Evidence Based Medicine: MRCEM Primary questions with answers

Evidence Based Medicine carries 10 of the 180 questions in MRCEM Primary. The topic covers types of data, averages, standard deviation, distributions, hypothesis testing, type I and type II errors, power, statistical tests, correlation and survival analysis. It also covers risk and odds, diagnostic test characteristics, forest plots, study designs, systematic reviews and bias. Emergency physicians use this to judge new research and interpret test results.

5 free questions below, each with the answer and the full explanation · 128 in the full bank · Last updated after the 27 October 2026 MRCEM Primary sitting.

1. Question 1 · Statistical Tests

   Which of these is an example of a parametric test:

   1. AMann-Whitney U-test
   2. BKruskal-Wallis test
   3. CWilcoxon matched pairs test
   4. DAnalysis of variance (ANOVA) test
   5. EFriedman's test
   Show answer and explanation

   Answer: D. ANOVA is a parametric test.

   • A: Non-parametric.
   • B: Non-parametric.
   • C: Non-parametric.
   • E: Non-parametric.

   Full explanation: Statistical Tests

   Significance tests are mathematical functions which compute an observed value of a test statistic from experimental data. The p value required to determine significance is derived from this test statistic.

   The type (categorical or quantitative) and distribution (normal, non-normal) of data, the number of samples, and whether the data is paired or unpaired determines which significance test should be used. In practice, computer programs are used to do these calculations.

   Unpaired data refers to data from two groups which have different members. Paired data refers to data from the same individuals taken at different time points.

   Parametric and Non-Parametric Tests

   Parametric tests are used to compare samples of normally distributed data. Non-parametric testing is used to compare samples of non-normally distributed data.

   Parametric tests (normal data)	Non-parametric tests (non-normal data)
   Unpaired student t-test: used to compare the means of two independent groups of values	Mann Whitney U-test: used to compare the medians of two unpaired groups of values
   Paired student t-test: used to compare the means of values obtained from a single group at two different points in time	Wilcoxon matched pairs: used to compare the medians of two paired groups of values
   One-way Analysis of variance (ANOVA): used to compare the means of more than two unpaired groups of values	Kruskal-Wallis test: used to compare the medians of more than two unpaired groups of values
   Repeated-measures ANOVA: used to compare the means of more than two paired groups of values	Friedman's test: used to compare the medians of more than two paired groups of values

   Tests for Categorical Data

   The chi-squared test is used to compare the proportions of two or more unpaired groups with a particular attribute e.g. the proportion of people suffering a stroke in two groups receiving a statin or no statin to see if they are statistically different. An odds ratio of 1, or whose 95% CI includes the value of 1, implies that there is no difference between the two groups.

   For small-sized samples (fewer than five observations in any variable), Fisher's exact test can be used.

   The McNemar's test is used to compare proportions of two or more paired groups of values.

2. Question 2 · Systematic Review

   Which of these is NOT a benefit of systematic review:

   1. ASystematic review is quicker and less costly to perform than a new study.
   2. BSystematic review improves the reliability and accuracy of recommendations.
   3. CSystematic review has greater power to detect effects of interest.
   4. DSystematic review allows results to be generalised to a wider population and broader setting.
   5. ESystematic review is not affected by bias.
   Show answer and explanation

   Answer: E. Systematic review is not immune to bias; it is minimised by including unpublished and foreign-language studies.

   • A: True: quicker and cheaper than a new study.
   • B: True: improves reliability of recommendations.
   • C: True: pooling increases power.
   • D: True: allows wider generalisation.

   Full explanation: Systematic Review

   A systematic review (SR) is a formalised and stringent process of searching and identifying, critically evaluating, and selecting for inclusion based on predefined quality criteria, all relevant studies (both published and unpublished) surrounding a particular clinical question. Systematic reviews are the gold-standard source of research evidence in the hierarchy of research evidence.

