Background: Oncology comanagement of hospitalized cancer patients is increasingly common. We (1) and others (2-4) have previously demonstrated that when compared with traditional oncologist-only staffing, hospitalist co-management can improve on well-recognized quality outcomes including length of stay, early discharge rates and 30-day readmission rates. Yet, the composition of an oncology hospitalist practice is incompletely understood and further opportunities for practice improvement are poorly defined.
Methods: All Smilow Cancer Hospital solid tumor admissions from 07/01/2021-03/14/2022 with complete data across all covariates (n=962 hospitalizations distributed across 691 unique patients) were included. Independent variables included age, sex, race or ethnicity, cancer type, insurance coverage, severity of illness index (SOI) and discharge disposition. Outcomes included length of stay (LOS), timing of discharge and 30-day readmission rate. All variables were extracted from the medical record. Mixed linear (LOS, timing of discharge) or Poisson (30-day readmission rate) models with a random intercept variable for patient were conducted. Multivariable models incorporated all covariates.
Results: In multivariable models, LOS was independently associated with cancer type, severity of illness, and discharge disposition. LOS in patients with head and neck cancers (13.64 days (95% CI: 11.44, 15.85)) exceeded all other cancers by two-fold (p< 0.0001). LOS was significantly higher in patients with higher (8.76 days (95% CI: 7.52, 10.01) vs lower SOI (6.17 days (4.78, 7.55); p< 0.0001) and patients being discharged to skilled nursing facilities (SNF) (11.25 days (95% CI: 9.45, 13.06) as compared with patients being discharged home with self-care (4.24 days (95% CI: 2.97, 5.50); p< 0.0001).Significant delays in timing of discharge were observed among patients being discharged to SNF (4:17 pm (95% CI: 3:37 pm, 4:58 pm) or inpatient hospice (4:10 pm (95% CI: 3:31 pm, 4:49 pm) compared with patients being discharged home with self-care, (3:07 pm (95% CI: 2:39 pm, 3:36 pm), p=0.0002). Patients with higher SOI (4:03 pm (95% CI: 3:36 pm, 4:31 pm)) were discharged later than lower SOI patients (3:38 pm (95% CI: 3:07 pm, 4:09 pm) p=0.02), independent of discharge disposition. While the overall 30-day readmission rate was 27.86% (95% CI: 25.03%, 30.69%), patients discharged to home (1.87% (95% CI: 0.26%, 13.73%) or inpatient (1.04% (95% CI: 0.14%, 7.57%) hospice were unlikely to return to the hospital.
Conclusions: Multiple opportunities for improving care were identified. Shifting discharge to home rather than skilled nursing facility has the potential to reduce LOS, improve time of discharge and reduce readmissions. Matching discharge disposition to patients’ care goals and increasing hospice discharge will further reduce readmissions. Better understanding the root cause for extended LOS among head and neck cancer patients may help optimize inpatient care and reduce LOS in this patient group.