Administrative Health Care Annotated Bibliography

Searching: SocINDEX with Full Text Choose Databases

Basic Search Advanced Search Search History

Search AN 94723399 Clear

ImagesAuthors:

Source:

Document Type:

Subject Terms:

Author-Supplied Keywords:

Abstract:

Author Affiliations:

ISSN:

DOI:

Accession Number:

Images:

Contents

OBSERVATIONAL RESEARCH

EXPERIMENTAL RESEARCH

QUASI- EXPERIMENTAL RESEARCH

MAKING CAUSAL INFERENCES FROM RESEARCH

CONCLUSIONS

Notes

Footnotes

REFERENCES

Research Designs and Making Causal Inferences From Health Care Studies.

Flannelly, KevinJ. (AUTHOR) [email protected] Jankowski, KatherineR. B. (AUTHOR)

Journal of Health Care Chaplaincy. 2014, Vol. 20 Issue 1, p25-38. 14p.

Article

*Medical care research Research methodology

causality chaplaincy epidemiologic studies research research design

This article summarizes the major types of research designs used in healthcare research, including experimental, quasi-experimental, and observational studies. Observational studies are divided into survey studies (descriptive and correlational studies), case-studies and analytic studies, the last of which are commonly used in epidemiology: case-control, retrospective cohort, and prospective cohort studies. Similarities and differences among the research designs are described and the relative strength of evidence they provide is discussed. Emphasis is placed on five criteria for drawing causal inferences that are derived from the writings of the philosopher John Stuart Mill, especially his methods or canons. The application of the criteria to experimentation is explained. Particular attention is given to the degree to which different designs meet the five criteria for making causal inferences. Examples of specific studies that have used various designs in chaplaincy research are provided. [ABSTRACT FROM AUTHOR]

Copyright of Journal of Health Care Chaplaincy is the property of Taylor & Francis Ltd and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Journal of Health Care Chaplaincy Center for Psychosocial Research, Massapequa, New York, USA

0885-4726

10.1080/08854726.2014.871909

94723399

Choose Language Translate

Research Designs and Making Causal Inferences From Health Care Studies.

This article summarizes the major types of research designs used in healthcare research, including experimental, quasi- experimental, and observational studies. Observational studies are divided into survey studies (descriptive and correlational studies), case-studies and analytic studies, the last of which are commonly used in epidemiology: case-control, retrospective cohort, and prospective cohort studies. Similarities and differences among the research designs are described and the relative strength of evidence they provide is discussed. Emphasis is placed on five criteria for drawing causal inferences that are derived from the writings of the philosopher John Stuart Mill, especially his methods or canons. The application of the criteria to experimentation is explained. Particular attention is given to the degree to which different designs meet the five criteria for making causal inferences. Examples of specific studies that have used various designs in chaplaincy research are provided.

Keywords: chaplaincy; epidemiologic studies; research; research design; causality

Traditionally, healthcare research has been divided into two categories: observational research and experimental research, or simply experimentation (Dawson-Saunders & Trapp, [11]; Mausner & Kramer, [28]). Observational research is a broad category that includes several different types of research designs, including surveys, case studies, and specific types of epidemiological study designs that will be discussed later (Dawson-Saunders & Trapp, [11]). A third type of research design is also recognized in health care and the social sciences (Cook & Campbell, [ 9]; Kleinbaum, Kupper, & Morgenstern, [25]; Macnee & McCabe, [26]), although it is ignored in many textbooks on health research: that is, quasi- experimental research (see Table 1).

TABLE 1 Research Designs in Descending Order of Their Strength of Evidence

Experimental Research (Experiments–Randomized Control Trials) Quasi-Experimental Research  Observational Research  Analytic Studies   Prospective Cohort   Retrospective Cohort   Case-Control  Survey Studies   Correlational (cross-sectional and longitudinal)   Descriptive (cross-sectional)  Case Studies

OBSERVATIONAL RESEARCH Observational research provides a fertile ground for thought. This type of research often yields evidence that supports questioning commonly accepted beliefs, and it can provide new insights and new ways of thinking about causes and effects. It is exceptionally helpful in the development of new theory and study of new fields of inquiry. Observational research has the potential to be done quickly, with uncomplicated designs, and minimal monetary investment.

