A key ingredient in preventing HIV/AIDS is condom programming. The three principal ways in which condoms are distributed are through government programs, subsidized social marketing programs, and the commercial sector. Government programs typically involve donated products that are distributed through the established government health infrastructure. Under this approach, condoms are distributed free of charge through government hospitals, clinics, and rural health posts. The reach of public-sector programs is frequently extended through the use of outreach workers or distribution to key target groups, such as the military or schools.

Most countries with severe HIV/AIDS problems have also developed social marketing programs to provide supplemental means of condom distribution. These programs typically use subsidized products that are distributed primarily through traditional commercial outlets, such as pharmacies and small shops. Non-traditional commercial vendors, however, are increasingly being used to provide access to harder-to-reach groups and at venues closer to where sexual encounters are likely to take place. Less traditional outlets include bars, hotels, vending machines, taxi drivers, truck stops, gas stations, and kiosks.

Social marketing programs use the tools and techniques of commercial marketing to promote products having a social benefit, in this case condoms for AIDS prevention. The framework most commonly used to think about the various program components is the four Ps marketing-mix framework. The four Ps comprising this framework are products, place (distribution), promotion, and price. Promotion, a major component of most condom social marketing programs, includes mass media advertising, public relations activities, trade promotions, and provider training. Because the programs are subsidized, social marketing prices fall in between the free government products and full commercial prices. Prices are typically set to be appropriate for middle to lower-middle socioeconomic (or C and D class) target markets. Finally, market research is typically used to design, monitor, and evaluate the various components of the program1.

A third, less-developed approach for condom programming involves encouraging the commercial sector to become more actively engaged in promoting condom availability.

Commercial distributors do not make much profit on condom sales, so innovative ways, such as co-marketing, offering customer service training, or public relations events and contests, are often used to gain their participation.

Condom use has been dramatically increasing throughout the developing world. In Kenya, for example, condom distribution has increased from less than 10 million, in the mid-1980s, to more than 50 million, by the mid-1990s. Today, worldwide condom use could be as high as 600 million per year. To better understand the impact of condom programming on health outcomes and on preventing HIV/AIDS, it is important to evaluate how condoms are being used. Toward that aim, the following key questions are the most important to consider:

• How many condoms are being distributed?
• Who is using condoms?
• With which partners are people using condoms?
• With what consistency are people using condoms?
• What are the barriers to condom use?
• How effectively are condoms being used?
• How cost-effective are different condom distribution programs?

This chapter examines various methodologies and techniques that can be used to answer these questions.

The Use Of Indicators In Measuring Condom Program Effectiveness

In the early 1990s, WHO/GPA developed a set of Priority Prevention Indicators (PIs) to assist National AIDS Control Programmes evaluate their country programs. A few of these indicators dealt specifically with condom use and availability. WHO/GPA field-tested these indicators and prepared procedure manuals to guide country implementation. However, these indicators were not widely adopted for routine use by country programs.

Subsequently, under USAID's EVALUATION Project, another set of indicators for HIV and sexually transmitted infections (STIs) were developed and documented as a part of a handbook on reproductive health indicators2. Recently, two new efforts prepared handbooks of indicators for HIV/AIDS/STI programs. One effort was led by USAID's MEASURE/Evaluation Project, the other by UNAIDS. Both used a series of working groups to develop, refine, document, and field test the indicators.

Table 5-1 shows the recommended UNAIDS indicators that are specific to condom programming for AIDS at the national level3. As in the previous cases, the working groups have developed protocols for assessing these indicators.

Other programs have also used condom service statistics or sales, consistency of condom use, use effectiveness, attitudes and misconceptions regarding condoms, and matching of target markets with actual user profiles as additional indicators of program performance.

Evaluating Condom Use, Knowledge, And Attitudes Toward Condoms, And User Characteristics

According to the UNAIDS manual on monitoring and evaluating national HIV/AIDS prevention programs, "indicators of sexual behavior and condom use are probably the most important

of all indicators in monitoring HIV prevention programmes and evaluating their success." These indicators are best assessed through surveys. The following sections discuss various types of surveys and what information is best obtained from each.

