ORIGINAL ARTICLES: CONGENITAL HEART SURGERY

Accepted for pu

Presented at thSurgeons, Orlan

Address correspCardiovascular PFifth St S, Ste 60

� 2014 by ThePublished by

CONGENITAL HEART SURGERY:

TheAnnals of Thoracic SurgeryCMEProgramis locatedonlineathttp://www.annalsthoracicsurgery.org/cme/home. To take the CME activity related to this article, you must have either an STS member or an individual non-member subscription to the journal.

ART

CONGEN

ITALHE

The Importance of Patient-Specific PreoperativeFactors: An Analysis of The Society of ThoracicSurgeons Congenital Heart Surgery DatabaseJeffrey Phillip Jacobs, MD, Sean M. O’Brien, PhD, Sara K. Pasquali, MD, MHS,Sunghee Kim, PhD, J. William Gaynor, MD, Christo Ivanov Tchervenkov, MD,Tara Karamlou, MD, Karl F. Welke, MD, Francois Lacour-Gayet, MD,Constantine Mavroudis, MD, John E. Mayer, Jr, MD, Richard A. Jonas, MD,Fred H. Edwards, MD, Frederick L. Grover, MD, David M. Shahian, MD, andMarshall Lewis Jacobs, MDJohns Hopkins All Children’s Heart Institute, Saint Petersburg, Florida; Division of Cardiac Surgery, Department of Surgery, JohnsHopkins University School of Medicine, Baltimore, Maryland; Duke Clinical Research Institute, Duke University School of Medicine,Durham, North Carolina; Department of Pediatrics and Communicable Diseases, C. S. Mott Children’s Hospital, University ofMichigan, Ann Arbor, Michigan; Children’s Hospital of Philadelphia, Philadelphia, Pennsylvania; Montreal Children’s Hospital, McGillUniversity, Montreal, Quebec, Canada; Division of Pediatric Cardiac Surgery, Benioff Children’s Hospital, University of California SanFrancisco, San Francisco, California; Children’s Hospital of Illinois, Peoria, Illinois; Royal Brompton Hospital, London, United Kingdom;Children’s Hospital Boston, Harvard University Medical School, Boston, Massachusetts; Children’s National Heart Institute, Children’sNational Medical Center, Washington, DC; Shands Jacksonville, University of Florida, College of Medicine–Jacksonville, Jacksonville,Florida; University of Colorado Denver, School of Medicine, Aurora, Colorado; and Massachusetts General Hospital, Harvard MedicalSchool, Boston, Massachusetts

Background. The most common forms of risk adjust-ment for pediatric and congenital heart surgery used todayare based mainly on the estimated risk of mortality of theprimary procedure of the operation. The goals of thisanalysis were to assess the association of patient-specificpreoperative factors with mortality and to determinewhich of these preoperative factors to include in futurepediatric and congenital cardiac surgical risk models.

Methods. All index cardiac operations in The Societyof Thoracic Surgeons Congenital Heart Surgery Database(STS-CHSD) during 2010 through 2012 were eligible forinclusion. Patients weighing less than 2.5 kg undergoingpatent ductus arteriosus closure were excluded. Centerswith more than 10% missing data and patients withmissing data for discharge mortality or other key vari-ables were excluded. Rates of discharge mortality forpatients with or without specific preoperative factorswere assessed across age groups and were comparedusing Fisher’s exact test.

blication July 9, 2014.

e Fiftieth Annual Meeting of The Society of Thoracicdo, FL, Jan 25–29, 2014.

ondence to Dr Jeffrey P. Jacobs, MD, Andrews/Daicoffrogram, Johns Hopkins All Children’s Heart Institute, 6017, Saint Petersburg, FL 33701; e-mail: jeffjacobs@msn.com.

Society of Thoracic SurgeonsElsevier

Results. In all, 25,476 operations were included (over-all discharge mortality 3.7%, n [ 943). The prevalence ofcommon preoperative factors and their associations withdischarge mortality were determined. Associations of thefollowing preoperative factors with discharge mortalitywere all highly significant (p < 0.0001) for neonates, in-fants, and children: mechanical circulatory support, renaldysfunction, shock, and mechanical ventilation.Conclusions. Current STS-CHSD risk adjustment is

based on estimated risk of mortality of the primary pro-cedure of the operation as well as age, weight, and pre-maturity. The inclusion of additional patient-specificpreoperative factors in risk models for pediatric andcongenital cardiac surgery could lead to increased preci-sion in predicting risk of operative mortality and com-parison of observed to expected outcomes.

(Ann Thorac Surg 2014;98:1653–9)� 2014 by The Society of Thoracic Surgeons

eaningful analysis of outcomes requires adjustment

Mfor differences in case-mix across hospitals. The importance of this concept is magnified in the current eraof public reporting of outcomes. Public reporting withoutrisk adjustment is misleading and can lead to risk aver-sive practices, with efforts made to avoid caring for the

Dr Mayer discloses a financial relationship withMedtronic.

