論文ID: CJ-22-0107
Background: In modern critical care, extracorporeal membrane oxygenation (ECMO) is crucial in the management of severe respiratory and cardiac failure. Nationwide studies of the relationship between hospital volume and outcomes of ECMO use are unavailable.
Methods and Results: Using Taiwan’s National Health Insurance Research Database, we identified 11,734 adult patients who received ECMO support in 101 hospitals between January 1, 2001, and December 31, 2017. Outcomes included in-hospital mortality, 1-year mortality, and ECMO-related complications. Cox proportional hazards model, locally estimated scatterplot smoothing, and restricted cubic spline regression were used to analyze the volume–outcome relationship. The overall in-hospital mortality rate was 65.5%, and the 1-year mortality rate was 70.6% in this database. The 101 hospitals were divided into 4 groups based on annual volume. The in-hospital and 1-year mortality rates were significantly lower in the high-volume group (annual volume >40) than in the low-volume group (annual volume <10).
Conclusions: For critical care, high-volume hospitals have superior short-term and mid-term outcomes. To make the medical system equitable and reasonable, establishing a rapid and efficient nationwide referral system should be considered.
In modern critical care, extracorporeal membrane oxygenation (ECMO) is crucial in the management of severe respiratory and cardiac failure.1,2 ECMO provides temporary pulmonary or cardiac support, thus increasing the time available for physicians to conduct further examinations or treatment. Moreover, ECMO is indicated in postcardiotomy cardiogenic shock, heart and/or lung transplantation, major trauma, and even extracorporeal cardiopulmonary resuscitation.3–8
Over the past decades, the number of hospitals that offer ECMO has rapidly grown owing to improved techniques and extended indications, particularly for adults.9 However, the administration of ECMO support requires an experienced and multidisciplinary team for high-quality care. In a previous study, this complicated support system was often associated with a volume–outcome relationship in which high-volume centers had better outcomes.9–14 The Extracorporeal Life Support Organization (ELSO) Registry data indicated a strong association of a higher hospital-level ECMO volume with lower in-hospital mortality in neonates and adults.9 Furthermore, Tay et al revealed a similar relationship between hospital ECMO volume and in-hospital mortality in Korea.13 Those studies reported a similar relationship between in-hospital mortality and hospital annual volume, but not the mid-term outcome and other sequela after ECMO support.
The National Health Insurance (NHI) scheme in Taiwan is a government-operated single-payer program that has covered 99.6% of the entire population since 1995 and was established to minimize socioeconomic bias. It has been reimbursing the expenditure of ECMO and almost all medical disbursement during hospitalization since 2002.15,16 Taiwan’s National Health Insurance Research Database (NHIRD) contains records on in-hospital outcomes, long-term longitudinal follow-up outcomes, and other details of patients who received ECMO support which reimburses the expenditure of almost all medical disbursement, including surgeries, medication and even lifesaving procedures.17,18 This population-based study used the NHIRD to determine the relationship between the clinical outcome and hospital annual volume.
This open, observational, longitudinal cohort study used hospitalization records from the NHIRD to compare the outcomes of patients who first-ever received ECMO between January 2001 and December 2017. In total, 16,735 patients were identified. Patients were excluded if they were <20 years old or had missing demographic information (Figure 1). The NHI’s claim order code 68036 or the International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) procedure code 39.65 and International Classification of Diseases, 10th Revision, Procedure Coding System code 5A15223 were applied to extract data of patients who received ECMO in all 101 hospitals in Taiwan in which the procedure was performed during the study period. We excluded the years in which a particular hospital did not perform any ECMO procedures to obtained the annual volume of ECMO procedures performed at each hospital by dividing the total number of ECMO procedures by the number of years in which ECMO was performed. For example, 393 ECMO procedures were performed at a particular hospital between 2001 and 2017, but not between 2001 and 2003. With 14 as the number of years in which ECMO was performed, the annual volume of ECMO at that hospital was 28. We divided the population according to such integer values based on the distribution of the data. The 101 hospitals were divided into 4 groups: volume 1 (annual volume >40), volume 2 (annual volume between 20 and 40), volume 3 (annual volume between 10 and 20), and volume 4 hospitals (annual volume <10).
