As part of our Journal Club summaries our JC Chairs (Drs. Lisa Calder and Ian Stiell @EMO_Daddy) have been tasked with explaining Epidemiological concepts so that everyone in our department can analyze the literature and appraise articles on their own. For this Blog post we have all the “Epi Lesson” as they relate to “Prognosis and Miscellaneous Articles”. That completes our 5 part series on Epi Lessons!


Characteristics of a Good Article About Prognosis         

Dr Ian Stiell     May 2105
As described in the iconic JAMA publication, User’s Guide to the Medical Literature, here are some key questions to ask yourself about an observational study of prognosis.  1) Regarding the validity of the results: a) was the sample of patients representative?, b) were the patients sufficiently homogeneous with respect to prognostic risk?, c) was follow-up sufficiently complete?, and d) were outcome criteria objective and unbiased? 2) In interpreting the results: a) how likely are the outcomes over time? and b) how precise are the estimates of likelihood? 3) How can I apply the results to patient care: a) Were the study patients and their management similar to those in your practice?, b) was the follow-up sufficiently long?, and c) can you use the results in the management of patients in your practice?

Impact of Clusters in Analyses

Dr. Christian Vaillancourt
Most statistical analyses assume that all observations (data) are independent from each other. On occasion, groups share similarities (for e.g. patients of a same physician) which makes the contribution of each patient slightly less valuable or different from the others. The amount of similarity within a group is called “intra-class correlation” and can be measured statistically. This can be used to calculate a slightly larger required sample size, or to adjust analyses accordingly.

Performance Bias                                 

Dr. Venkatesh Thiruganasambandamoorthy
Performance bias in this study refers to potential systematic differences in the care provided between groups other than the interventions of interest (point-of-care ultrasound). After enrolment, blinding (or masking) of physicians to the ultrasound results (specifically with regards to cardiac activity) might have reduced any bias introduced by the vigour of resuscitation which in itself affects outcomes. Hence, effective blinding would have ensured that the groups (with and without cardiac activity) received similar amount of attention and treatment.

Prognosis Articles in Emergency Medicine

Dr. Ian Stiell    November 2015
Emergency physicians generally do not counsel patients on long-term outcomes. We are, however, concerned about prognosis in several ways. First, we need to know what the short-term outcomes are likely to be for patients with common serious conditions like heart failure, COPD, etc. Risk stratification tools can help us, particularly with regards to disposition and follow-up. Second, we need to know the evidence that a given treatment may lead to better outcomes in the very common situations where there have been no RCTs or an RCT is unlikely to be conducted. For example, many aspects of CPR, particularly by lay bystanders will be difficult to evaluate by RCTs; in such cases, cohort studies can be useful.

Propensity Score Matching                                         

Dr. Lisa Calder   March 2013
Large population cohort studies are really the only method available for searching for associations between drugs and rare but serious adverse effects. Because there are so many potential confounders, high quality studies will attempt to adjust for these – propensity score matching is one method of confounder adjustment. Propensity score matching involves calculating conditional probabilities of having a certain outcome given the presence of pre-selected covariates (e.g. risk factors for a given outcome) and matching exposed and control subjects with similar propensity scores.

Prospective Cohort Studies

Dr. Jeff Perry
Prospective cohort studies should be as they say “prospective”.  However, there are many variations of these in the medical literature.  Some are prospectively identified cases with retrospective data abstraction.  This may be appropriate, if all data are reliably recorded (e.g. sex, age, CT done), however, if collecting clinical data, these data are much less reliably recorded.  This may then become a medical record review/historical cohort study, with all the limitations and missing data associated with this study design.

Spectrum Bias

Dr. Venkatesh Thiruganasambandamoorthy
For clinicians the probability and the seriousness of underlying condition(s) previously reported dictates how aggressively they investigate to identify the serious condition. Seriousness depends both on the condition (e.g symptomatic sustained ventricular tachycardia is lot more important than one premature ventricular beat with similar QRS morphology) and the treatment if one would treat the condition (the former one might need a defibrillator insertion and latter one nothing). In the PESIT study, two important components likely led to spectrum bias: the authors did not report the characteristics all patients presented the emergency with syncope (which likely is different from other centers) and second the decision to hospitalize patients would not be the same at other centers (or where you practice). If the study population reported in a study is different from your practice population, then physicians cannot apply the results of the study in their practice.

