Stats And Education

Question 1

Curriculum development initial steps

Answer 1

Problem identification
General needs assessment
Targeted needs assessment
Develop goals and objectives

Question 2

Goals vs objectives

Answer 2

Goal: broad educational aims, statements of purpose
Objective: need to be precise and measurable

Question 3

5 elements of educational objective

Answer 3

1. Who (target audience)
2. Will do (verb)
3. How much/how well (adjective that describes performance)
4. Of what (noun that describes a criterion)
5. By when (noun that describes conditions of performance)

At the end of the didactics session (5), the medical students (1) will demonstrate knowledge of 3 diagnostic presentations of ADHD (4) by answering post didactic questions (2) with a passing rate of higher than 80% of the questions (3)

Verbs used in (2) should be clear and up to little interpretation
Good: list, explain, answer
Bad: know, understand, appreciate

Question 4

3 different types of learner objectives (and verbs/phrases associated with each)

Answer 4

1) Cognitive/knowledge: concerned with the cognitive expertise of the learner. Hierarchical cognitive capacity from knowledge (“list”), to comprehension (“explain”), to application (“illustrate”), to analysis (“predict”) to synthesis (“propose”) to evaluation (“validate”)

2) attitudinal/affective: focused on attitudes of learner. Objectives may include phrases such as “rate as important” or “rate as valuable”

3) psychomotor: skills/competence based or behavioral/performance based (procedure, interviewing technique) “demonstrate”, “imitate” “perform”

Question 5

Hidden curriculum

Answer 5

Transmission of the culture of the workplace (norms, values, attitudes) to learners through observations of individual practitioners and group interactions

Question 6

Components of feedback

Answer 6

Timely, specific, aligned with learner goals
Delivered in safe space
Incorporating areas of strength and growth to identify next steps for learning

Question 7

Problems and challenges with feedback

Answer 7

Perceived lack of time
Fear of delivering negative feedback
Lack of specificity

Question 8

Sensitivity

Answer 8

How often a positive task correctly identifies those who have the disease. Helps rule out disease.

Sensitivity = true positive / (true positive + false negative) = true positive / total number with disease

Question 9

Specificity

Answer 9

How often a negative result correctly identifies those who do not have the disease. Helps rule in disease.

Specificity = true negative / (true negative + false positive) = true negative / total without disease

Question 10

Positive predictive value

Answer 10

Probability of disease and patient with positive test

Positive predictive value = true positive / (true positive + false positive) = true positive / all of the positives

Question 11

Negative predictive value

Answer 11

Probability of not having disease if test is negative

Negative predictive value = true negative / (true negative + false negative) = true negative / all of the negatives

Question 12

Positive and negative likelihood ratios

Answer 12

Positive likelihood ratio = sensitivity / (1- specificity)

Negative likelihood ratio = (1-sensitivity) / specificity

For example, if positive likelihood ratio is nine, then a positive test is seen nine times more in patient with disease than patient without disease

Question 13

Which depends on prevalence, and which is independent of prevalence?
-Predictive values
-Likelihood ratios

Answer 13

Predictive values depend on prevalence. If prevalence is high, a positive test is more likely to be a true positive.

Likelihood ratios are independent of prevalence

Question 14

Incidence

Answer 14

Number of new cases that develop in a population over a certain period of time. Does not take into account number of cases already present.

Question 15

Prevalence

Answer 15

Total number of cases measured in particular point in time. Function of both incidents and duration of disease.

Question 16

How do the following influence incidence and prevalence?
-improved diagnostic accuracy
-primary prevention

Answer 16

Improved diagnostic accuracy, increases both incidence and prevalence

Primary prevention decreases incident and eventually decreases prevalence

Question 17

Relative risk
-definition
-Study design
-calculation

Answer 17

Compares the probability of developing an outcome between two groups over a certain period of time. Within a certain period of time, how many times more likely are exposed people to develop a particular event, than unexposed?

Prospective study design

Relative risk = risk of disease in exposed / risk of disease in unexposed

Question 18

Odds ratio
-definition
-Study design
-calculation

Answer 18

Compares the chance of exposure to a particular risk factor in cases and controls. How many times more likely are diseased people to be exposed to a particular risk factor compared to non-diseased people?

