- © 2000 by American Society of Clinical Oncology
Predicting Costs of Stem-Cell Transplantation
- From the Departments of Adult Oncology and Biostatistical ScienceDana-Farber Cancer Institute, and Department of Medicine, Brigham and Women’s Hospital, Harvard Medical School, Boston, MA.
- Address reprint requests to Stephanie J. Lee, MD, MPH, Center for Outcomes and Policy Research, Dana-Farber Cancer Institute, 44 Binney St, Boston, MA 02115; email stephanie_lee{at}dfci.harvard.edu
Abstract
PURPOSE: Few studies have formally evaluated the relationship between costs, baseline patient characteristics, and major complications of stem-cell transplantation. We sought (1) to determine whether obtaining baseline information enabled identification of patients whose treatments would be the most costly and (2) to estimate inpatient costs for managing specific transplantation complications.
PATIENTS AND METHODS: We collected inpatient costs and clinical information for 236 consecutive patients undergoing transplantation at a single institution between July 1, 1994, and February 20, 1997. Multivariable linear regression was used to evaluate the associations between baseline patient characteristics and costs of hospitalization for initial transplantation and between clinical events and such costs.
RESULTS: The median initial inpatient cost in 1997 dollars was $55,500 for autologous transplantation (range, $28,200 to $148,200) and $105,300 for allogeneic transplantation (range, $32,500 to $338,000). When only baseline variables were considered, use of a mismatched allogeneic donor and year of transplantation were significant predictors of costs. No characteristics predicted which patients would incur the highest 10% of costs. When clinical events were considered, infection and in-hospital death were associated with higher costs in autologous transplant recipients ($18,400 and $20,500, respectively), whereas infection, veno-occlusive disease, acute graft-versus-host disease, and death were predicted to add between $15,300 and $28,100 each to allogeneic transplantation costs.
CONCLUSION: We were not able to identify before transplantation the patients whose treatments would be the most costly. However, the association between clinical complications and higher costs suggests that prevention may have significant economic benefits. Interventions that decrease these complications may have favorable cost-benefit ratios even if they do not affect overall survival.
LOWERING COSTS while preserving quality of care and clinical outcomes is an increasingly important goal for transplantation programs. Several studies have documented the success of cost-cutting efforts, such as substitution of less costly technology when there is no difference in clinical results1 or decreasing length of stay by accelerating engraftment2-7 or moving care to the outpatient setting.8 However, even among patients treated similarly at a single institution, there is great variability in hospitalization costs. The patient characteristics and clinical events associated with this variability are being explored in an increasing number of studies.9,10 From a programmatic perspective, it may be useful in a capitated system to identify before transplantation the patients whose treatments will be the most costly, so that higher reimbursement rates can be negotiated. From a research perspective, cost estimates for transplantation complications would allow projection of the financial benefits of preventive strategies.
To identify the major patient characteristics and transplantation complications predictive of hospitalization costs, we analyzed clinical and financial transplantation data from a single institution (Brigham and Women’s Hospital, Boston, MA) for patients undergoing transplantation between July 1, 1994, and February 20, 1997. We considered type of transplant, demographics, and disease characteristics, as well as major clinical events such as acute graft-versus-host disease (GVHD), infections, veno-occlusive disease (VOD), diffuse alveolar hemorrhage (DAH), and death. We sought to answer two questions: First, can baseline patient characteristics, known before admission for transplantation, predict which patients will incur the highest costs? This information could allow risk adjustment or assist transplantation programs in addressing fiscal concerns. Second, what are the costs of specific transplantation complications when the occurrence of other events is controlled for? These estimates would allow predictions of the economic impact of these complications.
PATIENTS AND METHODS
Patients
Two hundred thirty-six consecutive patients undergoing stem-cell transplantation for hematologic disorders at Brigham and Women’s Hospital in Boston, MA, between July 1, 1994, and February 20, 1997, were included in this study. There were 53 autologous transplant recipients, two syngeneic transplant recipients, 103 related-donor transplant recipients, and 78 unrelated-donor transplant recipients. Of the related donors, 96 were siblings, three were offspring, and four were parents. One patient received an allogeneic transplant following relapse after autologous transplantation; only the allogeneic transplantation occurred during the period of analysis and was included.
