Limits...
A mathematical model for predicting glucose levels in critically-ill patients: the PIGnOLI model.

Zhang Z - PeerJ (2015)

Bottom Line: Insulin rate was significantly associated with blood glucose reduction (coefficient: -0.52, 95% CI [-1.03, -0.01]).Conclusion.The study developed the PIGnOLI model for the initial insulin dose setting.

View Article: PubMed Central - HTML - PubMed

Affiliation: Department of Critical Care Medicine, Jinhua Municipal Central Hospital, Jinhua Hospital of Zhejiang University , Zhejiang , PR China.

ABSTRACT
Background and Objectives. Glycemic control is of paramount importance in the intensive care unit. Presently, several BG control algorithms have been developed for clinical trials, but they are mostly based on experts' opinion and consensus. There are no validated models predicting how glucose levels will change after initiating of insulin infusion in critically ill patients. The study aimed to develop an equation for initial insulin dose setting. Methods. A large critical care database was employed for the study. Linear regression model fitting was employed. Retested blood glucose was used as the independent variable. Insulin rate was forced into the model. Multivariable fractional polynomials and interaction terms were used to explore the complex relationships among covariates. The overall fit of the model was examined by using residuals and adjusted R-squared values. Regression diagnostics were used to explore the influence of outliers on the model. Main Results. A total of 6,487 ICU admissions requiring insulin pump therapy were identified. The dataset was randomly split into two subsets at 7 to 3 ratio. The initial model comprised fractional polynomials and interactions terms. However, this model was not stable by excluding several outliers. I fitted a simple linear model without interaction. The selected prediction model (Predicting Glucose Levels in ICU, PIGnOLI) included variables of initial blood glucose, insulin rate, PO volume, total parental nutrition, body mass index (BMI), lactate, congestive heart failure, renal failure, liver disease, time interval of BS recheck, dextrose rate. Insulin rate was significantly associated with blood glucose reduction (coefficient: -0.52, 95% CI [-1.03, -0.01]). The parsimonious model was well validated with the validation subset, with an adjusted R-squared value of 0.8259. Conclusion. The study developed the PIGnOLI model for the initial insulin dose setting. Furthermore, experimental study is mandatory to examine whether adjustment of the insulin infusion rate based on PIGnOLI will benefit patients' outcomes.

No MeSH data available.


Related in: MedlinePlus

Graphical presentation of the BG predicted by the model including FP terms (red line) and the model with linear terms (blue line).Both models appeared similar in predicting BG. The initial BG was controlled at its mean value of 195.9 mg/dl.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4465940&req=5

fig-1: Graphical presentation of the BG predicted by the model including FP terms (red line) and the model with linear terms (blue line).Both models appeared similar in predicting BG. The initial BG was controlled at its mean value of 195.9 mg/dl.

Mentions: The parsimonious model was fitted to address the problem of instability. Graphical presentation showed that although the interaction term was statistically significant, the magnitude was of marginal clinical significance (Fig. 3 in Supplemental Information 2). Therefore, I opted not to incorporate interaction terms in the parsimonious model. Figure 1 shows the scatter points predicted by FP model and simple linear model, and the two lines were close to each other. Visual inspection of the graph indicates the use of parsimonious model would not compromise the prediction accuracy of the model.


A mathematical model for predicting glucose levels in critically-ill patients: the PIGnOLI model.

Zhang Z - PeerJ (2015)

Graphical presentation of the BG predicted by the model including FP terms (red line) and the model with linear terms (blue line).Both models appeared similar in predicting BG. The initial BG was controlled at its mean value of 195.9 mg/dl.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4465940&req=5

fig-1: Graphical presentation of the BG predicted by the model including FP terms (red line) and the model with linear terms (blue line).Both models appeared similar in predicting BG. The initial BG was controlled at its mean value of 195.9 mg/dl.
Mentions: The parsimonious model was fitted to address the problem of instability. Graphical presentation showed that although the interaction term was statistically significant, the magnitude was of marginal clinical significance (Fig. 3 in Supplemental Information 2). Therefore, I opted not to incorporate interaction terms in the parsimonious model. Figure 1 shows the scatter points predicted by FP model and simple linear model, and the two lines were close to each other. Visual inspection of the graph indicates the use of parsimonious model would not compromise the prediction accuracy of the model.

Bottom Line: Insulin rate was significantly associated with blood glucose reduction (coefficient: -0.52, 95% CI [-1.03, -0.01]).Conclusion.The study developed the PIGnOLI model for the initial insulin dose setting.

View Article: PubMed Central - HTML - PubMed

Affiliation: Department of Critical Care Medicine, Jinhua Municipal Central Hospital, Jinhua Hospital of Zhejiang University , Zhejiang , PR China.

ABSTRACT
Background and Objectives. Glycemic control is of paramount importance in the intensive care unit. Presently, several BG control algorithms have been developed for clinical trials, but they are mostly based on experts' opinion and consensus. There are no validated models predicting how glucose levels will change after initiating of insulin infusion in critically ill patients. The study aimed to develop an equation for initial insulin dose setting. Methods. A large critical care database was employed for the study. Linear regression model fitting was employed. Retested blood glucose was used as the independent variable. Insulin rate was forced into the model. Multivariable fractional polynomials and interaction terms were used to explore the complex relationships among covariates. The overall fit of the model was examined by using residuals and adjusted R-squared values. Regression diagnostics were used to explore the influence of outliers on the model. Main Results. A total of 6,487 ICU admissions requiring insulin pump therapy were identified. The dataset was randomly split into two subsets at 7 to 3 ratio. The initial model comprised fractional polynomials and interactions terms. However, this model was not stable by excluding several outliers. I fitted a simple linear model without interaction. The selected prediction model (Predicting Glucose Levels in ICU, PIGnOLI) included variables of initial blood glucose, insulin rate, PO volume, total parental nutrition, body mass index (BMI), lactate, congestive heart failure, renal failure, liver disease, time interval of BS recheck, dextrose rate. Insulin rate was significantly associated with blood glucose reduction (coefficient: -0.52, 95% CI [-1.03, -0.01]). The parsimonious model was well validated with the validation subset, with an adjusted R-squared value of 0.8259. Conclusion. The study developed the PIGnOLI model for the initial insulin dose setting. Furthermore, experimental study is mandatory to examine whether adjustment of the insulin infusion rate based on PIGnOLI will benefit patients' outcomes.

No MeSH data available.


Related in: MedlinePlus