A Importância da aplicação dos Scores e do conceito do Heart Team para a cirurgia cardíaca.

(1)

A Importância da aplicação dos

Scores e do conceito do “Heart

Team” para a cirurgia cardíaca.

Mauro Paes Leme

[email protected]

(2)

F.W.Taylor

(Administração Científica)

1856 - 1915

• é preciso estudar cientificamente o

trabalho como objeto de pesquisa séria;

• é preciso desenhar as ferramentas e os

processos de trabalho para a dimensão

humana;

• é preciso preparar as pessoas para usar as

ferramentas da forma mais eficiente que se

conhece.

(3)

Benchmarking

(Xerox)

• Processo de comparação do desempenho

entre dois ou mais sistemas;

• Método sistemático de procurar os melhores

processos e idéias inovadoras que conduzem

a um desempenho superior (Christipher E.

Bogan);

• Curiosidade histórica: marcas (“marks”)

deixadas nos bancos de descanço(“bench”),

indicando a distância de Roma, capital do

Império.

(4)

Benchmarking

Modelos de Predição

• processo contínuo;

• procura fornecer informações valiosas;

• processo de aprendizagem;

• permite análise dos processos;

• ferramenta de gestão, aplicável aos processos da

organização (alocação de recursos).

(5)

A experiência clínica individual e a habilidade

do profissional que emprega esses

instrumentos é indispensável para uma

correta

_{interpretação}

dos resultados.

• Meta-análise recente verificou-se que o EuroSCORE e o Parsonnet apresentam melhor performance em

termos de discriminação, acurácia e calibração dentre 14 modelos de predição de risco para permanência

prolongada em UTI após cirurgia cardiovascular, apesar de ambos terem sido originalmente desenvolvidos

para predizer mortalidade.

• Apesar disso, o EuroSCORE é o mais utilizado internacionalmente, inclusive no Brasil e, de maneira geral,

tem se mostrado eficiente mesmo quando aplicado a populações não europeias,

apesar de apresentar

limitações.

Editorial

Modelos de Predição de Risco: são eles Realmente Necessários?

Risk Prediction Models: Are they Really Necessary?

Domingo M. Braile1,2,3_{, Rosangela Monteiro}3,4_{, Ricardo Brandau}3_{, Fabio B. Jatene}4

Faculdade Estadual de Medicina de São José Rio Preto1_{, São José do Rio Preto, SP; Faculdade de Ciências Médicas Unicamp}2_{, Campinas, SP;}

Escritório Editorial da Revista Brasileira de Cirurgia Cardiovascular3_{, São José do Rio Preto, SP; Instituto do Coração (InCor) HC-FMUSP}4_{, São}

Paulo, SP - Brasil

Av. Juscelino K. Oliveira, 1505 - Jardim Tarraf I - 15091-450 - São José do Rio Preto, SP - Brasil

E-mail: [email protected]

Artigo recebido em 15/11/10, revisado recebido em 15/11/10, aceito em 18/11/10.

Palavras-chave

Procedimentos cirúrgicos cardíacos/mortalidade, medição de risco/métodos, prognóstico.

Os modelos de predição de risco têm ocupado cada vez mais espaço nas publicações científicas e também no dia a dia de profissionais e instituições médico-hospitalares1-9_.

Criados inicialmente com o objetivo de analisar a probabilidade de complicações e óbitos de pacientes submetidos a intervenções, esses modelos de predição possibilitam realizar um balanço dos riscos e benefícios do procedimento. Apesar de nenhum sistema de predição ser suficientemente abrangente para estimar o resultado específico para cada paciente, a estratificação de risco possibilita a pacientes e médicos conhecerem o provável risco de complicações ou óbito para o grupo de indivíduos com perfil de risco similar, submetidos ao procedimento proposto, colaborando na tomada de decisões.

Além disso, esses modelos multivariados de avaliação de risco têm sido aplicados na comparação do desempenho de instituições ou mesmo de profissionais individualmente, configurando como uma forma objetiva de mensurar a qualidade dos serviços de saúde, e auxiliando na adequação da alocação de recursos. Ainda que sejam alvo de muitas críticas, os modelos de avaliação de risco, obviamente, são superiores à comparação de números absolutos, tais como taxas de mortalidade, na avaliação de desempenho de grupos ou hospitais.

