UMEM Educational Pearls - By Mike Witting

“I’m not going to the hospital, my father died in a hospital.”

In planning a study it’s a good practice to consider what confounding variables you may need to look out for.

Confounding variables are associated with the predictor (independent) and outcome (dependent) variables, but they are not in the causal chain. In the above example, disease is likely the predictor variable, death is the outcome variable, and going to the hospital is a confounder. Of course, this assumes the death was not iatrogenic; then the hospital would be in the causal chain.

Patients may be selected for interventions based on severity of disease, functional status, education level, and other factors, and these may be confounders.

Confounding can be addressed at the design stage, by:

• Specification – excluding patients with the confounder (often not feasible)
• Matching – selecting cases and controls matched by confounding variable levels
• Randomization – randomly select patients for an intervention and hope confounding variables will balance out

It can be addressed in the analysis stage by:

• Stratification – analyzing data in strata defined by confounding variable levels
• Adjustment – mathematically adjusting for the confounding variable (usually by regression)

Show References

Free Biostatistics Software Options

EpiCalc 2000 – available for Windows. Capabilities include sample size calculation, inferential statistics (p values, 95% CI), simple stratified analysis, paired and independent analyses. Right clicking allows you to do many things. (http://www.brixtonhealth.com/epicalc.html). I've had the most experience with this one.

Epi Info – supported by CDC. Available for Windows and for hand-held. (https://www.cdc.gov/epiinfo/index.html)

P/S – Power and Sample Size Calculation – available for Mac and Window. Supported by Vanderbilt biostatistics. (https://biostat.app.vumc.org/wiki/Main/PowerSampleSize)

Show References

Needed for sample size determination

Power – (1-beta), where beta is the risk of a type 2 error – rejecting the accepting the null hypothesis when it is true – this is usually selected to be 0.8 or 0.9.

Significance (alpha), the chance of making a type 1 error – accepting the alternate hypothesis when the null hypothesis is true. This is usually selected to be 0.05.

One-tailed or two-tailed – is the null hypothesis one of no difference (experimental arm not better or worse) or one-sided (experimental arm not better)?

Effect Size. This is the challenging part. This is the size of the difference in outcomes you’re looking for. 

  For continuous outcomes (example – difference in pain scores). You’ll need an estimate for the variation in the scores between presentations, or the standard deviation. You can get this from a literature estimate or a from small local measurement, say of 10 patients or so.

  For a dichotomous outcome (example – percentage of successes), you can usually estimate the percentage in one group and choose the difference you are looking for.

The effect size has a big effect on the sample size. Generally, cutting the effect size in half increases the sample size by fourfold.

Statistical software - next pearl.

Starting a study? Frame your research question in the PICO format:

Patients (consider severity of presentation, setting, demographics)

Intervention (either something you propose or something in use)

Comparison (another group, the same group without intervention, or a benchmark)

Outcome (a measurement)

This classic format has been used to evaluate studies, as in Journal Club (by our esteemed Dr. Wilkerson), as a literature search tool, or by the Cochrane review. 

Starting with a PICO research question can help you narrow your focus and maintain it.

Show References

Considering starting a research project? Apply the FINER criteria:

Feasible

                  Do you have the resources to study this? Enough patients? Support?

Interesting

                  Does it interest you enough to devote your time to it? Does it interest colleagues?

Novel

                  Would it provide new findings, or confirm, refute, or extend prior findings?

Ethical

                  Can you think of a way to ethically study it?

Relevant

                  Consider possible outcomes of your research. Could the study advance care or policy?

Show References