Development and Validation of a Nomogram Incorporating Colloid Osmotic Pressure for Predicting Mortality in Critically Ill Neurological Patients

Abstract

Backgrounds: The plasma colloid osmotic pressure (COP) values for predicting mortality are not well-estimated. A user-friendly nomogram could predict mortality by incorporating clinical factors and scoring systems to facilitate physicians modify decision-making when caring for patients with serious neurological conditions. Methods: Patients were prospectively recruited from March 2017 to September 2018 from a tertiary hospital to establish the development cohort for the internal test of the nomogram, while patients recruited from October 2018 to June 2019 from another tertiary hospital prospectively constituted the validation cohort for the external validation of the nomogram. A multivariate logistic regression analysis was performed in the development cohort using a backward stepwise method to determine the best-fit model for the nomogram. The nomogram was subsequently validated in an independent external validation cohort for discrimination and calibration. A decision-curve analysis was also performed to evaluate the net benefit of the insertion decision using the nomogram. Results: A total of 280 patients were enrolled in the development cohort, of whom 42 (15.0%) died, whereas 237 patients were enrolled in the validation cohort, of which 43 (18.1%) died. COP, neurological pathogenesis and Acute Physiology and Chronic Health Evaluation II (APACHE II) score were predictors in the prediction nomogram. The derived cohort demonstrated good discriminative ability, and the area under the receiver operating characteristic curve (AUC) was 0.895 [95% confidence interval (CI), 0.840-0.951], showing good correction ability. The application of this nomogram to the validation cohort also provided good discrimination, with an AUC of 0.934 (95% CI, 0.892-0.976) and good calibration. The decision-curve analysis of this nomogram showed a better net benefit. Conclusions : A prediction nomogram incorporating COP, neurological pathogenesis and APACHE II score could be convenient in predicting mortality for critically ill neurological patients.

Keywords: colloid osmotic pressure; critically ill neurological patients; mortality; nomogram; predicting.

Figure 1

Scatter plot of COP values correlated to neurological ICU mortality. The blue line indicates the trend line with a gray shadow representing the 95% confidence interval. The plot visualizes that the mortality risks descends as the COP values ascend in neurological ICU patients. COP, colloid osmotic pressure.

Figure 2

Receiver operating characteristic curve analyses of prediction for the mortality in the development and validation cohort. AUC, the area under the receiver operating characteristic curve; CI, confidence interval.

Figure 3

Nomogram predicting the probabilities of mortality for neurological critically ill patients. To obtain the nomogram-predicted probability, locate patient values on each axis. Draw a vertical line to the point axis to determine how many points are attributed for each variable value. Sum the points for all variables. Locate the sum on the total point line to assess the individual probability of mortality in the neurointensive care unit. APACHE II, Acute Physiology and Chronic Health Evaluation II. The unit of colloid osmotic pressure is mmHg.

Figure 4

Calibration plot for nomogram in the (A) development cohort and (B) validation cohort. (A) The 45° dashed line (“Ideal”) represents ideal predictions, the plot illustrates the accuracy of the best-fit model (“Apparent”) and the bootstrap model (“Bias-corrected”) for predicting ICU mortality. The ticks across the x-axis represent the frequency distribution of the predicted probabilities. (B) The blue dashed line denotes perfect calibration. A smoothing curve (green solid line) and the calibration curve (red solid line) are also overlaid. The distribution of calculated predicted probabilities is overlaid along the horizontal axis. A subset of various statistics useful for validating the model are also shown. Dxy: Somers' Dxy rank correlation between p (predicted possibilities) and y (actual outcome = 0 or 1). C(ROC): the ROC area. U: Unreliability index, for testing unreliability. Brier: Brier score, average squared difference in p (predicted possibilities) and y (actual outcome = 0 or 1).

Figure 5

(A) Decision curve analysis of increasing COP in patients with the model nomogram. The y-axis measures the net benefit. The green line represents the model nomogram. The blue long-dashed line represents the assumption that all patients undertake post-pyloric tube placement. Thin red dashed line represents the assumption that no post-pyloric patient undertakes tube placement. The net benefit was calculated by subtracting the proportion of all patients who are false positive from the proportion who are true positive, weighting by the relative cost of forgoing administration compared with the negative consequences of an unnecessary administration. Threshold probability is the probability of survival from which an intensivist considers that he decides an intervention measure to increase the COP. The decision curve showed that if the threshold probability of a ICU survival is 5% or above, which explicitly covers the range of clinically reasonable threshold probabilities (probability of success >50%), using the nomogram in the current study to predict ICU survival adds more benefit than the administer-all scheme or the administer-none scheme. For example, if the personal threshold probability of a ICU survival is 50% (i.e., the intensivist would opt for administration if the patient's probability was 50%), then the net benefit is 0.045 when using the nomogram to make the decision of whether to start the administration, with added benefit than the administer-all scheme or the administer-none scheme. (B) Clinical impact curve of model nomogram. The red curve (number of high-risk individuals) indicates the number of people who are Classified as positive (high risk) by the model at each threshold probability; the blue curve (number of high-risk individuals with outcome) is the number of true positives at each threshold probability. Clinical impact curve visually indicated that nomogram conferred high clinical net benefit and confirmed the clinical value of the nomogram.