Recruitment and Sampling Methods in Community-Based Infectious Disease Research

1
Recruitment and Sampling Methods in
Community-Based Infectious Disease Research

• Crystal M. Fuller, PhD
• Infectious Disease Epidemiology Methods (P8462)
• Spring 2002

2
Lecture Overview

• I. Why study high-risk populations using a
  community-based research approach?
• II. Traditional sampling methods and sample
  selection
• III. Special issues in studying high-risk
  populations and infectious diseases (i.e., HIV,
  HCV, TB)
• Sampling and recruitment methods adapted used to
  gain external validity in studying hidden or
  difficult-to-reach populations
• Conducting research among young IDUs
• The REACH II Study, Baltimore, Maryland

3
High Risk Populations

• Infectious Disease Population at risk
• TB Homeless Injection drug users (IDUs)
• HIV Injection drug users
• Non-injection drug users
• Sex workers
• Men who have sex w/ men
• (MSM minority subgroups)
• Sex partners of high-risk individuals
• HCV Injection drug users

4
HIV Prevalence among Adult IDUs

• YEAR AUTHOR
• 1987 Marmor et al.
• 1990 Caussy et al.
• 1990 Vlahov et al.
• 1990 Koblin et al.
• 1993 Metzger et al.
• 1996 Wiebel et al.
• 1987 Chaisson et al.
• 1997 Cheng et al.

• CITY (N) HIV
• New York 290 48
• New Jersey 926 35
• Baltimore 2,616 25
• Worcester 874 19
• Philadelphia 88 18
• Chicago 850 25
• San Francisco 281 10
• Los Angeles 3,854 2

5
HIV Prevalence among New and Young IDUs

• 1988 van den Hoek et al. Amsterdam 15.6
  (Injecting lt 5 years)
• 1989 Friedman et al. New York 22.0
  (Injecting lt 5 years)
• 1990 Vlahov et al. Baltimore 15.1
  (Injecting lt 18 months)
• 1992 van Ameijden et al. Amsterdam 29.2
  (Injecting lt 2 years)
• 1996 Garfein et al. Baltimore 13.9
  (Injecting lt 1 years)

6
HIV Incidence among Adult IDUs

• 1998 17 U.S. Cities 1.5 per 100 PY (Fisher et
  al.)
• 1995 Baltimore 4.0 per 100 PY (Nelson et
  al.)
• 1990 Chicago 4.9 per 100 PY (Wiebel et al.)
• 1990 San Francisco 1.9 per 100 PY (Moss et al.)

7
HIV Incidence among Young, New IDUs Year
Author Sample
Characteristic Incidence
(Per 100 PY)

• 1992 Nicolosi et al. Italy (lt 20 yrs of
  age) 16.1 (gt 20 yrs of age) 4.9
• 1995 Nelson et al. Baltimore (men lt 34 yrs
  of age) 2.2 (men gt 34 yrs of age)
  1.4 (women lt 34 yrs of age)
  3.2 (women gt 34 yrs of age) 1.4
• 1998 Doherty et al. Baltimore (lt 22 yrs of
  age) 4.8 (gt 22 yrs of age) 2.0

8
Whos at highest risk for HIV among IDUs?

• Younger age
• Short-term injecting career
• African American and Latino/a ethnicity
• MSM (minority subgroups)
• Female

9
Implementing a Research Design Community-Based
Approach

• Definition of Community-Based Research
• Epidemiologic research which involves participant
  recruitment, screening, enrollment, and data
  collection efforts performed within a
  non-institutional, community setting
  incorporating members of the community where the
  research is being conducted.
• Community-based clinic (STD clinic)
• Community-based organization
• Storefront-type setting in high-risk community

10
Traditional Sampling Methods

• Probability sampling
• Simple random sampling
• Stratified random sampling
• Cluster sampling
• Requires that all individuals in the total
  population sample have a known chance of
  selection into the study sample

