Abstract & Commentary
Can We Prevent the Avoidable Readmission When Time Is of the Essence?
By Deborah J. DeWaay, MD, FACP
Assistant Professor, Medical University of South Carolina, Charleston, SC
Dr. DeWaay reports no financial relationships in this field of study
SYNOPSIS: There are four predictors for a potentially avoidable readmission due to end-of-life issues: number of admissions in the past year, opiate prescription at discharge, neoplasm and Elixhauser comorbidity index.
SOURCE: Donzé J, Lipsitz S, Schnipper J. Risk Factors For Potentially Avoidable Readmissions Due To End-of-life Care Issues. J Hosp Med 9(5): 310-314. May 2014
There are multiple problems that lead to potentially unnecessary readmissions of patients at the end of their life, including but not limited to poor pain management, a lack of advance directives, and unwanted overtreatment. End of life care is expensive, and uses a disproportionate amount of resources; the 6% of Medicare beneficiaries who die each year use 30% of the resources. The authors of this study sought to identify patient risk factors that would predict when end-of-life issues would lead to a 30-day potentially avoidable readmission (PAR).
This study is a nested case-control study comparing non-readmitted controls to patients with PARs. All patient admissions from the Brigham and Women’s Hospital were collected for one year. Patients admitted for observation, those who died before discharge, those who left against medical advice or those who were transferred to another hospital were excluded from the study. The primary outcome was a 30-day PAR from any end-of-life issue as determined by a validated algorithm. A PAR was defined as a readmission secondary to a diagnosis known at the initial admission or from a treatment complication. An unavoidable readmission was defined as a planned admission for scheduled care (for example, chemotherapy), rehabilitation treatments, or for a new condition not present during the initial hospitalization. Nine senior resident physicians were trained to review a random sample of the PARs and determine if they were secondary to end-of-life issues as defined as a hospitalization from a terminal condition (life expectancy < 6 months) not appropriately addressed leading to a readmission. The authors collected information on an extensive list of risk factors including demographics, source of index admission, length of stay, number of hospital admissions, caregiver at discharge, number of medications at discharge, opioid prescription at discharge, and selected comorbidities (for example, diabetes, heart failure and neoplasm).
The authors compared the PARs secondary to end-of-life issues with the control group who had no readmissions in 30 days after their initial admission using a bivariate analysis of all collected potential risk factors. The Student t test was used for continuous variables and the Pearson χ2 test was used for categorical variables. The authors subsequently performed a multivariable logistical regression analysis on the variables found to be significant in the initial analysis.
The investigators reviewed 12,383 hospitalizations and excluded 17% for a variety of reasons. Of the remaining 10,275 hospitalizations, 22% were followed by a 30-day readmission; 8% were followed by a readmission that was deemed as potentially avoidable. A random sample of these PARs were taken (n=534). 15% (n=80) of the sample was secondary to end-of-life issues. A palliative consult was obtained on 20% (n=16) of these PARs secondary to end-of-life issues. After the multivariate analysis, four risk factors were identified as being associated increased PARs from end-of-life issues (control vs. PAR due to end of life): presence of neoplasm (33.9% v. 86.3%: p <0.001), an opiate prescription at the time of discharge (33.2% v. 73.8%: p <0.001), Elixhauser score (median score: 8 v. 23 p <0.001), and the number of admissions in the previous 12 months (1 v. 2 <0.001). The Elixhauser comorbidity index is an index that is calculated via administrative data using the 30 comorbidities in the ICD-9 coding manual into a single numeric score that summarizes disease burden and is predictive of in-house mortality. Of note, age was not a significant risk factor associated with a PAR from end-of-life issues.
There are several limitations to this study. First, the authors looked at readmissions at 3 affiliated hospitals, which account for 80% of the readmissions for patients that go to these facilities. Therefore, patients readmitted to other centers are not accounted for. Second, there are other risk factors, such as cognitive status, which may also be important but cannot be ascertained using administrative data. Finally, the generalizability to hospitals which are small, community-based or rural is questionable.
COMMENTARY
Discussing end-of-life issues with patients is a skill that is crucial for hospitalists to have. In addition, hospitalists often have these discussions when the patients did not discuss these issues with their primary care physician. However, it is common for hospitalists to sometimes avoid the issue because they are dealing with short length of stays and a patient with whom there isn’t a long-term relationship. Many patients do not need extensive end-of-life discussions with their hospitalist. They have a limited illness that requires short-term treatment and their comorbidities are controlled. However, there is a group of patients with multiple comorbidities, multiple admissions and no end-of-life planning that do need extensive discussions with their hospitalist. In my opinion, the most difficult conversations are those with patients that are dying from multisystem organ dysfunction. There isn’t one illness that will cause them to pass away, so bringing up end-of-life planning can be difficult, especially with the patient who is in denial about his or her global health. The risk factors described in this article are a good way for hospitalists to screen their patients for who really needs these discussions, challenge us to have the hard conversations, and consult palliative care. Hospitalists need to be equipped with having these discussions, however, because I suspect few hospitals have the palliative care resources to consult on every patient that has these risk factors.
References
- Van Walraven, Carl; Austin, Peter C.; Jennings, Alison; Quan, Hude; Forster, Alan J. (2009). "A Modification of the Elixhauser Comorbidity Measures into a Point System for Hospital Death Using Administrative Data". Medical Care 47 (6): 62633.
