Week 4 (lecture 3 pt 2)

Question 1

Explain relative risk or risk ratio

Answer 1

Relative risk or risk ratio (RR)
Risk = # of subjects in group with unfavorable event / total # of subjects
RR = risk in treatment group / risk in control group
RR = 1: no difference in risk of the outcome between groups
RR > 1: greater risk of the outcome in treatment group
RR < 1: lower risk of the outcome in treatment group

Question 2

Relative risk reduction (RRR): What does it calculate?

Answer 2

Calculates how much the risk is reduced with the treatment group
RRR = (% risk in control group - % risk in treatment group) / % risk in the control group
RRR = 1 – RR

Question 3

True or false: RR and RRR are proportional differences between treatment and control groups. They cannot describe absolute risk

Answer 3

True

Question 4

Explain Number needed to treat (NNT)

Answer 4

How many patients must be treated for one patient to benefit over a specific period of time
NNT = 1 / (risk in control) – (risk in treatment)
NNT = 1 / ARR

Question 5

Explain Number needed to harm (NNH)

Answer 5

Number of patients receiving treatment for a specific period of time that will have an adverse effect (harm)
NNH = 1 / ARR
-Can use absolute value
-Round down to avoid understating potential harm

Question 6

Define relative risk
Define Relative risk difference or relative risk reduction
Define Absolute risk difference

Answer 6

1) Ratio between the rate of the outcome in the treatment and control groups
2) Comparison of risk between treatment vs. control
3) Net benefit beyond control group observation

Question 7

Summarize NNT and NNH

Answer 7

Number needed to treat
Refers to the number of patients that must receive the treatment in order to for one patient to experience a desired outcome
Preferably a low number
Number needed to harm
Reflects the numbers of patients that must receive the treatment in order to for one patient to experience an adverse outcome
Preferably a high number

Question 8

Summarize relative risk and relative risk reduction

Answer 8

Relative risk
RR = [ A/(A+B) ] / [ C / (C +D) ]
Relative risk reduction
RRR = 1 - [ A/(A+B) ] / [ C / (C +D) ]
RRR = ARR / [ C / (C +D) ]

Answer 9

Absolute risk reduction
Control group rate = C / (C +D)
Experimental group rate = A / (A+B)
ARR = (% risk in control group) – (% risk in treatment group)
Number needed to treat
NNT = 1 / ARR

Question 10

Give some basic facts abt cohort studies

Answer 10

Useful in monitoring aspects that would be unethical to replicate in an RCT
May include large number of patients which will increase the likelihood of observing rare events
Longer periods of time observed compared to RCTs

Question 11

Cohort studies:
1) Define closed cohort
2) Define fixed cohort
3) Define open cohort

Answer 11

1)Closed cohort: set group of individuals who are followed forward in time to determine if they develop the disease outcome
2) Fixed cohort: a closed cohort with specified follow-up time periods
3) Open cohort: group with variable follow-up time periods

Question 12

True or false: Cohort studies can be Prospective or
Retrospective

Answer 12

True

Question 13

Multiple outcomes can be evaluated with cohort studies; give examples

Answer 13

Death, diseases, quality of life, and adverse effects of treatment can all be outcomes within a single cohort study

Question 14

Cohort studies: Give 2 Measures of association

Answer 14

Risk ratios – Fixed cohorts
[A / (A+B)] / [C / (C+D)]
Rate ratios – Open cohorts
(A / person-time in the exposed group ) / (C / person-time in the unexposed group)

Question 15

What is the math for risk difference in cohort studies?

Answer 15

[A / (A+B)] - [C / (C+D)]

Question 16

Cohort studies: Explain Attributable fraction of the exposed / attributable risk percent

Answer 16

Expresses the risk difference relative to the risk in the exposed group
Measure indicates the proportion of risk in the exposed group that would not have occurred in the exposed group if they were not exposed
The higher the ARP, the greater burden the exposure contributes to the risk of the outcome in the exposed group
([A / (A+B)] - [C / (C+D)]) / [A / (A+B)]

Question 17

Cohort studies: Explain Prevented fraction in the exposed

Answer 17

For an exposure that is protective for a disease outcome, the proportion of potential cases in the exposed group which were prevented by being exposed can be measured
1 – RR = 1 - [A / (A+B)] / [C / (C+D)]

Question 18

Give the basics of Case-Control Studies

Answer 18

1) Compares the frequency of exposure among subjects or “cases” that experience an outcome event, most commonly a disease, and “controls” who do not have the outcome event or disease
2) Subjects selected on the basis of the outcome event
3) Once cases and controls are assembled, the researcher can also collect data on a wide range of exposures, and estimate their effect on the outcome
4) Cases may include newly diagnosed cases (incident cases) or persons with existing disease (prevalent cases)
-A sample of controls, or those without the event or disease, should then be selected to estimate the exposure distribution in the base population

Question 19

Case-Control Studies:
1) Define Population-based case-control
2) Define Clinic-based case-controlled

Answer 19

1) Identifies cases and controls in a defined base population
2) Cases are typically selected from individuals with the relevant disease of interest at a hospital or clinic and controls are selected from the same institution without the disease

