Justin Kindy, FSA, MAAA; David Roer, MD; Robert Wanovich, PharmD; and Stephen McMurray, MD
Patients with end-stage renal disease (ESRD) require extensive, complex, and costly medical care. Life-sustaining dialysis treatments, often administered as thrice-weekly hemodialysis, rid the body of the toxins and fluids normally removed from the blood by the kidneys. In 2013, although patients with ESRD represented less than 1% of the US Medicare population, they accounted for approximately 7% of Medicare fee-for-service expenditures, with an annual average cost of $84,550 per patient.1
On average, patients with ESRD were hospitalized 1.69 times per year in 2013, and approximately one-third of those hospitalized were readmitted within 30 days of discharge,1
making hospitalizations a major cost driver in this population.
Intensive patient management by dialysis providers and their partner nephrologists has been shown to improve outcomes among patients with ESRD2
and may simultaneously reduce costs. For example, the method used to access the patient’s circulatory system for dialysis has a strong association with outcomes: arteriovenous grafts (AVGs) or arteriovenous fistulae (AVFs) have a lower rate of infections and other complications compared with central venous catheters (CVCs). Thus, patient management at the time of dialysis initiation3
and initiatives to increase the use of AVFs and minimize use of CVCs4
have been associated with improved outcomes and lower costs. Patient management programs can also enhance adherence to complex medication regimens.5
Mechanisms for delivering coordinated care to patients with ESRD include federally funded experimental models such as the CMS ESRD Management Demonstration Project,6
special needs plans,7
and, most recently, ESRD Seamless Care Organizations. In the private sector, payer–provider partnerships have been proposed as a means to provide high-quality and cost-effective care for patients with chronic illnesses, including ESRD.8
Recently, a payer and a provider (a subsidiary of a large dialysis organization) initiated a contractual partnership with the goal of improving care and reducing costs for patients with ESRD who were receiving their dialysis treatments at the provider’s facilities in Pennsylvania. The payer's claims data combined with the provider’s analytic capabilities enabled the use of risk stratification and predictive models to target patients for specific interventions. These included supplemental care pathways and clinical protocols targeting specific objectives, such as transitioning patients from CVCs to AVFs, reducing hospitalizations, and decreasing the frequency of readmissions. Field-based renal nurse care managers supported fluid management, immunizations, nutritional supplementation, and end-of-life counseling. Both Medicare Advantage (MA) and commercial insurance members were eligible for the program, which used a shared savings model to create economic incentives for the partners. Incentive payments were contingent on the provider meeting or exceeding United States Renal Data System quality benchmarks for rates of AVF and CVC use, influenza and pneumococcal pneumonia vaccination, and hospital readmissions. Savings were calculated by comparing the actual total cost of care with a predetermined, actuarially derived expected total cost of care. Incentive payments constituted a contractually stipulated portion of the calculated savings. Here, we report an observational analysis of outcomes and costs of care during the first 2 years of the partnership.
We conducted an observational analysis of outcomes and costs among a payer’s plan members receiving dialysis through a specific provider. The payer’s MA and commercial insurance members were analyzed separately. Two consecutive 12-month periods (April 1, 2013-March 31, 2014 and April 1, 2014-March 31, 2015) following initiation of the partnership were compared with a 12-month baseline period (July 1, 2011-June 20, 2012). We excluded patients who 1) were in the top 1% in terms of total annual aggregated nondialysis costs, 2) were missing dialysis/inpatient/skilled nursing facility (SNF) claims in at least 50% of the identified eligible member months, 3) had dialysis/inpatient/SNF claims of less than $1500 (commercial) or $1000 (MA) per member per month (PMPM) in at least 50% of eligible member months, or 4) received a kidney transplant (exclusion began in the month that the transplant was received).
This retrospective, observational, noninterventional analysis of outcomes in patients in the partnership was conducted using existing deidentified patient data; therefore, according to the HHS regulations, 45 CFR Part 46, this study was exempt from institutional review board or ethics committee approval. We adhered to the Declaration of Helsinki and informed consent was not required.
Demographic information and clinical data in each study period were abstracted from the electronic health records of the dialysis provider. Comorbidities were determined from patient utilization within the payer claims database using International Classification of Diseases, Ninth Revision
diagnosis codes. Payer claims were used to determine healthcare costs.
Outcomes and Analysis
We considered several outcomes that were based on benchmarks set by the National Kidney Foundation, including vascular access type, vaccination rates, and hospital readmission rates. Vascular access type among hemodialysis patients was defined as AVF, AVG, or CVC based on the access type used for the majority of treatments in each month and was expressed as the proportion of patients using each access type in each study period. Rates of pneumococcal vaccination were considered as the proportion of patients who had received the vaccination in the past 5 years or had received 2 doses ever. Influenza vaccination rates were determined as the annual period prevalence of patients (aggregated proportion) who received vaccinations between September 1 and March 31 within each study year. Hospital readmission rate was defined as the proportion of patients readmitted within 30 days of discharge following an inpatient stay.
We further considered healthcare costs, hospitalization and emergency department (ED) visit rates, length of stay, and hospitalized days. PMPM costs were considered, excluding dialysis treatment costs, and a 90-day claims run-out period was used to allow consistent data capture. Nondialysis medical costs included 4 categories: 1) inpatient hospital care costs, excluding inpatient care following kidney transplant; 2) professional medical care costs, defined as costs attributed to physician visits; 3) prescription drug costs; and 4) all other medical care costs, including but not limited to those of laboratory tests, SNFs, EDs, and home health care. Hospitalization and ED visit rates were calculated as the number of events per 1000 patient-years; length of hospital stay (days) and costs per admission were considered as the mean values in each study period. The total number of hospitalized days in the period was also determined.