Conditioning Regimens, GVHD Prophylaxis, and Routine Care
A variety of conditioning regimens were used during the study period, chosen on the basis of transplant type and research protocols. One autologous transplant recipient underwent total-body irradiation (TBI) as part of his preparative regimen. The remainder of the autologous transplant recipients received chemotherapy-only regimens, primarily cyclophosphamide, etoposide, and carmustine.11 The majority of allogeneic transplant recipients received cyclophosphamide (1,800 mg/m2 for 2 days) and underwent TBI (14 Gy in eight divided doses) with or without administration of cytarabine.12 Patients with aplastic anemia (AA) received cyclophosphamide and antithymocyte globulin. Allogeneic transplant recipients unable to receive radiation underwent chemotherapy-only conditioning with busulfan (16 mg/kg divided over 4 days) and cyclophosphamide (750 mg/m2 bid for 4 days). GVHD prophylaxis consisted of short-course methotrexate and cyclosporine or tacrolimus (FK-506) therapy with or without corticosteroids.
The two syngeneic transplant recipients were placed in the autologous transplant group for purposes of analysis. Syngeneic transplant recipients were conditioned with cyclophosphamide therapy and TBI but did not receive any GVHD prophylaxis.
Patients received trimethoprim-sulfamethoxazole or equivalent pneumocystis pneumonia prophylaxis during the week before stem-cell infusion. The day of infusion was designated day 0. Gut decontamination was achieved with oral antibiotics in allogeneic transplant recipients. All patients received prophylactic acyclovir from day 5 until day 30 if they were herpes simplex virus (HSV)–seropositive, whereas allogeneic transplant recipients received prophylactic ganciclovir on day 30 if their absolute neutrophil count was greater than 1,000/μL and either patient or donor was cytomegalovirus (CMV)-seropositive. No other bacterial, viral, or fungal prophylaxis was given. Patients were treated with broad spectrum antibiotics at the time of their first neutropenic fever and with antifungal agents as appropriate.
Patients were housed in either HEPA-filtered or laminar airflow rooms in an intensive care setting; patients requiring aggressive, invasive monitoring and interventions did not require transfer to separate units. Discharge criteria, which did not change over the course of the study, included engraftment, adequate oral intake, and control of all medical problems. No patients underwent outpatient transplantation.
Predictors Known Before Transplantation
Patients were classified as having non–Hodgkin’s lymphoma, Hodgkin’s disease, multiple myeloma, acute myelogenous leukemia, acute lymphoblastic leukemia, chronic myelogenous leukemia, chronic lymphocytic leukemia, myelodysplastic syndrome or myeloproliferative syndrome, or AA or paroxysmal nocturnal hemoglobinuria. Patients were also divided according to level of risk before transplantation. Good-risk patients either had AA, paroxysmal nocturnal hemoglobinuria, myeloproliferative syndrome, refractory anemia, or chronic myelogenous leukemia in chronic phase or were in first complete remission from other diseases. Poor-risk patients (1) were partially responsive, had refractory disease, or were in relapse at the time of transplantation or (2) had advanced myelodysplastic syndrome or secondary acute myelogenous leukemia. Intermediate-risk patients were patients who were having at least second complete responses or who were in other disease states associated with intermediate risk.
Patients were also classified by whether they received peripheral-blood progenitor cells alone or in combination with marrow and by the year in which bone marrow transplantation occurred. HSV and CMV serologic status and ferritin levels were documented before admission. Patients who were CMV-seronegative and whose donors were also CMV-negative were distinguished from all other patients. Weight and serum creatinine and total bilirubin levels were recorded on admission, and creatinine clearance was calculated. Data on Karnofsky performance status and CD34+-cell doses were not obtained prospectively in most cases and thus these variables could not be evaluated in the model.
HLA typing was performed serologically for class I antigens and either serologically or molecularly for class II antigens. Patients and donors were classified as matched or mismatched on the basis of HLA-A, -B, and -DR typing.
