Behavioural Surveillance Surveys - The Wisdom of Whores
Behavioural Surveillance Surveys - The Wisdom of Whores Behavioural Surveillance Surveys - The Wisdom of Whores
Using the chi-square test for trend, it turns out that both trend lines are statistically significant. The chi-square test statistic for the trucker’s condom use trend is 67.84, with a p-value of
Changes in sampling methodology over time are almost certain to result in selection bias, making trends over time much more difficult to interpret than when a consistent methodology is used. Even if sampling methodology remains the same, changes in the political or social climate may have important influences on who is included in a survey. A crackdown on illegal immigrants may completely change the profile of the sex worker population between one survey and the next, for example. Differences in risk behavior over time may not reflect behavior change in this case, but may simply mean the population measured is no longer the same. It should be noted that rapid population turnover is typical of many of the sub-populations of interest to HIV prevention programmes. As long as the population sampled relates in the same way to the wider population to which results will be extrapolated, this will not increase sampling error. It will, however, affect the interpretation of trends over time. One important potential source of selection bias is refusal bias. BSS asks people about personal and often illegal behaviors. It is absolutely essential, therefore, that the purpose of the study be carefully explained to selected respondents, and that their full consent to participate be obtained before questioning begins. Refusal bias arises when those who refuse to participate have different behaviors than those who agree. In the case of BSS, refusal bias may well underestimate true levels of risk behavior, because people may avoid participating because they do not want to admit to behaviors that they recognize are risky. Survey reports should always state what proportion of the selected sample refused participation, and should give socio-demographic profiles of refusers where available. One way of attempting to gauge the importance of selection bias is to collect some basic socio-demographic variables from selected participants, including those who refuse to participate in the study. These variables can be compared with the same variables in respondents in earlier survey rounds (as well as in those who refuse to take part) to see if any systematic differences are discernable. Variables that will help judge the likely magnitude of selection bias will depend on the local situation but may include nationality, province of origin, length of time associated with the site, etc. Measurement bias Another source of potential error in surveys is error in the measurement of the variables. This can arise when survey forms are not clear, for example when the meaning of a question in the local language is open to different interpretations. Interviewer bias can also affect the correct measurement of behavior. Some interviewers are more judgmental than others, and people may be more or less willing to report stigmatized behaviors to an interviewer depending on their attitude. Indeed people’s unwillingness to tell the truth about their sexual or drug-taking behavior may be the most important source of measurement bias. It is not clear to what extent this is likely to change over time. Where an interviewer has to turn a verbal answer into one that fits in to a coded category, they may also influence results, by consistently preferring one category over another when answers are ambiguous. Many of the difficulties associated with measurement bias can be minimized through comprehensive training of interviewers and survey staff, and active supervision. 84 C H A PTER 7 B EHAV I OR A L S U R V EI L L A NC E S U R V EY S
- Page 43 and 44: In looking at behaviors of hard-to-
- Page 45 and 46: Maps derived from program planning
- Page 47 and 48: Selecting primary sampling units (c
- Page 49 and 50: ...when measures of size are not av
- Page 51 and 52: If there is no reason to believe th
- Page 53 and 54: Figure 3 : Decision tree for first-
- Page 55 and 56: Implications of alternative samplin
- Page 57 and 58: Table 5 : Values of Z 1-α and Z 1-
- Page 59 and 60: Table 6 : Sample size requirements
- Page 61 and 62: equired. In this case, the pros and
- Page 63 and 64: Should one- or two-tailed z-score v
- Page 65 and 66: Other measurement issues for BSS Th
- Page 67 and 68: A “low-tech” solution to the pr
- Page 69 and 70: 5 Weighting in multi-stage sampling
- Page 71 and 72: Figure 6 : Procedures for calculati
- Page 73 and 74: Calculating weights from sampling p
- Page 75 and 76: Calculating standard errors with mu
- Page 77 and 78: 6 Adapting and using questionnaires
- Page 79 and 80: Some attempts have in the past been
- Page 81 and 82: Informed consent Confidentiality an
- Page 83 and 84: GUIDELINES FOR REPEATED BEHAVIORAL
- Page 85 and 86: Recommended Methods of Statistical
- Page 87 and 88: where n is the sample size in the d
- Page 89 and 90: Table 2 : Reported number of non-re
- Page 91 and 92: False conclusions: the danger of co
- Page 93: Analysis of trends in behavior over
- Page 97 and 98: 8 Using the data collected to impro
- Page 99 and 100: Improving prevention programs As a
- Page 101 and 102: Finally, information about HIV and
- Page 103 and 104: Figure 9 : HIV and STD prevalence,
- Page 105 and 106: 9 Indicators This guide identifies
