The Specialty Group for Cardiothoracic Surgery, Hospital Authority … First Integrated... · 2019. 5. 16. · The Specialty Group for Cardiothoracic Surgery, Hospital Authority Hong

The Specialty Group for Cardiothoracic Surgery,Hospital Authority Hong Kong

FirstIntegrated report oncardiac surgical outcome

2015-2016

Prepared by

Prof. MJ UnderwoodDr T AuDr HL Cheung

for The Specialty Group for Cardiothoracic Surgery, Hospital Authority Hong Kong

Robin Kinsman BSc PhD

Dendrite Clinical Systems

The Specialty Group for Cardiothoracic Surgery, Hospital Authority Hong KongFirst integrated report on cardiac surgical outcome 2015-2016

3

Acknow

ledgements

Preface

Explanatory text to be inserted

Prof. Sir Bruce Keogh


4

Cont

ents

Contents

Preface 3

Foreword 5

Introduction 6Methodology 6Concept of risk 6Outcome: presentation 6

Analysis

Cardiac surgery activity 8

Cardiac surgical outcomes: mortality 10Overall 10Coronary artery surgery 12Isolated valve surgery 16Other procedures 18

Coronary artery surgery: detailed analysis 20Patient demographics 20Operative procedures 20Morbidity 21

Conclusions 30

Appendices

Codes for CABG used for cross check of mortality 32

The logistic euroSCORE: calculation and definitions 32


5

Introduction

Foreword

This is the first joint report on cardiac surgical activity and outcomes from the three centres in Hong Kong providing cardiac surgical service, Queen Mary Hospital (QMH), Queen Elizabeth Hospital (QEH) and the Prince of Wales Hospital (PWH).

Data collection for cardiac surgical activity and outcomes was commenced in 2006 using a commercially available cardiac database system, which allows integration with the Hospital Authority Clinical Management System (CMS) and also allows sophisticated risk-adjusted outcome analysis. The project was piloted at PWH and start-up funding was from a grant obtained from the Professional Development Agency from the Government of the HKSAR. This was then expanded so currently all three units use the same system and have independent servers for data collection. The whole system and infrastructure have been designed so that we can move towards real-time monitoring of outcomes, which means identifying outlying trends before they become clinically significant.

This has been a step wise process. While intuitively we support openness and transparency there can be unintended consequences of data publication, and an underlying fundamental is the trust of the clinicians in both the data and the analysis process, which takes time to accrue. Working through the Specialty Group for Cardiothoracic Surgery (SG CTS), definitions have been agreed and appropriate risk-scores adopted. These risk scores have been validated independently using the Hong Kong population. A database validation exercise has been completed whereby independent analysis of random cases from each database and checking the accuracy of the recorded factors used to generate risk scores has been completed, and each database deemed fit for purpose.

The use of a commercial database also gives us access to external data analysis and international benchmarking against other norms. We performed a feasibility study last year whereby historical data for coronary artery surgery was independently extracted from our servers, the data was merged and it was benchmarked using risk-adjusted outcomes from the United Kingdom (UK) Society for Cardiothoracic Surgery. This demonstrated that external benchmarking could be achieved using an external collaborator and during the process patient data confidentiality was maintained.

The purpose of this report was to expand that process and look at all cardiac surgical activity for the combined calendar year of 2015 and 2016 and to include a detailed look at coronary bypass surgery as the marker operation for activity and outcomes. This would be done independently and benchmarked against some United Kingdom norms. After written permission from the three centres data for the allotted time period was remotely and independently extracted from our servers in Hong Kong. Initial assessment was made about missing outcome and risk score data and the preliminary results fed-back. Any missing data was then corrected and ratified and a second extraction was performed. Data analysis of activity and outcomes was then performed by an independent external body (Dendrite, UK) an expert in cardiac surgical data analysis. It is this analysis that forms the basis of this report. Mortality data for coronary artery surgery was also cross checked using Hong Kong’s Clinical Data Analysis and Reporting System (CEDARS) after cleaning for coding miss-matches and found to be 100% accurate.

This is the first time we have reported merged data for cardiac surgical activity and outcomes in Hong Kong and the first time this has been done using an independent 3rd party. As stated previously we set up this project so we can move to real-time monitoring of outcomes, but to achieve this we need to be able to utilise the capability of our system and capture more data automatically from the CMS avoiding the delay of personnel-led data entry. The concept of annual reporting results in historical feedback, which we feel in less useful than having real-time data to analyse.

Occasional reports such as this, however, may still be useful to both the medical and patient community. It demonstrates for the first time that cardiac surgical outcomes in Hong Kong are well within international norms and demonstrates the advantages of having access to independent data analysis and benchmarking. The journey of data collection, outcome analysis and public reporting is a complex one. We have step wise addressed all the concerns and put into place underlying mechanisms which now allow us to generate this report whilst maintain the trust of clinicians and the safety of patients data. The next phase of our outcome monitoring programme development should enable us to have an intrinsic system, which will allow us to identify and correct and deviations of performance at an early stage and as such this report and our future plans should provide reassurance for patients, clinicians, managers and commissioners that the cardiac surgery programme in Hong Kong is performing to international standards, and that the SG in CTS is actively looking to monitor and further improve its standards.

Prof. MJ Underwood, Dr T Au, Dr HL Cheung: on behalf of the Specialty Group in Cardiothoracic Surgery

Robin Kinsman, Data Wizard, Dendrite Clinical Systems, United Kingdom


6

Intr

oduc

tion

Introduction

Methodology

The purpose of this exercise was to build on the previous feasibility exercise in 2017 and have an external, internationally benchmarked assessment of cardiac surgical activity and outcomes in Hong Kong. It focuses on the outcomes of 2,560 patients who had cardiac surgery in the assessment period.

