Bias in Epidemiological Studies - Literature review Example

Name: Institution: Tutor: Date: PART I Comment on the capacity of cross sectional, case control and cohort studies to provide support for a possible causative association between a risk factor and a disease outcome. Introduction According to Nijsten & Stern (2012), results from observational research have often received significant criticism owing to arguments that they are susceptible to various influences from unpredictable incomprehensible factors. Alternatively, recent developments have challenged this particular view, revealing comparable outcomes between randomized controlled trials and observational studies. Cross sectional studies, case control studies and cohort studies form important categories of observational study designs that are useful in assessing the correlation existing between disease outcomes and risk factors in epidemiological studies. However, as highlighted earlier, a major question seems to be their capacity to provide support for a possible causative association between a risk factor and a disease outcome. This particular paper generally intends comment on the capacity of cross sectional, case control and cohort studies to provide support for a possible causative association between a risk factor and a disease outcome. Faced with a research study issue regarding the correlation between a likely risk factor and a disease, an epidemiologist must go for the most suitable strategy in order to solve the issue. Various circumstances ought to be taken into account including the rate of occurrence of the disease, time between exposure and manifestation of a given disease, the importance of the study issue, whether the risk factor is linked to one or other more diseases, available funding for the research, ethical issues, among others (Meirik,2012, Pp.1). In considering the highlighted factors into consideration, the researcher is able to find a research strategy or strategies that he or she deems appropriate for his or her research. Cohort studies generally select its subjects on the basis of their risk status. In this case, individuals, usually two groups, one consisting of those who have the disease (the case group) are compared with another group consisting of those who are disease negative (the control group). The subjects are studied over a period of time in order to evaluate their future outcome status and are normally used to ascertain possible risk factors. Despite having various advantages and limitations, Meirik (2012, Pp.1) argues that the conclusions derived from findings of cohort studies are usually much stronger than those of case control studies. A significant advantage of this type of observational study can be argued based on the fact that both its prospective as well as retrospective designs often sets up a temporal correlation between the exposures and the outcomes, as a result, ensuring the exposure measurements are not biased by outcomes and minimizing the possibility that a correlation is effect-caused. In general, I am of the opinion that, just as argued by the Centers for Disease Control and Prevention (2004,p.1), cohort studies offer stronger support for causality compared to case control and cross sectional studies. My opinion is based on the fact that other its prospective as well as retrospective designs often sets up a temporal correlation between the exposures and the outcomes, this study design seems to be having the five criteria essential in establishing a cause and effect relationship including strong correlation,consistency,plausibility,temporality and the biological gradient. Case control studies, on the other hand, choose its subject for study with reference to the subjects’ disease status. Usually a comparison involving a group of various individuals who have the disease (the case group) with those who do not have the disease (the control group) are carried out. The control group, however, has to originate from the target population where the cases were obtained. This type of observational study often refers back in an attempt to identify the possible exposures that both the case and control groups may have faced. According to Ho et.al (2008, Pp.1679), cross sectional studies are just as strong as cohort studies. Its most significant advantage is based on the fact that its studies are often derived from the sample of a major population and therefore does not have to depend on individuals presenting themselves for a medical treatment. On the other hand, its disadvantages consist of the fact that it is argued not to represent individuals who take part in a study. Besides, Hung (2006, Pp.31) argues that when employed as a means of disease prevalence assessment, this observational study design appears not to be effective especially in situations where the disease rates are very small. Case control studies are also argued not to effectively represent disease recurrent conditions as the disease/condition may be at its peak or inactive when the research study is being conducted. In fact, Van den Broeck & Brestoff (2013, p.26) argues that for exposure to be a cause, it has to have occurred before the outcome, a requirement often known as the basic temporality criterion, with the opposite condition or situation known as the reverse causality, which denotes when an outcome has a casual-effect on exposure. Van den Broeck & Brestoff (2013.p.26) claim this is of serious concern especially where it is challenging to evaluate the order of events, for instance, in a number of cross sectional studies. In general, cross sectional studies determines mostly the association and not causality. Case control studies involve the epidemiologists operating in reverse, from the outcome to the assumed cause hence known as retrospective studies. The subjects are often chosen based on either disease presence or absence or the