Biostatistics and Preventive Medicine Flashcards
Bias (Recruiting participants)
Selection bias
Colon Cancer Screening
Screening is by colonoscopy after the age of 50 every 10 years
What happens to incidence and prevalence if additional Federal research dollars are targeted to a specific condition
- I = no change
- P = no change
Diagnostic Odds Ratio [DOR]
= LR+ / LR-
Late-look bias
Definition, association(s), solution(s)
- Severely diseased individuals are not uncovered
- Early mortality
- Stratify by severity
Dependent Probability
P = P(A) * P(B I A)
Number needed to treat (NNT)
= 1 / ARR
Sensitivity (Recall, True Positive Rate [TPR])
= TP / (TP + FN)
= 1 - False Negative Rate [FNR]
- SNOUT
Incorrect results of statistical hypothesis
- Type I error (alpha): stating that there is an effect or difference when none exists (null hypothesis incorrectly rejected in favor of alternative hypothesis). (H0,H1)
- Type II error (beta): stating that there is not an effect or difference when one exists (null hypothesis is not rejected when in fact it is false). (H1,H0)
Proportionate Mortality Rate (PMR)
Deaths from cause / All deaths
Selection bias
Definition, association(s), solution(s)
- Sample is not representative
- Berkson bias (population selected from hospital), healthy worker effect (study population is healthier than general population), Non-response bias (people included in a study are different from those who are not)
- Randomization and independent sample
Crude Rate
Actual measured rate for the whole population
Correct results of statistical hypothesis
- Stating that there is an effect or difference when one exists (null hypothesis [H0] is rejected in favor of alternative hypothesis [H1]). (H1,H1) which equals power (1-beta)
- Stating that there is not an effect or difference when none exists (null hypothesis not rejected). (H0,H0) which equals a correct result
Binomials
- Every term expanded is nCr x^n-r * y^r
- We can find a term that contains the factor x^r in an expansion of (x+y)^n by using nCn-r * x^r * y^n-r
What happens to incidence and prevalence if behavioral risk factors are reduced in the population at large
- I = decrease
- P = decrease
t-test
- 1 interval and 1 nominal
- 2 groups only
Randomized Controlled Trials (RCTs)
Definition, Advantages, Disadvantages
- Experimental, prospective study in which subjects are randomly assigned to a treatment or control group. Could be single or double blinded study
- Ad:
- Minimize bias
- Potential to demonstrate relationships because exposure is assigned randomly, which minimize confounding
- Dis:
- Costly and time consuming
- Some interventions (like surgery) are not amenable to masking
Absolute Risk Reduction (ARR)
= [c/(c+d)] - [a/(a+b)]
What happens to incidence and prevalence if number of persons dying from the condition increases
- I = no change
- P = decrease
Odds Ratio (OR)
= (ad) / (bc)
What happens to incidence and prevalence if new effective vaccine gains wide spread use
- I = decrease
- P = decrease
Outcomes (definitions)
Results of each trial
Positive Predictive Value [PPV] (Precision)
= TP / (TP + FP)
- Varies directly with prevalence
False Discovery Rate [FDR]
= FP / (FP + TP)
Type II error (beta)
- Also known as false-negative error
- It is related to statistical power (1-beta), which is the probability of rejecting the null hypothesis when it is false.
