Perioperative mortality for management of hepatic neoplasm: a simple risk score

To develop a population-based risk score for stratifying patients by risk of in-hospital mortality following procedural intervention for hepatic neoplasm.

There has been growing support for the value of surgical management of hepatic neoplastic disease, both primary and metastatic. Advances in surgical and ablative technologies have contributed to a decrease in the mortality associated with these procedures. However, multiple patient-, disease- and treatment-related factors can contribute to perioperative morbidity and mortality.

Using the Nationwide Inpatient Sample from 1998 to 2005, a retrospective cohort of patient-discharges for hepatic procedures with a concurrent diagnosis of hepatic primary or metastatic neoplasm to the liver was assembled. Procedures were categorized as lobectomy, wedge resection, or enucleation/ablation. Logistic regression and bootstrap methods were used to create an integer score for estimating the risk of in-hospital mortality using patient demographics, comorbidities, procedure type, tumor type, and hospital characteristics. A randomly selected sample of 80% of the cohort was used to create the risk score. Testing was conducted in the remaining 20% validation-set.

In total, 12,969 patient-discharges were identified. Overall in-hospital mortality was 3.45%. Predictive characteristics incorporated into the model included: age, sex, Charlson comorbidity score, procedure type, hospital type, and type of neoplasm. Integer values were assigned to these, and used to calculate an additive score. Five clinically relevant groups were assembled to stratify risk, with a 36-fold gradient in mortality. Rates in the groups were as follows: 0.9%, 2.5%, 6.8%, 17.6%, and 35.9%. In the derivation set, as well as in the validation set, the simple score discriminated well, with c-statistics of 0.76 and 0.70, respectively.

An integer-based risk score can be used to predict in-hospital mortality after hepatic procedure for neoplasm, and may be useful for preoperative risk stratification and patient counseling.