After written permission all three cardiac centres allowed remote access to their cardiac databases. Data for the calendar years 2015 and 2016 were remotely extracted and merged by Dendrite. Initial feedback was given on missing data for outcomes and any risk factor included in generating the logistic euroSCORE. During this process outcome data was nearly 100% complete, missing euroSCORE data was fed back to the units for completion and then a re-extraction of data was completed. Data for calculating the euroSCORE was 97% complete.

Cross-checking of all mortality and outcomes using HA data not feasible due to the complex coding mechanism for administrative data, but coronary surgery was used as a surrogate. There were 5 apparent coronary deaths accounted for by CEDARS but not included in the Dendrite database; but, on cross-checking, they were identified as not actually being isolated CABG procedures. The mortality matches for all three centres was then 100%.

The complete, validated data on this patient cohort was then analysed by Dendrite using some United Kingdom indices and benchmark indicators.

Concept of risk

Some patients are more unwell than others and some need more complex operations. Hence, hospitals and consultants perform a range of cardiac operations and the type of patients they operate on can differ. So that we can make fair comparisons between them, the survival rates have been risk adjusted to take into account how ill the patient was and the complexity of the operation.

Risk adjustment is done using complex mathematical methods, to effectively show what the survival would have been had all patients been similar and all operations been of average difficulty. The risk that any one patient will not survive surgery is dependent upon a number of different factors, some of which can be quantified. Risk scoring systems for patients undergoing cardiac surgery attempt to take some of these factors and turn them into a numerical risk score, which represents the probability of death or some other outcome for an individual. Over time, a variety of risk stratification systems have evolved, ranging from simple additive systems to highly complex statistical algorithms. These provide the basis for rational and meaningful comparisons of outcomes between groups of patients and institutions.

The risk adjustment method that has been used for adult cardiac surgery is an adaptation of the euroSCORE model. This model has been adapted to make sure that it is an accurate predictor of survival overall. This means that it discriminates well between patients with higher and lower risk in general. However, no risk model is perfect. Because it is difficult for risk adjustment to fully account for these specialist practices, we should be very careful about drawing conclusions based on risk adjusted survival rates alone and any outcomes should be interpreted with caution.

We use the euroSCORE as a risk model, it has been tested on the Hong Kong population and performs well in its discriminatory capability for death / survival, but over time it has drifted in its calibration (its accuracy in refecting risk for a defined group of patients). To account for this the United Kingdom Society has calculated a complex correction co-efficient to adjust for calibration drift and this has been used for the current data on isolated CABG.

Outcome: presentation

Survival rates are expected to be closer to the national average when more procedures have been carried out. This is because when there are more procedures a death affects the overall survival rate less. When fewer procedures have been done, even a single death (which could be due to chance), affects the overall survival rate much more. There will always be some variation between hospital survival rates because of the differences between patients and operation types. Using only a national average as the standard can make it difficult to tell whether a survival rate that sits above the national average is higher than we would expect it to be or not.

In presenting outcomes we use a variety of methods to look at risk adjusted mortality. Whilst they all employ complex statistical analysis, outcome reporting is not a true science and its limitations should be recognised. Each methodology has advantages and disadvantages so we have used multiple methods to ensure consistency in analysis and hence have represented the data in differing formats. Audit is different from scientific investigation and it is inappropriate to expect the same statistical tools to be useful in both cases. In audit, it is reasonable


7

Introduction

to inspect the data without any prior assumptions rather than to test a particular hypothesis. If the results are satisfactory, seeking differences between units or seeking to rank them, is not helpful.

We have used funnel plots to graphically demonstrate outcomes in terms of mortality. These funnel plots also show control limits; the curved lines on the charts that give them the funnel shape. The horizontal line in the middle of the funnel represents the average mortality rate. Control limits show the lowest mortality rate we would expect based on the number of operations performed and their difficulty. If a mortality rate is between the two control limits near the top and bottom of the graph, it is an expected mortality rate, and any variation above or below the average can be put down to chance alone. If the mortality rate is above the upper control limit, it is higher than expected. This may mean a number of things, including problems with the quality of the data submitted to the audit, specialist practice that can’t be properly risk adjusted, chance, or poor quality of care. If the mortality rate is below the lower control limit it is better than expected. It is important to remember that the surgeons whose mortality rates are included on this site work as part of a larger clinical team. These teams include anaesthetists, junior medical staff, perfusionists, nurses, pharmacists and physiotherapists.

We have also used control plot calculators to numerically demonstrate outcomes and make a statistical assessment of any difference between observed and expected outcomes using the euroSCORE as the risk modulator. This methodology takes into account not only the mean calculated risk but also the standard deviations around which the mean value sits, which is important if there are large ranges of expected mortality within the analysis group.