Case control study design

Odds ratio = odds of exposure in diseased / odds of exposure in not diseased

Question 19

Correlation coefficient

Answer 19

Rangers from minus one to plus one. The plus or minus tells direction of association. Closer to minus one or plus one tells the strength of association. Correlation coefficient does not imply causation.

Question 20

Attributable risk
-what it measures
-Calculation

Answer 20

Measures access incidents of disease due to a particular factor or exposure

Attributable risk = (incidence in population with the risk factor) - (incidence in population without risk factor)

Question 21

Number needed to treat

Answer 21

The number of people that need to be treated in order to prevent one event

Absolute risk difference = control event rate - experimental event rate

Number needed to treat = 1/ARD

Question 22

Leadtime bias

Answer 22

Increase in survival due to earlier detection and not due to successful intervention or improved prognosis

Question 23

Length time bias

Answer 23

Screening test, preferentially, detects, less aggressive form and increase in apparent survival time

Question 24

Null and alternative hypotheses

Answer 24

Null hypothesis: states that there will be no difference in outcome in the study and control groups
Alternative hypothesis : exposure is in someway related to the outcome. If there is a statistically significant difference in outcome between the groups.

Question 25

P value

Answer 25

Way of expressing the result of a studies statistical significance. It represents the probability of finding an association by chance 11 does not exist. (ie that the null hypothesis is true) The smaller, the P value the more confident we can be. P P less than or equal to .05 is considered statistically significant

Question 26

Power

Answer 26

The power of the study is the probability that it can detect a treatment effect if one is present

Question 27

Type one error

Answer 27

Concluding that there is a difference in outcomes or there is an association when there is not. Exposed by investigating the P value. Alpha = the probability of committing a type one error

Question 28

Type two error

Answer 28

Concluding that there is no difference in outcomes or there is no association when there is one Exposed by investigating the power Beta = probability of committing a type two error

Question 29

Three things that power depends on

Answer 29

1) alpha level. Decreased alpha level. (stronger significance criteria.) decreases the power 2) sample size: increased sample size increases the power. More subjects increases the probability of detecting a difference. 3) magnitude of difference in outcome between the groups. Subtle differences are more difficult to detect.

Question 30

Confidence interval

Answer 30

The 95% confidence interval provides an interval of values within which we can be 95% confident that the true prevalence lies. If the 95% confidence interval for a reported relative risk or odds ratio does not include one. There is a less than 5% chance that the observed association is due to chance

Question 31

Calculating the confidence interval

Answer 31

Need to know the mean standard deviation Z score and sample size Calculate the standard error of the mean which is equal to the standard deviation divided by the square root of the sample size Multiply the standard error of the mean by a Z score. For a 95% confidence interval it is 1.96. Then obtain confidence limits, which is the mean plus or minus above

Question 32

Study designs Are two or more groups compared? If no then …

Answer 32

Case report or case series

Question 33

Study designs Which is determined first? The exposure the outcome or are they at the same time? If exposure is first then is it assigned by investigators ? If yes, then it is a…

Answer 33

Clinical trial

Question 34

Study designs Which is determined first? The exposure the outcome or are they at the same time? If exposure is first then is it assigned by investigators ? If no, then it is a…

Answer 34

Cohort study

Question 35

Prospective versus retrospective cohort study

Answer 35

Prospective cohort study follows a cohort overtime with exposure status, determined now and see development of disease in the future Retrospective cohort studies review past records to determine exposure status and then compare disease in incidence Strengths of retrospective are that they can be conducted more quickly and cheaply, and they are more efficient for diseases with long latency. Limitations are that they rely on data from existing records that can be missing or incomplete and temporal relationships may be more difficult to determine. Strengths of prospective studies or that there’s more control over data collection on exposures outcomes and other factors. It is also easier to determine the temporal relationship between the exposure in the outcome. Limitations are that there are more time-consuming and expensive and you may need to follow subjects for a long time

Question 36

Study designs Which is determined first? The exposure the outcome or are they at the same time? If the outcome is determined first, then it is a ….

Answer 36

Case control study Address the disease exposure relationship by comparing the exposure status of the cases versus the controls . Look at the disease cases and the non-disease controls and then look back to compare the risk factor frequency.

Question 37

Study designs Which is determined first? The exposure the outcome or are they at the same time? If at the same time then

Answer 37

Cross-sectional study Simultaneous measurement of the exposure and the outcome. The temporal relationship between the two is not always clear.

