Epidemiologic Study Designs and Plans

Epidemiologic Study Plan Steps

• List study participants
• Design the study using a formal approach. This guides data collection, analysis, and interpretation
• Experimental study design is a type
• Analytical study design is a type
• Descriptive study design is a type
• Employ a measurement approach
• Use exposure versus outcome variables
• Address statistical issues, like the impact of sample size of data management on data validity and reliability

Basic Epidemiology Study Designs

• Experimental study design involves investigators influencing the exposure of study subjects.
• Descriptive study design examines who, what, when, and where.
• Analytical study design examines why and how.

Recruiting Study Participants

• Probability sampling methods use random selection for strong statistical inferences about the whole group
• Non-probability sampling methods use non-random selection based on convenience or certain criteria for easy data access
• Population shares at least one characteristic
• Target population is determined by sampling criteria
• Accessible population refers to those available to a researcher, who participate in the study
• Sample refers to the small group used to represent the general population

Variable Types

• Exposure variables influence health outcomes and are risk factors, also called explanatory, predictor, or independent variables
• Outcome variables are diseases, events, behaviors, or conditions that change due to a change in exposure, also called dependent variables
• Effect modifiers occur when the exposure and outcome relationship depends on a third variable
• A moderator can impact the strength (magnitude) and direction of the relationship between exposure and outcome variables
• A confounder is a systematic error when a factor is associated with a disease outcome independently of the exposure
• Confounders must be risk factors for the outcome
• Confounders must be associated with exposure
• Confounders cannot be an intermediate between

Sample Size and Study Power

• Larger sample sizes mean higher study power
• Larger sample sizes increase data validity

Data Types in Epidemiology

• Nominal data is categorical with no inherent ordering (e.g., marital status, sex)
• Ordinal data is categorical with inherent ordering (e.g., preference rating)
• Numerical data is quantitative, including discrete and continuous types (e.g., number of fractures)

Statistical Methods

• Frequency shows the number of observations for a given variable
• Relative frequency shows the number of observations for a specific level of a variable compared to the total
• Formulas are used to calculate ratios, proportions, prevalence, incidence, and mortality rates

Crude vs. Adjusted Rate

• Crude rates are calculated without restrictions on who is counted in the numerator or denominator
• Adjusted rates account for confounding influences like age or gender when comparing two groups
• Adjusted rates allow comparisons between populations by controlling for factors influencing health outcomes

Prevalence vs. Incidence

• Prevalence measures all new cases in a population at a specific point in time
• Incidence measures all new and existing cases in a population at a specific point in time
• Point prevalence is a measure at a single time point
• Period prevalence is measured over a period of time

Descriptors of Person Data

• Person data describes age, gender, and sex
• Place data focuses on geographic location, comparisons between or among groups after migrations
• Time data looks at seasonal illness and duration of illness

Population Pyramids

• Expansive pyramids have high birth and death rates
• Stationary pyramids have similar numbers or percentages across age groups
• Constricting pyramids have fewer younger individuals

Graphical Trends in Epi

• Epidemic curves show the frequency of new cases over time based on onset date
• The curve's shape relative to a disease's incubation period can indicate the epidemic's source
• Secular trends represent long-term changes in health states or events
• Short-term trends are brief, unexpected increases in health states or events
• Cyclic trends involve periodic increases and decreases in health states

Types of Surveillance

• Passive surveillance is reported by providers, inexpensive, simple, and varies in quality/completion
• Active surveillance involves agencies reaching out to providers for more complete reporting, used in epidemiologic studies

Descriptive Study Designs

• Ecological studies aggregate data on the population or community level with strengths in well-defined sources but limitations in individual-level inferences
• Case reports provide in-depth analysis of individuals with unusual diseases
• Case studies involve small groups with similar diagnoses, providing detailed descriptions of diseases, which can apply findings to larger studies but cannot establish causation
• Cross-sectional studies compare disease outcomes to a variable of interest at a single point, strengths in good availability for rare outcomes but limitations in establishing causation or sequence

