Numbers are a dirty business

Celia Dugger, whose articles I still find frustrating, has written about the global decline in child mortality over the past twenty years. Dugger commits a popular sin in journalism: reporting absolute numbers instead of ratios or percentages. Giving absolutes can be intentionally misleading – how often have you seen a headline which reads “greatest number of job losses since 1425?”

Luckily, Karin Grepin jumps in with a detailed discussion of the numbers:

The announcement was that there has been a decline in the number – or level – of child deaths, which I thought was an unusual metric to report. The number of deaths is a function of the number of women, the number of births per woman, and proportion of children that die. When I think child mortality, I think just the last of these components. Fertility has been on the decline and it could very well be that we now have less deaths because there are just less births. But as it turns out, this is not what happened because of a nifty little phenomenon that demographers like to call “population momentum”. Since there are more women alive, we still have more births even with lower fertility. We have actually seen a nearly proportional decline in the actual under 5 mortality rate (deaths per live birth).

Read the rest of Grepin’s discussion of the article on here blog here.

This is also an excellent opportunity to mention Grepin’s working paper on the negative impact of HIV/AIDS targeted funding on the general health sector available here, which is a must-read.

Sine qua non


Sorry, I couldn't resist

Cecilia Dugger’s recent article in the NYtimes highlights South Africa’s faltering efforts to fight HIV/AIDS through large-scale male circumcision. It’s a decent read, although Dugger’s angle is tangent to that of the current medical consensus: mass circumcision is an extremely cost-effective preventative intervention and should be pursued with zeal.

A quick primer for the unfamiliar: in 2005 a randomized trial in Uganda revealed compelling evidence that circumcision reduces a man’s chance of contracting aids by roughly 50%. A round of follow-up trials confirmed that, at least within the span of the trial, that circumcised men were better-protected. The WHO declared the intervention a worthwhile one, and with the past few years we’ve seen unprecedented levels of mass circumcision around the continent.

This is mostly fine and well – the intervention has been proved to be effective in the short term – I’ve always had a few concerns though (I’ll warn you that, as usual, I’m out of my depth here – this is why we need a health blogger!):

A recent study in the Lancet reported what many have suspected: male circumcision does absolutely nothing to prevent male-to-female transmission of HIV. Thus, while circumcision will obviously protect women indirectly by reducing the chances their spouse will contract HIV (at least, in the limited time span of the randomized trial), it will do nothing for them if their spouse ends up contracting the virus, or already has it.

This essentially means that the first line of defense against HIV remains the condom. Health of workers understand this, and so continue to push condom use alongside the circumcision. However I’ve seen little effort to estimate the long run behavioral effects of the intervention. The original Ugandan study determine that the intervention didn’t substantially chance behavior, but this was back when the intervention wasn’t known to be effective. Now since circumcision is known to be partially effective, many men may now decide to trade off some of that protection for extra pleasure. A good measure of the elasticity of promiscuity would go a long way (I never thought I’d ever write elasticity and promiscuity in the same sentence).

There are some troubling anecdotal signs that men may be using the circumcision as their only means of protection. This quote from Dugger’s article stands out:

Even without government involvement, demand for the surgery, performed free under local anesthetic, has surged over the last year here at the Orange Farm clinic.

When was the last time you read a story about excess demand for condoms in the South African press? These men are clearly not all going to go on using condoms – it’s actually quite likely that circumcision is acting as a substitute for condom use – men would rather take the snip and accept a little risk.

This isn’t necessarily the case, but we’ll never really know unless we start observing behavior in the long run (most of the medical trials are abandoned prematurely due to ethical reasons – when the intervention clearly works the health workers are compelled to offer the control t treatment). We need to be sure that (i) we aren’t creating perverse incentives and (ii) we aren’t crowding out condom use.

Intentionally missing the point

BBC News is running an article about the recent arrest of Chansa Kabwela, the editor of the Zambian newspaper The Post.

Zambia’s public health system recently suffered a pay strike – one of the many woman that were unlucky enough to go into labour during this time ended up giving birth in the street, to a baby who ended up dying. Someone took photos of the whole event. Kabwela, apparently struck by the human cost of the crisis displayed explicitly in the photos, sent them on to government ministers.

How did the government respond? They arrested Kabwela and charged her with spreading indecent material, punishable with imprisonment of up to five years. Even worse, the penal code doesn’t clearly define what obscene material is.

A vibrant, intelligent and critital media is both one of the most needed and lacking institutions in this part of the world, where ‘democratic’ governments routinly crack down on a press they never really understood.

If only you knew the power of nagging

Give quiche a chance

I’m naturally a bit skeptical of ground-level interventions that don’t involve cash, needles or textbooks. Anything that involves dubiously-titled training or “empowerment”  sets off my very cynical alarm bells. However, I’m beginning to be persuaded by the evidence that targeted information campaigns work.

First there was Pedro Vicente and Paul Collier’s study on a randomised anti-violence campaign staged prior to the 2007 Nigerian elections, showing significant reductions in the treated districts. Then there was the Heckle and Chide’s study of minibuses in Kenya: a random treatment group were given posters advising passengers to speak up if the minibus drivers drove dangerously (which is pretty much what minibus drivers are born to do). The treatment group saw sizable declines in insurance claims, including those for injury and death.

Now there is a soon-to-be-published paper by Martina Björkman and Jakob Svensson, offering a unique randomised intervention:

  1. Assess local health providers and inform the communities on their relative performance using ‘report cards’,
  2. Encourage these communities to form groups to monitor local health performance.
  3. Sit back and see what happens.

A year after the intervention, a repeat study revealed that the treated communities had: harder working health providers, higher rates of immunization and significantly reduced rates of child mortality and underweight children, all with the same levels of funding.

The best part of the study was the lack of investigation into what the communities were doing to make changes – (there is some rough evidence that the communities were more active in electing and dissolving the local provider management committees). My guess is that a fair amount of nagging was involved.

I’ve come to believe that a crucial part of development is strengthening the accountability link between citizens and their government (not to be confused with enforcing accountability externally), especially when the citizens face a trade-off for enforcement (in this situation, that trade-off is time spent hassling health workers).

A few questions remain:  is it persistent (or would health workers become more resistant to this informal accountability over time?) Is this scalable? Which part of the intervention was key: the information transfer allowing for yardstick comparisons between district, or the “empowerment” workshops? My hunch is the former.

(Bonus points to those that got the Red Dwarf reference).