From the Division of STD Prevention, Centers for Disease Control and Prevention, Atlanta, Georgia
Correspondence: Matthew Hogben, PhD, Mail Stop E-44, Centers for Disease Control and Prevention, Atlanta, GA 30333. E-mail: firstname.lastname@example.org.
Received for publication August 1, 2008, and accepted September 19, 2008.
WITH MANY CHANGES, INCLUDING advances in health, comes speculation about the possible damaging and unintended consequences of introducing change. In many preventive interventions, unintended consequences follow a particular form called disinhibition or risk compensation. As typically understood, these concepts are operationalized via individuals who, once feeling protected against 1 health risk, engage in other risky behavior that puts them at risk for the same or other health problems. For example, commentary from the 1960s and 1970s cited effective STD treatment as a risk for increased sexual behavior.1 As well, recent debate over mandatory human papillomavirus vaccinations for school entry has brought similar fear that young women might subsequently feel free from concern about the cancer-causing disease and have more unprotected sex with more partners at an earlier age.2–4 In this issue of Sexually Transmitted Diseases, Greg et al. present data from a HIV prevention trial testing the safety and efficacy of oral tenofivir for women in Ghana,5 which we will use to prompt some thoughts about the differences between disinhibition and risk compensation and their scope and relationship to intervention.
First, we note there are important conceptual differences between disinhibition and risk compensation with different implications for intervention strategies. Disinhibition derives from psychological terminology; it occurs when people stop trying to avoid risk to themselves or others. Probably the most widely known examples in sexual behavior are centered around the disinhibiting effects of alcohol; an inebriated person may be sexually incautious or aggressive because he or she no longer “cares” about the risk of sexual exposure.6,7 Other examples are drawn from people who feel they cannot avoid a harm and then no longer even try to do so.8 In both examples, the outcome is behavioral disinhibition through lack of caring, although the causes (alcohol vs. perception of unavoidable risk) are very different. Risk compensation, on the other hand, is best understood from a more cognitive perspective. The term applies to those whose diminished susceptibility via a given preventive intervention permits them to increase other risk behaviors. Although both terms are often used interchangeably in the literature, what we have defined as risk compensation is the most common source of concern for those conducting interventions.
Guest et al.'s analysis of HIV preexposure prophylaxis in this issue5 is a good starting point to examine these phenomena more closely because the women in the trial, largely sex workers, (a) had high potential for disinhibition via greatly elevated–“unavoidable”–risk of HIV acquisition and (b) high potential for risk compensation via their economic motive for sex. Specifically, the authors examined changes in risk behavior among HIV negative, sexually active women after enrollment in the trial, with data gathered across up to 12 visits over the course of 6 months. For the purposes of these analyses, Guest et al. stratified the sample by baseline sexual risk characteristics (most risky 25% vs. all others); as both control and intervention arm participants took a pill daily, changes in risk would not easily be attributed to knowing one was on antiretroviral prophylaxis. They found that participants reported decreases in unprotected sex, with the steepest declines among women categorized (via number of partners and proportion of unprotected sex acts) in the highest risk group. Through qualitative interviews with a subset of the women, the authors were able to attribute decreased risk behavior at least partly to counseling accompanying the drug regimen.
All this is good news about “changes in sexual behavior.” The intervention provides another demonstration that most members of groups whose sexual health prospects have become more assured do not subsequently attempt to damage said prospects: that is, no widespread risk compensation was seen. Similarly, penicillin “preexposure prophylaxis” for syphilis, tested in Louisiana between 1997 and 1999, and a review of sexual behavior for HIV positive people on antiretroviral therapy yielded the same conclusions.9,10 Inference about disinhibition is more indirect, although the results are consistent with reduced disinhibition as the women's protective behaviors increased with the advent of the preventive intervention. As with risk compensation, the qualitative data were supportive of such inferences but without ruling out alternative explanations.
We do not suggest that risk compensation and disinhibition are always negligible effects in interventions. If one lesson seems to be that neither disinhibition nor risk compensation is a widespread, population-level phenomenon in terms of sexual risk, another is that behaviors consistent with both phenomena can sometimes be found. For example, a Ugandan trial of a condom promotion campaign in which men in the intervention arm were taught technical use skills found increased condom use, but also increased numbers of partners, among those exposed to the intervention.11 In the same review that demonstrated no overall differences in sexual activity among HIV positive people who were or were not receiving antiretroviral therapy,9 some who viewed therapy as reducing transmission risk did have increased numbers of partners. Findings such as these in the least illustrate that within situational contexts and/or individual belief schemes, the issues easily become complex. Those interested in changing public health behavior should at least be aware of the possibility of risk compensation or disinhibition and be prepared to address any negative unintended consequences of their interventions.