Enrollment of the population-based cohort of 16,735 adult patients who received extracorporeal membrane oxygenation (ECMO) support for the first time between 2001 and 2017 according to the Taiwan National Health Insurance Research Database.
We tracked the comorbidity history for all patients for at least 3 years before the index admission in the NHIRD. The baseline characteristics and ECMO-related complications were identified using ICD-9-CM codes for the index hospitalization and prior hospitalizations. The outcomes of interest in this study included hospital annual volume, ECMO-related complications, in-hospital and 1-year mortality rates, and all-cause mortality. The patients were followed up from their index admission to December 31, 2017, or death; no missing data were found in the database during the follow-up period. Detailed information regarding the NHIRD has been reported previously.15–18
Statistical AnalysisThe χ2 test and analysis of variance were used to compare the differences in patients’ characteristics, outcomes, and ECMO-related complications among the various volume groups. The Cox proportional model and its hazard ratio (HR) were used to analyze the dose–response relationship between high- and low-volume groups. Moreover, locally estimated scatterplot smoothing (LOESS) regression was used to model the nonlinear relationship for the annual ECMO volume and mortality rate among study hospitals.19,20 A two-sided P value <0.05 was considered statistically significant. All statistical analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC, USA).
From January 1, 2001 to December 31, 2017, a total of 11,734 patients aged >20 years received their first-ever ECMO in Taiwan. The mean age of the overall cohort was 57.4 years. The preceding indication of ECMO was postcardiotomy cardiogenic shock in 43.7%; cardiogenic shock, myocarditis, or acute myocardial infarction in 26.2%; and respiratory indication in 20.4%. The overall in-hospital mortality rate was 65.5% and 1-year mortality rate was 70.6%.
The 101 hospitals were divided into 4 groups according to their annual volumes. The numbers of hospitals in the high-volume group, volume 2, volume 3, and low-volume group were 4, 11, 12, and 74, respectively. No significant difference between patients was observed among the 4 groups in terms of demographics, patient economic status, hospital urbanization level, and clinical and pre-ECMO characteristics (Table 1). In the high-volume group, all 3,351 patients (100%) received ECMO in medical centers, whereas in the low-volume group, 2,024 patients (84.47%) received ECMO in regional hospitals. There was a higher number of patients with postcardiotomy cardiogenic shock in the high-volume group, and a higher number of patients with respiratory failure in the low-volume group.
| Variable | Volume 1 | Volume 2 | Volume 3 | Volume 4 | P value |
|---|---|---|---|---|---|
| Annual >40 (n=3,551) |
Annual 20–40 (n=3,572) |
Annual 10–20 (n=2,212) |
