How to Calculate Denominators to Track Indicators from Health Facility Data (Part 1 of 2)

Health facility data typically have provided only the numerator for indicators of interest, which makes it difficult to interpret time trends. But it is possible to calculate denominators based on two methods: population projection estimates and service utilization.  This article explains how to calculate denominators using population data, and a separate post describes how to calculate denominators using service utilization data. The most accurate method will depend on the quality of the available data in a particular setting.

Below are some indicators included in this resource.


Population-based estimates

Projection of national population census data can be used to estimate the denominators for health facility indicators, since these datasets report the population by age and sex for the country and subnational areas. However, census projections may have errors in the population count, in the projection assumptions about fertility, mortality, and migration, or both. Projections for small areas such as districts are particularly prone to inaccuracies. This is complicated by the fact that censuses are conducted only one per decade at the most.

The steps for calculating denominators based on census projections and assessing their validating are shown below.

Compiling population projections

The DHIS 2 platform for each country includes population projection indicators for total population, age less than 1, age less than 5, women aged 15-49, and the total number of live births. Population projection is also an output of the data quality assessment described in a separate blog post.

Compute demographic parameters

Tools for computing population growth rate, crude birth rate, and crude death rate are available in Excel and Stata. Below are the formulas for calculating population growth, crude birth rate, and crude death rate.

Evaluate the demographic parameters

It is important to evaluate the demographic parameters according to consistency over time; large variation between years would not be expected. It is also useful to compare estimates with the United Nations population projections. Results for crude birth rates can be compared to Demographic Health Survey Program’s STATcompiler.

In the example  graph below, we can see that DHIS 2 and UN estimates are similar for women age 15-49 and children under age 1 for all years. However, for children under age 5, DHIS 2 estimates were higher than UN estimates in 2019-2021, while UN estimated a higher number of live births every year. Comparing ratios at the subnational level can also yield useful insights.

Compare coverage for highest coverage interventions

If the estimated denominators are correct, they should be close to the numbers reported for interventions know to have high coverage. The population-based Demographic Health Surveys can be used to identify interventions with the highest population level coverage in a country; usually these are first antenatal visit (ANC1) and first dose of diptheria-pertussis-tetanus vaccine (DPT1). For ANC1, the denominator is number of pregnancies; use the formula below to estimate the number of pregnancies based on the number of live births.


Similarly, for DPT1/BCG coverage, we need to estimate the number of eligible infants from the live births projection in DHIS2, using the formula shown below. For this, we subtract the number of infants who died within six weeks of birth from the number of live births. The neonatal mortality rate is used to estimate the number of infants who died. For example, if there were 1000 live births, and neonatal mortality rate was 30 per 1000 live births, then there would be 970 eligible infants.

Graphing the number of ANC1, DPT-1 and projected live births will show if the estimates are consistent. We can see in the example graph below that the number of projected live births is lower than DPT-1 utilization but higher than ANC-1.





Thus, when coverage is estimated using the project live births as a denominator, the values exceed for DPT-1, while they may underestimate ANC1 coverage compared to DHS survey data:


Since this is a common problem, another method of calculating denominators is available, as described in the next article .

This article builds on previous articles which describe the analysis workshop held in June 2022 and explain how to assess data quality in routine aggregated health facility data.

Resources related to this article:

  • A PowerPoint presentation describing how to complete this analysis.
  • Stata code that produces a log file, Excel outputs, and graphs for national and subnational levels. Make sure to indicate your working folder, specify your country, and change default values as relevant. For the graphs, you may need to modify the labels to adapt to your specific country.
  • An Excel workbook template. Based on Niger data, it includes data sheets and analysis sheets with graphs. The analysis sheets include formulas that need to be adapted to your country’s specific.
  • Dataset of United Nations population projections.

Additional resources are available on the health facilities data & analysis center page.