The variable life-adjusted (VLAD) plots show the difference between expected and actual cumulative mortality. VLAD shows whether a units performance is above or below what might be expected. This mortality-scoring system accumulates penalties for each death and rewards for every survivor, based on the inherent risk of peri-operative death of each case concerned. VLAD provides a graphical display of risk-adjusted survival figures for individual surgeons or units over time and could be modified to monitor performance over a range of treatments and outcomes. The variable life-adjusted display is a plot that shows the difference between the cumulative expected mortality and the deaths that actually occurred. VLAD gives a tally of how far a unit or surgeon’s cumulative mortality figures are above or below the prediction, taking into account the expected risk associated with the particular caseload. Every case in the series is plotted from left to right on a horizontal axis. Essentially, the line moves up for survivors and down for deaths. For each case there is a previous risk predicted by euroSCORE that determines the magnitude by which the graph ascends or descends. For each patient who survives, the plot ascends by an amount equal to the estimated probability of death. For an in-hospital death, the plot descends by an amount equal to the estimated probability of survival. If the patient was at high risk of peri-operative death, the units mortality figures are not unduly penalised, but mortality figures are more heavily penalised when a low-risk patient dies. VLAD shows the difference between the expected and actual cumulative mortality. VLAD is not a formal statistical testing procedure, rather it is a method that assists in the process of assimilating complex information on patterns of mortality. We have not used confidence intervals on these plots because they would imply that we have defined what is or is not acceptable performance. VLAD plots should be used with common sense to avoid over interpretation of the results, but can be very useful when looking at larger numbers of procedures and trends over time.


8

Ana

lysi

s

Fig. 1.01 Hong Kong cardiac surgery: calendar years 2015 & 2016

Hospital

PWH QEH QMH All

Card

iac

proc

edur

e

Isolated CABG 266 319 231 816

CABG & valve 42 39 68 149

CABG, valve & other 4 9 29 42

CABG & other 3 9 39 51

Isolated valve 257 237 273 767

Valve & other 54 72 175 301

Other 88 134 212 434

All 714 819 1,027 2,560

Analysis

Cardiac surgery activity

Two thousand five hundred and sixty patients had cardiac surgical procedures during the calendar years 2015 and 2016. Fig. 1.01 shows the distribution of these procedures by operative category and is presented as absolute numbers. The categories are grouped so comparative analysis on the major operative groups can be made with greater statistical power. The category other contains procedures outside the main operative categories. This is a complex group and the bulk of procedures represented include mainly aortic procedures, predominantly emergency and other complex operations such as cardiac transplantation.

Each of the three cardiac units in Hong Kong has different ways of managing the patient journey through fixed facilities and resource allocation. Access to operating theatres, the way patients are managed on intensive care, high dependency and the ward differ in all three centres and hence impact on throughput in terms of operative numbers in different ways.

Fig. 1.02 shows the same data presented as a percentage of overall activity. The most striking change is the percentage of coronary surgeries performed, which is on average around 30% for Hong Kong. Ten years ago, coronary surgery accounted for nearly 70% of activity and this reduction over time is probably a reflection of advances in percutaneous therapies as an initial treatment therapy. Overall cardiac surgery numbers, however, have remained static over time, reflecting increases in valvular surgery, complex combined operations and also complex aortic surgery, which has steadily risen.


9

Analysis

Fig. 1.02 Cardiac procedures at each hospital; calendar years 2015 & 2016

PWH QEH QMH

Perc

enta

ge o

f ope

ratio

ns

Isol

ated

CA

BG

CABG

& v

alve

CABG

, val

ve &

ot

her

CABG

& o

ther

Isol

ated

val

ve

Valv

e &

oth

er

Oth

er

Cardiac procedures

40%

30%

20%

10%

0%


10

Ana

lysi

s

Fig. 1.03 Hong Kong cardiac surgery: In-hospital mortality and procedure; calendar years 2015 & 2016

In-hospital mortality

Counts Mortality rate

Alive Dead Risk 1 Observed Risk-adjusted

Card

iac

proc

edur

es

Isolated CABGPWH 262 3 4.1 1.1% 0.8%QEH 312 7 8.6 2.2% 1.8%QMH 226 5 7.2 2.2% 1.5%

CABG & valvePWH 40 2 2.1 4.8% 4.5%QEH 35 4 1.5 10.3% 23.5%QMH 64 3 3.7 4.5% 3.7%

CABG, valve & otherPWH 4 0 0.3 0.0% 0.0%QEH 7 2 0.6 22.2% 50.9%QMH 26 3 2.6 10.3% 11.8%

CABG & otherPWH 2 1 0.6 33.3% 44.2%QEH 5 4 1.6 44.4% 66.1%QMH 35 4 4.2 10.3% 9.8%

Isolated valvePWH 256 1 9.2 0.4% 0.0%QEH 227 10 7.6 4.2% 5.5%QMH 262 11 13.0 4.0% 3.4%

Valve & otherPWH 47 7 6.1 13.0% 14.6%QEH 65 7 5.4 9.7% 12.3%QMH 172 3 10.6 1.7% 0.5%

OtherPWH 73 3 9.5 3.9% 1.3%QEH 109 25 14.3 18.7% 28.6%QMH 190 20 18.6 9.5% 10.2%

AllPWH 684 17 32.0 2.4% 1.3%QEH 760 59 39.6 7.2% 10.4%QMH 975 49 59.8 4.8% 4.0%

Cardiac surgical outcomes: mortality

Overall

The crude mortality for Hong Kong was 4.8% for all cardiac procedures which is comparable with international standards. Risk adjusted, the observed versus expected mortality (calculated using the euroSCORE) was 41.6 versus 43.8 so, overall, Hong Kong is performing as expected in terms of mortality related to the expected outcomes using our scoring model. Fig. 1.03 shows the mortality for the operative sub-groups and gives a representation of observed versus expected (risk score) for each. As stated previously, caution must be taken when looking at sub-groups with less than 100 procedures as the statistical analysis may not bear scrutiny with small patient groups that may contain a heterogeneous mix of complex cases, which do not calibrate well with our scoring system. With this in mind we have looked in more detail at the isolated coronary surgery and isolated valve groups, which are represented by significant number, making the statistical analysis more robust. We have also