Annual <10 (n=2,399) |
||
| ECMO indication | <0.0001 | ||||
| Postcardiotomy | 1,798 (50.63) | 1,532 (42.89) | 939 (42.45) | 853 (35.56) | |
| Cardiogenic shock, myocarditis or AMI | 748 (21.06) | 1,064 (29.79) | 572 (25.86) | 687 (28.64) | |
| Respiratory | 645 (18.16) | 607 (16.99) | 534 (24.14) | 607 (25.3) | |
| Trauma | 114 (3.21) | 147 (4.12) | 56 (2.53) | 138 (5.75) | |
| Other | 246 (6.93) | 222 (6.22) | 111 (5.02) | 114 (4.75) | |
| ECMO duration (days) | 6.05±8.98 | 5.07±4.47 | 5.15±4.44 | 5.08±4.72 | <0.001 |
| Hospital level | <0.0001 | ||||
| Medical center | 3,551 (100) | 2,707 (75.78) | 1,157 (52.31) | 372 (15.53) | |
| Regional/district hospital | 0 (0) | 865 (24.22) | 1,055 (47.69) | 2,024 (84.47) | |
| Urbanization level | <0.0001 | ||||
| Urban | 15 (0.45) | 30 (0.92) | 6 (0.3) | 19 (0.91) | |
| Suburban | 2,337 (70.16) | 2,124 (65.43) | 1,396 (68.8) | 1,236 (58.97) | |
| Rural | 979 (29.39) | 1,092 (33.64) | 627 (30.9) | 841 (40.12) | |
| Surgical year | <0.0001 | ||||
| 2000–2005 | 347 (9.77) | 160 (4.48) | 92 (4.16) | 28 (1.17) | |
| 2006–2009 | 719 (20.25) | 693 (19.4) | 394 (17.81) | 311 (12.96) | |
| 2010–2013 | 1,183 (33.31) | 1,261 (35.3) | 794 (35.9) | 791 (32.97) | |
| 2014–2017 | 1,302 (36.67) | 1,458 (40.82) | 932 (42.13) | 1,269 (52.9) | |
| Age (years) | 56.65±16.04 | 57.74±15.86 | 57.98±15.69 | 57.41±15.39 | 0.005 |
| Monthly income, USD | <0.0001 | ||||
| 0–596 | 1,129 (31.98) | 988 (28.01) | 581 (26.42) | 675 (28.41) | |
| 610–760 | 1,029 (29.15) | 1,244 (35.27) | 791 (35.97) | 861 (36.24) | |
| >800 | 1,372 (38.87) | 1,295 (36.72) | 827 (37.61) | 840 (35.35) | |
| Comorbidities | |||||
| Hypertension | 1,691 (47.62) | 1,777 (49.75) | 1,123 (50.77) | 1,202 (50.1) | 0.079 |
| Diabetes mellitus | 1,131 (31.85) | 1,153 (32.28) | 738 (33.36) | 771 (32.14) | 0.683 |
| Heart failure | 763 (21.49) | 656 (18.37) | 450 (20.34) | 317 (13.21) | <0.0001 |
| Prior myocardial infarction | 422 (11.88) | 426 (11.93) | 260 (11.75) | 206 (8.59) | <0.001 |
| Peripheral arterial disease | 455 (12.81) | 474 (13.27) | 282 (12.75) | 281 (11.71) | 0.363 |
| Prior stroke | 307 (8.65) | 362 (10.13) | 207 (9.36) | 250 (10.42) | 0.075 |
| Chronic kidney disease | 1,279 (36.02) | 1,299 (36.37) | 805 (36.39) | 794 (33.1) | 0.039 |
| Liver cirrhosis | 125 (3.52) | 89 (2.49) | 48 (2.17) | 57 (2.38) | 0.005 |
| Coagulopathy | 129 (3.63) | 133 (3.72) | 65 (2.94) | 68 (2.83) | 0.138 |
| COPD | 327 (9.21) | 317 (8.87) | 189 (8.54) | 189 (7.88) | 0.335 |
| Depression | 148 (4.17) | 128 (3.58) | 92 (4.16) | 106 (4.42) | 0.386 |
| CCI score | 2.74±2.32 | 2.64±2.29 | 2.47±2.09 | 2.38±2.17 | <0.001 |
| Pre-OP anti-HTN medication | |||||
| ACEI/ARB | 1,231 (34.67) | 1,300 (36.39) | 820 (37.07) | 839 (34.97) | 0.190 |
| β-blocker | 1,100 (30.98) | 1,045 (29.26) | 683 (30.88) | 642 (26.76) | 0.002 |
| CCB | 780 (21.97) | 815 (22.82) | 542 (24.5) | 576 (24.01) | 0.098 |
| α-blocker | 121 (3.41) | 113 (3.16) | 71 (3.21) | 58 (2.42) | 0.173 |
