Types of observational studies
     Case-Control
          Retrospective
          Subjects with disease of interest are compared to an otherwise similar group that is disease free
          Information about risk factor exposure is collected
          Meant to determine associations between risk factors and disease occurrence
          Can determine an ODDS RATIO but not incidence
          Very popular in exploring an exposure-disease association, because they are relatively cheap and less time-consuming than cohort studies.
          One major drawback is the fact that risk cannot be derived directly from the results.
          The odds ratio IS NOT the same as relative risk, although it can sometimes be approximately equal.
               Relative risk can be calculated in follow-up studies by comparing the risk in exposed individuals to the risk in unexposed individuals. 
               Direct calculation of the relative risk is not possible in case-control studies, because the study design DOES NOT include FOLLOWING PEOPLE OVER TIME.
               If the PREVALENCE OF THE DISEASE IS LOW, the odds ratio APPROXIMATES the relative risk. This statement is called the "rare disease assumption". In other words, if the outcome of a case-control study is not common in the population, the odds ratio is close to the relative risk.

     Case series
          Can be helpful in determining the natural history of uncommon conditions but provides no information about disease incidence
          
     Cohort
          Prospective observational
          Groups are selected based upon presence or absence of risk factors
          Subjects are followed over time to determine the INCIDENCE of something

     Cross-sectional
          Takes a sample of individuals from a population at one point in time
          Allows determination of a disease's PREVALENCE (total number of cases in a population at a given time).
          
Types of study designs
     Factorial design
          Two or more different interventions, each with different variables or endpoints (see figure)
     Cluster analysis
          Grouping of different data points into similar categories
          Usually involves randomization at the level of groups rather than at the level of individuals
     Cross-over study
          A group of participants is randomized to one treatment for a period of time and the other group is given an alternative treatment for the same period of time
          At the end of the time period, the two groups then switch treatments for another set period of time
     Parallel study
          Randomizes one treatment to one group and a different treatment to the other group

Types of bias
     Hawthorne effect
          The tendency of a study population to affect the outcome because these people are aware that they are being studied. This awareness leads to a consequent change in behavior while under observation, thereby seriously affecting the validity of the study. Hawthorne effect is commonly seen in studies that concern behavioral outcomes or outcomes that can be influenced by behavioral changes. In order to minimize the potential of the Hawthorne effect, studied subjects can be kept unaware that they are being studied; however, this may pose ethical problems. Randomized control trials have a sense of uncertainty and risk due to randomization, which may be more potent behavior modifiers than mere observation.
     Confounding
          The process of MATCHING is an efficient method to control confounding and is frequently used in case-control studies.
          Initial step involes selection of matching variables, which should always be the potential confounders (e.g. age, race).
          Cases and controls are then selected based on the matching variables such that both groups have a similar distribution in accordance with the variables.
     Selection bias
          Obtaining patients that DO NOT REFLECT the populations of interest
     Susceptibility bias
          A subgroup of selection bias that occurs when the treatment regimen selected for a patient depends on the severity of the patient's condition. 
          This type of bias fails to take into account other confounding variables that may be accounting for the patient's condition.
          This bias also negates the benefits of randomization, which usually avoids selection bias and confounding variables.
     Information bias
          Occurs due to the imperfect assessment of association between the exposure and outcome as a result of errors in the measurement of exposure and outcome status.
          It can be minimized by using standardized techniques for surveillance and measurement of outcomes, as well as trained observers to measure the exposure and outcome.
     Lead time bias
          Happens when two interventions are compared to diagnose a disease, and one intervention diagnoses the disease earlier than the other without an effect on the outcome (such as survival).
          This would make it appear that the intervention prolonged survival when it really just diagnosed th disease sooner.
     Measurement bias
          Occurs from poor data collection with inaccurate results, which is not what is described in this case.
     Ascertainment bias
          Results from mislabeling exposed/unexposed or cases/controls
     Recall bias
          Occurs when a study participant is affected by prior knowledge to answer a question.
          More common in case-control studies rather than randomization.
     Observer's bias
          Occurs when the observer may be influenced by prior knowledge or details of the study that can affect the results.
          May be avoided by BLINDING or measuring objective outcomes (such as death).

Statistical Tests
     Two sample t-test
          Commonly employed to compare two means
          Requires the two mean values, the sample variances, and the sample size.

     Two sample z-test
          Can be used to compare two means
          Population (not sample) variances are employed, but since population variances are usually not known, this test is not often used.

     ANOVA (i.e. analysis of variance)
          Used to compare three or more means

     Chi-square test
          Appropriate for categorical data and proportions

     Meta-analysis
          Epidemiologic method of pooling data from several studies to do an analysis having relative big statistical power.

Other issues
     Latent period
          A natural phenomenon, NOT a bias
          Chronic diseases may have a very long latent period, and extended periods of continuous exposure may be necessary to affect the outcome.

Content on this page has not been checked for accuracy. The author's credentials have not been confirmed. Always use multiple peer-reviewed scholarly sources and good clinical judgement in your decision making.

The Compendium platform was created by David Kwinter, MD, CCFP(EM).

Dr. Kwinter is a Cayman Family Doctor. He runs a Family Medicine Clinic at Cayman Medical in George Town, Grand Cayman, Cayman Islands.