Can cardiovascular clinical characteristics be identified and outcome models be developed from an in-patient claims database?

Feasibility analysis of conducting observational studies with the electronic health record

Background

The secondary use of electronic health records (EHRs) promises to facilitate medical research. We reviewed general data requirements in observational studies and analyzed the feasibility of conducting observational studies with structured EHR data, in particular diagnosis and procedure codes.

Methods

After reviewing published observational studies from the University Hospital of Erlangen for general data requirements, we identified three different study populations for the feasibility analysis with eligibility criteria from three exemplary observational studies. For each study population, we evaluated the availability of relevant patient characteristics in our EHR, including outcome and exposure variables. To assess data quality, we computed distributions of relevant patient characteristics from the available structured EHR data and compared them to those of the original studies. We implemented computed phenotypes for patient characteristics where necessary. In random samples, we evaluated how well structured patient characteristics agreed with a gold standard from manually interpreted free texts. We categorized our findings using the four data quality dimensions “completeness”, “correctness”, “currency” and “granularity”.

Results

Reviewing general data requirements, we found that some investigators supplement routine data with questionnaires, interviews and follow-up examinations. We included 847 subjects in the feasibility analysis (Study 1 n = 411, Study 2 n = 423, Study 3 n = 13). All eligibility criteria from two studies were available in structured data, while one study required computed phenotypes in eligibility criteria. In one study, we found that all necessary patient characteristics were documented at least once in either structured or unstructured data. In another study, all exposure and outcome variables were available in structured data, while in the other one unstructured data had to be consulted. The comparison of patient characteristics distributions, as computed from structured data, with those from the original study yielded similar distributions as well as indications of underreporting. We observed violations in all four data quality dimensions.

Conclusions

While we found relevant patient characteristics available in structured EHR data, data quality problems may entail that it remains a case-by-case decision whether diagnosis and procedure codes are sufficient to underpin observational studies. Free-text data or subsequently supplementary study data may be important to complement a comprehensive patient history.

Appropriateness of ICD-coded Diagnostic Inpatient Hospital Discharge Data for Medical Practice Assessment

Objectives: We performed a systematic review to investigate the quality of diagnostic hospital discharge data (DHDD) in order to gain insight in the usefulness of these data for medical practice assessment. We investigated the methods used to evaluate data quality, factors that determine data quality and its consequences for medical practice assessment.

Methods: We selected studies in which both completeness (or sensitivity: SENS) and correctness (or positive predictive value: PPV) were measured. We used the random-effects model to calculate mean SENS and PPV and to explore the effect of a number of covariates.

Results: The 101 included studies were very heterogeneous. We distinguished six typical study designs. We found a mean SENS of 0.67 (95%CI: 0.62– 0.73) and PPV of 0.76 (95%CI: 0.73– 0.79); SENS was significantly lower for comorbidity and complication studies than for some single disease studies. PPV was significantly higher for Scandinavian countries than for other countries. Recoding compared to re-abstracting of the medical record as a gold standard gave a significantly lower PPV. Diagnostic data were considered appropriate by the authors of the studies for quality of care purposes when both SENS and PPV were at least 0.85. Only 13% of the studies fulfilled this criterion.

Conclusions: Variability in quality of care between settings can easily be overshadowed by variability in data quality. However, the use of DHDD by physicians to evaluate their own medical practice may be useful. But only if physicians are willing to critically interpret the meaning of the information for their medical practice assessment.

Analysis of the Association of Clubhouse Membership with Overall Costs of Care for Mental Health Treatment

We examined whether frequency of attendance at the B'More Clubhouse was associated with lower mental health care costs in the Medicaid database, and whether members in the B'More Clubhouse (n = 30) would have lower mental health care costs compared with a set of matched controls from the same claims database (n = 150). Participants who attended the Clubhouse 3 days or more per week had mean 1-year mental health care costs of US $5697, compared to $14,765 for those who attended less often. B'More Clubhouse members had significantly lower annual total mental health care costs than the matched comparison group ($10,391 vs. $15,511; p < 0.0001). Membership in the B'More Clubhouse is associated with a substantial beneficial influence on health care costs.