Em sua maioria, os sistemas de predição desenvolvidos em cirurgia cardíaca foram desenvolvidos a partir de grandes populações de pacientes, resultantes, muitas vezes, de estudos multicêntricos. A partir desses dados são estabelecidos escores de risco baseados nos fatores identificados como preditores de óbito ou complicações.

O fato é que, desde o primeiro escore de risco a se tornar amplamente conhecido - o índice Parsonnet, na década

de 80 do século passado -, uma grande variedade desses instrumentos tem sido proposta, dentre eles o Cleveland Clinic score, French score, Pons score, Ontario Province score, o Society of Thoracic Surgery (STS) Scoring System, o EuroSCORE e o Bernstein-Parsonnet.

Embora não exista um modelo de estratificação de risco ideal, esse deveria reunir as seguintes características: facilidade de implementação, objetividade, acurácia na predição da mortalidade observada e ter uso difundido.

Em meta-análise recentemente publicada10_{, os autores}

verificaram que os EuroSCORE e o Parsonnet apresentam melhor performance em termos de discriminação, acurácia e calibração dentre 14 modelos de predição de risco para permanência prolongada em UTI após cirurgia cardiovascular, apesar de ambos terem sido originalmente desenvolvidos para predizer mortalidade.

Diversos centros passaram a aplicar o EuroSCORE, entretanto surgiram resultados discrepantes entre a mortalidade esperada e a observada, especialmente em pacientes de alto risco. Parolari e cols11,12_{, dentre outros autores, apontam que}

o EuroSCORE superestima a mortalidade.

Em decorrência dos avanços no cuidado perioperatório em cirurgia cardiovascular, muitos pacientes que morreriam no período em que o EuroSCORE e Parsonnet foram criados, agora sobreviveriam, mas ainda teriam alta probabilidade de desenvolver complicações. Assim, considerando-se que as técnicas cirúrgicas e de cuidado pós-operatório estão em constante evolução, bem como o perfil dos doentes, os escores devem ser dinâmicos e submetidos a constante atualização. O STS score é atualizado quase que anualmente, enquanto o EuroSCORE somente agora, mais de 10 anos após sua proposição, está sendo submetido a sua primeira revisão. Vários autores defendem que o EuroSCORE está ultrapassado e que os resultados da cirurgia melhoraram significativamente na última década, especialmente entre os idosos13-15_{. Apesar disso, o EuroSCORE é o mais utilizado}

internacionalmente, inclusive no Brasil16-28_{e, de maneira}

geral, tem se mostrado eficiente mesmo quando aplicado a populações não europeias, apesar de apresentar limitações.

Dentro desse contexto, o desenvolvimento de um escore nacional, com base no perfil de risco dos pacientes submetidos à cirurgia cardíaca no Brasil e refletindo verdadeiramente a nossa prática clínica, possivelmente será o próximo passo.

Muitos grupos ainda apresentam barreiras na adoção de escores de risco na prática clínica diária, dentre outros motivos, por preferirem empregar as informações clínicas de

677

(6)

Cirurgia Cardíaca - “Scores” de Risco

• Parsonnet Score

• Cleveland Clinic Score (1986)

• STS Score

(2007, atualmente 4.4 milhões de pacientes)

• EuroScore/EuroScore II

(aditivo e logístico)

• The Mayo Clinic Risk Score (APTC e CRM)

• Ontario Province Risk Score

• Bernstein-Parsonet Score

• Ambler Score (cirurgia valvar, UK)

(7)

STS Score

2007

• Desenvolvido para avaliar 7 tipos de

operações ou combinações de

cirurgias específicas;

• Calcula o risco de mortalidade,

permanência hospitalar, AVC,

ventilação prolongada, mediastinite,

insuficiência renal e reoperação.

I. Introduction

The Society of Thoracic Surgeons’ risk models predict the risk of operative mortality and

morbidity after adult cardiac surgery on the basis of patient demographic and clinical

variables. The models are primarily used to adjust for case mix when comparing outcomes

across institutions with different patient populations. Such comparisons are provided in the

Database reports received by STS Database participants. The STS models are also used by

physicians and patients as tools for understanding the possible risks of surgery. As these

risks are solely statistical estimates, they should be supplemented by the professional

judgment of the patients’ healthcare provider, particularly their cardiac surgeon.

