Pediatric Extracorporeal Membrane Oxygenation Mortality Is Related to Extracorporeal Membrane Oxygenation Volume in US Hospitals

Pediatric Extracorporeal Membrane Oxygenation Mortality Is Related to Extracorporeal Membrane Oxygenation Volume in US Hospitals

j o u r n a l o f s u r g i c a l r e s e a r c h  a p r i l 2 0 1 9 ( 2 3 6 ) 1 5 9 e1 6 5 Available online at www.sciencedirect.com ScienceDirect...

298KB Sizes 0 Downloads 2 Views

j o u r n a l o f s u r g i c a l r e s e a r c h  a p r i l 2 0 1 9 ( 2 3 6 ) 1 5 9 e1 6 5

Available online at www.sciencedirect.com

ScienceDirect journal homepage: www.JournalofSurgicalResearch.com

Pediatric Extracorporeal Membrane Oxygenation Mortality Is Related to Extracorporeal Membrane Oxygenation Volume in US Hospitals Dani O. Gonzalez, MD, Yuri V. Sebastia˜o, PhD, Jennifer N. Cooper, PhD, Peter C. Minneci, MD, MHSc, and Katherine J. Deans, MD, MHSc* Center for Surgical Outcomes Research, The Research Institute at Nationwide Children’s Hospital, Columbus, Ohio

article info

abstract

Article history:

Background: Our objective was to examine extracorporeal membrane oxygenation (ECMO)

Received 27 August 2018

utilization and determine whether pediatric-specific and overall ECMO volumes are asso-

Received in revised form

ciated with mortality rates.

15 November 2018

Methods: State Inpatient Databases from 17 states were queried for ECMO admissions

Accepted 21 November 2018

during 2008-2014. Hospitals in which >90% of their ECMO patients were 18 y old were

Available online xxx

considered pediatric ECMO centers. Hospital overall ECMO volumes were calculated as the average annual number of admissions, of any age, and categorized as <6, 6-14, 15-30, and

Keywords:

>30. Multivariable analyses were conducted to examine the impact of ECMO volume on

Pediatric

pediatric in-hospital mortality.

ECMO

Results: There were 4546 pediatric ECMO admissions across 84 hospitals. Most patients

Volume

were neonates (59.9%), and the most common indication for ECMO was neonatal respira-

Mortality

tory failure (20.1%). Approximately 35% of hospitals offering pediatric ECMO averaged <6

Survival

annual ECMO admissions. Centers with >30 annual ECMO admissions had significantly lower mortality than hospitals with lower ECMO volume. Among the high-volume centers, pediatric ECMO centers had significantly lower mortality rates than high-volume nonpediatric ECMO centers (17.4% versus 38.2%). Conclusions: A high proportion of hospitals performing pediatric ECMO have a low number of annual ECMO admissions. Pediatric centers with high volume had the lowest riskadjusted mortality rates for pediatric ECMO. ª 2018 Elsevier Inc. All rights reserved.

Introduction Extracorporeal membrane oxygenation (ECMO) is a resourceintensive technology that requires organized and experienced health care teams. Recent reports suggest an association between ECMO volume and mortality rates, with higher volume associated with lower mortality.1-5 Although controversial, a minimum acceptable ECMO volume has been suggested to

maintain clinical proficiency and justify the institutional support necessary for a comprehensive program. The Extracorporeal Life Support Organization recommends that centers perform at least 6 ECMO procedures per year to maintain clinical expertise. The International ECMO Network recommends 20 cases per year.2 Freeman et al.3 conducted a review of the Pediatric Health Information System (PHIS) including data from children’s hospitals and concluded that a

* Corresponding author. Center for Surgical Outcomes Research, The Research Institute at Nationwide Children’s Hospital, 700 Children’s Drive, FB 3A.3 Columbus, OH 43205. Tel.: þ1 614 722 3066; fax: þ1 614 722 6980. E-mail address: [email protected] (K.J. Deans). 0022-4804/$ e see front matter ª 2018 Elsevier Inc. All rights reserved. https://doi.org/10.1016/j.jss.2018.11.043

160

j o u r n a l o f s u r g i c a l r e s e a r c h  a p r i l 2 0 1 9 ( 2 3 6 ) 1 5 9 e1 6 5

minimum volume of 22 cases per year was associated with improved mortality. The objective of this study was to examine ECMO utilization across a diverse set of the US ECMO centers to determine whether pediatric-specific and overall ECMO volumes are associated with mortality rates.

