• Ichiro Hisatome, Katsuyuki Tomita, Ryohei Kato, Tomoyuki Ikeuchi, Hirokazu Touge, Toshihiro Hamada, Tamotsu Goto, Haruaki Ninomiya, Naoyuki Otani, Masanari Kuwabara, Satoshi Miyazaki, and Akira Yamasaki.
• Department of Cardiology Medicine, NHO Yonago Medical Center, Japan.
• Intern. Med. 2024 Sep 11.

AbstractBackground Hypouricemia, defined as a serum uric acid (SUA) level ≤2 mg/dL, could be a risk factor for death in hospitalized patients. However, how explanatory variables can explain hypouricemia as an objective variable in a logistic regression analysis remains unknown. Purpose To predict the risk factors for hypouricemia in hospitalized patients using a robust Bayesian logistic (RBL) model. Methods This study retrospectively enrolled patients who visited Yonago Medical Center between April 2020 and March 2021. The association between potential risk factors and hypouricemia was analyzed using the RBL model in Python-modulated PyMC3. The final model was selected based on the lowest Watanabe-Akaike information criterion (WAIC). Results Of the 618 patients, 64 (10.4%) had hypouricemia. Based on the model according to the lowest WAIC, independent risk factors for hypouricemia were febuxostat [odds ratio (OR) 5.46, 95% confidence interval (CI) 2.32-13.4], amino acids in parenteral nutrition (OR 5.19, 95% CI 1.62-15.1), TMP-SMX (OR 4.20, 95% CI 1.66-10.9), emaciation (OR 3.48, 95% CI 1.75-7.21), and serum sodium level (OR 0.90, 95% CI 0.84-0.96). Conclusion The RBL model predicted amino acids in parenteral nutrition, TMP-SMX, emaciation, and low serum sodium levels for hypouricemia, in addition to the authentic risk factor febuxostat.

7 keywords...

hide…