Table 3 Summary of previous published literature reviews (n = 31 reviews) of barriers and/or strategies to increase disadvantaged group representation in health research

Marpsat & Razafindratsima [152]

Narrative with 60 included studies

General ‘hard-to-reach’

No sampling frame – difficulties sampling hidden groups and rare populations

1. Time-location sampling.

Each non-probability sampling method described advantages and disadvantages of each strategy presented with no preferences recommended.

2. Respondent driven sampling.

3. Capture-recapture (or contact-re-contact).

Malekinejad et al. [151]

Systematic review with quantitative analysis with 123 included studies from Africa, Asia, Europe, Latin America, Oceania

HIV related (injecting drug users, sex workers and homosexual men)

Sampling for HIV surveillance studies

Respondent drive sampling

All but 13 studies reached >90% of their intended sample size and only six failed to reach equilibrium for key demographics suggesting representativeness. The majority of studies used formative research, face-to-face interview formats, three referrals per participant, a single interview site for data collection, and incentives.

Aldana & Quintero [153]

Narrative review with 18 included studies

General ‘hard-to-reach’

Sampling for surveillance studies

1. Venue-based time-location sampling.

Advantages and disadvantages of each strategy discussed.

2. Targeted sampling.

Targeted sampling which requires ethnographic data provides important qualitative information.

3. Respondent driven sampling

Venue-based sampling allowed a systematic recruitment of participants and produced probability samples but only of the visits to the venues included in the sample.

Respondent driven sampling based on social networks permitted calculation of population estimates.

Atkinson & Flint [150]

Narrative review with 22 included studies

General ‘hard-to-reach’

Sampling in general

Snowball sampling

Described advantages and disadvantages of snowball sampling with no comparison with other methods. Recommended particularly for qualitative research.

Peterson et al. [34]

Narrative review with 32 studies

Illicit drug users

Sampling following discharge from treatment

Targeted sampling (i.e., non-probability sampling method requiring identification of high yield locations).

Comparison with ‘in treatment’ samples suggested representative samples obtained using targeted sampling.

Faugier & Sargeant [154]

Narrative review with about 30 studies (number not reported, estimated from reference list)

Sex workers, HIV, substance abuse populations

Sampling issues

Non-random methods of sampling such as snowball sampling.

Snowball sampling, although not probability or random sampling, has been used for research with hard to reach groups.

Andresen et al. [138]

Narrative review with about 100 studies (number not reported, estimated from reference list)

‘Low-frequency’ populations in public health surveillance surveys

Sampling groups with small numbers in the population results in small samples of hard to reach groups in population surveys

Improving analysis using:

Discussed advantages and weaknesses of each strategy for increasing small-group participation in population surveys. Recommendations included:

• Aggregating data – by location or time

Partnering with agencies (e.g., state hospital associations);

• Spatial smoothing

Bilingual interviewers;

• Small area estimation

Including non-random sampling methods to enhance samples;

• Exact statistics

Changes in statistical methods.

• Provider profiling methods

Using non-probability sampling strategies:

• Convenience sampling

• Snowball methods

• Publicly available or commercial lists

Improving recruitment methods:

• Extending survey modes (to include face to face)

• Staff training

• Incentives

• Involve community members

• Bilingual interviewers and surveys

Data collection methods:

• Use qualitative methods and participatory research

• Multi-method surveys

Kalton [149]

Narrative and selective review with about 90 studies (number not reported, estimated from reference list)

Rare (low prevalence) populations

Sampling

Over-sampling

Options discussed include:

• Disproportionate stratified sampling;

• Two-phase sampling;

• Use of multiple frames;

• Multiplicity sampling;

• Location sampling;

• Panel surveys;

• Use of multi-purpose surveys.

Recommended use of more than one sampling method and tailoring based on study research question.

Dhalla & Poole, [147]

Systematic review with 19 studies from OECD countries and 39 studies from non-OECD countries

Hard-to-reach participants for HIV vaccine trials

Low recruitment rates into HIV vaccine trials

Two types of participant motivation to participate in HIV vaccine trials: altruistic (social benefits) and personal benefits)

Motivators which may help develop and tailor recruitment strategies included: (altruistic) to protect partner/community/others; help stop spread of AIDS; help research; help find cure; (personal) protection from HIV; because friends are; to enjoy sex; monetary and non-cash incentives; doctor/military suggestion; personal recognition; free health care.

Swanson & Ward [131]

Systematic review with qualitative synthesis with 107 studies (about 50% were on the barriers to recruitment only)

American ethnic minorities

Recruitment into clinical trials

Effective and ineffective methods outlined

‘Effective’ methods for recruitment based on methodologically superior studies included:

• Community partnerships

• Community leaders

• Community involvement in development

• Formative qualitative research

• Gifts and incentives

• Provide transport

• Community advisors

• Sustainable interventions

• Grants to health services

• Establish networks of doctors

• Cultural/family tailoring

• Recruitment materials in target language

• Tracking databases

• Active recruitment with research staff to help completion of forms

• Train staff in cultural issues

• Employ community residents as part of research

UyBico et al. [141]

Systematic review with quantitative analysis and consideration of studies’ methodological quality with 20 included studies

Vulnerable populations

Recruitment into intervention health research

1. Social marketing (e.g. mass mailing, telephone calls, mass media).

Successful strategies included:

2. Community outreach (via churches, community organisations, presentations/meetings, community events, door to door canvassing).

