Reading the evidence
Research guides: reading and applying evidence
Practical guides to appraising research, interpreting results and deciding whether evidence answers your question.
A research paper can answer a narrow question well and still be the wrong paper for your decision. A trial may compare treatments you cannot obtain. A diagnostic study may recruit patients whose symptoms are quite different from those seen in your setting. Before getting drawn into a p-value or an impressive chart, write down what you need to know.
These guides help you work through that reading task. You can start with a practical checklist, follow a particular study design, or explore what happens when research becomes a recommendation. The examples use invented data so that you can follow the reasoning without mistaking the numbers for findings about a real treatment or test.
Choose the question before the checklist
Ask whether you are trying to estimate a treatment effect, interpret a test, predict an outcome or understand someone's experience. Each question calls for different evidence. Random allocation can help a trial compare interventions fairly. It does not tell you whether an interview study explored its participants' experiences carefully.
Find the relevant guide below and use its questions as prompts for reading. A reporting checklist tells you what authors should describe. A risk-of-bias assessment asks how a study's methods could distort a particular result. A well-presented report can still have serious limitations, while missing information can leave you unable to judge what happened.
Put a result back into its study
Keep the population, comparison, outcome and follow-up period beside the estimate. A relative reduction means little without the original event risk. A confidence interval describes uncertainty under a statistical model; it cannot remove confounding or repair incomplete follow-up. Different definitions of an outcome can also change what a comparison means.
Several guides include a worked example. Check the denominators and units as you read it. When you try a related calculator, enter the same quantities and compare the result with the explanation. The purpose is to understand the calculation and its assumptions before using it with another dataset.
Decide what the evidence can support
One study is rarely the whole answer. A systematic review can show how findings compare across studies, but its conclusions depend on the evidence it found and the way it handled differences between those studies. A recommendation adds further judgements about benefits, harms, resources and the people affected.
The guides on certainty, guidelines and applying evidence explain those additional steps. They also separate a useful estimate from a decision that fits an individual situation. Clinical examples teach research interpretation. They do not determine a diagnosis or select treatment for a particular person.
Keep a short reading record
Finish by recording the question, the best estimate, the main uncertainty and the next piece of information you need. That might be an outcome definition in a protocol, a more relevant population, a longer follow-up period or an independent validation study. Naming the gap is more useful than giving an entire paper an unexplained score.
If you are new to the subject, begin with the critical appraisal checklist and the evidence-based practice workflow. Then choose the guide that matches your study. The statistics glossary and calculators provide support when the numerical details need a closer look.