Cost of Heart Failure in Patients Receiving beta-Blockers and Angiotensin-Converting Enzyme Inhibitors

OBJECTIVE

Numerous studies support the benefit of beta-blockers and angiotensin-converting enzyme inhibition (ACE-I) in the management of heart failure. However, the real-world cost of heart failure in patients who take these medications is not well documented; furthermore, it is unclear if heart failure costs remain significant when current, appropriately aggressive care is delivered.

DESIGN

This study describes 1-year medical costs in patients hospitalised for heart failure who received these therapies, alone or in combination.

METHODS

The study population was derived from 2.5 million patients with at least 3 years' continuous eligibility in Pharmetrics((R)), an integrated claims and pharmacy database on approximately 25 million covered lives from 40 US health plans. The enrolment period was from 1 January 1996 to 31 December 2000. Costs included all recorded payments over a 1-year period. A total of 3073 patients (age >18 years) hospitalised with heart failure were identified (mean [+/- SD] age 72 +/- 13 years; 46% female).

RESULTS

The 1-year cost was $US16 786 in patients who received neither ACE inhibitors nor beta-blockers as compared with $US19 567, $US22 785 and $US27 078 in patients who received ACE inhibitors, beta-blockers or both drugs at maximum dosage, respectively (p < 0.001) [year of costing 2000]. Follow-up costs were substantial, representing almost twice the initial hospitalisation cost. Adjusted for age, sex, diabetes mellitus, coronary disease, hypertension and renal failure, costs remained significant in heart failure patients who received ACE inhibitors and/or beta-blockers.

CONCLUSIONS

The 1-year cost of therapy for patients with heart failure is substantial, and there remains considerable need for more effective therapy to reduce the societal economic burden.

Prediction of long-term mortality after percutaneous coronary intervention in older adults: results from the National Cardiovascular Data Registry

BACKGROUND

The purpose of this study was to develop a long-term model to predict mortality after percutaneous coronary intervention in both patients with ST-segment elevation myocardial infarction and those with more stable coronary disease.

METHODS AND RESULTS

The American College of Cardiology Foundation CathPCI Registry data were linked to the Centers for Medicare and Medicaid Services 100% denominator file by probabilistic matching. Preprocedure demographic and clinical variables from the CathPCI Registry were used to predict the probability of death over 3 years as recorded in the Centers for Medicare and Medicaid Services database. Between 2004 and 2007, 343 466 patients (66%) of 518 195 patients aged ≥65 years undergoing first percutaneous coronary intervention in the CathPCI Registry were successfully linked to Centers for Medicare and Medicaid Services data. This study population was randomly divided into 60% derivation and 40% validation cohorts. Median follow-up was 15 months, with mortality of 3.0% at 30 days and 8.7%, 13.4%, and 18.7% at 1, 2, and 3 years, respectively. Twenty-four characteristics related to demographics, clinical comorbidity, prior history of disease, and indices of disease severity and acuity were identified as being associated with mortality. The C indices in the validation cohorts for patients with and without ST-segment elevation myocardial infarction were 0.79 and 0.78. The model calibrated well across a wide range of predicted probabilities.

CONCLUSIONS

On the basis of the large and nationally representative CathPCI Registry, we have developed a model that has excellent discrimination, calibration, and validation to predict survival up to 3 years after percutaneous coronary intervention.

A comparison of comorbidities obtained from hospital administrative data and medical charts in older patients with pneumonia

Background

The use of comorbidities in risk adjustment for health outcomes research is frequently necessary to explain some of the observed variations. Medical charts reviews to obtain information on comorbidities is laborious. Increasingly, electronic health care databases have provided an alternative for health services researchers to obtain comorbidity information. However, the rates obtained from databases may be either over- or under-reported. This study aims to (a) quantify the agreement between administrative data and medical charts review across a set of comorbidities; and (b) examine the factors associated with under- or over-reporting of comorbidities by administrative data.

Methods

This is a retrospective cross-sectional study of patients aged 55 years and above, hospitalized for pneumonia at 3 acute care hospitals. Information on comorbidities were obtained from an electronic administrative database and compared with information from medical charts review. Logistic regression was performed to identify factors that were associated with under- or over-reporting of comorbidities by administrative data.

