Literature Review: Observational Retrospective Cohort Studies

Nurses are increasingly expected to understand and conduct research and engage in disciplined studies benefiting nursing and its patients (Polit & Beck, 2017). Research findings from rigorous studies provide strong evidence for informing nurses’ decisions and actions and nurses are accepting the need to base specific actions on research evidence (Polit & Beck, 2017). As well, the call for cutting-edge research to meet individual, group and societal needs around the world is more urgent (Reio, 2016). Understanding the research methods used in the studies that guide best practice are necessary for nurses to be able to adequately interpret results and fully understand how the knowledge translates to practice. These better understandings will contribute to efforts to adapt and manage change. The purpose of this literature review is to fully discuss the nonexperimental study design using the existing literature on methodologies.

An ancestry search was conducted using the Cumulative Index of Nursing and Allied Health Literature (CINAHL) database using the search terms “methodology”, “research design”, “observational research”, “retrospective design” and “cohort studies”. Along side, “Quantitative Applications in the Social Sciences: Research designs” by P.E. Spector (1981), the most relevant observational research method articles and books were included in this literature review. Historic books, as Spector’s (1981) were included as a result of the quality of the published work describing research methods. Articles pertaining directly to retrospective and cohort designs were included, however articles using observational research of any sort as a method of specific study were excluded. The literature is organized in a way that discusses observational research broadly, and then more narrowly to its subsets of retrospective and cohort designs. Strengths and limitations of observational research are included at the end of the literature review, along with a comparison to other research methods.

Origin

Observational research has its roots in the naturalistic observational methods of naturalists who study nonhuman animals or ethnographers who study human behavior (Brandes, 2008). The use of observational research of various kinds can be found in all the social sciences and in fields that range from business to biology (Brandes, 2008). The term cohort has a military root and consists of bands or groups of people marching forward in time from an exposure to one or more outcomes (Grimes & Schulz, 2002). Specifically, retrospective cohort studies have almost as long a history as the prospective studies, with one dating back to 1933 about a black population in Tennessee (Doll, 2001). From 1933 to today, nonexperimental designs are used in research when the researcher either wants to describe a group or examine relationships between pre-existing groups (Salkind, 2010).

Get Help With Your Essay

If you need assistance with writing your essay, our professional essay writing service is here to help!
Find out more about our Essay Writing Service

Nonexperimental/Observational Design

Nonexperimental research designs are either quantitative, qualitative or mixed method (Reio, 2016). Nonexperimental or observational designs manipulate variables that could normally not be manipulated ethically and explore causal relationships when experimental work is not possible (Polit & Beck, 2017). Generally, when using observational research, many more observations or data points than typical experimental approaches are used but may not all be able to be studied (Brandes, 2008).

These studies are considered nonexperimental because they don’t involve a random assignment of participants to group or have an active introduction or manipulation of an intervention by the researcher (Cook & Cook, 2008). Variables in nonexperimental studies can be related in terms of direction, positive or negative, and strength, which is the consistency with which the variables correspond with each other (Cook & Cook, 2008). Control can be exercised in the selection of cases to include in the observational study, where cases can be chosen to meet certain specified criteria and control is achieved by selection only certain values of control variables and observing the variables of interest (Spector, 1981).

Lastly, erroneous utilization of imprecise and mistaken language must be avoided in nonexperimental research because these studies, when published, become part of the scientific literature and is a steppingstone further research related to the topic area (Reio, 2016).

Retrospective Cohort Study Design

While researchers of nonexperimental studies attempt to find associations between variables, retrospective studies specifically link past and present phenomena in which a researcher begins with a dependent variable and then examines whether it is correlated with one or more previously occurring independent variables (Polit & Beck, 2017). Retrospectively, eligible subjects are identified, a cohort is composed, and exposures are assessed (Euser, Zoccali, Jager & Dekker, 2009). This data is collected from existing records or documents and can immediately be analyzed. As well, many retrospective studies are cross-sectional, with data on both the dependent and independent variables at a specific time (Polit & Beck, 2017). This means that data for the independent variables are a recollection.

A retrospective cohort study allows the researcher to describe a population over time or obtain a preliminary measure of association for the development of future studies and interventions (Salkind, 2016) and can be completed with considerably fewer resources (Hartung & Touchette, 2009). The exposure and outcome of information in a cohort study are identified retrospectively using administrative data sets through reviews of patient charts, interviews, etc. (Salkind, 2016). In retrospective cohort studies, relevant information about the cohort members and outcomes of interest are obtained from past records (Doll, 2001), where comparisons are made between the experience of an exposed group to some factor within another group that has not been exposed (Grimes & Schulz, 2002). If not studying exposure, other retrospective cohort studies involve using previously existing data set or medical records to analyze what occurred following cohort assignment (Hartung & Touchette). The hallmark of all cohort studies is the following of groups through time with ultimate ascertainment of their development, and whether studying exposure or not, are still considered powerful study designs (Hartung & Touchette). The outcomes of cohort studies must be defined in advance and should be clear, specific, and measurable (Grimes & Schulz, 2002). The identification of outcomes should be comparable in every way to both study groups to avoid information bias, and if studying exposure, in a cohort study, the exposure should be quantified by degree (Grimes & Schulz, 2002).