1. Logistic euroSCORE ÷ 2.


11

Analysis

Fig. 1.04 All cardiac surgery: Crude mortality; calendar years 2015 & 2016

Upper 99% alert line Upper 99.9% alarm line Average

Lower 99% alert line Lower 99.9% alarm line

Crud

e m

orta

lity

rate

0 200 400 600 800 1,000 1,200

Number of operations

9%

8%

7%

6%

5%

4%

3%

2%

1%

0%

looked at the other category using control chart analysis. Fig. 1.04 shows a non-risk adjusted funnel plot for mortality for all cardiac procedures performed in the assessment period. At first glance it seems like there is an outlier whose mortality is lying between the 99% and 99.9% alert lines. However, to account for this and some of the conflicts mentioned above Fig. 1.05 shows the VLAD plot for all cardiac procedures per unit using the logistic euroSCORE as the risk calculator. This shows that all three units are performing in line with the predicted risk for the patients they are treating and emphasises the need for risk adjustment if a fair and true presentation of mortality is to be achieved.

Fig. 1.05All cardiac surgery: VLAD plot;

risk set as logistic EuroSCORE; calendar year 2015 & 2015

PWH QEH QMH

Live

s sav

ed /

patie

nts

0 200 400 600 800 1,000 1,200


70

60

50

40

30

20

10

0

-10


12

Ana

lysi

s

Fig. 1.06 Hong Kong isolated first-time CABG: In-hospital mortality; calendar years 2015 & 2016



Alive Dead Risk 1 Observed Risk-adjusted

Hos

pita

l PWH 262 3 4.1 1.1% 0.9%

QEH 311 7 8.5 2.2% 1.0%

QMH 225 5 7.1 2.2% 0.9%

All 798 15 19.8 1.8% 1.0%

UK financial years ending 2014-2016 50,108 646 1.3%

Coronary artery surgery

Overall crude mortality was 1.8% (798 cases). Fig. 1.06 shows risk adjusted mortality in table form. This is in line with international standards (United Kingdom mortality of 1.3%, fiscal years 2014-2016, 50,108 patients). Fig. 1.07 shows the risk adjusted mortality for first-time isolated CABG compared to a published standard from the United Kingdom and is represented in a funnel plot. All three centres in Hong Kong have performed according to expected standards with no outlier. This analysis is confirmed using control chart analysis as shown in Fig. 1.08. This CABG cohort included all cases whether elective, urgent or emergency and Fig. 1.09 shows the outcomes when broken down according to operative urgency. Again care must be taken when interpreting outcomes in low-volume groups, and emergency coronary surgery in particular represents a salvage situation that cannot be adjusted for by risk scoring. Overall, outcomes in all operative urgency categories are well within that expected. The risk scores for this analysis were subjected to a coefficient factor correction to improve calibration. Fig. 1.10 shows the distribution of isolated CABG cases by euroSCORE category. There are small differences between the units, but with overlap of the error bars the distribution appears even, with PWH operating on more patients in the lower risk group. These variations will have been accounted for by using the re-calibrated risk score in the outcome analysis.

Fig. 1.11 shows the VLAD plot for CABG outcomes using the re-calibrated euroSCORE. As stated this provides a good graphical representation of any deviation in outcomes over time. All three units are performing in accordance to their individual case mix.


13

Analysis

Fig. 1.08 Hong Kong cardiac surgery: Control chart analysis on in-hospital mortality after isolated CABG; calendar years 2015 & 2016

PWH QEH QMH

Cont

rol c

hart

dat

a

Operations 262 312 226

Deaths 3 7 5

Observed rate 1.1% 2.2% 2.2%

Predicted rate 1 4.1% 8.6% 7.2%

Lower limits

2SD 1.9% 5.4% 4.0%

3SD 1.5% 4.2% 2.7%

4SD 0.4% 2.9% 1.3%

Upper limits

2 SD 6.9% 11.9% 10.6%

3SD 8.0% 13.8% 12.8%

4SD 9.9% 15.7% 15.0%

Performanceslightly low definitely possibly

probably normal

better than expected

better than expected

Fig. 1.07First-time isolated CABG: Risk-adjusted mortality compared to a published

standard from the United Kingdom; calendar years 2015 & 2016

Upper 99% alert line Upper 99.9% alarm line UK average 2014-16 (n=50,754)


Risk

-adj

uste

d m

orta

lity

rate

0 50 100 150 200 250 300 350


6%

5%

4%

3%

2%

1%

0%


14

Ana

lysi

s

Fig. 1.09 Honk Kong cardiac surgery: In-hospital mortality and urgency for isolated CABG; calendar years 2015 & 2016



Alive Dead Unspecified Risk 1 Observed rate RA rate

Hos

pita

l & o

pera

tive

urg

ency PWH

Elective 215 2 1 2.7 0.9% 0.7%

Urgent 45 1 0 1.3 2.2% 1.7%

Emergency 2 0 0 0.1 0.0% 0.0%

QEH

Elective 229 3 0 4.0 1.3% 1.0%

Urgent 49 2 0 1.7 3.9% 4.7%

Emergency 34 2 0 2.9 5.6% 3.9%

QMH

Elective 170 2 0 3.5 1.2% 0.7%

Urgent 52 2 0 2.8 3.7% 2.7%

Emergency 4 1 0 0.9 20.0% 21.0%


15

Analysis

Fig. 1.10 Isolated CABG: Logistic EuroSCORE; calendar years 2015 & 2016

PWH (n=266) QEH (n=319) QMH (n=231)

