Why STATcompiler data may be different from DHS final reports
The STATcompiler, by definition, is a tool designed for comparisons across countries and over time. To that end, individual country data are recalculated to match standard definitions. This often prompts questions, as users wonder why the data they see in STATcompiler don't always match the data presented in an individual final report. If you are noticing that the data in STATcompiler are different from the data presented in a final report table, check the following:
- Different time frames. Time frames for different indicators have changed over the years. For example, some maternal health indicators are based on births in the 3 years before the survey, while others are based on the 5 years before the survey. Some countries present both. Make sure your time frames are the same.
- Different denominators: Some indicator definitions have also changed as to the population of interest. For example, some indicators on youth and HIV activity present data based all young people age 15-19;others focus on only the sexually active youth. Similarly, some countries only ask married women about sexual activity, while most ask ALL women. Employment indicators used to be based on those who were currently employed, but new indicators are based on those who have been employed in the last 12 months. Some surveys interview women and men older than 49, but STATcompiler totals are always based on the standard age group of 15-49. Check your denominator.
- Country-specific definitions. Many indicators, though based on a standard, are adjusted in final reports due to country context. For example, in one country, churches are considered a "formal" source of condoms, but this is exceptional, and is not included in the STATcompiler definition of formal condom sources. Commonly used country-specific indicators include:
- Skilled attendance at delivery (i.e. who is considered a skilled attendant?)
- Improved sources of water and sanitation (i.e. which are considered to be safe drinking water sources?)
- Recommended treatments for illnesses can vary by country as Ministries of Health issue different guidance based on disease prevalence, resistance, and programming priorities.
Special note on Maternal Mortality Ratios:
These same questions are particularly relevant with the addition of the Maternal Mortality Ratio to the STATcompiler database. There are two main reasons why the maternal mortality ratios printed in individual final reports may be different than those presented in STATcompiler:
- Some surveys have collected maternal mortality information through the woman's questionnaire only, while others have collected sibling data from both women and men. STATcompiler maternal mortality statistics are based on results from the woman's questionnaire only.
- Surveys that interview only ever-married women calculate maternal mortality differently than the standard DHS survey that interviews all women.
- The time frame is also variable: some surveys use a 7 year MMR, while others use 10 year rates, and some use a 5 year rate. For the purposes of comparability, the STATcompiler defines MMR as deaths in the 7 years before the survey . Therefore, not all of the MMR data in STATcompiler will match the MMR figures presented in final reports.
The DHS Program is authorized to distribute, at no cost, unrestricted survey data files for legitimate academic research. Registration is required for access to data.
Guide to Using Datasets