Philosophical Underpinnings and Theoretical Framework

The philosophical underpinning guiding observational research is post-positivism. Using a post-positivist paradigm, there is still a belief and desire to understand reality, except there is recognition to the notion that it is impossible to reach total objectivity (Polit & Beck, 2017). A post-positivist recognizes the way scientists think and work and believe that scientific reasoning and common-sense reasoning are essentially the same process (Trochim, 2020). Post-positivists believe there is no difference in kind between the two, except scientists, follow specific procedures to assure that observations are verifiable, accurate and consistent (Trochim, 2020).

Theoretical frameworks are structures that hold or support a theory in a research study and introduces and describes the theory which explains why the study problem exists (Polit & Beck, 2017). Although theoretical frameworks can be applied to observational research, they are not normally used in retrospective cohort studies that focus on administrative or medical data collection. This type of research uses a quantitative approach when analyzing data, however, if there are theories or concepts that are relevant to the topic being studied, the idea of using a theoretical framework can be revisited. Researchers tend to see nonexperimental research as being useful in the early stages where a hypothesis can be developed and a relationship between two variables can be predicted by theory (Reio, 2016). However, is not appropriate for theory validation because this type of research is not capable of eliminating the factors that jeopardize the validity of experimental designs (Reio, 2016). Sometimes researchers fail to formally acknowledge or describe their framework being used and it may be difficult to figure out what the researcher thought was “going on”, however, in qualitative studies the frameworks become embedded in the study (Polit & Beck, 2017).

Strengths and Limitations

Many experimental designs are so full of confounding variables that causal inferences cannot be made with reasonable confidence, and there are nonexperimental observational designs that can establish causal chains of events (Spector, 1981). Systematic observation can allow powerful causal conclusions, and the efficacy of these designs should be respected. Practically all modern knowledge of astronomy comes from nonexperimental observation (Spector, 1981).

An advantage of the retrospective study is that it is simple, quick and inexpensive to conduct, requires fewer subjects and is the only feasible method for studying something with a long lag between exposure and outcome (Salkind, 2016). This is prevalent in social science research as it would often be unethical to manipulate an independent variable (Cook & Cook, 2008). By using a retrospective design, researchers could study multiple exposures and multiple outcomes within one cohort as it is relatively easy to pick up associations between many exposures and outcomes (Euser et al., 2009). Retrospective studies also allow for easy hypothesis generation, quantitative and specific, as a result of being able to study multiple variables and outcomes, which then can subsequently be tested in randomized control trials (RCTs). As well, they are a very time efficient and elegant way of answering new questions with existing data (Euser et al., 2009).

 Specifically, retrospective cohort studies are the best way to ascertain both the incidence and natural history and are useful for the investigation of multiple outcomes (Grimes & Schulz, 2002). Especially regarding studying exposures, cohort studies are useful for studying rare exposures, reduce the risk of survivor bias and allow the calculation of incidence rates, relative risks and confidence intervals (Grimes & Schulz, 2002).

A significant limitation to a retrospective design, is that researchers have no choice but to work with what has been measured in the past, and as a result, may present more errors due to confounding and biases (Euser et al., 2009). It is also not possible to establish causal effects using a nonexperimental, retrospective design. The variable’s exposure has not been allocated randomly and the association found may end up being explained by other variables that also have an association with the outcome (Euser et al., 2009). Nonexperimental research can utilize observational methods and involve the collection of data with far less direct manipulation of conditions or subjects and the lack of direct manipulation or control in the nonexperimental design can cause problems in the interpretation of results because the investigator is often unable to assure that extraneous factors are properly handled (Spector, 1981). As well, variables that are related to the independent variables may account for observed differences in the dependent variable, giving the false appearance of a valid answer to a research question (Spector, 1981). Some specific sources accounting differences are history, instrument reactivity, unreliability and invalidity of instruments, differential subject loss and bias in assignments of subjects to treatments (Spector, 1981).

