In-hospital complications of autologous ... - Wiley Online Library

4 downloads 4096 Views 167KB Size Report
Jan 18, 2008 - Significant complications were documented for >50% of admissions. Infectious ... lando, Florida, May 13-17, 2005. Address for .... Cancer Institute's Surveillance, Epidemiology, and ..... of technology, outcome, and cost.
1096

In-Hospital Complications of Autologous Hematopoietic Stem Cell Transplantation for Lymphoid Malignancies Clinical and Economic Outcomes From the Nationwide Inpatient Sample

Jeffrey A. Jones, MD, MPH1 Muzaffar H. Qazilbash, MD2 Ya-Chen T. Shih, PhD1 Scott B. Cantor, PhD1 Catherine D. Cooksley, DrPH1 Linda S. Elting, DrPH1

BACKGROUND. Autologous hematopoietic stem cell transplantation (auto HSCT) is standard of care therapy for multiple myeloma and Hodgkin and non-Hodgkin lymphomas in front-line and salvage settings, respectively. Complications remain common, but population-based estimates of their frequency and relative contribution to cost are not available.

METHODS. A retrospective cohort comprised of 8891 patients with multiple myeloma and lymphoma admitted to US hospitals for auto HSCT over a 2-year pe-

1 Health Services Research Section, Department of Biostatistics, The University of Texas M. D. Anderson Cancer Center, Houston, Texas.

riod (2000-2001) was extracted from the Nationwide Inpatient Sample (NIS).

2 Department of Blood and Marrow Transplantation, The University of Texas M. D. Anderson Cancer Center, Houston, Texas.

ICD-9-CM codes. Mean hospital charges were examined by outcome and trans-

Patient characteristics, vital status, and total hospital charges were obtained directly from the NIS. Transplant characteristics and outcomes were identified by formed into cost by using Medicare cost-to-charge ratios. Factors associated with hospital cost, length of stay, and in-hospital mortality were explored by using multivariate regression.

RESULTS. The mean hospital cost for auto HSCT during this period was $51,312. Significant complications were documented for >50% of admissions. Infectious complications ( 60%) and stomatitis ( 40%) were the most frequent, and both were associated with increased hospital costs (range, $15,000 to $50,000). In-hospital mortality was rare (75% of the autologous procedures reported to the American Bone Marrow Transplantation Registry in 2003.8 Technical advances in both conditioning regimens and supportive care have significantly decreased treatment-related mortality after auto HSCT.8 Economic advances have been slower to follow. Despite evidence for modest cost declines over time, autologous HSCT remains a resource-intensive procedure.9 Recent estimates of the direct medical cost for autologous HSCT for the lymphomas range from $20,000 to $60,000.10–13 Reported costs of autologous HSCT for multiple myeloma generally fall within this same range.14–17 Estimates from European and Canadian centers are generally lower than those reported from US facilities.10 However, direct comparisons between studies are thwarted not only by cross-national differences in healthcare service provision and reimbursement but also patient populations, time horizon of analysis, and specific period of observation.13 Yet despite the marked differences in cost estimates, relatively few studies have systematically assessed the factors explaining this variation. To the extent that average costs have declined over time, savings have generally been attributed to the increased use of peripheral blood (vs bone marrow) stem cells and attendant declines in time to hematological recovery.10,13 With respect to interpatient cost variation, however, limited data suggest that adverse events substantially contribute to the cost of the procedure. Clinical complications such as infection and in-hospital mortality, rather than baseline patient characteristics, were found to predict significantly higher cost in a single-institution retrospective study of patients undergoing autologous HSCT for hematological malignancies.18 Similarly, results from a recent prospective, multicenter cost analysis of patients undergoing allogeneic-sibling stem-cell transplantation concurred in finding that transplantrelated complications were more important risk factors for higher cost than baseline status.19 Given the widespread adoption of autologous transplantation in the treatment of lymphoid malignancies, we sought to better understand the frequency with which adverse events complicate hospital admissions for autologous HSCT and the extent to which these complications contribute to hospital cost. To that end, we performed a retrospective cohort study in a national sample of multiple myeloma and lymphoma patients admitted to US

1097

hospitals for the purpose of autologous HSCT between the years 2000-2001. We report nationallevel estimates of the incidence of common adverse events, their relation to patient and treatment characteristics, and their impact on hospital cost. We use multivariate methods to explore the independent contribution of these various factors to resource utilization (hospital cost, length of stay) as well as inhospital mortality.