Intermediate Predictors: Clinical Events During Transplantation
Clinical events were collected by chart review by one of the authors (S.J.L.) without knowledge of costs. Acute GVHD was graded by physicians using standard methods,13 and patients were divided into those with grade 0 to 1 disease and those with grade 2 to 4 disease. The diagnosis of VOD was made by the attending physician. Patients with documented bacterial, fungal, or significant viral infections were compared with those without fevers or with fevers but with cultures negative for organisms. Patients with either documented DAH14,15 or suspected DAH that warranted therapy were compared with those without DAH. Length of stay was the time from the date of admission for transplantation to the date of discharge.
Costs
Total charges for the transplantation hospitalization were obtained from the accounting department, broken down by hospital department. These charges were converted to costs using the departments’ ratios of costs to charges, and the costs were adjusted to 1997 dollars using the medical care component of the consumer price index.16 High-cost outliers were defined as costs in the highest decile. No attempt was made to exclude or adjust costs for patients receiving aggressive care in their final days, although the degree of support may have varied depending on patient and family wishes.
All costs incurred before admission for transplantation were excluded—including costs of any mobilization chemotherapy, administration of growth factors, stem-cell procurement, or central venous access placement—for the following reasons: (1) in the case of autologous transplant recipients, some of these procedures were performed outside our hospital system, and (2) donor type and, when applicable, stem-cell source were included in the model, adjusting for any baseline differences in costs of peripheral-blood collection and marrow collection.
No costs were included after initial hospital discharge, because initial inpatient costs are the overwhelming cost driver in the first 100 days and because we wished to provide a conservative estimate of costs of complications. Thus only the period from hospital admission for conditioning to hospital discharge was considered.
Data Analysis
Separate analyses were performed for autologous and allogeneic transplant recipients because factors contributing to costs were expected to differ between these populations. Logistic regression was used to identify characteristics associated with incurring of costs in the highest decile. Multiple linear regression was used to investigate the association between clinical events and total costs.17 All tests were two-tailed and a P value of .05 was considered significant. Analyses were limited to patients having complete covariate information. In the search for predictors of costs, only patient and transplantation factors that would be known before hospital admission were considered in baseline models. To estimate the costs of complications, we included clinical events in the full models.
Baseline data for autologous transplant recipients included age, sex, disease type and stage, HSV serologic status, weight, creatinine clearance, total bilirubin level, year of transplantation, and use of peripheral-blood stem cells (PBSCs). Baseline data for allogeneic transplant recipients included patient and donor ages and sexes, patient and donor CMV status, degree of match, type of donor, disease type and stage, use of TBI, HSV status, weight, creatinine clearance, total bilirubin level, ferritin level, and year of transplantation. The full models also considered occurrence of VOD, DAH, GVHD, infection, and death. Administration of high-dose steroids, mechanical ventilation, and other interventions used to treat complications were not included in the explanatory variables. Length of hospital stay was highly correlated with total overall costs (partial r2 = .76 to .77) and was not included in the models, because it was a proxy for costs.
RESULTS
Baseline Predictors: Patient and Transplantation Characteristics
Patient demographics are listed in Table 1. Among allogeneic transplant recipients, several factors differed between related- and unrelated-donor transplant recipients. Related-donor transplant recipients were more likely to be older (P = .006) and fully matched (92% v 81%, P = .022). Only six allogeneic transplant recipients (3%) received PBSCs, and all of these transplantations occurred in patients with related donors. Unrelated-donor transplant recipients were more likely to undergo TBI (P = .010). There was no difference between related and unrelated recipients in the percentage of CMV-seronegative donor/patient pairs.
Baseline Patient and Transplantation Characteristics
Although the proportion of transplants from mismatched donors was roughly constant over the period of this study (P = .735), more poor-risk patients received transplants in the later years (P = .013). No association between poor risk and the likelihood of receiving a transplant from a mismatched donor was detected (P = .244).
Intermediate Predictors: Clinical Events
The clinical events that occurred during transplantation are summarized in Table 2. Incidences of DAH, VOD, and infection; in-hospital mortality; length of stay; and overall costs were significantly higher for allogeneic transplant recipients than for autologous transplant recipients. Between related- and unrelated-donor allogeneic transplant recipients, there were no differences in these outcomes, and populations were pooled. Six autologous transplant recipients (11%) and 40 allogeneic transplant recipients (22%) required mechanical ventilation, which was significantly associated with mortality; all six autologous transplant recipients and 37 allogeneic transplant recipients (93%) died after intubation. Because mechanical ventilation was considered a therapeutic intervention, it was not included in the regression analysis. Six (11%) of the 55 autologous transplant recipients and 53 (29%) of the 181 allogeneic transplant recipients died during their transplantation hospitalization. The major cause of death was regimen-related toxicity, although two autologous and 10 allogeneic transplant recipients died from infection and five allogeneic transplant recipients died from GVHD. Length of stay decreased among autologous transplant recipients during the period of the study (P = .048) but did not change significantly among allogeneic transplant recipients (P = .65).
