SciELO - Scientific Electronic Library Online

 
vol.15 issue3Reflections on nursing teaching in the post-modernity era and the metaphor of a theory-practice gapThe PICO strategy for the research question construction and evidence search author indexsubject indexarticles search
Home Pagealphabetic serial listing  

Services on Demand

Journal

Article

Indicators

Related links

Share


Revista Latino-Americana de Enfermagem

On-line version ISSN 1518-8345

Rev. Latino-Am. Enfermagem vol.15 no.3 Ribeirão Preto June 2007

http://dx.doi.org/10.1590/S0104-11692007000300022 

UPDATING ARTICLE

 

An overview of research designs relevant to nursing: Part 1: Quantitative research designs

 

 

Valmi D. SousaI; Martha DriessnackII; Isabel Amélia Costa MendesIII

IPhD, APRN, BC, Assistant Professor, College of Health and Human Services, The University of North Carolina at Charlotte, United States of America, e-mail: vdsousa@uncc.edu
IIPhD, APRN, BC, Postdoctoral Research Fellow in Clinical Genetics, College of Nursing, The University of Iowa, United States of America, e-mail: martha-driessnack@uiowa.edu
IIIPhD, RN, Full Professor, University of São Paulo at Ribeirão Preto College of Nursing, WHO Collaborating Centre for Nursing Research Development, Brazil, CNPq Researcher 1A, e-mail: iamendes@eerp.usp.br

 

 


ABSTRACT

This three part series of articles provides a brief overview of relevant research designs in nursing. The first article in the series presents the most frequently used quantitative research designs. Strategies for non-experimental and experimental research designs used to generate and refine nursing knowledge are described. In addition, the importance of quantitative designs and the role they play in developing evidence-based practice are discussed. Nursing care needs to be determined by the results of sound research rather than by clinical preferences or tradition.

Descriptors: research; nursing research; quantitative analysis; methodology; nursing


 

 

INTRODUCTION

A research design is the framework or guide used for the planning, implementation, and analysis of a study(1-2). It is the plan for answering the research question or hypothesis. Different types of questions or hypotheses demand different types of research designs, so it is important to have a broad preparation and understanding of the different types of research designs available. Research designs are most often classified as either quantitative or qualitative. However, it is becoming more common for investigators to combine, or mix, multiple quantitative and/or qualitative designs in the same study(3).

Quantitative research designs most often reflect a deterministic philosophy that is rooted in the post-positivist paradigm, or school of thought. Post-positivists examine cause, and how different causes interact and/or influence outcomes. The post-positivist paradigm adopts the philosophy that reality can be discovered, however only imperfectly and in a probabilistic sense. The approach is typically deductive - where most ideas or concepts are reduced into variables and the relationship between or among them are tested(1,3). The knowledge that results is based on careful observation and measurement and interpretation of objective reality.

In contrast, qualitative research designs are rooted in the naturalistic paradigm. The approach to study is inductive, rather than deductive, and begins with the assumption that reality is subjective, not objective, and that multiple realities exist, rather than just one(1,3). When little is known about a particular phenomenon, experience, or concept, a qualitative design is often used first. Once concepts and/or themes are identified, or grouped into a theory, they can then be tested using a quantitative design or approach. Quantitative research designs primarily involve the analysis of numbers in order to answer the research question or hypothesis, while qualitative designs primarily involve the analysis of words.

 

RELEVANT QUANTITATIVE RESEARCH DESIGNS

Quantitative research designs adopt objective, rigorous, and systematic strategies for generating and refining knowledge(1-4). They primarily use deductive reasoning and generalization. Deductive reasoning is the process in which the researcher begins with an established theory or framework, where concepts have already been reduced into variables, and then gathers evidence to assess, or test, whether the theory or framework is supported(1). Generalization is the extent to which conclusions developed from evidence collected from a sample can be extended to the larger population(1).