   Systematic review achieves:

   • refinement and reduction - large quantities of information are refined and reduced to a manageable size
   • efficiency - SR is quicker and less costly to perform than a new study, it can prevent others from embarking on an unnecessary study and can shorten time between medical developments and their implementation
   • generalisability and consistency - results can often be generalised to a wider population in a broader setting than would be possible from a single study, consistencies in the results from different studies can be assessed and inconsistencies determined
   • reliability - SR aims to reduce errors and so tends to improve the reliability and accuracy of recommendations
   • power and precision (meta-analysis has greater power to detect effects of interest and provides more precise estimates than a single study)

   Meta-Analysis

   A meta-analysis the quantitative assessment of a systematic review. It involves combining the results of independent studies with common features to produce a more powerful overall estimate of effect. A meta-analysis can provide conclusive evidence for or against an intervention, even when individual studies are inconclusive.

   In meta-analysis of RCTs, the outcome data from each trial are usually illustrated as a series of odd ratios with their confidence intervals on a graph known as a forest plot.

3. Question 3 · Statistical Tests

   Which of these is an example of a parametric test:

   1. AMann-Whitney U test
   2. BStudent t-test
   3. CFriedman's test
   4. DKruskal-Wallis test
   5. EWilcoxon matched pairs test
   Show answer and explanation

   Answer: B. The Student t-test is a parametric test.

   • A: Non-parametric.
   • C: Non-parametric.
   • D: Non-parametric.
   • E: Non-parametric.

   Full explanation: Statistical Tests

   Significance tests are mathematical functions which compute an observed value of a test statistic from experimental data. The p value required to determine significance is derived from this test statistic.

   The type (categorical or quantitative) and distribution (normal, non-normal) of data, the number of samples, and whether the data is paired or unpaired determines which significance test should be used. In practice, computer programs are used to do these calculations.

   Unpaired data refers to data from two groups which have different members. Paired data refers to data from the same individuals taken at different time points.

   Parametric and Non-Parametric Tests

   Parametric tests are used to compare samples of normally distributed data. Non-parametric testing is used to compare samples of non-normally distributed data.

   Parametric tests (normal data)	Non-parametric tests (non-normal data)
   Unpaired student t-test: used to compare the means of two independent groups of values	Mann Whitney U-test: used to compare the medians of two unpaired groups of values
   Paired student t-test: used to compare the means of values obtained from a single group at two different points in time	Wilcoxon matched pairs: used to compare the medians of two paired groups of values
   One-way Analysis of variance (ANOVA): used to compare the means of more than two unpaired groups of values	Kruskal-Wallis test: used to compare the medians of more than two unpaired groups of values
   Repeated-measures ANOVA: used to compare the means of more than two paired groups of values	Friedman's test: used to compare the medians of more than two paired groups of values

   Tests for Categorical Data

   The chi-squared test is used to compare the proportions of two or more unpaired groups with a particular attribute e.g. the proportion of people suffering a stroke in two groups receiving a statin or no statin to see if they are statistically different. An odds ratio of 1, or whose 95% CI includes the value of 1, implies that there is no difference between the two groups.

   For small-sized samples (fewer than five observations in any variable), Fisher's exact test can be used.

   The McNemar's test is used to compare proportions of two or more paired groups of values.

4. Question 4 · Statistical Tests

   Which of these is an example of a non-parametric test:

   1. AMann-Whitney U test
   2. BWilcoxon matched pairs test
   3. CKruskal-Wallis test
   4. DFriedman's test
   5. EAll of the above
   Show answer and explanation

   Answer: E. All listed tests are non-parametric.

   • A: Non-parametric, but not the only one.
   • B: Non-parametric, but not the only one.
   • C: Non-parametric, but not the only one.
   • D: Non-parametric, but not the only one.

   Full explanation: Statistical Tests

   Significance tests are mathematical functions which compute an observed value of a test statistic from experimental data. The p value required to determine significance is derived from this test statistic.