Perc

enta

ge o

f pat

ient

s

<2.0 2.0-3.9 4.0-5.9 6.0-7.9 8.0-9.9 >9.9

Logistic EuroSCORE group

60%

50%

40%

30%

20%

10%

0%

Fig. 1.11Isolated CABG: VLAD plot;

risk set as logistic EuroSCORE ÷ 2; calendar year 2015 & 2015

PWH QEH QMH

Live

s sav

ed /

patie

nts

0 50 100 150 200 250 300 350


4

3

2

1

0

-1


16

Ana

lysi

s

Isolated valve surgery

Seven hundred and forty-five (745) isolated valve surgeries were performed. The overall crude mortality for Hong Kong was 2.8% with an observed versus expected mortality of 7.3 versus 9.9, which, again, is well within international and expected standards. Fig. 1.12 shows the funnel plot for mortality for this group, and all three centres fall well within the alert lines, demonstrating no outliers. At first glance there may seem to be differences, but when risk adjustment is made using control charts, we can see in Fig. 1.13 that all 3 units are performing as expected or better than expected according to the risk profile of the patients operated upon.

This is re-enforced in Fig. 1.14 where again we can see in risk-adjusted graphical form, all outcomes are within acceptable standards. This emphases the need to interpret crude mortality data with caution and, although not perfect, some risk adjustment to reflect case-mix is mandatory.

Fig. 1.12 Isolated valve surgery: Crude mortality; calendar years 2015 & 2016

Upper 99% alert line Upper 99.9% alarm line Average


Crud

e m

orta

lity

rate

0 100 200 300


9%

8%

7%

6%

5%

4%

3%

2%

1%

0%


17

Analysis

1. Logistic euroSCORE

Fig. 1.14Isolated valve surgery: VLAD plot;

risk set as logistic EuroSCORE; calendar year 2015 & 2015

PWH QEH QMH

Live

s sav

ed /

patie

nts

0 50 100 150 200 250 300


18

16

14

12

10

8

6

4

2

0

-2

Fig. 1.13 Hong Kong cardiac surgery: Control chart analysis on in-hospital mortality after isolated valve surgery; calendar years 2015 & 2016

PWH QEH QMH

Cont

rol c

hart

dat

a


Deaths 1 10 11


Predicted rate 1 9.2% 7.6% 13.0%

Lower limits

2SD 5.9% 4.4% 9.2%

3SD 4.3% 3.1% 7.2%

4SD 2.7% 1.8% 5.3%

Upper limits

2 SD 12.9% 11.5% 17.2%

3SD 14.8% 13.2% 19.5%

4SD 17.2% 15.4% 22.1%

Performancedefinitely definitely

better than expected as expected better than

expected


18

Ana

lysi

s

1. Logistic euroSCORE

Fig. 1.15 Hong Kong cardiac surgery: Control chart analysis on in-hospital mortality after other cardiac surgery; calendar years 2015 & 2016

PWH QEH QMH

Cont

rol c

hart

dat

a


Deaths 3 25 20


Predicted rate 1 9.5% 18.6% 10.5%

Lower limits

2SD 2.7% 11.9% 13.2%

3SD 1.4% 8.3% 10.5%

4SD 0.0% 5.5% 8.4%

Upper limits

2 SD 16.4% 26.6% 24.2%

3SD 20.6% 30.3% 27.4%

4SD 26.0% 34.9% 30.5%

Performanceslightly low

as expected as expected probably normal

Other procedures

As stated before, this category represents a heterogeneous group often pre-dominated by emergency aortic or cardiac transplant procedures. However, we think it is still an important category to give us some idea of performance whilst taking into account the caution required in applying statistical analysis.

Three hundred and seventy two procedures (372) fell into this category. The crude mortality was 10.7% for Hong Kong overall. The observed versus expected (logistic euroSCORE) mortality was 48 versus 42.4. Using control charts to account for risk and also the variation of risk scores within the groups, we can see in Fig. 1.15 that all 3 centres performed as expected in terms of outcomes in this heterogeneous and complex group.


19

Analysis


20

Ana

lysi

s

Coronary artery surgery: detailed analysis

As mentioned the aim of this report was to generate risk adjusted outcomes with international benchmarking. In the past, CABG dominated cardiac surgical activity and as a standard procedure was often used as a marker operation for assessment of quality and outcomes. As mentioned above, this has changed and CABG now represents around 30-40% of the total workload in Hong Kong. This being acknowledged, we have still looked at this cohort of patients in more detail and to report other markers of outcomes such as morbidity as well as the previously discussed mortality and gain insight into patient demographics and operative procedures.

