health economics, Water

Water supply & diarrhoea – latest systematic review and economic implications

An update to the WHO-led systematic review of the ‘Impact of drinking water, sanitation and handwashing with soap on childhood diarrhoeal disease’ by Wolf et al. (2018) was published in TMIH in May.

I re-read it last week with a water supply hat on, and was interested to see how they’ve improved on the 2014 version. The main difference, apart from including studies recently completed, is that they’ve updated the structure of the meta-regression to allow for separate results for “piped water, higher quality” and “continuous piped water”. This means there’s additional comparisons to be made (with more relevance for SDG6 “safely managed” definitions).

As background, their meta-regression approach allows estimation of service level transitions that have not been directly observed in studies. This builds on ‘network meta-analysis’, a technique increasingly being used in economic evaluations of health interventions where there are no head-to-head trials between options.

Below is my visualisation of the key results (building on the diagrams of the 2014 review’s results in this WHO publication). The numbers are percentage reductions in diarrhoea morbidity risk associated with each transition – explanation below the figure.

wolf et al* / ** see note at bottom of post

I made this based on table 3 of Wolf et al. 2018, calculating % reduction = 1 – risk ratio. In my view, the transitions most likely to happen in practice, as a result of investments, are the incremental ones. Therefore, I have put these in bold blue. Those transitions less likely are in bold back, and those fairly unlikely are in italics. By “unlikely”, I mean that those people remaining with unimproved water are now predominantly in rural areas, where the direct transition to a safely managed water supply (meeting SDG criteria for both quality and continuity) is unlikely to be affordable in many settings. This reduction (75%), which the authors estimated indirectly, would appear to be the reduction maximally achievable with water supply. Some more notes are at the bottom of this post.

What should we make of these results? I would make two observations, and then two arguments based on the economic implications.

  1. There is good evidence that improving people’s water supplies can reduce diarrhoea, which is among the top three contributors to the all-age disease burden amongst the two quintiles of countries with the lowest human development. For the transitions in the figure previously reported in the 2014 review, the only change is that the point estimate for improved off-plot to piped on-plot has fallen from 14% to 13%.
  2. The reduction varies by level of service attained. Piped water reduces diarrhoeal morbidity more than off-plot supplies, when provided at high quality and continuity. The continuity effect most likely happens via a direct increase in quality, though also via the fact that people are less likely to use secondary sources and improper storage.

What does this mean for the economics of water supply provision, specifically the comparison of investment options?

  1. There is an equity/efficiency trade-off to be made in investment prioritisation. Improving water quality in existing piped supplies has the largest single incremental effect (68%), and there are hundreds of millions of people in urban areas of developing countries using piped services which are not safely managed. Investments in this area are therefore important. However, any improvement in service level for populations using unimproved supplies should arguably be the highest priority on equity grounds, even if to an improved off-plot supply which has only an 11% reduction. Firstly, these populations are most likely to concentrated in rural areas where mortality risk from diarrhoea is highest. Furthermore, in such populations the 11% reduction is likely to be applied to a number of annual diarrhoea cases per capita which is higher than in most urban areas, where people also live closer to health services. Decisions obviously also take place not only on the effects side of the equation, but the cost side too.
  2. Broader benefits should also be taken into consideration. Time savings are often associated with a move from unimproved to off-plot improved supplies (and again in a move from there to on-plot piped). These have an economic value. Water quantity increases associated with a move to on-plot supplies decrease the opportunity cost of using water for handwashing, meaning it is more likely to happen. While improved continuity has a much smaller incremental impact (17%) than improved water quality (68%), improvements in continuity reduce wasted time and money invested in coping strategies and the use of secondary sources. These results are also based on a small number of studies (see below). In short, any decision should be based on a lot more than relative risk reductions.

Decision-makers are rarely faced with simple choices of ‘urban versus rural’, ‘pipes versus handpumps’, or ‘quality versus continuity’. The factors above are all built into the calculus of local government agencies and Ministries of Water when they make their investment plans. I would argue that the principle of “first a basic service for all” should be factored into any such decisions.


  • The bold blue figures are the ones for which we have studies with a direct comparison (though studies also exist comparing unimproved and piped) – see table 2 in the paper.
  • The asterisks denote that transitions to “piped, higher quality” rely on one study, and to “continuous supply” on two studies (against 6, 7 and 11 studies for other key transitions). So, these should be interpreted with caution.
  • I left out the results for point-of-use filters with safe storage to avoid cluttering the diagram, but the reductions for these are 48% (on an unimproved source). POU chlorination has limited impact after blinding is accounted for.
  • These are point estimates. Some of the upper bound 95% confidence intervals for risk ratios in table 3 are higher than 1, i.e. the result is not statistically significant.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s