Bio stats

Question 1

60% phlebitis in one group
75 % in the other

what is the absolute risk reduction?
Relative risk reduction?
Number needed to treat???

Answer 1

1. ARR = 15%
2. change (control group -intervention) /control group risk = 0.15/0.75 = 0.20
3. NNT = 1/ARR = 1/0.15 = 6.666 = 7!!!!

Question 2

positive likelihood ratio?

negative likelihood ratio?

in a study where 500 patients with thyroid cancer 800 patients with no thyroid cancer

what is the positive likelihod ratio associated with a serum midkine level of over 320.5. ??

note at this level, this included 375 of patients thyroid cancer testing positive and 120 patients without thyroid cancer.

Answer 2

+ve LR = probability diseased patient positive/probability of non diseased person testing positive

positive ratio, positive ratio, diseased, non diseased!!

-ve LR = probability of diseased patient negative / pobability of non diseased patient negative

negative, negative, diseased, non diseased

probability positive in diseased = 375/500 = 0.75

probability positive in non diseased = 120/800 = 0.15

+veLR = 0.75/0.15 = 5

Question 3

The negative predicitive value equation?

positive prediictive value?

if the NPV in population A is 98% and 85% in population B for alzheimers what does this mean?

Answer 3

NPV = probabilty truly dont have disease given a negative results = TN/TN+FN

PPV = probabilty you have the disease, given a positive results = TP/TP+FP

only gives information on disease prevalence. as (p)revalence increases, PPV increases and the opposite for NPV. that means population A has a low disease prevalence and so the proportion of patients without alzheimers disease is higher in population A!!!

NPV and PPV doesnt give any info on sensitivty and specificity

Question 4

study to determine risk factors for multidrug resistant TB
identify patients with this and patients with sensitive TB and look through records to identify risk factors for multi drug resistant TB such as HIV

then compare frequency of these risk factors amongst both groups

type of study?

Answer 4

case control!! - these are retrospective studies that start with the outcome/disease and non diseased groups, then look back to determine the risk factors. compare disease frequency

NOT retrospective cohort as you identify the risk first eg HIV and then look to see if they developed resistant TB. compare disease incidence

prospective cohort = risk factor, group without risk factor -> incidence

cross sectional = risk factor, group without risk factor -> prevalance

Question 5

Odds ratio>1 = increased risk from exposure

if confidence interval includes 1. (ONE) = statistically insignificant

compared to normal pregnancy, in a patient with ehlers dahnlos, if the Odds ratio is 3.5 for an amniotomy, confidence interval 2-6.2

what does this mean?

Answer 5

change = (3.5 - 1)/1 = 2.5 = 250% increased odds of amniotomy

Question 6

people with heart failure are divided into 2 qualitative/ categorical groups bases on type of care ((telemonitoring v usual care)

end point measured is death or readmission % (qualitative) in each roup

what statistical test can be used for this?

other examples of qualitiative variables = blood type

quantitative = temp, glucose level

Answer 6

chi square test!!!
evaluates relationship between 2 categorical variables and a depended categorical variable (outcome) is measured. requires a categorical variable used to divide participants into 2 separate groups contrast to t test where groups already independent

NOT anova as it compares the mean of a quantitative variable in at least 3 groups. eg study comparing serum ferritin levels in children, adolescents and adults

NOT correlation analysis as is used for 2 quantitative variables, eg hours of sleep and irritability score

NOT meta-analysis as it is a quantitative statistical technique used to combine and analyse data form several studies

NOT 2 sample T test as it compares the mean of quantitative variable between 2 independent groups eg comparing serum ferritin levels between males and females

Question 7

effect of concentration of CSF tau on brain atrophy (measured quantitatively using 2 outcomes = ventricular volume, and total brain volume)

what type of test?

Answer 7

linear regression!

testing effect of >/=1 explanatory variable (quant or qual) on 1 quantitative dependent variable (outcome)

NOT anova! as it is used to compare MEANNN of a quantitative dependent variable in several independent qualitative groups eg mean serum triglyceride levels in patients with low normal or high serum uric acid levels

Question 8

population level data eg county, state, country = ecological study

Question 9

if risk of hyperkalemia with follow up diuretic use =

hazard ratio = 0.41 and p value = 0.006

vs risk of hyperkalemia with follow up serum potassium level of 4.7 =
hazard ratio = 7.25 p value = 0.007

then the latter significantly increases the chances of hyperkalemia!!. look at the numbers not just the p values

review confidence intervals as well!!

Question 10

you are randomising to 3 different interventions/multiple (metoprolol, ramipril, amlodipine)

and then you are studing at least 2 variables. eg 2 different BP outcomes

what type of RCT study is this?

