Nursing Research and Observational Design
![Observational Research Design in Nursing Observational Research Design in Nursing](https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjUOH8hVmVrFwVn6JjXt2vc5t32Fjidz0cRLQH67k_yftUgQv4GPCGFIyd7dIAXw5RVT56jkibSCrM1j_StXmvCvXSuM73PGziZLKgDWqZ17J38AYc5SzAvEkxNe1jOD3MCwCig59ipC-vyGLsuwYF00lDsw8scqHxYWwWRIH0geFhX8LipTdeEbOkb/w640-h320/Observational%20Research%20Design%20(1).png)
![Observational Research Design in Nursing Observational Research Design in Nursing](https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjUOH8hVmVrFwVn6JjXt2vc5t32Fjidz0cRLQH67k_yftUgQv4GPCGFIyd7dIAXw5RVT56jkibSCrM1j_StXmvCvXSuM73PGziZLKgDWqZ17J38AYc5SzAvEkxNe1jOD3MCwCig59ipC-vyGLsuwYF00lDsw8scqHxYWwWRIH0geFhX8LipTdeEbOkb/w640-h320/Observational%20Research%20Design%20(1).png)
Observational Research Design,Purposes of Observational Designs,Types of Observational Designs According to Research Method,Longitudinal Comparative Design,Use in the Experimental Research to Avoid Biases,Selection Biases ,Measurement Biases,Causes of Biases,Conclusion.
Observational Research Design
Observational designs are non-experimental, quantitative designs. In contrast to experimental designs in which the investigator manipulates the independent variable and observes its effect, the investigator conducting observational research observes both the independent and the dependent variables.
In observational studies, variation in the independent variable is
due to genetic endowment, self-selection, or occupational or environmental
exposures. Because of the myriad sources of bias that can invalidate naturally
occurring events, rigorous designs and methods are required to minimize bias.
Observational designs should not be confused with. Observational methods of
data collection.
Purposes of Observational Designs
Observational designs are used when there is not enough knowledge
about a phenomenon to manipulate it experimentally. Sometimes research
involving human subjects is restricted to observational designs because of the
nature of the phenomenon; that is, experimental research is precluded for
ethical reasons.
Types of Observational Designs According to Research Method
Observational designs include quantitative, descriptive studies as well as analytical studies that are designed to test hypotheses. Descriptive, observational studies provide a basis for further study by describing and exploring relationships between variables, informing the planning of health services, and describing clinical practice for individual clients or groups of clients.
In contrast, analytic research is designed to test specific hypotheses in order to draw conclusions about the impact of an independent variable or set of variables on an outcome or dependent variable under scrutiny.
Observational
designs are classified as longitudinal or cross sectional. In a cross-sectional
study, all the measurements relate to one point in time; In the longitudinal
approach, measurements relate to at least two points in time.
A cross-sectional study, sometimes referred to as a correlational study, is conducted to establish that a relationship exists between variables. The term correlational refers to a method of analysis rather than a feature of the design itself.
Cross-sectional studies are useful if the independent
variable is an enduring or invariable personal characteristic, for instance,
gender or blood type. Cross-sectional studies are also useful for exploring
associations between variables.
Longitudinal Comparative Design
Longitudinal comparative designs are usually undertaken to explain the relationship: between an independent variable and an outcome. One type of longitudinal, comparative design is referred to as a cohort study. Although the investigator does not manipulate the independent variable, the logic and flow in a cohort study is the same as the logic of an experiment.
Subjects are measured or categorized on the basis of the independent variable and are followed over time for observation of the dependent variable. In a cohort study it is established at the outset that subjects have not already exhibited the outcomes of interest (dependent variable).
Thus, the time sequencing of events can be established. In other words, it can be demonstrated that the independent variable preceded the occurrence of the dependent variable.
Another type of longitudinal, comparative design is a case comparison study. In this design the flow is the opposite of a cohort study. Subjects are selected and categorized on the basis of the dependent variable (the outcome of interest). The purpose of the study is to test hypotheses about factors in the past (independent variables) that may explain the outcome.
Although case comparison designs are not prevalent in the nursing research literature,
they have great potential for studies of outcomes that occur infrequently.
Furthermore, this design is very efficient because it is possible to achieve
greater statistical power with fewer subjects than in other types of
observational designs.
Longitudinal comparative designs are also classified according to the time perspective of the events under study in relation to the investigator's position in time. A study is retrospective if, relative to when the investigator begins the study, the events under investigation have already taken place.
A study is prospective if the outcomes that are being investigated
have not yet taken place when the study is initiated. Various hybrid designs
are also possible; Referred to as ambit directional studies, they combine
features of both designs.
Use in the Experimental Research to Avoid Biases
As in experimental research, observational research designs and methods are selected with the aim of minimizing bias. Bias refers to distortion in the result of a study. A biased study threatens internal validity if the distortion is sufficient to lead to an erroneous inference about the relationship between the independent and dependent variable.
Potential sources of bias that can
threaten the internal validity of observational studies are those related to
selection, measurement, and confounding.
Selection Biases
Selection bias is a distortion in the estimate of effect resulting from
(a) flaws in the choice of groups to be compared
(b) inability to locate or recruit subjects selected into the sample, resulting in differential selection effects on the comparison groups
(c) subsequent attrition of
subjects who had initially agreed to participate, which changes the com
position of the comparison groups.
Measurement Biases
Measurement bias occurs when the independent variable or outcome (dependent variable) is measured in a way that is systematically inaccurate and results in distortion of the estimate of effect. Major sources of measurement bias are:
(a) a defective measuring instrument
(b) a procedure for ascertaining the outcome that is not sufficiently sensitive and specific
(c) the likelihood of detecting the outcome dependent on the subject's status on the independent variable
(d) selective recall or reporting by subjects
(e) lack of blind
measurements when indicated.
Causes of Biases
Because of the lack of randomization in a non-experimental study, uncontrolled confounding variables are a major threat to internal validity. Unless confounding factors are controlled in the design of the study or in its analysis, distortion in the estimate of effect will result.
A confounding factor operates through its association with both the independent and the dependent variables.
It can distort the results in either direction; that is, it can lead to an overestimation of the relationship between the independent and dependent variables by producing an indirect statistical association, or it can lead to an underestimate of the relationship between the independent and dependent variables by masking the presence of an association between the independent and dependent variables.
A distinction between confounding bias and other types of bias is that confounding is correctable at the design or analysis stage of the study, whereas bias due to selection and measurement problems are usually difficult or impossible to correct in the analysis. Confounding can be controlled or minimized at the design stage of the study by restricting the study sample or by matching the comparison groups.
At the
analysis stage confounding can be controlled or minimized by using a
multivariable approach to the statistical analysis to adjust for the
confounding factors or by examining the independent or dependent variable
relationship within specified levels or categories of the confounding factors
(stratified analysis). Confounding variables should not be confused with
mediator and moderator variables.
Conclusion
In the absence of randomization and manipulation, myriad sources of bias can influence observations and conclusions drawn from naturally occurring events, thus, rigorous observational designs and methods are essential.
Give your opinion if have any.