Table 3 Combined and non-census-based denominator estimation methods

Population census-based denominator estimation methods

Census-based projections

First, the population growth rate of the defined location obtained from the most recent census report or the subnational unit statistical office was applied to the population reported from the most recent census to obtain projected population for the time period of interest. The proportion of the total population that falls within the target population as reported from the most recent census is applied to the projected population to determine the population in need of the defined RMNCH intervention[24, 44,45,46]. This method was used by several authors in different ways depending on the RMNCH indicator of interest and mostly limited to preventive services.

Census may be too old, and projections may not be accurate particularly in LMICs. Using national level growth rate for subnational estimates and treating growth rate as constant overtime is problematic.

Non-census-based denominator estimation methods

Using other RMNCH reported data to estimate the target population for the indicator of interest

This approach first identifies service indicators with similar target population for which existing household surveys show near universal (100 %) coverage level.

The target population is then obtained from the facility data and adjusted slightly for the level of non-facility use and mortality in the population[5]. This method was used by Maina et al. and applies mostly to preventive coverage indicators[5]. To estimate the denominator for immunizations given at birth, reported ANC1 attendance can be used to estimate the number of births expected and vise verse[41]. Similarly, to estimate the coverage for DPT3 or OPV3, the target population can be estimated from the reported number of children who received DPT1, OPV1 or BCG within the same time period[5, 41].

Requires high coverage of related indicators; using regional level coverage to estimate denominators at the district level does not account for variation across districts.

Facility user-based denominators

This method is a multistep approach that is used to estimate the denominator for clinical interventions. First, a defined group of people accessing regular preventive care are first tested to determine if they have the condition of interest (prevalence of the condition among a subset of all who are seeking regular preventive service at a facility). The estimated prevalence of the condition among the subset tested is then applied to the total number of people accessing the defined preventive care to determine the total population of people in need of service for the tested condition. This method was used by Audureau et al. to estimate the coverage of antiretroviral drug coverage among HIV positive pregnant women for the purpose of preventing mother-to-child transmission. For instance, to estimate the denominator for the coverage of a single dose of nevirapine among HIV positive pregnant women, the authors multiplied the number of women attending ANC at each site by the observed HIV seroprevalence at each site (number of HIV-positive pregnant women/number of HIV tested pregnant women)[49].

Under-estimate pop in need by ignoring non facility users; unlikely appropriate for Low income countries where facility access and use is low

Using survey data

Like the facility user-based estimates, this approach applies survey reported prevalence of the condition of interest to the projected total population at the subnational level to estimate the population in need of service for the condition of interest. Saito et al. used this approach to estimate anti-retroviral therapy (ART) coverage for HIV infected children[50].

Requires frequent and accurate census data which is a challenge in LMICs; using regional level coverage from survey to estimate denominators at the district level does not account for variation across districts

Using previous coverage information

This approach is used for shorter campaign-based health interventions where the denominator is unknown but there is information from previous campaigns on the same intervention. The authors did not provide any additional explanation on how this was done[47].

Does not account for changes in the target population from when the last campaign was conducted.

Combined census and non-census-based approaches

Using previous coverage and census projections

Zuber et al. used this method to estimate the target population for a polio campaign program. First, the authors retrieved the number of people who received the intervention during the previous campaign, then multiplied the number of people who received the intervention during the last campaign by a multiplicative factor corresponding to the average annual population increase of the geographic location of interest. The authors did not provide details on how the annual population growth was estimated and whether the previous coverage was considered to be 100 %[24].

Requires data from previous campaign; assumes that coverage from previous campaign is 100 %; Using national level growth rate for subnational estimates may not be appropriate; no details on the estimation process

Census projections, previous coverage estimates, infant mortality and expert panel

This approach combines census projections with previous coverage data, infant mortality, and expert panel to estimate the denominator. First, the population in need is estimated using the population growth rate from the most recent census (as described earlier). Expert panel consisting of health administrators will then use previous coverage information and the number of reported deaths to adjust the census projected target population. This method was used by Mensah et al. to estimate measles coverage in Madagascar [51]. The authors did not provide details on how the previous coverage or infant mortality is used in this process.

In addition to the census based and previous campaign data limitations outlined above; the use of input from expert panel may result in arbitrary adjustment which may vary from year to year making it impossible to look at trends overtime or performance between subnational units