Methods Data source We queried the Healthcare Cost and Utilization Project (HCUP) State Inpatient Databases (SID), sponsored by the Agency for Healthcare Research and Quality. The SID contains data from all inpatient encounters in all or nearly all acute care hospitals in participating states. Therefore, the SID can provide a more comprehensive account of ECMO admissions in a given state, compared to tertiary hospital databases (e.g., PHIS), national samples (e.g., Nationwide Inpatient Sample), or the voluntary Extracorporeal Life Support Organization (ELSO) registry. In contrast to the PHIS, which only captures ECMO performed at participating children’s hospitals, and ELSO, which captures only voluntarily entered data from centers that elect to enter their data, the SID should capture nearly all ECMO encounters performed within a state regardless of which hospital type or whether the site is a contributing ELSO center. Data use agreements with HCUP were completed by the research team. SID data were deidentified by HCUP, which qualified for exemption from human subjects review by the Nationwide Children’s Hospital Institutional Review Board.

Cohort development We used available SID data from 17 states for the years 20082014 (See Appendix, Supplementary Table S1). The study population comprised admissions with an International Classification of Diseases, ninth revision, procedure code for ECMO (39.65) and discharged between January 1, 2008, and December 31, 2014. ECMO admissions for all ages were used to calculate the average annual number of ECMO admissions and the proportion of pediatric ECMO admissions for each hospital. The remainder of the analyses focused solely on pediatric ECMO admissions (i.e., patients aged 18 y at admission). Admissions that resulted in a transfer to a different hospital within 1 d after ECMO cannulation were excluded from the study. In addition, admissions for which there was insufficient information to classify patients into one of seven diagnostic groups (described below) were excluded. Combined, the exclusions comprised less than 8% of all pediatric ECMO admissions.

Outcome measures The outcome of interest was in-hospital pediatric ECMO mortality, defined as the proportion of pediatric ECMO admissions resulting in in-hospital death. The primary exposure of interest was overall ECMO center volume, defined as the average number of ECMO admissions, of any age (pediatric and adult), across the available data years for each hospital. Based on previous literature, the following ECMO center

volume categories were used for the study: <6, 6-14, 15-30, and >30 cases/y.1 The final category, comprising high-ECMO volume centers, was used as the reference category for comparison. Hospitals for which >90% of their ECMO patients were 18 y old were considered pediatric ECMO centers. The following patient characteristics were selected based on previous literature and clinical relevance to be examined as risk factors for in-hospital ECMO mortality: age, race, and ethnicity, health insurance, elective admission, transfer admission, presence of one or more pediatric complex chronic conditions,6 and diagnostic group (indication category for ECMO). Diagnostic groups were identified by adapting a previously published classification, which emulates indications for ECMO used by ELSO.3 Using patient age, diagnosis, and procedure codes, patients were classified into the following groups: congenital diaphragmatic hernia, neonatal or pediatric cardiac arrest, neonatal or pediatric cardiac disease, and neonatal or pediatric respiratory failure (See Appendix, Supplementary Tables S2-S3). Neonatal admission was defined as patient age of <30 d at admission for admissions in 16 of the 17 states included in the study. For the remaining state (Florida), age in days was not available for patients younger than 1 y at admission (infants). For Florida infants, the primary diagnosis for each admission was examined and classified as neonatal or pediatric. Patients for whom the admitting diagnosis was a neonatal condition were classified as neonates, and patients for whom the admitting diagnosis was a nonneonatal condition were classified as pediatric (See Appendix, Supplementary Table S4). Patients with an admitting diagnosis for a condition that can happen in any period of infancy (neonatal or otherwise) could not be classified as neonatal or pediatric and were, therefore, excluded from the analysis.

