India had a malaria strategy, a national framework, a budget line and a set of international funders. What it did not have was a number anyone trusted. I worked on this stakeholder audit with five colleagues, and the finding that reorganized the whole document was not a health finding at all.
Four figures for one year
For 2013 the government of India reported 440 malaria deaths.
The Medical Certification of Cause of Death report for the same year, released by the Registrar General of India in December 2015, put the figure at 5,428. A government committee chaired by the Indian Council of Medical Research estimated 40,297 deaths annually, roughly forty times the official number, and that report was never published by the government. A study in The Lancet estimated 205,000 malaria deaths a year in India, and found that 90 percent occurred in rural areas and 80 percent without proper medical attention.
Four published figures for the same disease in the same country, spanning a range of ten to a hundred times. Incidence was being under-reported by an amount that made the strategy unfalsifiable.
Why the miscount is the whole problem
State funding for malaria is allocated against reported incidence. Lower reported incidence means lower funding, which means fewer people, less surveillance and less medicine, which means fewer cases get recorded. The counting failure causes the funding failure, which causes the delivery failure. The document is structured as four failures rather than four opportunities, because opportunities let a reader nod and move on.
The funding numbers are stark once you follow them to the district. In one malaria-endemic district in Odisha, annual funding worked out at about ₹4 million for 1.5 million people, roughly ₹2.5 per person, of which about ₹1 per person survived establishment costs. On delivery, government statistics for 2013 showed that of 148,124 male health workers needed at medical sub-centers, only 83,241 positions were sanctioned and 52,215 were actually filled.
The finding came from the bottom of the org chart
I interviewed across the entire ecosystem, from the Principal Adviser for Health at NITI Aayog, the head of the national vector-borne disease program, AIIMS, the Public Health Foundation of India and Medicines for Malaria Venture, down to district collectors, district malaria officers, block supervisors and village health workers.
The sharpest thing in the report came from an ASHA worker in Junagadh Block, Kalahandi, Odisha. She was paid between ₹25 and ₹75 for malaria-related work, ₹75 if a blood slide came back positive, ₹25 for taking someone in for a test, while maternal and child health reporting paid her better for the same hours. So the person at the very end of the reporting chain, the one whose entry becomes the national number, was financially better off not finding malaria.
That is an incentive designed by accident, and neither compliance training nor communications spending fixes it. I have put a village health worker and a national policy adviser on the same evidentiary footing ever since, and let the evidence decide which one matters more on a given question.
What the audit produced
The stakeholder ecosystem was built up one layer at a time, starting with the central government and adding technical bodies, state governments, civil society organizations and pharmaceutical companies, so the reader saw the system assemble rather than meeting it finished. It ended in a matrix of roles, motivations and engagement vectors across those five groups, which is the part a client can actually act on: what each actor is for, what they want, and the one lever that moves them.
The outcome was a change of premise. The client had commissioned an audit of a health advocacy problem and received a diagnosis of a measurement problem, with the causal chain from miscount to funding to delivery written out and sourced. The recommendation that followed was to attack the data first, including using drug sales data to triangulate real incidence, because every other intervention is being sized against a number that is wrong.
What I would do differently
I would have built the drug-sales triangulation as a running estimate rather than proposing it, even in rough form on a single state. A recommendation to fix the instrument is easy to defer. A working instrument that disagrees with the official number every month is not.