Baseline and Tracking Surveys
Baseline and tracking surveys can be used to measure changes in condom use; use by type of partner; consistency of use; knowledge of correct use; attitudes toward condoms; barriers to condom use; and exposure to various program intervention, such as advertising and promotions. These surveys can be conducted among the general population or among target groups.

Baseline surveys are conducted before any program intervention commences. Their objective is two-fold: to collect background information useful in designing the program (formative research) and to collect baseline measures of key performance indicators, which will be used in subsequent evaluations.

Sampling and questionnaire design are the major research design issues. With these types of surveys, each wave involves drawing independent samples in each survey round (with the same set of parameters and from the same or an updated sample frame). This can be a costly and time-consuming exercise. Consequently, many surveys are narrowed in scope to include only those target markets that are of primary interest to the program. For example, surveys might be conducted among males in three urban areas as opposed to designing surveys that are nationally representative of all sexually active males. Furthermore, in order to make best use of HIV/AIDS surveillance data, attempts are frequently made to sample respondents around HIV sentinel sites.

Thought also needs to be given to the frequency of subsequent survey rounds. Experience has shown that, in the absence of a major and radically new program effort, condom use is unlikely to change significantly among the general adult population in a single year. Exceptions might be found among smaller, well-targeted groups, such as youth or certain high-risk communities. The frequency with which surveys are repeated should be determined by the amount of time in which significant changes are expected.

In the area of questionnaire design, significant progress has been made in the last ten years. There is now ample evidence that people will answer detailed questions about sexual behavior in questionnaires and that the data collected are corroborated by trends identified from other information sources, such as condom sales and STI prevalence. Several known biases exist, however. For example, there is evidence that women tend to underreport sexual activity outside of marriage in standardized surveys. Responses by males and females are often inconsistent. It can be important to corroborate survey information with other sources, if possible. Qualitative methods are often better at uncovering the details of condom use behavior. This information, however, can only be used to help interpret survey results and cannot be used to monitor indicators.

Two condom use issues have been particularly difficult to measure in structured questionnaires. The first issue involves trying to categorize the degree of risk someone is exposed to, by the type of sexual partners. Two approaches are being used. One is a time-based approach, that is, "regular" versus "non-regular" partners as defined by a relationship that has existed more than a year. A second approach classifies partners on the basis of cohabitation, namely living together or not living together.

The second issue involves measuring consistency of condom use. This is important because many condom promotional campaigns aim at getting more consistent condom use among non-regular or casual partners. In general population surveys, the tendency has been to ask for use (or non-use) at last sexual contact. It is felt that this avoids the bias associated with recall and self-reporting issues. For sub-population surveys, however, the trend has been to directly ask about consistent use, usually using some type of scaled answer such as always, sometimes, or never. Other lines of questioning, such as attitudinal issues and advertising recall, have been more successfully developed and are now usual components of market research tracking surveys.

The two key survey implementation issues include interviewer training and questionnaire pretesting. Both these elements can greatly improve the quality of results. A more detailed discussion of behavioral data collection, sampling issues, questionnaires, data validity, and qualitative evaluation approaches is presented in Chapter 8, "Uses of Behavioral Data for Program Evaluation."

Panel Studies
Panel studies are designed to examine changes over time for a given sample or panel of users or outlets. These studies consist of a series of interviews or observations of the same respondents or outlets over an extended period of time. For example, a panel of individuals could be asked to record their condom usage over the course of some time period. The basic characteristics of panel research are that it: (1) retains the same (or nearly the same) sample throughout the life of the study, and (2) measures elements at periodic intervals.

In contrast, cross-sectional surveys offer only a "snapshot" view of users. When replicated with successive samples, cross-sectional surveys may reveal trends at the level of population aggregates; but only panel studies (amenable to time-series analyses) are able to trace change in the knowledge, attitudes, and behaviors of individual users and may determine causality.