Statistical analysis Frequencies and percentages were used to describe patient characteristics in the study population and by category of overall ECMO center volume. The associations between individual patient characteristics and in-hospital death and between individual patient characteristics and ECMO center volume were tested using Pearson chi-squared test. We then determined the relationship between these variables and the odds of in-hospital death using random-intercept logistic regression modeling. We modeled the odds of in-hospital death among pediatric ECMO patients as a function of overall ECMO center volume, patient characteristics, and center percentage of pediatric ECMO cases (90%, >90%). A cutoff of 30 cases per year was used to characterize high- versus lowvolume ECMO centers and centers with >90% pediatric cases were considered pediatric centers and centers with <90% pediatric cases were considered nonpediatric centers. As an example, a center that performed >30 ECMO cases per year with <90% of the cases being pediatric would be categorized as a high-volume nonpediatric center, whereas a center that performed >30 ECMO cases per year with >90% of the cases being pediatric would be categorized as a high-volume pediatric center. The first model included ECMO center volume as the sole covariate to estimate the association between ECMO volume and mortality association. The next model added

161

gonzalez et al  pediatric ecmo volume and mortality

Table 1 e Pediatric ECMO study population characteristics (n [ 4546). Variable

Characteristic

Female Age

Race and ethnicity

Health insurance

n (%) 2046 (45.0)

<1 y, neonate*

2724 (59.9)

<1 y, nonneonatal infant

665 (14.6)

1-10 y

722 (15.9)

11-18 y

435 (9.6)

Non-Hispanic, white

1779 (39.1)

Non-Hispanic, black

722 (15.9)

Hispanic

707 (15.6)

Other, nonHispanic

653 (14.4)

Other/Unknown ethnicity

685 (15.1)

Medicaid

2503 (55.1)

Private

1745 (38.4)

Self-pay

58 (1.3)

Other/unknown

240 (5.3)

Elective admission

627 (13.8)

Transfer admission

2175 (47.8)

1 complex chronic conditionsy

2853 (62.8)

Diagnostic group

Postcannulation LOS (d)z (n ¼ 2752)

Transferred out

Table 1 e (continued ) Variable

Neonatal, congenital diaphragmatic hernia

556 (12.2)

Neonatal, cardiac arrest

528 (11.6)

Neonatal, cardiac disease

728 (16.0)

Neonatal, respiratory failure

912 (20.1)

Pediatric, cardiac arrest

779 (17.1)

Pediatric, cardiac disease

514 (11.3)

Pediatric, respiratory failure

529 (11.6)

<7

125 (4.5)

Characteristic

n (%)

Other type of facility

235 (5.2)

In-hospital death Center ECMO volume (average total cases/y)

1794 (39.5) <6

167 (3.7)

6-14

699 (15.4)

15-30

1827 (40.2)

>30

1853 (40.8)

State Inpatient Databases, 2008-2014. Frequencies (n) and percentages (%) shown. * Neonate: age 30 d or younger at admission or presence of neonatal condition (ICD-9 diagnosis). y Complex chronic condition categories: neurologic and neuromuscular, renal and urologic, gastrointestinal, hematologic/ immunologic, other congenital/genetic defect, malignancy, technology dependence, transplantation. z Limited to patients who were alive at discharge.

patient factors found to have an association with in-hospital death or with hospital volume of ECMO in univariable analysis at P < 0.20. To examine impact of being a pediatric versus nonpediatric ECMO center on the volumeemortality relationship, the final model included variables for overall ECMO center volume, center percentage of pediatric ECMO cases, and an interaction term between the two. Statistical significance was set at P < 0.05.

Results

8-14

247 (9.0)

15-30

806 (29.3)

30

1362 (49.5)

Unknown

212 (7.7)

Acute care hospital

378 (8.3) (continued)

There were 4546 pediatric ECMO admissions across 84 hospitals during the study period (Table 1). Most patients were neonates (59.9%), non-Hispanic white (39.1%), and insured by Medicaid (55.1%). The majority of patients had at least one complex chronic condition (62.8%). The most common indication for ECMO was neonatal respiratory failure (20.1%) followed by pediatric cardiac arrest (17.1%) and neonatal cardiac disease (16.0%). The postcannulation length of stay was at least 30 d in nearly 50% of patients who were alive at discharge. Most patients were managed at centers with an average of 15-30 annual ECMO admissions (40.2%) or > 30 annual ECMO admissions (40.8%). The overall mortality rate was 39.5%. Table 2 demonstrates the characteristics of the study population across the different volume categories. Approximately 35% of hospitals offering pediatric ECMO averaged <6 annual overall ECMO admissions, and only 15% of centers averaged >30 annual overall ECMO admissions during the study period. After risk adjustment, centers with lower volume had significantly higher mortality than hospitals with >30 annual ECMO admissions (Figure). Within the subgroup of high-volume centers, pediatric ECMO centers had