outcome under consideration, such that there are two groups, the case subjects (those with the health issue) and the control subjects (those without the health issue).These particular groups are thereafter compared in order to establish specific risk factors. According to the Centers for Disease Control and Prevention (2004,p.1), since they operate in reverse, case control studies are more likely to develop uncertainties as regards the sequential correlation between the risk factors and the outcomes. Other than this, Meirik (2012, Pp.1) argues that the conclusions derived from findings of Case control studies are usually weaker when compared to those of case control studies. In my opinion, think the case control studies do not meet the five criteria necessary to establish the cause and effect relationship, hence the question regarding its capacity to provide support for a possible causative association between a risk factor and a disease outcome. Conclusion From the above analysis what is evident is the fact that all the three major study designs have both their intrinsic strengths as well as limitations. This means that differences in terms of methodological quality amongst these study designs point out that results of given studies are most probably to be affected by various forms of bias when compared to others. It is thus important that investigators take into account the quality of given study designs when assessing the prospective causative association between a risk factor and a disease outcome. Work Cited Centers for Disease Control and Prevention (CDC). An Introduction to Epidemiology, Pp.1, 2004 Ho, Michael, Peterson, Pamela & Masoudi, Frederick. Key Issues in Outcomes Research: Evaluating the Evidence: Is There a Rigid Hierarchy? Circulation, (118), Pp.1675-1684, 2008 Hung, Bui.The most common causes of and risk factors for diarrhea among children less than five years of age admitted to Dong Anh Hospital, Hanoi, Northern Vietnam, Pp.1-88, 2006 Meirik, O.Cohort and Case-Control Studies: Unit for Epidemiological Research, Pp.1, 2012 Nijsten, Tamar & Stern, Robert. How Epidemiology Has Contributed to a Better Understanding of Skin Disease, Journal of Investigative Dermatology,(132), Pp.994–1002 Van den Broeck Jan & Brestoff, Jonathan. QR code for Epidemiology: Principles and Practical Guidelines, Springer, Pp.3-37, 2013 PART II Sources of Bias in Epidemiological Research, Potential Problems and How They Can Be Controlled Introduction Epidemiology is simply the science that deals with the study of health-related incidents affecting populations. Epidemiologists are often concerned with researching the origins of health issues and particularly issues relating to nutrition, risky human behaviours and environmental dangers (Bayona & Olsen, 2004, p.7). They generally gather information within the community and by way of data analysis, seek to discover the health risk factors, especially those that can be altered. Like every other science, epidemiology is rooted on the fundamental principle that accurate observations and measurements, along with thorough analysis with regards to existing knowledge, is actually the best way to conduct a given study. Even with this, Choi & Pak (2007, p.1) highlight that error factors that tend to reduce the credibility of given epidemiological studies are bound. In view of this therefore, the awareness of the possible sources of bias in epidemiological research is therefore vital for not only the researchers (when designing and carrying out studies) but policy-makers within the public health as well. According to Van den Broeck & Brestoff (2013, p.19), bias in epidemiology is defined as the systematic alteration of the correlation existing between a determinant and the outcome owing to an inadequacy within the research design. It stems from flawed trends during the process of data collection, data analysis, data interpretation, publication or data review. This often leads to overestimation or even underestimation of various frequency/effect measures. The basis of biased statistical outcomes may mostly be in the collection of data or information, in the choice of the sources of information or in the analysis of the collected data. Bias may therefore arise out of information bias, selection bias and confounding bias. There are a number of sources that lead to information bias including: Memory Lapses among Subjects Memory lapses among subjects in various case-control investigations requested to recollect past exposure to risk factors is a significant source of information bias in many epidemiological studies. According to Barratt & Kirwan (2009, p.1), in case-control investigations, data regarding exposure is often gathered retrospectively. This means that data quality is therefore greatly determined by the patients’ ability to correctly recollect past exposures. The consequence of this is that recall bias tends to occur when the provided information on exposure is dissimilar between the cases and controls, leading to either an overestimation or underestimation of the correlation between the outcome and exposure, hence bias. In this case, memory lapses among subjects in various case-control investigations forms one source of bias in epidemiological research. According to Barratt & Kirwan (2009, p.1), to control this, the exposure data should be collected from medical records or the research participants be blinded with regards to hypothesis being investigated. Use of sub-optimal measurement instruments As highlighted by Sanderson et.al, (2007, p.666), the assessment of quality as well as vulnerability to bias is important while not only interpreting primary research but while also carrying out systematic reviews or meta-analyses. On the other hand, as much as instruments used in assessing quality are well-described, little attention has often been given to related instruments used for