Cesar Victora gave a very interesting talk earlier today concerning the International Epidemiology Association’s position paper on the UN’s Sustainable Development Goals, which are currently being drafted (to replace the Millennium Development Goals post-2015). Victora is President of the IEA, for a few more hours at least (the new President takes office this evening). Many of his points were reiterated by the next speaker, Theodor Abelin, and in questions from the floor. There were no audible voices of dissent. (The talk reflects a fuller position paper, available here.)
The point that stayed with me most from Victora’s rich talk was the importance of relating goals to appropriate measurement techniques. My own interest in epidemiology has tended to focus on efforts to identify causes (“analytic” epidemiology), since causation is a natural magnet for philosophical interest. But measurement is also a focus of philosophical interest, and Victora nicely pointed out that “descriptive” epidemiology – the business of measuring things like maternal mortality rate, for example – is extremely important if these Sustainable Development Goals are to be effective. A country cannot be held to a goal that cannot be measured, and it cannot be fairly be held to a goal when progress towards that goal is estimated rather than measured.
For example, I was not surprised to learn that in many countries where maternal mortality is high, data on maternal mortality rates (MMRs) are scarce. What did surprise me was hearing about the calculations that some august international organisations perform in the absence of data. A calculation is performed involving GDP per capita, general fertility rate and skilled birth attendance. MMR is estimated as a function of these and perhaps some other similar variables. This means that if the country goes through a recession, the estimated MMR will automatically go up. – Perhaps is really will go up, but it seems strange to think of that calculation as a measurement, at least in the absence of extremely good evidence for the reliability of the estimating equation – evidence which, of course, we don’t have.
MMR is measurable, of course. The problem with MMR is simply a lack of data, and this problem afflicts a large class of conditions. As Victora put it in relation to snakebite: “Where we have snakes, we don’t have statistics, and where we have statistics, we don’t have snakes.”
However, Victora’s most penetrating critique of the SDGs concerned the setting of goals in the absence of clear ideas about how progress towards the goals will be measured. The health-related goal is as follows:
“Goal 3. Ensure healthy lives and promote well-being for all at all ages” (from the Outcome Document)
This overarching goal is broken down into 13 subgoals, some of which are very loosely specified. For instance, how are we to tell whether a country has managed to “strengthen prevention and treatment of substance abuse, including narcotic drug abuse and harmful use of alcohol”? Ironically, those goals that are most clearly specified are wildly unattainable, such as halving global deaths and injuries from road traffic accidents by 2020. Those that are not well specified present measurement challenges for epidemiologists.
This made me wonder whether a body like the IEA could itself set some “goals for the goal-setters” – that is, criteria which any health-related goal must meet if, in the professional opinion of the IEA, they are to be useful. The simplest such criterion would be that outcomes must be specified in terms of a recognised epidemiological measure (mortality, for instance). Another might be to accompany each goal with information (perhaps in a corresponding entry in an appendix) concerning the trend over the past similar period: so if the goal is the halve road traffic deaths in 15 years, or 25, information on the growth of road traffic deaths over the past 15 or 25 years might be included. Goals of this kind will always be political, but there might be agreement on a set of simple rules for setting such goals, and if such rules existed, this might pull epidemiologists closer in to the goal-setting process – a kind of politicking which, as one of the questioners pointed out, is not part of standard epidemiological training.