Usmle Biostatistics
BIOSTATISTICS
* Effect modification: effect of the main exposure on the outcome is affected by another variable. It is NOT a bias! * Case-control study: also known as retrospective study. Divided into “cases” and “controls.” If disease is rare, the odds ratio will approximate the relative risk (following subjects over time.) * Cohort study (Retrospective or prospective): divided into “exposed” and “non-exposed.” Study subjects are free of the outcome at the time the study begins. * Cross-sectional study: exposure and outcome are studied at one point in time. (Think: snapshot study) * Randomized control trial: gold standard for studying the efficacy of treatment or procedure. Subjects are randomized into experimental or control. Less bias and strong causal relationship. * Cross-over study: group of participants are randomized to one treatment for a certain period of time and the other with the alternative for that same certain period of time. Then after period ends, the groups switched treatments for the duration of the trial. * Parallel study: think “drug group vs. placebo group.” No other variables are measured
* Hazard ratio: the higher the ratio, that higher risk for hazardous events. If ratio is 1 (or value if closer to 1), then there’s little difference between the two entities * Factorial design studies: randomization of different interventions with additional study of 2 or more variables * Cluster analysis: randomization of different groups, not individuals
* Lowering the cut-off will increase its sensitivity * Confounder: an extraneous factor which has properties linking it to the exposure and outcome of interest * Probability if test is negative, 1-negative predicative value * Observer bias: investigator’s decision is adversely affected by knowledge of exposure status. Example: It is known the ESRD is more common in black population * Respondent bias: outcome obtained by patient’s
* Effect modification: effect of the main exposure on the outcome is affected by another variable. It is NOT a bias! * Case-control study: also known as retrospective study. Divided into “cases” and “controls.” If disease is rare, the odds ratio will approximate the relative risk (following subjects over time.) * Cohort study (Retrospective or prospective): divided into “exposed” and “non-exposed.” Study subjects are free of the outcome at the time the study begins. * Cross-sectional study: exposure and outcome are studied at one point in time. (Think: snapshot study) * Randomized control trial: gold standard for studying the efficacy of treatment or procedure. Subjects are randomized into experimental or control. Less bias and strong causal relationship. * Cross-over study: group of participants are randomized to one treatment for a certain period of time and the other with the alternative for that same certain period of time. Then after period ends, the groups switched treatments for the duration of the trial. * Parallel study: think “drug group vs. placebo group.” No other variables are measured
* Hazard ratio: the higher the ratio, that higher risk for hazardous events. If ratio is 1 (or value if closer to 1), then there’s little difference between the two entities * Factorial design studies: randomization of different interventions with additional study of 2 or more variables * Cluster analysis: randomization of different groups, not individuals
* Lowering the cut-off will increase its sensitivity * Confounder: an extraneous factor which has properties linking it to the exposure and outcome of interest * Probability if test is negative, 1-negative predicative value * Observer bias: investigator’s decision is adversely affected by knowledge of exposure status. Example: It is known the ESRD is more common in black population * Respondent bias: outcome obtained by patient’s