This overview is provided as background to help users of the online STS risk calculator

understand and interpret the results. Throughout this document, variable short names are

used frequently. Detailed information on the STS variables, including variable short names

and clinical definitions can be found at the STS website -

http://www.sts.org

under the STS

National Database tab, Data Managers Section. Brief definitions are also available by

clicking the “definitions” link on the risk calculator web page.

II. Surgical Procedures

The STS currently has three risk models: CABG, Valve, and Valve+CABG. The models

apply to seven specific surgical procedure classifications:

CABG model

1. Isolated Coronary Artery Bypass

(CABG Only)

Valve model

2. Isolated Aortic Valve Replacement

(AV Replace)

3. Isolated Mitral Valve Replacement

(MV Replace)

4. Isolated Mitral Valve Repair

(MV Repair)

Valve+CABG model

5. Aortic Valve Replacement + CABG

(AV Replace + CABG)

6. Mitral Valve Replacement + CABG

(MV Replace + CABG)

7. Mitral Valve Repair + CABG

(MV Repair + CABG)

See Table 3 below for detailed definitions of these procedure classifications.

NOTE: A predicted risk value will NOT be calculated for any procedure that does not fall into

one of these precisely defined categories.

STS Adult Cardiac Surgery Database Risk Model Variables – Data Version 2.61

(8)

STS Score

Separado em 3 grandes coortes com mais de

100 mil pacientes cada

I. Introduction

The Society of Thoracic Surgeons’ risk models predict the risk of operative mortality and

morbidity after adult cardiac surgery on the basis of patient demographic and clinical

variables. The models are primarily used to adjust for case mix when comparing outcomes

across institutions with different patient populations. Such comparisons are provided in the

Database reports received by STS Database participants. The STS models are also used by

physicians and patients as tools for understanding the possible risks of surgery. As these

risks are solely statistical estimates, they should be supplemented by the professional

judgment of the patients’ healthcare provider, particularly their cardiac surgeon.

This overview is provided as background to help users of the online STS risk calculator

understand and interpret the results. Throughout this document, variable short names are

used frequently. Detailed information on the STS variables, including variable short names

and clinical definitions can be found at the STS website -

http://www.sts.org

under the STS

National Database tab, Data Managers Section. Brief definitions are also available by

clicking the “definitions” link on the risk calculator web page.

II. Surgical Procedures

The STS currently has three risk models: CABG, Valve, and Valve+CABG. The models

apply to seven specific surgical procedure classifications:

CABG model

1. Isolated Coronary Artery Bypass

(CABG Only)

Valve model

2. Isolated Aortic Valve Replacement

(AV Replace)

3. Isolated Mitral Valve Replacement

(MV Replace)

4. Isolated Mitral Valve Repair

(MV Repair)

Valve+CABG model

5. Aortic Valve Replacement + CABG

(AV Replace + CABG)

6. Mitral Valve Replacement + CABG

(MV Replace + CABG)

7. Mitral Valve Repair + CABG

(MV Repair + CABG)

See Table 3 below for detailed definitions of these procedure classifications.

NOTE: A predicted risk value will NOT be calculated for any procedure that does not fall into

one of these precisely defined categories.

STS Adult Cardiac Surgery Database Risk Model Variables – Data Version 2.61

(9)

III. About the Current Models

The current models were developed during the fall of 2007 using STS Adult Cardiac Surgery

Database records for surgical procedures taking place between January 1, 2002 –

December 31, 2006. Risk models were developed for the nine endpoints defined in Table 1:

Table 1. Definition of STS Risk Model Outcomes

Endpoint

Description

Operative Mortality

STS v2.61 Sequence number 3050 (MtOpD):

Operative mortality includes both (1) all deaths occurring during the hospitalization

in which the operation was performed, even if after 30 days; and (2) those deaths

occurring after discharge from the hospital, but within 30 days of the procedure

unless the cause of death is clearly unrelated to the operation.

Permanent Stroke

STS v2.61 Sequence number 2830 (CNStrokP):

Postoperative stroke (i.e., any confirmed neurological deficit of abrupt onset caused

by a disturbance in cerebral blood supply) that did not resolve within 24 hours.

Renal Failure

STS v2.61 Sequence number 2890 (CRenFail):

Acute or worsening renal failure resulting in one or more of the following:

1. Increase of serum creatinine to > 2.0, and 2x most recent preoperative creatinine

level.

2. A new requirement for dialysis postoperatively.

Prolonged

Ventilation

> 24 hours

STS v2.61 Sequence number 2860 (CPVntLng):

Prolonged pulmonary ventilator > 24 hours.

