ASSIGNMENT Essay Example

  • Category:
  • Document type:
  • Level:
  • Page:
  • Words:

Research Methods

Research Methods


Currently health professionals broadly read, contribute and take part in medical research more than before. The world of research has turned out to be a vital element of the training that is received by the health personnel. It is significant that each health professional has fundamental knowledge of methods of research as well as vocabulary relevant to research. Analytic investigations examine a relationship between an issue of concern and other variables, and potential causative factors are evaluated (Ogino et al, 2012). If the investigation starts with an exposure (like use of oral contraceptive) and follows females for some years to determine outcomes (like ovarian cancer) then the study is considered a cohort study. On the contrary if the analytical investigation starts with an outcome (ovarian cancer) and traces the exposure, like oral contraceptive use, then this study is considered a case-control study. A cross-sectional study identifies both outcomes and exposure at one instance (Ogino et al, 2012). This paper will further discuss the types of research studies and further discuss the types of biases in epidemiological studies.

Observational Studies

Also known as non-experimental studies, observational studies observe anything that occurs naturally over time (Liamputtong, 2010). If the investigation has a control or comparison cluster it is considered analytical, if not then it is considered descriptive (Liamputtong, 2010). Analytical studies look at a relationship between an issue of concern and other variables, as well as potential causative factors that are studied (Liamputtong, 2010). Some of the analytical studies are discussed below.

Cohort Study

A cohort study follows 2 or more groups ahead from exposure to result (Thomas & Polgar, 2008). This kind of investigation can be carried out through going forward in time (prospective or future cohort study) or on the other hand, through going reverse in time to entail the cohorts and follow them up until the current (retrospective cohort study) (Thomas & Polgar, 2008). A cohort investigation is the best method of identifying occurrence and a disease’s natural history, and may be used in examination of multiple outcomes following a single exposure, particularly when there is rare exposure (Thomas & Polgar, 2008). Nonetheless, this kind of investigation is less functional for examination of uncommon occurrences or the ones that take long duration to develop (Liamputtong, 2010). For instance, a design of cohort study would be helpful for investigating radioactive exposure’s effects on children after a radioactive accident such as the Chernobyl disaster (Liamputtong, 2010). The significant characteristic of every cohort study is that it tracks individuals from exposure to result (Williams, 2008). Researchers carrying out this study must, hence, go ahead in time from the current or go reverse in time to select their cohort (Williams, 2008). Therefore it may be a retrospective of prospective cohort study. Retrospective is also known as a trohoc study (Williams, 2008).

A cohort investigation moves from exposure to outcome, either way (Liamputtong, 2010). For instance, the Chernoby disaster’s prospective cohort study would have began at the point of disaster and tracked the various exposure outcomes in due course in children that were nearly alike apart from history of the exposure (Berg & Latin, 2008). Correspondingly, in a study of retrospective cohort of the disaster the investigator would search for medical records of disaster’s time to establish the status of exposure of patients in a hospital and then search for the death registry of the nation to establish the basis of death in the unexposed and exposed (Ogino et al, 2012).

There are some significant weaknesses and strengths associated with cohort study. Since an exposure is recognized at the onset, there can be an assumption that the exposure came before the outcome. In case control study, recall bias is of less concern (Berg & Latin, 2008). The cohort study facilitates calculation of accurate occurrence rates, attributed risks and relative risks (Berg & Latin, 2008). Nevertheless, for rare events study or occurrences that require years to come up, this kind of study design may be slow to generate outcomes and hence prohibitively costly (Ogino et al, 2012). Follow up loss is a great challenge. Without standard contact with participants of the study, it might not be easy to approximate if, and when, an individual developed the illness of concern (Berg & Latin, 2008). For this problem to be overcome the researcher may apply relative risk (also referred to as risk ratio). A risk ratio is a comparative comparison that provides information regarding the association’s strength between the disease and exposure, and is most valuable for etiological study (Thomas & Polgar, 2008).

