Automatic detection of cirrhosis in hospitalized patients: a pragmatic experience

We evaluated the relevance of a systematic automatic detection of cirrhosis using biochemical markers in hospitalized patients.

We automatically calculated three free biochemical tests (APRI, Fib-4, and Forns) in patients consecutively hospitalized in our university hospital between July and September, 2010. Patients >18 years not known to suffer from chronic liver disease, were contacted to undergo liver stiffness measurement (LSM) as a reference diagnostic tool. To limit false positives, we required at least one APRI≥2 (indicating cirrhosis) and Fib-4>3.25 and/or Forns>6.9, without obvious overestimation.

A total of 10,035 APRI, 9903 Fib-4, and 1250 Forns were available in 4074 patients. The fibrosis tests were independently influenced by the location of the patient, especially Cardiology (Lower Forns) and Hematology/Oncology Departments (higher APRI, Fib-4, and Forns). Overall, 101 patients (2.48%) were suspected to have cirrhosis. LSM identified two cases of cirrhosis (LSM>13 kPa). In intent-to-diagnose analyses, the highest positive predictive values of the APRI, Fib-4, and Forns for the diagnosis of cirrhosis were 1.98, 1.98, and 11.76%, respectively. The positive predictive value never exceeded 50% in per-protocol analyses when considering patients with numerous positive results of the fibrosis tests.

In hospitalized patients, automatic detection of cirrhosis on the basis of APRI, Fib-4, and Forns was inefficient because of too many false-positive results.