Results

The prevalence of almost all comorbidities obtained from administrative data was lower than that obtained from medical charts review. Agreement between comorbidities obtained from medical charts and administrative data ranged from poor to very strong (kappa 0.01 to 0.78). Factors associated with over-reporting of comorbidities were increased length of hospital stay, disease severity, and death in hospital. In contrast, those associated with under-reporting were number of comorbidities, age, and hospital admission in the previous 90 days.

Conclusions

The validity of using secondary diagnoses from administrative data as an alternative to medical charts for identification of comorbidities varies with the specific condition in question, and is influenced by factors such as age, number of comorbidities, hospital admission in the previous 90 days, severity of illness, length of hospitalization, and whether inhospital death occurred. These factors need to be taken into account when relying on administrative data for comorbidity information.

Validation of a case definition to define chronic dialysis using outpatient administrative data

BACKGROUND

Administrative health care databases offer an efficient and accessible, though as-yet unvalidated, approach to studying outcomes of patients with chronic kidney disease and end-stage renal disease (ESRD). The objective of this study is to determine the validity of outpatient physician billing derived algorithms for defining chronic dialysis compared to a reference standard ESRD registry.

METHODS

A cohort of incident dialysis patients (Jan. 1-Dec. 31, 2008) and prevalent chronic dialysis patients (Jan 1, 2008) was selected from a geographically inclusive ESRD registry and administrative database. Four administrative data definitions were considered: at least 1 outpatient claim, at least 2 outpatient claims, at least 2 outpatient claims at least 90 days apart, and continuous outpatient claims at least 90 days apart with no gap in claims greater than 21 days. Measures of agreement of the four administrative data definitions were compared to a reference standard (ESRD registry). Basic patient characteristics are compared between all 5 patient groups.

RESULTS

1,118,097 individuals formed the overall population and 2,227 chronic dialysis patients were included in the ESRD registry. The three definitions requiring at least 2 outpatient claims resulted in kappa statistics between 0.60-0.80 indicating "substantial" agreement. "At least 1 outpatient claim" resulted in "excellent" agreement with a kappa statistic of 0.81.

CONCLUSIONS

Of the four definitions, the simplest (at least 1 outpatient claim) performed comparatively to other definitions. The limitations of this work are the billing codes used are developed in Canada, however, other countries use similar billing practices and thus the codes could easily be mapped to other systems. Our reference standard ESRD registry may not capture all dialysis patients resulting in some misclassification. The registry is linked to on-going care so this is likely to be minimal. The definition utilized will vary with the research objective.

Assessment of cost and health resource utilization for elderly patients with heart failure and diabetes mellitus

BACKGROUND

Our aim was to examine the health resource utilization and cost of care associated with heart failure (HF) and diabetes mellitus (DM) for elderly Medicare enrollees.

METHODS AND RESULTS

A retrospective case-control design was used to identify 4 groups of elderly patients with HF and DM (n = 498), HF only (n = 1089), DM only (n = 971), and no-HF and no-DM (n = 5438) using an administrative database of a large urban academic health care system. Demographic, diagnostic, health resource utilization, and cost (reimbursement) data were obtained from the Medicare claims database for the years 2000 and 2001. Disease states were identified by ICD-9 codes. Costs and health resource utilization were compared across the groups. The mean total costs were highest for the group with HF and DM ($32,676), and second highest for the HF only group ($22,230). In multivariable models that adjusted for potentially influential covariates, the group with HF and DM had a 3-fold increase in total cost compared with the group without DM and HF (relative total cost = 4.51, 95% confidence interval 3.82-5.31).

CONCLUSIONS

The presence of DM has a substantial influence on the costs for managing older patients with HF. An integrated approach to management may be needed.