MATERIALS AND METHODS Data Sources Data for the study were drawn from the Nationwide Inpatient Sample (NIS), a component of the Healthcare Cost and Utilization Project sponsored by the Agency for Healthcare Research and Quality. The NIS is the largest nationwide hospital inpatient care database in the US and the only national hospital database with charge information on all patients regardless of payer, including the uninsured, and persons covered by Medicare, Medicaid, or private insurance. Approximating a 20% stratified sample of US community hospital admissions, the NIS contains data from nearly 7 million hospital stays from more than 1000 hospitals and is weighted at the encounter level to permit population-level estimates of inpatient treatment characteristics.20 Diagnosis, Treatment, and Outcome Coding By using International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes, we identified a cohort of patients admitted to US hospitals for hematopoietic stem cell transplant (HSCT) during a 2-year period (2000-2001).21 From that group, we further selected the population treated for a primary diagnosis of multiple myeloma, Hodgkin disease, or non-Hodgkin lymphoma (NHL). We then characterized each transplantation procedure by stem cell source (peripheral blood, bone marrow) and conditioning regimen (with or without total body irradiation). Because comorbid medical illnesses are known to influence the incidence and the severity of adverse outcomes after stem cell procedures, we calculated Charlson comorbidity scores for each patient by using the Dartmouth-Manitoba adaptation for administrative data.22,23 Methods were further adapted to exclude cancer diagnoses, which are, here, considered separately. Adverse outcomes that complicated each admission were tabulated, and their incidence was calculated by diagnosis. Specific outcomes and their ICD9-CM codes are summarized in Table 1. The associations between baseline patient characteristics (age,

1098

CANCER

March 1, 2008 / Volume 112 / Number 5

TABLE 1 ICD-9-CM Codes Used to Identify Patients and Outcomes Group

Category

ICD-9-CM codes

NIS data type

Diagnosis

Non-Hodgkin lymphoma Hodgkin lymphoma Multiple myeloma Autologous Peripheral blood Bone marrow Total body irradiation Febrile neutropenia Bacteremia Other documented infection

196.x, 200.2x, 202.xx, V10.7x 201.xx 203.xx 41.00, 41.01, 41.04, 41.07, 41.09 41.04, 41.07, 41.05, 41.08 41.00, 41.01, 41.09, 41.02, 41.03 92.24, 92.26, 92.77, 92.29 288.0 38.xx, 790.7 001.xx-037.xx, 039.xx-135.xx, 487.1, 487.8, 490, 465.xx., 466.xx, 595.0, 595.9, 595.89, 681.xx-682.xx 99.15 96.01–96.05 528.x

Diagnosis Diagnosis Diagnosis Procedure Procedure Procedure Procedure Diagnosis Diagnosis Diagnosis

Transplant type Stem cell source Conditioning Outcomes

Parenteral nutrition Intubation Stomatitis

sex, comorbidity), treatment characteristics (stem cell source, pretransplant conditioning), and each adverse outcome were assessed. NIS diagnosis and procedure codes could not be used to ascertain disease status at time of transplant or use of many specific medical interventions (eg, erythropoiesisstimulating agents or leukocyte growth-factors). However, because interventions now exist for prophylaxis against several frequent adverse events,— specifically, infectious outcomes (febrile neutropenia, bacteremia, and other infections) and mucositis—we were interested to determine the relation between these potentially preventable ‘‘intermediate’’ outcomes and downstream events such as use of total parenteral nutrition, intubation, and death.

