Letter to Editor: Using Proper Methods to Identify Patients With Cirrhosis in Administrative Databases Is Crucial to Correctly Predict Outcomes

Potential conflict of interest: Dr. Swain advises and received grants from Gilead, Intercept, Allergan, and Novartis. He is on the speakers' bureau for Abbott. He received grants from CymaBay, GSK, and Genkyotex. To the Editor: We would like to congratulate Mumtaz et al.1 for their important study attempting to develop and validate a risk score to predict 30‐day hospital readmission in patients with decompensated cirrhosis using the US nationwide readmission database (NRD). Identifying patients with cirrhosis at high risk for readmission is critical for developing processes to effectively address this problem. However, flaws in patient identification challenge the utility of the risk score outlined in this study. Specifically, the authors failed to discuss their observed high readmission rates in the context of published data from the large multistate study by Tapper et al.2 The NRD consists of patient data from multiple state inpatient databases (SID). Tapper et al. used SID data from five large geographically diverse states to describe 12.9% to 24.2% 30‐day readmission rates in patients with 1‐3 cirrhosis decompensation features, respectively. In contrast, Mumtaz et al. report a baseline 30‐day readmission rate of 27%—an unexpected significant difference in readmission rates, given that the NRD is based on SID. This variation could be due to the case definition of decompensated cirrhosis. Specifically, the authors defined decompensated cirrhosis as the presence of cirrhosis plus any of the following: ascites, hepatic encephalopathy, variceal bleeding, or spontaneous bacterial peritonitis. However, in reviewing their coding, the International Classification of Disease Ninth Edition (ICD‐9) clinical modification codes 348.30, 348.39, and 780.97 were included, which describe general encephalopathy or altered mental status. The authors cite two studies to support the use of these codes. However, no study has in fact validated these codes in patients with cirrhosis. Similarly, the authors used nonvalidated codes for coagulopathy. There are important studies describing the validity of ICD‐9 codes to identify patients with cirrhosis.3 The use of well‐validated codes to identify patients with cirrhosis in administrative databases is critical for accuracy of conclusions made from using the data. We re‐extracted the cohort of patients with cirrhosis from the NRD 2013 (n = 14,325,172) using the authors' codes and methodology (cirrhosis plus at least one decompensation condition) and identified 107,317 patients with decompensated cirrhosis, compared with 175,761 patients used for the authors' analysis. Therefore, it is very likely that other nonvalidated decompensation features such as coagulopathy were included in their case definition. An additional concern is the inclusion of patients with cirrhosis who were readmitted for non‐cirrhosis‐related conditions, such as epilepsy or pneumonia, which would increase readmission rates.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.