MODULE 2 · LESSON 2.2
Structuring a Health Dataset
Estimated time: 20 minutes
2.2.1 What is a dataset?
A dataset is an organised collection of related information.
For example, a dataset might contain monthly reports from five health facilities. Each row represents one facility's report for a particular month. Each column represents a characteristic of that report.
| Facility | Month | ANC visits | Deliveries |
|---|---|---|---|
| Central Clinic | January | 80 | 25 |
| Riverside Clinic | January | 65 | 18 |
| Hilltop Clinic | January | 72 | 21 |
| Central Clinic | February | 90 | 28 |
| Riverside Clinic | February | 70 | 20 |
This is a structured dataset because the information follows a consistent arrangement.
2.2.2 Understand records and variables
Record: a single observation represented by a row. In our example, the first record describes Central Clinic in January.
Variable: a characteristic represented by a column. Examples include Facility, Month, ANC visits, and Deliveries.
Understanding these terms will help you work with Excel, DHIS2, Power BI and other analytical tools.
2.2.3 What makes a dataset difficult to analyse?
| Facility | Month | Patients |
|---|---|---|
| Central Clinic | Jan | 120 |
| central clinic | January | 135 |
| Riverside Clinic | Jan | 95 |
| Riverside | January | 100 |
Do you notice the problem? The same facility appears under different names. January is also written in different ways.
A computer may treat Central Clinic and central clinic as separate text values in some analytical operations. Similarly, Riverside Clinic and Riverside may be treated as different facilities. This can produce misleading summaries.
Before analysis, we need to standardise the data using an authoritative facility list and a consistent reporting-period format.
Do not automatically merge two facility names just because they look similar. Confirm that they refer to the same facility.
2.2.4 Rules for structuring health data
- Use one header row.
- Give each column a clear, unique heading.
- Put each observation on a separate row.
- Avoid merging cells inside the main dataset.
- Avoid inserting blank rows between records.
- Use consistent date and text formats.
- Store numerical values as numbers, not text.
- Keep notes and explanations outside the main data table.
- Use defined categories and facility identifiers where available.
- Document important definitions and changes.
Why avoid merged cells? Merged cells may make a report look attractive, but they can make sorting, filtering and importing data more difficult. For analysis, a simple rectangular table is usually better.
2.2.5 Practise structuring a dataset
A fictional health officer receives the following information: Central Clinic reported 40 deliveries in April. Riverside Clinic reported 32 deliveries in April. Hilltop Clinic reported 27 deliveries in April.
Organise the information into a table with three columns: Facility, Month, and Deliveries.
Suggested answer
- How should the three reports be arranged?
Lesson 2.2 summary
A good dataset is organised consistently, with one observation per row and clearly defined variables in columns.