Case-Control Study

As stated above, a cohort method’s drawback to study disease causes with low occurrence is that lengthy and large investigations might be needed to provide sufficient statistical power (Liamputtong, 2010). Case-control design is an alternative that avoids this hardship (Liamputtong, 2010). In a study of case-control patients who developed an illness are detected and their precedent exposure to alleged aetiological causes is evaluated against that of controls, which do not have the illness (Liamputtong, 2010). This facilitates approximation of odds ratios although not of attributable risk.

Studies of case control are particularly helpful for results that take an extended period to develop, like cancer and cardiovascular disease (Liamputtong, 2010). These studies frequently demand less effort, money, and time compared to cohort studies. Control need to be similar to situations in every significant respect other than for having the result in question (Liamputtong, 2010). Since studies of case-control do not have denominators, researchers are not able to calculate rates of incidence, attributable risks, or relative risks (Somerville et al, 2012). Rather, odds ration are the association measures used, when there is uncommon outcome (e.g. most cancers) the odds ratio offers a good substitute for the actual relative risk (Liamputtong, 2010).

Cross Sectional Study

Liamputtong, 2010). (Liamputtong, 2010). The problem’s prevalence, instead of the incidence, is documented in a cross-sectional review, and each link needs to be interpreted with caution (Liamputtong, 2010). Additionally, it can be used to investigate the relation between variables, for instance between lung cancer and hemoptysis (Liamputtong, 2010). Such information may be helpful in assessment of health status as well as a population’s needs (Liamputtong, 2010). It offers a picture of a population’s health experience at a specific time, like the national survey of family health (Also referred to as prevalence survey, a cross-sectional study measures a symptom’s prevalence, a symptom’s determinants, or both, in a populace at one position in time or over a small stage of time

Studies of cross sectional are carried out to inspect the absence or presence of illness and the absence or presence of an exposure at specific period (Rothman et al, 2008). Because both outcome and exposure are determined at similar period, the temporal relation involving the two may not be clear, and therefore the causation from these types of studies cannot be proven (Rothman et al, 2008). For instance, an assumption that a cross sectional investigation discovers obesity to be more frequent among females with arthritis instead of those without arthritis. Did extra load of weight on joints result in arthritis, or did women who have arthritis unwillingly become inactive and subsequently obese? This kind of question is not answerable in a cross sectional investigation. Other cross sectional studies applications lie in health care planning (Liamputtong, 2010). For instance, an occupational physician arranging for a programme of coronary prevention may desire to know the commonness of various risk factors within the labour force under his management so that his intervention could be tailored accordingly (Rothman et al, 2008).

Causality’s Assessment

The aim of these investigations is to establish the association between a disease and exposure with precision and validity (Somerville et al, 2012). One of the very difficult assignments in epidemiological investigation is to assess if associations involving disease and exposure derived from studies of observational epidemiology are of causal character or not (Somerville et al, 2012). It is imperative that findings from studies of analytical epidemiology are critically examined prior to making any causality judgment (Liamputtong, 2010). Additionally, outcomes from one distinct epidemiological research only exceptionally offer conclusive proof of a causal association between disease and exposure. One aspect concerned with the relationship between disease and exposure is the association’s strength. A strongly increased relative risk is most probable to reflect a causal relationship compared to a moderately or slightly elevated risk (Liamputtong, 2010).

Consistency of outcomes across studies carried out with various methodologies and in various environments, is another aspect. Specificity is another aspect that exposure brings about a specific illness, e.g. the surveillance that smoking cigarette is linked to the respiratory tract’s squamous cell carcinoma. A significant condition is events’ sequence: the potentially causal factors ought to come before the effect i.e. the disease. Biological gradient or dose-response relationship is another feature. For instance, massive sunlight exposure is more possible to bring about melanoma in vulnerable people compared to moderate or little sunlight (Liamputtong, 2010).

Sources of Bias in Epidemiological Research

Bias is the systematic error within the study’s protocol that results in a false relationship between disease and exposure (Somerville et al, 2012). Hence it is basically an error that the investigator commits in the conduct or design of a research. One of the kinds of biases is selection biases. This is a flaw as a result of systematic dissimilarities in features between the people selected for research and the ones that are not (Liamputtong, 2010). The other kind of biases is information/observation bias. This is an error in measuring outcome or exposure data that leads in information’s diverse quality (accuracy) between comparison clusters (Liamputtong, 2010).

