Coverage estimates¶

Background and purpose¶

Objective of the module¶

The Coverage Estimates module quantifies health service coverage by integrating adjusted administrative service volumes from the Health Management Information System (HMIS), population projections from the United Nations World Population Prospects (UN WPP), and household survey data. While the module currently draws on Demographic and Health Surveys (DHS) and Multiple Indicator Cluster Surveys (MICS), it is designed to accommodate other nationally representative survey sources as they become available. The module estimates the share of the target population that received a given health service, providing a standardized measure of service reach for use in monitoring, comparison, and downstream analysis. The module is structured in two components.

Part 1 constructs target population denominators using multiple methodological approaches and evaluates their performance by comparing resulting coverage estimates with available survey reference values for each health indicator.

Part 2 allows users to review and adjust denominator selections based on programmatic considerations and to extend survey-based coverage estimates over time using trends derived from administrative data, where survey data are not available.

Together, these components convert administrative service volumes into standardized coverage estimates that can be examined over time and across geographic levels, and used in analytical and monitoring contexts.

Analytical rationale¶

Health service coverage is a core metric for assessing health system performance and equity. While Module 2 produces adjusted service volumes, these figures on their own do not indicate the extent to which services reach the populations they are intended to serve. Coverage estimates place service delivery in context by relating service volumes to population need.

This module addresses key challenges in estimating coverage, including:

• Multiple data sources: Integrates HMIS data with survey data

• Denominator uncertainty: Different methods for estimating target populations may yield different results; the module systematically evaluates options

• Temporal gaps: Surveys occur every 3-5 years; the module projects estimates for intervening years using administrative trends

• Subnational analysis: Enables coverage monitoring at national, provincial, and district levels

Key points¶

Part 1 and part 2 explained¶

Part 1: Denominator calculation and selection

• Calculates target populations (denominators) using multiple approaches: HMIS-based (from ANC1, delivery, BCG, Penta1) and population-based (UN WPP)

• Compares coverage estimates from each denominator against survey data

• Automatically selects the "best" denominator for each indicator by minimizing error

• Outputs: Denominator datasets and combined results showing all options

Part 2: Denominator selection and survey projection

• Allows users to override automatic selections and choose specific denominators

• Calculates year-over-year coverage trends from administrative data

• Projects survey estimates forward using HMIS trends to fill temporal gaps

• Outputs: Final coverage estimates combining HMIS, survey, and projected values

Analytical workflow¶

Overview of analytical steps¶

Part 1: Denominator calculation and selection¶

Step 1: Load and prepare data sources The module begins by loading three data sources and ensuring they are compatible. HMIS data is aggregated from monthly to annual totals. Survey data is harmonized (DHS prioritized over MICS) and forward-filled to create continuous time series. Population data is filtered to the target country.

Step 2: Calculate multiple denominator options For each health indicator, the module calculates several possible target populations:

• Service-based denominators: Using HMIS volumes divided by survey coverage (e.g., if 10,000 women received ANC1 and survey says coverage is 80%, estimated pregnancies = 10,000/0.80 = 12,500)

• Population-based denominators: Using UN population projections and birth rates

• Each denominator is adjusted for demographic factors (pregnancy loss, stillbirths, mortality rates) to match the indicator's target age group

Step 3: Calculate coverage for each denominator The module computes coverage by dividing the service volume by each denominator option. This produces multiple coverage estimates per indicator, each based on a different population assumption.

Step 4: Compare to survey benchmarks Each coverage estimate is compared to survey data using squared error calculation. The survey serves as the benchmark since it is based on representative household sampling.

Step 5: Select the best denominator The denominator producing the lowest error (closest match to survey) is automatically selected as "best." The selection prioritizes HMIS-based denominators over population projections to ensure data is driven by observed service delivery.

Step 6: Generate outputs The module saves denominator datasets for transparency and combined results files showing coverage from all denominators plus the selected best option.

Step 7: Repeat for subnational levels If subnational data is available, the process repeats for administrative level 2 (e.g., provinces) and level 3 (e.g., districts), with fallback mechanisms to handle missing local survey data.