Frequency of electrocardiographic recordings in patients presenting with angina pectoris (from the Investigation of National Coronary Disease Identification)

The appropriate progression of diagnostic testing for acute angina has been the topic of several recommendations by the American College of Cardiology and American Heart Association (ACC/AHA). We measured how frequently electrocardiography (ECG) is provided as the initial cardiac diagnostic test, as recommended for patients with new angina. Using an insurance database representing 2% of the U.S. adult population, we identified patients undergoing a new cardiac diagnostic process for angina. Rates of initial ECG were stratified by age, gender, co-morbid disease, and care setting. Of 4.4 million patients, 18,139 met the entry criteria by presenting with anginal symptoms for testing. A substantial portion (35%, 95% confidence interval [CI] 34% to 35%) did not receive the initial ECG recommended by expert guidelines. Patients treated in emergency departments received an initial ECG more frequently (91%, CI 90% to 92%) than patients tested in outpatient settings (61%, CI 60% to 62%; risk ratio [RR] 0.67, CI for RR 0.65 to 0.68) or in inpatient hospital settings (34%, CI 32% to 37%; RR 0.38, CI for RR 0.36 to 0.40). Slightly lower rates of initial ECG were observed in men (RR 0.93 vs women, CI for RR 0.91 to 0.95) and patients over 64 years (RR 0.93 vs younger patients, CI for RR 0.91 to 0.95). Total diagnostic costs averaged $954 when testing began with the recommended ECG versus $1,233 when testing did not. In conclusion, ECG is not universally obtained as the initial test for patients presenting with anginal symptoms despite evidence-based recommendations for such use. Clinicians should be aware that suboptimal use of ECG in certain settings may hinder investigations of heart disease.

Comparison of mortality risk adjustment using a clinical data algorithm (American College of Surgeons National Surgical Quality Improvement Program) and an administrative data algorithm (Solucient) at the case level within a single institution

BACKGROUND

There is great interest in efficiently evaluating health care quality, but there is controversy over the use of administrative versus clinical data methods. We sought to compare actual mortality with risk-adjusted expected mortality in a sample population calculated by two different methods; one based on preexisting administrative records and one based on chart reviews.

STUDY DESIGN

We examined a sample of patients (n = 1,234) undergoing surgical procedures at an academic teaching hospital during 1 year. The first risk-adjustment method was that used by the National Surgical Quality Improvement Program, which is based on dedicated medical record review. The second method was that used by Solucient, LLC, which is based on preexisting administrative records.

RESULTS

The ratio of observed to expected mortality for this population set was higher using the National Surgical Quality Improvement Program algorithm (1.1; 95% CI, 0.8 to 1.5) than using the Solucient algorithm (0.9; 95% CI, 0.6 to 1.2) but neither estimate was notably different from 1.0. Similarly, when observed to expected mortality ratios were calculated separately for each quartile of mortality, there were no marked differences within quartiles, although minor differences with potential importance were noted. Fit was comparable by age categories, gender, and American Society of Anesthesiologists' categories. A number of actual deaths had higher predicted mortality scores using the Solucient algorithm.

CONCLUSIONS

Risk-adjusted mortality estimates were comparable using administrative or clinical data. Minor performance differences might still have implications. Because of the potential lower cost of using administrative data, this type of algorithm can be an efficient alternative and should continue to be investigated.

On the efficacy of programmatic searching of medical claims for the occurrence of hospital admissions for coronary artery disease

This study was conducted in order to test the proposition that medical claim records, when searched electronically, can be reliably used to locate individual, disease-specific hospital admissions. For the study, admissions for coronary artery disease (CAD), self-reported by employer-sponsored recipients of chronic disease management (DM) services, were verified against physician-compiled medical records. Confirmed events were then subjected to electronic searching of the corresponding medical claim records using a variety of conditional requirements for included types of evidence. At maximum sensitivity (92.6%), the search algorithm positively identified 126 of 136 verified admissions while falsely identifying 1,025 others. At maximum specificity (98.7%), the algorithm positively identified 55 of 136 while falsely identifying 13. The maximum value of the true positive to false positive ratio was 4.47. The maximum Youden index value was obtained by requiring that the diagnostic intensity (proportion of event-related claims having a CAD-related diagnosis code) have a minimum value of 0.20. The study concluded that an admission search algorithm applied to typical commercial medical claims generated results that are unsatisfactory for the determination of admission incidence in the CAD population. While the methods may be sound, they fail to overcome the weaknesses of the searched data.