Utilization and Cost Calculations Length of stay is calculated and reported by the NIS as the number of days elapsed from admission to discharge for a single hospitalization. We then calculated mean length of stay for each diagnosis group. Total hospital costs were estimated from NIS charges. We next converted the reported charges to costs using Medicare cost-to-charge ratios for urban centers, a composite of both capital and operating expenses.24 Ratios for urban centers were chosen not only because most facilities performing stem cell procedures are located in metropolitan areas, but also because the generally lower (vs rural) ratios provide more conservative estimates of total cost. Costs obtained from these calculations were then inflated to 2003 US dollars ($US) by using the consumer price index for medical care.25 We calculated mean hospital costs for each diagnostic group. Costs were also considered by outcome group; we report mean costs for admissions in which

Procedure Procedure Diagnosis

each individual adverse outcome, any single adverse outcome, and no adverse outcomes occurred.

Statistical Analysis Bivariate comparisons between diagnostic, treatment, and outcome groups were made using chi-square statistics for categorical variables. Means and 95% confidence intervals were calculated for length of hospital stay and hospital costs and compared by cancer diagnosis and transplant characteristic. To determine factors predicting higher costs and longer length of stay, we constructed multivariate linear regression models for both outcomes. Continuous and categorical variables with more than 2 categories were recoded as dichotomous categorical variables to facilitate interpretation of the regression coefficients. Independent variables incorporated in the models included patient characteristics (age, >65 vs 65 years; sex, men vs women; comorbidity score, 0 vs 1), treatment factors (peripheral blood vs marrow stem cells, total body irradiation vs no radiationcontaining conditioning), and presence and/or absence of individual adverse events (neutropenic fever, bacteremia, other documented infection, stomatitis, intubation, use of total parenteral nutrition, death). Because cost data were positively skewed, we used logarithmic transformations as the dependent variable in regression models for hospital costs.26 To explore the independent contribution of these same factors to in-hospital mortality, we first performed univariate comparisons followed by multivariate logistic regression analysis. Independent variables included in this model were the same as described above for the linear regression models. Odds ratios and 95% confidence intervals were calculated for each postulated predictor variable.

Auto Stem Cell Transplant Complications/Jones et al.

1099

TABLE 2 Patient and Treatment Characteristics of Study Cohort by Diagnosis

Variable Patient characteristics Age, y Mean [95% CI] >65 Sex* Men Women Comorbidity score 0 1 Transplant characteristics Stem cell source Bone marrow Peripheral blood Conditioning TBI No TBI

Myeloma N 5 2762

Non-Hodgkin N 5 2937

Hodgkin N 5 1219

Total N 5 6918

No. (%)

No. (%)

No. (%)

No. (%)

55.2 [54.4, 56.0] 408 (14.8)

48.0 [46.1,49.7] 335 (11.4)

34.1 [32.4,35.8] 15.5 (1.3)

43.4 [41.3,45.5] 758 (11.0)

1618 (58.6) 1144 (41.4)

1835 (62.5) 1097 (37.4)

699 (57.4) 520 (42.6)

4152 (60.0) 2761 (39.9)

2356 (85.3) 406 (14.7)

2568 (87.4) 370 (12.6)

1119 (91.8) 100 (8.2)

6043 (87.4) 876 (12.7)

216.2 (7.8) 2546 (92.2)

279 (9.5) 2659 (90.5)

123 (10.1) 1096 (89.9)

618 (8.9) 6301 (91.1)

320 (11.6) 2442 (88.4)

366 (12.4) 2572 (87.6)

108 (8.9) 1111 (91.1)

794 (11.5) 6125 (88.5)

TBI indicates total body irradiation; 95% CI, 95% confidence interval. * Patient sex was not recorded for 5 admissions in the non-Hodgkin group.

