Construction Industry Research and Policy Center
This paper reports on the analysis of fatal events in the construction industry which occurred in calendar year 1997. Four earlier studies1 by Construction Resources Analysis (CRA) analyzed the causes of fatal events in this industry in 1991-1992, 1993-1994, 1995, and in 1996.
The data analyzed in this report, provided by OSHA, consist of narrative descriptions of the 604 fatal events resulting from accidents which occurred in construction during calendar year 1997. As in the earlier studies, fatal events included in the OSHA data which resulted from heart attacks, fainting, etc., when not caused by the construction activity, were omitted from the analysis; this resulted in the omission of 25 records in 1997. Although the Occupational Safety and Health Act of 1970 requires employers to report fatalities to OSHA within eight hours of the occurrence of the event, all records were not available at the time of writing (it is estimated that 98 percent of the fatal event records were in the data base analyzed), so the results reported upon here do not allow a year-to-year analysis of changes in the absolute number of fatal events.
Each narrative record consists of a brief description of the event leading to the fatality. The event descriptions were read by the three authors and each was classified into one of 29 causative categories. The narrative descriptions varied greatly in clarity and thoroughness, so when the authors could not agree on the cause of an event the event was classified as "other/unknown". Some of the narratives provided thorough descriptions of the actions leading to the fatality, but many were so brief and lacking in detail that it was difficult to extract a specific cause.
1 An Analysis of Fatal Events in the Construction Industry. 1991-1992 (1993), An-Analysis of Fatal Events in the Construction Industry. 1993-1994 (1995), An Analysis of Fatal Events in the Construction Industry, 1995 (1996), and An Analysis of Fatal Events in the Construction Industry. 1996 (1997). Construction Resources Analysis, University of Tennessee, Knoxville.
In this analysis the authors originally planned to classify each fatal event according to: (1) type of construction (new or addition, alteration or rehabilitation, maintenance or repair, demolition, other); (2) estimate of total project value (seven dollar-value categories beginning with "under $50,000" and ending with 120,000,000 and over"); (3) 17 end-use categories, such as "single-family housing, multi-family building, " "commercial building, 19 street or highway," etc.; and (4) the construction operation being performed that caused the fatal event (a list of 71 construction operations such as "backfilling and compacting," "cutting concrete pavement," "erecting structural steel," "installing equipment (HVAC and other," etc.). Lists of these variables and their definitions were provided to OSHA by CRA, and coding for them was subsequently required in an IMIS screen for each fatal construction event inspected by OSHA. However, CRA's review of these data revealed that coded data for an event were often internally inconsistent and often did not comport with corresponding narrative descriptions. Consequently, it was determined by OSHA and CRA that the poor quality of the coded data precluded its inclusion in this report. The data analyzed in this report, therefore, are restricted to the narrative descriptions of the fatal events where the authors were able, in most cases, to classify the events with moderate certainty according to 29 types of causes, essentially the same types as were used in CRA's previous fatality studies. (See Table 1.)
Table 1. Construction Fatality Event Causes, 1997
A. Distribution of Fatal Events by Cause
Table 1 shows the cause classification system, the number of times each cause represented a fatal event in 1997 and the relative frequency of each cause. It can be seen that "fall from/through roof' led all other causes in number of fatal events (72 or 11.9 percent of total fatal events), followed by "fall from/with structure (other than roof)" (55 or 9.1 percent). The third leading cause was "crushed/run over of non-operator by operating construction equipment" (49 or 8.1 percent). The fourth leading cause was "electric shock by equipment contacting power source" (46 or 7.6 percent); the fifth leading cause was "lifting operation" (33 or 5.5 percent); and, the sixth leading cause was "fall from/with ladder" (32 or 5.3 percent). The number and relative frequencies of the remaining causes of the 604 fatal events analyzed may be read directly from Table 1.
Table 2 shows a comparison of the ranks of the causes in 1997 with the average rank of the causes of fatal events during the period 1991 - 1996. It can be seen that the overall rank pattern of the causes in 1997 is very similar to the rank pattern in 1991 - 1996. An overall statistical comparison of the correlation of the rank in 1997 with the average rank in 1991 1996 was calculated using a Spearman rank correlation procedure.2 The correlation obtained was +.93, p C .0001, indicating that the ranks of the causes in the two time periods are highly positively correlated, i.e., did not change significantly between 1991 1996 and 1997. Since averaging the 1991-1996 ranks removed inter-year variance, a somewhat lower correlation would be expected between 1996 and 1997 ranks of causes, i.e., a measure of the short-term cycle as opposed to a longer-term trend. However, the Spearman rank correlation between 1996 and 1997 causes was calculated and found to be higher, +.94, p V .0001, indicating that the pattern changed even less between 1996 and 1997 (although no statistical significance is inferred)
The correlation result is not surprising given that the general composition of construction output, and therefore the mix of construction operations required to produce the output, was probably very similar during the time periods examined. This interpretation implies that the rank of a cause is a function of the magnitude of exposure to the cause and/or the inherent danger associated with the cause.
It should be noted that the annual rate of fatal events involving accidents on construction sites (inspected by OSHA and reported as of June in the following year) increased between 1991 - 1996 and 1997, averaging 582 fatal events per year in the 1991 - 1996 period and 604 fatal events in 1997, a 3.8 percent increase3. However, construction employment increase during the two periods under study, averaging 4,898(000) in 1991 - 1996 and 5,629(000) in 1997, which means there was a decrease in fatal events per 1000 employees, i.e., .119 in the 1991 - 1996 data and .107 in 1997, a 10.1 percent decrease4. However, a comparison of fatal events per 1000 in 1996 (as reported in mid-June of the following year) with 1997 data shows an increase of 1.9 percent.
2Sidney Siegel, Nonparametric Statistics for the Behavioral Sciences (New York: McGrawHill Book Co., Inc., 1956), p.219.
3The reader is cautioned not to infer statistical significance to annual increases or decreases in the number of accident-caused fatal events on construction sites for two reasons: (1) the data file supplied to CRA by OSHA contained only the fatalities investigated by OSHA (but probably not even all of those); and (2) the authors deleted from the data illness fatalities not related to accidents, e.g., heart attacks and seizures, and fatal events involving construction employees but not on construction sites or the contractor's yard, e.g., off-site auto accidents.
Appendix A: Definition of Causes
* Includes fatalities resulting from asphyxiation/fire/explosion/drowning of trapped operators.