Part 2: Denominator selection and survey projection¶

Step 1: User configuration Users review Part 1 results and configure denominator selections for each indicator. Options include using the automatic "best" selection or overriding with a specific denominator based on programmatic knowledge.

Step 2: Filter to selected denominators The module filters Part 1's combined results to include only user-selected denominators, creating a focused dataset for analysis.

Step 3: Calculate coverage trends Year-over-year changes (deltas) in HMIS-based coverage are calculated. This shows whether coverage is increasing, decreasing, or stable over time.

Step 4: Identify survey baseline For each geographic area and indicator, the most recent survey observation is identified as the baseline anchor point for projections.

Step 5: Project survey estimates forward The module extends survey coverage estimates into years without surveys by applying HMIS trends. The projection uses: Last survey value + (Current year HMIS coverage - Survey year HMIS coverage). This preserves the survey calibration while incorporating observed trends.

Step 6: Combine all estimates The final output merges three types of estimates:

• HMIS-based coverage: Direct calculation from service volumes and selected denominators

• Original survey values: Actual household survey observations

• Projected survey coverage: Survey estimates extended using HMIS trends

Step 7: Save final outputs Results are saved with standardized column structures for each administrative level, ready for visualization and reporting.

Workflow diagram¶

Key decision points¶

1. Selection of denominators

In Part 1, the module automatically selects denominator options based on their alignment with available survey reference values. In Part 2, users may review and override these selections based on programmatic knowledge or analytical priorities. The choice of denominator determines whether coverage estimates are primarily anchored to observed service delivery patterns (HMIS-based denominators) or to demographic projections (population-based denominators).

2. Treatment of gaps between surveys

Household surveys are conducted at irregular intervals, typically every three to five years. In Part 1, survey values are forward-filled between survey years, implicitly assuming constant coverage until the next survey observation. In Part 2, coverage is projected forward using trends derived from HMIS data, allowing changes in service delivery to be reflected in periods without survey data.

3. Use of national versus subnational survey data

For immunization indicators only, when subnational survey estimates are not available, the module applies national survey values to subnational units as a fallback. This approach assumes that national immunization coverage rates are broadly representative at subnational levels, an assumption that may not hold in all settings. This fallback mechanism is not applied to other indicators, such as maternal or child health services, for which subnational analysis requires locally observed survey data.

4. Adjustment of denominators for target populations

Each health indicator corresponds to a specific target population (for example, pregnant women for antenatal care or infants for childhood vaccination). The module applies sequential demographic adjustments—such as pregnancy loss, stillbirths, and mortality—to align denominators with the relevant target population for each indicator.

Data processing and outputs¶

Input integration

The module integrates three primary data sources: annualized HMIS service volumes aggregated by geographic unit; household survey coverage estimates harmonized across survey rounds and forward-filled to create continuous time series; and population projections filtered to extract age- and sex-specific populations relevant to each health indicator.

Denominator construction

Using the relationship between reported HMIS service volumes and survey-based coverage estimates, the module derives HMIS-implied denominators representing the population size consistent with observed service delivery and survey coverage levels. These denominators are further adjusted to reflect indicator-specific target populations through sequential demographic corrections, including pregnancy loss, stillbirths, and mortality.

Coverage calculation

Multiple coverage estimates are calculated by dividing service volumes by alternative denominator options, including population-based, HMIS-implied, and hybrid approaches. Each coverage estimate is evaluated against survey reference values to assess plausibility and to inform denominator selection for each indicator.

Temporal projection

For years beyond the most recent survey observation, coverage estimates are projected forward by combining the last observed survey value with trends derived from HMIS data.

Analysis outputs and visualization¶

The FASTR analysis generates coverage estimate visualizations at multiple geographic levels:

1. Coverage calculated from HMIS data (national)

National-level coverage trends comparing HMIS-derived estimates against survey benchmarks.

2. Coverage calculated from HMIS data (admin area 2)

Coverage patterns at an intermediate subnational level (admin_area_2), highlighting geographic variation in service delivery across regions.

3. Coverage calculated from HMIS data (admin area 3)

Coverage estimates at a finer subnational level (admin_area_3), supporting more localized monitoring and identification of subnational disparities.