- Page 107 and 108: The standard questionnaires are acc
- Page 109 and 110: INDICATORS Page INDICATORS FOR YOUT
- Page 111 and 112: INDICATORS Page INDICATORS FOR MEN
- Page 113 and 114: KEY INDICATORS What follows is a fu
- Page 115 and 116: The indicator uses promoted data. O
- Page 117 and 118: To be counted the numerator for thi
- Page 119 and 120: Strengths and limitations Some meas
- Page 121 and 122: Adult Indicator 5 Consistent condom
- Page 123 and 124: Adult Indicator 7 (men only) Number
- Page 125 and 126: Adult Indicator 9 Consistent condom
- Page 127 and 128: Adult Indicator 11 Population seeki
- Page 129 and 130: Adult Indicator 12 Exposure to inte
- Page 131 and 132: Youth Indicator 2 No incorrect beli
- Page 133 and 134: Youth Indicator 4 Youth sexually ac
- Page 135 and 136: Youth Indicator 6 Number of sexual
- Page 137 and 138: Youth Indicator 8 Consistent condom
- Page 139 and 140: Youth Indicator 10 Commercial sex a
- Page 141 and 142: Youth Indicator 14 Exposure to inte
- Page 143 and 144: Female Sex Worker Indicator 2 No in
Changes in sampling methodology over<br />
time are almost certain to result in selection<br />
bias, making trends over time much more<br />
difficult to interpret than when a consistent<br />
methodology is used.<br />
Even if sampling methodology remains the<br />
same, changes in the political or social climate<br />
may have important influences on who is<br />
included in a survey. A crackdown on illegal<br />
immigrants may completely change the pr<strong>of</strong>ile<br />
<strong>of</strong> the sex worker population between one<br />
survey and the next, for example. Differences<br />
in risk behavior over time may not reflect<br />
behavior change in this case, but may simply<br />
mean the population measured is no longer<br />
the same. It should be noted that rapid<br />
population turnover is typical <strong>of</strong> many <strong>of</strong> the<br />
sub-populations <strong>of</strong> interest to HIV prevention<br />
programmes. As long as the population<br />
sampled relates in the same way to the wider<br />
population to which results will be<br />
extrapolated, this will not increase sampling<br />
error. It will, however, affect the interpretation<br />
<strong>of</strong> trends over time.<br />
One important potential source <strong>of</strong> selection<br />
bias is refusal bias. BSS asks people about<br />
personal and <strong>of</strong>ten illegal behaviors. It is<br />
absolutely essential, therefore, that the purpose<br />
<strong>of</strong> the study be carefully explained to selected<br />
respondents, and that their full consent to<br />
participate be obtained before questioning<br />
begins. Refusal bias arises when those who<br />
refuse to participate have different behaviors<br />
than those who agree. In the case <strong>of</strong> BSS,<br />
refusal bias may well underestimate true levels<br />
<strong>of</strong> risk behavior, because people may avoid<br />
participating because they do not want to<br />
admit to behaviors that they recognize are<br />
risky. Survey reports should always state what<br />
proportion <strong>of</strong> the selected sample refused<br />
participation, and should give socio-demographic<br />
pr<strong>of</strong>iles <strong>of</strong> refusers where available.<br />
One way <strong>of</strong> attempting to gauge the<br />
importance <strong>of</strong> selection bias is to collect some<br />
basic socio-demographic variables from selected<br />
participants, including those who refuse to<br />
participate in the study. <strong>The</strong>se variables can be<br />
compared with the same variables in respondents<br />
in earlier survey rounds (as well as in those who<br />
refuse to take part) to see if any systematic<br />
differences are discernable. Variables that will<br />
help judge the likely magnitude <strong>of</strong> selection<br />
bias will depend on the local situation but<br />
may include nationality, province <strong>of</strong> origin,<br />
length <strong>of</strong> time associated with the site, etc.<br />
Measurement bias<br />
Another source <strong>of</strong> potential error in surveys<br />
is error in the measurement <strong>of</strong> the variables.<br />
This can arise when survey forms are not<br />
clear, for example when the meaning <strong>of</strong> a<br />
question in the local language is open to<br />
different interpretations.<br />
Interviewer bias can also affect the correct<br />
measurement <strong>of</strong> behavior. Some interviewers<br />
are more judgmental than others, and people<br />
may be more or less willing to report<br />
stigmatized behaviors to an interviewer<br />
depending on their attitude. Indeed people’s<br />
unwillingness to tell the truth about their<br />
sexual or drug-taking behavior may be the<br />
most important source <strong>of</strong> measurement bias.<br />
It is not clear to what extent this is likely to<br />
change over time.<br />
Where an interviewer has to turn a verbal<br />
answer into one that fits in to a coded category,<br />
they may also influence results, by consistently<br />
preferring one category over another when<br />
answers are ambiguous.<br />
Many <strong>of</strong> the difficulties associated with<br />
measurement bias can be minimized through<br />
comprehensive training <strong>of</strong> interviewers and<br />
survey staff, and active supervision.<br />
84<br />
C H A PTER 7 B EHAV I OR A L S U R V EI L L A NC E S U R V EY S