Challenges of Identifying Similar Boat Accidents of Types Not Classified by USCG’s BARD

U.S. Coast Guard’s Boating Accident Report Database (BARD) remains the primary source for those trying to identify individual recreational boating accidents of a certain type in the United States. It is compiled from accident reports sent to USCG by the individual states and territories.

The Coast Guard’s Annual Boating Statistics report does a nice job of summarizing annual BARD data and providing statistics for the types of accidents the Coast Guard has individual classifications / categories for as seen in our discussion of their 2012 report. You can see many types of accidents are classified in this annual report including capsizing, collision with fixed object, collision with floating object, collision with submerged object, ejected from vessel, falls overboard, grounding, person struck by boat, and person struck by propeller.

As noted several times earlier on this site, the uninitiated may be confused by the presentation of the data (accidents listed as a series of three events) in the annual boating statistics report, but the information is there if you know how to interpret it.

However, those seeking similar statistical information for types of accidents not specifically classified by BARD and reported in USCG’s annual boating statistics report, are left to using BARD and building their own counts, as are those seeking detailed information on individual accidents of a certain type.

For example, we used BARD and other tools to develop lists of several types of accidents not specifically classified by BARD including:

Researchers, Safety Experts, Manufacturers, Safety Advocates, Legal Cases, Regulators, and others use BARD to identify specific accidents of types not classified by BARD to:

  • Attempt to estimate the frequency of this specific type of accident
  • Better understand how this specific type of accident comes about
  • Look for commonalities that might be exploited in mitigating such accidents
  • Estimate the success of proposed solutions
  • Study the severity of injuries and number of fatalities resulting from these accidents
  • Identify specific accidents of a certain type in which their products (boats, marine drives) were involved
  • Develop lists of similar accidents to encourage the industry to fix that specific problem
  • Develop lists of similar accidents or estimate frequency to raise awareness of certain types of accidents and encourage changes in behaviors
  • Develop lists of similar accidents to encourage entrepreneurs to develop solutions
  • Estimate the distribution of these accidents by boat type, drive type, by boat manufacturer, by drive manufacturer, etc.
  • Study the impact of past regulations on these accidents
  • Estimate the impact of proposed regulations on these accidents
  • Confront the industry in court with a list of similar accidents when they say this is the first one ever
  • As part of ongoing safety programs by manufacturers to investigate the involvement of their products or similar products in boating accidents
  • To study the geographical distribution of this specific type of accident
  • To identify the proportion of children or youth involved in this specific type of accident
  • To identify and monitor new types / clusters of accidents not yet on the radar screen, such as how PWC and tubing accidents went unnoticed at first.

Problems Encountered When Trying to Identify Similar Accidents in BARD

While using BARD requires a lot of patience, some database software skills, and acknowledging severe under reporting (not all accidents that should be reported are reported), it remains the only game in town.

BARD underwent a major change in format beginning with the data for 1995. Plus data since 1995 is available in a much easier to interpret format. Earlier data was generally entered as codes into data fields requiring you to lookup or memorize countless codes to interpret a single accident record.

But it is now 2014, 1995 was over 15 years ago and is plenty far enough to go back for most investigations.

So what is the problem?

Individual accident records include a large number of data fields spread over a few different data tables. Among the data fields is one for the “Accident Narrative”. This field contains a brief verbal description of what happened. Some “Accident Narratives” include private information (names, phone numbers, etc.). Therefore, the “Accident Narrative” field is not available in the annual version of BARD made public, referred to as Public BARD.

Recognizing the shortcomings of not having a verbal description of the accident available, beginning with Public BARD 2008, USCG began including a “Redacted Narrative” field. This is the information in the “Accident Narrative” data field with any private information redacted. While some 2008 or later boat accident records do not include a “Redacted Narrative”, the existing “Redacted Narratives” provide a tremendous tool for researchers trying to identify accidents of a certain type not classified by BARD.

For example, when trying to identify accidents in which a boat struck a dredge pipe, BARD classification data can be used to identify accidents in which the boat struck a floating, submerged, or fixed object as Event 1, Event 2, or Event 3. The “Redacted Narratives” for those accidents can then be read or word searched for keywords such as pipe, dredge, line, dredging, etc. to identify which “Reacted Narratives” to read closer to see if the boat struck a dredge pipe.

While this is wonderful, it leaves us without access to “Accident Narratives” or “Redacted Narratives” for accidents from 1995 through 2007.

Our own efforts have turned to news media reports, boating forum discussions, reports made directly to us, legal records, and other sources to try to fill in the earlier data. We turn to the same sources to identify accidents that may not have been reported to USCG or accidents that were reported to USCG but did not identify the exact type of accident we are looking for in their “Redacted Narrative”. Once we are able to identify an accident in the media or other non-BARD sources, we are sometimes able to use the date, geographical location, and other details to identify the accident in BARD. This is extremely time consuming vs. being able to directly search BARD redacted narratives.

Yet, More Problems

Researchers face several more problems in addition to not having access to “Redacted Narratives” prior to 2008. Among them are:

  • Many states to not allow their individual accident reports to be listed in Public BARD. This peaked in 2010 Public BARD when over 20 states plus several territories refused to allow their individual accident data to be included in Public BARD.
  • No access to accidents listed in BARD that did not meet BARD criteria (the accidents that are not included in the releasable Public BARD such as accidents that turned out to be state or federal government boats, commercial boats, boats being tested by a dealer mechanic, involved in a crime, etc.) Much can still be learned from these accidents. They represent instances of these same things happening (struck dredge pipe, fell off bow of pontoon boat, etc.). USCG maintains these accident reports in their version of the BARD database, but they are not available in Public BARD because they did not meet the criteria to be listed in BARD.
  • The current year’s reported accidents cannot be seen until Public BARD is released the next spring/summer, but they can be seen by USCG.
  • Researchers Can Only See the Tip of the Iceberg

    To better illustrate this issue we created an iceberg graphic that illustrates researchers are only able to search the very tip of the iceberg when trying to identify similar accidents that are not of type specifically classified used by BARD. Click on the chart to view a much larger pdf version.

    Searching BARD for similar accidents

    Searching BARD for Similar Accidents

    These shortcomings significantly limit the ability of those trying to reduce the frequency of specific types of accidents that do not fall into existing categories of accidents reported by BARD.

    The shortcomings also preclude the ability of researchers to make an estimate of how many accidents of this type may be included in BARD but not visible to the researcher.

    We will be addressing this issue further, in future posts.

    Leave a Reply