0003-4975/$36.00http://dx.doi.org/10.1016/j.athoracsur.2014.07.029

1654 JACOBS ET AL Ann Thorac SurgREFINING RISK ADJUSTMENT 2014;98:1653–9

CONGEN

ITALHEA

RT

sickest of patients who might benefit the most fromcardiac surgery [1, 2].

The most common forms of risk adjustment for analysisand reporting of outcomes from pediatric and congenitalcardiac surgery in use today are based mainly on the esti-mated risk of mortality of the primary procedure of theoperation [3–5]. With increased availability of robust clin-ical data, it should now be possible to add a variety ofspecific patient characteristics to pediatric and congenitalcardiac surgical risk models, including prematurity, chro-mosomal abnormalities (such as trisomy 18 and trisomy21), syndromes (such as Marfan and Noonan), noncardiaccongenital anatomic abnormalities (such as omphaloceleand congenital diaphragmatic hernia), and preoperativefactors (such as preoperative mechanical circulatory sup-port and preoperative mechanical ventilation to treatcardiorespiratory failure). The term “preoperative factors”is used rather than “preoperative risk factors” because notall of these preoperative factors are associated with risk.Importantly, the relative degree of risk attributable to eachof these preoperative factors depends on the age groupsand procedural groups under consideration, and on theendpoints being considered (be they mortality, length ofstay, complications, or others). The purpose of this analysisis to assess the association of patient-specific preoperativefactors with mortality, and to determine which of thesepreoperative factorsmaybe considered for future inclusionin pediatric and congenital cardiac surgical risk models.

Patients and Methods

Data SourceThe Society of Thoracic Surgeons Congenital Heart Sur-gery Database (STS-CHSD) was used for this study. The

Table 1. Inclusionary and Exclusionary Criteria Applied to Obtain

Inclusionary and Exclusionary Rules Applied

Start with all operation in STS-CHSD through January 1, 20131. Exclude PDA closure in 2.5 kg or less and exclude organ

procurement2. Include only cardiac operations3. Include only index operation of a hospitalization4. Include neonates, infants, children, adults5. Include surgery year 2010, 2011, or 20126. Include data version 3.0 only7. Include only demographic data collected with version 3.08. Include operations with STAT Mortality Score9. Exclude operation with missing mortality data

10. Exclude STS database participants with >10% missing mortalityrate (after above steps)

11. Exclude STS database participants with >10% missing rate ofpreoperative factors, noncardiac congenital anatomicabnormalities, chromosomal abnormalities, syndromes,prematurity (neonates and infants), or number of priorcardiothoracic operations (after above steps)

CHSD ¼ Congenital Heart Surgery Database; PDA ¼ patent ductus arteriofor Cardio-Thoracic Surgery; STS ¼ The Society of Thoracic Surgeons.

STS-CHSD is the largest database in the world of patientswho have undergone congenital and pediatric cardiacsurgical operations. It is a voluntary registry that containspreoperative, operative, and outcomes data for all pa-tients undergoing congenital and pediatric cardiovascularoperations at participating centers. As of January 2014, theSTS-CHSD contains data on more than 292,000 surgeriesconducted since 1998 at 120 hospitals in North America,representing more than 90% of all US centers performingcongenital heart surgery and more than 90% of all pedi-atric and congenital cardiac operations in the UnitedStates [6]. Coding for this database is accomplished byclinicians and ancillary support staff using the Interna-tional Pediatric and Congenital Cardiac Code [7, 8] and isentered into the contemporary version of the STS-CHSDdata collection form (version 3.0) [9]. The definitions of allterms and codes from the STS-CHSD used in this article,including all of the preoperative factors evaluated in thisanalysis, have been standardized and published [9].Evaluation of data quality includes the intrinsic verifica-tion of data, along with a formal process of in-personsite visits and data audits at approximately 10% of allparticipating centers each year conducted by a panel ofindependent quality personnel and pediatric cardiac sur-geons [10]. The Duke Clinical Research Institute serves asthe data warehouse and analytic center for all STS nationaldatabases. Approval for the study was obtained from theDuke University Medical Center Institutional ReviewBoard as well as from the Quality Measurement Task Forceof the STS Workforce on National Databases.

Study PopulationAll index cardiac operations in the STS-CHSD during2010 through 2012 were eligible for inclusion. Patients

Study Cohort

RecordsExcluded

RemainingRecords

No. ofPatients

No. ofParticipants

. 256,854 168,776 11112,140 244,714 157,078 111

43,330 201,384 143,362 11128,935 172,449 143,253 111

17 172,432 143,240 111119,960 52,472 47,352 104

0 52,472 47,352 1048,379 44,093 39,658 1041,348 42,745 38,685 1041,919 40,826 36,924 1044,916 35,910 32,437 93

10,434 25,476 23,019 72

sus; STAT ¼ The Society of Thoracic Surgeons–European Association

Table2.