Survey studies collect information, or data, from individuals using questionnaires or face-to- face interviews. Interviews are typically used when it is important to delve more deeply into issues or individual experiences than can be done using standard questionnaires (Dawson- Saunders & Trapp, [11]). The purpose of the survey dictates the kinds of questions it asks. Social surveys usually collect information about people's attitudes and opinions about social issues, whereas health surveys collect information about height, weight, blood pressure, symptoms of disease, and so forth. Almost all survey studies collect information about the attributes or characteristics of the individuals that they survey, such as age, gender, and marital status. Naturally, these characteristics vary from person to person. A survey of U.S. adults, for example, may question people anywhere from 18 to 100 years of age. The same participants may be males or females, and they may be married, or unmarried. In scientific language, these attributes are called variables because the attributes vary along some dimension. Indeed, the scientific term for anything that varies along a dimension is a variable, including attitudes, opinions, and measures of health and health outcomes.

The first modern social survey study was conducted by interviewing household members in London during the 1880s. The survey's results appeared in the 1889 book, Life and Labour of the People of London, and a series of books that followed (Marsden & Wright, [27]). Social surveys have been ubiquitous in the United States since the 1950s, and U.S. health surveys have steadily increased since then, as well. Today, survey research is the most commonly used research method in the social sciences (except psychology), and it is widely used in the health sciences.

It is not surprising, therefore, that surveys are the most common type of study method used in chaplaincy research (Galek, Flannelly, Jankowski, & Handzo, [14]). During the last decade (2000–2009), survey studies in chaplaincy have explored a number of topics, asking: patients about their satisfaction with chaplaincy care; hospital administrators about the roles and functions of chaplains in their institutions; and chaplains about their interventions and the spiritual needs of their patients (Galek et al., [14]). Many survey studies related to chaplaincy fall into the category of descriptive studies, because they simply describe the attitudes, behaviors, health outcomes, and so forth of the people surveyed.

The Journal of Health Care Chaplaincy (JHCC) published a number of survey studies in the past few years that we think are good examples of descriptive studies. One is a survey that asked chaplains at several prominent U.S. hospitals about their access to medical records (Goldstein, Marin, & Umpierre, [17]). Another analyzed the survey responses of over 200 chaplains to questions about the challenges, rewards, and frustrations of working in the U.S. Veterans Health Administration (Beder & Yan, [ 4]). Large-scale descriptive studies have been used in epidemiology to measure the incidence and prevalence of diseases, other health problems, and behaviors in a population (Kleinbaum et al., [25]). Descriptive studies are mainly used in epidemiology when little is known about the occurrence or etiology (i.e., cause or origin) of a disease. Incidence and prevalence are different measures of the rate of occurrence or presence of diseases, among others, which will be discussed in a later article on research methodology.

The second major category of survey studies is the correlational study, which measures the correlations or relationships among different variables. Health researchers often look at the correlations between variables measuring personal characteristics and experiences and variables measuring health outcomes to see if the former are related to health outcomes. The first statistical procedure to measure correlations between variables was developed by Charles Darwin's cousin Francis Galton to measure inherited similarities in physical attributes (Galton, [15]). Galton's colleague, Karl Pearson developed the mathematical formula for the correlation coefficient, which is still used today (Dawson-Saunders & Trapp, [11]). The correlation coefficient measures the strength and direction of association between two variables.

When descriptive studies and correlational studies are conducted by surveying people at a certain point in time, they are called cross-sectional studies because their measures are recorded at a cross-section in time (Kleinbaum et al., [25]). As cross-sectional studies measure all of the study's variables at the same time they can only tell us how things are and what is occurring at the time the survey was conducted. Descriptive studies and correlational studies are often lumped together as cross-sectional studies in the medical literature, although correlational studies typically provide more information about relationships between variables.

JHCC has published several correlational studies in the past few years. A recent pilot study of spirituality and anxiety in palliative care patients is a particularly good example because it analyzed the collected data using a statistical procedure called correlation (Gaudette & Jankowski, [16]). Correlational analysis, which is the basis for the term correlational study, measures the degree that two things are related to, or associated with, each other. In this study, correlations were used to measure the degree to which: (a) anxiety was related to beliefs about God, and (b) anxiety was related to spiritual practices, such as meditation. The study also made limited use of a related statistical procedure developed by Galton and Pearson called regression analysis. An excellent example of the capacity of correlational studies to shed light on complex issues is a large-scale study of Swiss patients that used a regression analysis to examine the extent to which patient satisfaction with chaplaincy care was influenced by various factors, such as patient age, gender, religion, health status, and hospital length of stay (Winter-Pfandler & Morgenthaler, [38]). Correlation, regression, and other measures of association will be explained in later articles.