| Thiazide | 130 (3.66) | 106 (2.97) | 69 (3.12) | 67 (2.79) | 0.222 |
| Loop diuretics | 727 (20.47) | 645 (18.06) | 442 (19.98) | 343 (14.3) | <0.0001 |
| Potassium-sparing diuretics | 385 (10.84) | 279 (7.81) | 184 (8.32) | 129 (5.38) | <0.0001 |
| Vasodilator | 849 (23.91) | 801 (22.42) | 527 (23.82) | 457 (19.05) | <0.0001 |
| Nitrate | 677 (19.07) | 619 (17.33) | 387 (17.5) | 337 (14.05) | <0.0001 |
| Number of anti-HTN drugs | 1.69±1.85 | 1.60±1.77 | 1.68±1.80 | 1.44±1.67 | <0.001 |
| Pre-OP other medication | |||||
| Statin | 625 (17.6) | 677 (18.95) | 433 (19.58) | 480 (20.01) | 0.090 |
| Antiplatelet | 793 (22.33) | 834 (23.35) | 536 (24.23) | 481 (20.05) | 0.004 |
| Warfarin | 235 (6.62) | 187 (5.24) | 106 (4.79) | 86 (3.58) | <0.0001 |
| NOACs | 8 (0.23) | 11 (0.31) | 2 (0.09) | 3 (0.13) | 0.272 |
| OHA | 651 (18.33) | 697 (19.51) | 469 (21.2) | 474 (19.76) | 0.064 |
| Insulin | 219 (6.17) | 251 (7.03) | 144 (6.51) | 157 (6.54) | 0.538 |
| Digoxin | 248 (6.98) | 188 (5.26) | 128 (5.79) | 80 (3.33) | <0.0001 |
| Medical resources utilization | |||||
| Outpatient | 24.29±19.83 | 22.23±18.66 | 23.37±19.83 | 23.99±21.43 | <0.0001 |
| Emergency | 2.46±2.85 | 2.21±2.10 | 2.17±2.06 | 2.18±2.90 | <0.0001 |
| Inpatient | 2.80±2.31 | 2.33±1.93 | 2.22±1.88 | 1.91±1.52 | <0.0001 |
Values are given as n (%) or mean±standard deviation. ACEI/ARB, angiotensin-converting enzyme inhibitor/angiotensin-receptor blocker; AMI, acute myocardial infarction; CCB, calcium channel blocker; CCI, Charlson’s Comorbidity Index; COPD, chronic obstructive pulmonary disease; ECMO, extracorporeal membrane oxygenation; HTN, hypertension; NOAC, non-vitamin K antagonist oral anticoagulant; OHA, oral hypoglycemic agent; OP, operation; USD, US dollar.
From the early period (2000–2005) to the late period (2014–2017) of this study, ECMO use in the low-volume, volume 3, volume 2, and high-volume hospitals increased by 4432.1% (from 28 to 1,269 cases), 913% (from 92 to 932 cases), 811.3% (from 160 to 1,458 cases), and 275.2% (from 347 to 1,302 cases), respectively (Figure 2). With the increase in ECMO volume in Taiwan year by year, the procedure was no longer concentrated in high-volume hospitals. The low-volume hospitals comprised 52.9% of total ECMO volume in the late period (2014–2017).
Distribution of the hospital volume groups in Taiwan during the study period (2001–2017).
The in-hospital mortality rate was significantly higher in the low-volume hospitals (volume 4, 67.1%) than in volume 3 (66.9%), volume 2 (65.9%), and high-volume (volume1, 63.3%) hospitals (Table 2). Furthermore, this trend of the relationship was noted for in-hospital mortality. Patients in high-volume hospitals have the longest stays in intensive care unit and hospital, with prolonged ventilator support. ECMO-related complications were higher in the high-volume group, including new-onset stroke, new-onset dialysis, and more blood product transfusions, than in the low-volume group.
