What is data for health?

The first few posts on Where There Is No Data have set the scene by discussing the challenges faced by marginalised populations in HIV programmes.  We’ve also talked about the fact that having good data isn’t always enough – policy decisions tend to be influenced as much by politics and prejudices as they are by evidence.  Nonetheless, while data isn’t everything, getting better data is important.  It can begin to shine a light on problems that have been ignored; and when there is political commitment to tackling these problems, it can help make sure that health programmes are properly designed.

Policy makers, funding agencies and NGOs often talk about different types of data and research, but not everyone is familiar with these.  When we talk about data for health programmes what do we mean?

  • Data that describes the burden of health problems including the prevalence and incidence of these conditions, and how they are distributed among different sections of the population.  In most countries, data on HIV prevalence (i.e., the percentage of the population that is infected with HIV) is derived from a range of sources: surveys of a representative sample of households during which respondents are asked to participate in anonymous HIV testing; surveys of specific population groups such as pregnant women, men who have sex with men, sex workers (these are often called “sentinel surveillance” surveys); and routine data – for instance, the proportion of people volunteering for HIV testing who are found to be HIV positive.  None of these methods provides a complete picture of HIV prevalence but, if all of them are used fairly regularly, they can help give a good overview of what is happening.  In some countries, special studies are carried out to measure HIV incidence (i.e. the rate of growth of the epidemic in a given period of time such as a year), but this type of study is rarely carried out at national level.
  • Prevalence and incidence studies can also help to identify how a health problem is distributed in the population – whether it affects men or women more, whether some age groups are disproportionately affected, and which behaviours or sub-populations are most affected.  A lot of this information can be gained from household surveys and routine data.  However, because in many societies people are reluctant to talk openly about sex and sexuality, and in particular because of the stigma against behaviours such as sex work, they are often under-represented.  Moreover, knowing how a health problem is distributed in the population is not just a matter of knowing the prevalence of the problem in each sub-population group.  It also means knowing the size of each these sub-population groups is.  This is useful for planning programmes and ensuring that resources are allocated in the right places.  Once again though there are particular challenges in estimating numbers of marginalised populations.  Sex between men, sex work, and drug use are behaviours not identities, and people with these behaviours often have good reasons to avoid being counted or included in surveys.  In many locations there has been little or no research on these groups and informal, community level research is needed before any formal surveys can be carried out.
  • Information about what makes people vulnerable or at risk.  The same surveys described above can help provide an indication of why some people are more vulnerable or at risk than others.  In the context of HIV, researchers often ask respondents about their sexual behaviour, condom use and so on – although once again, peoples’ responses to these questions are not always reliable.  Moreover it is not enough to know whether people have good enough knowledge about HIV or whether they use condoms or whether they have access to health care.  It is also important to know why these things happen.  Is it because programmes are not reaching them? What role does stigma and discrimination play?  Getting the answers to these questions often requires a different approach: one that engages much more with the people concerned, and ideally one that is led by them.  These factors also often vary from place to place and can change over time, so it is important to have mechanisms that enable communities to collect, understand and act on local data in a regular way.
  • Data on what types of programme or “intervention” are effective.  Many policy makers rely on experimental research methods to provide evidence on the effectiveness of programmes or interventions for preventing or treating health problems.  Although they are expensive to conduct these studies help provide an estimate of how effective different approaches are.  There is considerable debate surrounding the reliability of this sort of study for evaluating social change programmes, although their use in this area is growing.  In any case, because they are experiments, generally conducted in controlled conditions, they will not necessarily be as effective or work in the same way when implemented at scale.  For this reason programmes have to find ways of continually monitoring their impact and of identifying any unintended consequences.  Once again a combination of large scale survey and routine data and more qualitative, community based approaches is needed.
  • There is increasing interest in good quality programme related data – for instance, data on the coverage of programmes (how many people they reach and who these people are); data on how programmes are funded (where does the money come from? Communities? The government? Donors?); and data on what it costs to implement different types of programme.  Good data on costs, combined with good data on effectiveness can be used to make sure the most cost-effective programmes get funded – in other words that available resources have the biggest possible impact.