Analytic studies are a third category of observational studies that are employed in epidemiological research (Katz, [21]; Kleinbaum et al., [25]; Mausner & Kramer, [28]). The primary purposes of analytic studies are to identify risk factors associated with a disease and to estimate the degree to which these factors influence a disease. The term "risk factor" is used in epidemiology instead of "cause" to refer to a variable that is believed to contribute to the development of a disease (Kleinbaum et al.). The three major research designs of analytic studies are the case-control study, the retrospective cohort study, and the prospective cohort study (Katz; Kelsey, Thompson, & Evans, [22]; Kleinbaum et al.).

A case-control study assembles a group of people who have a disease, or other health outcome (the cases), and a group of similar people who do not have the disease (the controls), and compares the life histories of the individuals in the two groups to try to discover why the disease developed in the cases but not the controls (Dawson-Saunders & Trapp, [11]; Katz, [21]). The first case-control study in the United States appears to have been the investigation of an outbreak of scarlet fever in Flint Michigan in 1924. The cause of the outbreak was determined, surprisingly, to be ice cream (Morabia, [32]). Many case-control studies since then have attempted to trace the connection between chronic diseases and exposure to environmental risk factors that occurred years ago, which is a much more difficult task (Archer, [ 1]; Mezei & Kheifets, [30]; Teschke et al., [37]).

Cohort studies compare groups of people to determine what may have caused a disease or other health outcome. Whereas study participants in case-control studies are assigned to groups on the basis of a health outcome, such as lung cancer, participants in cohort studies are assigned to groups based on some factor that is suspected of being a precursor, or risk factor, for a health outcome. The cohort is simply a group of people with something in common who remain part of this studied group over time (Dawson-Saunders &Trapp, [11]). For instance, all the people who were born in a certain year, say 1995, would form the 1995 age-cohort. Commonly used cohorts in epidemiological studies are people working in particular industries or occupations (Kelsey et al., [22]). The purpose of cohort studies is to assess if past exposure (or lack of exposure) to health risks (e.g., cigarette smoking, or working with asbestos) predict some specified health outcomes—or other outcomes of interest (e.g., lung cancer) (Kelsey et al., [22]).

Like case-control studies, retrospective cohort studies look into the past to find the risks for disease, and study groups of people to look into the future to see if the outcome of interest occurs. While case-control studies and retrospective cohort studies try to determine how past situations and events affected current and future health outcomes, respectively, prospective cohort studies attempt to predict how current circumstances affect future outcomes. In prospective cohort studies, the intent is to identify groups before (or very soon after) one group is exposed to a health risk-factor. The groups are then followed into the future to see if the health outcome of interest emerges over time (Katz, [21]; Kelsey et al., [22]; Kleinbaum et al., [25]). The Framingham Heart Study, which began in 1948, was one of the first prospective studies conducted in the United States. This study, which coined the term "risk factor," examined blood pressure, cholesterol, smoking and other personal characteristics as risk factors for coronary heart disease in men (Berridge, Gorsky, & Mold, [ 6]; Kelsey et al.).

Case studies, which have a long history in medicine, form the fourth and last major category of observational research. Case studies also have a long history in chaplaincy. Anton Boisen used them as teaching tools in clinical pastoral education (as cited in Asquith, [ 2]). A case study is generally defined as an in-depth investigation of one person, institution, or social group (Merriam, [29]; Rubinson & Neutens, [35]). Case studies of patients are still used in medicine, and they can be an important first step in developing more methodologically sophisticated studies (Dawson-Saunders & Trapp, [11]; Fitchett, [13]; Kelsey et al., [22]).

Fitchett ([13]) has argued that they are an ideal form of research for chaplains. In the past three years, JHCC has published three case studies of patients, which readers should find useful. The first is the case of a woman with advanced metastatic breast cancer (Cooper, [10]), the second is the case of woman with recurrent leukemia (King, [24]), and the third is the case of a male out-patient with Parkinson's disease who was being treated for depression (Risk, [34]).

EXPERIMENTAL RESEARCH Experimental research is often considered to be the more complicated type of research even though this is not always true (Cook & Campbell, [ 9]). Direct manipulation of environmental factors often occurs in experimental research, and the interference with, and manipulation of, environmental factors requires careful thought about ethics, especially when human beings are involved as participants in the research. Experimental research can provide the best support for cause and effect explanatory relationships (Campbell & Stanley, [ 8]; Greenhalgh, [18]).

