← Medical education

Understanding study proportions and missing data

A worked example distinguishes enrollment, measurement and observed findings.

Start with a worked denominator

Teaching example, not a result from the library: a study enrolls 20 people, tests 12 and finds 6 abnormal results. Among those tested, the proportion is 6/12 = 50%.

The other 8 cannot be assumed normal. Dividing by 20 answers a different question and does not establish the abnormality rate in everyone enrolled.

Why can one person appear more than once?

One participant may contribute several variants, two gonadal specimens or repeated hormone measurements, and may appear in later publications.

Variant rows, specimens, case records and independent participants therefore cannot be added as if they were the same unit.

The same marker can be measured under different conditions

Hormone results depend on age, sampling time, medication, assay and units. A basal value and a post-stimulation value are not directly interchangeable.

Check these conditions before interpreting an apparent difference between studies.

Sources: [2] Endocrine Society · CAH guideline resources (measurement context)

How the evidence matrix helps

The matrix places populations, measured denominators, results, units and source locations side by side. A blank is a prompt to check the paper's reporting.

Library entry counts describe the collection. Reported sample sizes describe the populations in the original studies.

With 20 enrolled, 12 tested and 6 abnormal, are the other 14 normal?

No. Only 6 tested people are known to have normal results; 8 have no result in this example.

Read the research: source studies in the library

Sample sizes are as reported in the source studies. Follow each publication link to verify the source.

Open the evidence matrix Version des données v1.4.0

Original studies and references