# 2236 Predicting Conversion in Laparoscopic Colorectal Surgery: A
Logistic Regression Model.
Tonia M. Young-Fadok, Heung W Lee, Charles H. Darby, Rochester, MN
AIM: Conversion from laparoscopic to open colorectal operation results in
loss of patient benefits and increased costs. Our aim was to identify factors
that predict conversion.
METHOD: From 7/92-6/99, 315 laparoscopic colorectal procedures were
performed; 69 were converted. Chart review identified factors potentially
associated with conversion: age, gender, body mass index (BMI), prior operation,
diagnosis, site of lesion, operative procedure, date of operation,
and immunosuppressives. Variables were categorized on a-priori considerations.
Univariate screening identified predictors significantly associated
with conversion at the 0.25 level in order not to exclude potentially important
predictors. Univariate logistic regression models investigated linear
relationships between categorical variables and conversion; both linear
and quadratic relationships were explored between continuous predictors
and conversion. Interaction factors were sought. Stepwise selection was
used to choose predictors to be included in the final prediction model.
RESULTS: In the multivariate model, risk of conversion was significantly
associated with BMI, age, diagnostic category (diverticular disease versus
non-diverticular), and site of lesion (see Table). The final model was: ln(P/
(1-P)) = 0.06BMI + 0.03AGE + 0.88DIAG + 0.49SITE - 6.8996 where (P = risk
of conversion). The c-value for this model was 0.71. The exponential of the
resultant value X (i.e. ex) gives the odds ratio of conversion (P/(1-P)), and
from this the risk of conversion, P, can be calculated.
CONCLUSION: A logistic regression model was developed to predict risk
of conversion. The risk is increased with increasing BMI and age, plus diverticulitis.
This model will be employed clinically to guide operative approach
and counsel patients.
|