Prevalence

ofCom

mon

Preoperativ

eFa

ctorsan

dTh

eirAssociatedDischarge

Mortality

Poten

tial

Preop

erativeRiskFa

ctors

Neo

nates

(n¼

5,630)

Infants

(n¼

9,500)

Child

ren(n

¼8,742)

Adults(n

¼1,604)

MortalityRatea

(pValue)

MortalityRatea

(pValue)

MortalityRatea

(pValue)

MortalityRatea

(pValue)

Ove

rallmortalityrate

9.88

%2.91%

0.92%

1.93%

Preop

erative/preproceduralmechan

ical

circulatory

supportb

20/37¼

54.1%

(<0.0001

)12

/49¼

24.5%

(<0.00

01)

13/48¼

27.1%

(<0.00

01)

0/4¼

0.0%

(0.999

9)Sh

ock,

persisten

tat

timeof

surgery

42/125

¼33

.6%

(<0.0001

)9/47

¼19

.2%

(<0.00

01)

9/29

¼31

.0%

(<0.00

01)

1/4¼

25.0%

(0.075

2)Ren

aldysfunction

31/119

¼26

.1%

(<0.0001

)17

/79¼

21.5%

(<0.00

01)

6/53

¼11

.3%

(<0.00

01)

2/26

¼7.7%

(0.088

2)Mechan

ical

ventilation

totrea

tcard

ioresp

iratoryfailu

re27

1/17

92¼

15.1%

(<0.0001

)83

/723

¼11

.5%

(<0.00

01)

21/146

¼14

.4%

(<0.00

01)

3/12

¼25

.0%

(0.001

3)Gastrostomypresent

3/18

¼16

.7%

(0.413

2)41

/654

¼6.3%

(<0.00

01)

13/364

¼3.6%

(<0.00

01)

0/5¼

0.0%

(0.999

9)Sh

ock,

resolved

attimeof

surgery

63/421

¼15

.0%

(0.000

6)10

/179

¼5.6%

(0.040

7)8/59

¼13

.6%

(<0.00

01)

1/7¼

14.3%

(0.127

9)Coa

gulation

disorder,h

ypoc

oagu

lablestateseco

ndary

tomed

ication

5/11

¼45

.5%

(0.002

6)5/44

¼11

.4%

(0.008

6)2/50

¼4.0%

(0.076

4)1/36

¼2.8%

(0.508

6)

Hyp

othyroidism

6/23

¼26

.1%

(0.021

1)9/190¼

4.7%

(0.125

8)4/113¼

3.5%

(0.019

7)2/66

¼3.0%

(0.367

4)Preop

erativeneu

rologicdefi

cit

10/47¼

21.3%

(0.022

)12

/155

¼7.7%

(0.001

8)8/246¼

3.3%

(0.001

8)4/47

¼8.5%

(0.011

4)Preop

erativeco

mplete

atriov

entricularbloc

k10

/50¼

20.0%

(0.027

6)2/47

¼4.3%

(0.398

4)5/161¼

3.1%

(0.015

8)0/57

¼0.0%

(0.624

7)Stroke

,CVA,o

rintracranialhem

orrh

age

grad

e>2duringlifetim

e7/41

¼17

.1%

(0.117

9)10

/100

¼10

.0%

(0.000

6)4/102¼

3.9%

(0.014

)2/45

¼4.4%

(0.215

2)

Seizure

duringlifetim

e7/51

¼13

.7%

(0.342

6)9/187¼

4.8%

(0.120

4)10

/291

¼3.4%

(0.000

3)3/48

¼6.3%

(0.062

9)

aNum

berof

deaths

divide

dby

numbe

rwith

factor

(n/n)e

qualsmortalityrate.

bSu

chas

intraaortic

ballo

onpu

mp,

ventricu

larassist

device,extracorporealm

embran

eox

ygen

ation,

orcardiopu

lmon

arysu

pport.

CVA

¼cerebrov

ascu

laracciden

t.

1655Ann Thorac Surg JACOBS ET AL2014;98:1653–9 REFINING RISK ADJUSTMENT

CONGEN

ITALHEA

RT

weighing less than 2.5 kg undergoing isolated closure ofpatent ductus arteriosus (PDA) were excluded. Patientswith missing data for discharge mortality were alsoexcluded from the analysis. Centers with more than 10%missing data for discharge mortality or other key vari-ables including preoperative factors were also excluded.Table 1 documents the inclusionary and exclusionarycriteria applied to obtain the final study cohort. The finalstudy cohort included 25,476 index cardiac operationsperformed in 23,019 patients at 72 centers from January 1,2010, to December 31, 2012.

