Unit 3: Sample Size, Sampling Methods, Duration and Frequency of Sampling

1
Unit 3 Sample Size, Sampling Methods, Duration
and Frequency of Sampling
3-3-1
2
Warm Up Questions Instructions

• Take five minutes now to try the Unit 3 warm up
  questions in your manual.
• Please do not compare answers with other
  participants.
• Your answers will not be collected or graded.
• We will review your answers at the end of the
  unit.

3-3-2
3
What You Will Learn

• By the end of this unit you should be able to
• explain the process of sampling
• determine who is eligible for a sample
• choose the appropriate sampling scheme for
  specific situations
• plan for the duration and frequency of sampling
  to be conducted

3-3-3
4
The Sampling Process

• Appropriate and consistent sampling methods
  ensure that sero-surveys are representative of
  populations in which they are conducted.
• Representative samples resemble the true
  population.
• They ensure that surveys are consistent from year
  to year, and are consistent among sites.

3-3-4
5
Components of the Sampling Process

• Sample size, or number of individuals to include
  in the sample
• Sampling scheme, or procedures for choosing
  individuals to be included in the sample
• Duration and frequency of sampling, how long to
  sample and how often to sample

3-3-5
6
Sample Size

• The number of subjects included in a sentinel
  surveillance sample is guided by the need to
• determine trends in HIV prevalence over time
• identify sub-populations at high risk for
  infection
• Sample size needs to be large enough to be able
  to detect the difference between two prevalence
  estimates.
• Statistically, this is referred to as the margin
  of error (for example, 3).

3-3-6
7
Sample Size, Cont.

• The standard statistical approach for determining
  the sample size requires
• estimate of HIV prevalence
• the margin of error considered acceptable
• the level of confidence desired
• Typically, sample size calculations are done by a
  countrys national AIDS control programme
• Minimum sample sizes should be calculated for
  each sentinel site for each survey period.

3-3-7
8
Calculating Sample Size with Epi Info

• The STATCALC feature of Epi Info software
  provides a user-friendly sample size calculator
  for setting specific target sample sizes.
• Learn more about Epi Info and download the
  software for free at this site
  http//www.cdc.gov/epiinfo

3-3-8
9
Determining Sample Size in Practice

• A minimum sample size between 250 and 400
  patients per site is recommended.
• A margin of error of 3 and 95 statistical
  confidence can be reached.
• This is the average sample size that can be
  obtained in a typical clinic over 8 to 12 weeks.
• This provides sufficient statistical power to
  determine trends in HIV prevalence over a 3-year
  period.

3-3-9
10
Figure 3.1. Formula to Determine Sample Size

• z? is a factor that corresponds to the desired
  confidence interval (for a 95 confidence level,
  z? 1.96)
• P is the expected proportion of patients with the
  outcome of interest (in this case, P expected
  HIV prevalence).
• W is the width of the interval, for example the
  width for a margin of error of /- 3 is 0.06.

3-3-10
11
Sample Size Rules of Thumb

• The closer the estimated prevalence is to 50 the
  larger the sample size that will be required to
  achieve the same confidence interval width.
• The sample size needed gets larger as the desired
  margin of error gets smaller.
• A sample size must be practical to achieve in the
  course of a few months.

3-3-11
12
Sampling Schemes

• There are three basic sampling schemes that are
  used at sentinel sites
• consecutive sampling
• systematic sampling
• simple random sampling
• Each scheme has advantages disadvantages.

3-3-12
13
Consecutive Sampling

• In consecutive sampling, select every patient
  that meets the inclusion criteria until you get
  to the required sample size or the survey period
  is over.
• Example Include every ANC patient that meets the
  inclusion criteria that comes into the ANC clinic
  during the month of January.

3-3-13
14
Systematic Sampling

• In systematic sampling, make a list of every
  patient who meets the inclusion criteria. Select
  every nth person from list until required sample
  size is reached or survey period is over.
• Example Make a list of all ANC attendees that
  meet the inclusion criteria, and select every
  third patient until you have 256 participants or
  the survey period is over.

3-3-14
15
Simple Random Sampling

• In simple random sampling, generate a list of
  random numbers. Match numbers to a list of
  patients who meet study criteria until the
  required sample size is met.
• Lists of random numbers can be taken from random
  number tables, or can be generated by computer or
  other methods.

3-3-15
16
Table 3.1. Advantages and Disadvantages of
Consecutive Sampling

• Advantages
• relatively easy to employ
• less opportunity for intentional or unintentional
  manipulation by clinic staff or errors due to
  confusion

• Disadvantages
• not based on randomisation
• may be variations in who is seen at ANCs at
  different times of the year (for example during
  the rainy season women may have problems coming
  from rural areas to the clinic)

3-3-16
17
Table 3.1. Advantages and Disadvantages of
Systematic Sampling

• Advantages
• more likely than consecutive sampling to produce
  representative sample if done correctly

• Disadvantages
• more difficult to do correctly
• requires more attention to procedural details
• non-random selection of the first patient or
  failure to sample every nth patient will make
  sample non-random
• method may also take longer to fill sample size

3-3-17
18
Table 3.1. Advantages and Disadvantages of Simple
Random Sampling

• Advantages
• most likely to produce a true sample of the total
  population if done correctly

• Disadvantages
• complex, prone to errors and confusion, and may
  take longer to conduct
• requires a census or numbered list of potential
  survey participants

3-3-18
19
Duration of Surveys

• HIV prevalence in a given area is likely to
  change over time.
• Ideally, HIV prevalence is measured at a single
  point in time, in a point estimate.
• Practically, point estimates are usually
  impossible to obtain.
• In order to best approximate a point estimate,
  sampling time period should be limited.