In principle, selecting a panel is no different from selecting a sample for a cross-sectional survey. Both are designed to represent a universe of users, the general population, or of outlets.

The first step involves recruiting respondents into the panel and obtaining reasonable assurance that they will cooperate throughout the duration of the study. Most panel studies offer some type of incentive to help ensure continued participation. Examples of incentives include small gifts, cash, and special lotteries for panel members. Even the best-run panels, however, will suffer some respondent attrition. Some members will undoubtedly move, be temporarily away, or otherwise be inaccessible or difficult to find. This is a problem unique to panel designs and special efforts need to be made to minimize these losses.

Bias resulting from attrition can be taken into account when interpreting panel data by comparing the characteristics of respondents lost with the characteristics of those who remain. For example, if losses from mobility occur more frequently among younger people than among older ones, findings can be weighted accordingly.

Although panel studies are complex and often costly, they offer several advantages. Obtaining repeated information from respondents over time increases accuracy and permits a more precise determination of behavioral processes, such as use habits, which are derived from patterns over extended periods of time. Most important, panels make it possible to analyze shifts and changes and to link them to the prior events that have brought them about. At the same time, panels have their problems, such as the conditioning of members through repeated contact and loss of members over time. Yet the panel remains one of the more powerful tools of causal analysis.

Omnibus or Continuing Surveys
Omnibus or continuing surveys are another source of data for measuring changes in users over time. An omnibus survey is a survey that is already being conducted by an established market research firm, such as Gallup or Research International. Such surveys are likely to exist in the more commercially developed countries. These surveys usually contain nationally representative samples of households and are conducted at regular intervals, such as monthly, bi-monthly, or quarterly. Each survey contains a core set of demographic and socioeconomic questions used for categorizing respondents. Commercial clients then pay to have a set of market-specific questions added (for example, questions pertaining to product or brand use, image, attitudes; reactions to pricing and availability; advertising recall). Clients only pay for the survey round in which they participate. They receive responses to the core questions as well as their market-specific questions. This format is popular among consumer products companies, such as Lever Brothers and Colgate Palmolive. Questions on condom knowledge and attitudes have been added to many omnibus surveys, although explicit questions on condom use might be considered less appropriate for this format. The main difference between panel and omnibus studies is that panel studies conduct repeated interview waves with the same set of respondents, whereas omnibus surveys involve repeated interview waves with different individuals, though they are drawn from the same sample.

The major advantage of participating in an omnibus survey is that it is usually less expensive than conducting a separate survey, since the costs are shared among several clients. Thus, it often provides an inexpensive way to collect survey data for several points in time in order to track changes in knowledge, attitudes, and practices. The major limitation of omnibus surveys is that the sample cannot be specified. If a country desires to sample only urban men, for example, this sub-sample will have to be extracted from the full omnibus sample. This may result in small sample sizes and correspondingly lower measurement precision.

Results from omnibus surveys conducted in Mexico and Jamaica show that knowledge of condoms as a method of AIDS prevention increased after the launch of AIDS media campaigns. Figure 5-1 shows the increase in knowledge of condoms as a prevention measure between waves 2 and 3 of the Mexican survey, which corresponds to both the launch and accompanying controversy surrounding that country's national AIDS media campaign4.

Consumer Intercept Surveys
A consumer intercept can take place in several ways. One way is to have an interviewer briefly screen clients/customers as they leave an outlet. If they qualify, a request for an interview can be made on the spot. Another approach is to have the staff or salesperson screen and indicate appropriate persons to interviewers. Alternatively, the staff person, after appropriate screening, can ask those who qualify if they would be willing to be contacted in their home for a subsequent interview. If prevalence were very low, it would be less time consuming and more cost-effective to record the names and addresses of qualifying respondents and arrange to interview them later in their homes than to intercept them for an immediate interview.

