What went wrong? A seriously evidence-based checklist, but when implemented, no benefit achieved.
The problem was certainly worth tackling. Uttar Pradesh has a population of 204 million; if a country rather than a province of India, it would be the 5th largest in the world. Uttar Pradesh is poor and rural, with some dreadful health indicators: neonatal mortality is 32 per 1000 births and maternal mortality an appalling 258 per 100 000 births - the comparable figures for England are 2.6 and 8.5.
So implementing a checklist of proven interventions should have worked, but it didn't. Why not? There is a telling remark in the discussion section:
'Quality of care in control sites, as measured through birth attendants’ adherence to practices, was far lower than previously recognised.'
If things were that bad, maybe they needed more focus on the behavioural intervention. The researchers probably framed their task as implementing a checklist; what they were actually trying to achieve was substantial behaviour change, starting from a very low baseline, in a very difficult target group (the birth attendants). Perhaps a model based on Michie's behaviour change wheel would have worked better.
The other problem is that the checklist contained 28 'Essential' items. Maybe they really were all Essential, but that's a lot of change to cope with. Could you do 28 new things? Here is an interesting critique of checklist design and use.