Preliminary validation of new prognostic scoring system in patients with invasive meningococcal disease

Introduction

Invasive meningococcal disease (IMD) caused by an invasive strain of Neisseria meningitidis group C:2a:P1,2,P1,5 and lately by an invasive strain group B:2a:P1,2 as well appeared in the Czech republic in 1993 and completely changed severity of the disease. Incidence and severity of the disease in the West Bohemian region is the highest in our country. We have published preliminary results of our several studies of prediction factors of this disease. Haemocoagulopathy seemed to be of great and reliable prediction value. Fibrinogcn, Quick, platelets and AT III were found as the most reliable factors in patients treated in the interdisciplinary ICU .

Purpose of the study

Prospectively to develop and preliminarily validate a model for the probability of hospital death in patients with meningococcal invasive disease.

Methods

Defined continuous variables (fibrinogen, Quick, platelets and AT III) and categorical variable (survival) were collected. A preliminary model was developed on the base of 82 patients data. Logistic regression was used for the development of the model, which was evaluated by the receiver operating characteristic (ROC) analysis. The mortality rate preliminary validation is the outcome was studied.

Studied group characteristics

Unlike previous studies we studied now the group of patients with IMD treated in the whole West Bohemian region. Six patients from total number of 88 patients were excluded from the study due to incomplete data (n = 82).

Results

(1) Logistic regression model has validity coefficient = 0.469: P(Death=1)= 1/(1+EXP(Sß₀ + ß_iX_i))

(2) Receiver operating characteristic for logistic model - ROC report: (see Table overleaf).

Set up of optimal value of P(Death=1) by logistic regression model was based on minimal number of the false death predictions (C) along with maximum correct death predictions (A), (D=survival prediction, B=false death prediction). We looked up in ROC report the value of P(Death=1) with maximum ratio between sensitivity and false positivity. The value P(Death=1) = 0.256 increased the percentage of correctly classified patients on the level of 91.46%. The area under the ROC curve is 0.869. By the substitution to the next equation enables simply calculate the preliminary validated prediction. Result = 0.256 means the prediction of the death probability 91.46%.

P(Death=1)= 1/(1+EXP(1.200-0.020×TROMBO-0.448×QUICK-0.592×ATIII+0.228×FBG))

Table 1

Table 2