Experimental research in medicine was formalized in the mid nineteenth century through the work of Claude Bernard (Bernard, [ 5]), although medical research had been conducted for centuries prior to that (Singer, [36]). A few years before Bernard's book was published, the philosopher John Stuart Mill ([31]) published his ideas about scientific reasoning, experimental methods, and inferring causality from research. For our purposes, Mill's key proposals may be summarized as five criteria for drawing causal inferences from experimentation (Mill, himself, did not number them.). First, the presumptive cause, or causal agent, must precede the effect in time. The cause must occur before the effect. Second, the effect must occur whenever the presumptive cause is present (Mills' Method of Agreement). The effect always happens if the cause has occurred. Third, the effect must not occur when the presumptive cause is absent (Mills' Method of Difference). Fourth, the presumptive cause must be isolated from other potential causes of the effect. Fifth, to ensure that the presumptive cause is isolated from all other potential causes, it must be produced artificially, which in this situation precludes observing the natural occurrence of the causal agent.

Mill ([31]) proposed two other methods for identifying causal relationships:[ 2] The Method of Concomitant Variations and the Methods of Residues. Mill also recognized that there may be multiple causes of an effect. Current statistical methods can help identify multiple causes, but we are not going to treat this topic here.

The renowned scientist Louis Pasteur was conducting studies that employed Mill's experimental methods around the same time that Mill was writing about them (Singer, [36]). A famous experiment by Pasteur was designed to test the theory of spontaneous generation of gas. It can be viewed as a simple test of whether fermentation requires bacteria to occur. To test this, Pasteur created a solution, in which bacteria normally produced fermentation, in a sealed flask. Next, he heated the flask to kill any bacteria and other living matter already in the flask. Then, he waited for months to see if fermentation occurred when no bacteria were present (the method of difference). The method of difference confirmed that fermentation (the effect) did not occur without the presence of live bacteria (the causal agent). When Pasteur eventually unsealed the flask to expose it to bacteria (the causal agent) in the air, fermentation (the effect) occurred in the flask within hours (the method of agreement).[ 3] Mill strongly advised that researchers use both methods, in what he called the Joint Method of Agreement and Difference, and this is what Pasteur did sequentially.

Modern experimental research adheres to Mill's five criteria to make it possible to draw causal inferences from experimental results (Plutchik, [33]). The first criterion is applied in experimentation by presenting or administering an experimental treatment (Mill's presumptive cause), and then measuring its effect or outcome. Thus, the presumptive cause precedes the intended effect. In human research, the outcome of interest is usually measured before the treatment is administered, to see if it is already present to some degree (Campbell, [ 7]; Campbell & Stanley, [ 8]).

The second and third criteria are applied by using Mill's Joint Method of Agreement and Difference, simultaneously, in which some participants receive the treatment (the experimental condition) and some participants to do not receive the treatment (the control condition) (Campbell, [ 7]; Campbell & Stanley, [ 8]). The fourth criterion is followed by carefully trying to control for variations in environmental variables that may affect the outcome (called extraneous variables) other than the experimental treatment (Campbell, [ 7]; Kidder & Judd, [23]). Human and animal experimentation further controls for extraneous variables, in the form of personal differences, by randomly assigning participants to the experimental and control conditions (Edwards, [12]; Kidder & Judd).

The fifth criterion is an inherent part of experimentation in that the presence or absence of a treatment (which is somewhat oddly referred to as the independent variable) is artificially manipulated by the researcher, as is the case when the treatment is a surgical procedure (Kleinbaum et al., [25]). The level of treatment is also manipulated in many experiments, such as drug treatments that consists of different doses of a drug (Kleinbaum et al.), In such experiments, one might predict that the strength of the outcome would vary by dose, which is an example of Mill's Method of Concomitant Variations (Mill, [31]).

Adherence to these five criteria is what defines an experiment (Plutchik, [33]). In modern language, the defining characteristics of experiments are said to be: manipulation of the independent variable (treatment), randomization of participants to conditions, and control of extraneous variables (Kidder & Judd, [23]). Experiments in medicine that adhere to these principles have come to be called "randomized control trials" (RCTs) (Dawson-Saunders & Trapp, [11]; Katz, [21]).