Data CollectionData collected included patient age, preoperative factors,primary procedure, and discharge mortality. The STS-CHSD includes variables pertaining to a variety of pa-tient characteristics including age, weight, age-for-weightZ score, prematurity, chromosomal abnormalities, syn-dromes, noncardiac congenital anatomic abnormalities,and preoperative factors. This analysis focuses on theassociation of patient-specific preoperative factors withdischarge mortality. This analysis does not evaluateother patient characteristics, including age, weight, age-for-weight Z score, prematurity, chromosomal abnor-malities, syndromes, and noncardiac congenital anatomicabnormalities.

Risk models for future use in the STS-CHSDwill include a variety of categories of patient character-istics, including prematurity, chromosomal abnormal-ities, syndromes, noncardiac congenital anatomicabnormalities, and preoperative factors. The presentstudy is confined to the factors coded in the STS-CHSDunder the heading “preoperative factors,” which arepatient preoperative “status’ factors such as preopera-tive mechanical circulatory support and preoperativerenal dysfunction, among many others. These preoper-ative status factors can be distinguished from other pa-tient characteristics, including patient-related geneticand structural factors such as chromosomal abnormal-ities, syndromes, and noncardiac congenital anatomicabnormalities. Ultimately, plans are eventually toinclude in future risk models in the STS-CHSD boththese preoperative status factors and other patientcharacteristics, including patient-related genetic andstructural factors such as chromosomal abnormalities,syndromes, and noncardiac congenital anatomic abnor-malities. (A separate article is now in preparation thatdiscusses the details of incorporation of these patient-related genetic and structural factors into such a riskmodel.)

Data AnalysisDischarge mortality was determined for patients with orwithout preoperative factors in the STS-CHSD. Of 34preoperative factors for which data are collected in STS-CHSD version 3.0, specific preoperative factors wereincluded in the analysis if their age group–specific prev-alence was greater than 2% or if the number of associateddeaths was 20 or more (Table 2). These cutoffs (or“thresholds for inclusion”) were selected based on the

1656 JACOBS ET AL Ann Thorac SurgREFINING RISK ADJUSTMENT 2014;98:1653–9

CONGEN

ITALHEA

RT

expert consensus of the investigative team to eliminaterarely occurring preoperative factors that are rarelyassociated with mortality before discharge. The rationalefor these thresholds for inclusion was to avoid a situationwherein information that might be considered for inclu-sion in a future risk model might reflect the influence of asmall number of events that could conceivably haveoccurred at random. The actual death rates of patientswith each preoperative factor and patients withouteach preoperative factor were compared using Fisher’sexact test.

Institutional Review Board ApprovalThe Duke University Health System Institutional ReviewBoard approved the study and provided a waiver ofinformed consent. Although the STS data used in theanalysis contain patient identifiers, they were originallycollected for nonresearch purposes and the risk to pa-tients was deemed to be minimal [11].

Results

The final analysis included 25,476 operations, with anoverall discharge mortality of 3.7% (n ¼ 943). Table 2shows the prevalence of preoperative factors that wereanalyzed and their associated discharge mortality. Theassociations of discharge mortality with the followingpreoperative factors were all highly significant (p <0.0001) for all three pediatric age categories (neonates,infants, and children): mechanical circulatory support,renal dysfunction, shock, and mechanical ventilation.

Other individual risk factors were highly significant(p < 0.0001) in one or more age groups, but not in all threepediatric age categories. The only preoperative factorsfound significant (p < 0.05) for adults with congenitalheart disease were preoperative mechanical ventilationand preoperative neurological deficit. Meanwhile, anumber of other preoperative factors, such as preopera-tive gastrostomy, were significant for infants and chil-dren, but not for neonates.

Comment

Current risk adjustment in the STS-CHSD is based onestimated risk of mortality of the primary procedureof the operation, as well as on age, weight, and prema-turity. Our analysis has demonstrated the significant as-sociation of certain patient-specific preoperative riskfactors with discharge mortality after surgery for pediatricand congenital cardiac disease. Based on our analysis, theinclusion of additional patient-specific preoperative fac-tors in risk models for pediatric and congenital cardiacsurgery could lead to increased precision in predictingrisk of operative mortality and comparison of observed toexpected outcomes.