All analyses were conducted by using the survey analysis command set of Stata/SE version 9.2 (Stata Corporation, College Station, Tex), which accounts for sample weights in variance estimates by using Taylor-series linearization. To ensure the accuracy of variance estimates in subgroup analyses, appropriate subpopulation commands were used. All tests of statistical significance were 2-sided, and significance was assessed at P < .05.

RESULTS Baseline patient and treatment characteristics are detailed in Table 2. A total of 8891 patients with a diagnosis of multiple myeloma, Hodgkin disease, or non-Hodgkin lymphoma were admitted to United States hospitals for the purpose of hematopoietic stem cell transplant during the 2-year study period. Most patients underwent autologous procedures (77.8%), representing 6918 inpatient admissions. On average, the patients were young and relatively healthy compared with the general population diagnosed with these disorders reported in the National Cancer Institute’s Surveillance, Epidemiology, and End Results (SEER) statistics.27 Only in the myeloma cohort was the average patient aged older than 50 years (mean, 55.2 years), 5 to 10 years older than the average age for transplanted patients with nonHodgkin and Hodgkin lymphoma, respectively. The proportion of patients older than 65 years was likewise highest in that group (14.8%; P < .001 ). Comor-

bidity (Charlson score, 1) was relatively rare but significantly more common in the myeloma and non-Hodgkin groups (P 5 .041). Nearly 90% of patients had no significant comorbid illness in the Charlson index, and greater than 95% had 1 comorbidity regardless of diagnosis. With respect to procedural characteristics, peripheral blood stem cells were used far more frequently than bone marrow (91.1% vs 8.9%). Although bone marrow was more commonly used for Hodgkin patients than either of the other 2 diagnostic groups, the difference was not significant (P 5 .58). Pretransplantation conditioning included total body irradiation ranging from 8.9% to 12.4% of admissions, a difference that was also not statistically significant (P 5 .73).

Clinical Outcomes Figure 1(Top) depicts the incidence of in-hospital complications by diagnosis group. Overall, the incidence of adverse events examined did not differ significantly by diagnosis. More than 40% of patients experienced no significant complication during the hospitalization in which the transplant was performed. Infectious complications (febrile neutropenia, bacteremia, and other documented infection) were the most commonly reported adverse events. Whereas febrile neutropenia was frequently encountered (range, 39.9% myeloma to 44.4% non-Hodgkin), bacteremia (range, 12.1% Hodgkin to 16.5% my-

1100

CANCER

March 1, 2008 / Volume 112 / Number 5

FIGURE 1. (Top) and (Bottom): Incidence and mean hospital costs of outcomes complicating hospital admissions for autologous HSCT in lymphoid malignancies. Error bars indicate 95% confidence intervals. FN indicates febrile neutropenia; Other Infect, other documented infection; TPN, total parenteral nutrition.

eloma), and other documented infections (range, 2.1% Hodgkin to 6% non-Hodgkin) were far less common. After fever and infection, stomatitis was the most common adverse event, occurring in approximately 40% of study patients. total parenteral nutrition and intubation were distinctly uncommon, and in-hospital mortality was rare, occurring in 2.1%, 4.4%, and 2.6% of Hodgkin, non-Hodgkin, and myeloma patients, respectively. In general, baseline patient characteristics were not associated with the incidence of specific adverse outcomes in this cohort, but there were some notable exceptions. For instance, patients over the age of 65 years were more likely to develop stomatitis during the transplant admission (68.7% vs 58.8%; P 5 .056). Older age was not associated with any of the other identified outcomes. Comorbid medical