observational epidemiological research (Barratt & Kirwan, 2009, p.1). As such, variation or the usage of sub-optimal measurement instruments is argued to be one major source of information bias in epidemiological research. Usage of such instruments often results in incorrect estimates which affects the study results. In order to control this issue, it is necessary that investigators use standardized or calibrated instruments in the studies Bias by researchers or research personnel This forms a significant source of information bias in epidemiological research. According to Barratt & Kirwan (2009, p.1), a number of research personnel and researchers have been known to simply fabricate research data either deliberately or unconsciously in an attempt to favour results that they believe in, a factor that leads to a false estimate in terms of the association existing between exposure and outcome. To control this, observers ought to not only be blinded to the research hypothesis being investigated but also an individual’s exposure and disease status. Dissimilar standards within different biochemical laboratories Variation in the underlying standards within various biochemical standards also forms a significant source of information bias in epidemiological research. This can however be controlled by adhering to developed universal standards used in research. Misunderstanding or misinterpretation of Questions by given subjects This is also a common source of information bias while conducting epidemiological studies. This is often as a result of the subjects being interrogated or filling a research questionnaire misunderstanding or misinterpreting the questions asked. The bias can also be as a result of the subjects’ inability or refusal to offer correct responses due to personal reasons (Van den Broeck & Brestoff, 2013, p.35). This issue can be controlled by framing interview situations so as to encourage openness among the subjects being interviewed or in situations involving filling of questionnaires. Missing data required for an analysis Missing data intended for a variable required in an analysis forms a potential source of selection bias in epidemiological studies of many kinds. The issue can be very severe especially in analyses involving significantly large numbers of variables (Van den Broeck & Brestoff, 2013, p.35). For instance, regression procedures normally exclude a whole observation when the value of any of the variables within regression is missing. This is able exclude large percentages of observations even when just 5 percent value of any one given variables is missing, as a result, inducing selection bias. In order to control this issue, there is need for the avoidance of the systematic differences in the missing data rates. Lack of precision of the sampling frame This forms a major source of selection bias in epidemiological studies. This often produces non-representative samples causing the parameter estimates to be different from the one existing. As a source of selection bias, Delgado-Rodríguez & Llorca (2004, p.636) argues that it can be controlled by measuring the variables that influence selection on all the study subjects. Research Design Study designs also form a significant source of selection bias arising especially as a result of poor or inappropriate description of eligible target population. This is often witnessed in cases where for instance the patients collected do not actually represent cases derived from the population. According to Delgado-Rodríguez & Llorca (2004, p.639), this particular source is attributed to many possibilities including competing risks, length-biased sampling, and healthcare access bias among others. As a source of selection bias, it can also be controlled by measuring the variables that influence selection on all the study subjects. Co-occurrence or the mixing of effects of unconnected factors This is also a common source of confounding bias in epidemiological studies that often leads to spurious results, a factor that leads researchers to conclude an existence of a statistical association when in real sense it not actually existence or otherwise the non-existence of an association when in real sense it is actually present. To control this, Barratt & Kirwan (2009, p.1) suggest that significant attention ought to be given both at the design stage and analysis stage so that these particular effects are minimized. Conclusion The above discussion has without a doubt highlighted a number of sources of bias in epidemiological studies that arise from the systematic alteration of research results. These alterations are an issue especially when an epidemiologist is assessing the correlation between the risk factor and a health issue. Whether a given risk factor goes undetected or a given health condition is misidentified, the implications may be very dangerous for the public. Detecting bias is possible and therefore ought to be a key priority in any epidemiological research. Being set from the start in order to avoid the many pitfalls is normally the best resolution. Work Cited Bayona, Manuel & Olsen, Chris. Observational Studies and Bias in Epidemiology, Pp.7-36, 2004 Barratt, Helen & Kirwan, Maria. Bias in Epidemiological Studies, Pp.1, 2009 Choi, Bernard & Pak, Anita. Understanding and Minimizing Epidemiologic Bias in Public Health Research, Canadian Journal of Public Health, Vol.96, No.4, 2005 Delgado-Rodríguez, M & Llorca, J.Bias, Journal of Epidemiology Community Health, (58), Pp.635-641, 2004 Sanderson et.al.Tools for assessing quality and susceptibility to bias in observational studies in epidemiology: a systematic review and annotated bibliography, International Journal of Epidemiology, Volume 36, Issue 3, Pp. 666-676, 2007 Van den Broeck, Jan & Brestoff, Jonathan. QR code for Epidemiology: Principles and Practical Guidelines, Springer, Pp.3-37, 2013 Read More