Outcomes
Predicting Costs Using Pretransplantation Information Only
Median costs in 1997 dollars for initial hospitalizations for autologous and allogeneic transplantations were $55,500 (range, $28,200 to $148,200) and $105,300 (range, $32,500 to $338,000), respectively. Excluded were costs of donor identification, collection, and pretransplantation testing. Total costs for autologous transplant recipients decreased during the study period (P = .0323), whereas allogeneic transplantation costs increased (P = .0190). There was no correlation between year of transplantation and number of complications or length of stay for either autologous or allogeneic transplant recipients (data not shown). The only other statistically significant pretransplantation predictor of costs was the use of mismatched donors, among allogeneic transplant recipients ($25,300, P = .0365). There was no correlation between costs and disease type or risk status, use of PBSCs, or patient age or sex. Weight, creatinine clearance, and total bilirubin level on admission were also not correlated with costs and were excluded from further analysis (data not shown).
Autologous and allogeneic transplant recipients with costs in the highest decile were identified. This group comprised allogeneic transplant recipients with costs of more than $191,672 and autologous transplant recipients with costs higher than $99,187. No baseline data, including disease risk status, age, and type of donor or match, predicted which patients fell into this group.
Costs of Complications
The full models estimating the costs of complications included both baseline data and clinical events. For autologous transplant recipients (Table 3), earlier year of transplantation (P = .0305), infection ($18,400, P = .0135), and type of disease (P = .0441) were predictive of higher costs. Death was associated with an increased cost of $20,500 (P = .0799). In these data, we did not find an association between use of PBSCs and costs. The r2 of the full model was .6372.
Predictors of Costs in Autologous and Syngeneic Transplantation*
Among allogeneic transplant recipients (Table 4), increased costs were associated with more recent year of transplantation (P = .0084), acute GVHD ($28,100, P = .0005), VOD ($18,600, P = .0137), infection ($15,300, P = .0403), and death ($20,200, P = .0195). We did not find an association between costs and disease risk status, donor type, mismatch, use of TBI, or stem-cell dose. Disease type was of borderline significance, with a P value of .0793. The r2 of the full model was .4362.
Predictors of Costs in Allogeneic Transplantation*
Listed in Table 5 are actual median costs incurred by allogeneic transplant recipients according to the number of these complications. Median costs increased in a linear fashion with increasing numbers of complications. The 54 patients with none of these complications incurred a median cost of $90,100 (range, $66,600 to $126,200), whereas the 25 patients with three complications incurred a median cost of $153,500 (range, $89,300 to $338,000). The number of complications was not correlated with the year of transplantation (P = .522).
Actual Costs Incurred by Patients Undergoing Allogeneic Transplantation, by Number of Transplantation Complications
DISCUSSION
In this large, single-institution study, no clinical information available before transplantation accurately predicted inpatient hospitalization costs or which individuals would incur costs in the highest decile. Rather, we found that higher costs were best explained by occurrence of major complications (acute GVHD, VOD, infection, or death), which increased costs by approximately $20,000 per event.
One explanation for our findings is that strict patient selection criteria or small sample size might have limited our ability to detect associations. However, we note that this study population represents 21/2 years of activity at a typical large transplantation center. If a factor was not found to be significant in this cohort, its practical relevance in predicting costs at an institutional level is questionable. The best explanation for our results is that transplantation complications drive up costs, whereas baseline patient characteristics have a limited ability to predict who develops these complications.
