A Predictive Risk Stratification Model for Delirium After Major Urologic Cancer Surgeries
Sources of Funding: None
Introduction
Post-operative delirium is a common complication in the elderly and contributes to increased healthcare costs, mortality, cognitive decline, and hospital length of stay. No definitive pre-operative risk prediction model for patients undergoing major urologic cancer surgeries is currently available.
Methods
Using the Premier Hospital Database, we retrospectively identified patients who had undergone radical prostatectomy (RP), radical nephrectomy (RN), partial nephrectomy (PN), and radical cystectomy (RC) from 2003 to 2013. Post-operative delirium was defined using International Classification of Disease, Ninth Revision (ICD-9) codes, as well as post-operative use of antipsychotics, sitters, and restraints. Potential pre-operative risk factors of delirium were extrapolated from patient, hospital, and surgical characteristics. A pre-operative delirium risk prediction score was developed from our multivariate model. Its performance was quantified using Receiver Operating Characteristic (ROC) analysis. All analyses were survey-weighted and clustered by hospitals to achieve estimates generalizable to the US population.
Results
We identified 165,387 patients representing a weighted total of 1,097,355 patients from 490 hospitals who had undergone RP, RN, PN, or RC. Our model revealed a wide range of clinical and demographic factors that significantly contribute to the risk for post-operative delirium (Figure A). Our delirium risk score was associated with the development of post-operative delirium (Odds Ratio: 1.31, 95% CI 1.29-1.33, p <0.001, Figure B), and it demonstrated good discrimination in the prediction of delirium (Receiver Operator Characteristic [ROC] area = 0.76, 95% CI, 0.76-0.77, Figure C). The ability of the risk score to predict delirium was consistent across surgical subgroups, and the risk score was also predictive of the duration of delirium (Incidence Rate Ratio = 1.07, 95% CI 1.04-1.11, p<0.001).
Conclusions
The preliminary results of our pre-operative delirium risk prediction tool are promising given its consistency with published delirium risk factors and ease of use. Further validation of this model will shed insight about its clinical utility to identify patients at high-risk of post-operative delirium who may benefit from early therapeutic intervention.
Funding
None
Ross Krasnow
Tammy Hsieh
Adam Kibel
James Rudolph
Benjamin Chung
Steven Chang