Shit - Behavioural Flashcards
Measurement for case-control
OR = ad/bc
= have it and risk * not and don’t / have it and don’t * don’t have it but do
Measurement for cohort
RR = (a/a+b)/ (c/c+d)
= risk of disease when you have exposure / risk of disease when you don’t have exposure
Phases of clinical trials
I = safe (healthy ppl) II = work - efficacy, AE, dose III = good/better - common AE IV = stay - rare/LT AE
sensitivity =
= 1 - FN
specificity =
= 1 - FP
AR =
AR = (a/a+b) - (c/c+d)
= risk of disease when you have exposure - risk of disease when you don’t have exposure
= (RR - R unexposed)/RR
RRR =
RRR = 1-RR
ARR =
ARR = (c/c+d) - (a/a+b)
= risk of disease when you don’t have exposure - risk of disease when you have exposure
NNT
NNT = 1/ARR
NNH
NNH = 1/AR
Precision
= reliability
decrease with random errors
Accuracy
= validity
decrease with systematic errors
Berkson bias vs. healthy worker bias
Berkson = study population from hospital is sicker than general population
HWB = study population is healthier than the general population
lead-time vs latent period
lead time = find earlier so think it increases survival. Decrease effect with “back-end” survival measures
Latent period = time between intervention and effect. Measuring too soon could prevent you from seeing the result because it hasn’t happened yet
SD %s
1SD = 68% 2SD = 95% 3DS = 99.7%
Positive skew =
Tail on positive side; Mean > median
Nevative skew =
Tail on negative side; Mean
Power =
1-beta
= correctly rejecting the null
Increase power (decrease beta) via:
- increased “n”
- increased expected effect size
- increase precision of measurement
Type I error
Falsely rejecting the null (FP error)
aka saw a difference when there is none
Type II error
Failing to reject a false null (FN error)
aka didn’t see a difference when there really is one
CI:
If CI includes there numbers, then fail to reject null:
Mean difference: 0
OR or RR: 1
Overlap between 2 groups’ CIs
T-test
comparing 2 groups, numerical data
ANOVA
Comparing ≥3 groups, numerical data
Chi-squared
Comparing ≥2 groups, categorical data (i.e. ethnicity)
Quarternary prevention
Identify patients at risk for unnecessary treatments
Medicare vs medicaid
medicare = old, disabled, ESRD Medicaid = very low income
Parts of medicare:
A = hospital B = basic clinic stuff C = A+B via external company D = drugs
Therapeutic privilege
Withhold patient information if it will severely harm them or undermine their decision-making capacity
i.e. woman tells you she will kill herself if she has breast cancer because she watched her mom die from it you find a lump during exploratory surgery and take it out without permission
Transfusion on minors of jahovas parents?
YES
Power of attorney hierarchy
pt picks someone
pt can revoke this even if they are not competent
Surrogate decision makers
spouse > adult child > parents > adult siblings > other relative
APGAR
Appearance (colour) Pulse (100) Grimace (reflex to stimulus) Activity (tone) Respiration (quality)
Object permanence age
9 months
Positive likelihood ratio
sens/1-spec
Negative likelihood ratio
1-sens/spec
Parameters affected by disease prevalence
PPV and NPV
ex: higher prevalence = more TP = higher PPV
higher prevalence = less TN = lower NPV
most resistant measure of central tendency to outliers
mode
Least vs greatest financial risk methods of doctor reimbursement
Least risk = fee for service: get paid for every test, so order more tests
Highest risk = capitation: get paid a set amount per patient and insure the costs for tests, so more primary/secondary care to prevent diseases and catch them early to avoid expensive tests and surgeries
Most preventable cause of disease and death in the USA
SMOKING!
MCC cancer in women vs MCCD cancer in women
MCC = breast > lung > colon MCCD = lung > breast > colon
matching is used to decrease what type of bias:
confounding
attrition bias is a type of what form of bias
Selection bias
(anything that causes a systematic difference between the groups terms of treatment response or prognosis
Crossover study design:
case control where you switch the groups half way
Limitation = effect from one treatment carrying over. Limit this with a washout period
Ecological study
Use populations, not people
Power =
= the study’s ability to find a difference when one does exist
= 1-beta
(so beta = the probability of finding no difference when there is one i.e. failing to reject a false null)
Hawthorne effect =
Participants altering their behaviour when they know they are being studied