Patient demographics

Fig. 1.16 shows the distribution of patients according to age. Intuitively age is a well-recognised risk factor for both mortality and morbidity. There are similar distributions relative to age groups across all three units with overlapping confidence intervals suggesting equivalence. It does show, however, that on average in Hong Kong over 25% of CABG patients are over the age of 70 years, a significant proportion who have an associated increase in risk profile. Fig. 1.17 shows all three units have a similar distribution according to gender, which is important as females are considered at being of higher risk for mortality and complications. Fig. 1.18 shows the distribution of diabetes and hypertension across the three units. Again, these are important pre-operative risk factors, which may impact. Hypertension is present in over 83% of our patient population and almost 50% have some form of diabetes. Fig. 1.19 shows the percentage of patients undergoing surgery with left main stem disease and overall this approaches 40% in Hong Kong. Fig. 1.20 shows the number of diseased vessels in the patients having operative intervention. This is seen as a surrogate for complexity of coronary artery disease and seems to be uniformly distributed overall. Fig. 1.21 shows the symptom status of this patient cohort. As expected in the modern era of cardiac surgery, few patients have minimal symptoms. Previous myocardial infarction is an important risk factor as it may reflect severity of myocardial damage and Fig. 1.22 shows the number of previous myocardial infarctions in our patient population. Around 60% of our cohorts present having had a previous infarct with an even distribution across the three cardiac centres. The timing of surgery in relation to a previous infarct is an important risk factor, the sooner an operation is required reflects the severity of the coronary disease and may have a profound impact on outcomes. Fig. 1.23 shows the distribution for the three centres in Hong Kong. It is difficult to look at precise differences with overlapping confidence intervals, but the data here shows that relatively few patients in Hong Kong require surgery within 24 hours of an infarct (usually an emergency or salvage procedure); however, a significant number are operated upon within 1-30 days. These latter patients represent a high risk group but our outcomes suggest they are being managed well. All the above may have an impact on heart function measured in this report as ejection fraction. The poorer the heart function the greater the perceived risk. Fig. 1.24 shows the distribution according to heart function and again small differences between units are unlikely to be clinically significant.

Operative procedures

Fig. 1.25 shows the number of bypass grafts per patient per hospital. Again perceived differences are unlikely to reach statistical significance, but over 80% of our population require 3 or more bypass grafts, a reflection of the severity of the coronary disease we are now treating surgically. Use of the left internal mammary artery (LIMA) as a bypass graft to the left anterior descending artery is known to improve long-term survival and outcomes and its percentage use is often used as a quality surrogate for coronary surgery. It is sometimes not used in very high risk patients so its utilization will rarely reach 100%. Fig. 1.26 shows the absolute numbers in terms of using the LIMA and Fig. 1.27 shows this in graphical form with error bars. Both QMH and PWH use the LIMA in approaching 95% of patients. The LIMA usage rate is lower in QEH but, importantly, the data for LIMA usage in their database was not complete, which might explain this deviation . Myocardial protection is an important component of any coronary operation and Fig. 1.28 shows the different techniques used in Hong Kong. There is relatively little off pump surgery performed with approaching 99% of cases utilizing cardiopulmonary bypass. Cardioplegia is uniformly used as myocardial protection with almost 100% adoption of blood as the protection medium. Intermittent antegrade infusion is the baseline strategy, but there are individual unit differences in the temperature at which it is used. This has no impact on outcomes, and individual units choose their own best methods of myocardial protection, but we can say apart from temperature there is uniformity in the technique in Hong Kong.


21

Analysis

Morbidity

The focus of this report as stated was on risk-adjusted outcomes in terms of mortality for cardiac surgery in Hong Kong as part of our journey to real-time monitoring. Morbidity following surgery though is important as it might impact on the patient’s quality-of-life and also some instances of morbidity can reflect the quality of care and service being given. For patients following coronary surgery we picked three outcomes to look at: new permanent stroke, re-operation for bleeding, and the need for new dialysis in the post-operative period. A post-operative stroke obviously has implications for a patient’s quality-of-life and hence is an important factor to measure. The frequent need for a second operation to control bleeding may reflect some deviation in the team’s performance in terms of surgical technique or blood product usage, and so is often used as a quality indicator. The need for the institution of new dialysis for renal failure is also seen as a surrogate for some deviation in post-operative care at a number of levels. These three variables are shown in Fig. 1.29. As we can see, with overlapping confidence limits there is no significance difference between the three units and the absolute numbers for each morbidity recorded are low and well within international standards.


22

Ana

lysi

s

Fig. 1.16 Isolated CABG: Age at surgery; calendar years 2015 & 2016

PWH (n=266) QEH (n=318) QMH (n=230)Pe

rcen

tage

of p

atie

nts

<51 51-55 56-60 61-65 66-70 71-75 76-80 >80

Age at surgery / years

28%

24%

20%

16%

12%

8%

4%

0%

Fig. 1.17 Isolated CABG: Gender; calendar years 2015 & 2016

PWH (n=266) QEH (n=319) QMH (n=231)

Perc

enta

ge fe

mal

e pa

tient

s

PWH QEH QMH

Hospital

20%

16%

12%

8%

4%

0%


23

Analysis

Fig. 1.18 Hong Kong cardiac surgery: Other risk factors in isolated CABG; calendar years 2015 & 2016

Hospital

PWH QEH QMH All

Diabetes

No 119 154 114 387

Yes 147 164 117 428

Unspecified 0 1 0 1

Diabetes rate 55.3% 51.6% 50.6% 52.5%

Hypertension

No 61 47 20 128

Yes 205 237 211 653

Unspecified 0 35 0 35

Hypertension rate 77.1% 83.5% 91.3% 83.6%

Fig. 1.19 Hong Kong isolated CABG: left main stem disease; calendar years 2015 & 2016

Left main stem disease

No Yes Unspecified Rate

Hos

pita

l PWH 186 80 0 30.1%

QEH 180 132 7 42.3%

QMH 123 108 0 46.8%

All 489 320 7 39.6%


24

Ana

lysi

s

Fig. 1.21 Isolated CABG: Angina; calendar years 2015 & 2016

PWH (n=266) QEH (n=315) QMH (n=231)

Perc

enta

ge o

f pat

ient

s

CCS 0 CCS 1 CCS 2 CCS 3 CCS 4

Angina

60%

50%

40%

30%

20%

10%

0%

Fig. 1.20Isolated CABG: Number of diseased coronary vessels;

calendar years 2015 & 2016

PWH (n=266) QEH (n=317) QMH (n=231)

Perc

enta

ge o

f pat

ient

s (lo

g sc

ale)

