Page 145 - Haematologica August 2018

End-of-life care for myeloma
unit [ICU] admission) are thus endorsed as indicators of high quality EOL care.14,15
Although little is known about EOL care for patients with myeloma, several studies have demonstrated that patients with hematologic cancers have low rates of hos- pice enrollment and high rates of aggressive medical care at the EOL compared to patients with solid malignan- cies.16-19 Rates of hospice use have increased for patients with blood cancers in the past decade; however, there have also been concomitant rises in “late” enrollment (gen- erally defined as ≤ 3 days before death) and aggressive EOL care.20-22 Growth in hospice use that is largely driven by late enrollment is less meaningful, as patients are being admitted to hospice primarily to manage their death, rather than to obtain palliative benefits.23
These trends may not apply to patients with myeloma. Myeloma shares many characteristics with other blood cancers (e.g., bone marrow failure leading to transfusion dependence), has other features that are similar to advanced solid malignancies (e.g., high prevalence of pain, incurability), and still others that are unique (high likeli- hood of renal disease and dialysis). In this context, we aimed to characterize EOL care among older patients with myeloma. We hypothesized that there would be an increase in hospice use over time; however, given the tra- ditional palliative needs of this population at the EOL and the known incurability of myeloma, we also hypothe- sized that we would not see increases in late enrollment.
Methods
Data Source
We used the National Cancer Institute’s Surveillance, Epidemiology, and End Results cancer registry linked to Medicare healthcare claims (SEER-Medicare). This database provides cancer registry data from 18 geographic areas, representing 28% of the population of the USA, linked to billing claims for Medicare ben- eficiaries.24 At the time of this analysis, the database included diag- noses through 2013 and billing claims through 2014. The Dana- Farber/Harvard Cancer Center Office for Human Research Studies deemed the study exempt from review.
Cohort Assembly
Outcomes
Hospice use: was defined as the presence of at least one hospice claim (outpatient or inpatient). We defined “late” enrollment as enrollment ≤ 3 days before death.
Aggressive EOL care: was defined as the occurrence of at least one of the following indicators: 1) chemotherapy use within 14 days of death, 2) ≥2 emergency department (ED) visits within 30 days of death, 3) ≥2 hospitalizations within 30 days of death, 4) hospital stay >14 days within 30 days of death, 5) at least one ICU admission within 30 days of death, and 6) death in a hospital.12,21
These measures are well-established indicators of potentially sub- optimal EOL care.
Covariates
In addition to sociodemographic characteristics, we examined comorbidity using the Deyo adaptation of the Charlson Comorbidity Index26 in the twelve months before death, transfu- sion-dependence (presence of two or more claims for transfusions in the last 30 days of life),20 and dialysis-dependence (presence of two or more claims for dialysis in the last 30 days of life).
Statistical Analyses
We assessed univariable associations of patient characteristics with outcomes of late hospice enrollment and experiencing at least one indicator of aggressive care using Chi-square tests. We then fit multivariable logistic regression models to characterize factors independently associated with the two aforementioned outcomes. Only covariates with P<0.05 in univariable analysis were included in the models. Trends in overall hospice use, late enrollment, and receipt of at least one indicator of aggressive EOL care were depicted visually using locally weighted scatterplot smoothing, where the day of death was the unit of analysis for the plot. We assessed significant trends over time using the Cochran- Armitage test, which tested for a monotonic change (i.e., increase or decrease) across the ordered years of death. In a separate mul- tivariable logistic regression model that included hospice use as a covariate, we examined the relationship between hospice enroll- ment and medically aggressive EOL care. Two-sided P values <0.05 were considered statistically significant. All analyses were performed using SAS version 9.4 (Cary, NC).
Results
Patient Characteristics
This study cohort included 12,686 myeloma decedents. The median age at diagnosis was 77 years. About half of the cohort was male (49.5%) and most were white (80.6%; Table 1). Of the total cohort, 7.3% were transfu- sion-dependent in the last 30 days of life, and 10.5% were dialysis-dependent. Median survival was 17.6 months (interquartile range [IQR] 5.2 to 39.1 months).
Hospice Use
Among the entire cohort, 6111 (48.2%) received hospice care. The median length of stay in hospice was 13 days (IQR 5 to 45 days). The majority (79.6%) used outpa- tient/home hospice services, while 19.4% used inpatient hospice services, and 1.0% used both. Among those who enrolled, 17.2% spent ≤ 3 days in hospice. In univariable analysis, patients who were transfusion-dependent were more likely to enroll in hospice ≤ 3 days before death com- pared to those who were not transfusion-dependent (36.5% vs. 16.0%, P<0.001; Table 2). Dialysis-dependence was also associated with late hospice enrollment (32.3% vs. 16.0%; P<0.001). These findings remained consistent in multivariable analysis: patients who were transfusion- dependent were more likely to enroll in hospice late (odds ratio [OR] 3.02, 95% confidence internal [CI] 2.39 – 3.82). Similarly, patients on dialysis were also more likely to enroll late (OR 2.22, 95% CI 1.79 – 2.76). Other factors significantly associated with enrolling late in hospice included male sex, living in urban areas, higher comorbid- ity scores, and surviving less than a year after myeloma diagnosis (Table 3). There was a significant increase in
We identified patients ≥65 years diagnosed with myeloma or plasmacytoma between 2000 and 2013, who were deceased by December 31, 2013. We excluded patients who died within 30 days of diagnosis. To ascertain complete claims history, patients had to have been continuously enrolled in Medicare Parts A and B with no health maintenance organization enrollment during the twelve months before death. We excluded patients diagnosed with myeloma at death or autopsy, and those who had end-stage renal disease or disability at diagnosis.25 Figure 1 and the Online Supplementary Methods detail the cohort assembly.
haematologica | 2018; 103(8)
1381