illness was significantly associated with an increased risk for intubation compared with patients without comorbid illness (5.2% vs 1.5%; P 5 .001). However, comorbidity did not predict other complications in bivariate analysis. Comorbidity was uncommon in the cohort as a whole, suggesting that patients undergoing autologous HSCT for these diagnoses are likely healthier than the larger populations from which they are drawn. Peripheral blood stem cells (PBSC) have previously been reported to decrease the time to hematopoietic reconstitution after myeloablative chemotherapy. In the cohort studied here, PBSC transplants were associated with a nonsignificant decreased risk for developing bacteremia (14.4% vs 20.2%; P 5 .064) but with a similar risk for febrile neutropenia (32.2% vs 39.8%; P 5 .136) or other documented infection (3.1% vs 4.3%; P 5 .58). Stem cell source was not associated with risk for any of the other outcomes that we studied. On the other hand, conditioning with regimens that included total body irradiation markedly increased the risk for stomatitis (66.7% vs 36.7%; P < .001) and, to a lesser extent, increased the risk for neutropenic fever (43.3% vs 31.7%; P 5 .016). We further explored the relation between ‘‘preventable’’ intermediate outcomes and downstream adverse events. We did not detect an association between stomatitis and most of the downstream events studied (bacteremia, neutropenic fever, documented infection, or intubation; data not shown). However, stomatitis was associated with a nonsignificant increase in the risk for total parenteral nutrition (52.6% vs 38.3%; P 5 .095). Bacteremia, on the other hand, was associated with a marked increase in risk for use of both total parenteral nutrition (20.3% vs11.4%; P 5 .0082) and intubation (7.4% vs 1.1%; P < .001).

Economic Outcomes Length-of-stay data are summarized in Table 3. The mean length of stay for all diagnoses was 19.6 days and did not differ significantly among groups. Infusion of peripheral blood stem cells did not result in significantly shorter hospital admissions when compared with procedures in which bone marrow cells were used. Conversely, conditioning with total body irradiation did lead to significantly longer hospitals stays. Overall, hospitalizations were approximately 6 days longer in patients conditioned with irradiation, a statistically significant difference that was observed in all diagnostic groups. Hospital costs by diagnosis are given in Table 3. The mean hospital cost for all patients in the cohort was US $51,312. Costs for the 3 groups were similar,

Auto Stem Cell Transplant Complications/Jones et al.

1101

TABLE 3 Mean Length of Stay and Hospital Costs by Diagnosis and Transplant Characteristics Myeloma Mean length of stay, d (95%CI) Overall By stem cell source Peripheral blood Bone marrow By conditioning regimen TBI No TBI Mean hospital costs, 2003 US$ (95% CI) Overall By stem cell source Peripheral blood Bone marrow By conditioning regimen TBI No TBI

Non-Hodgkin

Hodgkin

All diagnoses

17.8 (15.7–20.0)

21.3 (19.0–23.7)

19.2 (16.9–21.6)

19.6 (17.8–21.3)

17.8 (15.4–20.2) 18.1 (16.3–19.8)

20.9 (18.6–23.3) 25.1 (20.8–29.5)

19.3 (16.9–21.6) 19.2 (11.0–27.4)

20.8 (19.1–22.5) 25.2 (22.1–28.3)

25.3 (19.5–31.2) 16.8 (14.7–18.9)

24.9 (19.8–20.0) 20.8 (18.3–23.4)

29.4 (23.4–35.3) 18.3 (16.2–20.4)

26.7 (23.0–30.3) 20.5 (18.9–22.1)

$42,863 (36,012–49,714)

$58,673 (50,979–$66,367)

$52,993 (45,352–$60,634)

$51,312 (45,753–$56,871)

$43,059 (35,658–50,460) $40,603 (34,339–46,867)

$57,445 (49,792–65,118) $70,578 (54,596–86,560)

$53,003 (45,442–60,565) $52,902 (26,872–78,932)

$57,037 (50,970–63,103) $74,976 (60,756–89,197)

$68,578 (46,604–90,551) $39,668 (33,469–45,868)

$67,916 (49,292–86,539) $57,455 (49,521–65,390)