| Variable | Volume 1 | Volume 2 | Volume 3 | Volume 4 | P value |
|---|---|---|---|---|---|
| Annual >40 (n=3,551) |
Annual 20–40 (n=3,572) |
Annual 10–20 (n=2,212) |
Annual <10 (n=2,399) |
||
| In-hospital death | 2,246 (63.25) | 2,353 (65.87) | 1,479 (66.86) | 1,609 (67.07) | 0.006 |
| New-onset stroke | 184 (5.18) | 185 (5.18) | 74 (3.35) | 57 (2.38) | <0.0001 |
| New-onset ischemic stroke | 132 (3.72) | 122 (3.42) | 47 (2.12) | 37 (1.54) | <0.0001 |
| New-onset hemorrhagic stroke | 69 (1.94) | 76 (2.13) | 29 (1.31) | 27 (1.13) | 0.008 |
| Fasciotomy or amputation | 81 (2.28) | 63 (1.76) | 32 (1.45) | 52 (2.17) | 0.101 |
| Prolonged ventilation (≥7 days) | 3,486 (98.17) | 3,484 (97.54) | 2,190 (99.01) | 2,350 (97.96) | 0.001 |
| New-onset dialysis | 1,733 (48.8) | 1,727 (48.35) | 1,017 (45.98) | 986 (41.1) | <0.0001 |
| Deep wound infection | 744 (20.95) | 733 (20.52) | 591 (26.72) | 597 (24.89) | <0.0001 |
| PRBC (U) | 11.24±11.02 | 13.50±12.11 | 11.39±11.22 | 9.75±9.60 | <0.0001 |
| FFP (U) | 11.12±12.19 | 12.07±12.18 | 11.19±12.33 | 8.86±9.35 | <0.0001 |
| Platelets (U) | 9.34±11.97 | 7.51±10.06 | 8.21±10.97 | 8.33±10.43 | <0.0001 |
| Ventilator (days) | 12.20±11.08 | 9.26±9.24 | 10.20±9.10 | 9.66±9.28 | <0.0001 |
| ICU duration (days) | 12.41±10.00 | 11.55±9.88 | 11.09±9.23 | 9.40±8.94 | <0.0001 |
| Hospital stay (days) | 34.69±39.30 | 25.97±31.15 | 25.50±31.47 | 19.70±26.18 | <0.0001 |
Values are given as n (%) or mean±standard deviation. ECMO, extracorporeal membrane oxygenation; FFP, fresh frozen plasma; ICU, intensive care unit; PRBC, packed red blood cells.
The patients who received ECMO at high-volume institutions had a significantly lower 1-year mortality rate than those who received ECMO at low-volume institutions (67.5% vs. 73.4%, P<0.001; Table 3). In the subgroup analysis, patients in the low-volume hospitals had an increased risk of 1-year death than those in the high-volume hospitals (HR: 1.11, confidence interval: 1.04–1.18, P=0.002) (Figure 3). Although the patients in high-volume hospitals had a lower all-cause mortality rate than those in low-volume hospitals, there was no significant difference (74.2% vs. 76.9%, P=0.06). The patients in the all-cause death group were followed up for a mean period of 1.3±2.68 years (range: 0–16.9 years). Long-term complications were observed in the high- and low-volume groups, such as revascularization, acute myocardial infarction, ischemic stroke, hemorrhage stroke, heart failure, and respiratory failure. The number of patients requiring permanent dialysis due to endstage renal disease (ESRD) were even lower in the high-volume group.
| Variable | Volume 1 | Volume 2 | Volume 3 | Volume 4 | P value |
|---|---|---|---|---|---|
| Annual >40 (n=3,551) |
Annual 20–40 (n=3,572) |
Annual 10–20 (n=2,212) |
Annual <10 (n=2,399) |
||
| All-cause deaths | 2,635 (74.2) | 2,717 (76.06) | 1,694 (76.58) | 1,845 (76.91) | 0.060 |
| 1-year deaths | 2,397 (67.5) | 2,531 (70.86) | 1,591 (71.93) | 1,761 (73.41) | <0.0001 |
| Surviving to discharge | 1,305 (36.75) | 1,219 (34.13) | 733 (33.14) | 790 (32.93) | 0.006 |
| All-cause deaths after discharge | 389 (29.81) | 364 (29.86) | 215 (29.33) | 236 (29.87) | 0.995 |
| CV death | 143 (10.96) | 145 (11.89) | 100 (13.64) | 88 (11.14) | 0.305 |
| Revascularization (PCI or CABG) | 21 (1.61) | 29 (2.38) | 9 (1.23) | 14 (1.77) | 0.269 |
| Acute myocardial infarction | 26 (1.99) | 37 (3.04) | 14 (1.91) | 24 (3.04) | 0.187 |
| Ischemic stroke | 37 (2.84) | 24 (1.97) | 15 (2.05) | 15 (1.9) | 0.383 |
| Hemorrhagic stroke | 18 (1.38) | 12 (0.98) | 11 (1.5) | 9 (1.14) | 0.721 |
| All-cause readmission | 892 (68.35) | 773 (63.41) | 452 (61.66) | 460 (58.23) | <0.0001 |
| Admission for heart failure | 111 (8.51) | 104 (8.53) | 60 (8.19) | 63 (7.97) | 0.966 |
| Respiratory failure | 140 (10.73) | 133 (10.91) | 73 (9.96) | 99 (12.53) | 0.421 |
| ESRD requiring permanent dialysis* | 5 (0.56) | 17 (2.01) | 19 (3.57) | 51 (8.7) | <0.0001 |
Values are given as n (%). *Patients who received hemodialysis in this admission were excluded. CABG, coronary artery bypass graft; CV, cardiovascular; ESRD, endstage renal disease; PCI, percutaneous coronary intervention.