A study by Bay and his colleagues (Bay, Beckman, Trippi, Gunderman, & Terry, [ 3]) is one of a few published experimental studies in the chaplaincy research literature. Positive health outcomes, such as positive religious coping and reduced negative religious coping were studied in cardiac patients over a short period of time. Many controls were instituted to isolate the effects of a chaplaincy intervention on cardiac patients. The chaplain's intervention followed an exact schedule and one chaplain delivered the intervention. The participants' characteristics were controlled, including the requirements that participants speak and understand English, that participants have only one of two kinds of heart surgery, and that participants agree to be randomly assigned to the treatment condition (chaplaincy) or control condition (no chaplaincy). Participants were required to attend all the chaplain's meetings and complete all the questionnaires. Increased positive religious coping and reduced negative religious coping were found in the treatment group compared to the control group. A more detailed discussion will be found on the possible conclusions drawn from these findings in a future article delving deeper into experimental design.

QUASI-EXPERIMENTAL RESEARCH The term quasi-experimental was introduced in 1963 in the context of conducting research in educational and social settings (Campbell & Stanley, [ 8]). Campbell and Stanley's book explored the problems that limit one's ability to make causal inferences from "true" experimental studies and studies that lack some of the key features of "true" experiments, a topic that was initially addressed by Campbell ([ 7]). Although some people refer to quasi- experiments as natural experiments, quasi-experimental designs are used to study the effects of many things that are not acts of nature. The term quasi-experiment or quasi- experimental design is typically used to describe studies in which the independent variable is not manipulated by the researcher and the "participants" are not randomly assigned to conditions (Kleinbaum et al., [25]). However, it also applies to research in which there is no control group, such as quality improvement research.

Although there is little available information on the extent to which quasi-experiments are used in medicine or other healthcare fields, a 2005 review of research on infectious diseases identified 73 quasi-experimental studies that were published in three journals in just two years (Harris, Lautenbach, & Perencevich, [19]). Over three-quarters of the studies did not have a control group.

A study of students in clinical pastoral education (CPE) published in JHCC is a good example of a quasi-experiment (Jankowski, Vanderwerker, Murphy, Montonye, & Ross, [20]). Student growth in chaplaincy skills was assessed for groups of students taking a shorter intensive unit or a longer, extended unit of CPE. Students completed questionnaires at the beginning and the end of each unit of CPE, which measured their pastoral skills, emotional intelligence, and self-reflection. Controlling for major demographic variables, and the two types of CPE units, students with no prior CPE and fewer years of professional ministry, were found to experience more positive change in their pastoral skills. Controlling for demographic variables, students in intensive CPE experienced greater increases in emotional intelligence and self-reflection. However, the ability to draw causal conclusions from the study is limited by the lack of a control group and the fact that students were not randomly assigned to the two types of CPE units. While differences in the demographic characteristics of the two groups were controlled for statistically, other, unmeasured characteristics of students in the intensive units could account for the findings.

MAKING CAUSAL INFERENCES FROM RESEARCH The types of research we have discussed, traditionally, have been viewed as a hierarchy with respect to the quality of evidence they provide for assessing causal inferences (Greenhalgh, [18]). Experimentation forms the top of the hierarchy because it establishes the temporal order needed to infer causation, and controls for other alternative explanations of what might have caused the observed outcome. The goal of experimentation is to reduce, if not eliminate, the possibility of some alternative explanation for the apparent effect of treatment, other than the treatment (Edwards, [12]; Katz, [21]).

If the expected outcome is observed in the experimental group and not the control group one can logically conclude that the outcome is due to the experimental treatment. The possibility that the outcome is attributable to other variables that could have affected the participants at the same time as the treatment was administered is eliminated to the degree that the experimental and control conditions are identical except for the treatment (Edwards, [12]; Kidder & Judd, [23]). Because participants are randomly assigned to conditions, or groups, the differences in outcomes between the groups cannot be due to differences in the personal characteristics of the participants in the groups because random assignment should ensure that the characteristics of the participants are, on average, comparable in the groups (Edwards, [12]; Greenhalgh, [18]; Mausner & Kramer, [28]). Although Mill thought multiple experiments, or experimental trials, in which a presumptive cause was present or absent, might be required to be able to determine causality, experiments in modern medicine and other fields often consist of a single trial during which the participants in the experimental condition receive the experimental treatment (Greenhalgh, [18]).

