ReCAP: Insights Into the Potential Preventability of Oncology Readmissions

QUESTION ADDRESSED: Are oncology readmissions preventable? If so, what resources and changes in practice or culture would be required to reduce readmissions? CONCLUSION: Three independent reviewers analyzed 72 hospital readmissions and found that 22 (31%) of the 72 readmissions were preventable. The most common causes of preventability were overwhelming symptoms in patients who qualified for hospice but were not participating and insufficient communication between patients and the care team about symptom burden. The most common reason for nonpreventability were high symptom burden among patients not appropriate for hospice or for whom aggressive outpatient management was inadequate despite extensive efforts (Table). Readmissions from nursing facilities—where there is little oncology supervision—accounted for 35% of the total. METHODS: Standardized criteria to define preventability/nonpreventability were developed before data collection began. The records of sequential nonsurgical readmissions were reviewed independently by two experienced oncology reviewers. When the reviewers disagreed about assignment, a third reviewer broke the tie. Seventy-two readmissions from 69 patients were analyzed. The first two reviewers agreed that 18 (25%) were preventable and that 29 (40%) were not. A third reviewer found four of the split 25 cases to be preventable, so the consensus preventability rate was 22 (31%) of 72. BIAS, CONFOUNDING FACTOR(S), DRAWBACKS: A large minority of readmissions can be viewed as a failure of some aspect of the medical care system: symptom management, communication, psychosocial support, education or expectation management. The exact ratio of preventable to nonpreventable readmissions is less important than the finding that many are preventable with better outreach to frail or vulnerable patients and more rigorous or effective goals of care discussions. The findings are consistent with the small number of other studies of readmissions, all judged retrospectively. Such efforts are inherently subjective, but we attempted to minimize bias by creating standard definitions of preventability (Table) and by using independent assessments, avoiding an open consensus process that introduces additional types of bias. REAL-LIFE IMPLICATIONS: Some hospital readmissions may be preventable, depending on the conditions and social situation of the patients. Unfortunately, there are no ideal methods for determining preventability of hospital readmissions. Analyses of coded administrative data allow for large data sets, but such methods are silent about the appropriateness or potential preventability of the readmission. Coded data necessarily overlook patient-level issues such as fear, frailty, social isolation or symptom burden, and ignore a patient’s desire for aggressive cancer care. Indeed, some readmissions in oncology are a consequence of continued aggressive therapy that is requested by patients or families and is rendered due to the “shared decision making” process. Chart review, although limiting the sample size, allows more insights into the patient-level and social factors associated with readmissions as well as gaps in the care process, but not all. It cannot determine, for example, if a decision not to opt for hospice care was primarily motivated by patient attitudes, oncologist approach or some combination. Although these data include only 30-day readmissions, the same sort of issue likely pertain to all unplanned admissions and to emergency department visits as well. Oncology programs are encouraged to study their own patterns of unplanned admissions and readmission in order to learn about care gaps. Greater outreach to at–risk patients as in a medical home might prevent many unplanned admissions. Finally, we note that most studies of oncology readmissions have focused on physician assessment of causes with less attention on the patient perspective about reasons for unplanned admission. Such a study is ongoing and will complement these findings. [Table: see text]