If seeking the cooperation of a commercial pharmacy or shopkeeper, it is usually necessary to provide the proprietor with some incentive. Often, a small cash payment is provided for each interview. If interviews are to be conducted on the spot, sometimes incentives, such as a free pack of condoms, are also offered to respondents. An alternative approach is to provide the customer with a certificate that can be redeemed at the participating outlet. This single mechanism provides a benefit to both the proprietor and customer.

There are sampling issues involved in how to most appropriately select the sites where respondents are being intercepted and screened. If a sizeable number of outlets are to be selected, the preferred method is to choose outlets randomly. If only a modest number of sites are to be covered, the selection is often done purposely to ensure sufficient variability with regard to neighborhood characteristics, volume of customers, or other criteria.

Results can be used to profile users. The profiles are used to determine how well the program is reaching its target market. Results can also be used to better understand key attitudes and behaviors of condom users. It is conceivable that the intercept approach could also be used with clients of sex workers (SWs), although some type of special arrangement would have to be worked out with the SWs themselves.

Evaluating Condom Availability, Condom Quality, And Quality Of Customer Service

Condom availability is a prerequisite of condom use. Therefore, it is important to measure various elements of condom availability. The issue becomes more complex when availability is broken down into "available by whom" and "at what times," and when availability takes into account additional social barriers as well as issues of eligibility and affordability. The next sections outline standard procedures for measuring condom availability as well as the quality of the condoms themselves and the associated quality of service delivery.

Outlet Checks, Retail Audits, and Mapping Studies
Condom availability at the national level is best determined by interviews with key informants, such as government procurement officers, commercial importers/distributors, other donor agencies, and non-governmental organizations (NGOs). Logistics management information systems can be useful in measuring condom flows to various distribution or wholesale points within the distribution system. They are generally inadequate, however, to measure condom availability or sales at the retail level.

Availability at the retail level is usually measured through some type of outlet check or audit. Outlet checks generally are used to determine the number of outlets of each type that have condoms of different brands. Retail audits provide information on the number of condoms sold in each outlet over time. Sampling is typically handled in one of two ways. With the more traditional outlets or with outlets that are relatively few in number, the first wave is frequently a census of all outlets, for example, pharmacies or government clinics in a certain region. This can be compared against a list of "official" outlets. Official lists may miss outlets that are not licensed or those that are closed, but may also indicate outlets that have been missed by the census. Samples are then pulled from this census for subsequent tracking waves. At the end of the project, a census often repeated to see if the total number of outlets has changed. Data from censuses give an indication of the percent penetration of condom distribution, by type of outlet.

For less traditional outlets or outlets that are too numerous to be covered on a census basis, sites are selected on based on the advice of those who know the local situation well. Data on availability are then reported as the number of outlets where condoms were found, divided by the number of sites visited. Regardless of the sampling approach, it is obviously important to identify and sample (in some manner) all the different types of outlets where condoms are likely to be found, including places such as bars, hotels, gas stations, and market stalls.

The biggest expense associated with an outlet check is getting interviewers to the outlets. Once they are there, they can relatively easily check other items of interest to the program as well, such as the availability of literature, condom prices, and waiting time for purchases. Once in the field, interviewers can also readily pick up other pertinent information, such as black market sales, leakage of public-sector goods into the commercial sector, and illicit cross-border trading of condoms.

Retail audits differ from outlet checks in that the protocol includes measuring sales. This is done by visiting the same set of outlets at regular intervals and collecting information on beginning and ending period inventories as well as intervening deliveries. Sales for the period can then be calculated by adding deliveries to beginning period inventory and subtracting ending period inventory. Some type of retailer training and incentive is usually offered to ensure successful participation in the audit. The principal problem encountered with projecting sales from this approach is that it is difficult to select a representative sample. Sales by outlet vary greatly and this information is not readily available by some proxy measure beforehand. Another approach is to purchase retail-audit data from an established commercial market research firm. For example, IMS HEALTH, a company that collects and sells pharmacy sales data in many countries, has spent considerable time working out sampling schemes for various countries. Condoms, however, are not covered in all IMS country surveys, and commercial data vendors do not typically include sales from less-traditional outlets.