Answer 10

Factorial design!!!

not crossover study = one group gets a treatment, another group randomised to a separate one, and then they switch treatments/interventions

not parallell study = intervention one group, intervention/placebo the other. no multiple variables examined

not cluster analysis

Question 11

carbonated beverages consumed in each household (risk factor)
and estimated PREVALENCE of obesity in the household (outcome)

most likely study?

Answer 11

observational study!!! -> specifically a cross sectional study, which involves looking at data at a specific snapshot in time.

other observational studies = cohort, case control, case

contrast experimental studies which assess impact of an intervention. eg RCTS

Question 12

learn the diffeent types of bias. on your phone! eg berkson bias, prevalence/neyman bias

Question 13

If 95% CI for mean difference includes 0 = no significant difference

If 95% CI for relative risk or odds ratio includes 1 = no significant difference.

Question 14

participants change behaviour upon awareness of being observed. what type of bias is this?

Answer 14

hawethorne bias!!

Question 15

how do you calculate relative risk?
how do you calculate relative risk reduction?

absolute risk reduction?

NNT?

high yield!!

Answer 15

RR = rate treatment/rate control

RRR = 1 - RR

ARR = control ratee - treatment ratee

NNT = 1/ARR

rate indicates you are using proportions

Question 16

traditional schedule = 1,200 patient days, 60 non intercepted serious errors

intervention schedule = 1,250 patient days, 25 non intercepted serious errors

approximate proportion of decreased risk in non intercepted serious errors per patient day among interns in intervention schedule vs those in traditional schedule

Answer 16

RRR = control rate- treatment rate/ control rate

60/1200 = 0.05

25/1250 = 0.02

(0.05 -0.02)/0.05

= 0.6!!!!

Question 17

a study comparing medical therapy alone to medical therapy + CABG intervention

at the end some people in the medical therapy group underwent CABG (they had deteriorated to that point) and they want to add this to the group that had medical therapy + CABG for analysis

what type of bias would this introduce?

Answer 17

selection bias!!!

(when the treatment regimen selected for a patient depends on the severity of their disease = a type of selection bias called succeptibility bias!!! )

counfounding by indication may occur!!

surgery patients in medical therapy group may have had underlying confounders causing their deterioration and need for surgery

Question 18

sometimes in an abstract, if they
put *** next to values and the bottom says **p<0.01, this means that the values with the stars are the only statistically significant values!!!

Question 19

blinding is not always possible in RCTs!! such as diet interventions and surgeries intervention, in these cases they will not introduce bias/study results should not be avoided because of bias

Question 20

a confidence interval of 95% captures 95% of the distribution and has an alpha significance level of 1-0.95 = 0.05

confidence interval of 99% captures 99% of the distribution and has an alpha significance level of 1-0.99 = 0.01

so going from 95% CI to 99% CI = larger or wider data set

if a question says mean is 107 grams with a confidence interval 95% of 104-110, what is the most likley lower and upper limits for a confidence interval of 99% -> the answer would be the one with a wider lowe and upper limit eg 102 - 112

equally CI of 90% = alpha of <0.1

Question 21

an association between a risk factor and an outcome is more likely to be causal if its strength increases as as the exposure level increases

eg higher reduction in hba1c the more exercise you do.

likelihood of a causal relationship is not greatly increased because some confounders are controlled for because some may not have been controlled for!

Question 22

increasing (P) revalence increases (P) PV. but decreases NPV

oppositive effect when you decrease prevalence

Question 23

CAREFUL!! abstract with 3 pages with go to page 2 and go to page 3 yellow button

Question 24

NNT?

Answer 24

1/ARR

Question 25

randomization is done to control for what?

Answer 25

confounding

Question 26

a question such as % of maternal deaths in the breastfeeding group that can be attributed to breast feeding is asking for attributable risk percentage = (risk exposed - risk non exposed)/risk exposed x 100

Question 27

Research using information from patients medical charts only require what type of approval?

Answer 27

IRB approval!! but not informed consent because it is secondary analysis of existing data (I)RB = important!

Question 28

if a study has a relative risk of 2.01 and a 95% confidence interval of 0.8-3.1 -> that means p >0.05 as its not significant

Question 29

lower levels of cognitive dysfunction tend to be associated with lower levels of pain. but weak association what does this say about r? R is positive, but closer to 0 than 1 if r>0, then both levels increase, or decrease together. can draw it out

Question 30

if a power analysis has been conducted, that means sample size would be okay, even if small, and will not affect validity

Question 31

BP recorded 120 and and 100, then you record a value of 240 which one is likely to increase the most? mean median or mode?

Answer 31

mean! it is the most sensitive to outliers

Question 32

anxiety screening tool 86% sensitive, 78% specific on a sample of 400 students 25 % have anxiety. calculate NPV.