- To increase power and reduce beta error:
- Increase sample size
- Increase expected effect size
- Increase precision of measurement
Randomized Clinical Trials (RCTs) phases
- Phase I: small # of healthy volunteers to assess safety, toxicity, pharmacokinetics and pharmacodynamics
- Phase II: small # of patients with disease of interest to assess treatment efficacy, optimal dosing, adverse effects
- Phase III: large # of patients randomly assigned either to treatment under investigation or the best available treatment (or placebo) to compare the new treatment to the current standard of care
- Phase IV: post-marketing surveillance of patients after treatment is approved to detect rare or long term adverse effects. can result in treatment being withdrawn from market
Probability of mutually non-exclusive events
P(A union B)= P(A) + P(B) - P(A intersection B)
Repeated measures ANOVA
- 1 interval and 1 nominal
- More than 2 groups, linked data
Cross-Sectional Study
Definition, Advantages, Disadvantages
- People in the population are examined the presence of a disease of interest at a given point in time (prevalence study)
- Ad:
- Provide an efficient means of examining a population
- Can be used as a basis for diagnostic testing
- Can be used to plan which health services to offer and where
- Dis:
- Cannot determine causal relationships
- Risk or Incidence cannot be directly measured
Confidence Interval (CI)
- Range of values within which the true mean of the population is expected to fall, with a specified probability
- CI = mean +/- Z * (SEM)
- For the 95% CI, Z= 1.96 (95% CI corresponds to p=0.05)
- For the 99% CI, Z= 2.58
Osteoporosis Screening
- Every women should be screened with bone densitometry at the age of 65 by DEXA scan
- Prophylaxis with bisphosphonates to increase bone density
One-way ANOVA
- 1 interval and 1 nominal
- 2 or more groups
Diabetes Mellitus Screening
- Screening with fasting blood glucose (2 measurements over 125) or HbA1c < 6.5% for patients that have hypertension and/or hyperlipidemia
- No clear recommendation for age to start screening in general population
Disease Rate
of actual cases / # of potential cases
Number needed to harm (NNH)
= 1 / AR
Tertiary Disease Prevention
Aims to reduce the disability or morbidity resulting from disease like some treatments and surgeries
Prevalence types
- Point prevalence
- Period prevalence
Probability of an event
P(A) = n(A) / n(S)
Permutations
- Order is important
- When all objects are taken n!
- When number of objects taken at a time from n nPr = n! / (n-r)!
F-Score
- Harmonic mean of precision (PPV) and recall (Sensitivity)
- = 2 * [(precision * recall) / (precision + recall)]
Indications of Pneumococcal vaccination
- Everyone above age of 65
- Cochlear implant
- CSF leak
- Alcoholics
- One vaccine above 65 only
- Single revaccination after 5 years if the patient is immunocompromised or the first injection was prior to age 65
Sample points (definition)
Elements of the sample space
Relative Risk (RR)
= a/(a+b) / c/(c+d)
Trial (definition)
A repetition of experiment
Relative Risk Reduction (RRR)
= 1 - RR
Specific Rate
Actual measured rate for subgroup of population
Cohort Study
Assess what, Method(s) for data analysis
- Single risk factor affecting many diseases
- Relative risk to estimate risk
Independent Probability
P = P(A) * P(B)
Abdominal Aortic Aneurysm Screening
- All men above age of 65 with a history of smoking should be screened with an ultrasound
- Aneurysm should be repaired if it is wider than 5 cm
- Also screen 65-75 with a +ve family history
What happens to incidence and prevalence if long-term survival rates for the disease are increasing
- I = no change
- P = increase
Probability of mutually exclusive events
P(A union B)= P(A) + P(B)
Combinations
- Order is not important
- nCr = n! / r! (n-r)!