Administrative Versus Clinical Data for Coronary Artery Bypass Graft Surgery Report Cards

OBJECTIVE

The objective of this study was to compare the performance of a risk model for isolated coronary artery bypass graft (CABG) surgery based on administrative data with that of a clinical risk model in predicting mortality and identifying hospital performance outliers.

METHODS

Clinical data records from the California CABG Mortality Reporting Program for 38,230 isolated CABG patients undergoing surgery in 2000-2001 were linked to records in the California patient discharge data (PDD) abstract. Risk factors based on administrative data that mirrored clinical risk factors were developed using the condition present at admission indicator in the PDD to separate preoperative acute conditions from complications of surgery. Using logistic regression, risk model performance across data sources was compared along with hospital risk-adjusted mortality ranks and quality ratings.

RESULTS

The administrative data showed lower prevalence of risk factors when compared with the clinical data. The clinical risk model had somewhat better discrimination (C = 0.824) than the administrative model (C = 0.799). The clinical model yielded 17 outliers and the administrative model 16 with agreement on 12 hospitals' status. Performance of the administrative risk model was minimally affected by removal of information from prior admissions and removal of risk factors not confirmed in the clinical record.

CONCLUSIONS

Unique properties of the California administrative data, including the ability to distinguish acute preoperative risk factors from complications of surgery, permitted construction of an administrative risk model that predicts mortality on par with most published clinical models. Despite this, the administrative model identified slightly different hospital outliers, which may indicate somewhat biased assessments of hospital patient risk.

Review of the performance of methods to identify diabetes cases among vital statistics, administrative, and survey data

PURPOSE

The ability to identify prevalent cases of diagnosed diabetes is crucial to monitoring preventative care practices and health outcomes among persons with diagnosed diabetes.

METHODS

We conducted a comprehensive literature review to assess and summarize the validity of various strategies for identifying individuals with diagnosed diabetes and to examine the factors influencing the validity of these strategies.

RESULTS

We found that studies using either administrative data or survey data were both adequately sensitive (i.e., identified the majority of cases of diagnosed diabetes) and highly specific (i.e., did not identify the individuals as having diabetes if they did not). In contrast, studies based on cause-of-death data from death certificates were not sensitive, failing to identify about 60% of decedents with diabetes and in most of these studies, researchers did not report specificity or positive predictive value.

CONCLUSIONS

Surveillance is critical for tracking trends in diabetes and targeting diabetes prevention efforts. Several approaches can provide valuable data, although each has limitations. By understanding the limitations of the data, investigators will be able to estimate diabetes prevalence and improve surveillance of diabetes in the population.

Identifying physician-recognized depression from administrative data: consequences for quality measurement

BACKGROUND

Multiple factors limit identification of patients with depression from administrative data. However, administrative data drives many quality measurement systems, including the Health Plan Employer Data and Information Set (HEDIS).

METHODS

We investigated two algorithms for identification of physician-recognized depression. The study sample was drawn from primary care physician member panels of a large managed care organization. All members were continuously enrolled between January 1 and December 31, 1997. Algorithm 1 required at least two criteria in any combination: (1) an outpatient diagnosis of depression or (2) a pharmacy claim for an antidepressant Algorithm 2 included the same criteria as algorithm 1, but required a diagnosis of depression for all patients. With algorithm 1, we identified the medical records of a stratified, random subset of patients with and without depression (n = 465). We also identified patients of primary care physicians with a minimum of 10 depressed members by algorithm 1 (n = 32,819) and algorithm 2 (n = 6,837).

RESULTS

The sensitivity, specificity, and positive predictive values were: Algorithm 1: 95 percent, 65 percent, 49 percent; Algorithm 2: 52 percent, 88 percent, 60 percent. Compared to algorithm 1, profiles from algorithm 2 revealed higher rates of follow-up visits (43 percent, 55 percent) and appropriate antidepressant dosage acutely (82 percent, 90 percent) and chronically (83 percent, 91 percent) (p < 0.05 for all).

