Abstract

Significance to the VA: Complications following major cancer surgery occur in over 30% of cases and have a major impact on quality of life and disease-free survival. Having two or more psychosocial risks independently confers a 3.4-fold increase in the odds of a postoperative complication. While this association has been established, it is not yet known which psychosocial risks are most associated with surgical complication and which risks are acutely modifiable. Psychosocial vulnerability in the immediate postoperative period substantially complicates care delivery. Housing instability and other psychosocial risk factors (food insecurity, mental health comorbidities, and criminal justice involvement) are not routinely screened for in surgically treatable disease. The long-term goal of this proposal is to address gaps in cancer surgery care by developing psychosocial support interventions that reduce surgical complications ultimately improve cancer survival. Innovation and Impact: Although preoperative optimization programs recognize the importance of risk stratification and potential psychosocial needs, there is limited evidence for which risk factors to target and to what degree these risks are modifiable. This study will inform the best means to identify psychosocial risks among patients undergoing cancer surgery using existing VA data and identify those factors most strongly associated with perioperative complications. We intend to use these findings in future work to then identify a broad set of intervenable psychosocial risk factors for veterans undergoing cancer surgery. Specific Aims & Methodology AIM 1. Compare detection of housing instability and other psychosocial risk factors in veterans undergoing major cancer surgery using structured data elements versus natural language processing (NLP). Using a national cohort (10,000+ per year of analysis, 2008-2022) of elective cancer surgery patients, housing instability and other psychosocial risk factors will be assessed using conventional definitions of these risks in VA structured data fields. Identification of housing instability and other psychosocial risk factors will also be performed using an internally developed NLP algorithm that incorporates a supervised machine learning algorithm. Finally, 100 discordant cases (structured vs. NLP) will be used to compare performance. AIM 2. Compare the association of postoperative complications and alternative measures of housing instability and other psychosocial risk factors in veterans undergoing major cancer surgery. We will use the enhanced cohort from Aim 1 that combines surgical outcomes data with new definitions of housing instability and other psychosocial risk factors to test for the independent association between psychosocial risk and worse outcomes after surgery. AIM 3. To tailor and pilot via expansion of an existing intervention (renamed Surgical Vulnerability Program, SVP) to mitigate the effects of perioperative psychosocial risk on postoperative complications following major cancer surgery. Informed by Aims 1 and 2, we will recruit 30 patients enriched with psychosocial risks identified by the outperforming approach and tailor an existing intervention better support these patients through the perioperative period. Care partners will be interviewed to evaluate feasibility, acceptability, and perceived utility. Pathway to Translation/Implementation: A more accurate delineation of psychosocial risks and a clear association with postoperative complications will better characterize preoperative surgical risk as well as its potential modifiability. A “better” designation of psychosocial risk will facilitate targeting social support interventions in the pre- and postoperative period that mitigate their impact on surgical outcomes. Moreover, the use of NLP algorithms to identify novel psychosocial risk factors prior to surgery will allow for high- throughput, nationwide risk stratification and opportunity to intervention.

NIH Reporter Project Information: https://reporter.nih.gov/project-details/11348036


PUBLICATIONS:
None at this time.