Interpretation guide

For all coverage charts (outputs 1–3):

• Black line/points: Survey-based coverage (DHS/MICS) — the reference standard
• Grey line/points: HMIS-based coverage calculated from facility data
• Red line/points: Projected coverage extending survey estimates using HMIS trends
• Y-axis: Coverage percentage (0–100%)
• X-axis: Time period (years)

Geographic levels:

• Output 1: National-level trends
• Output 2: Admin area 2 (regional/provincial) breakdown
• Output 3: Admin area 3 (district) breakdown for local targeting

Detailed reference¶

Part 1: Denominator calculation (technical details)¶

Configuration parameters¶

The module begins with several configurable parameters that control the analysis:

COUNTRY_ISO3 <- "ISO3"                         # ISO3 country code (e.g., "RWA", "UGA", "ZMB")
SELECTED_COUNT_VARIABLE <- "count_final_both"  # Which adjusted count to use
ANALYSIS_LEVEL <- "NATIONAL_PLUS_AA2"          # Geographic scope

Analysis level options:

• NATIONAL_ONLY: National-level analysis only
• NATIONAL_PLUS_AA2: National + administrative area 2 (e.g., provinces)
• NATIONAL_PLUS_AA2_AA3: National + admin area 2 + admin area 3 (e.g., districts)

Demographic adjustment rates:

PREGNANCY_LOSS_RATE <- 0.03      # 3% pregnancy loss
TWIN_RATE <- 0.015               # 1.5% twin births
STILLBIRTH_RATE <- 0.02          # 2% stillbirths
P1_NMR <- 0.039                  # Neonatal mortality rate
P2_PNMR <- 0.028                 # Post-neonatal mortality rate
INFANT_MORTALITY_RATE <- 0.063   # Infant mortality rate
UNDER5_MORTALITY_RATE <- 0.103   # Under-5 mortality rate

Count variable options:

• count_final_none: No adjustments (raw reported data)
• count_final_outlier: Outlier adjustment only
• count_final_completeness: Completeness adjustment only
• count_final_both: Both adjustments (recommended)

Input data sources¶

Part 1 integrates three primary data sources:

1. HMIS Adjusted Data (from Module 2)

• National: M2_adjusted_data_national.csv
• Subnational: M2_adjusted_data_admin_area.csv
• Contains service volumes by indicator, area, and time period

2. Survey Data (DHS/MICS)

• Source: GitHub repository (unified survey dataset)
• Provides coverage benchmarks for comparison
• DHS data prioritized over MICS when both available

3. Population Data (UN WPP)

• Source: GitHub repository
• Provides population-based denominators
• Includes total population, births, under-1, and under-5 populations

Additional data context:

Population projections (UN WPP) Sourced from the United Nations World Population Prospects, these estimates provide age-specific and total population figures used to calculate denominators for coverage estimates. These projections account for demographic trends, including fertility, mortality, and migration.

Survey data - MICS MICS, conducted by UNICEF, provide household survey-based estimates for key health indicators, including coverage of maternal and child health services.

Survey data - DHS DHS, conducted by USAID, provide survey data on health service utilization, including immunization rates and maternal care coverage.

Core functions documentation¶

admin_area_1  admin_area_2  year  countanc1  countdelivery  ...  nummonth
Country_Name  Province_A    2020  12500      10200          ...  12
Country_Name  Province_A    2021  13000      10500          ...  11

denominator_adjusted = denominator × (nummonth / 12)

coverage = (service_volume / denominator) × 100

squared_error = (HMIS_coverage - survey_coverage)²

Priority 1: Independent denominators (non-reference, non-UNWPP) → lowest error
Priority 2: Reference-based denominators (only if no independent available)
Priority 3: UNWPP denominators (last resort fallback)

admin_area_1  year  indicator_common_id  denominator_best_or_survey  value
Country_Name  2020  anc1                 best                        85.3
Country_Name  2020  anc1                 survey                      84.2
Country_Name  2020  anc1                 danc1_pregnancy             85.3
Country_Name  2020  anc1                 dwpp_pregnancy              82.1

Statistical methods & algorithms¶

na.locf(survey_value, na.rm = FALSE)

Year:   2015  2016  2017  2018  2019  2020
Raw:    85.3  NA    NA    NA    87.2  NA
Filled: 85.3  85.3  85.3  85.3  87.2  87.2

Conceptual framework: Demographic cascades¶

Before presenting the specific formulas, it is important to understand the conceptual flow of denominator calculations. Denominators are derived through sequential demographic adjustments that reflect the biological cascade from pregnancy to specific health service target populations.

