Model for Predicting 30-Day Mortality for Cancer Patient in the Hospital

 

Patient selection: terminal cancer patient in the hospital

 

Parameters on admission:

(1) age in years

(2) functional status (difficulty vs no difficult with activities of daily living, ADL)

(3) body temperature in °F

(4) heart rate in beats per minute

(5) systolic blood pressure in mm Hg

(6) pulse oximetry in percent

(7) use of supplemental oxygen

(8) type of admission (elective vs emergent)

 

Parameter	Finding	Points
functional status	problem with ADL	1
	no problem with ADL	0
use of supplemental oxygen	no	0
	yes	1
type of admission	elective	0
	emergent	1

 

X =

= (0.033 * (age)) + (0.452 * (points for functional status)) - (0.146 * (body temperature)) + (0.019 * (heart rate)) - (0.012 * (systolic BP)) - (0.098 * (pulse oximetry)) + (0.862 * (points for supplemental oxygen)) + (0.601 * (points for type of admission)) +18.29

 

where:

• Patients were at greater risk if they were older, had tachycardia, were hypotensive, were hypoxic, needed supplemental oxygen or had an emergency admission.

 

probability of 30-day mortality =

= 1 / (1 + EXP((-1) * X))

 

Limitations:

• The parameters are not clearly defined but are left to inference.

• The discussion involves values of X rather than the derived probabilities.

 

Performance:

• The highest Youden score (based on page 2077) is 0.41 with a sensitivity of 63% and specificity of 78%. The C statistic was 0.79.

• The standard error was 6.79. The Homer Hosmer-Lemeshow chi-square 9.97 with p=0.27.