High Healthcare Utilization and Costs in Patients with Higher-Risk MDS/Low Blast Count AML Treated with Azacitidine in Ontario, Canada

Background: Azacitidine (AZA) use in higher-risk MDS has been adopted because it improves survival. Despite this, "real-world" data on the economic impact and resource utilization remains unknown. We used the Ontario provincial AZA MDS registry, which captures all AZA-treated patients in the province, to analyze "real-world" data on healthcare use, associated costs and their predictors in AZA treated higher-risk patients.

Methods: We linked the provincial MDS AZA registry (single-payer/universal access), which captures baseline characteristics and treatment response for all AZA-treated patients in Ontario, to population-based health system administrative databases. Only higher-risk MDS patients (IPSS intermediate-2, high) and low blast count AML (21-30% blasts) treated from May 30, 2010 to March 16, 2015 were included. Patients were followed for 24 months following first AZA treatment and censored at the earliest of 90 days after last AZA treatment, date of death, time of acute leukemia induction/allogeneic stem cell transplant or March 31, 2016. We estimated healthcare resource utilization and the mean (and overall) standardized 28-day healthcare cost in Canadian dollars ($1 CDN = 0.76 USD$). Quantile regression was used to explore predictors of cost. Negative binomial regression models were used to explore predictors for higher rate of emergency department (ED) visits, and for longer length of stay, with the natural logarithm of length of follow-up as an offset variable in each model.

Results: The registry had 652 higher-risk MDS and 225 low blast count AML patients (n = 877) with median follow up of 8 months (IQR 4-13). Median age was 73 years (IQR 66-79), 66.0% were male, 17.8% were secondary MDS and IPSS scores of those calculable were intermediate-2 (64.9%) and high-risk (35.1%). At the time of AZA initiation, 587 patients (66.9%) were transfusion dependent. The median number of cycles received was 6 (range 3 to 11) and median overall survival was 16.1 months (95% CI 13.9 to 18.3). Overall, 705 patients (80.4%) had at least 1 ED visit and 290 (33.1%) had an ED visit during their first cycle of AZA. In addition, 680 patients (77.5%) had at least 1 hospital admission with a mean hospital stay of 17.7 days (95% CI 16.3 to 19.1) over the entire study period. 141 patients (16.1%) required admission to an intensive care unit. Older age (Rate ratio [RR] = 1.33, 95% CI 1.09-1.62), rurality (RR=1.75, 95% CI 1.42-2.15), high IPSS score (RR=1.31, 95% CI 1.06-1.62), and increased comorbidity level were each independent predictors of increased ED visits; while higher comorbidity level (RR=1.51, 95% CI 1.08-2.11), high IPSS score (RR=1.39, 95% CI 1.01-1.92), and transfusion dependence (RR=1.51, 95% CI 1.13-2.01) were associated with longer hospital stays.

The overall mean cost was $146,675 per patient (95% CI $139,537 to $153,812) including AZA and $103,580 (95% CI 98,675 to 108,486) excluding AZA drug costs. The mean standardized cost per 28-day period per patient was $17,638 (95% CI $16, 870 to $18,407) with AZA and $13,450 (95% CI $12,730 to $14,170) without AZA drug costs. Inpatient admissions ($4,631, 95% CI $4,010 to $5,251) and non-physician outpatient cancer clinic costs ($6,092, 95% CI $5,851 to $6,333) were the major cost drivers. Excluding AZA costs, the mean standardized 28-day costs were higher in those receiving less than 4 cycles of AZA (n= 295) at $19,408 (95% CI $17,568 to $21,248), compared with those receiving 4 or more cycles (n= 582) at $10,430 (95% CI $10,069 to $10,790) with inpatient admissions as the major driver (mean $10,192, 95% CI $8,594 to $ 10,192 vs. $1,812, 95% CI $1,558 to $2,065). On multivariable analysis, only greater comorbid disease burden (β = $2,074, 95% CI $665 to $3,483) and transfusion dependence (β = $2,402, 95% CI $1,190 to $3,613) were associated with higher median standardized 28-day cost.

Conclusions: In our analysis of "real-world" patients with uniformly higher-risk MDS treated with AZA we demonstrate a significant economic impact above and beyond the cost of AZA alone. The costs are higher in patients who are transfusion dependent and have greater comorbidity and appear to be driven by inpatient care and outpatient non-physician ambulatory care. This group of patients are high users of healthcare resources with the majority having ED visits and inpatient admissions. These results will inform patients and providers about the "real-world" anticipated toxicities of AZA.