Quasi-experiments are lower on the hierarchy for several reasons. To the degree the temporal order of the presumptive cause and the observed outcome can be established, the temporal order required for inferring causality is not a problem. However, the nature of quasi- experiments poses several threats to the ability to infer cause-effects relationships from them. Because the researcher does not manipulate the independent variable (or treatment), the observed outcome could be the result of something that occurred about the same time as the presumptive cause. This possibility is not easily dismissed because quasi-experiments, unlike experiments, do not control for extraneous variables. The lack of random assignment of participants to experimental and control poses another threat since, without random assignment, the differential outcomes could be due to differences between participants in the conditions. Some quasi-experiments lack any group with which the experimental effects can be compared.

Cross-sectional surveys and case-studies are traditionally placed at the bottom of the hierarchy. The problem with cross-sectional surveys is that they cannot determine the temporal order of events that is needed to infer a cause-effect relationship between two variables, because the personal characteristics of survey participants and their health outcomes are measured at the same point in time (Kelsey et al., [22]; Mausner & Kramer, [28]). This inability of survey studies to establish the temporal relationship required to draw causal inferences can be corrected by conducting surveys of the same individuals at two or more points in time. This type of survey research is called a longitudinal study rather than a cross-sectional study.

Case-studies are placed at the low end of the hierarchy for two reasons: their inability to draw causal inferences and their lack of generalizability. First, case-studies of patients are very rarely, if ever, conducted before the patient has a disease or other health problem, so case- studies do not provide a time-line by which to determine whether a cause-effect relationship exists. Second, it is often difficult to generalize the circumstances of a particular person to other people. However, case-studies may be used to study the effects of treatment, regardless of the disease etiology, a point that is often lost in the medical literature. Moreover, as previously mentioned, case studies can be helpful in the development of more sophisticated studies of a topic (Dawson-Saunders & Trapp, [11]; Fitchett, [13]; Kelsey et al., [22]).

Analytical studies fall toward the middle of hierarchy. The case-control study is ranked just above cross-sectional surveys and case studies in the hierarchy because their design permits them to differentiate between cause and effect by demonstrating that exposure to some risk factor occurred prior to a disease. However, as the study may be conducted years after the exposure, documenting the temporal relationship between risk factors and health outcomes may be problematic (Kelsey et al., [22]; Mausner & Kramer, [28]). The required documentation of exposure to the risk factor, and/or the onset of the disease may be inadequate because of poor recoding keeping or poor memories of the events (Kelsey et al., [22]; Mausner & Kramer).

Even when the temporal relationship can be properly established, the causal link between exposure and disease is not conclusive. Case-control studies assume that the cases and controls are members of the same population, who only differ from each other with respect to exposure to the risk factor. If this is not true, the putative risk factor may not be the cause of the outcome, since the outcome could be the result of something else that is similar to the cases but different between the cases and controls (Kleinbaum et al., [25]). This was the basic criticism of the case-control studies that identified cigarette smoking as a risk factor for lung cancer. The alternative explanation was that people who smoke and people who develop lung cancer share some common characteristic, other than smoking, that was the cause of the lung cancer (Dawson-Saunders & Trapp, [11]). Despite their weaknesses, case- control studies are useful for studying rare diseases and identifying possible risk factors (Kelsey et al., [22]).

Retrospective cohort-studies, many of which have examined health problems associated with workplace risk factors (Kelsey et al., [22]), form the next rung of the hierarchy (Greenhalgh, [18]). Generally, cohort studies provide stronger evidence for cause and effect between risks and outcomes than case-control studies because the selection of cases is done before the adverse health outcome is observed (Dawson-Saunders & Trapp, [11]; Kelsey et al.). Hence, at least in theory, they can establish the temporal order required to draw causal inferences. However, because they attempt to look back in time to establish the temporal relationship needed to determine the cause-effect relationship between a risk factors and disease, they face some of the same problems as case-control studies. For example, they may not be able to (a) document the temporal relationship between risk factors and outcomes, or (b) identify and measure variables that may be risk factors for the disease apart from the supposed risk factor. Finally, the selection of the appropriate controls can be difficult, since the controls should be representative of the general population overall, and with respective to exposure to the risk factor (Kelsey et al., [22]; Mausner & Kramer, [28]).