Illustrative example: From pregnancy to DPT-eligible population

Consider how an estimated 10,000 pregnancies translate to the population eligible for DPT vaccination:

Starting point (pregnancies):           10,000
→ After pregnancy loss (3%):            10,000 × (1 - 0.03) = 9,700 deliveries
→ After twin adjustment (1.5% rate):    9,700 × (1 - 0.015/2) = 9,627 births
→ After stillbirths (2%):               9,627 × (1 - 0.02) = 9,435 live births
→ After neonatal deaths (3.9%):         9,435 × (1 - 0.039) = 9,067 DPT-eligible children

This cascade demonstrates how each demographic factor sequentially reduces the population size as we move through life stages. The detailed mathematical formulas in the following sections follow this same logic, but work in both directions:

• Forward cascade: Starting from earlier indicators (ANC1, Delivery) and adjusting toward later target populations
• Backward cascade: Starting from later indicators (BCG, Penta1) and working backwards to estimate earlier populations

The specific rates and formulas for each denominator source are provided in detail below.

HMIS-based Denominator Calculations¶

Denominators derived from ANC1

Starting from ANC1 service counts and survey coverage, we calculate:

Estimated pregnancies (base calculation):

Estimated deliveries (adjusted for pregnancy loss):

Estimated births (adjusted for twin births):

Estimated live births (adjusted for stillbirths):

Population eligible for DPT/Penta vaccines (adjusted for neonatal mortality):

Population eligible for MCV1 (adjusted for post-neonatal mortality):

Population eligible for MCV2 (adjusted for additional post-neonatal mortality):

Denominators derived from delivery

Starting from institutional delivery counts and survey coverage:

Estimated live births (base calculation):

Estimated births (adjusted for stillbirths):

Estimated pregnancies (adjusted for twin births and pregnancy loss):

Population eligible for DPT/Penta vaccines:

Population eligible for MCV1:

Population eligible for MCV2:

Note: Denominators derived from Skilled Birth Attendance (SBA) follow the same formulas as delivery denominators.

Denominators derived from BCG (National analysis only)

Starting from BCG vaccination counts and survey coverage:

Estimated live births (base calculation):

Estimated pregnancies (working backwards through demographic adjustments):

Population eligible for DPT/Penta vaccines:

Denominators derived from Penta1

Starting from Penta1 vaccination counts and survey coverage:

Population eligible for DPT/Penta vaccines (base calculation):

Population eligible for MCV1:

Population eligible for MCV2:

Denominators derived from live birth counts

When live birth data is directly reported in HMIS:

Estimated live births (base calculation):

Estimated pregnancies (working backwards):

Estimated deliveries:

Estimated births:

Population eligible for DPT/Penta vaccines:

Population eligible for MCV1:

Population eligible for MCV2:

UNWPP-based Denominator Calculations¶

Denominators derived from UN WPP (National analysis only)

Instead of using service volumes, these denominators are calculated directly from population projections and demographic rates:

Estimated pregnancies (from crude birth rate and total population):

Estimated live births (from crude birth rate):

Population eligible for DPT/Penta vaccines (under-1 population):

Population eligible for MCV1 (adjusted for neonatal mortality):

Population eligible for MCV2 (adjusted for post-neonatal mortality):

Adjustment for Incomplete Reporting:

When HMIS data contains fewer than 12 months of reported data in a year, all UNWPP denominators are scaled to match the reporting period:

This adjustment ensures denominators are comparable to service volumes that may only represent partial-year reporting.

