MFOM Statistics Revision: P Values and CIs

Understanding statistics is an important part of MFOM revision because occupational medicine questions often require you to interpret research findings rather than simply recall statistical definitions. The key areas to master include p values, confidence intervals, sensitivity, specificity, positive predictive value and negative predictive value, and knowing how these measures affect the interpretation of a study.

What statistics do you need to know for MFOM?

For MFOM revision, you should be comfortable interpreting common statistical measures in an epidemiological or occupational health scenario. This means being able to identify what a statistic tells you, recognise what it does not tell you, and apply it to the clinical or workplace question being asked.

The current FOM MFOM Part 2 SBA blueprint includes research and critical appraisal topics such as common statistical tests, p values, confidence limits, study designs, bias, confounding, association and causation, descriptive statistics, correlation and regression.

A useful way to organise your revision is into four groups:

• Diagnostic test performance: sensitivity and specificity
• Clinical or screening test interpretation: positive predictive value and negative predictive value
• Uncertainty around an estimate: confidence intervals
• Statistical evidence against a null hypothesis: p values

You should also understand related concepts such as prevalence, incidence, relative risk, odds ratios, standardised mortality ratios, correlation, regression and common measures of association.

For an MFOM question, do not stop at knowing the definition. Ask yourself:

1. What is being measured?
2. What population was studied?
3. What is the relevant comparison?
4. How precise is the estimate?
5. Could the finding have occurred by chance?
6. Could bias or confounding explain the finding?
7. Does the statistical finding have clinical or occupational significance?

That final question is particularly important. A statistically significant result is not automatically an important result for the worker, employer or occupational health service.

What are sensitivity and specificity?

Sensitivity describes how well a test identifies people who truly have the condition.

Specificity describes how well a test identifies people who truly do not have the condition.

The easiest way to remember the definitions is to start with the disease status rather than the test result.

Measure	Question it answers	Formula
Sensitivity	Of those who have the condition, how many test positive?	True positives / all with disease
Specificity	Of those who do not have the condition, how many test negative?	True negatives / all without disease
PPV	Of those who test positive, how many have the condition?	True positives / all positive tests
NPV	Of those who test negative, how many do not have the condition?	True negatives / all negative tests

A useful memory aid is:

Sensitivity = disease present → test positive

Specificity = disease absent → test negative

This distinction is commonly tested because candidates can confuse sensitivity and specificity when faced with a long vignette.

How should you interpret a highly sensitive test?

A highly sensitive test has a low proportion of false negative results.

This makes a sensitive test useful when the consequences of missing a case are important. If the test is highly sensitive, a negative result can provide useful evidence against the condition, although the strength of that conclusion also depends on the clinical setting and the test's characteristics.

A common revision mnemonic is SnNout: a highly Snsensitive test, when Negative, helps to outrule disease.

Do not treat this as an absolute rule. The interpretation of any test depends on the context in which it is being used.

How should you interpret a highly specific test?

A highly specific test produces relatively few false positive results.

This makes a highly specific test useful when you want to be confident that a positive result represents genuine disease or exposure-related abnormality.

The corresponding mnemonic is SpPin: a highly Specific test, when Positive, helps to include disease.

Again, the mnemonic is a memory aid rather than a substitute for understanding the underlying statistics.

What is the difference between sensitivity and positive predictive value?

This is one of the most important distinctions for MFOM statistics questions.

Sensitivity starts with people who have the disease.

It asks:

Of everyone who actually has the condition, how many tested positive?

Positive predictive value starts with people who tested positive.

It asks:

Of everyone who tested positive, how many actually have the condition?

This difference means that sensitivity and PPV are not interchangeable.

Consider a workplace screening test for a relatively uncommon condition. Even if the test has excellent sensitivity and specificity, a positive result may not always mean that the worker has the condition. When the underlying prevalence is low, false positive results can represent a substantial proportion of all positive results.

This is why PPV is influenced by prevalence.

Similarly, NPV is influenced by prevalence. When prevalence is low, the proportion of negative tests that are genuinely negative can be high.

For exam questions, therefore, look carefully at whether the question asks about:

• Performance of the test against a reference standard
• The probability that a person with a positive result has the disease
• The probability that a person with a negative result does not have the disease

These are different questions.

How do you calculate sensitivity and specificity from a 2 × 2 table?

You should be able to recognise a basic 2 × 2 diagnostic table.

	Disease present	Disease absent
Test positive	True positive	False positive
Test negative	False negative	True negative

From this table:

Sensitivity = true positives / true positives + false negatives

Specificity = true negatives / true negatives + false positives

PPV = true positives / true positives + false positives

NPV = true negatives / true negatives + false negatives

A common exam trap is to use the wrong denominator.

For sensitivity, the denominator is everyone who has the disease.

