Biostatistics Flashcards
(28 cards)
What is the difference between ratio, interval, nominal, and ordinal data?
Ratio - continuous, equal difference between values, a true and meaningful zero
- ex. age, height, weight, time, blood pressure
Interval - continuous, equal difference between values, but without a meaningful zero (0 does not equal none)
- ex. celsius and fahrenheit
Nominal - discrete (categorical), categories are in an arbitrary order (yes/no data)
- ex. gender, ethnicity, marital status, mortality
Ordinal - discrete (categorical), categories are ranked in a logical order, but the difference between categories is not equal
- ex. NYHA function class I-IV, 0-10 pain scale
when is mean, median or mode preferred?
mean - preferred for continuous data that is normally distributed
median - preferred for ordinal data or continuous data that is skewed
mode - preferred for nominal data
What is a range?
The differene between the lowest and highest value
(if highest value is 8 and lowest is 6, the range is 2)
With normally distributed data, what percent of your data will be within 1 SD of the mean? What about 2 SD of the mean?
68% of data will be within 1 standard deviation of the mean
95% of data will be within 2 standard deviations of the mean
What is a positive vs negative skew?
Positive (right) skew - more low values, the peak is on the left
- mode is lower, mean is higher, and median is right in the middle
Negative (left) skew - more high values, the peak is on the right
- mode if higher, mean is lower, and median is right in the middle
What is the null hypothesis for a study to demonstrate that one drug or treatment is superior than another?
Null hypothesis = no difference between groups
ex. metoprolol = placebo
The goal of the study is to disprove the null hypothesis and show that the new drug is superior to the old one
if no statistically significant difference -> accept null hypothesis
if there was a statistically significant difference -> reject null hypothesis or accept alternative hypothesis
What is the alpha level?
Alpha level is an error margin used as the standard for significance. It is commonly set at 5%, or 0.05. The P-value is compared to alpha to determine statistical significance.
if p-value is < alpha (p <0.05), the null hypothesis is rejected
if p-value is ≥ alpha (p ≥ 0.05), the null hypothesis is accepted
When analyzing confidence interval significance? When do you not want it to include zero vs. one?
When comparing difference (means) data, you do not want it to include zero, because that would mean no change. (pretty much just normal data)
When comparing ratio data (relative risk, odds ratio, hazard ratio), you do not want it to include one, because that would mean no difference.
What is type 1 vs. type 2 error?
Type I error: the null hypothesis is rejected in error (false positive)
- risk is determined by alpha and is related to confidence error
- alpha = the risk of a type I error
- if p-value < 0.05, we are 95% confident the result is correct, the risk of a type I error is <5%
Type II error: the null hypothesis is accepted in error (false negative)
- beta = risk of type II error (usually 0.1 or 0.2)
- power = 1 - beta
- power = probability of avoiding a type II error
What is risk? What is relative risk? What is relative risk reduction?
Risk = # of subjects with the unfavorable event / # of subjects in that arm of the study
Relative risk = risk in the treatment group / risk in the control group
- RR = 1, no difference in risk
- RR > 1, higher risk in the treatment group
- RR < 1, lower risk in the treatment group
Relative risk reduction = (% risk control group - % risk treatment group) / % risk in control group OR RRR = 1 - RR
- shows how much the risk if reduced in the treatment group vs. control group
What is absolute risk reduction (ARR)?
ARR - the absolute difference in outcome rates between 2 groups. This is the actual effect of the drug BEYOND the effect of the placebo
- ex. if ARR was 12% (or 0.12): for every 100 patients treated with metoprolol, 12 fewer patients will experience HF exacerbation
ARR = (% risk in the control group) - (% risk in the treatment group)
What is the number needed to treat (NNT)? What is NNH?
NNT - the number of people who need to be treated for a certain period of time in order for 1 patient to benefit
NNT = 1 / (risk in control group - risk in treatment group) OR 1 / ARR
- round up no matter what
NNH - number of people who need to be treated for a certain period of time in order for 1 patient to experience harm (ex. an adverse event)
NNT = 1 / (risk in control group - risk in treatment group)
- round down no matter what
What is odds ratio (OR)?
