# BEHAVIORAL SCIENCE

rename
eesohbel's
version from
2016-08-11 01:49

## Types of study

Question | Answer |
---|---|

case control | compares a group of people with a disease and without a disease. looks for risk factor. almost always retrospective. |

case series | descriptive study that tracks patients with a known condition |

cross-sectional study | assesses frequency of a disease at a point in time |

prospective cohort study | compares a group with a given risk factor to a group without the risk factor to see if there is an increased likelihood of developing disease. can be either prospective or retrospective |

cross over study | subjects are randomly allocated to a sequence of two or more treatments given consecutively. A washout (no treatment period) is often added between treatment intervals to limit cofounding effects |

Cross sectional study measures | disease prevalence |

Case-control study measures | odds ratio (small number of people with rare disease) |

cohort study measures | relative risk (bigger number) looks to see if increased exposure increased likelihood of disease |

odds ratio can be a close approximation | relative risk |

cohort study best for determining | incidence of a disease. comparing incidence of disease in 2 populations allows for calculation of relative risk |

cast control study best for | comparing risk factor frequency |

cross section best for | comparing disease prevalence |

critical distinction btwn case control and cohort | case control determines outcome first and looks for associated risk factor; cohort ascertains risk and then outcome |

## Clinical trial

Question | Answer |
---|---|

triple blinding | refers to additional blinding of researchers analyzing the data |

phase I study sample | small number of healthy volunteers |

purpose of phase I study | is it safe |

phase II study sample | small number of patients with disease of interest |

phase II study purpose | does it work |

phase III study sample | large number of people randomly assigned either to the treatment or best available treatment |

phase III study puprose | is it good or better |

phase IV study sample | post marketing surveillance of patients after treatment is approved |

phase IV study purpose | can it stay |

## Evaluation of diagnostic tests

Question | Answer |
---|---|

sensitivity | percentage of people with a disease who test positive for the diseae |

specificity | percentage of people with a disease who test negative |

sensitivity | TP/ (TP + FN) |

specificity | TN/(TN + FP) |

what is PPV | proportion of positive test results that are true positive |

PPV equation | TP/(TP+FP) |

NPV equation | TN/(TN+FN) |

what is NPV | probability that person actually is disease free given a negative test result |

low prevalence | high NPV and low PPV |

incidence | number of new cases/number of people at risk |

prevalence | number of existing cases/number of people at risk |

prevalence usually greater than incidence | chronic diseases |

prevalence = incidence | short duration of a disease |

case fatality rate | number of fatal cases/ total number of people with the disease |

## Quantifying Risk

Question | Answer |
---|---|

what is ODDS ratio | odds that the group with the disease was exposed to a risk factor (a/c) divided by the odds that the group without the disease was exposed (b/d) |

ODDS ratio | (a/c)/(b/d) |

what is relative risk | risk of developing disease in exposed group divided by risk in unexposed group |

RR | (a/(a+b))/(c/(c+d)) |

what is atributable risk | the proportion of disease occurences that are attributable to the exposure |

AR | (a/(a+b))-(c/(c+d)) |

Atributable risk percent | (RR-1)/RR |

what is absolute risk reduction | the difference in risk (not the proportion) attributable to intervention as compared to a control |

ARR | (c/(c+d))-(a/(a+b)) |

number needed to treat | 1/ARR |

number needed to harm | 1/AR |

increased precision | decreased standard deviation and increased statistical power |

external validity | results hold true for populationn at large |

internal validity | results hold true w/in the study |

berkson bias | study population selected from hospital is less healthy than general population |

lead-time bias | early detection is confused with increased survival |

confounding | relationship between exposure and outcome is distorted due to third variable that is associated w/both exposure and outcome |

effect modification | slightly different. variable actually enhances magnitude of risk factor for causign disease. example is estrogen and smoking |

hazards ratio | ratio of an event occurring in treatment versus non treatment arm. ratios higher than 1 indicate treatment arm had higher rate of events |

## Miscallaneous

Question | Answer |
---|---|

one standard deviation | 68% |

two standard deviations | 95% |

three standard deviations | 99.7% |

tail to right | positive skew |

positive skew | mean>median>mode |

tail to left | negative skew |

negative skew | mean |

null hypothesis | hypothesis of no relationship or difference |

alternative hypothesis | there is a relationship |

if p value is less than .05 | reject the null and accept the alternative |

type II error | occurs when researchers accept null hypothesis when it is false. Aka a relationship exists! |

type I error | when researchers reject the null, but the null is true. There is no relationship! |

power (1-B) | probability of rejecting a null hypothesis when it is truly false. Typically set at 80% and depends on sample size and difference between outcomes |