Finally, some countries measure condom availability at the retail level through mapping studies, which plot outlet census data on maps, either manually or electronically using some type of geographic information system (GIS) software. This provides spatial information and patterns pertaining to availability and accessibility. It is sometimes used in conjunction with national health surveys.

Condom Quality Testing
Condom quality can be measured at different points in the distribution system. The most important checkpoint, however, is the point closest to use–in other words, at the retail level. An outlet check or retail audit provides a good vehicle for collecting a sample of condoms for quality checking. WHO and USAID have standard protocols for checking condom quality5,6. This requires a minimum of equipment and trained staff.

Mystery Shopper Studies
Mystery shopper studies involve the use of researchers posing as shoppers to observe performance at service delivery points. The researchers are disguised observers in the sense that they do not identify themselves as researchers, but instead appear to be normal customers or clients. These types of studies are used to monitor or assess the quality of staff-customer interaction.

Mystery shoppers are partially scripted and act out the role of a typical condom customer/client. Immediately after the "shopping experience" they record their observations and information gathered on a questionnaire. Many questionnaire items are factual in nature, such as, "How many condoms did they give you?"

Others, however, involve subjective judgment and are recorded on some type of scaled basis. For example, the question, "Did the pharmacist maintain a neutral, non-judgmental manner with you?" could be scored on a five-point scale from very judgmental (1) to not at all judgmental (5). This raises the issue of inter-rater reliability, or the variability due to different researchers' subjective judgment of the factor being assessed. Good interviewer training can go a long way to reducing this problem. The training should include both a thorough understanding of the factors being assessed and actual practice scoring these behaviors in the field. Sometimes two observers are used in mystery shopper studies. One researcher does the role playing and the other just observes. Both researchers record their experience independently. Another way to increase the accuracy of mystery shopper scoring is to conduct two baseline waves, by different researchers/shoppers, followed by a series of tracking waves. Sometimes different roles or scenarios are acted out in the tracking waves. For example, a young person might have a different experience obtaining condoms than would an adult. The greater number of measurement points should increase the accuracy of the assessment. Care must be taken, however, to ensure that the staff do not become familiar with the researchers. This can be managed either by using different researchers on tracking waves or by leaving a long enough time interval between visits.

In social marketing programs, outlets with continuing good customer service are often given some type of incentive or award. Mystery shoppers can be used to monitor and evaluate this performance. In addition, results, particularly improvements desired, can be incorporated into follow-up training programs.

Analytic Techniques

Surveys are commonly analyzed using descriptive statistics, cross-tabulations, t-tests, and multiple regression. The strengths and limitations of each are well understood by researchers and will not be further discussed here. There are, however, three other analytic techniques that have not been as widely applied to this kind of research but probably should be used more frequently. These are cost-effectiveness analysis, time series analysis, and discriminant analysis. These are more sophisticated regression techniques that involve different ways of statistically controlling for outside variables in order to isolate the effect of a particular program intervention. They are very briefly summarized here.

Cost-effectiveness Analysis
The relative cost-effectiveness of different condom distribution or delivery systems can be evaluated using this type of analysis. Because AIDS prevention funds are limited, it is important to periodically examine the cost-effectiveness of different condom programming alternatives.

As mentioned in the introduction to this chapter, condom distribution systems fall into one of three categories–government programs, social marketing programs, or commercial distribution. Public and commercial distribution systems are more conventional and, therefore, are somewhat easier to understand. Social marketing programs, on the other hand, can be organized in a variety of ways. There are two principal ways–organizations specially created to manage social marketing programs or existing commercial distribution companies who manage social marketing programs. In some countries, governmental or quasi-governmental organizations administer the program.