Selection Biases’ Sources

There are various sources related to selection bias. Nonresponse bias takes place because people who do not give response to a call to take part in studies of research are normally different from the ones that respond (Liamputtong, 2010). For instance, participants seem to have lifestyle habits that are healthier, with lower mortality and smoking rates. Due to this they seem to be diverse from the target populace. Control selection bias would take place if the researcher does not apply similar criteria to choose controls and cases and these measures were associated with status of exposure (Somerville et al, 2012). This may be avoided through remembering that cases resemble controls in each way other than disease. In the diarrheal deaths’ example, if case arose from lower status of socio-economy and controls arose from higher status of socio-economy there would be bias in selection, since the status of socio-economy is by itself a determinant of rates of infection and to health services access (Somerville et al, 2012).

Self selection bias is another source of selection bias and can occur as a result of agreement or refusal to take part in what is connected to exposure or disease (Liamputtong, 2010). For instance, if the comparison cluster chosen for the investigation are principally volunteers, they might be more understand about issues of health and have very diverse habits from the group of exposure (Liamputtong, 2010). In a cohort study selection bias takes place when the selection of unexposed and exposed people is connected with developing the result of concern (Liamputtong, 2010). For instance children who develop heart diseases in their adulthood might have more issues like poor habits of eating than simply not exercising as the cause.

Exclusion bias takes place when in particular situations epidemiologic research exclude respondents to avoid confounding (Liamputtong, 2010). Confounding factor is the alteration of the approximated outcome of an outcome of an exposure, brought about by the availability of an extraneous effect linked to the outcome and exposure but not a transitional step in the causal conduit between outcome and exposure (Liamputtong, 2010). If the criteria of exclusion are different for controls and cases or different for nonexposed and exposed, an exclusion bias might be introduced. In 1974 investigators published the outcomes of a hospital-based case-control study where cancer of the breast was connected with reserpine’s use, a common treatment for hypertension at that period (Somerville et al, 2012). To make controls resemble the population of study more closely, women with medical conditions leading to reserpine’s prescribed use were not included in the control cluster. On the other hand, similar criteria of exclusion were not applied for the cases. Hence, association’s overestimation between reserpine and breast cancer was found (Somerville et al, 2012).

The study was replicated in 1985 and the investigators carried out 2 dissimilar analyses of their information. Initially, they involved every woman in their analysis. The data was then reanalyzed following exclusion of controls with cardiovascular condition. Their results indicated no relationship when every woman was included; however, their data indicated a powerful association when women with cardiovascular conditions from the control cluster were excluded (Somerville et al, 2012).

Information Biases’ Sources

When information acquired from research participants is systematically incorrect concerning the exposure or disease under investigation, information bias can occur (Liamputtong, 2010). Every time the information accuracy concerning exposure is dissimilar in controls and cases, a differential data bias takes place (Liamputtong, 2010). Such bias may possibly lead to an under- or overestimate of the actual relationship with regards to the circumstances. For instance, if there is underrecognition of exposure in cases, there will be an occurrence of a positive association’s underestimation, but if there is underrecognition of exposure in controls, there will be a positive association’s overestimation (Liamputtong, 2010).

There are various kinds of information bias that will be discussed in this paper. Frequently, medical records are used in epidemiologic research to abstract data (Liamputtong, 2010). Conversely, careful consideration needs to be provided to the data’s quality since medical records are designed for the purposes of diagnosis and treatment, not for investigation (Liamputtong, 2010). For instance, data are normally more inclusive once there is a clear establishment of diagnosis. Better standard on diagnosis is normally realized in patients who have chronic illness. Hence more inclusive information regarding exposures may be more commonly established in patients with chronic illness.

Biases also take place during interviewing. Interviewers seem to be more precise when they interview cases unlike when they interview controls (Ogino et al, 2012). When interviewers know that a participant is a ‘case,’ they will seem to forcefully find exposure, and classify indeterminate or vague responses as signifying exposure. This more accurate questioning enhances the data’s quality from cases. On the other hand, because the accurate questioning appears not to take place with a control participant, the data concerning exposure in controls will seem to be underestimated (Liamputtong, 2010). Hence the association’s overestimation will be the outcome.