• Social marketing was successful in 44% of studies.

3. Through health system (doctor referral, health centre recruitment, registry, patient records)

• Health system based strategies were successful in 40% of studies.

4. Referrals (by friends/family, other research participants)

• Referrals strategies were successful in 35% of studies.

• Community outreach was successful in 13% of studies

Lai et al. [142]

Systematic review with qualitative synthesis with 14 included studies

Under-represented populations

Recruitment into cancer clinical trials

1. Recruitment letters/flyers and telephone calls.

Only three studies reported efficacious strategies:

2. Incentives and gifts.

• Media campaign (compared to clinic registry recruitment)

3. Recruitment facilitators (doctors, insurers, businessmen and community organisations).

• Enhanced mailing process, church project sessions conducted by African Americans, letters and telephone reminders (compared to mailed survey only and telephone only)

• Companies providing researchers with names and phone numbers (compared with employees actively signing up at work)

Howerton et al. [143]

Systematic review with 18 studies

Under-represented populations (recruitment to cancer clinical trial)

Clinician characteristics, attitudes and practices

11 clinician level promoters of recruitment

Clinician factors such as communication style, lack of trust of research, lack of awareness, logistics and cost are barriers to the recruitment of under-represented populations in cancer clinical trials. Addressing those barriers through incentives (e.g., extra staff), provider training, institutional affiliation, helps improve attitudes and recruitment.

Wendler et al. [146]

Systematic review with 20 studies with meta-analysis

Racial and ethics minorities

Participants’ willingness to participate and attitudes towards research

Changing attitudes to research versus other pragmatic barriers to research.

No differences in willingness (as measured by consent rates of racial minority groups versus non-minority Whites) to participate in research by racial group. Attitudes towards research were positive. Efforts should focus on improving access to research not changing participant attitudes.

Hussain-Gambles et al. [14, 15]

Systematic and thematic review with about 38 studies (number not reported estimated from reference list)

Ethnic minority groups

Participant fear and mistrust, inappropriate exclusion criteria and study designs, costs to researchers, lack of ethnic staff, socio-cultural issues, cultural myths

N/A – review of barriers only

Under-representation caused by a combination of factors that need to be addressed. Potential strategies are discussed, but not included in the review (such as strategies to reduce fear and mistrust, education and training, community links and advocacy, outreach strategies, recruitment through primary health care) to address barriers.

Ford et al. [132]

Systematic review with 46 studies

Racial and ethnic minorities, older, rural and low socioeconomic status

Mistrust of research, perceived harms, costs, transport, lack of education, time, fear, family, provider attitudes and characteristics, communication, lack of protocols, religious/spiritual beliefs, low health literacy, culture

N/A – review of barriers only

Concluded under-represented groups face numerous barriers to participation in cancer-related trials.

Guiliano et al. [144]

Non-systematic narrative review

‘Minority’ groups

Structural, cultural and linguistic factors limiting participation in cancer research

N/A- review of barriers

Research where participants feel ownership, trust and receive results more likely to increase participation.

Miranda et al. [145]

Narrative review with about 48 studies (number not reported, estimated from reference list)

Low income Latinos

Lack of insurance, time, child care, and transport. Cultural barriers and beliefs in traditional non-medical healers.

1. Use of health services for recruitment.

Research needs to be culturally sensitive and to remove logistic barriers. Bilingual and bicultural staff should be part of the research team.

2. Bilingual and bicultural staff.

3. Developing culturally sensitive research materials.

Flory et al. [140]

Systematic review with 30 included studies

Low literacy

Participants limited understanding and literacy

1. Multi-media,

Extended discussion (educator or staff spending more time talking one-on-one to participants) was the most effective strategy according to methodologically superior studies.

2. Enhanced’ consent forms,

3. Extended discussions

4. Test/feedback (quizzing participants about the information)

5. Miscellaneous

Shavers-Hornaday et al. [139]

Non-systematic and narrative review with about 100 studies (number not reported, estimated from reference list)

African-Americans

Participant barriers (distrust, health care access and utilization, quality of care).

Outlined strategies based on 11 cancer trials that successfully recruited African/Americans

Effective recruitment strategies based on the results of 11 cancer trials that successfully recruited representative samples of African American include:

Investigator barriers (low recruitment and retention, cost, relationships with minority health professionals)/

• An active commitment to recruiting African American subjects;

• Community outreach programs and advertisements;

• Involvement of local churches and community organizations;

• Publicity campaigns directed at African Americans;

• Participant logistics such as convenient testing times, transport, convenient location;

• Use of incentives;

• Use of African American role models;

• Flexibility and willingness to change protocol

• Use of lay health workers;

• Door to door canvassing

Ndumele et al. [156]

Systematic review with 45 included studies

Minority populations with chronic disease

Recruitment into qualitative research

1. Health care setting

Based on 21 studies that provided data-based results: no discernible patterns of recruitment method that seemed associated with greater rates of participant recruitment.

