Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
com
© 2004
Original Article
ABSTRACT
Background: When developed in 1981 by Moore and colleagues the Penetrating Abdominal Trauma Index (PATI)
was used to identify trauma patients at risk of postoperative complications. Methods of trauma quantification have
been extensively developed but their outcome evaluations have been naïve, subject only to basic statistical analyses.
The accuracy of PATI was assessed by means of Receiver Operating Characteristics (ROC) curve analysis.
Aims: To evaluate PATI by means of ROC curve analysis and establish its sensitivity and specificity for predicting
the morbidity and mortality in penetrating abdominal trauma.
Settings and design: Patients who attended the University hospital trauma unit; prospective cohort study.
Methods: This was a prospective study over an 11-month period, patients with penetrating abdominal trauma who
underwent laparotomy were enrolled. Initial assessment of the patients was done following the ATLS guidelines.
Patients were stratified on the basis of those who developed complications, no complications and postoperative
mortality. PATI was calculated based on operative findings and the outcomes were measured on the basis of
complications or mortality in the postoperative period.
Statistical analysis: Receiver Operating Characteristics curve analysis and Logistic regression analysis using
PATI as independent variable.
Results: Eighty-nine consecutive patients underwent laparotomy for abdominal trauma, 49 for stab wounds and 40
for gunshot wounds. Patients with complications scored a mean PATI of 21.47; those who died scored a mean PATI
of 25.29. ROC curve analysis of PATI scores in patients who developed complications showed results of 42.1%
sensitivity and 91.4% specificity. Sensitivity and specificity for mortality prediction was 42.9% and 91.5% respectively.
Conclusion: The statistical evaluation of trauma scores should be performed using an adequate methodology to
avoid naive evaluations. ROC curve evaluation of this trauma score index allows a comprehensive study of the
instrument’s performance, avoids data simplification and permits cross-analysis of different trauma score indexes.
KEY WORDS
Penetrating Abdominal Trauma Index, ROC curve, instrument performance, Venezuela.
How to cite this article: Gomez-Leon JF. Penetrating abdominal trauma index: Sensitivity and specificity for morbidity and mortality by roc
analysis. Indian J Surg 2004;66:347-51.
The Penetrating Abdominal Trauma Index (PATI) is phases of trauma management for injury evaluation.
described as a method for quantifying the risk of Thus, pre-hospital triage instruments, in-hospital and
complications following penetrating abdominal trauma, outcome instruments are used to assess the level or
introduced in 1981 by Moore et al. PATI provided a type of care needed, for decision-making, and for
useful means to investigate and served as a tool in the evaluation of trauma outcome for policy decisions,
decision-making process when managing penetrating respectively; that is why performance evaluation of
abdominal trauma.1 In order to quantify trauma, injury these instruments have important clinical
severity instruments (ISI) are used in the different consequences. 2 ISI are evaluated via either naïve
Address for correspondence: Jose Francisco, Department of Surgery, Division of General Surgery, Hospital Universitario Dr. Luis Razetti,
Barcelona, Anzoategui, Venezuela. E-mail: jfgldoc@Hotmail.com
Paper Received: June 2004. Paper Accepted: October 2004. Source of Support: Nil.
evaluations which depend upon the prevalence of the severity of trauma outcome was also analyzed as a
disease or outcome of interest in the study population, dichotomous variable, thus, patients were stratified as
or calculations of sensitivity and specificity. Naïve survivors or deaths.
assessments suppose the instrument’s performance in
a single figure: accuracy prediction of events that are Statistics: Mean score results for the two subsets of
affected by prevalence-dependent populations that populations were obtained by analysis of variance
cannot be generalized, but can be valid within a specific (ANOVA) of the difference of points obtained by each
study. Logistic regression analysis and sensitivity- subset of patients i.e. complicated vs. non-complicated
specificity assessments avoid the aforementioned and survival vs. non-survival. For logistic regression
problems because they evaluate the instrument’s and ROC analysis PATI was used as the independent
performance in two separate populations, where variable; complications and trauma severity outcome
sensitivity equals the number of patients correctly (death) were used as binary dependent variables, that
predicted to live and specificity the number of patients is, binary response was set as complications present
correctly predicted to die, who actually die.3 Several and no complications present until discharge and
ISI have been correctly evaluated by means of these survival or no survival of patients. Logistic regression
figures, these include the Injury Severity Score, New analysis was performed to determine the association
Injury Severity Score, ICD-9 codes and Revised Trauma of the relative predictive power of the independent
Score. PATI lacks sensitivity, specificity or logistic variable with respect to the categorical dichotomous
regression analysis evaluations as evidenced in the complication/no complication and survival/death
available literature. It is being used as a powerful tool dependent variables. ROC curve analysis was
in our setting to asses penetrating abdominal trauma performed to summarize the instrument’s performance
management. The goal of this paper is to evaluate this with respect to discriminating patients who will or not
instrument’s behaviour, and study its ability to correctly complicate and those who will or will not survive. Thus,
prognosticate patients using the advantages of logistic mean PATI scores for each subset of the population
regression analysis and ROC curve analysis over the were plotted and sensitivity/specificity calculated by
traditional statistical techniques. means of non-parametric estimates of the Area under
the curve. Statistical analysis was done using SPSS®
MATERIALS AND METHODS 9.0. Significance was attributed to P values < 0.05.
