Characterization of High Mortality Probability Operations at National Surgical Quality Improvement Program Hospitals

The decision to operate on high-risk patients can be challenging. To our knowledge, there is little evidence available for guidance and wide variability in practice.¹ Surgeons may encounter clinical scenarios in which they believe that the chances of improving a patient’s life are minimal compared with the risks of surgery, yet the notion of futility in surgery remains controversial.²

To better understand extreme risk surgery and determine whether the National Surgical Quality Improvement Program (NSQIP) preoperative risk calculator can help to define futility, the NSQIP database was used to analyze patients who underwent surgery despite a high preoperative probability of death.

All patients in the NSQIP database who underwent surgery from 2007 to 2016 with an estimated preoperative probability of 30-day mortality greater than 75% were identified. The built-in NSQIP risk calculator, which calculates preoperative probability of mortality on a continuous numerical scale using outcomes of previous operations and patient and operative characteristics, was used.³ A 75% predicted mortality was selected as a cutoff level to explore a group at the end of the spectrum of predicted mortality in which the decision to operate would require special consideration. Variables included timing of mortality, postoperative diagnosis, comorbidities, and discharge destination.

This study was approved by the Yale University Institutional Review Board. The need for patient consent was waived as there was no direct patient contact and all records were deidentified. The study was conducted from December 1, 2018, to July 1, 2019.

In total, 3148 patients underwent surgery with a preoperative predicted mortality greater than 75% (0.06% of 5 457 202 operations during this time). Of these, 1007 patients (32.0%) were older than 80 years, 1500 (47.%) were totally functionally dependent, 2206 (70.1%) were receiving mechanical ventilation prior to surgery, 2326 (73.9%) had septic shock, and 698 (22.2%) had disseminated cancer. The most common diagnoses associated with extremely high-risk surgery were vascular insufficiency of the intestine (554 [17.68]), sepsis (412 [13.1%]), bowel perforation (274 [8.7%]), and Clostridium difficile colitis (192 [6.1%]); 828 patients (36.0%) had been transferred from a separate institution and 1716 procedures (86.7%) were performed by general surgeons (Table 1). Almost half of the patients had died by postoperative day 7 (1401 [44.5%]) (Table 2). There were few patients at the high end of predicted mortality, but even in the greater than 90% predicted mortality group, there were 4 outliers (9.5% of survivors with discharge information available) who were discharged home within 30 days.

This analysis was performed to better understand surgery that is performed on extremely high-risk patients; inform surgeons, patients, and families; and help to avoid low-value surgery in the future. These results may help to define a group of patients who undergo surgery with a low likelihood of benefit. In general, these individuals are elderly, dependent patients who develop emergent abdominal conditions with no good alternatives to surgery (eg, bowel ischemia or perforation) and often transfer from other institutions. These attributes suggest that there is an overlap at times between surgery and end-of-life care.

The outcomes of early in-hospital death and long hospitalizations among this cohort seem to be at odds with the preferences for deescalation of care for patients with terminal conditions.⁴ Ongoing efforts have focused on the use of palliative care and shared decision-making to avoid low-value procedures toward the end of life,⁵ but a key ingredient in successful deescalation of treatment is time,⁶ which is usually limited in emergency surgical situations. The present results may suggest that families and patients who are totally dependent or already have disseminated cancer be encouraged to consider what their wishes may be in the event of a major surgical emergency and perhaps include these decisions formally in advanced directives before acute illness.

In contrast, this analysis demonstrates that NSQIP preoperative characteristics are not sufficient to predict total futility, as there are inaccuracies and outliers at the end of the spectrum. The predicted mortality was considerably higher than the observed mortality in the most high-risk groups. Ultimately, the ability to extrapolate tools to predict patients who absolutely will not benefit from surgery will be an ongoing issue. Future study of high-risk patients is needed to explore additional factors associated with costs, do-not-resuscitate status, and quality of life following surgery. Our inability to include these factors are limitations of the present study.

Accepted for Publication: August 2, 2019.

Corresponding Author: Kevin Y. Pei, MD, Department of Surgery, Texas Tech University Health System, 3601 Fourth St, Lubbock, TX 79430 (kevin.pei@ttuhsc.edu).

Published Online: October 30, 2019. doi:10.1001/jamasurg.2019.3750

Author Contributions: Drs Resio and Pei had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Resio, Chiu, Pei.

Acquisition, analysis, or interpretation of data: Resio, Zhang, Pei.

Drafting of the manuscript: Resio, Pei.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Resio, Chiu, Zhang.

Supervision: Pei.

Conflict of Interest Disclosures: None reported.

Meeting Presentation: This study was presented at the Association of Veterans Affairs Surgeons Annual Meeting; Seattle, Washington; April 28, 2019.