0 1 2 3 0 1 2 3 0 1 2 3

PWH QEH QMH

Number of diseased vessels and hospital

100%

10%

1%

0.1%


25

Analysis

Fig. 1.22 Isolated CABG: Number of previous MIs; calendar years 2015 & 2016

PWH (n=266) QEH (n=319) QMH (n=231)

Perc

enta

ge o

f pat

ient

s

0 1 >1 0 1 >1 0 1 >1

PWH QEH QMH

Number of previous MIs and hospital

60%

50%

40%

30%

20%

10%

0%

Fig. 1.23 Isolated CABG: Timing of prior MIs; calendar years 2015 & 2016

PWH (n=168) QEH (n=179) QMH (n=132)

Perc

enta

ge o

f pat

ient

s

<24

hour

s

1-30

day

s

31-9

0 da

ys

>90

days

<24

hour

s

1-30

day

s

31-9

0 da

ys

>90

days

<24

hour

s

1-30

day

s

31-9

0 da

ys

>90

days

PWH QEH QMH

Number of previous MIs and hospital

60%

50%

40%

30%

20%

10%

0%


26

Ana

lysi

s

Fig. 1.24 Isolated CABG: Ejection fraction; calendar years 2015 & 2016

Good Fair PoorPe

rcen

tage

of p

atie

nts

PWH QEH QMH

Hospital

80%

70%

60%

50%

40%

30%

20%

10%

0%

Fig. 1.25 Isolated CABG: Number of bypass grafts; calendar years 2015 & 2016

PWH (n=266) QEH (n=293) QMH (n=231)

Perc

enta

ge o

f pat

ient

s

1 2 3 >3 1 2 3 >3 1 2 3 >3

PWH QEH QMH

Number of bypass grafts and hospital

90%

80%

70%

60%

50%

40%

30%

20%

10%

0%


27

Analysis

Fig. 1.26 Hong Kong cardiac surgery: LIMA usage in isolated CABG; calendar years 2015 & 2016

LIMA used

No Yes Unspecified LIMA rate

Hos

pita

l PWH 21 245 0 92.1%

QEH 99 187 33 65.4%

QMH 12 219 0 94.8%

All 132 651 33 83.1%

Fig. 1.27 Isolated CABG: LIMA usage; calendar years 2015 & 2016

PWH (n=266) QEH (n=286) QMH (n=231)

Perc

enta

ge L

IMA

usa

ge

PWH QEH QMH

Hospital

100%

80%

60%

40%

20%

0%


28

Ana

lysi

s

Fig. 1.28 Hong Kong cardiac surgery: Protection of the heart in isolated CABG; calendar years 2015 & 2016

Hospital

PWH QEH QMH All

Prot

ecti

on o

f the

hea

rt

Bypass

No 4 9 2 15

Yes 262 309 220 791


Bypass rate 98.5% 97.2% 99.1% 98.1%

Cardioplegia on bypass

Non-cardioplegia 1 0 8 9

Cardioplegia 261 118 206 585


Cardioplegia rate 99.6% 100.0% 96.3% 98.5%

Solution

Blood 261 112 205 578

Crystalloid 0 1 1 2

Unspecified 0 5 0 5

Blood use rate 100.0% 99.1% 99.5% 99.7%

Temperature

Warm 2 108 178 288

Warm & cold 1 1 3 5

Cold 258 3 1 262


Warm temperature rate 0.8% 96.4% 97.8% 51.9%

Infusion mode

Antegrade alone 261 107 206 574

Ante & retrograde 0 6 0 6

Unspecified 0 5 0 5

Antegrade alone rate 100.0% 94.7% 100.0% 99.0%

Timing

Intermittent 260 112 176 548

Continuous 1 0 0 1


Intermittent rate 99.6% 100.0% 100.0% 99.8%


29

Analysis

Fig. 1.29 Isolated CABG: Post-operative outcomes; calendar years 2015 & 2016

PWH QEH QMH

New

per

man

ent s

trok

e ra

te

PWH

(n=2

66)

QEH

(n=3

11)

QM

H (n

=224

)

Hospital

1.4%

1.2%

1.0%

0.8%

0.6%

0.4%

0.2%

0.0%

Re-o

pera

tion

for b

leed

ing

rate

PWH

(n=2

66)

QEH

(n=3

18)

QM

H (n

=223

)

Hospital

5%

4%

3%

2%

1%

0%

New

dia

lysi

s rat

e

PWH

(265

)

QEH

(n=2

87)

QM

H (n

=222

)

Hospital

2.0%

1.6%

1.2%

0.8%

0.4%

0.0%


30

Ana

lysi

s

Conclusions

1. This is the first integrated report on cardiac surgical outcome in Hong Kong that has been subject to external data analysis and international benchmarking.

2. This report reflects the commitment of the three cardiac surgical units to openness and transparency of outcomes as we move toward real-time monitoring and has been achieved with the support of the Specialty Group for Cardiothoracic Surgery in Hong Kong under the umbrella of the Hospital Authority.

3. It represents a step-wise process in which all three centres have agreed uniform definitions of risk factors, initiated database inspections for data completeness, accuracy and previously performed a feasibility study using historical data for coronary surgery. Mortality for coronary surgery in this report has been cross-checked with administrative data to ensure accuracy of reporting.

4. It demonstrates that overall cardiac surgical outcomes in terms of mortality are well within risk prediction, and Hong Kong performs well when international benchmarking is used.

5. Satisfactory outcomes in terms of mortality has also been shown in the sub-groups of coronary bypass surgery, isolated valve surgery and in a group representing heterogeneous high-risk surgery.