For specificity, the denominator is everyone who does not have the disease.

For PPV, the denominator is everyone with a positive test.

For NPV, the denominator is everyone with a negative test.

If an SBA gives you a table, identify the denominator in words before calculating anything. This simple step can prevent many avoidable errors.

What is a confidence interval?

A confidence interval (CI) provides information about the uncertainty or precision surrounding an estimate.

For example, imagine a study reports a relative risk of 1.40 with a 95% confidence interval of 1.10 to 1.80.

The point estimate is 1.40, but the confidence interval shows the range of values compatible with the data under the assumptions of the statistical method used.

A narrower confidence interval generally indicates greater precision than a wider interval.

This is important because two studies can report the same point estimate but have very different confidence intervals.

For example:

Study	Relative risk	95% CI
A	1.50	1.40–1.60
B	1.50	0.80–2.80

The point estimate is identical, but Study A provides a much more precise estimate than Study B.

The wide interval in Study B indicates considerable uncertainty around the point estimate.

How do you interpret a confidence interval around a ratio?

When interpreting a confidence interval, pay attention to the null value.

For measures such as relative risk, odds ratio and hazard ratio, the null value is generally 1.

If a 95% confidence interval for a ratio excludes 1, this is conventionally consistent with statistical significance at the 5% level under the relevant assumptions.

For example:

• Relative risk 1.8, 95% CI 1.3–2.5: the interval excludes 1.
• Relative risk 1.8, 95% CI 0.9–3.2: the interval includes 1.

The second result does not provide conventional evidence of a statistically significant association at the 5% level.

However, do not simply look at whether the interval crosses the null value. Consider the size and clinical relevance of the estimate, as well as study design, bias, confounding and precision.

What if the confidence interval includes the null value?

If the confidence interval for a ratio includes 1, the result is not conventionally statistically significant at the corresponding two-sided significance level when a 95% CI is being used.

For example, an odds ratio of 1.4 with a 95% CI of 0.95–2.05 includes 1.

This means the data are compatible with no association as well as with a positive association.

It does not prove that there is no association.

That distinction is important in exam questions. A non-significant result should not automatically be described as evidence that the exposure has no effect.

What does a confidence interval mean for a difference?

Not every measure uses 1 as its null value.

For measures based on a difference, the null value is usually 0.

For example, suppose a study reports:

Mean difference = 4.2 units, 95% CI 1.1 to 7.3

The interval does not include zero, so the result is conventionally statistically significant at the 5% level under the relevant assumptions.

If the result were:

Mean difference = 4.2 units, 95% CI −0.8 to 9.2

the confidence interval would include zero.

A useful MFOM revision habit is therefore to ask:

What is the measure, and what is its null value?

Measure	Typical null value
Relative risk	1
Odds ratio	1
Hazard ratio	1
Correlation coefficient	0
Mean difference	0
Risk difference	0

What is a p value?

A p value is used to assess how compatible the observed data are with a specified null hypothesis, under the statistical model and assumptions used.

A small p value provides evidence against the null hypothesis.

A commonly used threshold is p < 0.05, although the choice of threshold should not be treated as a substitute for interpretation.

For example:

p = 0.03

means that, assuming the null hypothesis and relevant model assumptions are true, the observed result or a more extreme result would have a probability of 3% under the specified statistical framework.

It does not mean:

• There is a 3% probability that the null hypothesis is true.
• There is a 97% probability that the treatment or exposure has an effect.
• The finding is clinically important.
• The study is free from bias.
• The result will necessarily be reproduced in another population.

These distinctions are classic exam territory.

How do p values and confidence intervals relate?

P values and confidence intervals provide related but different information.

A p value gives information about the compatibility of the data with the null hypothesis. A confidence interval provides information about the estimated effect and its precision.

For example:

Odds ratio 2.0, 95% CI 1.2–3.3, p = 0.008

This tells you more than simply saying that the result was statistically significant.

You can see:

• The estimated association is an odds ratio of 2.0.
• The confidence interval indicates the precision of the estimate.
• The interval does not include 1.
• The p value provides evidence against the null hypothesis.

When revising, practise interpreting all three pieces of information together rather than memorising p-value thresholds in isolation.

How should you approach an MFOM statistics SBA?

Statistics questions can look intimidating because the vignette may contain a large amount of information. In practice, you can often simplify the problem by identifying exactly what the question is asking.

Use this approach:

1. Identify the study question. Is it asking about diagnosis, association, effect size, precision or statistical significance?
2. Identify the statistic. Is it sensitivity, specificity, PPV, NPV, relative risk, odds ratio, mean difference or another measure?
3. Identify the denominator or null value. This is particularly important for diagnostic tests and confidence intervals.
4. Read the confidence interval. Look at its width and whether it crosses the relevant null value.
5. Interpret the p value. Do not turn it into a probability that the hypothesis is true.
6. Consider clinical importance. Statistical significance and clinical significance are not the same.
7. Consider bias and confounding. A statistically significant association can still be misleading if the study has important methodological limitations.