Include (but not limited to) causes such as ARDS, pulmonary edema, and/or any

patient requiring mechanical ventilation > 24 hours postoperatively.

Deep Sternal

Wound Infection

STS v2.61 Sequence number 2780 (CIStDeep):

Deep sternal infection, within 30 days postoperatively, involving muscle, bone,

and/or mediastinum REQUIRING OPERATIVE INTERVENTION.

Must have ALL of the following conditions:

1. Wound opened with excision of tissue (I&D) or re-exploration of mediastinum

2. Positive culture

3. Treatment with antibiotics.

Reoperation

for any reason

STS v2.61 Sequence numbers 2720 (CopReBld), 2730 (COpReVlv), 2740

(COpReGft), 2750 (COpReOth), 2760 (COpReNon):

Reoperation for bleeding/tamponade, valvular dysfunction, graft occlusion, other

cardiac reason, or non-cardiac reason

Major Morbidity or

Operative Mortality

A composite endpoint defined as any of the outcomes listed in the first six rows of

this table.

Short Stay:

PLOS < 6 days *

Discharged alive and within 5 days of surgery

Long Stay:

PLOS >14 days

Failure to be discharged within 14 days of surgery

*NOTE: The definition of the short patient length-of-stay endpoint differs from previous

versions of the STS risk model. In the new definition, patients must be discharged alive in

order to receive credit for a PLOS < 6 days.

STS Adult Cardiac Surgery Database Risk Model Variables – Data Version 2.61

Os modelos preditivos

do STS Score foram

desenvolvidos em 2007

contemplando os 9

“endpoints” ou

desfechos (18-110

anos).

I. Introduction

The Society of Thoracic Surgeons’ risk models predict the risk of operative mortality and

morbidity after adult cardiac surgery on the basis of patient demographic and clinical

variables. The models are primarily used to adjust for case mix when comparing outcomes

across institutions with different patient populations. Such comparisons are provided in the

Database reports received by STS Database participants. The STS models are also used by

physicians and patients as tools for understanding the possible risks of surgery. As these

risks are solely statistical estimates, they should be supplemented by the professional

judgment of the patients’ healthcare provider, particularly their cardiac surgeon.

This overview is provided as background to help users of the online STS risk calculator

understand and interpret the results. Throughout this document, variable short names are

used frequently. Detailed information on the STS variables, including variable short names

and clinical definitions can be found at the STS website -

http://www.sts.org

under the STS

National Database tab, Data Managers Section. Brief definitions are also available by

clicking the “definitions” link on the risk calculator web page.

II. Surgical Procedures

The STS currently has three risk models: CABG, Valve, and Valve+CABG. The models

apply to seven specific surgical procedure classifications:

CABG model

1. Isolated Coronary Artery Bypass

(CABG Only)

Valve model

2. Isolated Aortic Valve Replacement

(AV Replace)

3. Isolated Mitral Valve Replacement

(MV Replace)

4. Isolated Mitral Valve Repair

(MV Repair)

Valve+CABG model

5. Aortic Valve Replacement + CABG

(AV Replace + CABG)

6. Mitral Valve Replacement + CABG

(MV Replace + CABG)

7. Mitral Valve Repair + CABG

(MV Repair + CABG)

See Table 3 below for detailed definitions of these procedure classifications.

NOTE: A predicted risk value will NOT be calculated for any procedure that does not fall into

one of these precisely defined categories.

STS Adult Cardiac Surgery Database Risk Model Variables – Data Version 2.61

(10)

STS adult cardiac

database,

version 2.73. July 1, 2011

• Doença hepática

• Irradiação prévia

• Necessidade de oxigênio

• Aorta em porcelana

• Fragilidade “Frailty”

Hipoalbuminemia e estado

nutricional ?

Malignidade ?

sábado, 18 de agosto de 12

(11)

Resultados em cirurgia cardiovascular: oportunidade

para rediscutir o atendimento médico e cardiológico no

sistema público de saúde do país.

“Mortality related to cardiac surgery in Brazil, 2000-2003 (JTCVS 2008)”.

Editorial, RBCCV 2008

• interpretação dos dados (70% das cirurgias, DATASUS);

• traduz a competência das instituições e do sistema;

• a utilização dos Scores de risco possibilita a correção dos

resultados de gravidade,

se adequadamente

calibrados.

