Assessment of the Modified Frailty Index and Risk Factors for Point-of-Origin Discharge Failure (PODF) Following Lobectomy
Victoria Yin*2, Sean C. Wightman1, Scott M. Atay1, Takashi Harano1, Anthony W. Kim1
1Division of Thoracic Surgery, Department of Surgery, Keck School of Medicine of USC, Los Angeles, CA; 2Keck School of Medicine of USC, Los Angeles, CA
Objective: To compare the predictive power of three models for point-of-origin discharge failure (PODF) following lobectomy: Model A with patient characteristics and comorbidities identified from bivariate analyses, Model B with the modified Frailty Index-5 (mFI-5), and Model C with mFI-5 and surgical approach (minimally invasive vs open).
Design: Retrospective cohort study using 2017-2020 American College of Surgeons-National Surgical Quality Improvement Program (ACS-NSQIP) database. Patients who were admitted from home and discharged to any other location (rehab, separate acute care, skilled care, unskilled/other facility) were considered to have PODF. Multivariate logistic regression was used to model PODF. Separate models were used to assess mFI-5 as a predictor because its calculation (one point for each: dependent functional status, diabetes, chronic obstructive pulmonary disease, congestive heart failure, and hypertension) shared many comorbidities with Model A. mFI-5 was used as an ordinal independent variable in the models. Predictive power was assessed using the C-statistic.
Setting: Hospitals included in the ACS-NSQIP.
Patients: 16,425 patients who underwent lobectomy (Current Procedural Terminology code 32480 or 32663), were admitted from home, and postoperatively discharged to home or another facility.
Main Outcome Measures: Odds ratio for PODF and C-statistic for models of PODF.
Results: Of the 16,425 patients included in the study, 748 (4.6%) patients experienced PODF.
Older age, hypertension, dyspnea, and history of chronic obstructive pulmonary disease were preoperative characteristics significantly associated with discharge to a non-home location in Model A (p<0.05 for all, see Figure). Each one-day increase in length of stay was estimated to increase the odds ratio for PODF 1.12 times (p<0.001). The only protective factor identified was minimally invasive approach (p<0.001). Model A had the best predictive power of the models tested, with C-statistic of 0.786. In Models B and C, a one-point increase in mFI-5 was significantly associated with greater odds of PODF (p<0.001 for both models). Model B, which assessed mFI-5 alone, had a C-statistic of 0.631, indicating fair predictive power. The model significantly improved with the addition of surgical approach as an independent variable (C-statistic 0.674, likelihood ratio p<0.001).
Conclusions: These findings suggest that decreasing a patient"ôs risk for PODF may involve optimizing the comorbidities identified and using a minimally invasive approach when reasonable, especially among older patients. While mFI-5 alone may fall short of predicting PODF, consideration of surgical approach and age can help surgeons practice appropriate expectation management regarding a patient"ôs probability of returning home following lobectomy.
Factors significantly associated with PODF among patients undergoing lobectomy in logistic regression models.
Back to 2023 Abstracts