by Jim Fisher
posted 10/14/2001

[Ed. note: This editorial is in no way intended to trivilaize any death or injury in our amusement parks. It is intended to illustrate just how sensational the coverage of amusement safety has been in the last few years. -J]

In 1954, a small volume by Darrell Huff entitled How to Lie with Statistics was first published. This book is still in print 47 years later. Huff details some of the many ways in which statistics may be misused and distorted either deliberately or accidentally. This book belongs on your bookshelf.

The Consumer Product Safety Commission (CPSC) either failed to read this book or read it too well. The CPSC has again released its annual report on injuries and deaths related to amusement rides, and once again it is meaningless and distorted. There are serious problems with this report in several areas.

Lack of a Meaningful Sample for Injuries

For any statistical study the first requirement is a good set of data. Otherwise the garbage in, garbage out principle applies.

The CPSC report is based on data from the National Electronic Injury Surveillance System. (NEISS) This is a sampling of 100 hospital emergency rooms in the United States. It is a stratified sample selected to include hospitals of various sizes and children’s hospitals. The number of hospitals included is less than 2% of the hospitals in the country.

NEISS seems to be a pretty representative sample of hospital emergency rooms, but that is all it is; a sample of "hospital emergency rooms." As such, it is a good way to sample injuries from hazards that are widespread, such as automobiles, baby cribs, garden hoses, beds, and televisions, all of which by the CPSC’s own numbers injure more people than amusement rides. These potential injury sources exist in millions of locations across the country. The injuries from these sources tend to be pretty uniformly distributed across the country.

NEISS is not a good way to sample injuries caused by hazards that exist in a limited number of locations, like the roughly 200 amusement parks in the US. This is especially true since the size of the amusement parks varies greatly. In fact, the most likely number of amusement parks sampled by NEISS is just four, though it is probable that the actual number sampled is not exactly four. If a hospital were added to the sample that received the emergency cases from just one large amusement park with an attendance of 3 million visitors it could be expected to raise the estimated injuries by 50%. (There are about 20 parks in the country that attract 3 million or more visitors.)

To use an extreme example of how distorted this type of sampling can be, suppose you wanted to look at injuries caused by complete-circuit roller coasters over 300 feet tall. Since there is only one coaster that meets this description, either NEISS includes the emergency room that receives patients from Cedar Point and will catch almost all of the these injuries, or NEISS does not include that emergency room and will probably won’t catch any of these injuries. (No criticism of Cedar Point intended, but I’m sure someone stubs his toe occasionally on Millennium Force.)

NEISS is actually a better way to sample injuries from mobile rides than from fixed rides. The mobile rides tend to be relocated each week or two during the season, increasing the number of locations at which the hazard exists. There are probably several thousand locations for carnivals during the course of a year compare to about 200 for amusement parks.

There is no other sampling available that is more meaningful at this time, so there is no meaningful data at this time.

Looking At the Injury Data

This year’s report from the CPSC gives only the estimated number of injuries. Last year’s report gave us a peek at the kind of data that is actually used to compose these reports. Here are the actual numbers of injuries that were reported to the CPSC by NEISS.

Year	Fixed Location	Mobile Ride	Mobility Unknown	Total Injuries
1999	87	44	36	167
1998	62	29	33	124
1997	63	29	25	117
1996	49	33	13	95
1995	60	43	18	121

Fixed location basically means an amusement park ride. Mobile ride means a carnival ride, and unknown means that the data did not show whether the ride was fixed or mobile. It is from these 200 injuries reported in the NEISS data that the CPSC projects a total of over 10,000 injuries nation wide.

Why the Increase in Reported Injuries?

Last year the CPSC also gave us a little insight into the source of the increased number of injuries reported from fixed rides. The report indicated that one hospital, identified as #58 by NEISS, was responsible for most of the increase in injuries reported. The report goes on to state that 10% of the total injuries came from this hospital in the early 90’s and 44% in the late 90’s. If we use these percentages to factor out the injuries from hospital #58 what do we discover? The number of injuries reported from the other 99 hospitals was unchanged during this period. The entire change came from one hospital!

In statistics this type of data point is called an outlier and is usually ignored. The CPSC argues that excluding this emergency room would make their sample not representative of emergency rooms. The lie that the CPSC is sneaking in is the idea that a representative sample of emergency rooms is also representative of amusement park injuries. As we’ve illustrated, this is hardly the case.

