Srpski Arhiv za Celokupno Lekarstvo (Jan 2019)

Performance of a calculator for diagnosing the cause of liver damage

  • Petrović-Subić Narcisa,
  • Kojić Miroslav,
  • Janković Slobodan M.,
  • Stefanović Srđan M.

DOI
https://doi.org/10.2298/SARH180504064P
Journal volume & issue
Vol. 147, no. 1-2
pp. 27 – 33

Abstract

Read online

Introduction/Objective. Making a calculator that would recognize patterns of abnormal liver function tests and link them to the most probable etiology could help clinicians in their initial orientation towards a definitive diagnosis in patients with liver damage. The aim of our study was to design, construct, and validate a calculator that based on a pattern of abnormalities in liver function tests of a patient with liver damage would propose the most probable etiology. Methods. Patterns of abnormal liver function tests for certain etiology of liver damage were extracted from distributions of actual values taken from reports in medical literature about patients whose etiology of liver damage was proven by reliable diagnostic tests. After setting up the calculator with the patterns extracted, its diagnostic value was checked under real-life conditions, on a sample of patients with liver damage whose etiology was established by the gold standard of diagnostics (biopsy or else). The calculator validation study was carried out at the Military Medical Academy in Belgrade during a two-year period (2015–2016). Results. For all tested diagnoses, the calculator demonstrated a highly significant difference between the area under the receiver-operator curves’ values and the value of 0.5 (p < 0.001), and high level of sensitivity (more than 90%, except for the model for chronic hepatitis) as well as relatively high specificity (more than 75%) were noted, indicating good ability of the calculator to detect etiology of liver damage. Conclusion. New calculators showed satisfactory sensitivity and specificity for revealing major liver damage etiologies. [Project of the Serbian Ministry of Education, Science and Technological Development, Grant no. 175007]

Keywords