Views: 1,269 | Downloads:
29
| Responses: 0
| XML | Respond to this article | Alert & updates | Request permissions | Email to a friend |
Original ArticleOpen Access
Predicting Outcome in Pediatric Near-drowning
Plubrukarn R 
,
Tamsamran S
,
Tamsamran S Background
: Near-drowning is common in children and has a high mortality rate. Some
survivors remain in a vegetative state after the accident and are a great burden to their family and society.
Objectives
: To find out whether outcomes on near drowning can be reliably identified early
in the course of illness.
Method
: Medical records of 72 children admitted to Queen Sirikit National Institute of Child
Health, Bangkok, Thailand, for treatment of near drowning from January 1993 to December 2001 were
retrospectively studied. Stepwise multivariate discriminant analysis was used to identify the power of
variables achieving highest overall accuracy in minimizing errors for predicting poor outcome in intact
survivors.
Results
: The patients were identified into three groups: functionally intact, vegetative and
dead groups, it was found that a combination of physical examination in the emergency department
(ED), the need for cardiopulmonary (CPR) in the ED, amount of adrenaline given during CPR, and high
blood sugar achieved an overall accuracy of 83 per cent. When categorizing patients into good outcome
versus
poor outcome (the combination of the vegetative and dead group were the poor outcome group).
The variables mentioned above achieved an overall accuracy of 98 per cent. Good outcome survivors
could be correctly predicted with no error, but error occurred when poor outcome survivors were pre-
dicted to be good outcome in 3 per cent. Glasgow coma
score~
5, the need for CPR in the ED and blood
sugar> 300 mg/dl were selected clinical variables found to have optimum predictive abilities with an
overall accuracy of 96 per cent, but showed an error of 6 per cent in predicting poor outcome from
functional intact survivors (unpredicted good outcome).
Conclusion
: From the present study discrimination analysis cannot accurately separate all
intact survivors from the vegetative groups, but can prospectively differentiate unpredicted good out-
come from vegetative or dead groups. When using only simple clinical classification systems, unpredicted
S502
: Near-drowning is common in children and has a high mortality rate. Some
survivors remain in a vegetative state after the accident and are a great burden to their family and society.
Objectives
: To find out whether outcomes on near drowning can be reliably identified early
in the course of illness.
Method
: Medical records of 72 children admitted to Queen Sirikit National Institute of Child
Health, Bangkok, Thailand, for treatment of near drowning from January 1993 to December 2001 were
retrospectively studied. Stepwise multivariate discriminant analysis was used to identify the power of
variables achieving highest overall accuracy in minimizing errors for predicting poor outcome in intact
survivors.
Results
: The patients were identified into three groups: functionally intact, vegetative and
dead groups, it was found that a combination of physical examination in the emergency department
(ED), the need for cardiopulmonary (CPR) in the ED, amount of adrenaline given during CPR, and high
blood sugar achieved an overall accuracy of 83 per cent. When categorizing patients into good outcome
versus
poor outcome (the combination of the vegetative and dead group were the poor outcome group).
The variables mentioned above achieved an overall accuracy of 98 per cent. Good outcome survivors
could be correctly predicted with no error, but error occurred when poor outcome survivors were pre-
dicted to be good outcome in 3 per cent. Glasgow coma
score~
5, the need for CPR in the ED and blood
sugar> 300 mg/dl were selected clinical variables found to have optimum predictive abilities with an
overall accuracy of 96 per cent, but showed an error of 6 per cent in predicting poor outcome from
functional intact survivors (unpredicted good outcome).
Conclusion
: From the present study discrimination analysis cannot accurately separate all
intact survivors from the vegetative groups, but can prospectively differentiate unpredicted good out-
come from vegetative or dead groups. When using only simple clinical classification systems, unpredicted
S502
Download:
PDF