   The type (categorical or quantitative) and distribution (normal, non-normal) of data, the number of samples, and whether the data is paired or unpaired determines which significance test should be used. In practice, computer programs are used to do these calculations.

   Unpaired data refers to data from two groups which have different members. Paired data refers to data from the same individuals taken at different time points.

   Parametric and Non-Parametric Tests

   Parametric tests are used to compare samples of normally distributed data. Non-parametric testing is used to compare samples of non-normally distributed data.

   Parametric tests (normal data)	Non-parametric tests (non-normal data)
   Unpaired student t-test: used to compare the means of two independent groups of values	Mann Whitney U-test: used to compare the medians of two unpaired groups of values
   Paired student t-test: used to compare the means of values obtained from a single group at two different points in time	Wilcoxon matched pairs: used to compare the medians of two paired groups of values
   One-way Analysis of variance (ANOVA): used to compare the means of more than two unpaired groups of values	Kruskal-Wallis test: used to compare the medians of more than two unpaired groups of values
   Repeated-measures ANOVA: used to compare the means of more than two paired groups of values	Friedman's test: used to compare the medians of more than two paired groups of values

   Tests for Categorical Data

   The chi-squared test is used to compare the proportions of two or more unpaired groups with a particular attribute e.g. the proportion of people suffering a stroke in two groups receiving a statin or no statin to see if they are statistically different. An odds ratio of 1, or whose 95% CI includes the value of 1, implies that there is no difference between the two groups.

   For small-sized samples (fewer than five observations in any variable), Fisher's exact test can be used.

   The McNemar's test is used to compare proportions of two or more paired groups of values.

5. Question 5 · Statistical Tests

   Which of these statistical tests is most appropriate for comparing means between two independent parametric variables:

   1. AMann Whitney U-test
   2. BKruskal-Wallis test
   3. CWilcoxon matched pairs test
   4. DUnpaired student t-test
   5. EChi-squared test
   Show answer and explanation

   Answer: D. The unpaired Student t-test compares means of two independent parametric samples.

   • A: Non-parametric equivalent.
   • B: Non-parametric test for more than two groups.
   • C: Non-parametric test for paired samples.
   • E: Compares categorical data, not means.

   Full explanation: Statistical Tests

   Significance tests are mathematical functions which compute an observed value of a test statistic from experimental data. The p value required to determine significance is derived from this test statistic.

   The type (categorical or quantitative) and distribution (normal, non-normal) of data, the number of samples, and whether the data is paired or unpaired determines which significance test should be used. In practice, computer programs are used to do these calculations.

   Unpaired data refers to data from two groups which have different members. Paired data refers to data from the same individuals taken at different time points.

   Parametric and Non-Parametric Tests

   Parametric tests are used to compare samples of normally distributed data. Non-parametric testing is used to compare samples of non-normally distributed data.

   Parametric tests (normal data)	Non-parametric tests (non-normal data)
   Unpaired student t-test: used to compare the means of two independent groups of values	Mann Whitney U-test: used to compare the medians of two unpaired groups of values
   Paired student t-test: used to compare the means of values obtained from a single group at two different points in time	Wilcoxon matched pairs: used to compare the medians of two paired groups of values
   One-way Analysis of variance (ANOVA): used to compare the means of more than two unpaired groups of values	Kruskal-Wallis test: used to compare the medians of more than two unpaired groups of values
   Repeated-measures ANOVA: used to compare the means of more than two paired groups of values	Friedman's test: used to compare the medians of more than two paired groups of values

   Tests for Categorical Data

   The chi-squared test is used to compare the proportions of two or more unpaired groups with a particular attribute e.g. the proportion of people suffering a stroke in two groups receiving a statin or no statin to see if they are statistically different. An odds ratio of 1, or whose 95% CI includes the value of 1, implies that there is no difference between the two groups.

   For small-sized samples (fewer than five observations in any variable), Fisher's exact test can be used.

   The McNemar's test is used to compare proportions of two or more paired groups of values.