Methodological Issues in the Estimation of Risk Factors of Non-Communicable Diseases in Large Scale Sample Surveys: An Indian Experience

Monday, 18 August 2014
Exhibit hall (Dena'ina Center)
Arvind Pandey, PhD , INDIAN COUNCIL OF MEDICAL RESEARCH, NEW DELHI, India
Himanshu K Chaturvedi, PhD , National Institute of Medical Statistics, New Delhi, India
Jitenkumar Singh, MS , National Institute of Medical Statistics, New Delhi, India
Atul Juneja, PhD , National Institute of Medical Statistics, New Delhi, India
Tulsi Adhikari, PhD , National Institute of Medical Statistics, New Delhi, India
Damodar Sahu, PhD , National Institute of Medical Statistics, New Delhi, India
INTRODUCTION: The complexity of data collection has been ever-increasing and becoming  challenges especially when  targeted to capture all possible variability  and achieve more precision in estimation of parameters in different dimension of large population. Despite of having several methodological options and its experiences, we have never been satisfied with the outcomes of our own designed study. In this context, we would like to discuss the methodology and design of a large population based NCD risk factors survey conducted in India and similar such design followed in other countries as per WHO.  

METHODS:  In India, a large population based survey was conducted  in seven states for estimation of non-communicable disease risk factors in different subgroups of population of each state. The states included Andhra Pradesh, Kerala and Tamil Nadu from south; Madhya Pradesh and Maharashtra from central; Uttarakhand from North and Mizoram from Northeast India covering larger geographical spread and about one third of total population of India. The survey methodology and design was finalized with the views to capture adequate required sample of individuals from both sexes in all five age groups starting from the age 15 with 10 years interval (15 to 65) for  rural and urban population. Multi-stage sampling method was used for selection of primary sampling units and households. At household level, Kish method was used for random selection of one individual from age group 15 to 54 years whereas all individuals aged 55-64 included in the sample.

RESULTS: Methodological complexity adopted in data collection was mainly to achieve precision in estimation of NCD risk factors for each subgroup.   Survey was conducted using the common protocol, but inadequate samples captured in older age subgroups resulted with low precision and reliability of estimated parameters.   

CONCLUSIONS: The deviation in the outcome from expected design invites the methodological challenge for statistical experts.

See more of: Monday Poster Session
See more of: Free Papers
<< Previous Abstract | Next Abstract >>