Study Design Selection

Why Study Design Matters

Different research questions require different study designs. Selecting appropriate designs ensures your review includes the most relevant and methodologically sound evidence.

Quantitative Study Designs

Randomized Controlled Trials (RCTs)

The gold standard for intervention studies. Participants are randomly assigned to intervention or control groups.

  • Best for: Testing intervention effectiveness
  • Strengths: Minimizes bias, establishes causality
  • Limitations: May lack external validity, expensive
  • Variants: Cluster RCTs, crossover trials, factorial designs

Cohort Studies

Follow groups over time to assess exposures and outcomes. Can be prospective or retrospective.

  • Best for: Long-term outcomes, rare exposures, prognosis
  • Strengths: Temporal sequence clear, multiple outcomes
  • Limitations: Expensive, time-consuming, potential confounding
  • Useful when RCTs are unethical or impractical

Case-Control Studies

Compare people with a condition (cases) to those without (controls), looking back at exposures.

  • Best for: Rare diseases, initial exploration of associations
  • Strengths: Efficient for rare outcomes, relatively quick
  • Limitations: Recall bias, selection bias, cannot establish causality
  • Less preferred than cohort studies when both are available

Cross-Sectional Studies

Assess exposure and outcome at a single point in time.

  • Best for: Prevalence, associations, survey data
  • Strengths: Quick, inexpensive, descriptive
  • Limitations: Cannot establish temporal sequence or causality
  • Often excluded from intervention reviews

Quasi-Experimental Studies

Intervention studies without randomization (e.g., pre-post designs, interrupted time series).

  • Best for: Real-world interventions, policy evaluations
  • Strengths: More feasible than RCTs in some settings
  • Limitations: Higher risk of bias, confounding
  • Consider when RCTs are insufficient or unavailable

Qualitative Study Designs

Phenomenology

Explores lived experiences and meanings.

  • Best for: Understanding patient experiences, perceptions
  • Methods: In-depth interviews, descriptive analysis
  • Example: 'Experiences of living with chronic pain'

Grounded Theory

Develops theories from systematically collected data.

  • Best for: Understanding processes, developing frameworks
  • Methods: Iterative data collection and analysis
  • Example: 'How patients navigate healthcare systems'

Ethnography

Studies cultures and contexts through immersion.

  • Best for: Cultural practices, organizational behavior
  • Methods: Participant observation, field notes
  • Example: 'Workplace culture in emergency departments'

Selecting Designs for Your Review

  1. 1Start with your research question type (intervention, prognosis, diagnosis, etc.)
  2. 2Identify the highest quality designs that answer your question
  3. 3Consider whether lower-quality designs should be included if high-quality is scarce
  4. 4Check similar published reviews to see what designs were included
  5. 5Be explicit about design hierarchy in your protocol
  6. 6Plan sensitivity analyses excluding lower-quality designs

Mixed Methods Reviews

Some reviews integrate both quantitative and qualitative evidence to address complex questions.

  • Effectiveness + experiences: RCTs for outcomes, qualitative for patient experiences
  • Implementation reviews: Quantitative for impact, qualitative for barriers and facilitators
  • Requires methods for synthesizing both types of evidence

Configuring in Scholara

Study design preferences are part of your protocol. They're set when the assistant builds your protocol during the Protocol phase, and can be edited from the protocol artifact in the side drawer or by asking the assistant directly — for example, "add quasi-experimental designs to the protocol" or "exclude case series from this review."

Your design choices are then used to:

  • Inform the search strategy so design-specific filters are applied where helpful
  • Drive abstract and full-text screening decisions when a study's design doesn't match
  • Organise included studies by design when you generate characteristics tables in the Analysis phase

When in doubt, err on the side of being inclusive initially. You can always exclude lower-quality designs in sensitivity analyses. However, combining very different designs (e.g., RCTs and case series) in meta-analysis is generally not recommended.