Association of the hospital’s annual ECMO volume with 1-year mortality risk. This hazard ratio (HR) analysis used the highest volume as the reference for comparison with other volume groups. CI, confidence interval; ECMO, extracorporeal membrane oxygenation.
The 101 institutions included in this cohort study had 1-year mortality rates ranging from 0% to 100%. However, high variability of the 1-year mortality rate was observed among patients at low-volume hospitals. The variability converged and corresponded with the increase in the annual volume of the hospitals. Higher volume corresponded to a lower 1-year mortality rate (Figure 4).
Scatter plot and regression relationship between annual ECMO volume and 1-year mortality rate. ECMO, extracorporeal membrane oxygenation.
The number of ECMO procedures has been rapidly increasing worldwide, and in in 2014 the ELSO Registry database revealed a growth in the need for ECMO support.9 In Taiwan, the number of patients who have received ECMO has grown exponentially in the past 2 decades. The first reason that the NHI scheme has been covering ECMO since 2002.18 The second reason is that the indications of reimbursement were widely expanded in 2009. Additionally, the availability of modern ECMO equipment, advanced implantation techniques, and high-quality critical care systems has contributed to the increase in ECMO use in Taiwan. Moreover, because of the decreased technical threshold, the procedure is no longer performed only in medical centers but also in regional hospitals. This phenomenon has rapidly popularized ECMO. According to the Taiwan NHI database, ECMO use increased >23-fold from 50 cases in 2001 to 1,176 cases in 2017. In the 4-year interval from 2014 to 2017, the number of patients who received their first-ever ECMO procedure in Taiwan was 4,961. The total population of Taiwan was 23.5 million on average. ECMO support prevalence was 5.27 per 100,000 people per year, which was very high compared with other countries, such as Korea (2.35 per 100,000 people per year), Japan (2.54 per 100,000 people per year) and Germany (5.21 per 100,000 people per year).13,21,22
However, the increase in the number of ECMO procedures has not been reflected in a survival outcome benefit.9–14 Overall, the in-hospital mortality rate in the past 2 decades in Taiwan has been 65.5%, similar to that in Korea (63.4%) and Germany (68.1%).13,14 In our study, the short-term outcomes were found to be superior in high-volume institutions than in low-volume institutions for either in-hospital mortality (63.3% vs. 67.0%, P=0.006) or 30-day mortality (64.9% vs. 69.6%, P<0.0001). Similarly, Barbaro et al analyzed the ELSO Registry database and found that higher annual hospital ECMO volume was associated with lower in-hospital mortality in adults.9 The same conclusion was also obtained in an analysis of a Korea nationwide database by Tay et al.13 Furthermore, Becher et al revealed similar findings from a nationwide database study in Germany.14
In the mid-term follow-up, we found that the 1-year mortality rate was lower in the high-volume hospitals than in the low-volume hospitals (67.5% vs. 73.4%, P<0.0001). Owing to the expertise of multidisciplinary support teams, better outcomes were observed among patients who underwent ECMO at high-volume hospitals. Nevertheless, the distribution of indications for ECMO was different for different hospital levels, which influenced the outcomes of the patients. Our previous study17 revealed that the in-hospital mortality rate among postcardiotomy patients with ECMO support was 61.7% between 2001 and 2011 per the Taiwan NHIRD. The short-term outcomes among in-hospital patients were better than those among the cohort in this study (in-hospital mortality rate: 65.5%). Moreover, the postcardiotomy patients were almost all concentrated in high-volume hospitals. Therefore, the heterogeneity of the population also contributed to the results.