Prospective cohort studies are ranked above retrospective cohort-studies for a number of reasons, based on the fact that they are truly longitudinal, and observe individuals across time rather than trying to recreate their life histories (Mausner & Kramer, [28]). The first reason, and the major advantage of prospective studies, is that the timing of exposure to risk factors and disease onset can be clearly established for individuals in the study (Mausner & Kramer). Second, they also are less likely to miss shared variables among the cases other than the risk factor that might cause the outcome, although the possibility of alternative causal agents always exist in observational studies (Dawson-Saunders & Trapp, [11]). Once such shared variables are identified, their effects can be controlled statistically; therefore, the effects of the hypothesized risk factor can be isolated from their effects on the health outcome. Third, prospective cohort studies can assess multiple health outcomes because they start at the time of exposure to the risk factor and follow individuals into the future, during which time more than one effect of the exposure may emerge (Kelsey et al., [22]; Mausner & Kramer, [28]).

CONCLUSIONS Healthcare research efforts begin humbly, through diligent observations of healthcare provision, utilization, and/or outcomes. These observations become increasingly more complex if the researcher intends to identify cause and effect relationships, whether to reduce or prevent disease or to improve education and practice. Attention must be paid to many variables, such as participants' characteristics, the setting of the research, and the timing of events. Time is crucial to establish cause and effect, and research findings should always be evaluated with attention to the criteria for drawing causal inferences from experimentation and other types of research. Knowledge of research methods and reasoning will decrease the likelihood that researchers will fall into the error of making cause and effect statements that are not justified by the methodology employed in a study.

Notes Footnotes 1 Notes

2 Collectively, Mill's five methods have been called Mill's Canons (Katz, [21]).

3 The fact that fermentation did not occur until air-borne bacteria entered the flask was taken as evidence that spontaneous generation had not occurred, which so undermined the theory of spontaneous generation that is was no longer taken seriously in scientific circles.

REFERENCES Archer, V. E. (1988). Lung cancer risks of underground miners: cohort and case-control studies. Yale Journal of Biology and Medicine, 61(3), 183–193.

Asquith, G. R. (1980). The case study method of Anton T. Boisen. Journal of Pastoral Care, 34(2), 84–94.

Bay, P. S., Beckman, D., Trippi, J., Gunderman, R., & Terry, C. (2008). The effect of pastoral care services on anxiety, depression, hope, religious coping, and religious problem solving styles: A randomized controlled study. Journal of Religion and Health, 47(1), 57–69.

4 Beder, J., & Yan, G. W. (2013). VHA Chaplains: challenges, roles, rewards, and frustrations of the work. Journal of Health Care Chaplaincy, 19(2), 54–65.

5 Bernard, C. (1865/1957). An introduction to the study of experimental medicine. New York, NY: Dover Publications.

6 Berridge, V., Gorsky, M., & Mold, A. (2011). Public health in history. Berkshire, UK: Open University Press.

7 Campbell, D. T. (1957). Factors relevant to the validity of experiments in social settingsPsychological Review, 54(4), 297–312.

8 Campbell, D. T., & Stanley, J. C. (1963). Experimental and quasi-experimental designs for research. Chicago, IL: Rand McNally

9 Cook, T. D., & Campbell, D. T. (1979). Quasi-experimentation: Design & analysis issues for field settings. Boston, MA: Houghton Mifflin.

Cooper, R. S. (2011). Case study of a chaplain's spiritual care for a patient with advanced metastatic breast cancer. Journal of Health Care Chaplaincy, 17(1–2), 19–37.

Dawson-Saunders, B., & Trapp, R. G. (1994). Basic and clinical biostatistics (nd ed.)., 2Norwalk, CT: Appleton & Lange.

Edwards, A. L. (1985). Experimental design in psychological research (Ed.), Fifth. New York, NY: Harper & Rowe.

Fitchett, G. (2011). Making our case(s). Journal of Health Care Chaplaincy, 17(1–2), 3–18.

Galek, K., Flannelly, K. J., Jankowski, K. R., & Handzo, G. F. (2011). A methodological analysis of chaplaincy research: 2000–2009. Journal of Health Care Chaplaincy, 17(3–4), 126–145.

Galton, F. (1888). Co-relations and their measurement, chiefly from anthropometrics data. Proceedings of the Royal Society of London, 45, 135–145.

Gaudette, H., & Jankowski, K. R. (2013). Spiritual coping and anxiety in palliative care patients: A pilot study. Journal of Health Care Chaplaincy, 19(4), 131–139.

Goldstein, H. R., Marin, D., & Umpierre, M. (2011). Chaplains and access to medical records. Journal of Health Care Chaplaincy, 17(3–4), 162–168.

Greenhalgh, T. (2001). How to read a paper: The basics of evidence based medicine. London: BMJ Books.

