Biostatistics Flashcards

1
Q

alpha

A

The selected threshold for rejecting the null hypothesis, typically a = 0.05

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
2
Q

p-value

A

calculated value from the data that is compared to the alpha (0.05)

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
3
Q

Confidence Interval

A

CI = 1 - a

p < 0.05 shows a “95% Confidence” that the conclusion from the results are correct

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
4
Q

Difference Data w/ Confidence Interval

A

Statistically Significant if CI Range does NOT INCLUDE ZERO (0)

95% Confident that the value lies within that range. If it includes ZERO, then there is NO difference (NOT SIGNIFICANT)

ex. Treatment = 46, Placebo = 8
Difference (95% CI) = 38 (18-58)

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
5
Q

Ratio Data w/ Confidence Interval

A

Ratio Data = Relative Risk, Odds Ratio, Hazard Ratio

Statistically Significant if CI Range does NOT INCLUDE ONE (1)

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
6
Q

Type 1 Error

A

False-Positives: null hypothesis was rejected, when it should have accepted

alpha = Type 1 Error

Ex. a = 0.5, if p < 0.5 then the probability of a Type 1 Error is < 0.5

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
7
Q

Type 2 Error

A

False-Negatives: null hypothesis was accepted, when it should have been rejected

beta = Type 2 Error; set by investigators (10-20%; 0.1-0.2)

Power Analysis used to determine sample size needed to detect a true difference (reduce Type 2 Error)

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
8
Q

Power

A

Probability that a test will REJECT the Null Hypothesis CORRECTLY (Power to AVOID a Type 2 Error)

Power = 1 - b

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
9
Q

Risk

A

in Treatment with ADR / # in Control with ADR

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
10
Q

Relative Risk

A

Tx Risk / Control Risk

RR > 1: Higher Risk in Tx
RR < 1: Lower Risk in Tx

“RR of Tx = 57%. Tx were 57% AS LIKELY to have ADR as Control”

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
11
Q

Relative Risk Reduction

A

How much Risk is REDUCED in the Treatment compared to Control

RRR = (Risk Control - Risk Tx) / Risk Control

RRR = 1 - RR

“RRR of Tx = 43%. Tx were 43% LESS LIKELY to have ADR as Control”

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
12
Q

Absolute Risk Reduction

A

ARR = % Risk Control - % Risk Treatment

“ARR = 12%. For every 100 patients Treated, 12 FEWER patients will have an ADR”

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
13
Q

Number Needed to Treat (NNT)

A

NNT for 1 patient to BENEFIT (avoid ADR)

NNT = 1 / (Risk Control - Risk Tx)

NNT = 1 / ARR

ROUND UP

“For every 9 patients Treated, 1 patient will BENEFIT (avoid ADR)”

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
14
Q

Number Needed to Harm (NNH)

A

NNH for 1 patient to have ADR

NNH = 1 / (Risk Control - Risk Tx) Absolute value

NNH = 1 / ARR Absolute value

ROUND DOWN

“1 additional case of ADR for every 90 patients taking Tx instead of Control”

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
15
Q

Odds Ratio

A

Odds of an Outcome occurring WITH Exposure, compared to Odds of an Outcome occurring WITHOUT Exposure (Retrospective)

Exposure / Outcome (+) / Outcome (-)
Present / A / B
Absent / C / D

A = (+) Outcome WITH Exposure
B = (-) Outcome WITH Exposure
C = (+) Outcome WITHOUT Exposure
D = (-) Outcome WITHOUT Exposure

OR = AD / BC

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
16
Q

Hazard Ratio

A

Used in Survival Analysis (similar to Relative Risk)

HR = % Treatment / % Control

17
Q

Odds Ratio & Hazard Ratio Interpretation

A

OR or HR > 1: Event Rate in Tx is HIGHER than Event Rate in Control

HR of 2 indicates there are TWICE as many ADRs in the Treatment Group

18
Q

Types of Statistical Tests: Continuous Data

A

Continuous Data = increase by same amount

1) Interval Data = Zero does not mean “Nothing” (Celsius)
2) Ratio Data = meaningful Zero (BP)

Types of Tests [T-tests]
1) One-sample t-test: One Sample compared to General Population

2) Paired t-test: One Sample used for pre/post-Tx (Patient is their own Control)
3) Student t-test: Two Samples (Tx vs Control)
4) ANOVA = 3 or more groups

19
Q

Types of Statistical Tests: Discrete (Categorical) Data

A

Discrete (Categorical) Data:

1) Nominal Data = gender, race, mortality
2) Ordinal Data = NYHA Class, Pain Scale

Types of Tests:
Comparing between Two Groups (Tx vs Control):
1) Chi-square Test
2) Fisher’s Exact Test

20
Q

Questions to ask when interpreting Sensitivity/Specificity:

A

1) If the result is POSITIVE, what is the probability of ACTUALLY having the disease?
2) If the result is NEGATIVE, what is the probability of TRUELY NOT having the disease?

21
Q

Sensitivity

A

True Positive: % a test identifies a patient WITH the Dx

Sensitivity = # Tested Positive / # True Positive

22
Q

Specificity

A

True Negative: % a test identifies a patient WITHOUT the Dx

Specificity = # Tested Negative / # True Negatives

23
Q

Intension-to-Treat analysis

A

Includes ALL patients in each group, even if the patient did NOT complete the trial according to study protocol

*Has real world estimate of treatment effect

24
Q

Per protocol analysis

A

Only for patients who COMPLETED according to study protocol

25
Q

Forest Plots

A

Confidence Intervals for:

1) Difference Data = NOT Significant if crosses ZERO
2) Ratio Data = NOT Significant if crosses ONE