$79,170 (52,422–105,917) $50,536 (44,072–57,001)

$76,782 (60,069–93,495) $56,242 (50,689–61,786)

TBI indicates total body irradiation; 95% CI, 95% confidence interval.

but costs for myeloma patients tended to be lower, likely owing to the nonsignificant trend toward shorter length of stay in that group. Despite evidence for faster hematological recovery, peripheral blood stem cells did not result in lower hospital costs in this cohort. However in our study, longer hospital stays associated with total body irradiation produced similarly greater hospital costs ($76,782 vs $56,242). Mean hospital costs for patients conditioned with total body irradiation were higher overall and in each diagnostic group, but only in the myeloma group was the difference statistically significant. Mean hospital costs were also assessed by adverse outcome (see Fig. 1(Bottom)). As expected, uncomplicated hospitalizations were the least expensive, costing less than the cohort average regardless of diagnosis. Conversely, incident infectious complications, stomatitis, and the use of total parenteral nutrition increased the mean cost of hospitalization by a variable amount above the base case that ranged from $15,000 to as much as $50,000 (see Fig. 1(Bottom)). Bacteremia, in particular, nearly doubled the cost of hospitalization for autologous HSCT compared with uncomplicated cases in both the nonHodgkin lymphoma and multiple myeloma groups. Notably, the rarest outcomes were the most costly. Admissions complicated by either death or the need for intubation resulted in hospital costs 3-fold to 4fold greater than admissions without adverse events, a difference which was highly statistically significant. In general, mean cost per hospital day (data not shown) was higher for each adverse outcome group, suggesting that higher mean costs were a function

not only of longer length of stay but also greater resource intensity. To better assess which specific patient factors, treatment characteristics, and adverse outcomes predicted longer hospitalizations and higher hospital costs, we fitted multivariate linear regression models. Because the incidence of adverse outcomes did not differ significantly by diagnosis (Fig. 1(Top)), the groups were collapsed for modeling purposes. The results of these regression analyses for length-of-stay and log-transformed hospital costs are presented in Table 4. Baseline patient characteristics continued to demonstrate no relation to the duration of hospitalization. Conditioning with total body irradiation, on the other hand, remained a predictor of greater length of stay in the multivariate model, increasing hospital stays by 4.9 days on average (P 5.001). Among adverse outcomes, bacteremia, other documented infection, need for total parenteral nutrition, and death all predicted longer hospitalizations. Hospitalization ending in death predicted a 14.6 day increase in the duration of inpatient admission. Only the absence of adverse events was significantly associated with a decreased length of stay (4.7 days; P 5.012). Outcomes of the regression analysis for logtransformed hospital costs were similar. The regression coefficients provided insight into the relative contribution of each characteristic and/or outcome to mean costs. For instance, preparative regimens including total body irradiation are predicted to increase the cost of hospitalization by >9%. Bacteremia or other documented infection each predicted

1102

CANCER

March 1, 2008 / Volume 112 / Number 5

TABLE 4 Multivariate Models for Mean Hospital Costs and Length of Stay Log-transformed mean hospital costsy

Mean length of stay* Variable

Coefficient

95% CI

P

Coefficient

95% CI

P

Age >65 y Sex, men vs women Comorbidity score, 0 vs 1 TBI conditioning PBSCT vs BMT No adverse outcome Febrile neutropenia Bacteremia Other infection Stomatitis TPN Intubation Death

1.42 0.68 1.77 4.92 21.11 24.66 1.11 6.45 10.50 20.16 3.47 21.09 14.57

20.41, 3.26 20.40, 1.76 20.58, 4.11 1.92, 7.92 23.46, 1.24 28.29, 21.03 20.48, 2.70 4.04, 8.86 4.35, 16.65 23.53, 3.21 0.98, 5.97 214.36, 12.18 3.83, 25.32

.128 .217 .139 .001 .354 .012 .170

Suggest Documents