Retrospective studies are susceptible to selection bias, where the study group has been formed out of interest and this can only be prevented by assuring a high percentage of participation and follow-up (Euser et al., 2009). This could also include loss to follow-up bias. Another bias retrospective studies are susceptible too is information bias; either non-differential bias or differential bias (Polit & Beck, 2017). As well, relative risk cannot be calculated and there are no data on prevalence or incidence (Salkind, 2016). These studies also must cope with confounders and difficulty in selecting control groups giving them a disadvantage by only obtaining a single outcome and relying on recall or records to determine exposure status (Salkind, 2016). Bias creates an association that is not true, whereas confounding describes an association that is true, but potentially misleading. Selection and information bias cannot be completely corrected after the completion of a study; therefore, their impact must be minimized during the analysis phase of the study. As well, by showing the potential confounder is not associated with either one of the exposures and outcome in the study, confounding can be ruled out. Methods to improve comparability include restriction and matching in the design phase and stratification, regression and propensity score in the data analysis phase.

Another limitation of retrospective studies is that the development and implementation of reliable codes can be time-consuming and expensive and often require huge data sets to achieve representative samples (Brandes, 2008). Lastly, the recollection of data on the independent variables is usually less accurate than the measurement of data occurring in the same period of time (Polit & Beck, 2017) and observational research can suffer from many threats to validity, requiring careful design interpretation (Hartung & Touchette, 2009).

Nonexperimental vs Experimental Research

 Although an RCT is considered the “gold standard” for producing reliable research (Reio, 2016) and there are several limitations to nonexperimental studies, research is not perfect and experimental research is not always the right choice when determining a study methodology. Experimental research is uniquely designed to address questions related to whether an intervention causes improved outcomes, and is time-consuming, expensive and restrictive (Cook & Cook, 2008). Observational research, however, is designed to answer different questions and is sometimes the only way researchers can explore certain questions. Retrospective cohort studies may be a more appropriate approach to studying unethical situations, rare conditions, or, if little is known about how the problem develops over time. As well, well-respected researchers note that nonexperimental research could be thought of as being more important than experimental research because we might not understand the links among variables that are not amenable to experimentation without it (Reio, 2016).

Conclusion

 In research nurses learn to remain skeptical, but not inflexible as consumers of research. We are introduced to nursing literature and several research methods to not only understand the strengths and weaknesses of each design, but also to support the critique of published nursing research. Nonexperimental research is a valuable method that allows researchers to manipulate variables that normally wouldn’t be able to be manipulated due to ethical considerations. As well, nonexperimental research guides hypothesis building and directs researchers to other questions that needs to be answered. Retrospective cohort studies form a suitable study design to assess associations between multiple exposures and multiple outcomes using existing data. Retrospective cohort studies have higher accuracy and higher efficiency as their respective main advantages (Euser et al., 2009). Although there are limitations to both nonexperimental designs and retrospective cohort study designs, one can see the benefit these studies have to offer and the future they have in nursing and other social science disciplines.

References

• Brandes, U. (2008). Observational Research. Design Dictionary, 275-279. doi: 10.1007/978-3-7643-8140-0_182
• Cook, B.G. & Cook, L. (2008). Nonexperimental quantitative research and its role in guiding instruction. Intervention in School and Clinic, Vol. 44 No. 2, pp. 98-104.
• Doll, R. (2001). Cohort studies: History of the method II. Retrospective cohort studies. Sozial- Und Präventivmedizin, 46(3), 152-160.
• Euser, A.M., Zoccali, C., Jager, K.J., & Dekker, F.W. (2009). Cohort studies: prospective versus retrospective. Nephron Clinical Practice. 113(3):c214-7. doi: 10.1159/000235241
• Grimes, D.A., & Schulz, K.F. (2002). Cohort studies: marching towards outcomes. Lancet, 359(9303), 341. doi: 10.1016/S0140-6736(02)07500-1
• Hartung, D.M., & Touchette, D. (2009). Overview of clinical research design. American Journal of Health-System Pharmacy, 66(4), 398-408. doi: 10.2146/ajhp080300
• Polit, D.F. & Beck, C.T. (2017). Nursing research: Generating and assessing evidence for nursing practice (10th ed.). Philadelphia, PA: Wolters Kluwer Health.
• Reio, Thomas G, Jr. (2016). Nonexperimental research: Strengths, weaknesses and issues of precision. European Journal of Training and Development, 40(8), 676-690. doi: 10.1108/EJTD-07-2015-0058
• Salkind, N. (2016). Encyclopedia of research design. Los Angeles, California; London: SAGE. doi: 10.4135/9781412961288
• Spector, P. E. (1981). Quantitative Applications in the Social Sciences: Research designs. Newbury Park, CA: SAGE Publications, Inc. doi: 10.4135/9781412985673
• Trochim, W.M.K. (2020). Positivism & Post-positivism. Sydney, Australia: Conjoint. Retrieved from https://socialresearchmethods.net/kb/positivism-and-post-positivism/