OR - probability of an outcome occurring with an exposure versus without the exposure
OR = [# that have the outcome w/ exposure (A) x # w/o the outcome, w/o exposure (D)] / [# w/o outcome w/ exposure (B) x # with outcome, w/o exposure (C)]
OR = [(with, with) x (without, without)] / [(without, with) x (with, without)]
ex. if OR was 1.23 w/ serotonergic ADs (outcome = falls w/ fracture), serotonergic ADs have a 23% increased risk of falls with fracture
What is a hazard ratio (HR)?
HR - rate at which an unfavorable event occurs within a short period of time
HR = hazard rate in treatment group / hazard rate in the control group
How do you interpret:
OR/HR = 1
OR/HR > 1
OR/HR < 1
OR/HR = 1, event rate is similar between groups
OR/HR > 1, event rate is higher in the treatment group
OR/HR < 1, event rate is lower in the treatment group
What is the difference between primary and composite endopoints?
Primary endpoint is the main result that is measured to see if the treatment had a significant benefit. (ex. death from CV causes OR nonfatal stroke OR nonfatal MI)
Composite endpoint combines multiple individual endpoints into one measurement. These have to be similar (ex. death from CV causes, nonfatal stroke, nonfatal MI, NOT A1c > 8% or something not that serious)
What tests can be used for normally distributed, continuous data? When would you use each?
If continuous and normally distributed, use a parametric test
If 1 group -> one-sample t-test
If 1 group with before & after measures -> dependent/paired t-test
If 2 groups (treatment & control) -> independent/unpaired student t-test
If 3 groups -> ANOVA
With discrete/categorical data, what tests should we use?
if 1 group -> chi-square test
if 1 group w/ before and after measures -> Wilcoxon Signed-Rank test
if 2 groups (treatment and control) -> chi-square test or Fisher’s exact test
What is correlation vs. regression?
Correlation - used to determine if one variable (ex. days of hospital stay) is related to another variable (ex. incidence of hospital-acquired infection)
- Spearman’s rank-order correlation is used for ordinal, ranked data
- Pearson’s correlation coefficient is used for continuous data (if r = 0, no correlation, if r = +1, -1 there is a strong positive or negative correlation)
Regression - used to describe the relationship between a dependent variable and one or more independent variable
- linear (continuous data), logistic (categorical data), Cox (categorical data in a survival analysis)
What is sensitivity vs. specificity?
Sensitivity - a true positive. 100% sensitivity means the test will be positive in all patients with the condition
Specificity - a true negative. 100% specificity means the test will be negative in all patients without the condition
What is the order of these studies from least to most reliable?
- case-controlled studies
- expert opinion
- systematic reviews and meta-analyses
- case series and case reports
- cohort studes
- randomized controlled trials
Least: expert opinion
Case series and case reports
Case-controlled studies
Cohort studies
Randomized controlled trials
Most: systematic reviews and meta-analysis
What is a case-control study vs. cohort study?
Case-control: A retrospective study that compares patients with a disease to those without the disease (ex. to see if there was an exposure to something)
Cohort study: prospective or retrospective. Compares patients who have exposure to patient who did not have an exposure (ex. to see if a patient got/gets a disease)
What is the ECHO model?
Looks at economic, clinical, and humanistic outcomes and compares them
Economic: direct, direct and intangible costs
- direct: how much does it cost (medical and non-medical)
- indirect: lost work time, low work productivity, morbidity cost from having the disease, mortality
- intangible: pain, suffering, anxiety, fatigue
Clinical outcome: medical events that occur as a result of the intervention
Humanistic: consequence of the intervention to the patient or caregiver (Ex. QOL, satisfaction)
What is the difference between average cost effectiveness ratio and incremental cost effectiveness ratio? What is the equation for incremental cost ratio?
Average cost effectiveness ratio - cost of one treatment independent of other treatments
- cost ratio = cost of outcome
Incremental cost effectiveness ratios (ICR) - change in costs and outcomes when comparing two treatments
ICR = (C2 - C1) / (E2 - E1)
- C, cost
- E, effects
