Skip to Content
DocumentationAnalytical FrameworkData Quality Assessment

Data quality assessment

Overview

High-quality health facility data form the foundation of reliable coverage estimates. This module systematically evaluates health facility data by considering the standard data quality metrics, which include:

  • Reporting completeness
  • Outlier detection
  • Internal consistency
  • Data missingness
  • Ratio calculations
  • Overall quality score

DQA metrics in the CD2030 analytical approach

Reporting completeness

For further reading on the scientific basis and methods used to assess reporting completeness, go to Reporting completeness: rationale.

MetricNumeratorDenominatorInterpretation
1aNumber of monthly facility reports receivedTotal expected reports (12 × number of facilities)≥90% is good; affects trend interpretation
1bDistricts with ≥90% reporting completenessTotal districtsIdentifies low-performing districts
1cDistricts with no missing values across all formsTotal districtsIdentifies the distribution of missing values across districts

Statistics for 1a and 1b are based on the mean of four reporting forms (ANC, delivery, immunization, OPD).

Sub-national reporting rate heat mapSub-national reporting rate heat map

National reporting rate and sub-national units with low reportingNational reporting rate and sub-national units with low reporting

Outlier detection

An extreme outlier is defined as a monthly value that is 5 times the median absolute deviation (MAD) from the monthly median value for a particular year. For further reading on the scientific basis and methods used to assess and adjust for extreme outliers, go to Extreme outliers: rationale.

MetricNumeratorDenominatorInterpretation
2aMonthly values not classified as outliersTotal monthly values≥99% expected
2bDistricts with no outliersTotal districts≥90% expected

Outliers are identified statistically. Always assess whether deviations reflect real-world events (e.g., campaigns) or data errors.

Outlier detection heat map by region and yearOutlier detection heat map by region and year

Internal consistency

  • Internal consistency assesses the degree to which related indicators are aligned within the dataset, based on expected epidemiological and service delivery relationships.
  • Within the CD2030 analytical framework, this step complements other DQA metrics by evaluating whether indicators that are biologically or programmatically linked follow plausible patterns.
  • Unlike standalone indicator checks, internal consistency focuses on relationships between indicators, ensuring that reported values reflect realistic service delivery pathways across the continuum of care.

The following metrics are considered when assessing internal consistency:

MetricNumeratorDenominatorInterpretation
3aANC1 reportedPenta1 reportedExpected ratio range
3bPenta1 reportedPenta3 reportedBased on survey data
3cDistricts within expected ratioTotal districtsWider acceptable range
3dDistricts within expected rangeTotal districtsSame as above

Consistency checks between ANC1 and Penta1 by yearConsistency checks between ANC1 and Penta1 by year

Data missingness

  • TBC

Data missingness heat map by indicator and regionData missingness heat map by indicator and region

Ratio calculations

  • TBC

Ratio plots of ANC1/Penta1 and Penta1/Penta3 by yearRatio plots of ANC1/Penta1 and Penta1/Penta3 by year

Overall quality score

MetricNumeratorDenominatorInterpretation
Overall scoreComputed automaticallyComputed automaticallyMean of indicators 1a, 1b, 2a, 2b, 3c, 3d

Overall data quality score table by yearOverall data quality score table by year

DataSuite automatically computes the overall data quality score.

Analysis SetupData Adjustment