Quantitative Studies
Introduction
Levels of Data
There are various levels of data.
Types of Study
- experimental studies
- observational studies
Experimental Studies
- prospective trials that are preferably randomized, controlled, and blinded
- usually use 'intention-to-treat' analysis
- limitations: subjects or conditions may not reflect the 'real world' and be of limited applicability
- cross-over studies can allow subjects to act as their own controls - ensure proper wash-out period
- n of one trials
Observational Studies
cohort
- normally prospective, or longitudinal
- starts with known exposure status and follows subjects over time to find disease outcomes
- good for studying rare exposures, estimates of timing
- selection of control group identical except for the things of interest is very difficult
- often expensive, costly, and suffers from participant loss
case-control
- starts with 'known' cases and controls and a number of potential casuative factors
- retrosepctive;
- very good for examining exposure history
- useful for rare diseases or long intervals between exposure and disease
- relatively quick and inexpensive
- high risk of recall bias and can be difficult to select appropriate controls
cross-sectional
- examines potential exposire at one point in time
- usually administered by survey
- limited usefullness except for looking for associations
ecological
- examines populations rather than individuals and deals with comparisons of rates
- used in preliminary stages of cancer research
- beware 'ecological fallacy'
unsystematic clinical observations (case series, case reports, personal opinion)
- interesting, but of limited value
Types of Data
Data is collected from an experimental population and is compared with the control population to test the hypothesis. Statistics allow us to test how unlikely it is that observed data does not come from the normal distribution of the control population.
Types of Data
categorical
- dichotomous
- nominal - names or classes (ie ethnicity)
- ordinal
numerical
- discrete
- continuous
- interval - equal spaces between measurements
- ratio - has a true zero
Dependent variables can be either continous or dichotomous, and the statistic test used depends on the type of data of the independent variable.
continuous dependent variables
statistical tests used:
- t-test - two independent variables
- ANOVA - more than two independent variables
- linear regression - continuous independent variables
- correlation - continous independent variables
dichotomous dependent variables
statistical tests used:
- Chi-square and Fisher's exact test - normally 2x2 tables
- logistic regression - any independent variable(s)
survival analysis
As participants can enter or leave studies at different time points, uneven observation periods are common with survival analysis. Person-time date can be used, but it assumes 1 person for 10 years is equal to 10 people observed for 1 year, which is not likely true.
More specialized tests include
- life table analysis - divides the observation interval into smaller periods and estimates probability for each
- Kaplan-meier curves - similar except that intervals are very small, and exact survival time is used
- Cox proportional hazards regression - predicts the time to outcome by comparing Kaplan-Meier curves while controlling for other variables; yields hazard ratios that can be interpreted as adjusted RRs.
Contraindications
Clinical Vignette 1
A person takes this treatment for this reason due to concerns re: condition
This happens.
How do you respond?
Clinical Vignette 2
Additional Resources
Topic Development
created: DLP, Aug 09
authors: DLP, Aug 09
editors:
reviewers: