Investigating Health Data Discrepancies

4.4.1 What is a data discrepancy?

A discrepancy occurs when two or more records or sources provide different values that are expected to be consistent or comparable. A facility register may show 120 recorded service contacts while the monthly summary reports 110. The difference is 10. Calculating the difference does not explain why it happened. We need to investigate.

4.4.2 Common causes of discrepancies

1. Data-entry errors. A value may have been entered incorrectly. For example, 150 may have been entered as 510.
2. Missing records. Some services may have been recorded in a register but omitted from the summary.
3. Duplicate records. A service may have been counted more than once.
4. Different reporting periods. One source may include the final day of a month while another does not.
5. Different definitions. One report may count visits, while another counts unique patients.
6. Differences in coverage. One source may include outreach services that another excludes.
7. Changes after submission. A report may have been corrected after the initial version was submitted.
8. Uncertain denominators. Population estimates may not accurately reflect the eligible population. This can affect calculated coverage indicators even when the numerator is correct.

4.4.3 A systematic investigation process

Suppose a fictional clinic register records 150 services, but the monthly report contains 140.

1. Confirm the measure. Are both figures intended to count the same service?
2. Confirm the period. Do both cover the same dates?
3. Check the source records, subject to appropriate access permissions.
4. Recalculate the total. Check the aggregation process and formulas.
5. Check for omissions or duplication.
6. Document the discrepancy, the investigation steps, the evidence and any correction.
7. If the evidence does not explain the difference, state that the discrepancy remains unresolved.

4.4.4 Worked example

After checking the records, the team discovers that ten entries were omitted during the preparation of the monthly summary. The team confirms the omission and follows the authorised correction procedure. The corrected summary now agrees with the verified source count. In a real setting, corrections should follow the organisation's reporting and audit procedures.