The importance of risk adjustment derives fromthe inadequacy of an analysis of outcomes using rawmeasurements of mortality, without adjustment forcomplexity. The mix of cases can vary greatly from pro-gram to program. Without risk adjustment, the analysis of

outcomes will be flawed. Therefore, the analysis of out-comes after surgery requires a reliable method of esti-mating the risk of adverse events based on case mix.However, formal risk modeling is challenging when manyof the individual operative procedures are performed insmall numbers in the entire cohort and very rarely in manyindividual centers. Complexity stratification provides analternative methodology that can facilitate the analysis ofoutcomes of rare operations. Complexity stratification is amethod of analysis in which the data are divided intorelatively homogeneous groups (called strata). The dataare analyzed within each stratum.The earliest forms of risk adjustment used by the

STS-CHSD were based on complexity stratification. Thedata are analyzed and reported within each stratum.The STS-CHSD currently uses three methods of proce-dural complexity stratification: (1) The Society of ThoracicSurgeons–European Association for Cardio-Thoracic Sur-gery (EACTS) Congenital Heart Surgery Mortality Cate-gories (STAT Mortality Categories); (2) Aristotle BasicComplexity (ABC) Levels; and (3) Risk Adjustment forCongenital Heart Surgery-1 (RACHS-1) Categories. Thesethree methods provide three different ways of groupingtypes of pediatric and congenital cardiac operations ac-cording to their estimated risk or complexity. The STATMortality Categories are empirically derived based on datain the STS and EACTS congenital heart surgery databasesand use five categories; the STAT Mortality Categoriesserve as the main complexity adjustment tool for theSTS-CHSD. The ABC method uses four categories. TheRACHS-1 method uses six categories, but functionally hasfive categories when applied to the STS-CHSD.The ABC levels were introduced into the STS-CHSD

in 2002. The ABC score is a measure of proceduralcomplexity that was developed by the EACTS/STS Aris-totle Committee and is based on expert opinion regardingthe potential for mortality, the potential for morbidity,and the technical difficulty of the operation [3, 4]. TheRACHS-1 categories were introduced into the STS-CHSDin 2006. The RACHS-1 categories are procedure-drivencategories developed to adjust for baseline case mix dif-ferences when comparing discharge mortality for groupsof patients undergoing pediatric congenital heart surgery.The RACHS-1 method was created using a combinationof judgment-based and empirical methodology [4]; it usesprocedural and patient level information as componentsto provide adjustment for the influence of differences incase mix on postoperative surgical mortality. The proce-dural component contains congenital cardiac surgicalprocedures categorized into six categories. The secondcomponent of the RACHS-1 method contains patientcharacteristics that may influence pediatric cardiac sur-gical outcomes, including age at surgery, prematurity(defined as less than 36 weeks) and major noncardiacstructural abnormalities (such as tracheoesophageal fis-tula) or major chromosomal abnormalities or syndromes(such as DiGeorge syndrome).The RACHS-1 and the ABC scores were developed at a

time when limited multiinstitutional clinical data wereavailable, and were, therefore, based in large part on

1657Ann Thorac Surg JACOBS ET AL2014;98:1653–9 REFINING RISK ADJUSTMENT

CONGEN

ITALHEA

RT

subjective probability (expert opinion). With the in-creasing availability of multiinstitutional clinical data,the STAT Mortality Categories were introduced into theSTS-CHSD in 2010. The STAT Mortality Categories arean empirically derived methodology of complexity strat-ification based on statistical estimation of the risk ofmortality from an analysis of objective data from the STS-CHSD and the EACTS Congenital Heart Surgery Data-base (EACTS-CHSD) [5]. The STAT Mortality Categoriesare a tool for complexity stratification that was developedfrom an analysis of 77,294 operations entered into theEACTS-CHSD (33,360 operations) and the STS-CHSD(43,934 patients) between 2002 and 2007. Procedure-specific mortality rate estimates were calculated using aBayesian model that adjusted for small denominators.Operations were sorted by increasing risk and groupedinto five categories (STAT Mortality Categories) that weredesigned to be optimal with respect to minimizingwithin-category variation and maximizing between-category variation. The STS and EACTS have transi-tioned from the primary use of Aristotle and RACHS-1 tothe primary use of the STAT Mortality Categories forthree reasons: (1) the STAT Mortality Score and Cate-gories were developed primarily based on objective datawhereas RACHS-1 and Aristotle were developed pri-marily based on expert opinion (subjective probability);(2) the STAT Mortality Score and Categories allow forclassification of more operations than RACHS-1 or Aris-totle; and (3) the STAT Mortality Score and Categorieshave a higher c-statistic than RACHS-1 or Aristotle [5].

One weakness of current systems of risk stratificationused for pediatric and congenital cardiac surgery is thelimited adjustment made for patient-specific factors. Withthe increased availability of verified clinical data, it is nowpossible to incorporate patient-specific factors into therisk models in the STS-CHSD. This analysis demon-strated which preoperative factors may be most impor-tant to consider when adjusting for patient-specific risk;these include mechanical circulatory support, renaldysfunction, shock, and mechanical ventilation. A bundleof these major preoperative risk factors could be incor-porated into future risk models for pediatric andcongenital cardiac surgery.