Table 1: Mean PATI scores in patients who Table 2: Logistic Regression results dependent
complicated and not complicated variable: Complication
PATI Variable B *Wald †
Model Chi-Square Exp(B)
Complications Mean N Std. Deviation PATI 0.9003 10.7654 13.366 1.0945
*No 11.24 70 8.33 *P=0.001; †P =0.0003
**Yes 21.47 19 12.97
Total 13.43 89 10.33 1.00
*95 % CI for mean PATI from 9.50 to 12.97
**95 % CI for mean PATI from 18.77 to 24.16
.75
20%
.50
15%
P e r c e n t c o m p l ic a t i o n
Sensitivity
.25
10%
0.00
5% 0.00 .25 .50 .75 1.00
Non Complic ation 1-Specificity
0% Complication
1Diagonal
- Specificity
segments are produced by ties.
2 6 11 15 19 24 32 45
4 9 13 17 21 26 41 Figure 2: ROC Curve PATI scores in the complication subset
Penetrating Abdominal Trauma Index
PATI results in the Death/Survivors subset of In our study, we could observe how increasing PATI
patients score values correlate with complications in the
Mean PATI score in patients who died was 25.29 ± postoperative period and even mortality secondary to
15.98 (CI 95% for mean PATI from 21.97 to 28.60), complications; this is congruent with the published
compared to the patients who survived (which account literature. Our aim was then to evaluate how this
for all those who complicated and did not complicate) instrument’s perfor mance fits with the standard
that was 12.41 ± 9.15 (95% CI for mean PATI from statistical techniques currently used to asses the
10.50 to 14.31) (Table 4). Figure 3 shows how survivors models.
are separated from non-survivors according to
increasing PATI scores, although, overlapping exists PATI has been used to measure injury severity in
as in almost all injury scores. The logistic regression abdominal trauma in order to assist the surgeon in
model depicts that patients with this score of 25 or categorizing the patients at risk of developing
more are 1.08 (95% CI 1.02 to 1.16) times more likely complications, and even in decision-making techniques
to die when compared to patients scoring less points. for repairing intra-abdominal organs according to its
Coefficient’s Wald statistic for PATI is 7.46, which is severity score.4-8
20%
Table 3: Coordinates of the Curve, PATI scores in
the complications subset 16%
Test Result Variable(s): PATI
Positive if Greater
P er c e n t S u r vi v o r s /D e a th
12%
Than or Equal To* Sensitivity 1-Specificity
1.00 1.000 1.000
8%
2.50 1.000 0.871
3.50 1.000 0.857
4.50 1.000 0.786 4%
Table 7: Complications and cause of death in the survival / not survival subset of patients
Count
Outcome
Survival Not survival Total
Complications Intra-abdominal abcess 1 1
Evisceration 2 2
Wound infection 3 3
Urinary tract infection 1 1
Urinary retention 1 1
ARDS 2 2
SEPSIS 2 2
MODS 1 1
Pulmonary embolism 1 1
Total 82 7 89
.25
indices could produce enough evidence to set
statements to encourage their use in the management
0.00
0.00 .25 .50 .75 1.00
of the trauma patients in our countries.
1-Specificity
Diagonal segments are produced by ties.
1 - Specificity REFERENCES
Figure 4: ROC Curve PATI scores in the outcome subset
1. Moore EE, Dunn EL, Moore JB, Thompson JS. Penetrating abdomi-
Several methods to asses abdominal trauma have been nal trauma index. J Trauma 1981;21:439-45.
2. Lett RR, Hanley JA, Smith JS. The comparison of injury severity
evaluated with adequate statistical techniques that instrument performance using likelihood ratio and ROC curve
quantify sensitivity and specificity, but they include analyses. J Trauma 1995;38:142-8.
invasive approaches as laparoscopy or image studies.9- 3. Rutledge R, Osler T. The ICD-9-based illness severity score: A
10 new model that outperforms both DRG and APR-DRG as predic-
Among the trauma indices that have used this tors of survival and resource utilization. J Trauma 1998;45:791-9.
methodology of evaluation are the injury severity score 4. O’Neill PA, Kirton OC, Dresner LS, Tortella B, Kestner MM. Analysis
and the new injury severity score.2 of 162 colon injuries in patients with penetrating abdominal
trauma: Concomitant stomach injury results in a higher rate of
infection. J Trauma 2004;56:304-12.
A Trauma index instrument’s behaviour in categorizing 5. Nelson R, Singer M. Primary repair for penetrating colon injuries.
a specific characteristic in trauma patients should be Cochrane Database Syst Rev 2003;(3):CD002247.
6. Mickevicius A, Klizaite J, Tamelis A, Saladzinskas Z, Pavalkis D.
evaluated by a methodology that avoids data [Penetrating colorectal trauma: Index of severity and results of
simplification and therefore bias in the conclusions treatment]. Kaunas: Medicina; 2003;39:562-9.
obtained by its application. 7. Bulger EM, McMahon K, Jurkovich GJ. The morbidity of penetrat-
ing colon injury. Injury 2003;34:41-6.
8. Sikic N, Korac Z, Krajacic I, Zunic J. War abdominal trauma: Use-
Logistic regression analyses allow evaluations on how fulness of Penetrating Abdominal Trauma Index, Injury Severity
an instrument’s performance can be graded with Score, and number of injured abdominal organs as predictive fac-
respect to the relative predictive power of a study tors. Mil Med 2001;166:226-30.
9. Ortega AE, Tang E, Fr oes ET, Asensio JA, Katkhouda N,
variable on the ef fect of a specific characteristic Demetriades D. Lapar oscopic evaluation of penetrating
outcome. ROC curve analysis evaluates an instrument’s thoracoabdominal traumatic injuries. Surg Endosc 1996;10:
performance in a comprehensive manner; as results 19-22.
10. Patel VG, Walker ML. The role of “one-shot” intravenous pyelo-
are summarized in a simple way, this methodology gram in evaluation of penetrating abdominal trauma. Am Surg
avoids data loss and simplification.2 1997;63:350-3.