CI | width of CI determines power of study |

the tighter the CI | more precise the result |

if the CI between 2 mean variables includes 0 | there is no significant diffference and accept the null |

if CI for odds ratio or relative risk includes 1 | accept the null. there is no relationship |

if CI for relative risk is greater than 1 | there is increased risk |

if CI for relative risk is below 1 | there is decreased risk |

if CI between 2 groups does not overlap | statistically significant difference exists |

if CI between 2 groups does overlap | usually no significant difference exists |

t-test | checks differences between means of two groups |

ANOVA | checks difference between means of three or more groups |

CHI squared test | checks difference between categorical variables, percentages or porportions |

r=1 | perfect correlation between two variables |

r=-1 | no relationship between two variables |

r>0 | there is a positive relationship between two variables |

r<0 | there is a negative relationship between two variables |

type II error | beta |

type I error | alpha |

## Section

Question | Answer |
---|---|

PPV and prevalence association? | PPV increases with increasing disease prevalence and decreases with decreasing disease prevalence |

NPV and prevalence association? | NPV decreases with increasing disease prevalence and increases with decreasing disease prevalence |

type I error | alpha. rejecting the null hypothesis when it is actually true |

type II error | beta. accepting the null hypothesis when it is actually false |

P < alpha | ideal, this is statistically significant and means rejecting the null |

P > alpha | fail to reject the null hypothesis |

R = 1 | perfect correlation |

R = 0 | no correlation at all |

R > 0 | + correlation so positive slope |

R < 0 | - correlation so negative slope |

1 standard deviation | 68% |

2 standard deviations | 95% |

3 standard deviations | 99.7% |

how does a larger sample size affect power? | increases power |

how does a larger sample size affect Confidence Interval? | decreases confidence interval |

precision is akin to | reliability |

accuracy is akin to | validity |

what factors make something reproducible? | precision and reliability |

what factors make something compared to the gold standard | accuracy and validity |

standard error of mean | SD / sq rt of sample size |

confidence interval | mean +/- Z (SEM) |

Z value? | 1.96 for 95% ; 2.58 for 99% |

incidence | number of new cases / number at risk |

prevalence | number of existing cases / total population |

positive skew | mode < median < mean (R skew) |

negative skew | mode > median > mean (L skew) |

Question | Answer |
---|---|

which study type looks at odds ratio? | case control |

what does case control find? | odds ratio |

what studies find relative risk? | RCT or cohort |

what do cohort studies find? | relative risk |

what do RCT studies find? | relative risk |

what do cross sectional studies find? | disease prevalence |

what studies find disease prevalence? | cross sectional |

snap-shot in time | cross sectional |

no controls | case series |

when prevalence is low, what happens to OR and RR? | they become more equal |

how do you increase precision? | decrease standard deviation and increase statistical power |

most affected by outliers? | mean |

least affected by outliers? | mode |

primary disease prevention | prevent (vaccination, condoms) |

secondary disease prevention | screening (pap) |

tertiary disease prevention | treatment (chemo) |

phase I trial | toxicity, safety, pharm dynamics and kinetics |

phase II trial | efficacy, dosing, adverse effects |

phase III trial | compare to standard |

phase IV trial | long term adverse effects |

NNT | 1/ARR |

NNH | 1/AR |

OR | (a/c) / (b/d) OR ad/bc |

RR | a/a+b / c/c+d |

AR | a/a+b - c/c+d |

ARR | c/c+d - a/a+b |

failure of precision | random error |

failure of accuracy | systematic error |

reproducibility of test results | precision aka reliability |

the test measures what it is supposed to | accuracy aka validity |

disease specific mortality | number of fatal cases/total population |

case fatality rate | number of fatal cases/# of people with disease |

external validity | results hold true for population at large |

internal validity | results hold true within study |

## BIAS

Question | Answer |
---|---|

study population selected form hospital is less healthy than general population | berkson bias (selection bias) |

study population is healthier than general population | healthy worker effect (selection bias) |

participating subjects differ from nonrespondents in meaningful ways | non-response bias (selection bias) |

awareness of disorder alters recall by subjects | recall bias |

common in retrospective studies | recall bias |

information is gathered in a way that distorts it | measurement bias |

subjects in different groups are not treated the same | procedure bias |

researchers belief in the efficacy of treatment changes the outcome | pygmalion effect (observer-expectancy bias) |

factor is related to both exposure and outcome | confounding bias |

early detection is confused with increased survival | lead time bias |

reduce selection bias? | randomization |

reduce recal bias? | decrease time from exposure to follow up |

reduce measurement bias? | use standardized method of data collection |

reduce procedure bias or pygmalion effect? | blinding and placebos |

reduce confounding bias? | multiple repeated studies, crossover studies, matching |

reduce lead time bias? | measure back-end survival |

change in behavior in a study group resulting from knowing they are being observed | hawthorn effect |