Study Design and Causality

Question 1

internal validity

Answer 1

Is an individual result valid? Or is it a result of experimental bias, random chance, or a confounding variable?

Question 2

external validity

Answer 2

Is a result causal and generalizable?

Question 3

Prospective studies

Answer 3

follow individuals forward in time and, at some point in the future, collect data on outcomes. These include prospective cohort studies and randomized controlled trials.

Question 4

Retrospective studies

Answer 4

use existing data collected at some point in the past. These include case control studies (participants selected on the basis of outcome present / absent, and then their exposures are compared), and retrospective cohort studies

Question 5

Cross-sectional studies

Answer 5

observational studies in which exposures and outcomes (e.g. a disease state) are measured simultaneously in a population at a single point in time

Question 6

Cohort studies

Answer 6

observational studies in which we follow a population forward in time collecting data on exposures and the development of outcomes. Most cohort studies are prospective.

Question 7

Hill’s Criteria

Answer 7

• Temporality
• Strength
• Consistency
• Specificity
• Dose-response (biological gradient)
• Plausibility
• Reversibility (or Experimental Evidence)
• Analogy

Question 8

Nominal or Categorical variables

Answer 8

Named but not necessarily ordered (may be dichotomous) - (e.g. sex, death, ethnic or cultural background).

Question 9

Ordinal variable

Answer 9

Necessarily ordered categories where the distance between each unit is not defined (e.g. military ranks; or, a satisfaction scale of poor, fair, good, very good, excellent).

Question 10

Interval - Discrete variable

Answer 10

Take on discrete (e.g. integer) values with equal magnitude between points (e.g. number of medications, days hospitalized)

Question 11

Interval - Continuous variable

Answer 11

May take on any value over a continuum (e.g. height or weight).

Question 12

Ecological study

Answer 12

Comparing the health or risk of disease between populations

Question 13

Case series study

Answer 13

Designed to describe characteristics of disease or exposure risks. Follows either the same disease or the same exposure. Brings out characteristics of disease and risk factors of particular populations.

Question 14

A fundamental feature of the cross-sectional study is that it . . .

Answer 14

. . . cannot be used to determine causality, by definition.

Question 15

Greatest challenges of case control studies

Answer 15

1. Recall of the patients (especially when studying something like foodborne illness)
2. Matching cases and controls well

Question 16

Interventional study

Answer 16

Usually randomized controlled. Allocation and control groups are assessed and compared to see the effects of the intervention. Must be double blinded.

Question 17

Systematic review vs meta-analysis

Answer 17

Systematic review means a summary of all studies done on a topic and their conclusions.

Meta-analysis is by definition an additional statistical analysis that compiles the raw data from many studies with similar questions and study design in order to create a compound dataset.

Question 18

Passive surveillance

Answer 18

• Routine reporting of health data
• Ex, disease registries

Question 19

Active surveillance

Answer 19

Health data actively sought out. such as during an outbreak. Also includes serosurveillance.

Think John Snow and Henry Whitehead.

Question 20

Sentinel Surveillance

Answer 20

Uses select institutions or groups as microcosms to provide generalizable information on disease trends and detect outbreaks. Not so useful for rare or uncommon events though.

Question 21

Rumour surveillance

Answer 21

Relies on unofficial forms of information, like social media. Nowadays since social media moves much faster than other information, processing it can pick up outbreaks or disasterous events faster than any other method.

Question 22

Syndromic surveillance

Answer 22

Surveills nonspecific factors which might clue in to an outbreak, such as

• Reports of common symptoms like fever, respiratory illness, gastrointestinal illness
• Purchase rate of generic medications
• Absenteeism from work and school

Usually relies on automated surfeillance

Question 23

Coherence in Hill’s Criteria

Answer 23

The cause-effect relationship does not conflict with what is known and there are no competing hypotheses

Question 24

If an outcome is very rare, the most practical choice for a study design is likely to be

Answer 24

Case control

Question 25

Zambia has one of the highest burdens of invasive cervical cancer, at 58.4 new cases per 100,000 person-years. This number describes the _______________ of invasive cervical cancer:

Answer 25

incidence rate

Question 26

In the measles vaccine study, at day 91 after immunization, 85.4% of children in the aerosol group, as compared with 94.6% in the subcutaneous group, were newly seropositive for measles. What type of outcome measure is this?

Answer 26

Cumulative Incidence . . . so they say. I say it and prevalence are equivalent here.

The differences between cumulative incidence and prevalence come from two sources: pre-existing condition (which counts as prevalence, but not incidence) and conditions which disappear before the end of the specified cumulative incidence period (seroconversion does not dissappear in this window). Ergo, in this case, cumulative incidence is equal to prevalence.