- nC0= 1
- nCn= 1
Observer-expectancy bias
Definition, association(s), solution(s)
- Researcher’s beliefs affect outcome
- Pygamlion effect (experimenter expectations inadvertently communicated to subjects, who then produce the desired effects
- Double blinded design
False Positive Rate [FPR] (Fall out)
= FP / (FP + TN)
- Type I error
Primary Disease Prevention
A method used to stop the disease before it starts like immunization and behavioral counseling
Experiment (definition)
An activity with an observable result
Normal Distribution
- Mean = Median = Mode
- 1 SD = 34% (68% on both sides)
- 2 SD = 47.5% (95% on both sides)
- 3 SD = 49.9% (99.7% on both sides)
Case-fatality rate
Deaths from cause / # of persons with the disease or cause
Positive Likelihood Ratio [LR+]
= TPR / FPR
Standardized Rate (Adjusted Rate)
Adjusted rate to make groups equal on some factor
Positive Skew Distribution
- Mean > Median > Mode
- Asymmetry with longer tail on right
What happens to incidence and prevalence if recovery from the disease is more rapid than it was 1 year ago
- I = no change
- P = decrease
Breast Cancer Screening
- Mammography between (40-50). Age of maximum benefit is > 50 is clear in decreasing mortality
- Screening is done every 2 years and can be stopped at age of 75
- Selective estrogen receptor modulators (SERMs) like tamoxifen and raloxifene used as a prophylaxis in patients who are free of disease but with multiple first degree relatives with breast cancer (at least 2) results in a 50% to 60 % reduction in breast cancer
- BRCA comes +ve, management remains undetermined
Hypertension Screening
For all above age of 18 every 2 years
Procedure (design) bias
Definition, association(s), solution(s)
- Parts of study do not fit together
- Non-comparable control group
- Random assignment
Case-Control Study
Definition, Advantages, Disadvantages
- A series of cases are identified and a set of controls are sampled from the underlying population to estimate the frequency of exposure in the population at risk of the outcome
- Ad:
- Use smaller groups than cohort, thereby reducing cost
- Can be used to study rare diseases and easily examine multiple risk factors
- Dis:
- Cannot calculate incidence or prevalence, but an odds ratio can be used to estimate a measure of relative risk
- Retrospective data may be inaccurate owing to recall or survivorship biases
Bias (Performing study)
- Recall bias
- Measurement bias
- Procedure bias
- Observer-expectancy bias
Precision and Errors
- Increased precision leads to increased statistical power (1-beta) and decreased standard deviation
- Random error decreases precision
Confidence Interval (CI) interpretation
- If the 95% CI for a mean difference between 2 variables includes 0, then there is no significant difference and H0 is not rejected
- If the 95% CI for odds ratio or relative risk includes 1, H0 is not rejected
- If the CIs between 2 groups do not overlap, that means statistically significant difference exists
- If the CIs between 2 groups overlap, that means no significant difference exists
Probability of complement of an event
P(A’)= 1 - P(A)
Negative Predictive Value [NPV]
= TN / (TN + FN)
- Varies inversely with prevalence
Prevalence
- # of existing cases / Total # of people in a population
- = Incidence * Duration of disease
Conditional Probability
P(B I A) = P(A intersection B) / P(A)
Variance
= (SD)^2
Incidence
of new cases / # of people at risk
Sample space (definition)
Set of all possible outcomes
False Omission Rate [FOR]
= FN / (FN +TN)
Cohort-Study
Definition, Advantages, Disadvantages
- A group of people is assembled, none of whom has the outcome of interest, but all of whom could potentially experience the outcome. Incidence of outcome events is compared in the 2 exposure groups
- Ad:
- The only way to directly determine incidence
- Can be used to assess the relationship of a given exposure to many diseases
- In prospective studies, exposure is elicited without bias from a known outcome
- Dis:
- Time consuming and expensive
- Assess only the relationship of disease to a few exposure factors
- Require many subjects, which makes it difficult to study rare diseases
Measurement bias
Definition, association(s), solution(s)
- Gathering info distorts it
- Hawthorne effect (subject’s behavior is altered because they are being studied)
- Control group/placebo group and using objective, standardized, and previously tested methods of data collection that are planned ahead of time
Negative Likelihood Ratio [LR-]
= FNR / TNR
What happens to incidence and prevalence if contacts between infected persons and non-infected persons are reduced for airborne infectious disease
- I = decrease
