This was a cross sectional study carried out in the Kalutara District, which has urban, semi-urban, rural and plantation sectors . Adults who were in the age group of 35 to 64 years were selected for the study. The sample size was calculated to determine a prevalence of 16%  with a margin of error at 3% and α error at 5%. Since a value for roh was not available from earlier studies a cluster factor of 2 was taken. The calculated sample size of 1147 was further inflated to 1262 in order to account for a drop-out rate of 10%, which was then rounded off to 1300. Stratified multistage cluster sampling was used. Since people of similar socioeconomic status tend to cluster together, the cluster size was restricted to twenty to accommodate more clusters in the study. This was expected to produce a wider scatter of the sample.
The first level of stratification was urban, rural and estate sectors. The rural and plantation sectors were over sampled in order to ensure adequate numbers of subjects with DM from these sectors. The final sample consisted of 600 individuals from the rural sector, 400 from urban sector and 300 from estate sector.
The primary sampling unit was the Grama Niladari Division (GND) (village officer division). GNDs were randomly selected within each sector. Probability Proportionate to Size sampling method was used and the probability of a GND being selected was proportionate to its population of 35 to 64 year olds. In each GND, individual households were selected randomly using the electoral registry.
Only those who had resided in the selected household for a continuous period of more than one year were included in order to ensure that all participants had an established lifestyle, related to the area of residence. All the adults who are eligible for the study in each household were listed and one individual was selected randomly. If the selected individual was not available, subsequent visits were made at a time he/she is available for data collection. Those living in institutions, pregnant and lactating females, and subjects on prolonged treatment with drugs known to cause diabetes mellitus were excluded.
An interviewer administered structured questionnaire was used as the study instrument. The tools used for the study were validated for Sri Lanka. The questionnaire was pre-tested and further improvements were made. The data collectors were trained in standard research methods and ethical principles. Fasting plasma glucose (FPG) levels were measured using venous blood samples after an overnight fast of 10 hours. Analysis was done at the Public Health Laboratory of National Institute of Health Sciences (Kalutara), using Clini Check Plus Mini Analyser using hexokinase-glucose 6 phosphate dehydrogenase method with colorimetry. Diabetes was defined as having one or more of the following criteria
Study participants with FPG level of =126 mg/dl 
Currently (within the past 4 weeks) on insulin
Currently (within the past 4 weeks) on hypoglycaemics
FPG ≥100mg/dl but <126mg/dl is considered as Impaired Fasting Glucose (IFG) . Measures were taken to improve the quality and accuracy of data in the design and implementation stages of the study. Steps were taken to minimize sampling and measurement errors including quality control of the laboratory analysis of FPG. These included using a block design to assess the variation between instruments and observers and measuring the correlation coefficients of inter and intra assay precision checks for plasma glucose.
Social status index was assessed as described by De Silva  and Unsatisfactory Basic Needs Index (UBNI) as described by Satharasinghe in 2008 .
After estimating crude prevalence, weight adjustments were made to correct for different probabilities. All results presented for the district are weighted and prevalence standardised for age and sex. Significance of prevalence of diseases and risk factors across different socio-economic strata were determined by chi square test for trend.