GPAtlasMRCGP · Applied Knowledge Test
Statistics is 10% of the AKT, and you can clear it in a day
Evidence and data interpretation is one of the two smallest domains on the paper. It is also the most self-contained. The clinical 80% is the whole of general practice and it takes months. This is a short list of ideas that keep coming back, and once you can do them, you can do them.
Statistics and admin are 20% of the paper between them. There is an admin day page too.
Of the AKT is evidence and data interpretation, under the RCGP blueprint of 80% clinical medicine, 10% evidence and data interpretation and 10% organisation and admin.
The real paper is 160 questions in 160 minutes, and so is each of our three mock exams.
Chapters in AKT Compass. Evidence, research and statistics is one of them in its own right, not a footnote on the end of a clinical chapter.
What the RCGP actually asks for
The College describes this part of the paper as a basic understanding and principles, such as interpreting graphs and charts as well as simple calculations. Read that carefully, because it tells you what to stop worrying about. You are not being asked to derive anything, choose a test for your own data or defend a method. You are being asked to read a chart, run a short calculation, and say what the result means for the patient in front of you.
The arithmetic is division and subtraction. What is really being tested is whether a number becomes a sentence in your head, and whether you notice when a result has been reported in whichever way makes it look biggest. Always confirm the current format and content of the paper on rcgp.org.uk.
What keeps coming back, and how long each part takes
Being honest about this matters more than being encouraging about it. Some of this domain really is a morning's work. Some of it is not, and knowing which is which is how you plan the day.
| Topic | What you are asked to do | How it behaves |
|---|---|---|
| Sensitivity and specificity | Read a two by two table and say what the test does when it is positive and when it is negative. | Quick win. Learn the square once and four of these topics come off it. |
| Positive and negative predictive value | The same square read the other way, plus what happens to the answer when the disease is rare. | Quick win to calculate. Give the prevalence point a few minutes of proper thought. |
| Absolute and relative risk | Turn two event rates into the two ways of reporting the same difference, and notice which one the abstract chose. | Quick win, and the most reliable mark in the domain. |
| Number needed to treat | One division, then one sentence about what it means for a single patient. | Quick win. Learn it directly after absolute risk reduction, because it is the same sum. |
| Incidence and prevalence | Tell them apart, and know what a longer illness does to each. | Quick win. |
| Confidence intervals | Say whether the interval crosses the line of no effect, and what a wide one is telling you about the study. | Quick to state, easy to state very slightly wrong. Worth saying out loud until it is exact. |
| p values and significance | Say what a p value does and does not mean, and why statistically significant is not the same as worth doing. | Needs care. The intuitive version is the wrong one, so this is unlearning, not learning. |
| Study designs | Match a described study to its name, and know what that design can and cannot show. | Needs care. This is read and recognised rather than worked out, so it rewards a second pass. |
| Bias and confounding | Name the bias from the description, and say what it did to the result. | Needs care. The names are the work, and they only stick through questions. |
| Screening | Lead time and length time bias, overdiagnosis, and what a programme has to satisfy before it is worth running. | Needs care. The arithmetic is easy and the ideas are slippery. |
| Forest and funnel plots | Read one and say what it shows, including what an asymmetric funnel suggests. | Quick win once you have read a few, and figures do appear on the real paper. |
| Survival curves and the other charts | Read the axes first, then the shape, then the gap between the lines. | Quick win, and the habit of reading the axes first is worth more than any formula. |
What the day actually looks like
Learn the two by two table once, properly
Sensitivity, specificity, positive predictive value and negative predictive value all come off the same square. Draw it the same way every single time and label it before you put any numbers in it. Most lost marks here are a transposed row, not a misunderstanding.
Learn the ways a difference can be reported
Absolute risk reduction, relative risk reduction and number needed to treat are one calculation looked at from three sides. Do them together and you will start noticing, automatically, when an abstract has quoted the relative figure because the absolute one is small.