11
Application of Traditional Sampling Methods in
Drug Use Research

• National Household Survey on Drug Abuse (NHSDA)
• Multi-stage probability sample
• Mail survey
• Monitor prevalence and incidence of drug use ?
  12 yrs of age ( 10 reporting illicit use)
• Monitoring for the Future (MTF)
• Probability sample
• School survey
• Monitor prevalence and incidence of drug use
  among 8th 12th grade students ( 1-2 reporting
  illicit use)

12
Community Epidemiology Work Group (CEWG)

• Community-level surveillance of drug abuse to
  determine nature and patterns of drug abuse and
  emerging trends
• A network composed of researchers from major
  metropolitan areas of the U.S. selected foreign
  countries
• Death certificates, Emergency Dept, treatment
  admissions, urinalysis (ADAM), drug seizures,
  price purity, distribution, etc. (DEA)
• Ethnographic research, focus groups, interviews,
  and other qualitative methods

13
Conducting Infectious Disease Research among
High-risk Populations

• Populations outside existing institutional and
  clinic settings - out of mainstream or
  hidden populations
• Populations within institutional settings
• Treatment settings
• Jails or prisons
• Emergency rooms
• STD clinics

14
Conducting Infectious Disease Research among
High-risk Populations

• No known sampling frame!

15
Conducting Infectious Disease Research among
High-risk Populations

• What happens when we use school surveys and
  household surveys?
• Major selections bias!!!
• Low proportions of illicit users attending school
• Low proportions of illicit users in households
  and may be less likely to report illicit use on a
  survey taken at home
• Homeless, transient, incarcerated, etc.

16
Conducting Infectious Disease Research among
High-risk Populations

• What happens when we use the institutional
  settings known to have large numbers of drug
  users (i.e., treatment programs)?
• Major selection bias!!
• Consecutive or convenience (grab) samples
• Enrollment of every individual entering a given
  institution in a stated study period
• Approximately 15 of all IDUs are in treatment at
  any given time
• Over-sampling of a more health-seeking
  population
• Exclusion of non-heroin IDUs

17
Conducting Infectious Disease Research among
High-risk Populations

• Sampling of hidden populations
• Street outreach methodology Community Health
  Outreach Worker (CHOW)
• Street-level trained staff, identify and
  penetrate drug user networks, identify high-risk
  neighborhoods, advocates and HIV educators
• Often ex-users, those who live within high-risk
  communities, and work in social services arena
• Late 1980s (San Francisco) and still used today
  nationally and internationally

18
Conducting Infectious Disease Research among
High-risk Populations

• What happens when we use street outreach
  convenience samples?
• Major selection bias!
• Non-treatment population
• Higher risk community of drug users
• Potential to obtain a more diverse population
• Drug type, duration of use, younger and older
  ages, in- and out-of-treatment, etc.

19
HIV Prevalence Rates among Treatment and
Community Samples

• 1989 (Metzger et al.) Philadelphia Treatment
  11.0
• Community 18.0
• 1990-91 (Robles et al.) San Juan Treatment 29.5
• Community 45.5
• (Alcabes et al.) Baltimore Treatment 27.3
• Non-treatment 24.3
• 1992-93 (Stark et al.) Berlin Treatment 3.9
• Community 20.7
• NEP 14.6

20
Sampling Methods Adapted to the Study of Hidden
Populations at High-risk for Blood-borne Pathogens

• Key-informant sampling
• Snowball or chain referral sampling
• Respondent-driven sampling
• Targeted sampling
• Capture-recapture sampling

21
Key-informant Sampling

• Key-informant Sampling
• Selecting knowledgeable respondents and asking
  them about behavior of others in their community
  and not about their own.
• In drug abuse research, such persons may include
  social workers, HIV or drug abuse counselors,
  public health workers, opinion leaders.
• Specifically designed to alleviate socially
  desirable responding which may arise in
  self-reports of sensitive information.

22
Potential Biases in Key Informant Sampling

• When professionals are used as key informants,
  their professional orientation may bias their
  responses.
• Opinion leaders may be better options
• Key Informants lack detailed knowledge of their
  peers.
• Key informants may not come in contact with a
  representative group of peers.
• Only useful in exploratory studies and collection
  of early pilot data.