Future DirectionsOne important result of this analysis is that beginningwith the Spring 2014 STS-CHSD feedback report, newSTS-CHSD risk models will be used that will include anumber of new patient-specific characteristics, includingchromosomal abnormalities, syndromes, noncardiaccongenital anatomic abnormalities, and preoperativefactors. The present analysis provides the rationale forselection of individual preoperative factors to be includedin these models. These new STS-CHSD risk modelswill improve the ability of the STS-CHSD to be used asa tool to improve the quality of surgical care deliveredto patients with pediatric and congenital cardiac disease[12–14].

In this analysis, the only preoperative factors foundsignificant for adults with congenital heart disease were

preoperative mechanical ventilation and preoperativeneurologic deficit. The STS is in the process of creating aspecific tool for surgical risk adjustment for adults withcongenital heart disease. Many adults with congenitalheart disease have unique preoperative factors, includingventricular dysfunction and pulmonary hypertension.Many of these preoperative factors in adults with con-genital heart disease tend to be quite different from thepreoperative factors in children. Eventually, age-specificrisk models will complement the overall risk models inthe STS-CHSD.In the January 2014 upgrade of the STS-CHSD,

several procedure-specific factors were added to thedata collection form. These new procedure-specificfactors pertain to the previously published benchmarkoperations [13] and should eventually facilitate thecreation of procedure-specific risk models for thesebenchmark operations.In reality, meaningful evaluation and comparison of

outcomes require consideration of both mortality andmorbidity, but the latter is much harder to quantify. TheSTAT Mortality Categories provide an empirically basedtool for analyzing mortality associated with operationsfor congenital heart disease [5]. The addition of patientcharacteristics (including prematurity, chromosomal ab-normalities, syndromes, noncardiac congenital anatomicabnormalities, and preoperative factors) can enhance riskadjustment using the STAT Mortality Categories.To complement the evaluation of quality of care in

pediatric and congenital cardiac surgery using the anal-ysis of risk-adjusted mortality, the STS has also devel-oped a tool to analyze risk-adjusted morbidity: theSTAT Morbidity Categories [15], which are based onmajor postoperative complications and postoperativelength of stay. Both major postoperative complicationsand postoperative length of stay were used becausemodels that assume a perfect one-to-one relationshipbetween postoperative complications and postoperativelength of stay are not likely to fit the data well. (Indeed,prolonged postoperative length of stay is not always aresult of complications but may be secondary to otherfactors such as sociodemographic variables and avail-ability of beds in post–acute care facilities, among others.)Incorporation of both major postoperative complicationsand postoperative length of stay allows creation of a muchmore informative model. The STAT Morbidity Categoriesprovide an empirically based tool for analyzing morbidityassociated with operations for congenital heart disease[15]. Future initiatives to assess quality and improveoutcomes using the STS-CHSD will adjust for bothmortality and morbidity based not only on the operationperformed but also on patient-specific factors.In conclusion, current STS-CHSD risk adjustment is

based on estimated risk of mortality of the primary pro-cedure of the operation as well as on age, weight, andprematurity. The inclusion of additional patient-specificpreoperative factors in risk models for pediatric andcongenital cardiac surgery could lead to increased preci-sion in predicting risk of operative mortality and com-parison of observed to expected outcomes.

1658 JACOBS ET AL Ann Thorac SurgREFINING RISK ADJUSTMENT 2014;98:1653–9

CONGEN

ITALHEA

RT

Dr Pasquali receives support from the National Heart, Lung,and Blood Institute (K08HL103631, principal investigatorDr Pasquali).

References

1. Shahian DM, Edwards FH, Jacobs JP, et al. Public reportingof cardiac surgery performance: part 1—history, rationale,consequences. Ann Thorac Surg 2011;92(Suppl):2–11.

2. Shahian DM, Edwards FH, Jacobs JP, et al. Public reportingof cardiac surgery performance: part 2—implementation.Ann Thorac Surg 2011;92(Suppl):12–23.

3. Jacobs JP, Lacour-Gayet FG, Jacobs ML, et al. Initial appli-cation in the STS congenital database of complexity adjust-ment to evaluate surgical case mix and results. Ann ThoracSurg 2005;79:1635–49.

4. Jacobs JP, Jacobs ML, Lacour-Gayet FG, et al. Stratification ofcomplexity improves the utility and accuracy of outcomesanalysis in a multi-institutional congenital heart surgerydatabase: application of the Risk Adjustment in CongenitalHeart Surgery (RACHS-1) and Aristotle Systems in The So-ciety of Thoracic Surgeons (STS) congenital heart surgerydatabase. Pediatr Cardiol 2009;30:1117–30.

