Data quality in routine health information systems: five checks that pay off

Routine health information systems collect enormous volumes of data from facilities every month. Managers use that data to plan, allocate resources and judge performance. If it is unreliable, every decision built on it is weaker, and nobody is told.

Data quality is not an abstract concept. It can be checked, and improving it often needs modest effort. Five checks cover most of the ground.

1. Completeness

Did every facility that should report actually report, and did each report fill in the required fields? Low reporting rates or many blank fields mean figures understate reality and comparisons mislead.

2. Timeliness

Was the report submitted by the deadline? Data that arrives weeks late cannot support timely decisions, such as responding to an outbreak or redistributing stock.

3. Internal consistency

Do the numbers make sense together? Examples include more women receiving a second antenatal visit than a first, vaccine doses given exceeding doses issued, or sudden spikes that no event explains. Simple automated validation rules in the system catch many such errors at entry.

4. Accuracy

Do reported figures match source records? Periodically recounting a sample from facility registers and comparing with what was reported, often called data verification, gives a measure of reporting accuracy and highlights facilities needing support.

5. Use

Is the data actually used? Data used in review meetings is scrutinised and corrected. Data that goes into a system and never comes out tends to deteriorate. Feedback to facilities, dashboards that managers open, and regular performance reviews all raise quality.

Beyond the checks

• Clear definitions. If two clinics count the same thing differently, totals mean little. Written indicator definitions and training prevent drift.
• Reliable denominators. Coverage figures depend on population estimates, which are often outdated or contested. Be explicit about which are used.
• Simple tools. Short, well-designed registers and forms reduce errors at the point of collection.
• Capacity. Data clerks and facility staff need training and mentoring, not just software.

Making it routine

A data quality assessment can reveal where problems concentrate, and a short action plan with named owners turns findings into improvement. Repeat it regularly and track progress.

Better data does not guarantee better decisions, but poor data nearly guarantees worse ones.