- P = decrease
Accuracy and Errors
Systematic error decreases accuracy
Receiver Operating Characteristic (ROC) Curve
- Plotting sensitivity (y axis) against 1 - specificity (x axis)
- The more the area under curve is the more better the test
- More sharp plot on the y axis more better the test isC
Attack Rate
Same as incidence but during an epidemic
Chi-square
- 2 nominal
- Any # of groups
Recall bias
Definition, association(s), solution(s)
- Subjects cannot remember accurately
- Retrospective studies
- Multiple sources to confirm info and decrease time from exposure to follow-up
Specificity (True Negative Rate [TNR])
= TN / (TN + FP)
= 1 - False Positive Rate [FPR]
- SPIN
Trade off between sensitivity and specificity
- Possible cutoff values are ( A: 100% sensitivity, B: practical compromise between sensitivity and specificity, C: 100% specificity)
- Lowering the cutoff value (from B to A) will increase FP and decrease FN thereby increases sensitivity and NPV while decreases specificity and PPV
- Raising the cutoff value (from B to C) will increase the FN and decrease the FP thereby increases specificity and PPV while deceases sensitivity and NPV
Negative Skew Distribution
- Mean < Median < Mode
- Asymmetry with longer tail on left
Case-Control Study
Assess what, Method(s) for data analysis
- Many risk factors for a single disease
- Odds ratio to estimate risk
What happens to incidence and prevalence if contacts between infected persons and non-infected persons are reduced for non-infectious disease
- I = no change
- P = no change
Events (definition)
Subset of the sample space
Matched pairs t-test
- 1 interval and 1 nominal
- 2 groups, linked data pairs, before and after
Cross-Sectional Study
Assess what, Method(s) for data analysis
- Association of a risk factor and a disease
- Chi-square to assess association
Accuracy [ACC]
= (TP + TN) / Total population
Lipid Screening
- Cholesterol and LDL measurement is recommended for healthy individuals when:
- Men are above age 35
- women are above age 45
- It is recommended for all patients with diabetes, hypertension, coronary artery disease or equivalent like carotid, aortic or peripheral artery disease
Standard error of th mean (SEM)
= SD / square root of n
Crude Mortality Rate
Deaths / population
Measures of dispersion
- Standard deviation
- Standard Error of the Mean (SEM): it is decreases as sample size (n) increases
What happens to incidence and prevalence if new effective treatment is initiated
- I = no change
- P = decrease
Pearson Correlation
- 2 interval
- Is there a linear relationship?
- r is always between -1 and +1. The closer the absolute value of r to 1, the stronger the linear correlation between the 2 variables
- Positive r value means positive correlation (variables are directly related)
- Negative r value means negative correlation (variables are inversely related
- Coefficient of determination = r^2 (value that is usually reported)
Prostate Cancer Screening
- 25% of patients with prostate cancer have normal PSA levels
- 25% of those with elevated PSA levels have no prostate cancer
- No benefit of PSA on mortality
- Do it if the patient asks for it
Type I error (alpha)
- Also known as false-positive error
- It is judged by p value. if p < 0.05, then there is less than a 5% chance that the data will show something that is not really there
Cause-specific mortality rate
Deaths from cause / population
Bias (Interpreting results)
- Confounding bias
- Lead-time bias
- Late-look bias
Attributable Risk (AR)
= [a/(a+b)] - [c/(c+d)]
Secondary Disease Prevention
The detection of the disease early in its course to reduce the associated morbidity and mortality like screening tests
Cervical Cancer Screening
- Pap smear is done from 21 to 65 years of age every 3 years
- Adding HPV testing to pap smear increases interval to 5 years
- HPV vaccine is routine for all women between ages of 11 and 26
- Chlamydia screen for women 15-25 years old
Indications of Influenza vaccine
- Everyone yearly
- Healthcare workers
- Pregnant women
False Negative Rate [FNR] (Miss Rate)
= FN / (FN + TP)
- Type II error
Confounding bias
Definition, association(s), solution(s)
- Unanticipated factors obscure results
- Hidden factors affect results
- Multiple/repeated studies, Cross-over studies (subjects act as their own controls), Matching (patients with similar characteristics in both treatment and control groups), Restriction and Randomization
Lead-time bias
Definition, association(s), solution(s)
- Early detection confused with increased survival
- Benefits of screening
- Measure “back-end” survival (adjust survival according to the severity of disease at the time of diagnosis)
