Megan McArdle points us to this scary report:
2,194 households containing 2,592 Ghanaian children under 5 y old were
randomised into a prepayment scheme allowing free primary care
including drugs, or to a control group whose families paid user fees
for health care (normal practice); 165 children whose families had
previously paid to enrol in the prepayment scheme formed an
observational arm. The primary outcome was moderate anaemia
(haemoglobin [Hb] < 8 g/dl); major secondary outcomes were health
care utilisation, severe anaemia, and mortality. At baseline the
randomised groups were similar. Introducing free primary health care
altered the health care seeking behaviour of households; those
randomised to the intervention arm used formal health care more and
nonformal care less than the control group. Introducing free primary
health care did not lead to any measurable difference in any health
outcome. The primary outcome of moderate anaemia was detected in 37
(3.1%) children in the control and 36 children (3.2%) in the
intervention arm (adjusted odds ratio 1.05, 95% confidence interval
0.66-1.67). There were four deaths in the control and five in the
intervention group. Mean Hb concentration, severe anaemia, parasite
prevalence, and anthropometric measurements were similar in each group.
Families who previously self-enrolled in the prepayment scheme were
significantly less poor, had better health measures, and used services
more frequently than those in the randomised group.
There appears to be an enormous Hawthorne Effect in this study.
Even the control group, with no “free access” had a dramatically better primary outcomes than was expected before the study was performed. This suggests that the mere act of checking on all those kids, asking about their health practices, and dutifully recording it all on clipboards, may have changed their behavior – aka the Hawthorne Effect. As a result, although the experimental group seems to have done twice as well as anticipated, relative to the pre-study baseline, the control group also improved, so their was no net difference detected.
From the report: It was assumed, based on data collected previously in the study area, that the prevalence of moderate anaemia among the control group would be 10% at the end of the 24-wk period of follow-up.
Yet the final outcomes (in table 4) was about 3% for the both the control and experimental groups, far less than the 10% anticipated.
Is the main story the huge effect of formally asking the study subjects about their health practices during Malaria season?
Not sure scary is the right word. Simplest explanation is that Ghana health care is worthless, but that’s not necessarily correct.
This seems similar to the RAND health insurance experiment where increasing deductibles/co-insurance led to lower utilization but the same health outcomes. Interesting that this type of result would might hold in a poor nation as well.
I think there are several themes here. One, the marginal value of care delivered for “free” might be very low. Two, health is not the same as health care. There’s a lot of evidence that sanitation, lifestyle, living conditions, education etc… have far more to do with health than the level of health care delivered.
Notice that it’s voluntary participation. Lower than expected rate of moderate anaemia may due to self-selection, rather than Hawthrone effect.
That said, the study does not measure any positive externality of public health care.
But there *was* a net difference detected in utilization of the healthcare, so if it had value we would still have expected some effect above and beyond the Hawthorne effect (which does seem very large here as you suggest).
I would second the simple explanation that the provided healthcare probably just isn’t any good on net. Considering that some chunks of healthcare in the US provide only very small net benefit for the cost (e.g. obstetrics) this isn’t implausible at all.
David L: By “no effect” in a study, one aways means that the confidence interval included the no-effect value. Whether that result is strong or weak depends on your prior beliefs about how big the postulated effect “should” be.
@Tom
“Anyone from a health economics background who could comment?”
Robin Hanson comments here
Out of curiosity, what was the marginal opportunity cost of healthcare? Even if no money is charged at the clinic, there are still costs to getting to the clinic, especially if one lives remotely with poor transportation.
David, this is gibberish. Have you actually had a formal class in statistics? Because you’re not even using the terminology correctly. We only know that in, say, a 95% confidence interval, 19 times out of 20 the parameter, mu, will lie within an interval around X-bar.
We cannot say there is a 95% probability that the true mean lies within this interval. That is simply false; would you say that there is a 95% probability that the dice are fixed if one rolls boxcars? Of course not. And this is particularly true for tests of small power(as student alludes to.)
Comments on this entry are closed.