Skip to Main Content

EBP Resources


Evidence-Based Medicine is the integration of best research evidence with clinical expertise and patient values.1

The EBP Cycle


1 (Sackett DL, Straus SE, Richardson WS, et al. Evidence-based medicine: how to practice and teach EBM. 2nd ed. Edinburgh: Churchill Livingstone, 2000.)


Overviews & Introductions

The 6s Pyramid

The 6s Evidence Hierarchy demonstrates the level of evidence within the different types of publications. As information travels up the pyramid, it becomes more synthesized and the evidence is stronger.  The higher levels of the pyramid include information that has been disseminated by several experts, whereas the lower levels of the pyramid have yet to go through an extensive review.

HLWiki Canada. (2016). Evidence-based Medicine - history. Retrieved from


Systematic Reviews


Point of Care


Practice Guidelines

Asking an Answerable Question


Using PICO TO Structure a Clinical Question

The PICO model offers one way to construct your question:

PICO Model for Formulating a Research Question
P = Patient or population Identify the patient groups and medical problem central to your question
I = Intervention Which intervention is being studied?
C = Comparative Intervention Which comparison interventions are being considered (if any)?
O = Outcome What would you like to measure or improve? e.g. reduced mortality, improved quality of life

Let's look at an example:

Situation: One of your area's goals is to reduce occurrence and improve treatment of bed sores. Dynamic bed support surfaces are being discussed.

This can be written into PICO categories:

P = Patients with pressure ulcers

I = Dynamic bed support surfaces

C = Static bed support surfaces

O = fewer pressure ulcers

From this we can state the answerable research question as: "For patients with pressure ulcers do dynamic bed support surfaces, when compared with static bed support surfaces, result in fewer pressure ulcers?"


What Study Designs Answer Which Questions?

Appraising Therapy Studies




patient Follow-up

  • Outcome data has been collected and analyzed for every patient who entered the trial
  • losses to follow up should be less than 20%, and evenly distributed between therapy and control groups
  • Are the reasons for patient drop-outs or loss of data explained 



  • Recruited patients should represent the target population; if possible could accomplish this by selecting them at random from the target population
  • The assignment of patients to treatment and control groups should be randomized, and the randomization method should be proveded;
    • method should actually result in randomness - (random number generator or coin flip, yes; alternating group A and group B, no)
  • The assignment of patients to treatment and control groups should be hidden from the patient, researchers, health care workers etc., and the method of hiding the allocation should be provided
    • method of randomization should not be able to be manipulated by anyone - if the randomization is by sealed envelope the envelopes have to be completely opaque; if the randomization is by computer program or coin flip, ideally it should be done by a third person who only knows the patient as a number and will not do "do-overs", etc.


Intention to treat

  • Patients should be analyzed in the groups to which they were randomized. 
    • For instance someone who was put in the treatment group who stopped taking the treatment should be analyzed in the treatment group, and not have her data moved into the control group
  • Were all randomized patient data analyzed?  If not was a sensitivity or "worst case scenario" analysis done?


Similar baseline characteristics

  • Were the groups similar at the start of the trial?
    • equally healthy or sick; similar representation levels of sexes, ages, races, etc.; similar representation of different economic groups; etc.
  • If important differences appear to exist, consider whether the difference is likely to favour one group over the other



  • Were patients, health care workers, raters/assessors, data analysts, and other study personnel  "blind" to treatment?
  • If blinding was not possible, how did the researchers try to mitigate potential bias?
    • blinded raters
    • objective outcome measures


Equal treatment

  • Were the patients in the two groups treated as equally as possible with the exception of the intervention being tested?
    • Think of things like: time spent with patient by health care workers, comfort of setting, time spent waiting, number of appointments, number of tests, etc.


  • How large was the treatment effect?
  • How precise was the treatment effect? 

ARR Absolute Risk Reduction
The difference between the risk of an event without the treatment and the risk of an event with the treatment.  

CER Control Event Rate
The risk of an event happening in the control / non-treatment group.

EER Experimental Event Rate
The risk of an event happening in the experimental / treatment group.

NNT Number Needed to Treat
The number of patients that would need to be treated in order to prevent one event.

RR Relative Risk  /  Risk Ratio
The ratio of the difference between the risk of an event without the treatment and the risk of an event with the treatment: 50% greater risk, 5 times the risk, etc. 

RRR Relative Risk Reduction
The ratio of the number fewer events in the experimental / treatment group compared to the control group


Appraising Diagnosis Studies


  • Did the clinicians face diagnostic uncertainty?
  • Was there a blinded comparison with an independent gold standard?
  • Was the patient sample representative of the patients to whom the test would be applied in clinical practice?
  • Was the gold standard test applied regardless of what the results of the experimental test were? (You want the answer to be "yes")
  • Did the person performing the gold standard test know the results of the experimental test or vice-versa? (You want the answer to be "no")
  • Were the methods for performing the test described well enough to reproduce them?


  • What is the Sensitivity of the test?
  • What is the Specificity of the test?
  • What is the Positive Predictive Value of the Test?
  • What is the Negative Predictive Value of the Test?

Sensitivity (Sn)
The percentage or proportion of people that have the condition AND test positive with the experimental test. 

Specificity (Sp)
The percentage or proportion of people that DO NOT have the condition AND DO NOT test positive with the experimental test. 

Positive Predictive Value (PPV)
The percentage or proportion of people that test positive with the experimental test who have the condition

Negative Predictor Value (NPV)
The percentage or proportion of people that do not test positive and do not have the condition.

Appraising Systematic Reviews


  • Did they establish the research question before taking any other steps?
  • Do they say where they looked for studies and how they looked for studies?
  • Is it unlikely that they missed any important relevant studies? Did they:
    • search the major databases
    • search for relevant "grey literature" (i.e. not published in a journal and indexed in a database.  Can include government reports, etc.)
    • search the references in relevant studies
    • contact experts / researchers - especially for unpublished studies
    • include subject headings and keywords in their strategies
    • look for papers in any language
  • Do the criteria used to select articles make sense, and were they defined before they started selecting studies? Should include:
    • patient characteristics
    • intervention(s) / exposure(s) of interest
    • outcome(s) of interest
    • often includes the type of study design
  • Do they make a list of the studies that were included, and a list of the studies that were excluded available?  Depending on who published the systematic review this might not be with the text of the review itself.
  • Were there more than one person doing study selection AND more than one person doing data extraction?  Was there a procedure for coming to a consensus?
  • Were the results similar (homogenous) between the different studies?
    • if there was variation (heterogeneity), authors may estimate whether the difference is significant with a chi-square test
    • if there was variation the authors should talk about why that might be the case
    • if there was variation the authors should use a "random effects model" or consider the appropriateness of pooling the results
  • Did they assess the likelihood of publication bias affecting the available data?  In other words, did they look to see if it was likely that only half of the studies with negative results were published compared to all of the studies with positive results? 
  • When they formed their conclusions did they take the rigour and quality of the studies into account?  (In other words, did they give higher quality studies more weight than poorer quality studies.)
  • Are potential conflicts of interest for both the systematic review and for the included studies stated?