research
Chiropractic (General)

Research: Know What You're Talking About

Dana Lawrence, DC, M. Med. Ed., MA; Christine Goertz, DC, PhD

Have you ever seen a patient in your office with multiple serious health problems you weren't sure exactly how to address? Or had someone walk into your office with a page they printed off the Internet, asking you if they should do as the article suggests and take glucosamine for knee osteoarthritis or try the DASH diet to lower their blood pressure? Or what about the patient who just doesn't respond to your care in the way you had hoped / expected? These are scenarios that occur every single day in the chiropractic office.

Evidence-based clinical practice (EBCP) often can provide the tools you need to effectively meet the needs of your patients in these instances. Let's get you started with a glossary of terms that are commonly used when discussing EBCP, defined using David Sackett's words: "the conscientious, explicit and judicious use of current best evidence in making decisions about the care of the individual patient." We have shied away from providing mathematical formulae, by and large, to focus on conceptual meanings of these terms. (You easily find information on how to calculate the terms; none is hard to do.) Let's take a look at a few of the common terms from different kinds of papers.

Papers That Examine Diagnostic Tests

Sensitivity: A measure of how well a diagnostic test identifies people with a target disorder. Imagine we had 100 people with knee osteoarthritis, and we had an orthopedic test that could accurately identify those with the condition half the time. This would be a sensitivity of 50 percent, equivalent to flipping a coin in terms of accuracy.

Specificity: A measure of how well a diagnostic test identifies people without a target disorder. It is sort of the reverse of sensitivity: Of all the people without the target disorder, how many test negative? Interestingly, the more sensitive a test is, the less specific it will tend to be; the more specific it is, the less sensitive it will tend to be.

Likelihood ratio of a positive test: The ratio of a positive test in those with the target condition to a positive test in those without the target condition. The larger the ratio is, the more positive it is; the more likely a positive test is really positive and will lead to you making a diagnostic decision to treat. It is defined as [sensitivity]/[1-specificity].

Likelihood ratio of a negative test: The ratio of a negative test in those with the condition to a negative test in those without the condition. It is defined as [1-specificity]/[sensitivity].

Validity: The degree to which a test measures what it claims to measure.

Reliability: The ability of a test to provide consistent results when repeated.

Papers That Provide Epidemiological Information

Incidence: The probability of a person being newly diagnosed with a condition of interest, typically measured over a one-year time span.
Prevalence: The proportion of people in a population who have a given condition at a specific point in time (point prevalence) or over a specific time frame (period prevalence).

Risk: An estimate of the proportion of unaffected persons in a population who will develop the disease of interest over a specified period of time. As you can imagine, this is closely related to incidence.

Cross-sectional study: A study done at one point in time. Typically, surveys are a form of cross-sectional study, in that they assess health status and risk at a single point in time. Cross-sectional studies may show associations, but cannot demonstrate causation.

Case-control study: In this type of study, subjects can be divided into cases (those with a condition of interest) and controls (those without). Then, one can retrospectively look at patient records for exposure to a risk factor, and can calculate the odds (using an odds ratio) that the exposure leads to the condition.

Cohort study: In this kind of study, a cohort of individuals, all without a condition of interest, are followed forward in time to examine the relationship between a risk factor and the development of a condition. In this case, a true risk ratio can be calculated, since there are newly diagnosed cases.

Odds ratio: Determined from a case-control study, this is a ratio of the odds of developing the condition in the group exposed to the risk factor, to the odds of developing it in the unexposed group. The larger the ratio, the more likely you are to develop the condition of interest when exposed to that risk factor.

Risk ratio: Determined from a cohort study, this is a ratio of the risk of developing the condition in the group exposed to the risk factor versus those in the unexposed group. As above, the larger the ratio, the greater the risk that exposure leads to the condition.

Risk difference: Also called attributable risk; the difference between risk in the exposed group and the unexposed group. Essentially, it is a measure of risk above baseline.

Number needed to treat: The number of patients needed to be treated to prevent one additional bad outcome. The best NNT one can have is 1; spinal manipulation in some studies has been found to have an NNT ranging from 1.8-2.4. As comparison, statins have an NNT of 104 for non-fatal heart attack over a five-year period, meaning you need to treat 104 people for five years to prevent one additional heart attack. NNT is calculated as 1/[risk difference].

Papers That Provide the Results of Clinical Research

Qualitative research: Research analyzing textual information, rather than numerical data.

Quantitative research: Research analyzing numerical data via statistical means.

Pragmatic research: Research looking at effectiveness; that is, research investigating interventions in real-world settings.

Comparative effectiveness research (CER): The direct comparison of two or more treatments to determine which has the best patient outcomes and less risk of harm.

Explanatory research: Research looking at efficacy; that is, research conducted under highly controlled conditions, unlike real-world conditions. This is better used to understand mechanism.

Sample size calculation: A method of determining the number of participants needed to properly power a study – to provide a reasonable level of comfort that the study results you found were due to study interventions, rather than purely by chance. We want only enough people to power the study, no more, since adding more subjects increases both risk and cost.

Placebo: An inert substance to which an active treatment may be compared.

Effect size: The size or magnitude of a difference following a treatment or intervention.

Sham: A physical procedure designed to look like a real intervention, but which offers no or little benefit.

Bias: Error in a study introduced by variables that may impact study outcome and are not taken into consideration when designing the project. Many kinds of bias exist. For example, investigators may impact study outcomes by giving participants in one group more encouragement or support in addition to the treatment being researched. Or you may have one blood-pressure cuff that takes accurate blood pressures and one that is off by 2 mmHg in a hypertension trial.

Crossover study: A type of study in which subjects act as their own controls. One group begins with the active intervention and later crosses over to the control, while the second group does the reverse.

Literature Reviews

Narrative review: The standard review you are familiar with – an overview on a given topic, usually written by an expert in that topic. It can be a good source of information for topics with which you may be less familiar, but is often selective in the information it presents.

Systematic review: A formal method of search, selection and synthesis, usually involving ranking the papers for quality and including only those that score well. This enhances rigor and removes a great deal of personal bias from the review process.

Meta-analysis: A complicated form of systematic review in which papers of enough similarity collapse their data into a single larger group of individuals for analysis. This provides a more realistic depiction of an effect size for a given intervention, since you are analyzing a larger pool of people than any of the individual studies alone.

Case Reports

Case report: The presentation of a single case and its diagnosis and progression. Case reports should have independent educational value in order to be worthy of publication.

Case series: A series of cases of similar nature, attempting to provide some information on trends related to diagnosis, prognosis or management.

Understanding the Evidence: Step #1 of Proper Application

A main goal of our DC column and blog, "Research – It's Not Just for Scientists Anymore," is to provide the tools you need to navigate clinical situations (such as the example we began this article with) when they occur.

Over the course of a small number of formal articles and a larger number of blog posts, we have provided you with a significant amount of information and frequently referred you to papers in the scientific literature for more information. We view evidence-based practice, in part, as a series of tools you can use to help you extract meaningful information from a high-quality paper in the scientific / chiropractic literature. But sometimes, it can be difficult to take it all in.

Even though research is not just for scientists anymore, we scientists have a pesky habit of using scientific language that can be difficult to interpret. We are trying to do better – really! In the meantime, we hope this evidence-based glossary helps.

August 2015
print pdf