Denominators derived from live birth estimates (secondary calculations)

After all primary live birth denominators are calculated (from ANC1, Delivery, BCG, Penta1, Live Birth Counts, and WPP), the module generates additional target population estimates for specific interventions by applying age-specific mortality adjustments:

Children aged 6-59 months (Vitamin A supplementation target population)

For each live birth denominator source, the estimated number of children aged 6-59 months is calculated:

Where:

• source represents any of: anc1, delivery, bcg, penta1, livebirths, or wpp
• The factor 4.5 represents the approximate duration (in years) of the Vitamin A target age range (6-59 months ≈ 4.5 years)
• Under-5 mortality rate adjusts for child survival to reach the 6-59 month age range
• Result: Estimated population of children aged 6-59 months eligible for Vitamin A supplementation

Infants under 12 months (fully immunized child target population)

For each live birth denominator source, the estimated number of infants under 12 months is calculated:

Where:

• source represents any of: anc1, delivery, bcg, penta1, livebirths, or wpp
• Infant mortality rate adjusts for survival to 12 months of age
• Result: Estimated population of infants under 1 year old eligible for full immunization assessment

These target population estimates are calculated automatically for all available live birth denominators, ensuring consistent methodology across different source indicators.

Workflow execution steps¶

Part 1 executes the following workflow for each administrative level (national, admin2, admin3):

Step 1: Load and validate input data

• Load HMIS adjusted data from Module 2 (national and subnational files)
• Load survey data from GitHub repository (unified DHS/MICS dataset)
• Load UN WPP population data from GitHub repository
• Validate ISO3 codes match across datasets
• Aggregate monthly HMIS data to annual totals
• Harmonize survey data (DHS prioritized over MICS)
• Forward-fill survey values to create continuous time series

Step 2: Calculate HMIS-based denominators

• For each health indicator with survey coverage data:
• Calculate base denominator: count ÷ survey_coverage
• Apply demographic cascades to derive related denominators
• Generate denominators from all available source indicators (ANC1, Delivery, BCG, Penta1, Live Births)

Step 3: Calculate WPP-based denominators

• Extract population projections for target country
• Calculate pregnancy estimates from crude birth rate
• Calculate live birth estimates
• Generate under-1 population denominators
• Apply mortality adjustments for vaccine-eligible populations
• Adjust for incomplete reporting periods (months reported < 12)

Step 4: Calculate secondary denominators

• For each *_livebirth denominator:
• Calculate Vitamin A denominator: livebirth × (1 - U5MR) × 4.5
• Calculate Fully Immunized denominator: livebirth × (1 - IMR)

Step 5: Calculate coverage estimates

• Divide HMIS service volume by each denominator option
• Create coverage estimates for all indicator-denominator combinations
• Preserve survey-based coverage as benchmark

Step 6: Select best denominator

• For each indicator, compare all denominator-based coverage estimates to survey data
• Calculate squared error: Σ(coverage_d,t - survey_t)²
• Select denominator with minimum error as "best"
• Apply preference rules (HMIS-based preferred over WPP)
• Flag denominators as "reference" if from same service

Step 7: Format and save outputs

• Save denominator files with source and target metadata
• Save combined results with all coverage estimates
• Mark best denominator for easy filtering
• Include survey values in output
• Create separate files for national, admin2, and admin3 levels
• Generate empty files with correct structure for unavailable admin levels

admin_area_1, [admin_area_2/3], year, denominator, source_indicator, target_population, value

admin_area_1, admin_area_3, year, indicator_common_id, denominator_best_or_survey, value

indicator_common_id, denominator_national, denominator_admin2, denominator_admin3

Part 2: Denominator selection and survey projection (technical details)¶

Purpose and Objectives¶

Part 2 serves three key purposes:

1. User-driven denominator selection: While Part 1 automatically selects the "best" denominator by minimizing error against survey data, Part 2 allows users to override this selection and choose specific denominators based on programmatic knowledge or policy priorities

2. Temporal trend analysis: Computes year-over-year changes (deltas) in coverage to understand service delivery trends over time

3. Survey projection: Projects survey-based coverage estimates forward in time using trends observed in administrative (HMIS) data, filling gaps where survey data is unavailable

User configuration¶

Users configure Part 2 through two key parameter sets:

