Meta-analysis finds computer technology-based HIV prevention interventions to be effective in changing behaviour

A meta-analysis of randomised controlled trials (RCTs) of computer technology-based HIV prevention programmes has found that such interventions compare favourably to those delivered by people. The study, reported in the January 2^nd 2009 issue of AIDS, assessed the results of twelve published and unpublished RCTs of interventions that sought to change sexual risk behaviour in individuals not already known to be HIV-positive.

A major challenge in widely implementing successful, conventional HIV prevention programmes is the cost of the human resources involved. Computerised interventions, once developed, can be offered in a variety of settings with minimal staffing. Another advantage is that the standardisation of content in an electronic medium ensures intervention fidelity - that is, consistent delivery of the intervention. At the same time, the content is customisable, and intervention designers can prepare content matching people’s individual risk characteristics.

The primary behavioural outcomes in the AIDS meta-analysis were condom use or unprotected sex. The twelve RCTs cumulatively enrolled 4639 people, with a median enrolment of 319 persons. Eleven took place in the United States and one in the Netherlands. All study results were published or presented between 2002 and 2008.

The interventions evaluated in the twelve RCTs collectively had a statistically significant effect on condom use/unprotected sex. The mean effect size (or way of measuring the difference between the two groups), using Cohen’s d, was 0.259 (95% confidence interval [CI] = 0.201, 0.317; Z = 8.74, p

The RCTs that reported on frequency of sex, number of sexual partners and incidence of sexually transmitted diseases also found positive effects on those markers of HIV risk behaviour.

The meta-analysis used results from the longest-term follow-up assessment reported for each RCT, with most of those assessments being conducted three to six months following the intervention.

The authors observe that the effect size for condom use/unprotected sex compares favourably to effect sizes reported in two meta-analyses of human-delivered HIV prevention interventions seeking to increase condom use. One of those meta-analyses reported a condom-use effect size of d = 0.18. The other reported a condom-use odds ratio of 1.13 to 1.64; the effect size for the computerised interventions converts to an odds ratio of 1.54.

The computer technology-based RCTs had a wide range of effect sizes, and analyses were performed to identify variables associated with higher intervention efficacy. Interventions that were tailored to individual risk characteristics had a greater effect than interventions targeting groups, and interventions with more sessions had a greater effect than those with fewer sessions.

Interestingly, characteristics of study populations did not appear to be a major factor in determining outcomes. Greater effect was seen in RCTs with fewer men, young people and white participants, but none of these trends were statistically significant. When female-only, male-only and mixed-sex interventions were compared, statistically significant differences were found, with female-only RCTs having the largest effect sizes, followed by male-only RCTs. Interventions targeting men who have sex with men (MSM) had a greater effect size than those targeting heterosexuals, but the difference was not statistically significant.

The authors’ observations about this issue provide a further indication of the potential value of computer technology-based interventions. “The studies included in the current meta-analysis were quite diverse in nature and included heterosexual, MSM, men, women, urban, rural, minority and majority populations,” they write. “The fact that interventions were successful with a number of such populations may bode well for the broad application of these types of interventions.”

The authors do not believe that publication bias (the tendency for journals to publish studies with positive outcomes) accounted for their findings because they calculated that the evidence of impact could only be cancelled out by a large number of studies with non-significant results.

The interventions assessed in the RCTs included six individually tailored interventions and four group-targeted interventions. Three interventions presented interactive decision-making simulations. In eight RCTs, interventions were delivered using on-site computers, and in three others interventions were delivered via the Internet. One intervention used computer-generated print materials that were individually tailored according to participants’ responses to survey items.

The authors note that the two RCTs with poor participant retention both examined Internet-based interventions. This leads them to suggest that more research is needed to determine how to effectively conduct RCTs online. They state: “Given that the Internet has emerged as a conduit for individuals to seek and find high-risk sexual partners, developing and testing interventions that can proactively reach out to such populations is a high priority.”