6. More detailed analysis of the coronary surgery population has shown some minor differences between units as regards patient demographics and operative procedures, but satisfactory outcomes have been maintained.

7. Markers of morbidity for patients undergoing coronary surgery are well within international standards are consistent amongst the three centres.

8. This report confirms that the infra-structure underpinning our data collection and analysis allows us to be subjected to external data extraction, analysis and benchmarking whilst maintaining patient confidentiality.


31

Analysis

Acknowledgments

Multi-disciplinary team work is required to provide a comprehensive cardiac surgical service. Much attention is naturally given to the surgical arm of this effort, but we must affirm the activity and outcomes presented here demonstrate par excellence the benefits of teamwork. None would have been possible without a wide range of associated personnel, including cardiology, anaesthetic and intensive care colleagues, junior medical colleagues, ward, intensive care and theatre nurses, perfusionists, physiotherapists, pharmacists, blood bank technicians, laboratory technicians and many other support personnel. Outcomes and service provision in this and future reports reflect the dedicated effort of all these professionals. All of their efforts are crucial in maintaining our outcomes particularly in the face of changing and increasingly high-risk patient profiles. We continue to face many pressures as we seek to deliver even higher quality care for our patients; all of the above mentioned professional groups have worked and continue to work above and beyond expectations to ensure that standards are consistently maintained. This is particularly pertinent when we are stretching the limits of current resource.

We would like to acknowledge all of the doctors, nurses, perfusionists and associated personnel who have contributed to data collection. We believe all personnel involved in care also have a duty to collect data and ensure quality and this has been taken on with vigour by the majority.

We thank colleagues, medical and administrative represented by the Cardiothoracic Specialty Group in Hong Kong for their support and we look forward to further collaborative efforts as we extend this process. We are grateful to all the HA IT team who continue to make a contribution and are now active in helping us integrate our databases and analysis modules so we can move to real-time outcome monitoring.

Importantly we would also like to thank Dr Peter Walton and his team at Dendrite for assisting us in the production of this important report.


32

App

endi

ces

Appendices

Codes for CABG used for cross check of mortality

• Equal To 36.10 (1) Coronary artery bypass graft

• Equal To 36.11 (0) Aortocoronary bypass of one coronary artery

• Equal To 36.12 (0) Aortocoronary bypass of two coronary arteries

• Equal To 36.13 (0) Aortocoronary bypass of three coronary arteries

• Equal To 36.14 (0) Aortocoronary bypass of four or more coronary arteries

• Start With 36.15 Single Internal Mammary-Coronary Artery Bypass

• Start With 36.16 Double Internal Mammary-Coronary Artery Bypass

• Start With 36.17 Abdominal-coronary artery bypass

• Start With 36.19 Other Bypass Anastomosis for Heart Revascularization

The logistic euroSCORE: calculation and definitions

For a given patient, the logistic euroSCORE is the predicted mortality according to the logistic regression equation, can be achieved with the following formula:

Predicted mortality =

where:

• e is the natural logarithm (2.718281828...)

• βo is the constant of the logistic regression equation (-4.789594)

• βi is the coefficient of the variable Xi in the logistic regression equation provided in the table below. Xi = 1 if a categorical risk factor is present and 0 if it is absent

For age, Xi = 1 if the patient’s age < 60 years; Xi increase by one point per year thereafter; hence, for age 59 or less Xi = 1, age 60 Xi = 2, age 61 Xi = 3, and so on.

This score is automatically calculated for each patient after data input into the Hong Kong Cardiac Surgical Database.


33

Appendices

Beta coefficients for the Logistic regression model of euroSCORE

Patient-related factors beta

age in years 0.0666354

female gender 0.3304052

chronic pulmonary disease 0.4931341 longterm use of bronchodilators or steroids for lung disease

extra-cardiac arteriopathy 0.6558917 any one or more of the following: claudication, carotid occlusion or >50% stenosis,

previous or planned intervention on the abdominal aorta,limb arteries or carotids

neurological dysfunction disease 0.841626 severely affecting ambulation or day-to-day functioning

previous cardiac surgery 1.002625 requiring opening of the pericardium

raised serum creatinine 0.6521653 >200 µmol l-1 pre-operatively

active endocarditis 1.101265 patient still under antibiotic treatment for endocarditis at the time of surgery

critical pre-operative state 0.9058132 any one or more of the following: ventricular tachycardia or fibrillation or aborted

sudden death, preoperative cardiac massage, preoperative ventilation before arrival in the anaesthetic room,preoperative inotropic support, intraaortic balloon counterpulsation or pre-operative acute renal failure (anuria or oliguria<10 ml hr-1)

Cardiac-related factors beta

unstable angina 0.5677075 rest angina requiring iv nitrates until arrival in the anaesthetic room

moderate LV dysfunction 0.4191643 left ventricular ejection fraction 30-50%

poor LV dysfunction 1.094443 left ventricular ejection fraction <30%

recent myocardial infarct 0.5460218 <90 days before surgery

pulmonary hypertension 0.7676924 systolic PA pressure>60 mm Hg

Operation-related factors beta

emergency 0.7127953 carried out on referral before the beginning of the next working day

other than isolated CABG 0.5420364 major cardiac procedure other than or in addition to CABG

surgery on thoracic aorta 1.159787 for disorder of ascending, arch or descending aorta

post-infarct septal rupture 1.462009

The Specialty Group for Cardiothoracic Surgery, Hospital Authority … First Integrated... · 2019. 5. 16. · The Specialty Group for Cardiothoracic Surgery, Hospital Authority Hong

Documents