Recall bias is another common type of information bias. Cases (people recognised as having the illness under investigation) seem to better recall previous exposures unlike controls (Liamputtong, 2010). For instance, women with malformed babies will remember better whichever event in the course of pregnancy unlike mothers of babies without malformations. It appears feasible that this is correct since people with a condition are more interested about recalling possible causes (Ogino et al, 2012). Hence recall bias will appear to overestimate the relationship of exposure with outcome to a risk factor.

Reporting bias is also a source of information bias. This type of bias happens when ‘cases’ stress the significance of exposures that they consider to be essential (Liamputtong, 2010). Occasionally this type of bias may be connected with occupational exposures that a patient wants to highlight because of an employee’s compensation or whichever benefit hence generating the association’s overestimation under investigation as it does not happen in controls (Liamputtong, 2010). In cross sectional investigations, bias may come up due to selection out of or into the population of study. For instance, in a hospital research, patients that stay for a shorter time in hospital have fewer likelihood of being involved in a cross-sectional study (Ogino et al, 2012).

Misclassification is the most frequent type of bias in epidemiological survey. It shows there is a flaw in the disease or exposure’s classification (Somerville et al, 2012). For instance, in a study of case control gauging exposure from childhood might misclassify subjects because of recall bias. Differential misclassification refers to flaws on one axis (disease or exposure) that are connected with the other axis (disease or exposure) (Somerville et al, 2012). For instance, if there is a flaw in classification of exposure, it is less or more expected to happen for a diseased person rather than a nondiseased person. It may decrease or increase the association’s strength between exposure and outcome. An example can be the consequence of childhood disorder in conduct on possibility of alcohol use disorders in adulthood. A characteristic aspect of disorder of conduct is lying; unexposed individuals are less likely to report alcohol use compared to individuals that are exposed.

Non-differential misclassification on the other hand implies the flaws on one axis, unconnected to the other axis (Ogino et al, 2012). For instance, if classification in exposure has an error, it happens with equal possibility among nondiseased and diseased people. It decreases association’s strength between exposure and outcome (Ogino et al, 2012). An example can be the consequence of birth weight on malnutrition risk. The scale used to evaluate weight is 500grams off. For both controls and cases, it is 500grams off.


This paper has discussed various types of analytical studies. It has been discussed that these studies are most helpful for examining a hypothesized link between adverse health outcomes and human exposure. The paper has also discussed and provided examples that there are various potential sources for flaws that can occur in systematic alterations of study outcomes. These alterations are an issue particularly when the investigator is approximating the relationship between a health issue and a risk factor. Whether a protective factor or a risk factor goes without detection, or a condition or behaviour is misrecognized as a protective or risk factor, the public can have serious implications in terms of behaviour that will eventually affect their health.


Ogino, S., King, E.E., Beck, A.H., Sherman, M.E., Milner, D.A. & Giovannucci, E. (2012). Interdisciplinary education to integrate pathology and epidemiology: Towards molecular and population-level health science. Am J Epidemiol, 176, 659-67.

Thomas, S. A., & Polgar, S. (2008). Introduction to research in the health sciences. London: Churchill Livingstone.

Williams, G. (2008). Introduction to research in the health sciences (5th edition). British Dental Journal, 204, 8, 467-467.

Berg, K. E., & Latin, R. W. (2008). Essentials of research methods in health, physical education, exercise science, and recreation. Philadelphia: Wolters Kluwer Health/Lippincott Williams & Wilkins.

Liamputtong, P. (2010). Research methods in health: Foundations for evidence-based practice. South Melbourne, Vic: Oxford University Press.

Somerville, M., Kumaran, K., & Anderson, R. (2012). Public health and epidemiology at a glance. Oxford: Wiley-Blackwell.

Rothman, K. J., Greenland, S., & Lash, T. L. (2008). Modern epidemiology. Philadelphia: Wolters Kluwer Health/Lippincott Williams & Wilkins.