Observational Designs (Portney)
Cohort subject selection -subjects must be... -how are subjects classified? -how to facilitate comparison?
...free of the target outcome at baseline but susceptible to developing it during the follow-up period By their level of exposure to the hypothesized risk factor Unexposed group should be as similar as possible to the exposed group in all factors related to the outcome except for the specific exposure under study
Case-control is especially useful for...
...studying rare disorders or conditions with long latency or incubation periods.
Challenges in retrospective studies
Available data must be sufficient to clearly establish each subject's exposure history and outcome status. When the target outcome is a transient condition, confirmation of prior outcome status can be challenging. Medical records and historical records may be incomplete, and a retrospective researcher is unable to exert direct control over the procedures that were used previously to measure target variables. Consequently, measurements may not align perfectly with the current study question or with current expectations for accuracy. Because the outcome status of participants is often known at the start of a retrospective study, investigators need to take action to protect against biased reporting or assessment.
Selection bias in case-control studies
Because subjects are purposefully selected based on whether a target disorder is present or not, selection bias is a special concern. Cases and controls must be chosen regardless of their exposure histories. If they are differentially selected on some variable that is related to the exposure of interest, it will not be possible to determine if the exposure is truly related to the disorder. When samples are composed of subjects who have volunteered to participate, self-selection biases can also occur.
Advantages to retrospective studies
Cheaper and faster than prospective studies, offering a more efficient way to investigate diseases with long latency or incubation periods.
Selection of controls in a case-control study/validity of study -What is the purpose of a control group in a case-control study? -What is a useful question to ask? -What about selection criteria?
Demonstrate how common the exposure is among people in the population from which the cases emerged Would the control subjects in this study have been eligible to serve as case subjects if they had developed the same target disorder? Any restrictions or criteria used to select the cases should also be used to select controls.
Most common use of observational design How is this done?
Draw causal inferences about the effects of a hypothesized risk factor By comparing the outcomes of a group of subjects who have been exposed to a risk factor to the outcomes of those who remain unexposed.
Confounding in case-control studies How to deal with confounders?
Effect of extraneous variables that are related to the exposure of interest and that independently affect the risk of developing the disorder. To examine the potential contribution of confounders, investigators must collect data on them and analyze their association with both the exposure and the disease.
Analytic cohort study
Endeavor to estimate the magnitude of the risk associated with an exposure Derived by comparing the incidence of new outcomes across groups of exposed and unexposed subjects
Advantage of cohort study
Establishing the correct temporal sequence of events = cohort studies involve only participants who, at the time exposure is ascertained, are free of the target outcome
How do observational studies contrast experimental trials?
Experimental trials -test the effectiveness of interventions -variable manipulation and controlled comparisons Observational studies -characterize unmodified relationships -analyzing how existing factors or individual characteristics are associated with health outcomes
Strengths of prospective studies
Exposure status is determined at the start of the study, without foreknowledge of any participant's eventual outcome. This is a strength of the prospective design since the lack of investigator or participant knowledge of the outcome minimizes the threat of biased data collection or reporting.
Observation bias in case-control studies -Why is it a concern? -What is observation bias? -What is interviewer bias? -Example -How to reduce this bias?
Foreknowledge of the outcome renders the design particularly susceptible to biased measurement by the assessor or biased reporting by the participant. Observation bias occurs when there is a systematic difference in the way information about the exposure is obtained from the study groups. Interviewer bias can occur when the individual collecting data elicits, records, or interprets information differently for controls than for cases. For example, an interviewer may be aware of the research hypothesis and might ascertain information to support this hypothesis. Because so many epidemiologic studies involve an interview or observation, this type of bias must be explicitly addressed by blinding interviewers to group and hypothesis, and by making data collection as objective as possible.
Recruiting controls in case-control study -Where can they be obtained from? -What is the benefit to special groups as controls? -What are the benefits of hospital-based controls? Disadvantages?
General population resources, such as using random-digit dialing, voter registration logs, membership directories, or Medicare beneficiary lists. Controls can also be recruited from the same hospital or clinic as cases; hospital-based controls are chosen from among patients who have been admitted for health conditions other than the one under investigation. Sometimes special groups can provide viable controls, such as family members, friends, or neighbors of the case subjects. Controls selected in this manner typically resemble the cases with respect to ethnic and lifestyle characteristics, thereby reducing the threat that these factors might confound a comparison of cases to controls during the analysis. Patients are readily available and similarly motivated. A possible disadvantage is that, like the cases, they too are ill. It is important to consider that a single exposure may be implicated in the etiology of more than one disease. Therefore, investigators in a hospital-based case-control study must ensure that a comparison of the exposure histories of ill cases to ill controls is not distorted by the dual role that exposure plays in both illnesses. Selecting hospital controls from among patients whose illnesses have a distinct etiology from that of the case subjects may be challenging, but it is essential to ensure a valid comparison.
What are analytical studies typically motivated by?