Quantitative research is most often about quantifying relationships between or among variables - the independent or predictor variable (s) and the dependent or outcome variable (s). Broadly, quantitative research designs are classified as either non-experimental or experimental (Table 1). Non-experimental designs are used to describe, differentiate, or examine associations, as opposed to direct relationships, between or among variables, groups, or situations. There is no random assignment, control groups, or manipulation of variables, as these designs use observation only. The most common non-experimental designs are descriptive or correlational studies.

 

 

Non-experimental designs are often further classified according to timing of data collection, cross-sectional or longitudinal, or according to the timing of the experience or event being studied, retrospective or prospective(1,5). In a cross-sectional study, variables are identified one point in time and the relationships between them are determined. In a longitudinal study, data are collected at different points over time. In a retrospective study, an event or phenomenon identified in the present is linked to factors or variables in the past. In a prospective study, or cohort study, potential factors and variables identified in the present are linked to potential outcomes in the future.

 

NON-EXPERIMENTAL RESEARCH DESIGNS

Non-experimental designs do not have random assignment, manipulation of variables, or comparison groups. The researcher observes what occurs naturally without intervening in any way. There are many reasons for undertaking non-experimental designs. First, a number of characteristics or variables are not subject or amenable to experimental manipulation or randomization. Further, some variables cannot or should not be manipulated for ethical reasons. In some instances, independent variables have already occurred, so no control over them is possible.

Non-experimental designs may resemble the posttest-only experiment. However, there is a natural assignment to the condition or group being studied, as opposed to random assignment, and the intervention or condition (X) is something that has happened naturally, not imposed or manipulated. The most common methods used in non-experimental designs involve exploratory surveys and/or questionnaires. Non-experimental designs are typically classified as either descriptive or correlational (Table 1).

Descriptive Designs

Descriptive, or exploratory studies are used when little is known about a particular phenomenon(1,6). The researcher observes, describes, and documents various aspects of a phenomenon. There is no manipulation of variables or search for cause and effect related to the phenomenon. Descriptive designs describe what actually exists, determine the frequency with which it occurs, and categorizes the information. Researchers pose Level I research questions(2,7-8) (Table 1). The results provide the knowledge base for potential hypotheses that direct subsequent correlational, quasi-experimental, and experimental studies. The two most common types of quantitative descriptive designs are: case control and comparative(1,6).

Case Control Studies. Case control studies involve a description of cases with and without a pre-existing condition or exposure. The cases, subjects, or units of study can be an individual, a family, or a group. Case control studies are more feasible than experiments in cases in which an outcome is rare or takes years to develop. This design is also known as a case report or case study.

Comparative Studies. Comparative studies are also called ex post facto or causal-comparative studies. These studies describe the differences in variables that occur naturally between two or more cases, subjects, or units of study. Researchers who use a comparative design normally pose hypotheses about the differences in variables between or among two or more units. The main difference between this approach and the quasi-experimental design is the lack of researcher control of the variables.

Correlational Designs

Correlational designs involve the systematic investigation of the nature of relationships, or associations between and among variables, rather than direct cause-effect relationships. Correlational designs are typically cross-sectional(1,6). These designs are used to examine if changes in one or more variable are related to changes in another variable(s). This is referred to as co-variance. Correlations analyze direction, degree, magnitude, and strength of the relationships or associations. The results from correlational studies provide the means for generating hypotheses to be tested in quasi-experimental and experimental studies. Researchers may pose Level I or II research questions(2,7-8) (Table 1). Three of the most common correlational designs include: descriptive, predictive, and model testing correlational design(1,6).

Descriptive Correlational Designs. Descriptive correlational studies describe the variables and the relationships that occur naturally between and among them.

Predictive Correlational Designs. Predictive correlational studies predict the variance of one or more variables based on the variance of another variable (s). As with experimental designs, the study variables are classified as independent (predictor) and dependent (outcome). However, these variables are not manipulated, but occur naturally.

Model Testing Correlational Designs. Model testing correlational studies examine, or pilot test, proposed relationships for a model or theory. As with experimental designs, the study variables are classified as independent (predictor) and dependent (outcome). However, the variables are not manipulated, but occur naturally.