Time-Dependent Variables

Dr. Christian Vaillancourt
Survival analyses model the time it takes before an event (outcome of interest) occurs. This time until event occurrence can be influenced and adjusted for one or more covariates. These covariates are either “static” i.e. they stay the same for the entire time period (for e.g. eye colour, or other covariate measured only once), or can be “time-dependent” and vary during the observation period until event/outcome occurrence (for e.g. taking a medication vs. not)…this implies the value is measured more than once.

Validity Lingo around Measurement Tools

Dr. Christian Vaillancourt
When describing the validity of measurement tools, various terms are commonly used:
Construct validity (which is of main concern), refers to how well a test measures what it claims to measure.
Criterion validity refers to the extent with which a test can predict an outcome now or in the future. Content validity refers to the extent to which a measure represents all facets of a given construct. Face validity is the extent to which a test is subjectively viewed by stakeholders as measuring the construct it claims to measure. 


Abstract Conclusions 

Dr. Lisa Calder     March 2013
The critical reader will be wary of overstated conclusions in a research abstract. Be sure to examine the results presented and ask yourself if the conclusions are truly supported by the methods described and the data presented. The same applies when writing a research abstract –ensure that your conclusions follow directly from the data you describe and are realistic given your chosen methods. 

AGREE-II Tool for Evaluation of Clinical Practice Guidelines  

Dr. Lisa Calder      February 2015
Clinicians frequently use clinical practice guidelines (CPGs) to inform their practice. Often these also form standards of care. It is important to critically appraise CPGs as you would other articles in the literature. One helpful, user-friendly tool is the AGREE-II instrument. Using this tool takes you through the appropriate steps of CPG development and dissemination and allows you to evaluate for important sources of bias such as appropriate stakeholder involvement and external review.

Clinical Practice Guidelines 

Dr. Ian Stiell    November 2015
As defined by the Institute of Medicine, “Clinical practice guidelines are statements that include recommendations intended to optimize patient care that are informed by a systematic review of the evidence and an assessment of the benefits and harms of alternative care options.” The development of high quality CPGs is resource intensive and many guidelines may be of lesser quality. Practitioners must be aware of what to look in assessing quality:
1) Was there a rigorous and transparent process for evaluating the evidence?
2) Were conflicts of interest properly dealt with?
3) Are the levels of evidence and strength of the recommendations clearly presented?
4) Is there a process for regularly updating the guidelines?

Database Research 

Dr. Lisa Calder    January 2013
Large databases provide a wealth of data to help answer questions which may never be answered with an expensive RCT. A key step, however, in database research is evaluating the accuracy of data entry and reassuring the reader that the data are high quality. This can be done by hand verifying a random sample of entries or re-verifying diagnostic codes, for example.

Quality Improvement Papers

Dr. Ian Stiell
Quality improvement interventions are designed to improve quality of care within an institutional context and are often complex, multi-component, and tailored to the local setting. These systematic, data-guided activities are designed to bring about immediate improvements in health care delivery at particular settings. Often QI initiatives set out to implement best practices that have previously been evaluated for effectiveness.

STROBE Statement 

Dr. Lisa Calder     April 2012
There are many important observational studies in emergency medicine and it is helpful to have a structured approach to critically appraise them. Using both the user’s guide to the literature and the STROBE statement can help you determine the validity of these studies

Study Setting and Subjects

Dr. Ian Stiell        September 2012
When reviewing a paper of original research, pay particular attention to the setting and the subjects to ensure that they are typical of your ED practice. Many studies enroll in-patients or outpatients in a specialty clinic setting and this often means that the findings cannot be easily extrapolated to the ED setting.