Transcript 10.26.04

Descriptive Studies:
Person, Place and Time
October 26 2004
Epidemiology 511
W.A. Kukull
Today’s quote
“…Disease does not occur randomly but in patterns which
reflect the operation of underlying causes…Knowledge of
these patterns…constitutes the major key to understanding
causation, and hence, devising methods of control and
prevention. In essence, the pattern of disease in populations
is described by the composite answers to three basic
questions: Who is attacked? Where does disease occur?
When does it occur?”
John P. Fox, MD, MPH
Descriptive Studies
• Observations of disease occurrence in
– Person
– Place
– Time
• Description
• Hypothesis
• Results
Hypothesis
Analytic Study
Description
– Hypothesis generation leads to analytic studies
Case Reports/Series
•
•
•
•
Focus on single patients or series of patients
Describe unusual occurrence of disease
Recognition of new diseases
Formulation of risk factor hypotheses
– 3 vinyl chloride workers with angiosarcoma
– 5 cases of Pneumocystis pneumonia in young
men
• Cannot test; no comparison group
Ecologic studies
(a.k.a. correlational studies)
• Population characteristics correlated with
disease occurrence
– cigarette consumption and CHD deaths
– Fluoride concentrations and fractures
• Strength: quick, inexpensive, use available data
• Weakness: cannot link exposures to
individuals; cannot control for confounding
Ecologic Studies
• Correlation (or “association”) is not
necessarily causation
– Lack of a correlation does not imply no
causation
• Ecologic study data represent average
exposure levels
– per capita alcohol consumption and CHD:
opposite findings ecologic vs cohort
Levels
• Levels of Measurement
– Who/what is measured
• Levels of Analysis
– What is the unit of analysis; group or individual
• Levels of Inference
– Is the inference level consistent with
measurement and analysis levels
Ecologic bias
• Ecologic bias: “failure of ecologic effect
estimates to reflect biologic effect at the
individual level…” (Morgenstern, 1998)
• Cross-level inference
• Other
Cross-sectional Surveys
• Exposure and disease are assessed
simultaneously-- “snapshot”
• Useful for prevalence, health planning
• Temporal relation not observable
– different conclusions at different times?
– reflect survival as well as etiology
– Example: inactive farmers and high CHD
Time
• Short term changes; point source epidemics
• Time specification:
– 36 cases of polio within 12 month of
tonsillectomy; 16 within 7 - 18 days
• Clustering in time
– Triggering exposures
– new statistical methods
Time (2)
Secular Changes
• Changes over long periods
– Completness of the data source
• In 1917, 35% of Death Certificates listed >1 cause;
in 1979, 73% listed >1 cause
• Reasons?
–
–
–
–
Changes in diagnostic ability and classification
Demographic changes; treatment and survival
Concomitant environmental circumstances
Change in natural history of disease
Time (3)
Secular change
• Generational trends or “Cohort” effects
– Are persons born at different times at different
risk?
– Is yearly cross-sectional mortality by age
misleading
– Not to be confused with “cohort studies”
Time
Cyclic Changes
•
•
•
•
•
•
Repetitive cycling e.g., by season
Flu and mortality
Suicide and season, day of week
Insect vectors; recreational activities
Most MI’s/strokes occur in morning hours
Usually evaluated by plotting occurrence of
cases (assuming stable population)
Place
• International comparisons: Infectious and
non-infectious diseases
– High stomach Ca in Japan and Iceland
– High CHD and colon Ca in US
– Provides a basis of broad classification for
countries as “probably” high or low
• Cross-National surveys: specific studies
standardized protocols
Place (2)
• Disease Mapping: intra- or inter-national
– John Snow and the Broad Street pump
– Proximity to shipyards and cancer
– Environmental contamination : Love Canal
• Boundary problem
– No prior hypothesis
– Prior hypothesis: set boundaries before
determining disease
– Observed cluster: don’t allow case occurrence
to set boundaries
Person:
Age effects
• Influences on Age Curves
– diagnostic difficulties and trends
– Stage of disease, incident or prevalent cases
– Early life mortality and diseases of aging
• Bimodality
– Hodgkins disease: two etiologies?
Person
• Gender differences
– ratio
– mortality: causes of death
– morbidity
• women higher for thyrotoxicosis, diabetes,
cholecystitis
• men: IHD(?), respiratory disease, PUD, accidents
– more ill health in women; earlier deaths in men
– Has it helped our understanding of disease?
Person
Race, Ethnicity, Religion
• Inconsistent definition in place and time
– Birth and Death certificate
– Hispanics: race, ethnicity or culture??
• Religion
– cultural background
– dietary practices
• Major gene diseases: Sickle cell; Tay-Sachs,
Thalassemia, alcohol dehydrogenase
Person
Migration: genes or environment
• Does a change in disease rates accompany a
change in environment?
– Cardiovascular disease; colon cancer
• Multi-generational studies
• Biases in Migrant studies
–
–
–
–
Diagnostic, reporting and coding practices
Migrants not representative of home country
Migration stresses may increase risk
Denominator inaccuracies
Person: other factors
• Socioeconomic Status
– restricted activity and disability days decline
with increasing income
– infant mortality impact
– No uniform definition; often based on
occupation; years of education
• Occupation: toxic exposures, accidents
• Marital Status: mortality lowest among
married people
Conclusion
descriptive studies
• Generate hypotheses
• Describe confounders
• Reveal biases
• Ease of execution vs interpretation