DENOMINATOR_SELECTION <- list(
  # PREGNANCY-RELATED INDICATORS
  anc1 = "best",                    # Options: "best", "danc1_pregnancy", "ddelivery_pregnancy", "dbcg_pregnancy", "dlivebirths_pregnancy", "dwpp_pregnancy"
  anc4 = "best",

  # LIVE BIRTH-RELATED INDICATORS
  delivery = "best",                # Options: "best", "danc1_livebirth", "ddelivery_livebirth", "dbcg_livebirth", "dlivebirths_livebirth", "dwpp_livebirth"
  bcg = "best",
  sba = "best",
  pnc1_mother = "best",
  pnc1 = "best",

  # DPT-ELIGIBLE AGE GROUP INDICATORS
  penta1 = "best",                  # Options: "best", "danc1_dpt", "ddelivery_dpt", "dpenta1_dpt", "dbcg_dpt", "dlivebirths_dpt", "dwpp_dpt"
  penta2 = "best",
  penta3 = "best",
  opv1 = "best",
  opv2 = "best",
  opv3 = "best",

  # MEASLES-ELIGIBLE AGE GROUP INDICATORS
  measles1 = "best",                # Options: "best", "danc1_measles1", "ddelivery_measles1", "dpenta1_measles1", "dbcg_measles1", "dlivebirths_measles1", "dwpp_measles1"
  measles2 = "best",

  # ADDITIONAL INDICATORS
  vitaminA = "best",                # Options: "best", "danc1_vitaminA", "dbcg_vitaminA", "ddelivery_vitaminA", "dwpp_vitaminA"
  fully_immunized = "best"          # Options: "best", "danc1_fully_immunized", "dbcg_fully_immunized", "ddelivery_fully_immunized", "dwpp_fully_immunized"
)

RUN_NATIONAL <- TRUE  # Always TRUE - national analysis is mandatory
RUN_ADMIN2 <- TRUE    # Enable/disable admin level 2 analysis
RUN_ADMIN3 <- TRUE    # Enable/disable admin level 3 analysis

Core functions and methods¶

coverage_deltas <- function(coverage_df, lag_n = 1, complete_years = TRUE)

project_survey_from_deltas <- function(deltas_df, survey_raw_long)

build_final_results <- function(coverage_df, proj_df, survey_raw_df = NULL)

Helper functions¶

Workflow execution steps¶

Part 2 executes the following workflow for each administrative level (national, admin2, admin3):

Step 1: Load data

• Load combined results from Part 1 for all admin levels
• Check which admin levels have data
• Extract survey data for use as projection baseline
• Display messages about data availability

Step 2: For each admin level

Sub-step 1: Filter by denominator selection

• Apply user's denominator choices using filter_by_denominator_selection()
• Message: Number of records selected

Sub-step 2: Compute deltas

• Calculate year-over-year coverage changes using coverage_deltas()
• Creates complete time series with gaps filled

Sub-step 3: Project survey values

• Use project_survey_from_deltas() to extend survey estimates
• Baseline is anchored to most recent survey
• Projections use cumulative deltas from HMIS trends

Sub-step 4: Build final results

• Combine HMIS coverage, projections, and original surveys
• Calculate final projected estimates using additive formula
• Preserve all metadata

Step 3: Standardize and save outputs

• Define required columns for each admin level
• Ensure all required columns exist (add as NA if missing)
• Order columns correctly
• Remove inappropriate admin level columns
• Save as CSV with UTF-8 encoding
• Create empty files for admin levels with no data

Output specifications¶

Part 2 produces three output files:

1. National Output: M5_coverage_estimation_national.csv¶

Columns:

• admin_area_1: Country name
• year: Year of estimate
• indicator_common_id: Standardized indicator code
• denominator: Selected denominator source
• coverage_original_estimate: Original survey-based coverage (NA for years without surveys)
• coverage_avgsurveyprojection: Projected survey coverage using HMIS trends
• coverage_cov: HMIS-based coverage estimate
• survey_raw_source: Survey source (e.g., "DHS 2018")
• survey_raw_source_detail: Additional source details

2. Admin Level 2 Output: M5_coverage_estimation_admin2.csv¶

Columns:

Same as national, plus:

• admin_area_2: Second-level administrative division name (e.g., province, region)