Answer 32

draw a table of anxiety no anxiety, positive test negative test and go from there. 75% x 400 = 300 dont have anxiety speccficity = 78% x 300 = 234 will test negative so thats how you do the calculations answer at end = 0.94

Question 33

if youre trying a new therapy + mycophenalate and also a new therapy + cyclophophamide but the latter group did not acheive significance like the first group. if they had more advanced disease cases at baseline or smallere sample size etc, you cant say only group 1 is recommended. recommendations based on subgroup analyses will be suboptimal also if they give you odds of acheiveing a response of 55% and 43% achieved a response on new therapy + MF vs 32% on standard therapy, the absolute diffrence is 43-32 = 11% and much lower!! so the answer was recomend new combination, but absolute difference in renal response is likely lower than 55%

Question 34

when transitioning between care homes, interventions that target pharmacy personell! is most effective at preventing adverse drug events

Question 35

what are the different phases of clinical trials and what does each one show?

Answer 35

SWIM! 1. safe? 2. work? 3. good or better? (improvement) 4. withdrawn from Market? (rare, long term adverse effects)

Question 36

if you are comparing a new treatment to placebo or standard treatment, what level of an RCT is this?

Answer 36

phase 3 trial !!

Question 37

a study where patients selected that have cancer, and controls where patients admitted to the hospital for other reasins. type of bias that may occur?

Answer 37

selection bias!!! berkson bias

Question 38

odds ratio is a good approximation of the relative risk in a study if?

Answer 38

the outcome is uncommon in the population eg incidence of TSS!

Question 39

what does pre-protocol analysis involve?

Answer 39

analysing data on only participants that completed the study. dont include dropouts!! this may overpredict results that may happen in the real world clinical setting opposite of intention to treat analysis

Question 40

type one vs type 2 error? decreasing sample size increases likelihood of which error?

Answer 40

1 = false positive type 2 = false negative type 2!

Question 41

NNT calculation based on abstract! -> its on your phone for review !

Question 42

case control study looking at association between bladder cancer and occupational exposure to fuel. so selecting cases with bladder cancer. who will be the controls?

Answer 42

participants WITHOUT bladder cancer, irrespective of their occupational exposure status

Question 43

if one study predicts the prevalence of a condition is 32% and the other 25% the study results do not contradict each other if one value falls within the confidence interval of the other

Question 44

equation for accuracy? if you are not given whole population, assume whole population is 100. check picture on phone on how to construct table and do the calculation!

Answer 44

accuracy = (TP + TN)/entire population it is basically the probabilty that someone is correctly classified

Question 45

how to construct odds ratio table and do calculation!!! -> check your phone for practice question! you alwasy construct the table as + - + -

Question 46

loss to follow up poses the risk of what type of bias?

Answer 46

selection bias!!!! specifically attrition bias

Question 47

power increases as significance level increases power increases as effect size increases!! eg if deifference in mean serum triglyceride levels is higher than expected power increases as sample size increases power increases as outcome variability decreases

Question 48

a confidence interval is created by taking a sample value (EG DIFFERENCE IN SAMPLE PERCENTAGES). and adding and subtracting a margin of error. so lets say percentage of deaths in hospital A is 3.6% and 5.1% in hospital B and values are significant the confidence interval is likely (-2.5% to -0.5%) because it has to include 3.6 - 5.1 = -1.5 the value has to be at the CENTER of the interval. so it is not (-4.5% to -0.3%)

Question 49

questions x 2 on your phone about relative risk reduction !

Question 50

a reciever operating characteristic curve plots sesntivity (y) against (1-specificity AKA false positives (x) ) on such a curve, the point that has the best diagnostic accuracy is the one that is high up on the y axis in terms of sensitivity but <- low on the x axis in terms of number of false positives

Question 51

if an iL-27 value of >/5 has a 95% specificity for blood stream infections 5% of patients without blood stream infections will be incorrectly identified. False positive = 100% - specificity

Question 52

case control studies can only evaluate 1 outcome!! but can evaluate for different risk factors!!

Question 53

studies with larger sample sizes have a lower probability of a type 2 error

Question 54

in an ecological study which deals with population level data, you can not make conclusions about causality, or conclusions for individuals/ at the individual level. you can only infer on associations

Question 55

in the study comparing kalaxin and warfarin, which of the following specific bleeding risks were most similar between the two groups -> answer is whichever one with the hazard ratio closest to 1 in this case it was major bleeding as hazard ratio = 0.96

Question 56

what can help confirm proper randomization?

Answer 56

table of baseline characteristics of participants

Question 57

Smaller sample size = higher SEM