Question 20

Case-Control Studies: Define Nested case-control

Answer 20

1) Type of population-based case-control study where the study is nested within a cohort
-Typically all the incident cases from the cohort form the case group, while controls are randomly selected from: 1) all subjects (case-base sampling), 2) subjects without the disease at the time the case was identified (density sampling), or 3) subjects who do not develop the disease over the entire follow-up period of risk (cumulative sampling)

Question 21

List the 3 types of case-control studies

Answer 21

1) Population-based
2) Nested
3) Clinic based

Question 22

The most basic approach to measuring exposure in cases and controls is to construct a __________ measure for categorizing exposed and unexposed groups
Currently exposed
Ever exposed
Formerly but not currently exposed

Answer 22

binary

Question 23

Explain Confounder adjustment in case-control studies

Answer 23

1) Matching – process of making the cases and controls similar (or balanced) with regard to this confounding factor so that there will be enough information in the analysis to adequately control for the factor
2) If the case-control study uses one-to-one matching, the unit of analysis becomes the matched case-control pair
3) To calculate the OR, each case-control pair is placed into one of four categories
-Two concordant pair categories where the exposure status was the same for both cases and controls
-Two discordant pair categories where the exposure status for the case was different than for the matched control

Question 24

When does bias occur?

Answer 24

Bias occurs when the estimate of the effect of the exposure on disease does not represent the true effect of the exposure on the disease outcome

Question 25

Confounding can occur in a variety of situations including what?

Answer 25

1) when an association between two variables (the exposure and disease outcome) is due to a common cause of the exposure and the disease outcome or 2) as a result of selection bias, such as from matching on a factor which is not a risk factor for the disease, but is associated with the exposure in a case-control study

Question 26

Differentiate primary and secondary data

Answer 26

Primary data = obtained directly from patients Secondary data = obtained from non-patient sources (e.g., medical records)

Question 27

Cross-sectional studies: Give the basics

Answer 27

Examine population characteristics at a cross-section (one point) in time Information about a population such as disease prevalence but may also be used to examine associations between an independent (exposure) and a dependent (outcome) variable Causality between an exposure and outcome cannot be established from a cross-sectional study

Question 28

Explain response rate and data analysis in cross-sectional studies

Answer 28

1) Response rate 60% response rate is considered a good benchmark in determining adequate survey response Johnson TP, Wislar JS. Response rates and nonresponse errors in ­surveys. JAMA. 2012;307(17):1805–1806. [PubMed: 22550194] Poor response rates limit generalizability 2) Data analysis Summarizing the characteristics of the population using means and percentages Cross-section in time analysis to show associations have large degree of biases and should be interpreted cautiously

Question 29

When are cross-sectional studies useful?

Answer 29

1) Efficient means of capturing descriptive information about a population at a given point in time 2) Because cross-­sectional studies obtain information on exposures and outcomes at the same point in time, they lack the ability to establish a temporal relationship between them 3) Associations found in cross-sectional studies should be viewed with skepticism and warrant a stronger study design to confirm a causal link

Question 30

Pre- and post-observational studies: What are they? Is there a lot of bias?

Answer 30

Researcher examines the effect of an exposure in a population by comparing observations in that population before and after the exposure Because observations are collected both before and after an exposure, the pre-post study design, eliminates the bias of temporal precedence found in cross-sectional studies

Question 31

Explain Pre- and Post-Observational study design

Answer 31

1) One-group ­pretest posttest study design; no control; no randomization 2) Observations are compared before and after an intervention in the population of individuals affected by the intervention Hypothetical example: Georgia Medicaid requires treatment for type II diabetes with metformin for three months prior to covering insulin glargine and then the next year the policy changes to a requirement of two years of metformin + / - sulfonylurea (e.g., glimepiride)+ / - DDP-4 inhibitor (e.g., sitagliptin) for two years prior to covering insulin glargine What effect will this have on outcomes? Increase micro- and macrovascular complications? Pharmacoeconomic impact? How much money did the state save?

Question 32

List some terms that are used in pre and post observational analysis

Answer 32

t-test to examine the mean change in the outcome between observation periods Variables could be included as covariates in the model to control for their effect on the outcome using multivariable linear regression Chi-square test to examine the probability of the outcome after the intervention compared to the period before the intervention

Question 33

Pre- and Post-Observational Studies: Give the strengths and limitations

Answer 33

1) Strengths: Eliminates the bias of temporality seen in cross-sectional study designs 2) Limitations: Potential biases such as history, maturation and regression to the mean History bias references the effect of a second factor occurring between the pretest and the posttest that is confused for the effect of the exposure of interest Maturation bias references the natural change in an outcome over time outside of the influence of the exposure of interest Regression to the mean references extreme scores on a variable at pretest which leads to more room for a score to approach the mean value than might be expected ­otherwise

Question 34

Ways to improve pre- and post-observational studies include?

Answer 34

1) Additional points of time observed Additional time points before the exposure allows the researcher to examine patterns of maturation occurring naturally outside of the influence of the exposure and control for these patterns 2) Nonequivalent comparison group Comparing Georgia Medicaid to a state that does not have restrictions on insulin glargine like Alabama Would New York or California be good comparison groups with Georgia?

Question 35

Time-Series Analysis: Give the design and analysis

Answer 35

Design Pre- and post-observational study with several points of observation Analysis Linear regression model where the predictor of interest is the observation time period

Question 36

What does an ecological study do?

Answer 36

Makes comparisons in groups of individuals rather than on individuals themselves