2013 Conference of the International Medical Geology Association (25–29 August 2013)

Paper No. 3
Presentation Time: 3:50 PM


ROYCHOWDHURY, Koel, Institute of Sustainability and Peace, United Nations University, 53-70 Jingumae 5-chome, Shibuya-ku, Tokyo, 150-8925, Japan and JONES, Simon, School of Mathematical and Geospatial Sciences, RMIT University, 414-418 Swanston Street, Building 8, Level 9, Room 1, Melbourne, 3000, Australia, roychowdhury@unu.edu

With a population of 1.21 billion and a GDP of more than US$1845 billion, India is the world’s largest democracy and the tenth largest economy. At the same time, India accounts for 21% of globally reported diseases. Though there has been a reduction in communicable diseases over the last two decades, two thirds of India’s total morbidity burden comes from non-communicable diseases. More than 6% of annual GDP is lost due to premature deaths and preventable illness in India (World Bank Report (2010)).

India’s health problems can be better tackled by utilizing disaggregated data to support planning and policy in the health sector. The Annual Health Survey (AHS) (2010 – 13) is conducted in nine Indian states with the highest population growth, including: Bihar, Jharkhand, Uttar Pradesh, Uttarakhand, Madhya Pradesh, Chhattisgarh, Orissa, Rajasthan and Assam. However AHS is mainly undertaken by means of household interviews. Due to logistical and geographic constraints it was conducted on a sample of only 20.1 million population from 4.1 million households.

This paper uses the data from the second round AHS along with night-time satellite images (captured by the VIIRS sensor on board the SUOMI group of satellites—the first ever cloud free night-time global composite captured for the sub-continent with a spatial resolution of approximately 500m) to predict health indicators across India. The day/night band of the VIIRS sensor provides detailed variation of radiance within settlements. Models are proposed to predict important metrics of health such as: crude birth rate, crude death rate, infant mortality rate, maternal mortality rate and gender ratio. Data from 210 districts was used to propose models (validated on the 74 districts of the nine selected states), which were then used to predict metrics for other districts. Results show that the models have an adjusted r2 = 0.75 for p < 0.05 for the metrics. Districts with more urban population such as Delhi, Mumbai, Chennai, Bangalore and Hyderabad have less incidence of mortality compared to rural populations. Maps were produced for these metrics for the country as a whole. The proposed method has utility in estimating metrics of health at a variety of sub-national spatial levels. This will enable promotion of affordable and efficient quality health services to remote areas of India.