23
Snowball Sampling

• Snowball or chain referral sampling
• A study sample based on referrals made among
  people who share or know of others who possess
  some characteristics that are of research
  interest. (Biernacki et al., 1981)
• Primarily used in sociologic (qualitative)
  research
• Allows for sampling of natural interactional
  units

24
Snowball Sampling

• Identify participants (opinion leaders or
  key-informants) to start the chain referral this
  is easily done with outreach workers or other
  study participants.
• Can accept participants who are part of a chain
  and those who are not.
• Important to keep track of index person who made
  the referral for later analysis on degree of
  clustering.
• May or may not want to divulge eligibility
  criteria to index persons.

25
Control the Types of Chains and Number of Cases
in each Chain

• Substantive consideration Once chains begin, an
  investigator may want to control the referral
  chains to help ensure that the general
  characteristics of the target population are
  reflected in the sample.
• Selective consideration May want to exclude or
  limit a certain subgroup (i.e., certain age
  group)
• Theoretical consideration What are some logical
  assumptions that can be made about certain groups
  (i.e., certain neighborhoods may use certain
  types of drugs and drugs purity may differ,
  etc.).

26
Generalizability of Snowball Sampling

• Can the findings be generalized to a wider
  population or is the sample limited?
• What is the extent to which snowball sampling is
  dependent on social networks and what types of
  networks?

27
Respondent Driven Sampling

• Respondent Driven Sampling (RDS)
• A type a chain referral sampling which involves
  an incentive system.
• Reduces some of the known biases of other chain
  referral methods
• Provides a means for controlling for the biases
  resulting from differences in the size of
  personal networks

28
Respondent Driven Sampling

• A researcher approaches and enrolls a participant
  who receives an incentive for undergoing the
  study interview.
• Same participant is also given an incentive for
  bringing in their peers as respondents.

29
Respondent Driven Sampling

• Two-step process to control sampling scheme
• 1) One or more members of the respondents group
  is promised a collective reward or threatened
  with a collective punishment based on whether the
  respondent complies.
• 2) Those members of the group respond to the
  secondary incentive by controlling the respondent
  group mediated control.
• RDS produces samples that are independent of the
  initial participants from which the sampling
  begins
• Does not matter whether the initial sample was
  chosen randomly or with known biases.

30
Targeted Sampling

• Targeted Sampling
• A purposeful, systematic method by which
  controlled lists of specified populations within
  geographical districts are developed, and
  detailed plans are designed to recruit adequate
  numbers of cases within each of the targets.
• This method uses aspects of street ethnography,
  theoretical sampling, stratified sampling, quota
  sampling, and chain referral sampling.
• (Watters et al., 1989)
• This method was designed to overcome deficiencies
  of chain referral models.

31
Targeted Sampling

• Two basic steps
• Field researchers (ethnographers) are deployed to
  map a target population
• Prevents the under-representation that often
  results from traditional sampling approaches.
• Field researchers (outreach workers) are deployed
  at diverse sites previously identified by the
  ethnographic mapping to recruit a known number of
  study participants.
• Further ensures that participants from different
  areas and subgroups will be better represented in
  the final sample.

32
Potential Biases in Targeted Sampling

• Recruiting participants during normal business
  hours
• Activities focused on more public locations
  (institutional bias may occur)

33
Targeted Sampling Additional Approaches

• What do we know about drug user/using settings
• Drug scenes are neither discrete nor public
• Drug activities (i.e., selling, purchasing,
  using) occur in concentrated areas
• Mostly dispersed occurring in private housing or
  vacant/abandoned buildings, or other nonpublic
  settings

34
Targeted Sampling Additional Approaches

• Extensive ethnographic approaches
• Weight drug scenes based on their intensity of
  use by each category of
• Drug user
• Time of the day
• Sample accordingly
• (could be random if numbers permit)

35
Capture-Recapture Methods

• Capture-Recapture
• A technique used to estimate the size of
  populations that are difficult to find and count,
  or populations that are in motion and can not be
  counted all at one time.
• (Sudman et al., 1988)

36
Capture-Recapture Methods

• Technique requires obtaining two or more
  independent observation on the same populations
  at the same times or based on different sources
  that represent the same population.
• The researcher needs to know three things
• of persons observed at the first time
• of persons observed at the second time
• Total of persons observed at both times
• Those captured must be tagged in some way so
  that it may be determined if they were included
  both captures.