5. O’Brien SM, Clarke DR, Jacobs JP, et al. An empirically basedtool for analyzing mortality associated with congenital heartsurgery. J Thorac Cardiovasc Surg 2009;138:1139–53.

6. Jacobs ML, Daniel M, Mavroudis C, et al. Report of the 2010Society of Thoracic Surgeons Congenital Heart SurgeryPractice and Manpower Survey. Ann Thorac Surg 2011;92:762–9.

7. International Pediatric and Congenital Cardiac Code.Available at: http://www.ipccc.net. Accessed December 30,2013.

8. Franklin RCG, Jacobs JP, Krogmann ON, et al. Nomenclaturefor congenital and paediatric cardiac disease: historical per-spectives and the International Pediatric and CongenitalCardiac Code. Cardiol Young 2008;18(Suppl 2):70–80.

9. STS Congenital Heart Surgery Database v3.0. Availableat: http://www.sts.org/sites/default/files/documents/pdf/congenitaldataspecificationsv3_0_20090904.pdf. AccessedJuly 4, 2014.

10. Clarke DR, Breen LS, Jacobs ML, et al. Verification of data incongenital cardiac surgery. Cardiol Young 2008;18(Suppl 2):177–87.

11. Dokholyan RS, Muhlbaier LH, Falletta J, et al. Regulatory andethical considerations for linking clinical and administrativedatabases. Am Heart J 2009;157:971–82.

12. Jacobs JP, Jacobs ML, Austin EH, et al. Quality measures forcongenital and pediatric cardiac surgery. World J PediatrCongenit Heart Surg 2012;3:32–47.

13. Jacobs JP, O’Brien SM, Pasquali SK, et al. Variation in out-comes for benchmark operations: an analysis of The Societyof Thoracic Surgeons Congenital Heart Surgery Database.Ann Thorac Surg 2011;92:2184–92.

14. Jacobs JP, O’Brien SM, Pasquali SK, et al. Variation in out-comes for risk-stratified pediatric cardiac surgical operations:an analysis of the STS Congenital Heart Surgery Database.Ann Thorac Surg 2012;94:564–72.

15. Jacobs ML, O’Brien SM, Jacobs JP, et al. An empirically basedtool for analyzing morbidity associated with operations forcongenital heart disease. J Thorac Cardiovasc Surg 2013;145:1046–57.e1.

DISCUSSION

DR CARL LEWIS BACKER (Chicago, IL): Jeff, congratulations.I want to point out that Dr Jacobs has been the chair of TheSociety of Thoracic Surgeons Congenital Heart Surgery Database(STS-CHSD) for the past 8 years and has done a terrific job inthat role. This is a very important analysis. We are now pushingthe envelope on our understanding of the influence of preoper-ative factors and their influence on the eventual outcome. Theexample of the patient with an anomalous coronary artery whogoes to the operating room on extracorporeal membraneoxygenation (ECMO) and in shock being completely differentthan the elective patient who shows up as an outpatient on themorning of surgery really says it all.

DR BOHDANMARUSZEWSKI (Warsaw, Poland): Jeff, this is anopportunity to thank you not only for this very interesting pre-sentation but also for the privilege of working together and beingyour partner as chair of the European Congenital Database.I have to say, there was nothing more important in my profes-sional life than the work that we’ve done together.

But looking to the future and thinking about how are we goingto progress with our work what do you think, Jeff, how should weaddress the issue of the stratification of the morbidity? I think theaudience should know that the next step is to implement to allthree databases the models that would analyze the morbiditybased on the morbidity scores.

DR JACOBS: Thank you Carl and Bohdan. Bohdan, your ques-tion is extremely important. Prior to answering this question, Iwould like to thank both you and Carl Backer for your very kindwords. I would also like to acknowledge my friend and partnerGus Mavroudis who founded the STS Congenital Heart SurgeryDatabase (STS-CHSD). Gus founded the STS-CHSD, Chaired

the STS-CHSD, and passed the leadership of the STS-CHSD onto me in 2006. Both Marshall Jacobs and Gus Mavroudis haveChaired the STS-CHSD prior to me. If not for the leadership andvision of Gus and Marshall, we simply would not have the STS-CHSD. I have had the pleasure of Chairing the STS-CHSD,fromJanuary 2006 through January 2014. As of this 2014 annualmeeting of STS, I am passing the Chairmanship of the STS-CHSD on to Marshall Jacobs. Marshall will return as Chair of theSTS-CHSD, after the 8 years that I had as Chair of the STS-CHSDwhen I was following in the footsteps of Gus and Marshall.Marshall Jacobs and Gus Mavroudis have been my friends andmentors in this activity, and I am massively thankful to them andappreciative of their support over the years. Other importantfriends and mentors in these initiatives have included FredGrover, Dave Shahian, Fred Edwards, and Rich Prager, as well asMartin Elliott, Francois Lacour-Gayet, Christo Tchervenkov, andBill Gaynor.During the time that I Chaired the STS-CHSD, and during the

