in their community (Reed et al. 2003).

Some parallels can be drawn between the messenger approach for social surveys and the use of gatekeepers by ethnographic researchers to study isolated communities. In both, the primary approach is to rely on informal contacts and stakeholders to access the community and establish collaboration or partnership with those who know the community well and can assist in creating a supportive atmosphere for the study (Maar et al. 2011; Mody et al. 2008). The messenger approach used in this survey research built on such collaboration and the adoption of context-specific local approaches.

Conclusion

The messenger approach used in this study allowed us to obtain a high response rate from otherwise underrepresented population groups in survey research. It is an effective, low-cost, and locally appropriate method to reach and recruit participants from isolated island communities. This approach has potential for small-scale surveys in other isolated island communities, especially the isolated settings of the SIDS. It can be regarded as an adaptation of the approaches used to reach isolated communities in qualitative research extended to a quantitative survey research context.

Footnotes

Acknowledgment

We acknowledge the cooperation and friendliness of the people who acted as messengers during our fieldwork for this research in Maldives.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

References

Aga Khan University. 2010. Project Report: National Micronutrient Survey 2007. Malé, Maldives: United Nations Children’s Fund and Ministry of Health.

Areán

P. A.

Alvidrez

Nery

Estes

Linkins

2003. Recruitment and retention of older minorities in mental health services research. The Gerontologist 43:36–44.

Averill

J. B.

2006. Getting started: Initiating critical ethnography and community-based action research in a program of rural health studies. International Journal of Qualitative Methods 5:1–8.

Cloos

Allen

C. F.

Alvarado

B. E.

Zunzunegui

M. V.

Simon

D. T.

Eldermire-Shearer

2010. “Active ageing”: A qualitative study in six Caribbean countries. Ageing & Society 30:79–101.

Connell

2010. Pacific islands in the global economy: Paradoxes of migration and culture. Singapore Journal of Tropical Geography 31:115–29.

Department of National Planning. 2010. Statistical Archive of Maldives. http://www.planning.gov.mv/statistics_archive/statistical_tables.html (accessed February 24, 2012).

Ejiogu

Norbeck

J. H.

Mason

M. A.

Cromwell

B. C.

Zonderman

A. B.

Evans

M. K.

2011. Recruitment and retention strategies for minority or poor clinical research participants: Lessons from the healthy aging in neighborhoods of diversity across the life span study. The Gerontologist 51:S33–45.

Emmel

Hughes

Greenhalgh

Sales

2007. Accessing socially excluded people—Trust and the gatekeeper in the researcher–participant relationship. Sociological Research Online 12. doi:10.5153/sro.1512.

Fua

S. J.

Manu

Takapautolo

Taufe’ulungaki

2007. Sustainable Livelihood and Education in the Pacific: Tonga Pilot Report. Suva, Fiji: Institute of Education, University of South Pacific.

10.

Groves

R. M.

1990. Theories and methods of telephone surveys. Annual Review of Sociology 16:221–40.

11.

Leipert

Reutter

1998. Women’s health and community health nursing practice in geographically isolated settings: A Canadian perspective. Health Care for Women International 19:575–88.

12.

Maar

M. A.

Lightfoot

N. E.

Sutherland

M. E.

Strasser

R. P.

Wilson

K. J.

Lidstone-Jones

C. M.

Graham

D. G.

. 2011. Thinking outside the box: Aboriginal people’s suggestions for conducting health studies with aboriginal communities. Public Health 125:747–53.

13.

Mammen

Sano

2012. Gaining access to economically marginalized rural populations: Lessons learned from nonprobability sampling. Rural Sociology 77:462–82.

14.

Mendez-Luck

C. A.

Trejo

Miranda

Jimenez

Quiter

E. S.

Mangione

C. M.

2011. Recruitment strategies and costs associated with community-based research in a Mexican-origin population. The Gerontologist 51:S94–105.

15.

Ministry of Planning and National Development. 2008. Analytical Report: Population and Housing Census 2006. Malé, Maldives: Ministry of Planning and National Development.

16.

Mody

Miller

D. K.

McGloin

J. M.

Freeman

Marcantonio

E. R.

Magaziner

Studenski

2008. Recruitment and retention of older adults in aging research. Journal of the American Geriatrics Society 56:2340–48.

17.

O’Connor

2011. The challenge of recruiting control groups. An experiment of different recruitment models in the control group of a clinical psychological postal survey. Quality & Quantity 45:743–50. doi:10.1007/s11135-010-9423-8.

18.

Ory

M. G.

Lipman

P. D.

Karlen

P. L.

Gerety

M. B.

Stevens

V. J.

Fiatarone Singh

M. A.

Buchner

D. M.

., and The Ficsit Group. 2002. Recruitment of older participants in frailty/injury prevention studies. Prevention Science 3:1–22.

19.

Palloni

McEniry

2007. Aging and health status of elderly in Latin America and the Caribbean: Preliminary findings. Journal of Cross-cultural Gerontology 22:263–85.

20.

Panapasa

Jackson

Caldwell

Heeringa

McNally

Williams

Taumoepeau

C. L.

. 2012. Community-based participatory research approach to evidence-based research: Lessons from the Pacific Islander American Health Study: Progress in community health partnerships. Research, Education, and Action 6:53–58. doi:10.1353/cpr.2012.0013.

21.

Pavlik

V. N.

Hyman

D. J.

Vallbona

Dunn

J. K.

Louis

Dewey

C. M.

Wieck

Toronjo

1996. Response rates to random digit dialling for recruiting participants to an onsite health study. Public Health Reports 111:444–50.

22.

Phillips

D. R.

Cheng

K. H. C.

2012. The impact of changing value system on social inclusion: An Asia-Pacific perspective. In From Exclusion to Inclusion in Old Age: A Global Challenge, eds. Scharf

Keating

N. C.

, 109–24. Bristol, UK: The Policy Press.

23.

Reed

P. S.

Long Foley

Hatch

Mutran

E. J.

2003. Recruitment of older African Americans for survey research: A process evaluation of the community and church-based strategy in the Durham elders project. The Gerontologist 43:52–61.

24.

Schoenfeld

E. R.

Greene

J. M.

S.-Y.

O’Leary

Forte

Leske

M. C.

2000. Recruiting participants for community-based research: The Diabetic Retinopathy Awareness Program. Annals of Epidemiology 10:432–40.

25.

Stineman

M. G.

Strumpf

Kurichi

J. E.

Charles

Grisso

J. A.

Jayadevappa

2011. Attempts to reach the oldest and frailest: Recruitment, adherence and retention of urban elderly persons to a falls reduction exercise program. The Gerontologist 51:S59–72.

26.

The World Bank. 2013. Data. http://data.worldbank.org/income-level/UMC (accessed February 2, 2013).

27.

United Nations Children’s Fund. 2012. Childinfo: Multiple Indicator Cluster Surveys. http://www.childinfo.org/mics.html (accessed March 14, 2013).

28.

Yancey

A. K.

Ortega

A. N.

Kumanyika

S. K.

2006. Effective recruitment and retention of minority research participants. Annual Review of Public Health 27:1–28.