Parameters:
(1) Glasgow coma score (ranges from 3 to 15)
(2) postoperative BUN in mg/dL
(3) number of associated organ injuries
(4) preoperative ALT in IU/L
(5) preoperative systolic blood pressure in mm Hg
(6) preoperative base excess in mEq/L
(7) preoperative platelet count per 10^4 per cubic mm (per µL)
(8) presence or absence of intestinal injuries (0 if absent, 1 if present)
where:
• A platelet count of 200,000 per µL would be 20 * 10^4 per cubic mm (per µL), or 20 in the equation.
• The reference ranges and methods for BUN and ALT were not given.
• A very negative base excess is associated with acidosis.
probability for overall death =
= 0.5830 – (0.03656 * (Glasgow Coma Score)) + (0.009954 * (postoperative BUN)) + (0.05494 * (number of associated organ injuries)) + (0.0001175 * (preoperative ALT)) – (0.002967 * (preoperative systolic blood pressure))
probability for hemorrhagic death =
= 0.1787 + (0.0001755 * (preoperative ALT)) - (0.03228 * (preoperative base excess)) - (0.01272 * (preoperative platelet count))
where:
• Equation 2 in the text shows (-0.0001755) and a positive 0.01272. However, the coefficients in Table 3 on page 391 are the opposite. If the equation as given is used, answers inconsistent with the text are returned.
probability for a septic death
= 0.4878 – (0.004259 * (preoperative systolic blood pressure)) + (0.212 * (intestinal injuries))
NOTE: I wondered if the y was the value for the logistic regression equation (see below), but when this is tried incorrect results are returned for the data in Table 1 on page 390.
probability of death =
= 1 / (1 + EXP((-1) * (y)))
Performance:
• Overall death: sensitivity 86%, specificity 100%, accuracy 98%.
• Hemorrhagic death: sensitivity 75%, specificity 100%, accuracy 98%.
• Septic death: sensitivity 67%, specificity 100%, accuracy 97%.
Limitations:
• The adjusted R^2 values were 0.34 for septic death, 0.67 for hemorrhagic death and 0.71 for overall death.
• The prognosis would be affected by surgical techniques employed.
• The absence of data on methodology for BUN and ALT may limit transportability.