3-3-19
20
Duration of Surveys, Cont.

• Give health workers in clinics precise dates on
  when to begin and end time-period for specimen
  collection.
• Many countries set period at or below 10 weeks.
• If clinic volume is low, sampling period may be
  longer.
• One requirement for inclusion of a site is the
  ability to fill sample size requirements in
  allotted sampling period.

3-3-20
21
Frequency of Surveys

• Select the sampling period as the same time each
  year (for example, 1 August through 30 September)
  and in the same sites.
• This ensures data are comparable from year to
  year, and place to place.
• To allow sufficient time for collation, analysis,
  interpretation and report writing, sero-surveys
  should be repeated every 1-2 years.

3-3-21
22
Frequency of Surveys, Cont.

• More frequent surveys do not usually contribute
  additional information that is useful for public
  health decisions.
• Among the most important considerations in
  determining the frequency of surveys is the
  availability of sufficient resources.

3-3-22
23
Inclusion and Exclusion Criteria

• Inclusion criteria characteristics required in
  study participants, in order to be considered for
  sample.
• Exclusion criteria characteristics of patients
  who should be excluded from sample, but who would
  otherwise be eligible.
• Goals of setting criteria
• minimising potential survey biases
• avoiding inclusion of a person multiple times

3-3-23
24
Table 3.2. Sample Inclusion and Exclusion
Criteria for ANC Surveys

• Inclusion Criteria
• age (only women 15 to 49 years old should be
  eligible for the sample)

• Exclusion Criteria
• has previously attended ANC during current
  pregnancy
• attendance at another ANC

3-3-24
25
Warm Up Review

• Take a few minutes now to look back at your
  answers to the warm up questions at the beginning
  of the unit.
• Make any changes you want to.
• We will discuss the questions and answers in a
  few minutes.

3-3-25
26
Answers to Warm Up Questions

• 1. Which of the following is a reason to have
  inclusion and exclusion criteria?
• a. Include as few participants as possible in the
  survey
• b. avoiding including the same person multiple
  times
• c. both of the above

3-3-26
27
Answers to Warm Up Questions, Cont.

• 2. True or false? The goal of sampling is to use
  the data from a sample of a population to
  estimate the HIV prevalence in the larger
  population. True

3-3-27
28
Answers to Warm Up Questions, Cont.

• 3. Which of the following is a decision that
  needs to be made at the beginning of a sampling
  procedure?
• a. the sample size
• b. the sampling scheme
• c. the frequency of sampling
• d. all of the above

3-3-28
29
Answers to Warm Up Questions, Cont.

• 4. True or false? As much as possible, the
  sampling period should be limited, in order to
  compare HIV prevalence over time. True

3-3-29
30
Answers to Warm Up Questions, Cont.
5. Match each sampling scheme with its
description

• _c_ consecutive
• _a_ systematic
• _b_ simple random

1. randomly selects the initial patient who meets
   inclusion criteria, and then selects every nth
   eligible patient thereafter
2. uses a computer or other method to generate
   random numbers that identify patients to be
   included in the sample
3. samples every patient that meets the inclusion
   criteria until the required sample size is
   achieved

3-3-30
31
Answers to Warm Up Questions, Cont.

• 6. Which of the above schemes is the most simple
  logistically, and best reduces the likelihood of
  selection bias? Consecutive

3-3-31
32
Answers to Warm Up Questions, Cont.

• 7. True or false? All subjects at the sentinel
  site who meet the inclusion criteria during the
  sampling period should be included in the survey.
  False

3-3-32
33
Answers to Warm Up Questions, Cont.

• 8. True or False? When surveys are repeated, they
  should be carried out in different sites from the
  initial survey and during a different time of the
  year. This helps give a clear picture of the
  epidemics scope. False

3-3-33
34
Small Group Discussion Instructions

• Get into small groups to discuss these questions.
  
• Choose a speaker for your group who will report
  back to the class.
• Take 15 minutes for this exercise.

3-3-34
35
Small Group Reports

• Select one member from your group to present your
  answers.
• Discuss with the rest of the class.

3-3-35
36
Case Study Instructions

• Try this case study individually.
• Well discuss the answers in class.

3-3-36
37
Case Study Review

• Follow along as we go over the case study in
  class.
• Discuss your answers with the rest of the class.

3-3-37
38
Questions, Process Check

• Do you have any questions on the information we
  just covered?
• Are you happy with how we worked on Unit 3?
• Do you want to try something different that will
  help the group?

3-3-38