Analytic Studies

• Analytic studies compare groups and test specific hypotheses

Analytic Study Designs

• Analytic studies answer how and why health-related states or events occur
• Experimental studies assign participants to exposures
• Observational studies do not intentionally expose participants or ask them to change behaviors
• Case control studies identify cases, then controls, and investigate past experiences, behaviors, or exposures
• Case control studies have outcomes known prior to exposure
• Selecting cases requires establishing diagnostic criteria
• Cases should represent all people affected by the disease
• Incident cases require waiting to recruit a study sample
• Prevalent cases have larger target pools available
• Controls must be selected independently of exposure and should represent the general population
• Population-based controls represent well-defined population
• Hospital-based controls are easily accessible
• Cohort studies are conducted on those without the disease, differing by exposure status over time to compare exposed vs. unexposed

Cohort Studies

• Exposure is known before outcome in cohort studies
• Retrospective cohort studies start in the past and follow up into less far past
• Prospective cohort studies start in the present and follow up into the future
• Notable cohort and case-control studies include the Framingham Heart Study, British Doctors Study, Nurses Health Study, and Black Women's Health Study

Experimental Studies

• Experimental studies see investigators influencing the exposure of study subjects
• Controlled trials assign individuals to exposure or non-exposure groups
• Community trials assign community groups to exposure or non-exposure groups
• Community interventions are designed for educational and behavioral changes at the population level
• Between group design compares outcomes between two or more groups with different levels of intervention
• Within group design compares outcomes observed in a single group before and after intervention

Blinding

• Single blinding means participants are unaware of their group assignment
• Double blinding means neither researchers nor participants know group assignments
• Triple blinding means researcher, participants, and those doing the analysis are unaware

Randomization

• Random assignment is intended to make intervention and control groups as similar as possible
• Chance is the only factor that determines group assignment
• Randomized controlled trials compare two or more interventions with random allocation and specific inclusion/exclusion criteria
• Randomized controlled trials demonstrate cause-and-effect
• Randomized controlled trials allow control of exposure levels and have high costs/ limited generalizability

Clinical Trial Phases

• Phase 1 trials evaluate safety, determine safe dosage, and identify side effects in a small number of participants
• Phase 2 trials test effectiveness, further evaluate safety using relatively small randomized blinded trials
• Phase 3 trials confirm effectiveness, monitor side effects, compare to other treatments, and collect information in large groups
• Phase 4 trials assess rare serious side effects and explore further therapeutic uses after therapy is approved

Research Ethics

• Research ethics are important in experimental research; example, the US Public Health Service Study on the Progression of Untreated Syphilis
• Maximize benefits
• Avoid risks
• Respect subjects
• Ensure fairness to all in society
• Participants can drop out of a trial at any time with no penalty

Hypothesis Testing

• Null hypothesis says there is no difference in the health outcome in exposed vs. non-exposed groups
• Alternative hypothesis says there is a difference is in the health outcome in exposed vs. non-exposed groups
• Type I error occurs when an investigator rejects a true null hypothesis
• Type II error occurs when a researcher fails to reject a false null hypothesis
• The likelihood of errors can be reduced by increasing the sample size
• Magnitude of association shows the direction (negative or positive) and strength of the association
• Larger numbers indicate stronger associations
• Confidence intervals are commonly used in research
• A confidence interval passing 1 usually indicates no statistical significance
• P value is the statistically tested number of how likely particular observations are true if the null hypothesis were true
• P < 0.05 shows statistical significance
• P > 0.05 is not statistically significant
• P = 0.05 is statistically significant

Odds ratio

• The odds ratio interpreation is used to describe the exposure as a risk factor or protection factor
• 1.0 indicates similar exposure among cases and controls

• 1.0 indicates greater exposure among cases (risk factor)

• < 1.0 indicates lower exposure among cases (protective factor)
• Statistical significance is a component