appendix 2.1:rds and tls an overview of probability-based sampling methods for key populations
TRANSCRIPT
Appendix 2.1:RDS and TLSAn Overview of Probability-Based Sampling
Methods For Key Populations
Session Overview Discussion of Sampling Hard to Reach
Populations Presentation, comparison, and discussion of:
Time Location Sampling (TLS) Respondent Driven Sampling (RDS)
How to Reach Hard to Reach Populations ….. Let’s Discuss!
What Makes RDS and TLS Unique for Sampling Key Populations?
They are probability-based sampling methods
Every respondent has a non-zero, known probability of being selected to participate in the study
With weighting, the sample can be made representative of the target population Characteristics of the sample are valid estimates
of the characteristics of the target population
Time Location Sampling (TLS) The method is known by several names
Venue Day Time (VDT) sampling Temporal-spatial sampling (TSS) Time Venue Sampling (TVS) Variation of Targeted Sampling (TS) Venue-Based Sampling (VBS)
The idea is to sample the target population at randomly selected venues where members of the population are known to congregate
Basic Principles of TLS Approximates random cluster sampling
1. Target population is divided into clusters or groups2. Groups of individuals are randomly selected for
sampling3. Individuals in selected groups are randomly
sampled
In TLS, the sampling clusters are known as venue-day-times (VDTs). VDTs are the: Places Day of the Week And times where target population can be found
Basic Steps for TLS1. Identify the universe of venues-day-time
(VDT) periods attended by the target population
2. Build a sampling frame of VDTs3. Randomly select VDTs for recruitment events4. Systematic Sampling: Intercept, consent,
interview, and VCT at event5. Data management6. Analysis, interpretation7. Use the data!
Venues-Day-Time (Cluster Sampling Frame)
Name Address
Phone
Mon Tues Wed Thurs Fri Sat Sun
Power Fitness
2230 18th St
6-10pm
6-10pm
West End Video
1984 West End Place
234-2390
8pm-12am
8pm-12am
8pm-12am
8pm-12am
Café Noir
28 Sheppard
834-4823
6-10pm
6-10pm 6-10 pm 6-10 pm10pm-12am
6-10 pm10 pm-12am
4 -8pm
Men’s Choir
1438 Oak St.
N/A 8-9pm
Creating a complete universe of venues-day-times is a lot of work!
Assumptions of TLS “Map universe” of all the places where and
when the target population can be found Randomly sampling enough places and times
provides everyone in the target population equal chance of being in study
If not equal chance, there are methods to “weight” sample according to who are more or less likely to be at the venues
Ethical Challenges in TLS Ensuring anonymity Returning results, ensuring appropriate care Drawing unwanted attention to safe havens
(police, public) Intoxicated participants Balancing random sample with outreach and
prevention
Chain-Referral (Snowball) Sampling Recruitment through a network
Participants recruit individuals from their personal network to participate
Strength: Reaches respondents who avoid public venues and
institutions May have greater coverage because respondents are
reached through their social networks Weakness:
This is a convenience sample rather than a probability sample
Characteristics of the sample are NOT valid estimates of the characteristics of the target population
Respondent-Driven Sampling (RDS) RDS combines sampling coverage and
probability sampling methods Combines a modified form of chain-referral with a
mathematical system for weighting the sample to compensate for its not having been drawn randomly
Based on the premise that peers are better able than outreach workers and researchers to locate and recruit other members of a hidden population
Unlike other chain referral methods, gives valid population point estimates with standard errors
Wave 1 Wave 2 Wave 3 Wave 4 Wave 5
Controlled Recruitment = Penetration / Sample Size Growth
Wave 1 Wave 2 Wave 3 Wave 4 Wave 5
Controlled Recruitment = Penetration / Sample Size Growth
Wave 1 Wave 2 Wave 3 Wave 4 Wave 5
Controlled Recruitment = Penetration / Sample Size Growth
Wave 1 Wave 2 Wave 3 Wave 4 Wave 5
Controlled Recruitment = Penetration / Sample Size Growth
Controlled Recruitment = Penetration / Sample Size Growth
Wave 1 Wave 2 Wave 3 Wave 4 Wave 5
Controlled Recruitment = Penetration / Sample Size Growth
Wave 1 Wave 2 Wave 3 Wave 4 Wave 5
RDS WeightingRDS has two types of weighting: Recruitment pattern weighting
Used to adjust for differential recruitment between groups
Network size weighting Well-connected individuals tend to be over-
sampled because many recruitment paths lead to them
Requirements of RDS Four requirements:
Document who recruited whom Recruiter and recruit must know one another Ration recruitment so a few cannot do all the
recruiting (i.e., three recruits/recruiter) Must ask recruiter and recruit about personal
network sizes
If a study does not include these features, it is not RDS
Steps in RDSIdentify, Recruit
and Interview Seeds
Provide seeds info on who and how to
recruit
Give 3 - 5 coupons to each seed
Recruits bring valid coupons to the study site; If eligible they
are recruited
Every new recruit is then
asked and given 3 - 5 coupons
The recruiter is rewarded for every coupon
redeemed
“Coupon Manager” tracks coupons/links recruiters and peers
RDS Assumption (Heckathorn)1. Individuals know each other as members of
the population in question2. Target population forms one single large
network3. Report network size accurately4. Recruit randomly from network5. Sampling with replacement
Advantages of RDS Maintains privacy of population Team can be centrally located Less logistical needs Ease of field operations Moderate formative research/mapping Target members recruit for you Reach less visible segments of population Good external validity Minimal number of additional questions needed Computer software available Generally lower cost
Challenges of RDS Coupons can be slow in being redeemed Population must be networked Must be able to verify group membership Must track links between recruiters and
recruits-coupon management Appropriate incentive levels Very difficult to deal with selective non
response bias Analysis is complex
Breakout (20 minutes): TLS vs. RDSWhich methodology might be more appropriate for the
target population here in the Bahamas?a) Do individuals in the target population frequent venues?b) Could the target population be identified at these venues?c) Could the target population be accessed at these venues
without drawing them unwanted attention?d) Might we miss some population groups at these venues?e) Are individuals in the target population well-networked?f) Are there distinct sub-populations?g) Do sub-populations interact with each other?h) Can members of the target population identify each
other?i) Will the target population be motivated to participate?
Thank You
Working Together to Plan, Implement, and Use
HIV Surveillance Systems