It is often assumed that social marketing programs offer a more cost-effective approach to delivering products because they have built-in cost recovery and harness existing commercial outlets. Within social marketing programs, it is further sometimes assumed that systems that

piggyback on established commercial distribution systems, as opposed to those that create their own distribution fleet, are more cost effective, though they are likely to reach fewer people. In an era of resource constraints, it is important to understand the relative cost-effectiveness of different programming approaches. For this reason, evaluations of condom programs should include analysis of cost-effectiveness.

The cost-effectiveness methodology involves looking at program costs in relation to program output (typically the number of condoms distributed) for alternative approaches. Total costs can be determined by adding direct costs and applying a proportion of indirect costs. Net costs are found by subtracting revenue from total costs. Alternatively, program costs can be obtained by tracing all direct costs and shadow pricing (estimating the value of) goods and services donated to the program. The number of condoms distributed to the retailer is typically the output measure used. These data are routinely collected as a part of logistics management information systems7, 8.

Cost effectiveness can also be examined as costs per the number of infections averted from condom programs and compared against other condom promotion strategies and other intervention strategies, such as blood screening or STI treatment9. Simulation models can be used to estimate the number of condoms required to avert one new HIV infection. Available simulation models range from easy-to-use models such as AVERT to large-scale simulation models such as iwgAIDS and STDSIM10,11,12.

Results from the iwgAIDS model, with data from East African countries, have been used to rank the relative cost-effectiveness of various targeted intervention strategies. Table 5-2 shows that interventions that target adolescents and high-risk groups are the most cost-effective11. Similar analyses might also be undertaken comparing social marketing with other types of interventions. For a more detailed discussion on the topics of impact modeling and cost-effectiveness analysis, see Chapter 17, "Guidelines for Performing Cost-effectiveness Analysis of HIV/AIDS Prevention and Care Programs."

Time Series Analysis
Time series analysis is a way to examine changes in a trend after a particular program intervention has taken place. For example, the number of condoms distributed to specialty outlets, such as bars, could be examined over time to see if the timing of any change in sales from bars corresponds to the effort to expand distribution to less-traditional outlets.

To do this analysis, a long time series of data are needed both before and after the intervention. The technique essentially involves fitting a trend curve to the data before the intervention and fitting a second to the data that corresponds to the period after the intervention. If there is a kink in the curve (if the slopes are different) in the "before" and "after" cases, it suggests (though does not prove) that the intervention program has had an effect on sales.

Discriminant Analysis
Discriminant analysis is used to identify the characteristics that best distinguish or differentiate groups from each other. For example, this technique could be used to identify the distinguishing characteristics of consistent versus inconsistent condom users. It might be revealed that inconsistent users are younger, poorer, and less well educated. Interventions could then be designed and targeted to this specific group to improve consistency of condom use. Discriminant analysis is similar to multiple regression but is appropriate when the dependent variable is dichotomous. For example, if we want to know whether users of condoms from a social marketing program are different from users of other condoms, the dependent variable would be "Uses the social marketing brand? (Yes/No)" Like multiple regression, we would include in the analysis a number of other variables that might explain whether a person uses the social marketing brand or not. These might include income, urban residence, knowledge of many brands, or access to a public supply source. The discriminant analysis would provide a equation showing how each of these explanatory variables relates to the probability that a person uses the social marketing brand. Discriminant analysis can also be used when there are more than two groups. For example, it could be used to determine if there are significant social and economic differences among those who use social marketing, public, and other commercial sources. One of the outputs of a discriminant analysis is a classification table that shows how the equation classifies all the people in the sample. Some would be correctly classified as using the social marketing brand or using some other brand. Others would be incorrectly classified by the discriminant equation. Some would be incorrectly classified as users of the social marketing brand when, in fact, they do not use it. In some cases, these misclassifications may indicate the characteristics of people who are likely to adopt the social marketing brand in the future or who could be convinced to use it most easily.

