Article Text
PDF
Abstract
Objective: The formerly observed volume–outcome relation for percutaneous coronary interventions (PCIs) has recently been questioned.
Design: We analysed data of the PCI registry of the Arbeitsgemeinschaft Leitende Kardiologische Krankenhausärzte.
Patients: In 2003 a total of 27 965 patients at 67 hospitals were included.
Results: The median PCI volume per hospital was 327. In-hospital mortality was 1.85% in hospitals belonging to the lowest PCI volume quartile and 1.21% in the highest quartile (p for trend <0.001). Two groups of patients were then compared according to their treatment at hospitals with either <325 PCIs (n = 5754) or >325 PCIs (n = 22 211) per year. Logistic regression analysis showed that a PCI performed at hospitals with a volume of >325 PCI/year was independently associated with a lower hospital mortality (OR = 0.67, 95% CI: 0.52 to 0.87; p = 0.002). If PCI was performed in patients with acute myocardial infarction there was a significant decline in mortality with increasing volume (p for trend = 0.004); however, there was no association in patients without a myocardial infarction.
Conclusions: This analysis of contemporary PCI in clinical practice shows a small but significant volume–outcome relation for in-hospital mortality. However, this relation was only apparent in high-risk subgroups, such as patients presenting with acute myocardial infarction.
- coronary angioplasty
- risk
- volume–outcome relation
Statistics from Altmetric.com
It is believed that technically challenging manual procedures will result in a better outcome if performed by experienced investigators at high-volume institutions.
This has been shown for many surgical procedures, such as coronary bypass surgery, pancreatic and thoracic surgery.1–4 Some of these analyses could also show that the volume–outcome relation is especially apparent for high-risk surgery and not evident in low-risk surgery.
In the case of percutaneous coronary interventions (PCI), early reports showed a statistically significant but relatively weak volume–outcome relation.5–11 This led to recommendations of different cardiac societies to perform PCIs only at institutions with a minimum volume of PCIs per year and also only by interventionalists with a defined minimum of PCIs per year.12–14
With the advent of routine coronary stenting, more recent studies reported no clinically relevant volume–outcome relations, thus challenging the previous recommendations of minimal procedural volumes.15 16
However, two recently published papers, reflecting actual clinical practice of contemporary PCI, reinforced the volume–outcome relation, with the limitation that it could only be observed in high-risk patients.17 18
To determine the influence of hospital PCI volume on the clinical outcome of PCI in current clinical practice, we analysed the data of registry of percutaneous coronary interventions of the Arbeitsgemeinschaft Leitende Kardiologische Krankenhausärzte (ALKK PCI registry).
METHODS
The PCI registry of the Arbeitsgemeinschaft Leitende Kardiologische Krankenhausärzte (ALKK, Stuttgart, Germany) contains all consecutive PCI procedures from an association of non-university centres in Germany since October 1992. Details of the organisation of the registry have been published previously.19 20 In brief: all interventions at the participating hospitals were prospectively enrolled on an intention-to-treat basis in the registry by telephone or fax to the coordinating centre in Kassel and since 2003 in Ludwigshafen. All complications occurring in the catheter room and throughout the hospital stay were prospectively documented.
For this analysis all patients from the year 2003 were analysed.
Definitions
Angioplasty was performed according to the standard protocol of each centre. The use of stents and other devices was left to the discretion of the treating physician. Success of primary angioplasty in patients presenting with ST segment elevation myocardial infarction (STEMI) was evaluated by the physician performing the intervention using the TIMI flow grade.21
End point
End point was in-hospital mortality. Death was defined as death from any cause after the PCI procedure.
Data acquisition
Data collection was performed with paper-based case record forms, electronic case record forms or directly via the internet to the data coordinating centre (Karl Ludwig Neuhaus data centre of the ALKK, Ludwigshafen). The data were checked for completeness and consistency and queries for the centres were generated.
Statistical analysis
The patient population is described by absolute numbers and percentages with respect to categorical variables, and by medians with quartiles or mean with standard deviation with respect to continuous variables. Interventional characteristics are reported for the first procedure performed in each patient during his or her hospital stay. The distribution of dichotomous variables was compared between groups by Pearson χ2, of ordered categorical variables by Cochran–Armitage test and of continuous variables by Mann–Whitney U test.
In order to look for a trend in the rates of in-hospital mortality, the hospitals were grouped into quartiles according to their annual PCI volume, and the Cochran–Armitage test for trend was performed. The same was done with quartiles according to annual volume of PCI for acute STEMI patients only. A binary classification was made into centres with PCI volume below the median (<325 PCIs/year) and above the median (>325).
As volume–outcome relation may be more pronounced in high-risk patients, centres were also compared exclusively for patients presenting with STEMI or NSTEMI, and further exploratory analyses were done for subgroups defined by various risk factors. In these comparisons, both the risk ratio and the risk difference were computed as effect measures.
