Observational Study Designs Summary

Observational Studies

Introduction

• Observational studies are a key part of epidemiology.

• Epidemiology studies populations (plants, animals, humans).

• Health, disease, and illness are not random; characteristics protect/predispose.

• Observational studies evaluate distribution of characteristics/events and associations between characteristics and health outcomes.

• Study Designs:

  • Experimental vs. Observational

  • Longitudinal vs. Cross-sectional

  • Prospective vs. Retrospective

Ecological (Correlational) Studies

• Use population-level data to compare associations across populations.

• Example: Cigarette consumption vs. lung cancer mortality rates.

• Advantages: Quick and easy using existing data.

• Disadvantages:

  • Cannot link risk factors to individuals with the disease.

  • Ecological fallacy.

  • Difficult to assess confounding variables.

Case Reports and Case Series

• Case report: Detailed description of a single case.

• Case series: Presents several similar cases.

• Focus on individual patients with defined clinical characteristics.

• Design: Simple description of clinical data without a comparison group.

• Objective: Describe new clinical phenomenon.

• Observations should be comprehensive and detailed.

• Inclusion criteria should be consistent across all patients.

• Data summaries include frequencies, proportions, means, medians, and standard errors.

• Interpretations should summarize the new phenomenon, reference previous observations, and suggest etiology or further studies.

• Advantages: Useful for forming hypotheses, planning natural history studies, and describing clinical experience; easy and inexpensive.

• Disadvantages:

  • Selection of cases may be biased.

  • Difficult to generalize results.

  • Findings may be chance happenings.

Single Time Point Studies

• Cross-sectional, Prevalence, and Incidence Studies.

• Contain individual-level data.

• Examine diseases, conditions, or characteristics in a defined population at a specific time.

  • Prevalence Rate: Number of persons with a disease at a time / Number in that population at risk at that time.

Case-Control Studies

• Subjects are selected based on "case" definition.

• Typically retrospective, looking backward in time.

• Compare persons with a disease to controls without the disease to identify potential etiologic factors.

• Useful for studying rare diseases.

• Three important criteria to minimize bias:

  • Cases representative of all patients with the disease.

  • Controls representative of the healthy population.

  • Information collected the same way from cases and controls.

• Addressing recall bias.

• Difficulties in Defining, Selecting, and Recruiting Controls.

Cohort Studies

• Cohort studies select study participants irrespective of disease status.

• Used to observe associations between exposures/risk factors and subsequent disease development.

• Information is collected and then disease outcomes accrue of time.

• Categories:

  • Retrospective (Nonconcurrent, Historical)

  • Prospective (Concurrent)

Odds Ratios, Risk Ratios, Relative Risks, and Attributable Risk

• Absolute risk: Risk in a certain group without comparison.

• $Odds = P/(1-P)$

• OR formula: $(a/c)/(b/d) = ad/bc$

• RR formula: $[a/(a+b)]/[c/(c+d)]$

• Attributable risk: Amount of disease attributed to a characteristic or exposure.

Mistakes, Misconceptions, and Misinterpretations

• Trusting Bivariate Associations Based on Observational Study Data can be misleading.

• Assuming Odds Ratios and Relative Risks Will Have a Similar Magnitude.

• Implying Causation

• Confusing Causation, Prediction, Association, and Confounding

• Assuming Observational and Randomized Studies Never Agree

Conclusions

• Observational studies are valuable alternatives, predecessors, and follow-ups to clinical studies.

• Good observational studies are vital to inform medical, public health, policy, and regulatory decisions.