Background: Unplanned readmissions among cancer patients are common and gastrointestinal (GI) cancer has some of the highest readmission volumes. Under the Affordable Care Act, hospitals have been getting penalized for excessive readmission rates. For now, the medical treatment of cancer is exempt from this measure. This is because the readmission profile of the cancer patient is still poorly understood. As with other disease states, not all cancer readmissions are preventable and there is no standardized and accurate way to ascertain this. Many believe that a readmission is a necessary encounter to address a patient’s worsened medical condition. For a cancer patient, association with more important outcomes such as survival has not been thoroughly studied.

Methods: We conducted a retrospective cohort study using Texas Cancer Registry linked with Medicare claims. We included patients diagnosed with gastrointestinal cancer between 2001 and 2010 and reviewed hospitalizations to short-stay acute care hospitals to determine the 30-day unplanned readmission rate. Patient claims over 2 years from the date of diagnosis were studied. Cox proportional hazard modeling adjusted by age, stage of disease and comorbidity index was used to compare survival of patients with zero unplanned readmissions to those with >=1 unplanned readmission. 

Results: A total of 31,736 patients contributing 73,391 hospitalizations were included. We found an unplanned readmission rate of 17.8% in our cohort. 27% of patients had at least one unplanned readmission during the observation period. Patients with zero unplanned readmission had longer length of stay than patients with >=1 unplanned readmission (6 days vs. 5 days respectively; p<.001). Across all types of GI cancer studied, having at least 1 unplanned readmission was significantly associated with poorer survival compared to patients who have zero unplanned readmission (p<.001). The covariates adjusted for, specifically age >80 (vs. 66-69years), comorbidity score >=2 (vs. zero comorbidity) and more advanced or unknown stage (vs. localized stage), were also associated with worse survival outcomes (all p-values <.05).

Conclusions: Studies correlating readmissions to survival in other diseases have shown conflicting data. Our findings show that 2-year survival is worse for patients with at least 1 unplanned readmission. It should be noted that patients who have zero unplanned readmission may have hospitalizations that are planned. This should be interpreted carefully. As much as this supports the notion that unplanned readmissions are associated with poor outcomes and should be prevented, it also implies that cancer patients who are readmitted are likely sicker and more frail than those not readmitted. It is possible that these readmissions are not preventable, and may be a marker of poor survival. Older age, high comorbidity score and advance cancer stage which have been shown in other studies to be predictors of readmission, were also shown in our study to be associated with worse survival outcomes. Policymakers should consider this if readmission measures are to be applied to the cancer population. Future studies should strive to provide a standardized definition for a preventable cancer readmission.