162

j o u r n a l o f s u r g i c a l r e s e a r c h  a p r i l 2 0 1 9 ( 2 3 6 ) 1 5 9 e1 6 5

Table 2 e Pediatric patient case mix by center volume in the ECMO study population (n [ 4546). Variable

Volume (average total cases/y)

Centers, n Centers w/>90% pediatric cases, n Pediatric cases, n Female, %

<6

6-14

15-30

>30

29 9

20

22

13

9

10

3

167

699

1827

1853

44.3

49.1

45.5

43.1

60.5

57.5

62.6

58.1

Age, % <1 y, neonate* <1 y, non-neonatal infant

12.0

14.2

12.8

16.8

1-10 y

11.4

18.7

15.9

15.2

11-18 y

16.2

9.6

8.7

9.9

43.7

37.3

42.8

35.8

Race and ethnicity, % Non-Hispanic, white Non-Hispanic, black

10.2

13.6

17.2

15.9

Hispanic

21.0

17.6

14.3

15.5

Other

25.2

31.5

25.7

32.8

46.7

55.2

55.9

54.9

Private

40.1

40.2

36.1

39.8

Other

13.2

4.6

7.9

5.3

14.4

14.6

7.8

19.3

28.1

36.3

49.3

52.5

55.7

59.1

64.0

63.5

11.4

8.9

15.0

10.9

7.8

14.0

9.3

13.4

18.6

16.7

15.6

15.9

Health insurance, % Medicaid

Elective admission, % Transfer admission, % 1 complex chronic conditions

y

Diagnostic group, % Neonatal, congenital diaphragmatic hernia Neonatal, cardiac arrest Neonatal, cardiac disease Neonatal, respiratory failure

22.8

17.9

22.8

18.0

Pediatric, cardiac arrest

17.4

20.6

15.5

17.4

Pediatric, cardiac disease

10.8

9.9

9.9

13.3

Pediatric, respiratory failure

11.4

12.0

12.0

11.2

Postcannulation LOS (d)z (n ¼ 2540) <14

24.0

11.8

12.4

16.7

15-30

40.6

33.3

33.2

29.3

30

35.4

54.8

54.5

54.1

Transferred out, %

18.0

8.0

12.0

16.6

Pediatric in-hospital death, %

41.3

44.6

42.5

34.4

State Inpatient Databases, 2008-2014. * Neonate: age 30 d or younger at admission or presence of neonatal condition (ICD-9 diagnosis). y Complex chronic condition categories: neurologic and neuromuscular, renal and urologic, gastrointestinal, hematologic/immunologic, other congenital/genetic defect, malignancy, technology dependence, transplantation. z Limited to patients who were alive at discharge; “Unknown” LOS category omitted (n ¼ 212).

significantly lower mortality rates than high-volume centers considered nonpediatric ECMO centers (17.4% versus 38.2%, P ¼ 0.0053; Table 3). Within the other three volume categories, there were no significant differences in mortality between pediatric and nonpediatric ECMO centers (Table 4). Among centers performing <30 ECMO cases per year, the risk-adjusted predicted mortality was similar for both pediatric and nonpediatric centers within each volume category (Table 4).

Discussion In this review of statewide databases of 17 states, a high proportion of hospitals performing pediatric ECMO have a low number of annual ECMO admissions. In addition, high-volume centers had the lowest mortality rates, with pediatric centers with high-ECMO volume having the lowest risk-adjusted mortality rates for pediatric ECMO. Children treated at lower-

163

gonzalez et al  pediatric ecmo volume and mortality

Fig e Adjusted association between ECMO volume and in-hospital death.