Multiple logistic regression analysis was used to adjust for baseline factors influencing in-hospital mortality. The effect estimates in the models are shown as odds ratios with 95% confidence intervals. The following variables were examined for inclusion into the model: age, gender, intervention for STEMI/NSTEMI, cardiogenic shock, acute heart failure on admission, 3-vessel or left main stem disease, renal insufficiency, diabetes mellitus, prior PCI and prior CABG. The final models were selected using stepwise procedures with significance levels of 0.05 for entry and 0.1 for removal; gender and diabetes were included in all models for clinical reasons. Hospital PCI volume was entered into the models dichotomised as <325 versus >325 PCIs/year, and alternatively as a continuous variable. A further logistic regression analysis was done for factors influencing hospital mortality in STEMI or NSTEMI patients only. As the outcomes of patients within the same hospital may not be independent, additional models were fitted that allowed for clustering of patients by using generalised estimating equations (GEE) with exchangeable working correlation structure.
The discrimination of the models was assessed by the C statistic. p Values <0.05 were considered significant. All p values are results of two-tailed tests. The calculations were performed using SAS©, version 9.1 (Cary, North Carolina).
RESULTS
In 2003 a total of 27 965 patients at 67 hospitals were included in the ALKK PCI registry. The median hospital stay was 4 days (quartiles: 2–9). The median PCI volume per hospitals was 327 (quartiles: 196–521). The distribution of the PCI volume at the participating hospitals is shown in fig 1.
Hospital mortality and PCI volume
In-hospital mortality for all PCI patients was 1.41%. It was 1.85% in hospitals belonging to the lowest PCI volume quartile and 1.21% in the highest quartile (p for trend <0.001) (fig 2).
There was a decline in mortality between the second (upper boundary 323 PCIs/year) and third quartiles and a further decline in the fourth quartile, whereas no difference was observed in in-hospital mortality between the first and second quartiles. Therefore we chose to compare two groups of patients according to their treatment at hospitals with either <325 PCIs (n = 5754) or >325 PCIs (n = 22 211) per year.
As shown in table 1, patients treated at hospitals with an annual PCI volume <325 were less often male (71.5% versus 73.3%, p = 0.006), less often presented with heart failure (3.4% versus 8.7%, p<0.001) and had less frequently a history of percutaneous coronary intervention (29.0% versus 32.9%, p<0.001) and prior coronary artery bypass grafting (7.5% versus 11.6%, p<0.001). They were more likely, however, to be treated for ST-elevation myocardial infarction (23.5% versus 22.3%, p = 0.053) and for non ST-elevation myocardial infarction (17.1% versus 14.6%, p<0.001) and less likely for stable angina pectoris (39.4% versus 45.3%, p<0.001).
Concerning angiographic characteristics, patients treated at hospitals with <325 PCIs/year more often had a normal left ventricular function (68.4% versus 66.2%, p = 0.006) and less often 3-vessel disease (27.2% versus 31.0%, p<0.001) (table 2). A final TIMI flow of less than grade 3 in ST-elevation myocardial infarction patients was observed more often (9.2% versus 8.2%, p = 0.017), total fluoroscopy time was longer (7.0 minutes median versus 6.0 minutes median, p<0.001) and the amount of contrast dye used higher (110 ml median versus 100 ml median, p<0.001).
These patients, however, were more often treated with unfractionated heparin, low molecular weight heparin, clopidogrel or ticlopidine as well as glycoprotein IIb/IIIa antagonists during the intervention (all p values <0.001) (table 3).
After adjustment for baseline factors, logistic regression analysis showed that a PCI performed at hospitals with a volume of >325 PCI/year was independently associated with a lower hospital mortality (OR = 0.67, 95% CI: 0.52 to 0.87; p = 0.002) (table 4). The effect of hospital volume was also present when included as a continuous variable (p = 0.004) and remained significant in the GEE models accounting for hospital clusters (p = 0.040 for the binary and p = 0.034 for the continuous variable). We also did analyses using a cut-off of 200 PCIs/year and 400 PCIs/year. The corresponding results for hospital mortality are: <200 PCI/year: 1.74% versus >200 PCI/year: 1.39%, OR = 1.25, 95% CI: 0.84 to 1.87, p value: 0.269; <400 PCI/year: 1.67% versus >400 PCI/year: 1.31%, OR = 1.28, 95%CI: 1.04 to 1.58, p value: 0.022.
