HMIS Data Adjustment and Scenario Creation

Handling outliers and incompleteness through multi-scenario analysis

INPUTS

Raw HMIS Data Service volumes by facility, indicator, month
Outlier Flags From Module 1 (DQA)
Completeness Flags From Module 1 (DQA)
↓
1

Data preparation

  • Merge datasets by facility, indicator, and month
  • Identify excluded indicators (e.g., deaths)
  • Check for low-volume indicators requiring filtering
  • Validate facility-indicator combinations
↓
2

Calculate rolling averages

  • Centered 6-month average (3 months before + 3 months after)
  • Forward 6-month average (next 6 months)
  • Backward 6-month average (previous 6 months)
  • Facility-specific historical mean (across all available data)
↓

Apply adjustment hierarchy

For outliers
  • Use centered average
  • Use forward average
  • Use backward average
  • Use same month last year
  • Use historical mean
For missing/incomplete
  • Use centered average
  • Use forward average
  • Use backward average
  • Use historical mean
↓

Create four scenarios

Scenario 1: None Original data (no adjustments)
Scenario 2: Outliers Only Adjust flagged outliers only
Scenario 3: Completeness Only Adjust missing/incomplete data only
Scenario 4: Both Adjustments Apply both outlier and completeness adjustments
↓

Geographic aggregation

Facility Level
Individual facilities
→
Subnational Level
Provinces, districts
→
National Level
Country total
Each geographic level maintains all 4 scenarios
↓

OUTPUTS

  • M2_adjusted_data.csv Facility-level data with all 4 scenarios
  • M2_adjusted_data_admin_area.csv Subnational aggregated data with all 4 scenarios
  • M2_adjusted_data_national.csv National-level aggregated data with all 4 scenarios
  • M2_low_volume_exclusions.csv Metadata on excluded indicators and facilities