How do we account for this kind of increase in injuries at one hospital? There are several possible explanations. A new amusement park may have opened near the hospital or an existing park greatly expanded. An amusement park near the hospital may have serious safety problems. The hospital that receives injuries from a large amusement park may have changed from somewhere else to hospital #58. Or, an amusement park near the hospital may have changed its policy from treating minor injuries at the first aid station in the park to sending them to the local emergency room. These are all possibilities, but the CPSC doesn’t offer any explanation.

Death Statistics

The data available for deaths is much better than the data for injuries, since nearly every amusement ride death is recorded. However, the CPSC report is haphazard even in its listing of deaths. The report is supposed to cover only injuries and deaths to consumers caused by mechanical amusement rides. By the CPSC’s own definition deaths to be specifically excluded include deaths of employees and deaths on water slides, which are not considered amusement rides for the purpose of the report.

The CPSC report includes an appendix that lists all of the deaths included since 1987. A review of the deaths listed by the CPSC shows that the list includes several deaths that do not meet the CPSC’s own definition of an amusement ride death. In 2000 one of the two deaths recorded was the death that occurred on a water slide in Connecticut. In 1997 there was a death included that was caused by an asthma attack, stated by the hospital report to have not been related to the ride. In 1995 a death caused by s guest falling down stairs while exiting the ride was included, as was a death from a preexisting condition. In 1993 a death to a construction worker assembling a carnival ride was included. Obviously the CPSC did not look at these deaths closely. In 1998 the CPSC says there were seven deaths, but their own listing of deaths only shows six deaths and one of those was not due to an amusement ride.

When this year’s report was released, CPSC Chairman Ann Brown referred to an increase in amusement ride related deaths. She said this even though her own department’s report stated that there was no significant trend in deaths. What do the actual numbers look like? It’s very difficult to determine any trends since there are so very few deaths that occur on amusement rides. However, I assembled the following list based on data from the CPSC and other sources.

Amusement Ride Deaths

Year	Fixed Location	Mobile Rude	Unknown	Total		Year	Fixed Location	Mobile Ride	Unknown	Total
2000	1	0	0	1		1986	0	4	1	5
1999	6	0	0	6		1985	1	1	0	2
1998	3	2	0	5		1984	2	3	0	5
1997	3	0	0	3		1983	4	1	0	5
1996	2	1	0	3		1982	1	1	0	2
1995	2	0	1	3		1981	3	3	2	8
1994	3	0	1	4		1980	3	1	1	5
1993	1	1	1	3		1979	3	4	0	7
1992	0	1	1	2		1978	10	2	3	15
1991	3	0	0	3		1977	1	2	2	5
1900	0	0	0	0		1976	2	0	7	9
1989	3	0	0	3		1975	1	0	0	1
1988	2	1	2	5		1974	0	1	5	6
1987	4	0	0	4		1973	0	1	2	3

If you don’t see any pattern over the last 20 years, you are correct. There isn’t any pattern. There is a decrease in the number of deaths compared to the 70’s, but overall, the number of deaths has remained the same or decreased over the last 30 years despite greatly increased attendance. So far this year there have been two deaths caused by amusement rides in the United States. The first death occurred on a portable ride regulated by the CPSC, not at an amusement park. The second death was caused by a woman ejected from her seat on a giant flume at Knott’s Berry Farm.

Inflatable Amusement Devices

This year the CPSC added a section to their annual report on inflatables. The CPSC says that the number of injuries on these has doubled in recent years. What the CPSC avoids is the fact that these devices have increased in popularity in recent years. If there are twice as many injuries but three times as many people using inflatables then they have actually become safer. The CPSC answers neither of the important questions about the safety of these rides. First, is the actual risk per use going up or down, and second, is the risk high or low?

Conclusions

The CPSC data set for injuries is meaningless because it doesn’t represent a proportionate number of hospitals near a proportionate number of amusement parks of various sizes. Any conclusions based on this data are equally meaningless.

The CPSC has too many errors in its death data. This indicates that they probably don’t bother to look at any of their data with any degree of care.

Deaths are down compared to the 1970’s. There is no trend in deaths over the last 20 years, which is positive considering that attendance is up.

So why does the CPSC keep issuing these reports? Why are they spending tax money to prepare reports on fixed amusement devices that the law does not allow them to regulate, and for which they have no decent data to base a report on?