volume ECMO centers have similar mortality regardless of whether the center was a pediatric or nonpediatric center. This study is the first to demonstrate an association between pediatric ECMO mortality rates and both overall ECMO volume and pediatric-specific ECMO volume. These findings raise concerns over maintenance of ECMO expertise at low-volume centers and suggest that regionalization of pediatric and neonatal ECMO may be beneficial. A high proportion of hospitals performing pediatric ECMO had a low number of annual ECMO admissions in this study. Previous reports have recommended a minimal acceptable ECMO volume to maintain clinical expertise.2,3 Freeman et al.3 performed a review of the PHIS database including ECMO admissions to children’s hospitals and reported that the minimal annual caseload associated with reduced mortality was 22 cases. In the present study, only 15% of centers averaged 30 annual ECMO admissions during the study period, with most centers having less than 15 annual admissions. These results suggest that many centers offering pediatric ECMO do not meet the minimum recommended annual volume to maintain expertise. Although the majority of studies throughout the literature have established a relationship between higher-ECMO volume and improved outcomes, several studies have found no association between higher volume and survival.7,8 In a review

of the PHIS database that included children receiving ECMO for cardiac operations, there was no association between ECMO volume and outcomes.8 In a review of the Nationwide Inpatient Sample that included adult patients receiving ECMO for respiratory failure, institutions with low volume had better outcomes than centers with medium and high volume.7 The patient populations in those two studies, however, are different from the cohort in the present study. ECMO is a complex technology that requires experienced teams and specialized protocols. One could logically postulate that centers with higher volume have more experience performing ECMO. Most large series throughout the literature support this.1-5 Our study provides further evidence to support the volumeemortality relationship with ECMO. Although it has been established that ECMO volume is associated with mortality, the question of interest in this study was whether pediatric-specific ECMO volume (versus overall ECMO volume) was associated with pediatric ECMO mortality. In this study, among the high-volume centers, ECMO centers that performed primarily pediatric ECMO had lower risk-adjusted mortality rates for pediatric ECMO than the high-volume centers that were considered nonpediatric. This supports regionalization of ECMO care within centers of excellence for pediatric ECMO, which has previously been proposed.4 Regionalization of ECMO is difficult because the

Table 3 e Adjusted odds ratio (OR) and mortality by combined hospital volume and pediatric case percent. Center ECMO volume* <30 30

Center % pediatric cases

OR (95% CI)

P

Predicted mortality

90%

3.49 (1.71-7.12)

0.0006

42.3 (36.2-48.7)

>90%

3.39 (1.67-6.89)

0.0007

41.6 (35.9-47.6)

90%

2.94 (1.38-6.27)

0.0053

38.2 (30.1-47.0)

>90%

1

Reference

17.4 (9.8-29.0)

Pediatric ECMO study population (n ¼ 4546). State Inpatient Databases, 2008-2014. OR and predicted mortality adjusted from a logistic mixed-effects model adjusting for elective admission, transfer admission, discharge year, pediatric complex chronic conditions, diagnostic group, and hospital percentage of pediatric cases>90%. * Average total ECMO admissions per year during the study period.

164

j o u r n a l o f s u r g i c a l r e s e a r c h  a p r i l 2 0 1 9 ( 2 3 6 ) 1 5 9 e1 6 5

Table 4 e Adjusted odds ratio (OR) and mortality by combined hospital volume and pediatric case percent. Center ECMO volume* <6

6-14

15-30 30

Center % pediatric cases

OR (95% CI)

P

Predicted mortality

90%

3.36 (1.35-8.32)

0.009

41.4 (27.6-56.8)

>90%

2.96 (1.20-7.34)

0.019

38.4 (25.2-53.7)

90%

3.47 (1.60-7.53)

0.0016

42.2 (32.9-52.2)

>90%

4.12 (1.80-9.43)

0.0008

46.5 (34.7-58.6)

90%

3.33 (1.57-7.05)

0.0017

41.2 (33.0-49.9)

>90%

3.37 (1.60-7.11)

0.0014

41.5 (33.5-50.0)

90%

2.93 (1.38-6.25)

0.0052

38.2 (30.2-47.0)

>90%

1

Reference

17.4 (9.8-29.0)

Pediatric ECMO study population (n ¼ 4546). State Inpatient Databases, 2008-2014. OR and predicted mortality adjusted from a logistic mixed-effects model adjusting for elective admission, transfer admission, discharge year, pediatric complex chronic conditions, diagnostic group, and hospital percentage of pediatric cases>90%. * Average total ECMO admissions per year during the study period.

