FLUCTUATION IN SEXUAL ACTIVITY, THE VALIDITY OF SEXUAL BEHAVIOUR ESTIMATES FOR SHORT TIME INTERVALS, AND HIV INTERVENTION EVALUATION IN RURAL ZIMBABWE.

Data on sexual behaviour over short time intervals are needed in HIV-related research for behavioural surveillance purposes (Amon et al., 2000), particularly where attempts are made to detect changes in behaviour resulting from intensive HIV prevention initiatives. In these and other circumstances, data on short time periods are also desirable as they are generally less affected by recall bias (Enel, Lagarde, & Pison, 1994).

However, these data are subject to a number of limitations. For example, the smaller numbers of events recorded over shorter time periods will tend to restrict the statistical power to detect differences in behavior between populations subgroups and temporal changes in behaviour, for any given sample size. A further problem is that bias can result from differences in the timing of interview dates where there are systematic or random temporal fluctuations in patterns of behaviour. This bias could distort comparisons of behaviour made between groups enumerated at slightly different time points and can obscure the true pattern of behaviour change that occurs between rounds in a longitudinal survey. The latter, in particular, would have serious implications for behavioural surveillance and, thus, for evaluation of HIV control interventions. At the population level, systematic fluctuations in behaviour unrelated to interventions could result in incorrect references being drawn as to the extent and direction of behaviour change. At the individual level, both systematic and random fluctuations would tend to obscure genuine changes in behaviour.

In recent studies in rural Zimbabwe (Gregson, Zhuwan, Anderson, & Chandiwana, 1998; Gregson et al., 2001), we noted a number of examples of periodic and irregular events that affected the wider ability of data on sexual behaviour collected for short time intervals prior to the interview date.

Sexual activity among men in formal sector employment was strongly correlated with time since last payment (see Table). Thus, those who had received their wages within the two weeks prior to interview were more likely to report sexual activity within that period than those who had not (OR, 1:35; p = 0.007). Married men living apart from their wives and single men resident in compounds on large-scale farming estates were also less likely to report any recent sexual activity. However, in focus group discussions and pocket chart voting exercises (Gregson et al., 1998), these men reported engaging in casual sexual relationships at the end of the month when they received their pay and were able to visit neighbouring towns and trading centres.

Table. Determinants of Any Sexual Activity Within the Last 2 Weeks
Among Employed Men

                                       All men        Married men
Respondent characteristics           OR       p      OR       p

Paid in the last 2 weeks             1.35   0.007    1.31   0.062
Married                             11.65   0.000     -       -
Cohabiting with a marital partner    2.71   0.000    2.83   0.000
Resident on an estate                0.74   0.010    0.89   0.464

Education level
 Higher                              2.21   0.010    2.87   0.009
 Secondary                           1.30   0.040    1.25   0.760
 Primary/none                        1.00            1.00

Age
 15-19                               0.63   0.077    0.46   0.257
 20-24                               0.97   0.857    1.03   0.899
 25-34                               0.98   0.920    0.95   0.760
 35-49                               1.00            1.00

N                                   1,769           1,070

                                    Unmarried men
Respondent characteristics           OR       p

Paid in the last 2 weeks             1.43   0.045
Married                               -       -
Cohabiting with a marital partner     -       -
Resident on an estate                0.61   0.004

Education level
 Higher                              1.41   0.553
 Secondary                           1.37   0.130
 Primary/none                        1.00

Age
 15-19                               0.74   0.485
 20-24                               1.10   0.812
 25-34                               1.19   0.656
 35-49                               1.00

N                                     688

Note. Effects of recent illness, other marital status, other places
of residence, interview method & HIV status were tested but were not
significant (p > 0.05).

We also observed that high-risk sexual activity for the spread of HIV infection was more intensive during public holidays when many male migrant workers return to their rural homes (at Christmas, often with wage bonuses). In contrast, some local religious groups practice sexual abstinence during religious festivals. For example, members of the African Apostolic Church of Johane Marange abstain from sex for a period of 1 month around the time of the Passover celebration. Thus, many polygynous men from this church reported no sexual partners in the previous 2 weeks when interviewed at this time.

Fluctuations in sexual activity can also be associated with irregular events: For example, a group of married men in Rusitu Valley gave traditional pregnancy-associated sexual abstinence on the part of their wives as a reason for having extramarital relationships and there was some evidence that these periods of abstinence were being curtailed in the face of the AIDS epidemic, as women sought to counter this (Gregson et al., 1998).

In each of these examples, individuals interviewed during a "high-risk period" for extramarital sexual activity would clearly be more likely to report such activity than those interviewed at other times. In longitudinal studies, individuals interviewed initially during a low-activity period could appear to have adopted safer behaviour when this is not in fact the case, and vice versa.

It is important, therefore, that close attention is paid to the possible extent and implications of bias in data on recent sexual behaviour at the design, data collection, and data analysis stages of behavioural and behavioural surveillance studies. Where feasible, interviews should be scheduled at comparable times of the year/month depending upon the exact time intervals for which data are required. Consideration could also be given to the inclusion of a question on time since last payday within questionnaire schedules so that this can be controlled for during data analysis.

REFERENCES

Amon, J., Brown, T., Hogle, J., MacNeil, J., Magnani, R., Mills, S., Pisani, E., Rehle, T., Saidel, T., & Sow, C. K. (2000). Behavioral Surveillance Surveys (BSS): Guidelines for repeated behavioral surveys in populations at risk for HIV. Arlington, VA: Family Health International.

Enel, C., Lagarde, E., & Pison, G. (1994). The evaluation of surveys of sexual behaviour: A study of couples in rural Senegal. Health Transition Review, 4 (Suppl.), 111-124.

Gregson, S., Nyamukapa, C., Garnett, G. P., Mason, P. R., Zhuwau, T., Careal, M., Chandiwana, S. K., & Anderson, R. M. (2001). Sexual mixing patterns and sex-differentials in teenage exposure to HIV infection in rural Zimbabwe. Manuscript submitted for publication.

Gregson, S., Zhuwau, T, Anderson, R. M., & Chandiwana, S. (1998). Is there evidence for behaviour change in response to AIDS in rural Zimbabwe? Social Science and Medicine, 46, 321-330.