Question 38

Randomization and blinding

Answer 38

Assign randomly to study groups with goal of creating similar distribution of known and unknown variables Blinding refers to exposure status hidden from patient and or investigator

Question 39

Intention to treat

Answer 39

Patient’s treatment status at point of randomization is analyzed. Preserve randomization and prevents bias due to selective noncompliance.

Question 40

The normal distribution

Answer 40

Mean equals median equals mode 68% lie within one standard deviation 95% lie within two standard deviations and 99% lie within three standard deviations

Question 41

Which measure of association is used in case control studies and which is used in cohort studies?

Answer 41

Odds ratio in case control Relative risk in cohort

Question 42

Parametric versus non-parametric versus binomial tests

Answer 42

Parametric is used if data are normally distributed, non-parametric is used when data are not normally distributed, and binomial is used if there’s two possible outcomes or categorical data

Question 43

Statistical test to use if you’re comparing one group with a hypothetical value

Answer 43

If normally distributed one sample T test If not, normally distributed a Wilcoxson Binomial chi square

Question 44

Statistical test to use if you are comparing two groups

Answer 44

If normally distributed unpaired or paired t test If not normally distributed Mann Whitney or Wilcoxin If binomial Fisher or chi square for larger sample

Question 45

Statistical test for comparing three or more groups

Answer 45

If normally distributed one way ANOVA If not normally distributed , Kruskal Wallis If binomial chi square

Question 46

Statistical test if quantifying an association between two variables

Answer 46

If normally distributed a Pearson correlation If not, normally distributed a spearman correlation If binomial contingency coefficient

Question 47

Statistical test for predicting a value from another measured variable

Answer 47

If normally distributed linear or non-linear regression If not normally distributed non-parametric regression If binomial simple logistic regression

Question 48

Statistical test for predicting value from several measured or binomial variables

Answer 48

If normally distributed multiple linear or non-linear regression If binomial, then multiple logistic regression

Question 49

Statistical test example. If you want to see if there is a difference in average test score is between two teaching methods, what test do you use?

Answer 49

T test (if normally distributed, otherwise mann Whitney)

Question 50

Statistical test example. If you want to see if there is a difference in the proportion of people who prefer brand a or brand B, what test do you use?

Answer 50

Chi square or fisher

Question 51

Statistical test example. If you want to see if there is a difference in average blood pressure across three different age groups?

Answer 51

If the data is normally distributed, ANOVA (otherwise Kruskal Wallis)

Question 52

Statistical test example. If you want to see the relationship between hours of studying and exam scores.

Answer 52

Pearson correlation if normally distributed. If not normally distributed spearman correlation.

Question 53

Selection bias

Answer 53

Subject selected or not representative of the study population Example examples include differential loss to follow up , non-response bias, or if selection of controls is related to the exposure

Question 54

Information bias

Answer 54

Information collected about or from the study subjects is erroneous Classification of disease or exposure that is more likely in one group versus the other group (data is more accurate for cases versus controls) Recall bias Ascertainment bias

Question 55

Confounding and 3 properties of confounders

Answer 55

Happens when the effect of the main exposure is mixed with the effect of extra extraneous factors Confounders must be 1) independent risk factor for the outcome (even in the absence of exposure) 2) associated with the exposure (distributed differentially across the exposure group) 3) is not part of the causal pathway, connecting the exposure to the outcome Example : the association between smoking and heart disease is confounded by age. Age is an independent risk factor for heart disease, regardless of smoking status. Age is associated with smoking in that smoking status is differentially distributed across age groups (more older people smoke). Age is not on the causal pathway between smoking and heart disease

Question 56

How to avoid confounding: three methods in the design phase and one in the analysis

Answer 56

Randomization: balance is known and unknown confounders Restriction : restrict study population to one category or level of the confounder Matching ; two subject in comparison group so similar or identical distribution of match factors Stratified analysis : examine the association within each strata of the confounder

Question 57

Moderator

Answer 57

A variable that influences the strength or direction of a relationship between the independent and dependent variables. Is not on the causal pathway. Example : if the independent variable is playing quietly before bed and the dependent variable is ease of morning waking, a moderator may be a quiet dark room.

Question 58

Mediator

Answer 58

A mechanism through which the independent and dependent variables are related. This is on the causal pathway. Example : if the independent variable is quietly before bed and the dependent variable is ease of morning, weak, a mediator may be time to sleep in initiation