time that Gus and Marshall did, one of our most valuable ac-tivities was our collaboration with Bohdan, Giovanni Stellin, andour colleagues from the European Association for Cardio-Thoracic Surgery (EACTS) and the European Congenital HeartSurgeons Association (ECHSA), as well as with Hiromi Kur-osawa and Arata Murakami, MD, of The Japan Congenital Car-diovascular Surgery Database (JCCVSD). Clearly, Bohdan, youare absolutely correct about the value of this collaboration. Weplace tremendously high value on this collaboration with ourinternational colleagues. We have done important researchtogether. This collaboration has been enjoyable, fun, andeducational – a true highlight of my professional life.Bohdan, your question is a very important question. My pre-

sentation today discussed strategies to measure risk-adjusted

1659Ann Thorac Surg JACOBS ET AL2014;98:1653–9 REFINING RISK ADJUSTMENT

CONGEN

ITALHEA

RT

mortality. Our goal was to determine the best way to report risk-stratified mortality using STAT Mortality Categories and aug-menting these STAT Mortality Categories with preoperativepatient characteristics, including noncardiac abnormalities, ge-netic abnormalities, syndromes, and the results of this study,patient specific preoperative “status” factors (such as preopera-tive mechanical circulatory support and preoperative renaldysfunction, among many others).

Certainly, risk-adjusted mortality is only part of the picture,and really the next step is to also be able to report risk-adjustedmorbidity. Marshall Jacobs has recently published a great paperabout the STAT Morbidity Categories, which is a tool to analyzerisk-adjusted morbidity associated with pediatric and congenitalheart surgery based on major postoperative complications andpostoperative length of stay. Both major postoperative compli-cations and postoperative length of stay were used because thesevariables provide related but not redundant information aboutmorbidity. I believe an important next step is to incorporateanalyses based on the STAT Morbidity Categories into the STSand the EACTS congenital heart surgery databases. In the future,we will be able to report risk-stratified morbidity using STATMorbidity Categories. We should then be able to augment theseSTAT Morbidity Categories with patient-specific variables, usinga strategy similar to the strategy that we presented today forreporting risk-stratified mortality.

DR PAUL KIRSHBOM (New Haven, CT): Great presentation,Jeff. I really enjoyed it. I wonder if you could give us a sense ofthe scale of the contributions of the different portions of the riskmodel. I mean, the STAT category, for example, might explain70% of the variability in mortality. What do the genetics andother preoperative factors add? What’s the relative scale of themodel contributors?

DR JACOBS: Thanks Paul. That is a great question that probablycannot be answered in the time that I have left at this podium.However, your question is the subject of an abstract that we aresubmitting to present at the 2014 annual meeting of The South-ern Thoracic Surgical Association. Hopefully, at the 2014 annual

meeting of The Southern Thoracic Surgical Association, we willpresent (and then publish in The Annals of Thoracic Surgery) thedetails of a new risk model, which will be based on this pre-sentation as well as the question you just asked: “What’s therelative scale of the model contributors?” How much should becontributed by the STAT Mortality Categories versus prematu-rity versus noncardiac congenital anatomic abnormalities versuschromosomal abnormalities versus syndromes versus preoper-ative factors. All of these variables are going to be factored intothe new risk model, the details of which will be the subject of asecond presentation and paper.

DR BACKER: So Jeff, how is this going to affect the U.S. News &World Report analysis that is generated by the STS? That reportis something that everyone is now focused on.

DR JACOBS: Yes, Carl, the U.S. News & World Report Chil-dren’s Hospital Rankings are of intense interest. Therefore, yourquestion is a great question. I think U.S. News & World Reporthas been reasonably and encouragingly responsive to work doneby the STS-CHSD. The leaders of the U.S. News survey havemade great efforts to mold their survey to what has been done bySTS. Therefore, we currently report to U.S. News & WorldReport mortality stratified by the STAT Mortality Categories.Previously, we reported mortality stratified by the RACHS-1Categories, when STS was primarily using RACHS-1 and Aris-totle. When STS evolved and transitioned to the primary use ofthe STATMortality Categories, U.S. News followed that lead andalso evolved and transitioned to the use of the STAT MortalityCategories.My hope is that after STS begins to report STAT stratified

outcomes adjusted for preoperative patient characteristics, thenU.S. News will follow our lead as well and also move to the newand improved STS risk model. So I believe that the changes thatwe make in the STS-CHSD will be incorporated in the U.S. News& World Report Children’s Hospital Rankings. This belief issupported by evidence from history where U.S. News & WorldReport has incorporated previous changes and upgrades in theSTS-CHSD.