The protocol for this matched case-control study was approved by the Institutional Review Boards of Columbia University and Johns Hopkins University. The cases consisted of air carrier crashes (i.e., “accidents”) investigated by the National Transportation Safety Board (NTSB). The NTSB is an independent federal agency charged by Congress with investigating every civil aviation crash in the United States. Controls were obtained from the incident data system of the Federal Aviation Administration (FAA). All aviation crashes are reported to the NTSB, and all incidents are recorded by the FAA. The NTSB records aviation crash investigation data using the core Factual Report (NTSB Form 6120.4) and a set of supplemental forms. The core Factual Report includes detailed information on crash circumstances, aircraft, the pilot involved in the crash and a written narrative outlining all factors contributing to the crash. Incident investigation and reporting follow the same procedures as crash investigation. Procedures for investigating and reporting aviation crashes and incidents are described in detail by the federal government (8
The federal government defines an aviation crash as “an occurrence associated with the operation of an aircraft in which any person experiences serious injury or death within 30 days of the aviation event or in which the aircraft receives substantial damage.” (8
) Serious injury refers to any injury resulting in a “fracture of any bone (excluding simple fracture of fingers, toes, or nose), causing severe hemorrhage, nerve, muscle, or tendon damage, involving any internal organ, second-or third-degree burns, burns affecting more than 5% of the body surface, and requiring hospitalization for more than 48 hours.” (8
) Substantial damage to an aircraft is defined as “damage or failure that requires major repair or replacement of the affected component.”(8
) An aviation incident is defined as an occurrence other than a crash associated with the operation of an aircraft which affects or could affect the safety of operations. (8
) Incidents encompass selected criminal acts reported to or by law enforcement agencies, emergency evacuations of aircraft, in-flight major component failures, and any reportable event that threatened or caused damage to aircraft or injury to persons (e.g., near mid-air collisions, pilot deviations, and maneuvers resulting in the loss of separation).
Cases and controls are limited to crashes and incidents occurring in the U.S. and involving scheduled or unscheduled flights governed by Title 14 Code of Federal Regulations Part 121 (14 CFR Part 121), and to aircraft made by three major manufacturers: Boeing, McDonnell-Douglas, and Airbus. Commuter air carriers and air taxis (14 CFR Part 135) and general aviation (14 CFR Part 91) were excluded from this study. Beginning in 1997, the definition of passenger-carrying aircraft covered under Part 121 was lowered from a capacity of 31 or more to 10 or more passengers. This change was effective at the same time in both cases and controls.
A total of 373 cases were selected from the 668 air carrier crashes identified from the NTSB aviation crash data system for the years 1983 through 2002. Of the excluded crashes, 57 occurred outside the geographic study area, 89 had missing information on pilot age or basic weather conditions, five were caused by criminal acts, and 144 involved aircraft made by companies other than Boeing, McDonnell-Douglas, and Airbus. We excluded crashes involving aircraft that were not made by the three major manufacturers to minimize the confounding effect of factors related to aircraft and flight characteristics and to focus on the major airlines. The 144 crashes that were excluded based on aircraft make involved planes manufactured by a variety of companies. These aircraft tended to be smaller than aircraft of the three major makes. The diversity of manufacturers of these aircraft also made it unfeasible to match cases and controls on the variable of aircraft make. Controls were matched with cases on calendar year of the index crash. A 1:2 case-control sampling ratio was used to increase the study power.
Age of the pilot-in-command was measured as a continuous variable in years and analyzed as a categorical variable: 29–39, 40–49, and 50–59 years. Pilot error was determined only for cases according to the probable causes of the crash as identified in the NTSB investigation reports. Total flight hours were categorized into four groups based on the approximate quartiles: ≤9000, 9000–11999, 12000–15999, and ≥16000 h. Total flight time in the last 90 days was categorized as: <120, 120–169, 170–199, and ≥200 h. Basic weather conditions were coded in the NTSB investigation reports as visual meteorological conditions (VMC) or instrument meteorological conditions (IMC).
The geographic entities examined in this study include nine FAA regions defined as follows: Eastern Region (District of Columbia, Delaware, Maryland, New Jersey, New York, Pennsylvania, Virginia and West Virginia); Central Region (Iowa, Kansas, Missouri, and Nebraska); Alaska; Great Lakes (Illinois, Indiana, Michigan, Minnesota, North Dakota, Ohio, South Dakota, Wisconsin); New England (Connecticut, Maine, Massachusetts, New Hampshire, Rhode Island, Vermont); Northwest Mountain (Colorado, Idaho, Montana, Oregon, Utah, Washington, Wyoming); Southern (Alabama, Florida, Georgia, Kentucky, Mississippi, North Carolina, South Carolina, Tennessee); Southwest (Arkansas, Louisiana, New Mexico, Oklahoma, Texas); and Western-Pacific (Arizona, California, Hawaii, and Nevada).
In the exploratory data analysis, the characteristics of cases and controls were compared using Miettinen’s method (21
) to account for the matched design. Conditional logistic regression models were used to estimate odds ratios (ORs) of crash involvement according to the individual variables. Total flight time in the last 90 days was not included in the multivariate analysis because of missing data for 39% of the case pilots and 16% of the control pilots. Both crude and adjusted ORs are presented with 95% confidence intervals (CI). Data analysis was conducted using the Statistical Analysis System (SAS) software (SAS Institute, Cary, NC).