Conclusion

The principal components of condom programming include condom distribution, condom promotion, and condom use. Indicators have been developed by different groups to measure various elements of availability and use. Availability can be measured through qualitative information interviews, management information systems (MIS) data, services statistics or sales, purchased commercial data, or some type of retail-level outlet check. The outlet check or retail audit typically provides the most reliable measure of condom availability. To thoroughly understand the issue of availability, however, additional issues, such as availability by whom and at what times, must be examined. This requires investigating relevant sub-groups as well as exploring questions of social barriers, accessibility, and cost. A combination of audits and surveys (or qualitative interviews) can be used to do this.

Elements of condom use are best measured through some type of survey, ideally repeated at different points in time. Survey protocols and questionnaire designs have been improving. In addition, good interviewer training and pretesting can help further improve and customize surveys. Sometimes insights gained from qualitative research can be helpful in interpreting quantitative findings.

Monitoring and evaluating condom programming can provide valuable guidance to HIV/AIDS prevention programs, in terms of making mid-course corrections as well in assessing overall program performance and in capturing lessons learned. The reliability of these assessments can be enhanced by cross-checking key findings with other available data sources.

References

1. Bertrand JT, Stover J, Porter R. Methodologies for evaluating the impact of contraceptive social marketing programs. Eval Rev 1989;13:323-354.
2. Dallabetta G, Hassig S, editors. Indicators for reproductive health program evaluation: final report of the subcommittee on STD/HIV. Family Health International/AIDSCAP, with Carolina Population Center, Tulane University and The Futures Group International; 1995.
3. UNAIDS/00.17E. National AIDS programmes: a guide to monitoring and evaluation. Geneva: Joint United Nations Programme on HIV/AIDS (UNAIDS); 2000.
4. Stover J, Bravo M. The impact of AIDS on knowledge and attitudes about condoms as a contraceptive method in urban Mexico. Int Fam Plann Perspect 1991;17:61-73.
5. WHO/RHT/FPP/98.15 and UNAIDS/98.12. The male latex condom: specifications and guidelines for condom procurement. Geneva: World Health Organization and Joint United Nations Programme on HIVAIDS; 1998.
6. McNeill ET, Gilmore CE, Finger WR, editors. The latex condom: recent advances, future directions. Research Triangle Park (NC): Family Health International; 1998.
7. Stover J, Heaton L. The costs of contraceptive social marketing programs implemented through the SOMARC Project. SOMARC/The Futures Group International; 1997.
8. Stallworthy G, Meekers D. An analysis of unit costs in selected condom social marketing programs, 1990-1996. Population Services International; 1998.
9. Stover J, Way PO. Impact of interventions on reducing the spread of HIV in Africa: computer simulation applications. Afric J Med Pract 1995;2:110-120.
10. Rehle T, Saidel T, Hassig S, et al. AVERT: a user-friendly model to estimate the impact of HIV/sexually transmitted disease prevention interventions on HIV transmission. AIDS 1998;12(Suppl 2):S27-S35.
11. Stanley EA, Seitz ST, Way PO, et al. The United States Interagency Working Group approach: The IWG model for the heterosexual spread of HIV and the demographic impact of the AIDS epidemic. In: The AIDS epidemic and its demographic consequences. Proceedings of the UN/World Health Organization Workshop on Modelling the Demographic Impact of the AIDS Epidemic in Pattern II Countries: Progress to Date and Policies for the Future. New York; 12-1Dec, 1989.
12. Van Vliet C, Homes KK, Singer B, Habbema JD. The effectiveness of HIV prevention strategies under alternative scenarios: evaluation with the STDSIM Model. In: Ainsworth M, Fransen L, Over M, eds. Confronting AIDS: evidence from the developing world. Selected background papers for the World Bank policy research report, Confronting AIDS: public priorities in a global epidemic. Brussels and Washington (DC): European Commission and World Bank; 1998. p. 207-238.