 

EXPERIMENTAL DESIGNS

Experimental designs typically use random assignment, manipulation of an independent variable(s), and strict controls(1,6,9). These characteristics provide increased confidence of cause-and-effect relationships. Random assignment means that each subject had equal chance to be assigned to either the control or experimental group. The use of random assignment of subjects attempts to eliminate systematic bias. Random assignment is different from random sampling. Random sampling means that each subject had an equal chance of being selected from a larger group to participate in the study. This approach is often used in survey research to facilitate generalization. However, it is the random assignment to different conditions that distinguishes a true experimental design. To be classified as true experimental, there must be randomization, a control group, and manipulation of a variable when examining the direct causal or predicted relationship between variables. When any one of these requirements is not met, the design is no longer a true experiment and is classified as quasi-experimental. Researchers typically pose Level III research questions(2,7-8) (Table 1).

True-Experimental Designs

True experimental designs examine the cause and effect relationships between independent (predictor) and dependent (outcome) variables under highly controlled conditions. The simplest of all experimental designs is the posttest-only control group. Other common true-experimental designs include the posttest only control group design, pretest-posttest control group design, Soloman four group design, and cross-over design(1,6,9).

Posttest Only Control Group Design. In posttest only control group design, subjects are randomly assigned (R) to either a control or an experimental group. The groups are not pretested. One group is exposed to a treatment (X) or series of different treatments (X1, X2), and then both groups are posttested (O).

 

 

Pretest-Posttest Control Group Design. In the pretest-posttest control group design, or classic experiment, subjects are randomly assigned (R) to either a control or experimental group. Both groups are pretested (O). The experimental group is exposed to a treatment (X) or different treatments (X1, X2), and then both groups are posttested (O).

 

 

Solomon Four-Group Design. In Solomon four-group design, subjects are randomly assigned (R) to one of four different groups. Two of the groups are pretested (O) and two are not. Only one pretested and one not pretested group are then exposed to a treatment (X). All of the groups are postested (O).

 

 

Cross-over Design. In the cross-over, or counterbalanced, switchover, or rotation design, subjects are given two treatments, one being the experimental treatment (XE), the other a control or reference treatment (XC). The subjects are randomly assigned to one of two groups. One group receives the experimental treatment first and the other group receives the experimental group second. After a period of time, sufficient to allow for any treatment effect to wash out (W), the treatments are crossed over. Multiple cross-over designs involve several treatments.

 

 

Quasi-experimental Designs

Quasi-experimental, like true-experimental designs, examine cause-and-effect relationships between or among independent and dependent variables. However, one of the characteristics of true-experimental design is missing, typically the random assignment of subjects to groups. Although quasi-experimental designs are useful in testing the effectiveness of an intervention and are considered closer to natural settings, these research designs are exposed to a greater number of threats of internal and external validity, which may decrease confidence and generalization of study's findings. The most common used quasi-experimental designs are: non-equivalent group pretest-posttest group design, control-group interrupted time series design, single-group interrupted time-series design, and counterbalanced design(1,6,9).

Non-equivalent pretest-posttest control group design. The non-equivalent pretest-posttest control group design is identical in many ways to the pretest-posttest control group design except that subjects are not randomly (NR) assigned to groups. Both groups are pretested (O) and posttested (O). However, only the experimental group is exposed to a treatment (X).

 

 

Control-group Interrupted Time Series Design. In the control-group interrupted time series design, groups are measured or tested repeatedly on the same variable over time. Again, there is no random assignment (NR) to groups. The experimental group is exposed to a treatment (X) at some point in the series while the control group is not.

 

 

Single-group Interrupted Time-Series Design. With the single-group interrupted time-series design, the researcher measures only one group repeatedly, both before and after exposure to a treatment (X).

 

 

Counterbalanced Design. The counterbalanced design is similar to the cross-over experimental design except that subjects are not randomly assigned (NR) to the different groups. All groups are exposed to all treatments. The most common counterbalanced design is the Latin square, where four different treatments are applied to four naturally assembled groups or individuals. Each of the groups or individuals is posttested after each treatment. The number of treatment and groups must be equal. The Latin square is shown here.