Harris, A. D., Lautenbach, E., & Perencevich, E. (2005). A systematic review of quasi- experimental study designs in the fields of infection control and antibiotic resistance. Clinical Infectious Disiseases, 41(1), 77–82.

Jankowski, K. R. B., Vanderwerker, L. C., Murphy, K. M., Montonye, M., & Ross, A. M. (2008). Change in pastoral skills, emotional intelligence, self-reflection, and social desirability across a unit of CPE. Journal of Health Care Chaplaincy, 15(2), 132–148.

Katz, D. L. (2001). Clinical epidemiology & evidence-based medicine. Thousand Oaks, CA: Sage Publications.

Kelsey, J., Thompson, W. D., & Evans, A. S. (1986). Methods in observational epidemiology. New York, NY: Oxford University Press.

Kidder, L. H., & Judd, C. M. (1959). Research methods in social relations (Ed.), Fifth. Fort Worth, TX: Holt, Rinehart & Winston.

King, S. D. (2012). Facing fears and counting blessings: A case study of a chaplain's faithful companioning a cancer patient. Journal of Health Care Chaplaincy, 18(1–2), 3–22.

Kleinbaum, D. G., Kupper, L. L., & Morgenstern, H. (1982). Epidemiolic research: Principles and quantitative methods. New York, NY: Van Nostrand Reinhold.

Macnee, C. L., & McCabe, S. (2008). Understanding nursing research (ed.), 2nd. Philadephia, PA: Lippincott, Williams & Wilkins.

Marsden, P. V., & Wright, J. D. (2010). Survey research and social science: History, current practice, and future prospects. In P. V.Marsden & J. D.Wright (Eds.), Handbook of survey research (nd ed., 2, pp. 3–25). Bingley, UK: Emerald Group Publishing.

Mausner, J., & Kramer, S. (1985). Epidemiology: An introductory text (nd ed.)., 2Philadelphia, PA: W.P. Saunders.

Merriam, A. B. (1988). Case study research in eduation: A qualitative approach. San Franciso, CA: Jossey-Bass.

Mezei, G., & Kheifets, L. (2006). Selection bias and its implications for case-control studies: a case study of magnetic field exposure and childhood leukaemia. International Journal of Epidemiology, 35(2), 397–406.

Mill, J. S. (1859). A system of logic, ratiocinative and inductive; bring a connected view of the principles of evidence and the methods of scientific investigation. New York: Harper & Brothers.

Morabia, A. (2013). Snippets from the past: Is Flint, Michigan, the birthplace of the case- control study?American Journal of Epidemiology, [Epub ahead of print].

Plutchik, R. (1968). Foundations of experimental research. New York, NY: Harper & Rowe.

Risk, J. L. (2013). Building a new life: A chaplain's theory based case study of chronic illness. Journal of Health Care Chaplaincy, 19(3), 81–98.

Rubinson, L., & Neutens, J. J. (1987). Research techniques in the health sciences. New York, NY: Macmillan.

Result List Refine Search 1 of 1

1,2

1,2

1

2

Find Similar Results using SmartText Searching.

Detailed Record

HTML Full Text

PDF Full Text (87KB)

Full Text

Cited References (38)

Tools

Google Drive

Add to folder

Print

E-mail

Save

Cite

Export

Create Note

Permalink

Listen

Translate

!Listen"

University Library

New Search Publications Subject Terms Cited References Images More

Sign In Folder Preferences Languages Help Exit

NY: Macmillan.

Singer, C. (1950). A history of biology (, Reviseded.). New York, NY: Henry Schuman.

Teschke, K., Olshan, A. F., Daniels, J. L., De Roos, A. J., Parks, C. G., Schulz, M., Vaughn, T. L. (2002). Occupational exposure assessment in case-control studies: Opportunities for improvement. Occupational and Environmental Medicine, 59(9), 575–934.

Winter-Pfandler, U., & Morgenthaler, C. (2011). Patients' satisfaction with health care chaplaincy and affecting factors: An exploratory study in the German part of Switzerland. Journal of Health Care Chaplaincy, 17(3–4), 146–161.

~~~~~~~~

By KevinJ. Flannelly and KatherineR. B. Jankowski

Reported by Author; Author

Copyright of Journal of Health Care Chaplaincy is the property of Taylor & Francis Ltd and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.

Result List Refine Search 1 of 1

Top of Page EBSCO Connect Privacy Policy A/B Testing Terms of Use Copyright Cookie Policy

© 2021 EBSCO Industries, Inc. All rights reserved.