Lecture 3: Analyzing, Communicating & Disseminating Surveillance Data Flashcards
Steps to analyze surveillance data
- know idiosyncrasies of the surveillance data set
- proceed from simplest to most complex
- recognize limitations of the data
- report findings to stakeholders and those who run the surveillance system
Types of analysis of surveillance systems
- descriptive analysis
- inferential analysis
- aberration detection
- demographic data analysis
Descriptive analysis
Looks at…
- Person (graph showing lyme disease cases in females vs males from age 5 to 80)
- Place (graph showing cluster of lyme disease in northeast and how that has been spreading within the northeast in the last 10 years)
- Time (shows lyme disease case incidence over time in the US)
- Measures (frequency, rates, central tendency, proportions)
Inferential analysis
- trend analysis
- survival analysis
- spatial and temporal aberration detection
- cross-sectional data
- analysis of data from complex surveys
Trend analysis
- Asks whether things are changing (getting better or worse) over time
- Two types of trend analyses
1) Monotonic trends: change at a constant rate over time
2) Non-monotonic trends: change in trends is inconsistent over time
Annual Percentage Change (APC or cAPC)
assumes a constant rate of change holds for the entire period
Segmented Annual Percent change (sAPC)
segmented analysis of changes in trend which:
- assumes the change in rates is constant over each time partition defined by the transition points
- assumes the change in rates varies among different time partitions
Annual percentage change (AAPC)
- summary measure of the trend over a pre-specified fixed interval which takes into account the trend tradition
- has advantage over the APC because it does not assume constant trend pattern that may not hold
- when trend is constant over the entire time interval, AAPC is reduced to both cAPC and sAPC
Calculation in monotonic trend analysis
-See lecture 3 slide 16
-relative difference of the proportions in each group
-AAPC= {Exp (Sigma wb)-1}100
w is length of each segment in years and b is slope for each segment in the desired range of years
-CI AAPC = {Exp log((AAPC/100)+1)+/_ 1.96(SQ Root(Sigma wsd)-1}*100
Non-monotonic trend analysis
- Regression analysis (difficult to explain)
- Jointpoint analysis (software models the number of times and when the trend changes)
Survival Analysis
- Asks whether there is a difference in survival rates between the groups being compared
- Kaplan Meier (a non-parametric method used to investigate the unadjusted survival times of a group without influence of other covariates in the model. Used when cox proportional hazards assumption is violated)
- Cox proportional hazards (a semi-parametric method used when there are many explanatory variables you must adjust and when some of these are continuous; assumes that hazard functions of different individuals remain proportional and constant over time)
Spatial aberration detection
- Asks whether there are 1) clusters of disease, 2) patterns in local rates observed in small areas, etc.
- 1) needs location data to determine clusters (calculate Tango’s index or use SaTScan)
- 2) needs data such as area case counts and population at risk, small-area estimation
- method to detect clusters (bayesian theory)
Spatial aberration detection: Tango’s index
measure of geographical closeness between regions (ranges from 0 to 1)
Spatial aberration detection: SaTScan
detects spatial or space-time clusters, tests if a disease is randomly distributed over space, time or over space and time
Temporal Aberrations
-Asks when do we have sufficient cases to determine that there is an increase in cases that warrants a PH intervention
-requires historical frequency of disease data
uses graphs of current and past experience, scan statistics, and time-series methods