Readers working on HIV and AIDS may be familiar with concepts such as “know your epidemic”, “know your response” and “strategic investment”.  The description of the different types of data described above is somewhat simplistic; however they are the basic building blocks behind these concepts.  It is particularly useful for community actors, key or marginalised populations to know about these types of data and concepts so that they can clearly articulate the gaps when speaking to policy makers.

We will explain all of these types of data in more detail on this site – including how reliable are, what sorts of skills are required to collect them, and the methods and the costs of collecting them.  There are a number of free online courses and resources where you can learn more about different research techniques and we will post links to these too.  Finally, we will talk about the importance of data generated by communities and the challenge of trying to get this sort of data taken seriously.


The “data paradox”

In this short post we discuss further some of the challenges faced in planning HIV programmes for key populations.  The post is based on discussions that were originally posted on our Facebook page.


A write-up of a discussion among sex workers about where they are most likely to face violence in their community, and what they can do about it. By FIMIZORE, Madagascar network of sex workers (c).

Stef Baral

So why are we in a situation where we have no data on some of the groups most affected by HIV? To characterize the widespread and generalized HIV epidemics found in mostly southern and eastern Africa, we have relied on surveillance systems that do not meaningfully assess the very populations known to be most affected by HIV in most parts of the world including sex workers, gay men and other men who have sex with men, people who use drugs, and trans populations. I understand the design and use of these systems–and there is no doubt that having a sampling frame made up of households is the most effective tool for population-based sampling to retrieve population-based estimates of HIV.

However, I also understand that for there to be household based transmission of HIV, there has to be sero-discordant relationship. And that tells me that there has to have been a primary infection event outside of the household. It is those risk factors for the extra-household primary infection events that we still seem to know very little about. And it seems to me that we may not have invested in this since maybe we don’t want to hear the answer.

Maybe it is because all around the world there are people that sell sex–as there always has been and always will be. All around the world, there are people who do not fit into heteronormative social expectations for sexuality and simple binary expressions of gender. And all around the world there are people who use drugs–some for fun and some because of untreated or undertreated mental health. And each of these people have specific needs for the prevention and treatment of the acquisition and transmission of HIV. But how can we fully understand these needs, if we don’t want to admit that populations exist everywhere. So we still, 30 some odd years later, find ourselves in a data paradox. A paradox where we know less about the needs of diverse populations in settings with the most stigma.

Does that mean we should not respond until we have “the data”? It took me a while to get how brilliant of a strategy this is–argue that the lack of data is a reason to not launch any programs or funding that may result in data. Ie, the lack of data feeds itself.

Matt Greenall

So the “data paradox” is this: decision-makers deny that most affected populations exist, or that they are relevant to the epidemic; so no research gets done on these populations; the lack of data feeds the denial; and so on.

But the other challenge we face is that decision-makers don’t always put resources towards programmes with key populations even when good data is available. Experience tells us that politicians, in particular, are very capable of ignoring data when it suits them to do so – and there are many countries where the quality of the data is reasonable enough but the investments in programming with the most affected groups are still all wrong. So while it is important to encourage better national level research and data that can eventually be plugged in to modelling and strategic planning exercises, we should also be aware that they won’t necessarily resolve everything.

Although there is no doubt that having the right laws and supportive leadership from the government or Ministry of Health makes a huge difference to how different problems get addressed, not all change comes from the top down. Similarly, the stigma and marginalisation that key populations face does not only come from the national level – it also has a lot to do with attitudes and behaviours of health care workers, law enforcement officers, and community members at the local level. These need to be addressed to. This discussion at the recent International AIDS Conference in Kuala Lumpur is an example of how research and programmes can be initiated at local level even without strong support from policy makers. I’ll also post something in a few days, and post a link here, about local, community-led research and how it can help groups get organised at local level.