Hypotheses that specify the particular exposures and outcomes to be studied and clarify the interpretations that will be offered for any found associations.
Misclassification of exposure in cohort studies
If an exposed person is mistakenly classified as unexposed or vice versa, the error is referred to as exposure misclassification.
Goal of observational studies: cause
Implicate factors that contribute to health outcomes but cannot provide conclusive evidence of a direct causal link
Challenges in cross-sectional studies
In establishing cause and effect, the time sequence of predictor and outcome variables is vitally important, a situation that can be clarified in longitudinal studies. In cross-sectional studies, however, it may not be possible to know whether the presumed cause (exposure) truly preceded the outcome.
Prospective studies
Information about one or more exposure variables (potential risk factors) is acquired through direct recording at the start of the study Exposed and unexposed subjects are identified, and both groups are followed forward in time to monitor subsequent outcomes.
Descriptive cohort study
Intent of chronicling the natural history of a disease
Selection of cases in case-control study: internal validity -What does internal validity of case-controls depend on? -What is case definition? -What can lead to misclassification errors and biased results?
Internal validity depends on the effects of case definition and case selection. Refers to the diagnostic and clinical criteria that identify someone as a case. False positives and false negatives
Retrospective cohort study
Investigation is initiated after some individuals have already developed the target outcome Investigator enters after exposure status has been determined and tracks data from the past to determine outcomes
Cross-sectional validity studies
Investigation of diagnostic test accuracy and other studies of concurrent validity When determining the accuracy of a diagnostic test to detect the presence or absence of a target health condition, the investigational test should be administered in close temporal proximity to the criterion test used to confirm the diagnosis. Diagnostic studies may be retrospective in so far as results from the index and reference tests are collected in the past (such as using medical records). Other diagnostic studies are prospective when data are not yet available from the index test, and must be newly collected by the investigators. In either instance, a diagnostic study is typically cross-sectional because current index test results are compared to current outcomes from the reference standard test. While it is not always feasible to administer both tests during the same clinic visit, it is important to keep the interval between test administrations brief so that a participant's health status does not fluctuate. Likewise, when estimating the concurrent validity of a measurement tool, recorded values are compared to values obtained using a reference standard instrument that is administered at approximately the same time. The cross-sectional design of these studies reflects an expectation that the health status of participants will remain unaltered during the brief interval between test administrations.
Prospective cohort study
Investigator contacts participants before any of them have developed the target condition Investigator enters the study to determine exposure status and follows subjects to determine who develops the disorder of interest
Prospective case-control
Investigator enrolls based on outcome status, but the investigator must wait for the cases to occur. Involve enrolling a specific selection of people and following that group while monitoring their health. Cases emerge as people who develop the disease or condition under investigation as the study progresses. Those unaffected by the disease form the control group.
Retrospective studies
Involve the examination of data that were previously collected
Advantage of longitudinal study
Its ability to document change and establish the correct time sequence of events, including confirmation that a suspected risk factor preceded its supposed outcome or effect
Practical challenges in prospective studies
Long-term commitment of time, funds, and resources
Cohort studies
Longitudinal The researcher identifies a group of subjects who do not yet have the outcome of interest Subjects are interviewed, examined, or observed to determine their exposure to factors hypothesized to alter the risk of developing the target outcome Exposed and unexposed cohort members are then monitored for a period of time to ascertain whether they do, in fact, develop the outcome. Comparisons are made of the frequency of newly developed outcomes across exposed and unexposed groups, indicating whether the hypothesized association is supported by the data.
Matching in case-control studies Matching and validity When might matching be difficult? What is an alternative strategy?
Match control subjects to a case on one or more characteristics, such as age, gender, neighborhood, or other characteristics with potential to alter the risk of the outcome (when #s are available) Greater similarity between controls and cases facilitates more valid comparison of their exposure histories; effective method of reducing or eliminating the possibility that the matched factors will confound the results; matching on a small number of factors does not eliminate the possibility that other differences between case and control groups may persist When feasible, researchers will try to recruit more than one control subject for each case to minimize bias and increase statistical precision during analysis. When many variables are relevant, such as gender, age, weight, or other health conditions. Propensity score matching = values are composite scores that represent a combination of several confounding variables. Scores are derived through a statistical process of logistic regression and can help to equate groups on baseline variables.
Challenges for cohort studies
Misclassification of exposure Bias and attrition
Case-control studies -What category? -How are groups selected? -What are cases? -What are controls? -What differences does the investigator look for? -How can comparisons be made? (2)
Observational On the basis of whether or not they have the health condition under study. Cases are those who have the target condition, while controls do not. If these two groups differ with respect to their exposure history or the presence of a risk factor. Cross-sectionally by looking at current characteristics or by looking backward in time via interview, questionnaire, or chart review
Observational studies, prediction, and prognosis
Observational studies are frequently undertaken to derive models that predict the likelihood that a particular treatment will succeed in the presence of one or more prognostic indicators.