However, no significant differences were observed between the 4 all-cause death groups. In the present analysis, the all-cause mortality rate was mostly caused by the 1-year deaths in these groups. In our previous study, we found that deaths were almost all concentrated in the first year of follow-up.17 No significant differences in all-cause death, readmission for any cause, and other long-term outcomes were observed between patients with and without ECMO support after cardiotomy. Most patients exhibited similar conditions throughout the 1st year of follow-ups. Therefore, the same results were obtained between the groups after discharge throughout long-term follow-up, and the differences between groups were not significant. The mortality rate after discharge and the long-term complications and other sequelae among patients at high-volume hospitals were similar to those at low-volume hospitals. The incidence of revascularization, acute myocardial infarction, ischemic stroke, hemorrhage stroke, and admission for heart failure and respiratory failure did not differ between the high- and low-volume hospitals.
To the best of our knowledge, this large-scale, nationwide database study is the first to reveal the relationship between hospital volume and both in-hospital and mid-term outcomes after discharge in patients who received ECMO support. Previous studies have mostly focused on in-hospital outcomes. Furthermore, because the NHI program reimburses almost all the expenditure for ECMO and medical disbursement in Taiwan, the bias of economic and social factors was reduced. Therefore, this nonselected nationwide population cohort study provides sufficient real-world data of the strong positive relationship between hospital volume and clinical outcome in patients who received ECMO support.
Study LimitationsThis study has some limitations on account of the nationwide database analysis design. First, certain clinical and procedural data were unavailable, such as laboratory information, echocardiography information, heart failure function classification, and ECMO mode (venoarterial or venovenous). Second, complication-related diagnosis and ECMO indications were based on ICD-9-CM codes and coding errors might have occurred. However, strict regulations regarding the ECMO procedure, catastrophic illness (e.g., stroke or ESRD), and medication covered by the NHI scheme result in claim expenditure only for those patients with a precise diagnosis. Therefore, the effect of the coding errors of diagnosis on our conclusion is limited. Third, although a large-scale statistical analysis minimized the bias in this retrospective study, unknown confounders and unmeasured factors will still occur. Further prospective clinical trials are required to confirm our findings. Finally, the different indications of ECMO influenced the outcomes among the volume groups. The heterogeneity of the patient population might have influenced the mortality rate in different volume groups, thereby limiting the findings of this study. Moreover, the division of the annual volume was arbitrary in this study. The hospital volume and medical care system differ in every country and region. We could only provide the experience in Taiwan and could not deduce a generalized result. Notwithstanding these limitations, we believe this study revealed the relationship between hospital volume and patient outcomes based on real-world data.
The ECMO is popularly used in the critical care system worldwide. Our nationwide population-based study revealed a strong positive relationship between the hospital annual volume and survival benefit, including short-term and mid-term outcomes. Medical system policies must be considered for the allocation of ECMO services. Moreover, the establishment of a rapid and efficient referral system across the country might improve the overall outcomes of these critical patients.
This study was based on data from the Taiwan National Health Insurance Research Database provided by the National Health Insurance Administration (Ministry of Health and Welfare of Taiwan) managed by the National Health Research Institutes of Taiwan. However, the interpretation and conclusions are those of the authors. The authors thank the Maintenance Project of the Center for Big Data Analytics and Statistics (Grant CLRPG3D0049) at Chang Gung Memorial Hospital for statistical consultation and data analysis. This work was supported by a grant from Chang Gung Memorial Hospital, Taiwan CFRPG3K0011, BMRPD95 (S.-W.C.). This work was also supported by the Ministry of Science and Technology grant MOST-108-2314-B-182A-141 (S.-W.C.).
No conflicts of interest for all authors.
This study received the approval of the Institutional Review Board (IRB) of Chang Gung Memorial Hospital (approval no. 201701250B0D001).
Please find supplementary file(s);
http://dx.doi.org/10.1253/circj.CJ-22-0107
