What would be the best study design to answer a clinical question that includes an intervention?

What Type of Question are You Asking?

After you have crafted your question using the PICO format, another step in the process to to identify the type of question you are asking, and identify the optimal type of research design to answer your question. 

Type of Question

Are you asking a question related to a diagnosis? There are four primary question types:

1. Therapy: Does the treatment work? How effective is the intervention?
2. Diagnosis / Diagnostic Test: What is the ability of the test to predict the likelihood of a disease?
3. Prognosis: What is the future course of the patient?
4. Harm/Etiology: What is the cause of the problem? What is the harmful effect of an intervention or exposure?

What Type of Study do you Need?

Type of Study

After you choose your question type, now its time to identity the research design. There are different types of research studies you can locate in the primary literature, and different research designs are better suited to answering different types of clinical questions.

The strength of the evidence increases the higher you are on the hierarchy.  Your risk of bias increases and the strength of evidence decreases the lower you are on the hierarchy.

Dahlgren Memorial Library. Evidence-based medicine resource guide: Clinical questions, PICO, & study designs. http://guides.dml.georgetown.edu/ebm/ebmclinicalquestions. Updated 2016. Accessed June 24, 2016.

Your mnemonic now reads:

• P  =Patient (population or problem)
• I = Intervention (prognostic factor or exposure)
• C = Comparison
• O = Outcome
• T = Type of Question
• T = Type of Study

Types of studies we are going to cover all fall under one of two categories - primary sources or secondary sources. Primary sources are those that report original research and secondary sources are those that compile and evaluate original studies.

Primary Sources

Randomized Controlled Trials are studies in which subjects are randomly assigned to two or more groups; one group receives a particular treatment while the other receives an alternative treatment (or placebo). Patients and investigators are "blinded", that is, they do not know which patient has received which treatment. This is done in order to reduce bias.

Cohort Studies are cause-and-effect observational studies in which two or more populations are compared, often over time. These studies are not randomized.  

Case Control Studies study a population of patients with a particular condition and compare it with a population that does not have the condition. It looks the exposures that those with the condition might have had that those in the other group did not.

Cross-Sectional Studies look at diseases and other factors at a particular point in time, instead of longitudinally. These are studies are descriptive only, not relational or causal. A particular type of cross-sectional study, called a Prospective, Blind Comparison to a Gold Standard, is a controlled trial that allows a research to compare a new test to the "gold standard" test to determine whether or not the new test will be useful.

Case Studies are usually single patient cases.  

Secondary Sources

Systematic Reviews are studies in which the authors ask a specific clinical question, perform a comprehensive literature search, eliminate poorly done studies, and attempt to make practice recommendations based on the well-done studies.

Meta-Analyses are systematic reviews that combine the results of select studies into a single statistical analysis of the results.

Practice Guidelines are systematically developed statements used to assist practitioners and patients in making healthcare decisions.  

In the Evidence Based Medicine paradigm, questions can be categorized into two types to help you grow your knowledge and practice. By determining the type of question, it will guide you to the type of evidence to review for an answer.

1. Background questions concern general knowledge of a condition, disease, process, etc.  These types of questions generally have only two parts:

• A question root (who, what, when, where, how, why) and
• a disorder, test, treatment, or another aspect of health care.  

Often these questions can best be answered by using a health sciences textbook or consulting a clinical point-of-care tool like DynaMed or UpToDate.

Examples:

• What is a first line treatment for heart failure?
• When should children receive the HPV vaccine?
• What gene mutation causes polycystic kidney disease?

2. Foreground questions are specific knowledge questions

• that affect clinical decisions and 
• include a broad range of biological, psychological, and sociological issues. 

Foreground questions are best suited to the PICO model as it captures the essential elements of your information need to help translate that question into a search query. To obtain answers, generally it requires a search of the primary medical literature in databases like PubMed or Embase.

Examples:

• In adults with heart failure, would adding warfarin to standard therapy reduce thromboembolism?
• Are patient education YouTube videos effective at changing the preferences of vaccine-hesitant parents?
• Does taking tolvaptan in adult patients with autosomal dominant polycystic kidney disease improve renal function and pain within six months of initiation of the medication?

Foreground questions may be further categorized into a type of clinical question such as treatment/therapy, diagnosis, prognosis, or etiology/harm that will also help in seeking out appropriate evidence types to address your inquiry. See more about question types on the CLINICAL FILTERS page.

What research design is best for intervention research?

What type of study design is appropriate to answer the clinical question?

What study type is best for answering a clinical question about treatment?

What is the suggested best study for questions on diagnosis?