(12)

• Brasil - 350 cir/milhão hab/ano - $ 290 (7,6% PIB)

• Europa - 900 cir/milhão hab/ano

• EUA - 2000 cir/milhão hab/ano - $2.725 (15,2%, PIB)

• Argentina - $ 380 (8,9% PIB)

• Resultados precisam ser interpretados além da competência individual:

suporte organizacional, estado evolutivo da doença, diagnóstico

acurado, cuidados intensivos pré e pós-operatórios, treinamento em

todas as áreas.

Resultados em cirurgia cardiovascular: oportunidade

para rediscutir o atendimento médico e cardiológico no

sistema público de saúde do país.

Walter J. GOMES, José Teles de MENDONÇA, Domingo M. BRAILE

RBCCV 2007

(13)

• Foram coletados dados de 19 mil pacientes adultos operados

consecutivamente em 128 centros de 8 países europeus;

• Analisados 68 fatores de risco pré-op e 29 fatores operatórios;

• A relação entre os diversos fatores de risco - análise univariada e

regressão logística. Identificados 17 fatores de risco “reais” (foram

atribuidos “Scores”), classificando em três grupos de risco: baixo

(Euroscore 0-2), médio (Euroscore 3-5) e alto (Euroscore ≥6);

• Observado a definição de cada fator de risco;

• Calibração mensurada comparando a mortalidade observada com a

esperada (teste de bondade de ajuste de Homer-Lemeshow);

• Acurácia, capacidade do modelo discriminar os pacientes que foram a

óbito dos sobreviventes, foi avaliada através da estatística-C ou propoção

das previsões concordantes (0,5 a 1,0), também denominada de área sob

a curva ROC.

O EuroSCORE como preditor de mortalidade em

cirurgia cardíaca. A construção do método de

predição.

(14)

Editorial

The EuroSCORE — 10 years later. Time to change?

Keywords: Aortic valve replacement; Risk-score; Euroscore

In this issue of the journal, Di Giammarco et al., from

Chieti, Italy

[1]

, report their evaluation of the performance

of the EuroSCORE calculator in the prediction of the 30-day

outcome after isolated aortic valve replacement (AVR), ‘in

order to assess its absolute reliability and usefulness as

selection criteria to percutaneous aortic valve implantation

(PAVI).’ With this aim, they carried out a retrospective

statistical analysis on 379 of their patients consecutively

submitted to isolated AVR in the past 10 years of surgical

activity. Their observed mortality was 5.2%, significantly

lower than the 9.4% expected mortality by the logistic

EuroSCORE. The discrepancy was particularly significant in

the latter 5-year period. They conclude that the ‘EuroSCORE

appears to be a not validable model in absolute and relative

risk prediction for isolated AVR. On this basis its use in

selecting candidates to PAVI should be carefully weighted.’

The emergence of PAVI has resulted in an unprecedented use

of the EuroSCORE to predict the results of conventional aortic

valve replacement, especially in the high-risk groups, which

would, thus, become the candidates for PAVI. Large series with

mean logistic scores above 20% suddenly appeared, where they

were not suspected before. Most surgeons had not seen so many

patients considered too high a risk for surgery. I certainly had

not. This, in turn, triggered many retrospective analyses of

series with patients who were recently submitted to AVR,

which, invariably, found that the EuroSCORE overestimates the

risk of isolated AVR

[2,3]

. This is the main message of the article

by Di Giammarco et al., but only this year there have been

another half a dozen papers with identical conclusions.

The group from Bad Oeynhausen, Germany, is about to

publish one based on 2757 patients submitted to AVR recently

in their centre

[4]

. Similarly, concluding that ‘patient

selection for interventional AVR cannot be based on the

EuroSCORE, because it lacks discrimination and

centre-specific calibration.’ These authors thus recommend

‘indi-vidual, surgical judgement that weighs institutional

exper-tise of high risk patients against possible reduction of

mortality by using interventional techniques.’ Although not

usually recognised, we all know that expertise, hence the

results thereof, vary from group to group and, within these,

from surgeon to surgeon. No risk score can cover all these

differences. Similarly, an analysis of a subgroup of 6305

patients submitted to isolated AVR registry of the German

Society of Thoracic and Cardiovascular Surgery from 2006 and

2007 revealed an overall hospital mortality of 3.9% whereas

the logistic EuroSCORE predicted 7.3%, which supports ‘the

substantial lack of predictive value of the EuroSCORE

[5]

.’

