BACKGROUND: National malaria death rates are difficult to assess because reliably diagnosed malaria is likely to be cured, and deaths in the community from undiagnosed malaria could be misattributed in retrospective enquiries to other febrile causes of death, or vice-versa. We aimed to estimate plausible ranges of malaria mortality in India, the most populous country where the disease remains common.

METHODS: Full-time non-medical field workers interviewed families or other respondents about each of 122,000 deaths during 2001-03 in 6671 randomly selected areas of India, obtaining a half-page narrative plus answers to specific questions about the severity and course of any fevers. Each field report was sent to two of 130 trained physicians, who independently coded underlying causes, with discrepancies resolved either via anonymous reconciliation or adjudication.

FINDINGS: Of all coded deaths at ages 1 month to 70 years, 2681 (3·6%) of 75,342 were attributed to malaria. Of these, 2419 (90%) were in rural areas and 2311 (86%) were not in any health-care facility. Death rates attributed to malaria correlated geographically with local malaria transmission ratesderived independently from the Indian malaria control programme. The adjudicated results show 205,000 malaria deaths per year in India before age 70 years (55,000 in early childhood, 30,000 at ages 5-14 years, 120,000 at ages 15-69 years); 1·8% cumulative probability of death from malaria before age 70 years. Plausible lower and upper bounds (on the basis of only the initial coding) were 125,000-277,000. Malaria accounted for a substantial minority of about 1·3 million unattended rural fever deaths attributed to infectious diseases in people younger than 70 years.

INTERPRETATION: Despite uncertainty as to which unattended febrile deaths are from malaria, even the lower bound greatly exceeds the WHO estimate of only 15,000 malaria deaths per year in India (5000 early childhood, 10 000 thereafter). This low estimate should be reconsidered, as should the low WHO estimate of adult malaria deaths worldwide.

Extensive media attention has been given to a study [1] in which use of the controversial technique of verbal autopsy (VA) has produced figures for malaria deaths very different from those arising from conventional methods. (See for example a Reuters report.) But does the lack of agreement between two alternative approaches mean, as the study authors argue, that it is the much higher figure resulting from the use of VA that is correct?

The study, published in the Lancet, was conducted in India where, based on routine case reports, the World Health Organization estimates that there were 10,000–21,000 malaria deaths in 2006. The VA researchers estimated that there are in fact around 205,000 deaths annually – some 13 times the WHO figure.

The researchers based their conclusions on interviews by non-medical field workers with families or “other respondents” about each of 122,000 deaths occurring during 2001–2003 in randomly selected areas of India. They obtained for each case a half-page narrative, plus answers to questions about the severity and course of any fevers. The data were interpreted by physicians trained in disease coding.