Download Role of Epidemiology and Statistics in Advanced Nursing Practice

Chapter 2: Role of Epidemiology and
Statistics in Advanced Nursing
Practice
Key Ideas, Definitions, and
Preliminaries (1 of 3)
• Epidemiology intersects with many disciplines.
– Statistics is one such discipline.
• Key idea of statistics is variability.
• Data can be thought of as facts.
• Data sets contain one or more variables.
– Example:
• Cases (rows in a spreadsheet)
• Facts about each patient (columns in a spreadsheet)
Key Ideas, Definitions, and
Preliminaries (2 of 3)
• Nominal data merely names.
– Example: the variable gender
• Ordinal data can be ranked.
– Example: severity of bedsores
• Interval data have property that differences
between values have clear interpretation.
– Example: age in years
Key Ideas, Definitions, and
Preliminaries (3 of 3)
• Ratio data have property that ratios of values
have clear interpretation.
– Example: patient with four bedsores has twice as
many as patient with two bedsores.
• All of these types of variables may occur in
epidemiological studies.
• Statistical data analysis involves determining
and describing the story a data set has to tell.
Role of Probability in Epidemiology
• Probability is study of laws of chance.
• In epidemiology:
– Probability is typically described as risk.
– Assignment of probability Is generally arrived at
empirically—through observation.
Statistical Inference and the Language
of Statistics (1 of 3)
• Statistical inference = drawing conclusions
from samples using statistical theory
(methods).
• Summary measures and graphical tools
– Mean, median, standard deviation (SD)
– Correlation coefficient
– Graphical tools include side-by-side box plots
(“box and whisker plots”) and scatter plots.
Statistical Inference and the Language
of Statistics (2 of 3)
• Inference
– Uses samples and statistical theory to draw
conclusions.
– Divided into three (overlapping) activities:
1. Estimation
2. Hypothesis testing
3. Model fitting (building)
– May involve confidence interval (CI) estimates.
Statistical Inference and the Language
of Statistics (3 of 3)
• Power
– Power of a test is measure of likelihood of
rejecting false hypotheses.
– A function of sample size.
– An effect size may be reported as well as a p value.
– Statistical significance and clinical significance
differ.
Measures of Disease Occurrence and
Risk (1 of 6)
• Incidence and prevalence
– Incidence proportion is pure number (unitless)
that might be thought of as a probability.
– Incidence rate takes into account the total
exposure of subjects during study period.
– Prevalence measures attempt to describe a
situation (phenomenon) at an instant in time.
Measures of Disease Occurrence and
Risk (2 of 6)
• Incidence and prevalence (cont’d)
– Inferences for incidence and prevalence
• Inference is process that uses data and statistical theory
to draw conclusions about populations based on
samples.
• Confidence interval (CI) estimates are used.
• Inferences about incidence rates are usually made
using the Poisson model.
– Stratification of epidemiological data is often
useful.
Measures of Disease Occurrence and
Risk (3 of 6)
• Survival analysis
– Originated in mortality studies.
• But time to any well-defined event can be studied.
– Kaplan-Meier estimation of survival function is
most commonly used nonparametric estimate of
survival function.
– Useful in comparing survival functions.
– Log-rank test can be used to test homogeneity of
survival curves.
Measures of Disease Occurrence and
Risk (4 of 6)
• Analysis of 2 × 2 tables
– Measures for comparing risks
• Relative risk (RR)
• Odds ratio (OR)
– Effect measures
• Attributable risk (AR)
• Etiological fraction (EF)
Measures of Disease Occurrence and
Risk (5 of 6)
• Regression modeling
– Building models to describe a phenomenon.
– These include linear models and logistic
regression.
– Linear models
• Most commonly used.
• Allow for various transformations of the explanatory
variables (such as quadratics or square roots).
Measures of Disease Occurrence and
Risk (6 of 6)
• Regression modeling (cont’d)
– Logistic regression
• Dichotomous logistic regression models the log odds of
the event of interest.
• One can use any mixture of interval and categorical
predictors.
• There are problems associated with the use of logistic
regression.
Summary of Epidemiological Measures
of Disease Risk
• See Tables 2-19 and 2-20.
• Vital statistics calculations are given in the
chapter.
• Data can be organized in two-dimensional
graphs created by statistical packages such as:
– Microsoft Excel
– PASW (SPS)
– SAS