Some other articles also recently published were directed at

validating the EuroSCORE in parts of the world other than

Europe and their conclusions are that the EuroSCORE does not

apply there

[6—9]

. However, I believe, it also does not apply in

Europe. For three main reasons: firstly, the EuroSCORE is

already outdated, as it was developed from data on patients

operated on almost a decade and a half ago, and the results of

surgery have improved significantly since, especially in the

elderly. Secondly, because the data originated from only eight

European countries and, from each one of these, only few

centres contributed. As with any type of statistical analysis, it

generated mean values which may even not serve all these

centres. As stated recently

[10]

, ‘the logistic EuroSCORE risk

stratification system was developed and validated within the

European population. There should be caution in the utilisation

of any particular risk stratification system outside the countries

of origins, and it is important to carefully evaluate the validity

of such system amongst foreign population,’ which means that

it may even not be applicable to all European countries.

Thirdly, and most important, the EuroSCORE was especially

developed for cardiac surgery in general, especially for

coronary re-vascularisation procedures, the majority of data

belonging to this group of patients, and not specifically for AVR.

Hence, it is now generally recognised that the EuroSCORE

does not reflect current results of cardiac surgery, including

coronary re-vascularisation, but especially AVR. Curiously, the

additive model of the EuroSCORE, initially published in 1999

[11]

, was modified in 2003 by the introduction of the logistic

model

[12]

, which is now the most widely used, but whose

values are even higher than those of the additive score.

Furthermore, it is important to stress that the model would not

apply, in any case, in any situation, before validation, to each

particular surgical group. We, in Coimbra, have recognised

that and have, recently, developed our own scores for

prediction of the risks of mortality and morbidity after

coronary surgery in our population and are about to do the

same for valve surgery. Others have gone in a similar path.

www.elsevier.com/locate/ejcts

European Journal of Cardio-thoracic Surgery 37 (2010) 253—254

1010-7940/$ — see front matter # 2009 European Association for Cardio-Thoracic Surgery. Published by Elsevier B.V. All rights reserved.

doi:

10.1016/j.ejcts.2009.08.017

‘EuroSCORE appears to be a not validable model in absolute and relative

risk prediction for isolated AVR. On this basis its use in selecting

candidates to PAVI should be carefully weighted.’

.

The group from Bad Oeynhausen, Germany, is about to publish one study

based on 2757 patients subm Similarly, concluding that ‘patient selection

for interventional AVR cannot be based on the EuroSCORE, because it

lacks discrimination and centre- specific calibration.’

The additive model of the EuroSCORE, initially published in 1999, was

modified in 2003 by the introduction of the logistic model, which is now

the most widely used, but whose values are even higher than those of

the additive score.Furthermore,

it is important to stress that the model

would not apply, in any case, in any situation, before validation, to each

particular surgical group.

Manuel J. Antunes*

Cardiothoracic Surgery, University Hospitals,

Coimbra.

Portugal

(15)

• Controle de qualidade em cirurgia é necessário para que se obtenha

de uma operação planejada o resultado esperado, em concordância

com o conhecimento médico atual.

• O primeiro passo para o controle de qualidade em cirurgia é a

configuração de uma base de dados.

• Os hospitais precisam ter os dados dos procedimentos nele

realizados, para promover melhora dos resultados cirúrgicos, corrigir

defeitos técnicos e estruturais e diminuir gastos desnecessários.

• As sociedades médicas são guardiãs das especialidades e por meio de

análise de dados podem localizar erros e corrigi-los.

• O controle de qualidade em cirurgia cardíaca depende da equipe

cirúrgica, da unidade hospitalar e do sistema de saúde.

Controle de qualidade em cirurgia

cardiovascular: um paradigma a ser atingido.

Quality control in cardiovascular surgery: a new paradigm

Rev Bras Cir Cardiovasc, 2007

(16)

• Operative risk was calculated using the STS Predicted Risk of Mortality, the EuroSCORE logistic and

additive algorithms, and the Ambler Risk Score.

• CONCLUSION:

The STS Predicted Risk of Mortality most accurately predicted peri- operative and

long-term mortality for the highest risk patients having aortic valve replacement.

Reliability of risk algorithms in predicting early and late operative outcomes in high-risk

patients undergoing aortic valve replacement.