3. Admin Level 3 Output: M5_coverage_estimation_admin3.csv¶

Columns:

• admin_area_1: Country name
• admin_area_3: Third-level administrative division name (e.g., district)
• year: Year of estimate
• indicator_common_id: Standardized indicator code
• denominator: Selected denominator source
• coverage_original_estimate: Original survey coverage
• coverage_avgsurveyprojection: Projected survey coverage
• coverage_cov: HMIS-based coverage
• survey_raw_source: Survey source
• survey_raw_source_detail: Source details

Methodological considerations¶

Code examples¶

# Set working directory
setwd("/path/to/module/directory")

# Load required libraries
library(dplyr)
library(tidyr)
library(zoo)
library(stringr)
library(purrr)

# Configure country
COUNTRY_ISO3 <- "KEN"  # Replace with your country code

# Use default analysis level (national + admin2)
ANALYSIS_LEVEL <- "NATIONAL_PLUS_AA2"

# Run Part 1
source("05_module_coverage_estimates_part1.R")

# Use country-specific mortality rates from DHS or other sources
PREGNANCY_LOSS_RATE <- 0.04      # Default: 0.03
TWIN_RATE <- 0.02                # Default: 0.015
STILLBIRTH_RATE <- 0.025         # Default: 0.02
P1_NMR <- 0.045                  # Default: 0.039
P2_PNMR <- 0.030                 # Default: 0.028
INFANT_MORTALITY_RATE <- 0.070   # Default: 0.063
UNDER5_MORTALITY_RATE <- 0.110   # Default: 0.103

# These parameters affect survival-adjusted denominators
source("05_module_coverage_estimates_part1.R")

# Override automatic "best" selection for specific indicators
DENOM_ANC1 <- "danc1_pregnancy"      # Use ANC1-based denominator
DENOM_PENTA3 <- "dwpp_dpt"           # Use WPP population estimate
DENOM_MEASLES1 <- "best"             # Keep automatic selection

# Run Part 2
source("06_module_coverage_estimates_part2.R")

# Part 1: Run national level only (faster)
ANALYSIS_LEVEL <- "NATIONAL_ONLY"
source("05_module_coverage_estimates_part1.R")

# Part 2: Will automatically skip subnational levels
source("06_module_coverage_estimates_part2.R")

# Part 1: Include admin3 level
ANALYSIS_LEVEL <- "NATIONAL_PLUS_AA2_AA3"
source("05_module_coverage_estimates_part1.R")

# Part 2: Will process all available levels
source("06_module_coverage_estimates_part2.R")

# Load coverage outputs
coverage_national <- read.csv("M5_coverage_estimation_national.csv")
coverage_admin2 <- read.csv("M5_coverage_estimation_admin2.csv")

# Filter to specific indicator
penta3_national <- coverage_national %>%
  filter(indicator_common_id == "penta3")

# Compare HMIS-based and survey-projected coverage
coverage_comparison <- penta3_national %>%
  select(year, coverage_cov, coverage_avgsurveyprojection, coverage_original_estimate) %>%
  mutate(
    hmis_survey_gap = coverage_cov - coverage_avgsurveyprojection,
    data_source = case_when(
      !is.na(coverage_original_estimate) ~ "Survey",
      !is.na(coverage_avgsurveyprojection) ~ "Projected",
      TRUE ~ "HMIS only"
    )
  )

# Identify admin2 areas with coverage below threshold
low_coverage_areas <- coverage_admin2 %>%
  filter(indicator_common_id == "penta3", year == max(year)) %>%
  filter(coverage_avgsurveyprojection < 80) %>%
  arrange(coverage_avgsurveyprojection)

Usage notes¶

Value add beyond standard DHIS2 analysis¶

While DHIS2 provides a robust foundation for data collection, storage, and basic visualization, FASTR builds on this foundation with additional capabilities: automatic data quality adjustment before analysis, advanced analytical methods including disruption detection and coverage projection, standardized visualizations using percent-change approaches, improved coverage estimation using survey-derived denominators, faster analytics cycles aligned with country decision-making timelines, and built-in capacity strengthening through reproducible methods.