37
Capture-Recapture Methods

• Assumptions
• Equal probability of being sampled during the
  study period
• Different individuals can be missed in the
  counting process for various reasons.
• Independence between counts observing an
  individual at one time has no effect on the event
  of observing the individual at a second.
• Being missed once can often be correlated with
  being missed a second time.
• Population size does not change in composition or
  size during the study period (closed population
  model).
• Population being studied changes between
  captures open population model need at least
  four captures.

38
What Happens in Real Life????

• Depends
• How hidden is the population?
• How much do you know as a researcher about the
  target population?
• Are there collaborations that can be developed to
  find out more about the target populations?

39
Whos at highest risk for HIV among IDUs?

• Younger and/or recently initiated IDUs!

40
Young, Recently Initiated IDUsBaltimore,
Maryland, 1997-1999REACH II Study

• Collaborative Injection Drug User Study (CIDUS
  II)
• Multisite Cohort Study
• Baltimore, MD
• Chicago, IL
• Los Angeles, CA
• New Orleans, LA
• New York, NY (Harlem)
• New York, NY (Lower East Side)

41
Recruitment Methodology

• Eligibility
• 15-30 years of age
• Injected at least once in past 6 months
• Injected 5 years
• Cohort study design
• Baseline, 6-month, 12-month follow-up visits

42
Sampling and Recruitment Methodology

• Convenience sample
• Modified target sampling
• Community street outreach
• Modified respondent driven / chain referral
• Enrolled each participant through a 2-stage
  screening process

43
Sampling and Recruitment Methodology

• Modified target sampling
• Community Advisory Board
• Social Services, Health Care Workers, CBO
  representatives, HIV and drug counselors, etc.
• Seasoned outreach staff
• Opinion leaders identified by outreach staff

44
Sampling and Recruitment Methodology

• Modified target sampling
• Identified 12 high-risk neighborhoods in
  Baltimore City known to be frequented by young
  IDUs
• Conducted mobile recruitment over a 2-year
  period 1997-1999 using a study van

45
Sampling and Recruitment Methodology

• Parked the REACH II Van in 4 different sites per
  week and revisited those same 4 sites each week
• Once a site dried up, we went to the next site
• Ensured that each site was utilized during summer
  months
• Extended hours in certain sites based on
  visibility of young users
• Goal Enroll every young IDU in Baltimore City!

46
Sampling and Recruitment Methodology

• Community street outreach
• Outreach staff went out and recruited potential
  participants within a 3-block radius of the
  mobile site
• Posted and distributed flyers on abandoned
  buildings, ER, homeless shelters, NEP, STD
  Clinics, treatment centers, etc.
• Conducted stage one of screening process

47
Sampling and Recruitment Methodology

• Modified respondent driven / chain referral
• Identified participants who had additional access
  to other young IDUs and gave incentives to the
  participant for bringing in young users.
• Shooting gallery managers
• Other young IDUs (lt 25 years of age)

48
Sampling and Recruitment Methodology

• Keep track of how each participant entered the
  study!
• Selection bias can be assessed and considered
  when interpreting study findings.

49
Study Findings

• Screened 276 young IDUs
• 82 were eligible and enrolled (n226)
• 61 female!
• 64 Black
• 11 HIV seropositive
• 65 HCV seropositive

50
REACH II Study Findings

• HIV seroincidence 6.6 per 100 PY
• HCV seroincidence 22.3 per 100 PY
• Extremely high-risk population!