The Journal of Thoracic and Cardiovascular Surgery 2008

Objective: Risk algorithms were used to identify a high-risk population for trans- catheter aortic valve

implantation instead of standard aortic valve replacement in patients with aortic stenosis. We evaluated

the efficacy of these methods for predicting outcomes in high-risk patients undergoing aortic valve

replacement.

Mean predicted operative mortality for the STS-PROM

high-risk group was 13.31% compared with 50.87%,

14.04%, and 19.03% for the LES, AES, and ARS high-risk

cohorts, respectively (

Table 3

).

Measured operative mortality was 18.75% (12/64) in the

STS-PROM high-risk cohort, 15.63% (10/64) and 11.90%

(10/84) in the LES and AES groups, respectively, and

13.40% (13/97) in the ARS group. The observed/expected

ratios calculated for perioperative mortality were 0.31, 0.85,

and 0.71 in the LES, AES, and ARS cohorts, respectively.

The STS-PROM group had an observed/expected mortality

ratio of 1.41. Longer-term follow-up showed that 45.3%

(29/64) of the patients identified by the STS-PROM as

being at high risk died within the follow-up period. Within

the LES and AES groups, 29.7% (19/64) and 33.3% (28/

84), respectively, of the patients identified as being at high

risk died during the follow-up period. Only 26.8% (26/97)

of the ARS patients were noted as having died. The

con-TABLE 2. Preoperative demographics of high-risk patients identified by various risk algorithms

STS-PROM (64) % LES (64) % AES (84) % ARS (97) % P value

Age (y)

79.5 ! 8.9

79.3 ! 6.3

78.6 ! 7.4

77.3 ! 7.7

.22

Ejection fraction

45.7 ! 14.6

43.0 ! 14.1

43.7 ! 14.2

41.1 ! 14.8

.31

Female gender

54.7

45.3

47.6

46.4 .70

Diabetes

42.2

26.6

31.0

41.2 .13

Renal failure

26.6

20.3

19.1

10.3 .062

Renal failure requiring dialysis

13.2

8.8

9.8

7.3 .85

Cerebrovascular accident

22.2

38.1

28.9

13.5 .004

Chronic lung disease

23.1

12.5

15.8

7.7 .82

Peripheral vascular disease

34.4

37.5

35.7

15.5 .004

Cerebrovascular disease

29.7

40.6

38.1

18.6 .008

Previous CABG

44.4

57.1

51.8

41.7 .22

Myocardial infarction

18.8

25.0

23.8

16.5 .49

Congestive heart failure

75.0

67.2

64.3

65.0 .51

Cardiogenic shock

6.3

4.7

3.6

3.1 .78

Arrhythmia

35.9

32.8

32.1

42.3 .48

NYHA class (III-IV)

71.9

60.9

63.1

64.6 .33

Status

Elective

31.3

53.1

56.0

43.3 .075

Emergency

1.6

1.2

3.1 Urgent

67.2

45.3

42.9

53.6 Intra-aortic balloon pump

14.1

17.2

16.7

11.5 .70

STS-PROM, The Society of Thoracic Surgeons Predicted Risk of Mortality; LES, EuroSCORE logistic; AES, EuroSCORE additive; ARS, Ambler Risk Score; CABG, coronary artery bypass grafting; NYHA, New York Heart Association.

TABLE 3. Risk calculations and mortality rates for four risk algorithms

All patients (638) STS-PROM LES AES ARS

Overall mean/minimum–maximum

4.26% 0.91–46.82

13.21% 1.51–93.32

7.46% 2–22

6.99% 0.4–39.7

High-risk groups

STS-PROM (n " 64)

LES (n " 64)

AES (n " 84)

ARS (n " 97)

90th percentile risk break point

8.38%

33.47%

12%

14.3%

Expected (predicted) mortality in top 10%

(mean/ minimum–maximum)

13.31% 8.38%–46.8% 50.87% 33.47%–93.32% 14.04% 12.0%–22.0% 19.03% 14.3%–39.7%

Observed perioperative mortality

18.75% (12/64)

15.63% (10/64)

11.90% (10/84)

13.40% (13/97)

Observed/expected (perioperative)

mortality

1.41

0.31

0.85

0.71 Observed late mortality

45.3% (29/64)

29.7% (19/64)

33.3% (28/84)

26.8% (26/97)

Observed overall mortality

64.06% (41/64)

45.31% (29/64)