Even risk compensation alone should be considered in a context broader than the postintervention environment. More appropriate, perhaps, is what Eaton and Kalichman call “risk equilibrium.”12 Eaton and Kalichman adapted a calibration model from another source,13 showing how assessment of risk compensation can be viewed in the context of a given individual's “set point,” the level of overall risk he or she finds acceptable to achieve a given goal (eg, sexual activity). When balancing costs and benefits of risk and preventive behavior, individuals adjust activities to maintain their set point. These adjustments become risk compensation if a prevention intervention reduces risk through the intervention. The term, risk compensation, is a perfectly accurate description of adjustment to equilibrium, but is simply an outcome of tension in a homoeostasis model analogous to, say, temperature regulation in mammals.
Expanding further, a snapshot of risk equilibrium at any given time point is not the whole picture because situational factors change, and people's goals are dynamic, nor are they wholly in charge of setting their risk levels. Guest et al.'s data remain informative: the female sex workers in the study found out they were actually HIV-negative and that tenofivir could reduce their likelihood of acquiring HIV. Therefore, the situational factors predicting infection likelihood changed, as did plausibly the women's perceptions of that likelihood and their subsequent goals. Guest et al.'s qualitative feedback, in which some respondents identified realizing they were HIV negative as a motivation for condom use and partner reduction, shows such logic. Assessing women's perceptions of risk, as with Guest et al. is also important to be sure the intervener understands the target's goals. For example, among women who are more attuned to avoiding pregnancy than STDs, introduction of highly efficient hormonal contraception (an intervention) is frequently accompanied by increasingly inconsistent condom use.14 What seems at face value to be an example of risk compensation is more properly interpreted as a misunderstanding of the target population's goals.
Finally, interventions should differ when addressing disinhibition versus risk compensation/equilibrium. Formative data, both quantitative and qualitative, help the researcher divine potential targets' experiences and goals; this helps reduce the odds of unintended consequences (here, risk compensation) and determine whether disinhibition is an extant phenomenon. Where disinhibition appears likely, the intervention will probably need to add elements that address motivation and possibly depression to the usual spectrum of tasks and choices. If risk compensation is expected, then the intervention needs to present parallel means to meet the targets' goals, convince the targets to modify their goals, or match their goals to intervention goals. All these modifications suggest a yet larger shifting of paradigms. Instead of treating unintended consequences such as risk compensation as purely negative, one could reframe the events as opportunities for further and more complete intervention. For instance, if people are indeed rethinking their risks or their motives, perhaps that is precisely the time to communicate positive health messages or to assess individual behavioral motives via motivational interviewing. If we are able to incorporate unintended consequences into interventions and to understand better why they occur, STD and HIV prevention will have improved some innovative and effective existing prevention efforts.
1. Willcox RR. A world look at the venereal diseases. Med Clin North Am 1972; 50:1057–1071.
2. Lo B. HPV vaccine and adolescents' sexual activity. BMJ 2006; 332:1106–1107.
3. Zimet GD, Liddon N, Rosenthal SL, et al. Psychosocial aspects of vaccine acceptability. Vaccine 2006; 24(suppl 3):S201–S209.
4. Kapoor S. The HPV vaccine and behavioral disinhibition. J Adolesc Health 2008; 42:105.
5. Guest G, Shattuck D, Johnson L, et al. Changes in sexual risk behavior among participants in a PrEP HIV prevention trial. Sex Transm Dis 2008; 35:1002–1008.
6. Leeman RF, Toll BA, Volpicelli JR. The drinking-induced disinhibition scale (DIDS): a measure of three types of disinhibiting effects. Addict Behav 2007; 32:1200–1219.
7. Fillmore MT, Weafer J. Alcohol impairment of behavior in men and women. Addiction 2004; 99:1237–1246.
8. Seligman MEP. Helplessness. San Francisco, CA: Freeman, 1975.
9. Crepaz N, Hart TA, Marks G. Highly active antiretroviral therapy and sexual risk behavior: a meta-analytic review. JAMA 2004; 292:224–236.
10. Farley TA, Cohen DA, Kahn RH, et al. The acceptability and behavioral effects of antibiotic prophylaxis for syphilis prevention. Sex Transm Dis 2003; 30:844–849.
11. Kajubi P, Kamya MR, Kamya S, et al. Increasing condom use without reducing HIV risk: results of a controlled community trial in Uganda. J Acquir Immune Defic Syndr 2005; 40:77–82.
12. Eaton LA, Kalichman SC. Risk compensation in HIV prevention: implication for vaccines, microbicides, and other biomedical prevention technologies. Curr HIV/AIDS Rep 2007; 4:165–172.
13. Wilde GJS. Target risk: Dealing With the Danger of Death, Disease and Damage in Everyday Decisions. Toronto, Ontario: PDE Publications, 1994.
14. Frank ML, Poindexter AN, Cox CA, et al. A cross-sectional survey of condom use in conjunction with other contraceptive methods. Women Health 1995; 23:31–46.