All outcomes were considered separately for commercial and MA members; no statistical comparisons were made.
Payer beneficiaries receiving dialysis at the provider’s facilities in Pennsylvania were eligible for enrollment in the program; approximately 80% to 85% of eligible patients elected to enroll. Patient characteristics by type and year are summarized in the eAppendix Table
(eAppendix available at ajmc.com
). Year to year, member characteristics were similar within each plan type. Overall, compared with commercial plan members, MA members tended to be older, have a greater burden of comorbidity, and have been on dialysis longer.
Analysis of plan members with respect to the vascular access and vaccination outcomes that were part of the program performance metric is shown in Figure 1
. Among both commercial and MA members, AVF and AVG utilization was greater in year 2 than at baseline, whereas CVC utilization was lower. Pneumococcal and influenza vaccination rates were consistently higher than 95% among both commercial and MA members across all study periods.
Analyses of hospitalizations and ED visits are shown in Figure 2A and 2B
. Among both member cohorts, the hospital admission rate declined over the study period. The length of stay for inpatient admissions also declined, from 6.4 days at baseline to 5.3 days in year 2 for commercial members and from 6.9 to 6.3 days, respectively, for MA members. The reductions in the hospital admission rate and length of stay resulted in a marked decline in the number of hospitalized days for plan members over the study period.
Hospital readmission rates were moderately lower in year 2 than at baseline. Among commercial plan members, the rate fell from 22% at baseline to 18% and 19% in years 1 and 2, respectively. The readmission rate was 29% at baseline among MA members, declining to 24% and 25% in years 1 and 2, respectively. The ED visit rate among commercial members declined from 1377 to 864 visits per 1000 patient-years from baseline to year 2. For MA members, it declined from 2178 to 1305 visits per 1000 patient-years over the same period.
Costs of medical care for program members also declined over the 2 years of the program (Figure 2C
). In aggregate, more than $5 million was saved relative to the year 1 and year 2 actuarially determined cost expectations, with $3.04 million derived from the commercial members and $2.01 million from the MA members. Across the primary cost categories and considering both years, commercial members experienced a 48% decline in inpatient care costs, a 38% decline in professional costs, and a 34% decline in other medical expenses. For the MA members, these costs decreased by 19%, 9%, and 12%, respectively. The only exception to this trend toward decreasing costs was prescription drug costs, which increased from baseline through year 2 in both member cohorts (by 32% for commercial members and 30% for MA members).
Management of patients with ESRD is difficult, requiring frequent and routine access to patients, partnership with nephrologists, specialized care protocols, and customized technology. In an effort to improve outcomes and lower costs, a dialysis provider and a payer created a shared savings incentive program. Members enrolled in the program were provided targeted services aimed at improving specific quality metrics and controlling nondialysis healthcare expenditures.
As shown here, this payer–provider partnership resulted in improved clinical care quality metrics and reduced medical costs relative to the baseline population. Improvements were observed in vascular access utilization, with an increase in AVF and AVG utilization and a decrease in CVC utilization among both commercial and MA members. Vaccination rates for influenza and pneumococcus were above 95% for both member cohorts across the study period. These vaccination rates are substantially higher than those most recently reported for the Medicare ESRD population generally (30% for pneumococcus1
and 75% for influenza9
Hospitalizations are a major cost driver among the ESRD population. Notably, the hospital admission rate, number of hospitalized days, ED visit rate, and 30-day readmission rate all decreased following the inception of the payer–provider partnership, and these decreases were accompanied by decreases in inpatient costs and healthcare costs overall. The hospital admission rate for MA program members in year 2 (1.3 admissions/patient-year) compares favorably with the rate for the Medicare hemodialysis population overall, which averaged 1.69 admissions per patient-year in 2013.1
The hospital readmission rate for MA program members in year 2 (25%) was markedly lower than the Medicare average (36.9% in 2013).1
Our findings are consistent with prior evidence demonstrating that care coordination can improve outcomes and may help contain costs for patients with ESRD.7,10
Although overall costs and inpatient costs declined among plan members, our analyses documented incremental increases in overall medication costs during years 1 and 2 compared with the baseline period. This result may be related to improved patient adherence to prescribed medications arising from enhanced clinical management. Alternatively, this observation may have been driven by a small number of patients receiving treatment with extremely expensive medications; further analysis will be required to fully understand this result.
Several factors may limit the generalizability of our findings to other ESRD populations. First and foremost, this was a retrospective observational analysis that was not designed to address cause-and-effect relationships. The patient population studied was defined by the patients’ choice of payer and provider and their willingness to enroll in the program. These factors, along with annual changes in membership and patient attrition, must be taken into consideration when interpreting our findings. Additionally, geography has been shown to influence healthcare spending,10
as well as dialysis access and modality1
; our study was limited to patients who received treatment in Pennsylvania.
The promising trends observed among members participating in this payer–provider ESRD population health partnership suggest that collaborations with shared incentives may be a valuable approach for improving ESRD patient outcomes and reducing care costs.
The authors acknowledge the medical writing assistance of Dena E. Cohen, PhD. The authors thank Carly Busch, Sarah Falkof, and John Plonka for assistance with data acquisition.