Workshop: Aerospace Epidemiology
The Aerospace Medicine Community has long been involved in the study of adverse outcomes, mishaps, and the terrible results of flights gone bad. For decades the profession has addressed the events surrounding mishaps as case reports with few summaries looking over many years. Those summaries that have been composed, many times, look at simple rates and draw few conclusions. Worse, many are done as annual summaries and as a result of an impotent denominator end up with careful analyses of what is commonly referred to in epidemiological circles as ‘noise.’ The data required for in depth study of the mishaps is ephemeral, difficult to obtain, and laborious to work with. Investigators reach inconclusive results in many cases because their work lacks adequate power. Even when the data is there and the power is there, proving causality remains elusive because of our inability to stratify well enough to achieve the Bradford-Hill criteria.
This course is designed to share the knowledge required to conduct sound epidemiological investigations into low occurrence rate events like aviation mishaps. Graduates of this seminar should leave with an understanding of the basic tools required to glean association from the available data, understand the criteria for proving causality, thoroughly understand the import of power (in a statistical sense) and have the ability to conduct epidemiologic calculations. Most importantly, they should have a sense for what is required to collect sound data for future analyses so that tomorrow’s aviation mishap epidemiologists can work to further improve safety. As I review this, in 2016, commercial civil aviation has become extremely safe, general aviation and military aviation are still experiencing preventable mishaps at unacceptable rates.
An understanding of epidemiology is not enough to be able to conduct sound aviation mishap investigations. The investigator must also embody a journeyman’s understanding of aeronautics, some knowledge of cognitive neuroscience, anthropometrics & physiology, a good concept of weather phenomena, a smattering of materials science & physics, a working knowledge of avionics and a concept of how all these things and more interdigitate to allow successful mission completion. The general knowledge of all thing’s aviation enables the epidemiologist to target specific hypotheses and peel back information to get at the likely causes for the mishaps. The knowledge provides the gut instinct that makes effective investigations possible. The same knowledge can also lead to bias. The epidemiological investigator would do well to remember the words of Dr. Jay P. Sanford who said of medical education: ‘half of what we have taught you is wrong and we do not know which half.’ Most members of AsMA have the knowledge required to analyze subsets of mishaps and have spent a lifetime acquiring it. To that knowledge, we would like to add the tools available to the epidemiologist.
Please login to register for this event
|
0800-0845 |
Intro to Aviation Mishap Epi & Admin
| Definitions |
| Denominators |
| Power I |
| Numerators |
Distributions Central Limit Theorem |
|
|
0850-0900 |
Open Questions and Answers/Discussion
|
|
0900-0945 |
| Distributions |
| Skewness & Kurtosis |
| Other Odd Distributions |
| EPI Cheer |
| Validity |
| Confidence Intervals |
| Parametrics |
| Hypotheses |
|
|
0950-1000 |
Open Questions and Answers/Discussion
|
|
1000-1045 |
| Bradford-Hill Criteria for Causation |
| Errors |
| Power II |
| Sampling, Sample Size & Study Limitations |
| Comparing Populations of Parametric Data |
| Controls |
|
|
1050-1100 |
Open Questions and Answers/Discussion
|
|
1100-1145 |
| Hypothesis Testing |
| P-Value |
| Z-Scores |
| The ‘T’ Test |
| 2 by 2 Contingency Tables |
| Taxonomy of the Table |
| Goodman & Kruskal's γ |
|
|
1150-1200 |
Open Questions and Answers/Discussion
|
|
1200-1250 |
Lunch
|
|
1250-1345 |
| Bias & Confounding |
| The Fisher Exact Test |
| Combining Multiple Contingency Tables |
| Mantel-Haenszel |
| Confidence Intervals |
| Understanding Relationship of α, p-value & CI |
|
|
1350-1400 |
Open Questions and Answers/Discussion
|
|
1400-1445 |
| Comparison of Relative Rates |
| Parametric Study Design & Analyses |
| Analysis of Variance (ANOVA) |
|
|
1445-1500 |
Open Questions and Answers/Discussion
|
|
1500-1545 |
| Non-Parametrics |
| Non-Parametric χ2 |
| Modeling |
| Study Design |
| Data Analysis |
|
|
1550-1600 |
Open Questions and Answers/Discussion
|
|
1600-1700 |
Practice Problems
|