Observational studies and diagnosis
Observational studies are used to derive clinical prediction rules for prognosis and to identify diagnostic indicators that accurately signal the presence or absence of a target condition. The accuracy of a set of diagnostic indicators may be modeled as a clinical diagnostic rule
Recall bias in case-control studies What might this bias result in? When is this of special concern?
Occurs when case subjects who have experienced a particular disorder remember their exposure history differently than control subjects who have not experienced the disorder. It is not unusual for individuals who have a disease to analyze their habits or past experiences with greater depth or accuracy than those who are healthy. This bias may result in an underestimate or an overestimate of the risk associated with a particular exposure. When cases are selected from hospitalized patients and controls are drawn from the general population, the problem of recall bias may be especially potent.
Retrospective case-control study
Participants are chosen based on their status as a case or control, and then prior exposure status is ascertained.
Challenges to internal validity in prospective studies
Related to measurement and confounding. Measurement validity can be threatened by testing effects since subjects undergoing repeated assessments may develop familiarity with testing procedures that could distort their performance. Attrition is also a potential threat in prospective studies involving a long period of monitoring. To minimize attrition, it is advisable to maintain regular contact with all prospective study participants. Confounding is a ubiquitous threat in observational research. In prospective designs, investigators must consider the possibility that other risk factors for the same health outcome may be unequally distributed across exposed and unexposed groups.
Cross-sectional studies When are variables measured? Can provide useful insight into...
Researcher takes a "snapshot" of a population, studying a group of subjects at one point in time concurrently, meaning that their time sequence cannot be confirmed. the current health status of a population, informing decisions about how best to allocate public resources.
Longitudinal studies
Researchers follow subjects through time, acquiring measurements at prescribed intervals
Challenges in case-control studies
Selection bias Observation bias Recall bias Confounding Matching
Disadvantage of cohort design Given this disadvantage, what design may be more appropriate?
Targeted outcome must occur with sufficient frequency to permit robust statistical comparisons of exposed and unexposed groups = the prospective cohort design may be inappropriate for studying rare outcomes. Large numbers of subjects would have to be followed over a long period to document new cases for analysis. To investigate risk factors for rare or slowly developing disorders = case-control design
Weaknesses of prospective studies
The investigator's inability to precisely anticipate how many participants will eventually develop the target outcome or how long it may take for the outcome to become apparent can render the prospective design impractical for investigations into rare or slowly developing health conditions.
Bias and attrition in cohort studies -Why are prospective cohort studies prone to attrition? -Bias may occur as a result of what? -How to reduce attrition? -How to deal with attrition?
Their longitudinal nature If the loss of subjects is related to either their exposure status, their outcome status, or both. Maintaining good rapport and regular communication with all cohort participants. When possible, look at the baseline measures among those who dropped out to see if they are different from those who were retained.
Recruiting cases in case-control study -In a population-based study... -In a hospital-based study...
Uses cases obtained from the general population; can be used to study disorders that are prevalent in the population or where access to data on large numbers of people can be affordably obtained; may support broader generalizability of the study's findings, but expensive and logistically challenging to verify a case diagnosis or confirm exposure history Cases are enlisted from patients in a medical institution; tend to be relatively easy to contact and recruit, and information on a participant's health history is readily obtained.
Differential misclassification What kind of bias can occur?
When errors are not randomly distributed across groups, the resulting misclassification may be differential or dependent on outcome status. Can result in an exaggeration of the true relationship between exposure and outcome. Such bias "away from the null" can occur, for example, when classification of exposure status is based on the inaccurate self-report of study participants.
Nondifferential misclassification How does this relate to measurement tools? What type of studies are particularly susceptible to this? What kind of bias can occur?
When exposed and unexposed subjects are misclassified with equal frequency/randomly Unreliable measurement tools are susceptible to random errors that can cause nondifferential misclassification. Clinical cohort studies, especially those that rely on unreliable clinical assessments to ascertain exposure status When exposure misclassification is both nondifferential and independent of the outcome for participants, the true relationship between exposure and outcome may be diluted, resulting in findings that are biased "towards the null."
When is a prospective approach most useful?
When the target outcome can be expected to occur fairly frequently and without an excessively long period of monitoring before detection is possible.
What is a cohort?
a group of individuals who are followed over time
How are existing groups identified in observational research?
by their shared history or current health status
What is the purpose of descriptive research?
characterize populations by examining the distribution of health-related variables within them
Observational studies may be classified as:
descriptive or analytical
What is the purpose of analytical reasearch?
examining those group differences in order to demonstrate how personal, environmental, behavioral, or genetic exposures help explain why one group's outcomes differ from another's
Types of cohort studies
prospective or retrospective
Longitudinal studies can be carried out...
prospectively or retrospectively
What do observational studies explore?
the causes, consequences, and predictors of disease or disability
What do observational studies characterize?
the impact of exposure to risk factors on individual or community health
