Biostatistics PDF

Chapter Twenty-Two: Biostatistics Key Definitions Epidemiology The study of a disease and its behavior within a population Endemic What is the expected amount of disease per unit time in that population? Epidemic More than 2 standard deviations above the normal expected rate Pandemic The entire worl...

Chapter Twenty-Two: Biostatistics Key Definitions Epidemiology The study of a disease and its behavior within a population Endemic What is the expected amount of disease per unit time in that population? Epidemic More than 2 standard deviations above the normal expected rate Pandemic The entire world is involved A brand-new disease to humans REMEMBER Screening will increase incidence and prevalence Key Rates Per thousand live births Birth Rate Number of live births Fertility Rate Number of live births by women in the fertile years (15-45) Infant Mortality Rate Number of infants who died within 1-12 months of life Neonatal Mortality Rate Number of newborns who died within the first 1 month of life Maternal Mortality Rate Number of moms who died during the birthing process per 100,000 live births Crude Mortality Rate Total number of deaths/populations Case Fatality Rate Number of people who died/total number of cases Attack Rate Number of people who got the disease/number of susceptible people in the population REMEMBER REMEMBER Perinatal = neonatal births + stillbirths weeks gestation REMEMBER Infant: 1-12 months old Neonate: less than one month old 648 Biostatistics Studies Case Study Done for the good of your colleagues Write-up of one interesting case Case Series Compile a series of case studies Look for trends Good for rare diseases Simply describes people with a certain disease May suggest the need for a retrospective look Clinical Correlation Case control study: • • People have the disease already What in the past were they exposed to get the disease? Clinical Correlation Control group: • People who do not have the disease but have the same demographics • Look for an odd • ratio for any cohort • What are the odds that this exposure was the cause of this disease? Cohort Study Cohort: a group with defined entry criteria Exposure known first, determining outcome 2 types: Prospective Choose a sample population; divide it into two groups based on presence or absence of a risk factor Follow the groups over time Measure incidence and prevalence More expensive Measures Relative Risk Retrospective Samples are chosen after the fact based on presence (cases) or absence (controls) of disease 649 Biostatistics Information gathered on risk factors Measure Odds Ratio Case Control Outcome known first, determining exposure Clinical Trials Look for an outcome given some kind of intervention Observe and intervene Single blinded: only patient does not know who is getting treatment Hawthorne Effect: when the patient knows, it will influence his behavior Double blinded: neither patient nor doctor knows who is getting treatment Gold Standard Keeps doctor from changing his behavior (observer bias) Pygmalion effect: If his behavior is so different that the calculation is affected Clinical Correlation Intervention makes a bigger difference when there is a higher prevalence of disease Meta-Analysis Compile studies on rare disease (because you can never get a complete cohort Look for a trend but cannot make any conclusions Not all studies are conducted with the same methodology Longitudinal Follow them over time for a defined period Lead time bias: Earlier detection of a disease appears as if there is a longer survival Latent-period bias: if study does not wait long enough for disease to occur Minimum of 10 years or more required for these types of studies Wait long enough to lead to a disease Early period has a lower prevalence of disease than later period has Cross Sectional Studies A prevalence survey Looks at the prevalence of a disease and risk factors 650 Biostatistics N O conclusions about the causes can be made A trend towards cause necessitates a prospective study Both exposure and outcome known at start of study Figure 22.1: 2 x 2 Table H ow to E valuate a Test M easurements of V alue S ensitivity Rules out disease TP TP +FN S pecificity Rules in disease TN TN +FP R EM EM B ER Abbreviations: TP : true positive TN : true negative FP : false positive FN : false negative P ositive P redictive V alue (P P V ) If a test comes back positive, what are the chances you really do have the disease? TP TP + FP Biostatistics Negative Predictive Value (NPV) If a test comes back negative, what are the chances you really do not have the disease? TN TN + FN Relative Risk (RR) How much more risk do you have than the general population? What is the risk the exposure will lead to the outcome? A A+B C C+D Attributable Risk (AR) How much of the relative risk is attributable to the exposure? RR—1 RR Odds Ratio What are the odds that the outcome is due to the exposure? AxD BxC Number Needed to Treat How many people need to be treated until one person benefits from the outcome? You want the NNT to be a low number 1 ARR Number Needed to Harm How many people are exposed to a risk factor or treated before one person is harmed from the outcome? You want the NNH to be a high number 1 AR 652 Biostatistics Likelihood Ratios Positive Likelihood Ratio: Sensitivity (1-Specificity) Negative Likelihood Ratio: (1-Sensitivity) Specificity Reliability /Precision Definition: how often do you get the same number Random error reduces reliability/precision Validity/Accuracy Definition: how close do you get to the gold standard number Systematic error reduces validity/accuracy Ex: Mis-calibrated equipment Figure 22.2: Precision/Reliability vs Accuracy/Validity 653 Biostatistics Confidence Interval Represents the confidence (95%) that the relative risk or odds ratio you calculated lies between the intervals Calculated after accounting for random error If the interval contains 1, then there is no statistically significant effect of exposure Example 1: CI: 1.24 - 1.89 Does not include 1, so it is statistically significant Example 2: CI: 0.92—1.06 Includes 1, so it is statistically not significant Biases in Research Selection bias o When entry criteria are not well defined Measurement bias o Human error o Can be affected by observer or patient Observer bias o Experimenter knows the exposure group and affects the results Lead-time bias Latent-time bias o Ending a study before the outcome has occurred Recall bias o Problem with memory o Occurs in retrospective studies Confounding bias o The largest bias o Anything that can confuse the issue o Match control/cohort as much as possible Observer bias o Observer modifies his/her own behavior Hawthorne bias o Patients know they are being observed and modify their behavior Berkson bias Population being observed is in a hospital Clinical Trials Phase I: testing safety in animals Phase II: testing in patient volunteers Phase III: test efficacy and document side effects in large group of patient volunteers Phase IV: post marketing studies 654 Biostatistics Summarizing Data Mean: average Median: middle number Mode: value that appears most often Bell shaped curve: mean = median = mode Positive Skew: mean > median > mode Negative Skew: mean < median < mode Figure 22.3: Bell-shaped Curve Statistically setting up a study Statistical Inference 1. Define the research question 2. Define the NULL hypothesis 3. Define ALTERNATIVE hypothesis P value Significant if p<0.05 Less than 5% chance that the results are due to chance Types of Errors Type I Error Alpha error The chance of rejecting the NULL hypothesis when it is actually true p<.05 Type II Error Beta error Accepting the null hypothesis when it is false = 1 – power Power is dependent on sample size Increasing sample size decreases type II errors 655 Biostatistics Statistical Testing T tests Compares means of two groups from a single nominal variable Pooled t- test Matched pair t- test Chi-square Compares the proportions of groups; looks for correlation between two factors Analysis of variance (ANOVA) Compares means of three or more groups Correlation analysis Figure 22.4: Positive and Negative Correlation 656 Biostatistics Figure 22.5: PPV and NPV 657 Chapter Twenty-Three: Cardiac Physiology 658