RISK RATIOS OF INJURIES BASED ON MECHANISM IN 137,500 SERIOUSLY
INJURED TRAUMA CENTER
PATIENTS
Purpose:
Mechanism is commonly used as a predictor of specific injuries. However, no
systematic study has been done. The weight attributable to mechanism is
determined
by the likelihood ratio (LR), the ratio of its true positive rate (TPR),
or presence
with a diagnosis, and its false positive rate (FPR), its presence in the
absence of
the diagnosis (LR=TPR/FPR).
Methods:
The relationships of injuries to mechanism was examined for the cohort of
exactly
137,500 seriously injured patients treated in Pennsylvania trauma centers
in the 8
years between the organization of their trauma system in 1986 and 1994.
The incidence
of each three-digit ICD9 and E-code were calculated. For all E-code/ICD9
code pairs,
true and false positive rates and likelihood ratios were calculated. The 95%
confidence limits were calculated for all values.
Results:
The 137,500 patients had 559,307 coded injuries (ave.=4.1); 133 injuries
occurred more
than once; 9 occurred in more than 10% of all patients, led by facial
lacerations in
29% and facial fractures in 20%. Injuries were caused by 103 mechanisms
more than
once; the most common, collisions of motor vehicles, was almost twice as
common (18%)
as the second, falls on level ground (9%). Of 6605 E-code/ICD9 code
pairs, mechanisms
with very significant associations with injuries included facial
fractures with
aircraft landings (LR increased the risk ratio 61 times, 95% CL=253-15
times).
Accidental falls down holes most dramatically increased the risk of vertebral
fractures (LR = 17, 95% CL=21-15) whereas suicidal jumps mostly increased
the risk of
foot fractures (LR = 13, 95% CL=16-11). Falls down stairs reduced the
risk of
injuries to the thoracic blood vessels from the baseline value (LR =
1/32, 95%
CL=1/10-1/98). Listings with 95% CL have been generated for all ICD9
codes, all
E-codes, and the likelihood ratios of all E-code/ICD9 combinations.
Conclusions:
Trauma registries provide clinically useful information about the
relationship between
mechanism and expected injuries when analyzed using risk ratios and
likelihood
ratios. Such information is available for 6605 combinations of
mechanisms and
injuries.
For more information, please contact:
John R. Clarke, M.D.*,
jclarke@gradient.cis.upenn.edu;
Melissa Kritikos**,
mkritiko@linc.cis.upenn.edu;
Bonnie Lynn Webber, Ph.D.**,
bonnie@linc.cis.upenn.edu;
Lawrence W. Bain, Jr.***; and
Waynes S. Copes, Ph.D.***
*Medical College of Pennsylvania and Hahnemann University;
**University of Pennsylvania; and
***Tri-analytics, Inc.
Return to the TraumAID home page.