Sample size calculation for anthropometry and mortality

[Yengi EmmanuelParticipant]

Dear forum members,

I kindly need your guidance. What standard SMART survey estimates should I use for estimating anthropometry and mortality sample sizes in an area were no survey has not been conducted in.

Thanks
Yengi Emmanuel

[SMARTKeymaster]

Hello,

There are several parameters you need to calculate sample size (estimated prevalence, desired precision, design effect for cluster surveys, average household size, proportion of children under 5, non-response rate, size of total population if small and level of confidence). Assuming you’re referring specifically to estimated prevalence/death rate that need significant consideration, continue reading below.

When calculating sample size, the Estimated Prevalence for the indicator of interest should be as close as possible to what you think the real prevalence will be.

1) If there is no previous information about the prevalence of your main indicator in the target population, gather all available information about the indicator of interest in the country/region. Look at different sources such as surveys in similar area/context and season, for example:

Previous survey in neighbouring areas
Surveillance data including proportions of children visiting clinics with malnutrition
Program staff from MoH, UNICEF, etc.
National demographic data.
MICS or DHS survey data done in the past or neighbouring district
Passive screening data from permanent health centre sites
Health workers, religious leaders

2) Consider how the prevalence has changed since the existing information was collected (has the situation deteriorated/improved?). Do not use previous survey results blindly without adjusting for differences between then and now. Justify your decision using contextual data such as:

Aggravating factors
Seasonality
Implementation of nutrition/feeding
Crises
Food security
Population movement

3) Determine the range of plausible values of prevalence and use the higher number (closer to 50%) to be conservative and achieve an acceptable precision.

Increasing your estimate of prevalence (closer to 50%) will increase your sample size. Similarly for mortality sample size calculation, estimate the crude death rate accessing available information to determine a range and choose the higher end to be conservative. Be careful not to grossly overestimate.

SMART Team

• This reply was modified 9 years, 12 months ago by  SMART.
• This reply was modified 9 years, 12 months ago by  SMART.

[Yengi EmmanuelParticipant]

Hi
Thank very much for your help

[Author]

Posts