Statistical ratio · BEGINNER LESSON
o/e score
An O/E score, short for observed-to-expected score, is a ratio that compares how often something was observed with how often it was expected to occur. In the sources, it is used both for event detection in safety data and for measuring how depleted a gene is for loss-of-function variants.
What it is and what is measured
An O/E score measures a comparison between two quantities: the observed amount and the expected amount. In vaccine safety work, the observed part is the count of a reported adverse event after vaccination, and the expected part is the number predicted from background rates, exposure data, and the risk window, which is the time period when an event is considered likely to happen. In hospital quality work, it compares the number of patients who received guideline-adherent care with the number predicted from patient mix. In genetics, the O/E score compares the number of observed rare loss-of-function variants in a gene with the number expected from population data; lower values mean fewer disruptive variants than expected.
[1] The use of observed-to-expected analyses as a signal detection tool ... [2] Observed-to-expected ratio for adherence to treatment guidelines as ... [3] Overview of specifications to the ACMG/AMP variant interpretation ...
How it works
The basic idea is simple: divide what was seen by what should have been seen under a reference model. If the ratio is above 1, more was observed than expected; if it is below 1, less was observed than expected. For genes, the observed/expected, or oe, score is used in population databases such as gnomAD to measure loss-of-function intolerance, meaning how strongly natural selection seems to remove damaging variants from that gene. A very low oe value suggests strong intolerance, while higher values suggest less depletion of such variants.
[2] Observed-to-expected ratio for adherence to treatment guidelines as ... [3] Overview of specifications to the ACMG/AMP variant interpretation ...
Why people use it
O/E scores help turn large datasets into a single comparison. In vaccine safety, they can refine a signal by checking whether a reported event occurs more often than background rates predict. In hospital studies, they can adjust for differences in patient mix. In genetics, they help researchers judge whether loss of function is a plausible disease mechanism for a gene, especially in autosomal dominant conditions with haploinsufficiency, where one working copy may not be enough.
[1] The use of observed-to-expected analyses as a signal detection tool ... [2] Observed-to-expected ratio for adherence to treatment guidelines as ... [3] Overview of specifications to the ACMG/AMP variant interpretation ...
Limitations
An O/E score is only as reliable as the data and assumptions behind it. For safety studies, the numerator and denominator must be estimated carefully, and composite event groups can hide important subtypes. Even precise-looking ratios can still carry uncertainty, so sensitivity analyses are used to test how results change when assumptions change. In hospital studies, unmeasured factors such as comorbidities or provider specialty can affect the ratio. In genetics, the oe score is most informative only for some gene-disease relationships, and it does not by itself prove that a variant or gene causes disease.
[1] The use of observed-to-expected analyses as a signal detection tool ... [2] Observed-to-expected ratio for adherence to treatment guidelines as ... [3] Overview of specifications to the ACMG/AMP variant interpretation ...
Simple interpretation diagram
Observed count -> divided by -> Expected count -> gives -> O/E score. If the score is greater than 1, the observed amount is higher than expected. If the score is less than 1, the observed amount is lower than expected. In gene studies, a lower oe score means stronger depletion of loss-of-function variants.
[1] The use of observed-to-expected analyses as a signal detection tool ... [2] Observed-to-expected ratio for adherence to treatment guidelines as ... [3] Overview of specifications to the ACMG/AMP variant interpretation ...
SEE THE IDEA
Observed-to-expected score in practice
What was actually counted in the data.
What a background model or reference predicts.
Observed divided by expected.
Above 1 means more than expected; below 1 means less than expected.
Observed and expected values are compared to produce a ratio that is then interpreted against 1.
The same ratio idea can be used in different settings, but the meaning of the observed and expected parts depends on the study.
TRY IT YOURSELF
Test your understanding
Choose an answer, then check the explanation.
PUT IT TOGETHER
Match each concept to its meaning
Drag a label onto its description, or select a label and then select a description. This also works with a keyboard.
4 pairs to match
Sources and further reading
This general lesson was generated from public source material. It explains a concept; it does not interpret a private paper or provide medical advice.