![](bg6.png)
relatively high beta coefficient in the regression, but
did not appear in any of these paths of the decision
tree. However, other variables with higher beta
coefficients in the regression, such as abdominal
diagnosis, lung diagnosis, SBP and temperature,
were also important in the tree.
This has implications for practice since clinicians
want to apply models of risk at the bedside. Often it
is not feasible to collect data on 20 variables, such as
those we found in the regression or cluster and
models that are easy to use to either rule out patients
who are not at risk of sepsis or those who are at risk
would be more useful. To test any model we have to
ensure that it is reliable and valid. Here we have
shown with the 30 patient accuracy test that our tree
is reliable and it approaches 100% validity. Our
analysis also illustrates the value of multiple
methods: 1) in our analysis, regressions can be used
to provide a broad estimate of risk, and 2) a more
precise estimate in this case can be made using a
decision tree. In a separate paper, we will compare a
cluster analysis approach to a decision tree. This is
outside the scope of this paper. A valid approach for
future research is the comparison of cluster analysis,
decision trees and regression analysis.
6 CONCLUSIONS
Multiple methods of analysing clinical data provide
different perspectives on models of risk of disease.
To develop robust models researchers may want to
consider regression to get a broad perspective on the
risk and utilize decision trees to provide more
parsimonious models.
This study has several strengths. This was a
prospective observational cohort and the
determination of sepsis used standard criteria. The
large sample size provided a large number of
variables that we could use for our analyses.
Future research will now entail testing the
decision tree paths in practice to determine which
path is most reliable and valid as well as completing
a more in depth measure of the accuracy of the
regression and decision tree models. We did not
have an opportunity to test other methods such as
cluster analysis, Bayesian methods or neural
networks, which we hope to do in the future.
ACKNOWLEDGEMENTS
We would like to thank Corey Hilliard for assisting
with manuscript preparation.
REFERENCES
American College of Chest Physicians/Society of Critical
Care Medicine Consensus Conference: definitions for
sepsis and organ failure and guidelines for the use of
innovative therapies in sepsis. (1992). Crit Care Med,
20, 864-874.
Alberti, C., Brun-Buisson, C., Burchardi, H., Martin, C.,
Goodman, S., Artigas, A. Sicignano, A., Palazzo, M.,
Moreno, R., Boulme, R., Lepage, E., & Le Gall, R.,
(2002). Epidemiology of sepsis and infection in ICU
patients from an international multicentre cohort study.
Intensive Care Med, 28, 108-121.
Angus, D.C., Linde-Zwirble, W.T., Lidicker, J., Clermont,
G., Carcillo, J., & Pinsky, M.R. (2001). Epidemiology
of severe sepsis in the United States: analysis of
incidence, outcome, and associated costs of care. Crit
Care Med, 29, 1303-1310.
Bernard, G.R., Vincent, J.L., Laterre, P.F., LaRosa, S.P.,
Dhainaut, J.F., Lopez-Rodriguez, A. Steingrub, J.S.,
Garber, G.E., Helterbrand, J.D., Ely, E.W. & Fisher,
C.J.Jr. (2001). Efficacy and safety of recombinant
human activated protein C for severe sepsis. N Engl J
Med, 344, 699-709.
Brun-Buisson, C., Doyon, F., Carlet, J., Dellamonica, P.,
Gouin, F., Lepoutre, A. Mercier, J.C., Offenstadt, G.
& Regnier, B. (1995). Incidence, risk factors, and
outcome of severe sepsis and septic shock in adults. A
multicenter prospective study in intensive care units.
French ICU Group for Severe Sepsis. JAMA, 274,
968-974.
Dellinger, R.P., Levy, M.M., Carlet, J.M., Bion, J., Parker,
M.M., Jaeschke, R., Reinhart, K., Angus, D.C., Brun-
Buisson, C., Beale, R., Calandra,T., Dhainaut, J.F.,
Gerlach, H., Harvey, M., Marini, J. J., Marshall, J.,
Ranieri, M., Ramsay, G., Sevransky, J., Thompson,
B.T., Townsend, S., Vender, J.S., Zimmerman, J.L. &
Vincent, J. L. (2008). Surviving Sepsis Campaign:
international guidelines for management of severe
sepsis and septic shock: 2008. Crit Care Med, 36, 296-
327.
Knaus, W.A., Draper, E.A., Wagner, D.P., & Zimmerman,
J.E. (1985). APACHE II: a severity of disease
classification system. Crit Care Med, 13, 818-829.
Knaus, W. A., Wagner, D. P., Draper, E. A., Zimmerman,
J. E., Bergner, M., Bastos, P. G., Sirio, C.A., Murphy,
D.J., Lotring, T. & Damiano, A. (1991). The
APACHE III prognostic system. Risk prediction of
hospital mortality for critically ill hospitalized adults.
Chest, 100, 1619-1636.
Letarte, J., Longo, C. J., Pelletier, J., Nabonne, B., &
Fisher, H. N. (2002). Patient characteristics and costs
COMPARISON OF ANALYTIC APPROACHES FOR DETERMINING VARIABLES - A Case Study in Predicting the
Likelihood of Sepsis
95