For more detailed revision of study design and bias, see our MFOM Epidemiology Revision Essentials.

What statistics mistakes should you avoid in MFOM revision?

Several statistical errors are particularly easy to make under exam pressure.

Confusing sensitivity with specificity

Remember that sensitivity starts with people who have the condition, whereas specificity starts with people who do not have the condition.

Confusing sensitivity with PPV

Sensitivity asks how well a test detects disease among those who have it. PPV asks how likely disease is among those who tested positive.

Saying that a p value is the probability that the null hypothesis is true

This is incorrect. The p value is calculated under the assumption of the null hypothesis and describes the compatibility of the observed data, or more extreme data, with that hypothesis.

Assuming a non-significant result proves there is no effect

A confidence interval may be wide because the study lacks precision. A non-significant result can therefore be compatible with a clinically important effect as well as little or no effect.

Ignoring the confidence interval

The point estimate alone does not tell you how precise the estimate is.

Treating statistical significance as clinical significance

A very small effect can be statistically significant in a large study. Conversely, an important effect may fail to reach statistical significance in a small or imprecise study.

Forgetting prevalence

Sensitivity and specificity are characteristics of test performance in relation to a reference standard, whereas predictive values depend strongly on the prevalence of the condition in the population being tested.

How can occupational medicine examples help with statistics?

Statistics becomes easier to remember when you connect it to occupational medicine.

Imagine you are evaluating a screening test for an occupational condition. A question might give you the number of workers with and without the condition, together with positive and negative test results.

You could be asked to calculate:

• Sensitivity
• Specificity
• PPV
• NPV

Alternatively, you might be presented with a study investigating an occupational exposure and asked to interpret a relative risk or odds ratio with its confidence interval.

A different question could give you a p value and ask what conclusion can appropriately be drawn.

The occupational context may involve respiratory disease, hearing loss, musculoskeletal disorders, workplace exposures or health surveillance. The statistical principles remain the same.

For other occupational medicine revision topics, our MFOM Occupational Hygiene Revision Essentials covers key occupational hygiene concepts that can appear alongside epidemiology and statistics questions.

How should you practise MFOM statistics questions?

Reading statistical definitions is useful initially, but question practice is where interpretation becomes more automatic.

When you answer a question, do not only record whether you got it right or wrong. Record why you chose your answer.

A useful error log might include:

Error	What to revise
Used the wrong denominator	2 × 2 diagnostic tables
Confused sensitivity and PPV	Diagnostic test terminology
Misread CI	Null values and precision
Misinterpreted p value	Hypothesis testing
Focused only on significance	Clinical versus statistical significance
Missed confounding	Epidemiological study interpretation

After several rounds of questions, patterns usually become clearer. If you repeatedly make the same type of error, target that concept rather than simply completing more random questions.

It is also useful to practise questions both by topic and in a mixed format. Topic-based questions help you learn a concept; mixed questions test whether you can recognise which statistical method is relevant without being told in advance.

Frequently asked questions

Is statistics difficult in the MFOM?

Statistics can initially seem difficult because several measures use similar terminology. The key is to understand what each statistic is actually asking rather than memorising isolated formulas. Repeated SBA practice can make recognition of common statistical concepts much quicker.

Do I need to memorise statistical formulas?

You should understand the basic calculations for measures such as sensitivity, specificity, PPV and NPV. You should also be able to interpret common statistical outputs, including confidence intervals and p values, rather than relying entirely on memorised definitions.

What is the most important difference between sensitivity and specificity?

Sensitivity measures the proportion of people with the condition who test positive. Specificity measures the proportion of people without the condition who test negative.

What does a 95% confidence interval tell me?

It provides information about the uncertainty and precision surrounding an estimated effect. When interpreting it, consider its width and whether it includes the relevant null value, while also considering the clinical importance of the estimate.

Does p < 0.05 mean that the result is clinically important?

No. Statistical significance and clinical significance are different concepts. A statistically significant result may represent a very small effect, while a clinically important effect may be estimated imprecisely and fail to reach conventional statistical significance.

Should I learn statistics separately from epidemiology?

It is useful to revise the underlying statistical concepts separately at first, but you should ultimately practise applying them within epidemiological and occupational medicine scenarios. This helps you recognise whether a question is testing calculation, interpretation, study design, bias or clinical relevance.

Statistics questions reward understanding rather than memorisation. Build confidence by learning the meaning of each measure, then test yourself with increasingly mixed SBA questions using our free sample questions, and when you are ready for further practice, explore our pricing.