severity of disease of children requiring ECMO and the large geographic area of the United States limit the ability of patients to be transferred to other centers. However, regionalization may be possible in areas where there are multiple ECMO centers within close proximity to each other, such as major metropolitan areas in the United States. Our results suggest that children requiring EMCO have lower mortality when they are treated at a high-volume ECMO center as compared with a lower-volume ECMO center. Furthermore, the beneficial effect of higher volume is accentuated if the child is treated at a high-volume pediatric center as compared with just a high-volume ECMO center. These results may explain the inconsistent results in the literature on the relationship between ECMO volume and outcomes. It was only at a cutoff of 30 cases per year that we were able to detect a benefit of ECMO volume on mortality. At ECMO volumes <30 cases per year, there was no difference in pediatric ECMO mortality across lower-volume categories and based on whether they were treated at a pediatric or nonpediatric ECMO centers. Using these results to improve practice will be difficult because only a minority of ECMO centers perform >30 cases per year. There were a number of limitations with this work, including the potential for miscoding within the statewide databases. Miscoding of diagnosis codes could lead to patients being misclassified within diagnosis groups. Moreover, the lack of pre-ECMO clinical data to accurately classify the indication for ECMO is a limitation. In addition, establishing the definition of low-, medium-, and high-volume ECMO centers presents a challenge; various studies have defined the ECMO volume categories differently. In this study, we used previously published criteria based on average annual ECMO cases per center as measure of institutional experience administering ECMO.1 Uncertainties remain about whether using alternative criteria to define center volume provides different findings in regard to association with mortality.8 A high proportion of hospitals performing pediatric ECMO have a low number of annual ECMO admissions, which raises concerns over maintenance of ECMO expertise. Pediatric centers with high ECMO volume had the lowest risk-adjusted mortality rates for pediatric ECMO, suggesting that

regionalization of pediatric and neonatal ECMO, where geographically feasible, may improve survival.

Acknowledgment Authors’ contributions: K.J.D. had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. D.O.G., Y.V.S., J.N.C., P.C.M., and K.J.D. all contributed to the following: study concept and design, acquisition, analysis, or interpretation of data, drafting of the article, critical revision of the article for important intellectual content, and statistical analysis.

Disclosure This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Appendix A. Supplementary data Supplementary data related to this article can be found at https://doi.org/10.1016/j.jss.2018.11.043.

references

1. Barbaro RP, Odetola FO, Kidwell KM, et al. Association of hospital-level volume of extracorporeal membrane oxygenation cases and mortality. Analysis of the extracorporeal life support organization registry. Am J Respir Crit Care Med. 2015;191:894e901. 2. Combes A, Brodie D, Bartlett R, et al. Position paper for the organization of extracorporeal membrane oxygenation programs for acute respiratory failure in adult patients. Am J Respir Crit Care Med. 2014;190:488e496. 3. Freeman CL, Bennett TD, Casper TC, et al. Pediatric and neonatal extracorporeal membrane oxygenation: does center volume impact mortality?*. Crit Care Med. 2014;42:512e519.

gonzalez et al  pediatric ecmo volume and mortality

4. Karamlou T, Vafaeezadeh M, Parrish AM, et al. Increased extracorporeal membrane oxygenation center case volume is associated with improved extracorporeal membrane oxygenation survival among pediatric patients. J Thorac Cardiovasc Surg. 2013;145:470e475. 5. Kirkland BW, Wilkes J, Bailly DK, Bratton SL. Extracorporeal membrane oxygenation for pediatric respiratory failure: risk factors associated with center volume and mortality. Pediatr Crit Care Med. 2016;17:779e788. 6. Feudtner C, Feinstein JA, Zhong W, Hall M, Dai D. Pediatric complex chronic conditions classification system

165

version 2: updated for ICD-10 and complex medical technology dependence and transplantation. BMC Pediatr. 2014;14:199. 7. McCarthy FH, McDermott KM, Spragan D, et al. Unconventional volume-outcome associations in adult extracorporeal membrane oxygenation in the United States. Ann Thorac Surg. 2016;102:489e495. 8. Rettiganti M, Seib PM, Robertson MJ, Wilcox A, Gupta P. Impact of varied center volume categories on volume-outcome relationship in children receiving ECMO for heart operations. J Artif Organs. 2016;19:249e256.