 

 

SELECTION OF QUANTITATIVE RESEARCH DESIGN

The selection of a research design is based on the research question or hypothesis and the phenomena being studied. A true-experimental design is considered the strongest or most rigorous with regard to establishing causal effects and internal validity. Internal validity is the control of factors within the study that might influence the outcomes besides the experimental intervention or treatment. A non-experimental design is generally the weakest in this respect. However, this does not mean that non-experimental designs are weak designs overall. They are weak only with respect to assessing cause-effect relationships and the establishment of internal validity. In fact, the simplest form of non-experiment, the one-time survey design that consists of one single observation (O), is one of the most common forms of research and, for some research questions, especially descriptive ones, is clearly a strong and most appropriate design.

 

CONCLUSION

Research is important to the nursing profession. It is designed to provide new knowledge, improve health care, and challenge current nursing practice with new ideas. Evidence-based nursing practice comes from the idea that the care we provide be determined by sound research rather than by clinician preference or tradition. Understanding how to select the best design to answer a research question or test a hypothesis is the first step in conducting meaningful research. This process assists nurses as they read and critique original research articles. Nursing practice is seldom changed based on one study. It is the accumulation of results from several studies, often using different research designs that provide enough evidence for change.

In the first article of this series, we have presented an introduction and overview to different quantitative research designs, including descriptive, correlational, true-experimental, quasi-experimental designs. Each design offers a unique approach or plan for answering a nursing research question. In the next article, qualitative research designs will be presented and discussed, providing nurses with even more choices of design. Finally, in the third article, the combination, or mixing of designs within one study, will be introduced. At the completion of this series, nurses will have an overview of relevant research designs for nursing research and be able to select an appropriate design as a framework or guide for a potential study.

 

REFERENCES

1. Burns N, Grove SK. The practice of nursing research: conduct, critique, and utilization. 5th ed. St Louis: Elsevier; 2005.        [ Links ]

2. Polit DF, Beck CT, Hungler BP. Essentials of nursing research: methods, appraisal, and utilization. 5th ed. Philadelphia: Lippincott; 2001.        [ Links ]

3. Creswell JW. Research design: qualitative, quantitative, and mixed methods approaches. 2nd ed. Thousand Oaks: Sage Publications; 2003.        [ Links ]

4. Carvalho V. Cuidando, pesquisando e ensinando: acerca de significados e implicações da prática da enfermagem. Rev Latino-am Enfermagem 2004 setembro/outubro; 12(5):806-15.        [ Links ]

5. Demo P. Pesquisa qualitativa: busca de equilíbrio entre forma e conteúdo. Rev Latino-am Enfermagem 1998 abril; 6(2):89-104.        [ Links ]

6. Walker W. The strengths and weaknesses of research designs involving quantitative measures. J Res Nurs 2005; 10(5): 571-82.        [ Links ]

7. Seers K, Crichton N. Quantitative research: Designs relevant to nursing and healthcare. NT Res 2001; 6(1): 487-500.        [ Links ]

8. LoBiondo-Wood G, Haber J. Nursing research: Methods, critical appraisal, and utilization. 5th ed. St Louis: Mosby; 2002.        [ Links ]

9. Blink P, Wood M. Advanced design in nursing. Thousand Oaks: Sage Publications; 1998.        [ Links ]

10. Cassidy CM, Hart JA. Methodological issues in investigations of massage/bodywork therapy: Part III: Qualitative and quantitative design for MBT and the bias of interpretation. J Bodywork and Movement Ther 2003; 7(3): 136-41.        [ Links ]

11. Shadish WR, Cook TD, Campbell DT. Experimental and quasi-experimental designs for generalized causal inference. New York: Houghton Mifflin Company; 2002.        [ Links ]

 

 

Recebido em: 21.6.1006
Aprovado em: 6.3.2007

Creative Commons License All the contents of this journal, except where otherwise noted, is licensed under a Creative Commons Attribution License