In-hospital mortality and PCI volume in different subgroups of patients
If the PCI was performed in patients without ST-elevation or non ST-elevation myocardial infarction, no association of the hospital PCI volume and hospital outcome was observed (p for trend between the quartiles = 0.84, fig 3). In STEMI or NSTEMI patients, however, there was a decline in mortality with increasing volume (p for trend 0.004), with the biggest decrease between the second and third quartiles (fig 4). After adjustment for other factors, logistic regression analysis showed that in patients presenting with STEMI or NSTEMI a PCI performed at hospitals with a volume of >325 PCI/year was independently associated with a lower in-hospital mortality (OR = 0.65, 95% CI: 0.49 to 0.86; p = 0.002) (table 5). For the continuous PCI volume, the effect was even more pronounced than in the total registry (p<0.001 in the independence model and in the GEE model).
The volume–outcome relation was maintained in acute STEMI patients, with in-hospital mortality of 7.77% in the lowest quartile, 6.27% in the second quartile, 4.86% in the third quartile and 4.84% in the highest quartile; p for trend = 0.04. A decreasing mortality rate was also observed in relation to the STEMI volume of the different hospitals: 7.45% mortality in the lowest quartile versus 4.98% in the highest quartile.
Further comparisons of different subgroups according to their treatment at hospitals with either <325 PCIs or >325 PCIs per year were performed (table 6). With increasing mortality risk of a given subgroup, the absolute risk reduction in favour of high-volume hospitals increased, with no, or not relevant, volume–outcome relation in low-risk subgroups.
DISCUSSION
The recent PCI guidelines of the American Heart Association and American College of Cardiology (AHA/ACC)13 recommend a minimum PCI volume for hospitals per year of 200 and a personal investigator volume of 75 PCIs per year. The European PCI guidelines22 avoid giving recommendations on operator or hospital minimum volumes. The AHA/ACC guidelines also include recommendations on minimum investigator (11/year) and hospital (36/year) volumes for primary PCIs.13
Most of those recommendations are based on analyses of volume–outcome relations based on data from PCI results of 1990–1997.5–11 These data demonstrated lower complication rates of PCIs performed at high volume centres by experienced physicians. These results, however, have been brought into question by more recent studies reflecting the introduction of stents and other technical improvements reducing the complication rate of PCI. Epstein et al15 conducted a retrospective analysis of the Agency for Healthcare Research and Quality’s Nationwide In-patient Sample hospital discharge database to evaluate in-hospital mortality among patients (n = 362 748) who underwent PCI between 1998 and 2000 at low (5 to 199 cases/year), medium (200 to 399 cases/year), high (400 to 999 cases/year), and very high (1000 cases or more/year) PCI volume hospitals. Crude in-hospital mortality rates were 2.56% in low-volume hospitals, 1.83% in medium-volume hospitals, 1.64% in high-volume hospitals, and 1.36% in very high-volume hospitals (p<0.001 for trend). Compared with patients treated in high-volume hospitals, patients treated in low-volume hospitals remained at increased risk for mortality after adjustment for patient characteristics (OR 1.21, 95% CI: 1.06 to 1.28). However, patients treated in medium-volume hospitals (OR 1.02, 95% CI: 0.92 to 1.14) and patients treated in very high-volume hospitals (OR 0.94, 95% CI 0.85 to 1.03) had a similar risk of mortality.
A study of Burton et al16 on 17 417 PCIs performed in Scotland between 1997 and 2003 found no influence of the annual PCI volume of the participating hospitals on 30-day death rate. However, increasing PCI volumes were associated with lower major adverse cardiovascular event (MACE) rates.
As a consequence, a Task Force of the German Society of Cardiology on quality assurance in invasive cardiology concluded that only a weak volume–outcome relation exists for contemporary PCI, and that there is no specific threshold to be recommended as minimum volume.14 However, they continued to give minimum volume recommendations mainly based on the common-sense view that experience is good for potentially harmful investigations. A similar statement is given in a recent publication of the German Institut für Qualität und Wirtschaftlichkeit im Gesundheitswesen (IQWiG)23 on volume–outcome relations for PCI. The authors also state, however, that in the case of PCI for acute myocardial infarction hospital mortality is lower when performed at high-volume hospitals.
In our analysis of a total of 27 965 PCIs performed at 67 hospitals in 2003 the median PCI volume per hospital was 327 (quartiles: 196–521). In-hospital mortality was 1.85% in hospitals belonging to the lowest PCI volume quartile and 1.21% in hospitals with the highest volumes (p for trend <0.001) with a marked decline in mortality between the second (upper boundary 323 PCIs/year) and third quartiles. After adjustment for other factors, logistic regression analysis showed that a higher annual PCI volume of hospitals was independently associated with lower hospital mortality. As the cut-off point used in the comparisons is in accordance with the observed data, the effect estimates may be too optimistic. Nevertheless, the results were also confirmed for PCI volume in quartiles and as a continuous variable, and support at least a general trend. Our analysis is not meant to suggest a certain threshold value.