Fix what a p value and a confidence interval mean
Say each definition out loud until it is exactly right rather than nearly right. These two are the questions people lose after revising them, because nearly right sounds convincing in your own head.
Read the designs and the biases as a list
Cohort, case control, randomised trial, cross sectional, systematic review, and the biases that go with each. You are matching a description to a name, so read them in one sitting and then test yourself rather than reading them again.
Read charts until they are boring
Forest plots, funnel plots, survival curves, box plots, scatter plots. Axes first, then the shape. The real paper puts figures in front of you, so practising on real ones is the point.
Then do questions until the arithmetic is automatic
This domain is unusual in that questions genuinely finish it off. The list is short enough that after a few dozen you are recognising the shape of the question before you have finished reading it.
How AKT Compass covers it
Evidence, research and statistics is one of the 20 chapters and it is built like the others, which is the whole point. It is not a summary sheet bolted on at the end.
Its own teaching notes
Written for this paper: what the guidance and the method actually say, the numbers worth knowing, and where candidates slip. They cover describing data and the charts that show it, counting disease and measuring what a treatment does, how good a test is and what screening does with it, significance and choosing a test, study designs and everything that distorts them, and appraising a paper.
Its own questions, in both formats
The real paper uses single best answer and multiple best answer, which gives six options with two right, and so do we. Statistics questions sit in the question bank under their own chapter and they appear in the mock exams at the proportion the blueprint gives them.
Its own charts to read
Real data figures, at the proportion the real exam carries them, because a statistics question you can answer without looking at the chart is not the question you will be asked.
Citations you can open
Each teaching note carries a citation you can check for yourself and the date we last reviewed it. We check every citation against the body that issued it, not against somebody's summary of it.
Nothing stands alone
Every question is anchored to a teaching point, so a question you get wrong leads straight to the teaching it came from, and nothing in the teaching goes untested.
Three mock exams
Each is 160 questions in 160 minutes, the real paper's own format, so you can see what your statistics marks look like under time pressure rather than at a desk on a quiet afternoon.
The AKT tab on the plans page shows what is open to buy today.
Common questions
How much of the AKT is statistics?
Evidence and data interpretation is 10% of the Applied Knowledge Test. The RCGP blueprint is 80% clinical medicine, 10% evidence and data interpretation and 10% organisation and admin, so statistics and admin are 20% of the marks between them.
What does the RCGP actually ask for?
The RCGP describes this part of the paper as a basic understanding and principles, such as interpreting graphs and charts as well as simple calculations. It is not a statistics course and you are not asked to derive anything. You are asked to read a chart, run a short calculation and say what the result means for a patient.
Can you really cover AKT statistics in one day?
The calculations and the definitions, yes. They are a short list that keeps coming back, and once you can do them you can do them. The parts that reward more than a day are study designs, the named biases and screening theory, because those are read and recognised rather than worked out. A day gets you most of the domain, and the rest is questions.
Do I need to be good at maths?
No. The arithmetic is division and subtraction. What the paper is really testing is whether you can turn a number into a sentence about a patient, and whether you notice when a result has been reported in the way that makes it look biggest.
How does AKT Compass cover statistics?
It has its own chapter, one of the 20, with teaching notes written for this paper, its own questions in both AKT item formats, and its own data charts to read. Every question is anchored to a teaching point, so a question you get wrong leads straight to the teaching it came from.
Being straight with you
- GPAtlas is not affiliated with or endorsed by the RCGP, and this is not the exam. We describe the blueprint and the format as the College publishes them, and you should confirm both on rcgp.org.uk before you book.
- Clearing statistics does not pass the AKT. It is 10% of the paper. The clinical 80% is still the work, and nobody can promise you a pass.
- What we are claiming is narrower than that, and it is true: this domain is small, it is self-contained, and it is the part of your preparation where a single focused day changes your score the most.
- If you want the other quick one, admin is the same size and mostly recall. That is the AKT admin day page.
Give it a day
The chapter, the questions and the charts are all in AKT Compass, and they are joined up, so a question you miss takes you to the teaching that explains it.