45.24% (38/84)

40.21% (39/97)

Kaplan–Meier survival estimate at 5 y

38.5% # 6.7%

53.7% # 7.2%

49.8% # 6.6%

55.1% # 6.2%

Overall mortality odds ratio,

8.06

3.27

3.45

2.76 High/low risk group

(4.63–14.01)

(1.92–5.57)

(2.14–5.57)

(1.75–4.36)

STS-PROM, The Society of Thoracic Surgeons Predicted Risk of Mortality; LES, EuroSCORE logistic; AES, EuroSCORE additive; ARS, Ambler Risk Score; CABG, coronary artery bypass grafting; NYHA, New York Heart Association.

Dewey et al

Evolving Technology

The Journal of Thoracic and Cardiovascular Surgery

●

Volume 135, Number 1 183

ET

(17)

• Variáveis disponíveis, entretanto

ignoradas, posteriormente observou-se

estarem associadas à mortalidade.

• Variáveis raras em alguns subgrupos de

pacientes.

• Variáveis que representam subgrupos

totalmente ausentes do conjunto de

pacientes sobre os quais os modelos

foram desenvolvidos.

Evolving

Technology

Reliability of risk algorithms in predicting early and late

operative outcomes in high-risk patients undergoing aortic

valve replacement

Todd M. Dewey, MDa_{, David Brown, MD}a_{, William H. Ryan, MD}a_{, Morley A. Herbert, PhD}b_{, Syma L. Prince, RN}a_,

and Michael J. Mack, MDa

Objective: Risk algorithms were used to identify a high-risk population for

trans-catheter aortic valve implantation instead of standard aortic valve replacement in patients with aortic stenosis. We evaluated the efficacy of these methods for predicting outcomes in high-risk patients undergoing aortic valve replacement.

Methods: Data were collected on 638 patients identified as having isolated aortic

valve replacement between January 1, 1998 and December 31, 2006, using The Society of Thoracic Surgeons (STS) database. Long-term survival was determined from the Social Security Death Index or family contact. Operative risk was calcu-lated using the STS Predicted Risk of Mortality, the EuroSCORE logistic and additive algorithms, and the Ambler Risk Score. Patients at or above the 90th percentile of risk (8.38% for STS, 33.47% for logistic, 12% for additive, 14.3% for Ambler) were identified as high risk. We then compared actual with predicted mortality and each algorithm’s ability to identify patients with the worst long-term survival.

Results: Operative mortality was 24 of 638 (3.76%). An additional 121 (19.0%)

patients died during the follow-up study period (mean 4.2 ! 2.7 years). Overall mortality was 145 of 638 (22.7%). Expected versus observed mortality for the high-risk group by algorithm was 13.3% versus 18.8% for STS, 50.9% versus 15.6% for logistic, 14.0% versus 11.9% for additive, and 19.0% versus 13.4% by Ambler. Long-term mortality, per high-risk group, was 64.1% in the STS Predicted Risk of Mortality, 45.3% in the logistic, 45.2% in the additive, and 40.2% in Ambler Risk Score. Logistic regression showed that the STS algorithm was the most sensitive in defining the patients most at risk for long-term mortality.

Conclusion: The STS Predicted Risk of Mortality most accurately predicted

peri-operative and long-term mortality for the highest risk patients having aortic valve replacement.

A

ortic stenosis (AS) is the most common valvular abnormality encountered in the United States, with a reported incidence of approximately 5 of every 10,000 adults.1_{The prevalence of AS is expected to markedly increase as}

the US population ages, with aortic valve calcification and stenosis affecting the health of larger numbers of patients.2-4

From the Cardiopulmonary Research Sci-ence and Technology Institute (CRSTI)a

and the Medical City Dallas Hospitalb_,

Dal-las, Tex.

Read at the Eighty-seventh Annual Meeting of The American Association for Thoracic Surgery, Washington, DC, May 5-9, 2007. Received for publication May 4, 2007; re-visions received Aug 29, 2007; accepted for publication Sept 12, 2007.

Address for reprints: Todd M. Dewey, MD, 7777 Forest Ln, Suite A323, Dallas, TX 75230 (E-mail: [email protected]). J Thorac Cardiovasc Surg 2008;135:180-7 0022-5223/$34.00

doi:10.1016/j.jtcvs.2007.09.011

180 The Journal of Thoracic and Cardiovascular Surgery●January 2008