Although the majority of patients with acute GVHD, VOD, or infections recover, we found that the cost of managing these complications is high even in the absence of fatal complications. Our contemporary cost estimates provide a basis for assessing possible economic benefits of preventive interventions in modern practice. For example, almost half of our allogeneic transplant recipients developed VOD, with an associated cost of $18,600. Our results suggest that if an intervention costing $5,000 per patient were performed in the allogeneic population and decreased the incidence of VOD by 44% or more, it could actually be cost saving. Thus, even expensive new technologies or interventions may be offset by a decrease in costs of complications if effectively used.
Some complications may be costly in the short term but could prevent additional costs later. For example, acute GVHD has been associated with decreased relapse rates,18-21 and it is possible that additional cost savings might accrue later to patients with acute GVHD who avoid relapsing. Our data could not directly confirm this hypothesis. However, our results do suggest that the costs associated with relapses at the lower rates attributable to acute GVHD would need to be less than $28,100 per patient if the initial costs are to be offset.
Most economic studies in transplantation have focused on either overall costs, departmental costs, or cost savings associated with specific treatment protocols, without examining the patient characteristics and clinical events associated with these costs.22-30 We found only one previous study in which costs of stem-cell transplantation complications were estimated. In that study, Griffiths et al31 evaluated patients undergoing allogeneic transplantation between 1980 and 1987. More recently, other investigators have begun to evaluate the relationship between patient characteristics, treatment, complications, and costs for high-cost procedures. Schulman et al10 estimated the cost savings associated with higher versus lower doses of CD34+ cell grafts in autologous transplantation. In our study, we did not have complete CD34+ data and were unable to evaluate this variable. Showstack et al9 studied 711 patients undergoing liver transplantation. They were able to identify several patient characteristics and clinical events associated with higher costs.
Some findings are surprising in light of work by other investigators. We could not demonstrate lower costs for autologous transplant recipients receiving PBSCs as other studies have.3-6,28,32 Given that 89% of autologous transplant recipients received PBSCs, we may not have been able to evaluate this variable effectively. Characteristics such as patient age, disease type or risk status, and type of donor were not correlated with costs in the baseline model although their influence on ultimate transplantation success is well documented.33-36 These prognostic factors may have less predictive power for the early complications we evaluated. Finally, costs for allogeneic transplant recipients increased with time whereas those for autologous transplant recipients decreased. Other investigators have noted a learning curve, with length of stay and costs decreasing with greater institutional experience.22 As programs reach a plateau on the learning curve, perhaps some costs may actually increase as less ideal candidates are accepted for transplantation or more aggressive supportive interventions are used. We found that the proportion of poor-risk allogeneic transplant recipients increased over the course of this study (P = .013).
A number of caveats must be mentioned. As in any single-institution study, some conclusions are specific to our center and reflect our specific mix of patients and practice patterns. Total costs may be particularly sensitive to the local practice environment and institution and should be interpreted with caution. However, the multivariate analysis controlled for many baseline patient and disease characteristics, and generalizable estimates for managing complications may be possible. A second caveat is that costs were collected from the time of conditioning through hospital discharge and represent the short-term costs of transplantation only.37 Information from beyond the initial hospitalization was excluded, to obtain the most straightforward and conservative estimates of the costs of acute complications. Additional hospitalizations to manage complications would increase these estimates. Third, certain costs, such as pretransplantation patient testing and collection from donors, were not included; however, these costs would not vary much between patients once donor type and stem-cell source were considered and most likely would not affect our conclusions. Finally, the best outcome is not always associated with the lowest costs. For example, the allogeneic transplant recipient with the lowest inpatient cost died on day 5 of regimen-related toxicity. Our study was not designed to assess the effectiveness of transplantation procedures for different indications, and it is important to consider both economics and clinical results when considering health care policy.
In summary, our work suggests that it is not yet possible to identify which patients will incur the highest initial inpatient costs. However, our findings serve as estimates of common complications seen in stem-cell transplantation and add further incentive to prevent infections, VOD, acute GVHD, and death, because success can be expected to have significant economic as well as clinical benefits.
Acknowledgments
Supported in part by National Institutes of Health grant no. CA75267-01 and the Amy Strelzer-Manasevit Scholars Program.
ACKNOWLEDGMENT
We thank Kim Cullen for her help in gathering cost information and the staff at Brigham and Women’s Hospital for the care of these patients.
- Received February 9, 1999.
- Accepted August 19, 1999.
