MethodLogical is now at

MethodLogical is now at

Due to some persistent technical issues we've been having with Blogger, we're now posting at Please update your RSS feeds, etc. For the time being, our new posts will automatically be mirrored here, but you'll have to visit the new site to comment.

Monday, November 29, 2010

Is HIV Testing a Waste of Time?

One of the pillars of almost all HIV prevention and reduction strategies is that we should test people for HIV – as many as possible. The CDC has set a goal of universal HIV testing for Americans in the “sexually active” age range of 13 to 65, and stressed that the same should apply worldwide. That article argues that universal testing will assist with surveillance – monitoring the spread of the disease – but given the time it takes to test people a random, representative sample would be much more effective for timely surveillance. If not to track the epidemic, then why test everyone? To help the infected, and encourage behavior change. When I volunteered for HIV-prevention NGO Students for International Change (now Support for International Change) back in 2004, the argument was that if you don’t know your status you can’t protect yourself or your partner.

The unquestioned assumption here is that test results will change people’s behavior (And I mean that in a personal sense – I never bothered to question it). In an article published in the American Economic Review, Rebecca Thornton* asked exactly that question, using condom purchases as her measure of risk-reduction. She ran a field experiment in which she randomly offered people in Malawi a monetary incentive to go to a makeshift “results center” clinic to learn their test results. By randomizing the incentive and varying the distance to the results centers, she is able to correct for potential selection biases in her results. In an observational study, it’s possible that only people who already know whether they have HIV are willing to get tested, in which case you’d see no change in behavior. After correcting for this problem, she does find a behavior change, but it’s not very promising: only HIV+ participants significantly increased their purchases of condoms, and by less than two on average. This isn’t very much: a cost-effectiveness analysis shows that any number of other strategies, such as treating comorbid STIs, would prevent more infections per dollar.

It’s certainly possible to raise issues with that study. First, maybe the results apply only to Malawi, or to the area where the experiment took place. Second, maybe we’re not accurately capturing all the potential risk reductions that people are engaging in. But testing as a way to change behavior faces an even more severe problem. As I noted in a previous post, the risk of HIV transmission peaks during the first 4-8 weeks of infection. The most common HIV tests look for the antibodies the immune system makes to kill off the virus; it’s no coincidence that these antibody tests won’t give a positive result for about 30 days – the “window period”. In the regions of Southern Africa where HIV is most prevalent, the pandemic is driven in large part by sexual encounters during that window period. That means that even if testing does change behavior, it won’t do much to stop new infections.

The HIV "Natural History" Diagram, showing viral load in red.
Note the discontinuity/break in the graph.
So what should we do instead? One solution, proposed by Granich et al. in The Lancet, is to stop trying to induce behavior change at all. Instead, they propose testing everyone and providing ARV treatment for everyone who tests positive, and their simulations project that that could end the pandemic in cost-effective fashion within forty years. If we want to focus on behavior change, the lesson is that we need to target the seronegative – those that either don’t have HIV yet or won’t yet test positive. The current strategy of widespread HIV testing in order to change behavior isn’t a complete waste of time – but it’s not nearly as effective as I used to think.

*Full disclosure: Rebecca is my (unofficial) advisor here at Michigan.


  1. Hey Kerwin- this is interesting. One thing you may be neglecting is that we can't treat HIV+ people until we know their status. So, pushing for testing is really a way of pushing for treatment. Also, since antiretrovirals reduce HIV transmission, this process may also lower spread of the disease. It is true that about 25% of infections occur during the "window period," but that leaves plenty of other infections that could be avoided. I guess the next question is this: is treatment as prevention a reasonable and cost-effective strategy?

  2. This comment has been removed by the author.

  3. Data regarding how being on ART influences infectivity are not yet conclusive. Where are the funds to test and treat everyone with an already flatlined budget for HIV/AIDS?

    The issue here is that we are focusing on treatment at the expense of prevention. Prevention programs have been highly successful at rapidly decreasing HIV incidence in many countries. The majority of infections in Africa are acquired through discordant couple situations. Why are we still testing people alone and not with their partners?

    Couples voluntary counseling and testing is an extremely cost-effective way that could be applied universally without straining already HIV-dominated health systems.

    Finally, is this even the discussion that we should be having? Another example of vertical programs narrowly focused on specific diseases rather than a "health" based approach.

    $1000 for ART and then death from diarrhea because no $.25 ORS?

  4. Adam, that's pretty much the exact point raised by the Granich et al. paper I linked to. One question is whether you could actually convince everyone to take the ARVs - no sure thing given the evidence on condom uptake. This is an area where economists can make themselves useful: HIV prevention in any form depends on setting up systems and incentives to get people to change their behavior in a particular way.

    Brad, is there evidence that VCT actually substantially changes behavior for serodisconcordant couples? It's possible but not easy to reconcile with the Thornton evidence for Malawi. Even if it does, we still need to worry primarily about non-marital partnerships and multiple concurrent partners. The evidence suggests those partnerships are the primary vector by which HIV spreads through the population, and they create a large share of serodisconcordant couples.


Note: Only a member of this blog may post a comment.