Objectives

Topics 1
Research Design

Topic 2
Designs for Nursing Research

Topic 3
Concepts Relevant to Design

Topic 4
Design Validity

Topic 5
Good Study Design

Topic 6
Modeling study design

References

TOPIC 2:
Designs for Nursing Research

Research designs can be divided in to two broad categories::

• Non-experimental research – In this category the researcher observes the phenomena as they occur naturally and does not intervene in any way. Types of designs in this category include:
  • Descriptive research
  • Correlational research
    
    
• Experimental research - In this category the researcher plays an active role by manipulating the independent variable (IV) in the form of delivery of a treatment or intervention. Experimental research is conducted to examine cause and effect. Two types of design fall into this category:
  
  
  • Experimental research –The following three elements must be present in a study for it to be considered an experimental design.
    • Random assignment of subjects to the control/comparison group or the experimental group
    • Manipulation of the IV by the researcher
    • Control of the research setting and situation by the researcher
      
      
  • Quasi-experimental research – Because of the ethical nature of conducting studies with human subjects, often an experimental design is not feasible. The quasi-experimental design must include:
    • Manipulation of the IV by the researcher,
      
      but it may lack:
      
    • A control group. However, in quasi-experimental design there is a nonequivalent comparison group. It is called a comparison group, instead of control group because members of the group were not randomly assigned to experimental or control conditions. Thus, there is no assurance that subject characteristics are equally and randomly distributed between groups, and this introduces a potential source of error.
    • Control of the research setting and situation.
      
      
      

• Descriptive
  • Typical Descriptive Design: Examines characteristics of a single sample (explores aspects of phenomena of interest).
  • Comparative Descriptive Design: Compares two or more groups that occur naturally in a setting (explores for differences).
  • Case Study: Intensive exploration of a single unit of study (a person, family, group, community, or institution).
  • Longitudinal descriptive (not covered in the book): Studies a sample of individuals over time to examine patterns of change, growth, or trends across time. 
• Correlational
  • Descriptive Correlational Design: Describes the relationship among variables in a particular sample. 
  • Multifactorial Correlational (not covered in the book):  Because we know reality in the world is multicausal, that is, there are many factors that influence an outcome, most correlational studies now are multifactorial and examine correlations between more than 2 variables simultaneously.
  • Predictive Correlational: Attempts to explore what factors predict (have an influence on) another variable. Because this design explores causality, the term independent variable is used to describe the predictor variables that are thought to predict the outcome variables (often called the dependent variable). (Please note: This use of the terminology independent variable and dependent variable is confusing because the terms have a much more exact meaning in quasi and experimental designs).
  • These studies can be done prospectively (with all data collected at the same time), but this is not a strong design. A strong design for a predictive correlational study collects data on the independent (predictor) variables at one point in time then examines the extent to which these variables relate to (predict) an outcome at another point in time.
  • Model Testing Design: This design is similar to a predictive correlational design but is testing a hypothesized causal model.  It requires a larger sample size and very tight measurement of variables. (Read this, but YOU ARE NOT RESPONSIBLE FOR LEARNING IT FOR THE MIDTERM.)
• Quasi-experimental
  • Similar to experimental design, but does not have the same level of control (e.g., are not able to control as many threats as possible). Aspects that might not be controlled are:
    • Not having a control group to which to compare results
    • Not randomly assigning participants to control or experimental group
    • Not having control over the intervention (e.g., using an intervention that is in place in practice).
  • Many different quasi-experimental designs exist. You should understand the concepts that make such a design different, but do not need to memorize each of the many different types of quasi-experimental designs.
• Experimental 
  • Provides the best method possible to examine a cause and effect situation.  Provides the greatest amount of control for examining causality.  Tries to reduce or eliminate (almost impossible) all factors that influence the dependent (outcome) variable other than the intervention (the independent variable). This is done by:
    • Sample selection for homogeneity (which we will discuss later)
    • Random sampling (randomly enrolling individuals from the population is not usually possible in health studies—we usually have a convenience sample.
    • Random assignment of participants into control and experimental group (so every eligible person has an equal chance of being assigned to the experimental/intervention or the control group—thereby theoretically evenly distributing into each group any characteristics of the sample that might have an impact on the outcome variable)
    • Controlling the intervention as tightly as possible and controlling as many other extraneous variables as possible. 
       
      
  • Experimental designs are varied.  Standardized randomized pre-test post-test design with control and experimental group, and randomized clinical trial are two commonly used experimental designs.
    
• Additional Points
  • Retrospective versus Prospective -
    • Descriptive and correlational studies can be retrospective (this means that the data being analyzed was either collected in the past as in chart data or that the data collection focuses on the past. Example: Interviewing mothers of 3-year-old prematurely-born children about their experiences when these children were in the NICU. 
    • Prospective means that the data was collected after the study was designed and focuses on the here and now.
  • Longitudinal - Almost any design can be longitudinal. 
    • This means that data were collected over time, not just once or twice. 
    • Descriptive longitudinal studies collect data from subjects about phenomena over weeks, months, or years. Many developmental studies are descriptive longitudinal studies that continue for the lifetime of the individuals enrolled!
    • These data are used descriptively and also are used for many different correlational, and especially multifactorial correlational, analyses to explore factors that influence positive and problematic developmental outcomes.
    • Experimental and quasi-experimental studies can collect outcome data longitudinally to determine whether the experiment (intervention, independent variable) has a long-term effect.

Learning research can be difficult because it is not exactly black and white,  which would make it easier to comprehend.  Designs are not as rigidly defined as the book indicates. A study can use a combination of designs (most sophisticated studies do). As noted above, a longitudinal descriptive study of developmental outcomes is undoubtedly going to also be a correlational study and will likely conduct predictive correlational analyses with their data.