The observed outcome difference of the total group, however, was driven by the high-risk subgroups; no significant association of the hospital PCI volume and in-hospital outcome was observed in patients without myocardial infarction. In STEMI or NSTEMI patients, however, there was a decline in mortality with increasing volume. After adjustment for other factors logistic regression analysis showed that a PCI performed at hospitals with a volume of >325 PCI/year was independently associated with a lower hospital mortality (OR = 0.68; 95% CI: 0.51 to 0.89; p = 0.006). Further analysis of subgroups divided by hospital volume above or below 325 PCIs per year showed that with increasing risk of a given subgroup, such as the presence of heart failure, 3-vessel disease or diabetes, the absolute risk reduction in favour of high-volume hospitals increased, with no significant volume–outcome relation in low-risk subgroups.
In our registry many data on intervention details are collected, which could help to explain the differences between low-volume and high-volume centres; a final TIMI flow of less than grade 3 in ST-elevation myocardial infarction patients was observed non-significantly more often (13.2% versus 12.5%, p = 0.490), reflecting a lower interventional success rate. However, the difference in final TIMI grade 3 flow is small and hardly explains completely the observed difference in in-hospital mortality in myocardial infarction patients. Further reasons for this difference might be that at low-volume hospitals more contrast dye was used (110 ml versus 100 ml median, p<0.001) and total fluoroscopy time was longer (7.0 minutes versus 6.0 minutes median, p<0.001). This at least indicates a longer duration of the PCI procedure giving the chance of more complications or being already the result of procedural complications.
Hospital stay was 5 days median in the first two quartiles compared with 4 days in quartiles 3 and 4 (p<0.001). This difference could have influenced the differences in hospital mortality, because high-volume hospitals could have transferred patients earlier to smaller spoke hospitals, thus giving the chance to miss complications. We do not have the information on which patients were transferred; however, the difference in in-hospital mortality was already apparent in the first days after the PCI. On the day of admission mortality was 0.34% in low-volume versus 0.25% in high-volume centres, in the patients discharged after the first day mortality was 0.27% versus 0.21% on day 2, and in the patients discharged after day 2 mortality was 1.26% versus 0.96%.
The lower mortality at high-volume hospitals may be due to multiple factors, such as better technical facilities (eg the availability of intra-aortic balloon pumping, on-site availability of cardiac surgery), which are associated with higher-volume centres, better intensive care facilities and so on. So the better results at high-volume hospitals may also be an indicator for better medical structures at larger hospitals in general.
Our results are well in line with those recently published by a French group.18 At 44 centres a total of 37 848 PCIs were performed between 2001 and 2002. Hospitals performing <400 PCIs/year were classified as low-volume hospitals. In emergency procedures in-hospital mortality was lower in the high-volume hospitals (6.75% versus 8.54%, p = 0.028), but no difference was observed in elective procedures (0.62% versus 0.62%, p = 0.99). In contrast to the French study, we evaluated the influence of risk factors, such as diabetes and renal insufficiency, as well as interventional characteristics, such as the number of diseased vessels, target vessel and fluoroscopy time. Our results confirm the findings of the French group, thus suggesting a concept of concentration of PCIs for myocardial infarction and other high-risk PCIs in specialised, high-volume centres, while accepting PCI in low-risk patients in even low-volume hospitals.
Our data are also supported by the analysis of Moscucci et al.17 In 18 504 consecutive PCIs performed by 165 operators in the year 2002, they found better outcomes in high-volume operators than in low-volume operators. This difference was apparent in low-risk as well as high-risk patients.
As cited above, even those studies which are taken to support the thesis of no relationship between outcome and hospital volume15 16 only failed to show a relationship concerning mortality, but MACE rates were significantly associated with hospital volumes in both studies. Furthermore, the overall mortality is very low in these studies, indicating that low-risk patients were treated. For the treatment of STEMI patients, many studies showed a significant volume–outcome relation.7 8 24–27
Limitations of the study
We did register only in-hospital events, not 30-day events. Therefore it may be difficult to compare our data directly with those of other studies with 30-day end points. However, most events occur early after PCI, when patients are usually still in the hospital, or they will have to stay in hospital for PCI-related complications, some occurring even more than 30 days after the index PCI, which are still counted in our registry. Other registries15 17 also reported in-hospital events, instead of 30-day events. We could not evaluate the influence of PCI volumes of single investigators, a factor which is important besides the volume of the hospital.17 We cannot exclude under-reporting of complications, since no formal audit was performed.
CONCLUSIONS
Even in current clinical practice of stent-supported PCI there still is a small but significant volume–outcome relation. This relation, however, is limited to high-risk patients, especially patients presenting with an acute myocardial infarction. At both types of hospital, however, continuing quality measurements on a hospital as well as on an individual basis will help to achieve and maintain low complication rates.
REFERENCES
Footnotes
Competing interests: None declared.