Tu et al reported a predictive model for asymptomatic spontaneous bacterial peritonitis in a patient with cirrhosis. This can help to identify a patient at risk. The authors are from the Chinese People's Liberation Army General Hospital in Beijing.
Patient selection: cirrhosis, ascites, asymptomatic for peritonitis (absence of fever, abdominal pain, tenderness and rebound pain), no sign of infection elsewhere
Parameters:
(1) MELD score
(2) PMN count in ascitic fluid, per microliter
(3) percent neutrophils in the peripheral blood
(4) hepatocellular carcinoma
(5) renal dysfunction
|
Parameter
|
Finding
|
Points
|
|
MELD
|
< 15.5
|
0
|
|
|
>= 15.5
|
0.312
|
|
PMN count
|
< 10.95 per µL
|
0
|
|
|
>= 10.95 per µL
|
0.263
|
|
percent neutrophils
|
< 66.15%
|
0
|
|
|
>= 66.15%
|
0.184
|
|
hepatocellular carcinoma
|
no
|
0
|
|
|
yes
|
0.233
|
|
renal dysfunction
|
no
|
0
|
|
|
yes
|
0.189
|
value of X =
= SUM(points for alll of the parameters) + 0.018
probability of spontaneous bacterial peritonitis =
= 1 / (1 + EXP((-1) * X))
Interpretation:
• The minimum value of X is 0.018, so the minimum probability is just over 50%.
Performance:
• The area under the ROC curve is 0.87.