2. Community organisations

3. Electronic or mailed invitations

4. Media

5. Word of mouth

6. Incentives

7. Recruits from existing databases

Hergenrather et al. [133]

Systematic review with qualitative synthesis with 31 included studies

Various vulnerable populations

Engaging communities and collecting data in acceptable methods

Photovoice (i.e., use of photographs to encourage group discussion).

Photovoice expanded representation and diversity of community members participating in health research.

Halcomb et al. [134]

Integrative review with about 40 studies (number not reported, estimated from reference list)

Culturally and linguistically diverse (CALD)

Challenges in focus group research with CALD populations

1. Involvement of key members of the target group;

Involvement of leader members of the target group; bilingual facilitators; consider particular CALD group as not all are the same are key considerations.

2. Logistical convenience;

3. Physical environment should be considered;

4. Bilingual facilitators;

5. Consider particular CALD group as not all are the same (e.g., incentives not acceptable to some).

Sheridan et al. [130]

Systematic review with 38 included articles

Low health literacy populations

Populations with low health literacy and language difficulties

Design features of health information interventions.

Multiple strategies are required to improve availability of health information and intervention for low literacy/numeracy groups. Design features found to improve participant comprehension:

• Presenting essential information first or on its own;

• Presenting disease risk or treatment benefit information using the same denominators

• Presenting numerical information in tables not text

• Adding icon arrays to numerical information

• Adding video to verbal narrative

Glazier et al. [157]

Systematic review with 17 included studies

Socially disadvantaged groups

Diabetes care interventions for socially disadvantaged groups

Patient, provider and health system interventions for diabetes

Positive intervention features were:

• Cultural tailoring

• Community or lay educators

• One-on-one interventions

• Behaviour related tasks

• Feedback

• High intensity and long duration

Negative intervention features were:

• Didactic teaching

• Focus on increasing (diabetes) knowledge

Yancey et al. [148]

Systematic review with qualitative synthesis and consideration of studies’ methodological quality with 95 included studies

African American and American ethnic groups

Recruitment and retention in general

1. Community involvement in research.

Based on methodologically superior studies, strategies recommended to increase recruitment and retention include:

2. Incentives and logistical aids (e.g., transport).

3. Cultural tailoring.

• Population based sampling strategies are unlikely to produce sufficient numbers

• Personal contact and mass media were efficacious for recruitment

• Non-restrictive eligibility criteria

• Community involvement more critical to retention than recruitment

• Timely incentive payments

• Cultural tailoring

Johnson et al. [155]

Systematic review with 6 included studies (RCTs only)

African American

Recruitment and retention into genetic and genomic studies

1. Population strategies (phone, mail or postcard);

• Phone recruitment (63%-91%) significantly better than mail or postcard (3%-19%).

2. community-based strategies (community engagement and partnerships);

• Community engagement produced mixed recruitment results (1% - 82%) with locals as recruiters important success factor.

3. Incentives

• Only 1 trial of incentives found no effect on retention.

Wallace & Bartlett [158]

Narrative review with about 38 studies (number not reported, estimated from reference list)

African American and Hispanic girls and women

Recruitment and retention

Recruitment:

Recommended using all of the strategies outlined.

• Building trust

• Familiarity and visibility

• Racial and ethnic concordance

• Convenience

Retention:

• Provide transport

• Language, literacy and culturally appropriate,

• Emphasising safety

• Flexibility

• Incentives

• Regular communication

• Veracity

Grove et al. [135]

Narrative review with 20 studies

Indigenous Australians

Recruitment and retention in longitudinal research

Community participation (and ownership), developing relationships.

Three studies reported successful recruitment and retention using community participation strategies.

Booth et al. [136]

Non-systematic (selective) narrative review with about 13 studies (number not reported, estimated from reference list)

Homeless people

Sampling and data collection, ethical and fieldwork issues

Key recommendations were:

1. Convenience (non-random) sampling.

• Use common sense to enhance practicality and reduce burden on participants;

2. Building trust.

• Cultural sensitivity;

3. Simple consent materials.

• Take non-threatening approach;

4. Incentives to participate.

• Appropriate language;

5. Cultural sensitivity.

• ‘Hanging out’ with target group;

6. Using a non-threatening approach.

• Provide feedback;

7. Providing feedback.

• Note style of clothing;

• Use informants and service providers;

• Incentives

Mathers & Cramer [137]

Narrative review with about 28 studies (number not reported, estimated from reference list)

Gay, Lesbian, Bisexual, Transgender (GLBT)

Recruitment & data collection: difficulties identifying ‘hidden’ samples. Researchers as ‘outsiders’. Researcher verbal and non verbal cues influencing survey responses.

Web and videoconferencing

Web and videoconferencing are private and non-threatening for hidden populations and should provide more accurate data.