CONCLUSIONS

Both algorithms had high false positive rates. Denominator construction (algorithm 1 versus 2) contributed significantly to variability in measured quality. Our findings raise concern about interpreting depression quality reports based upon administrative data.

Assessing data adequacy for clinical research: reliability and validity of a surgical database

As clinical databases are utilized more frequently for clinical research, it is essential that researchers assess the quality of databased information. While researchers have begun to report strategies to measure accuracy of databased information, knowledge remains limited. The purpose of this study was to assess the reliability and validity of databased information among selected study variables contained within a computerized coronary artery surgery clinical database using the written patient medical record as an external standard. Both reliability (N = 400) and validity (N = 100) samples were randomly selected from a databased sampling frame of 548 Medicare subjects who underwent coronary artery bypass grafting surgery in 1998. Reliability assessed by consistency rates were age (95%), race (94%), gender (99%), congestive heart failure (CHF) (60.5%), angina (91.5%), renal insufficiency (82%), hypertension (91.7%), diabetes mellitus (93.7%), chronic obstructive pulmonary disease (COPD) (75.5%), clinical status (97%), American Society of Anesthesiologists classification (99%), prior peripheral vascular surgery (15.5%), prior CABGS (99%), and duration of mechanical ventilation (87.5%). These percentages reflected a large portion of missing data for CHF, COPD, and prior peripheral vascular surgery. Validity assessed by sensitivity and specificity analyses were all greater than 80%. The majority of computerized databased information among selected study variables was the same information recorded in the written patient medical record. Using the same external standard to assess both reliability and validity was a significant limitation of this study, which resulted in the same measure of data adequacy by utilizing differing statistical methods.

Can pharmacy data improve prediction of hospital outcomes? Comparisons with a diagnosis-based comorbidity measure

OBJECTIVES

The performance of comorbidity measures derived from the hospital discharge abstract, the outpatient pharmacy record, and from both sources combined, were compared in predicting all-cause and unplanned hospital readmission and length of stay.

MATERIALS AND METHODS

Automated hospital and pharmacy data came from Kaiser-Permanente and included 6721 acute hospitalizations in Southern California from April 1993 to February 1995. The Deyo adaptation of Charlson's 17 comorbidities was derived from hospital discharge data and the 29 Chronic Disease Score (CDS) comorbidity markers were derived from outpatient pharmacy claims data. Logistic and OLS regression models were used to compare the performance of each measure in baseline models and to evaluate whether the CDS contributed additional explanatory power in a combined model.

RESULTS

The CDS was a significant predictor of unplanned readmission (C = 0.68) and LOS (Adjusted R(2) = 0.26) in multivariable models adjusted for baseline patient demographic and hospitalization characteristics. The Deyo measure was a significant predictor of all-cause readmission (C = 0.63), unplanned readmission (C = 0.68), and LOS (Adjusted R(2) = 0.26). When pharmacy-based disease markers were added to the Deyo baseline model, modest, statistically significant improvements in predictive power were noted in the unplanned readmission and LOS models.

CONCLUSIONS

The finding that both measures of comorbid disease demonstrated similar predictive power is noteworthy, because secondary diagnosis data document relevant illness in hospital patients and pharmacy claims data were never intended for that purpose. The results suggest that small improvements in model performance may come from combining both sources of data in models to predict hospital readmission and LOS.

Mortality and recurrent cardiac events after coronary artery bypass graft: long term outcomes in a population study

OBJECTIVE

To determine 30 day mortality, long term survival, and recurrent cardiac events after coronary artery bypass graft (CABG) in a population.

DESIGN

Follow up study of patients prospectively entered on to a cardiothoracic surgical database. Record linkages were used to obtain data on readmissions and deaths.

PATIENTS

8910 patients undergoing isolated first CABG between 1980 and 1993 in Western Australia.

MAIN OUTCOME MEASURES

30 day and long term survival, readmission for cardiac event (acute myocardial infarction, unstable angina, percutaneous transluminal coronary angioplasty or reoperative CABG).

RESULTS

There were 3072 deaths to mid 1999. 30 day and long term survival were significantly better in patients treated in the first five years than during the following decade. The age of the patients, proportion of female patients, and number of grafts increased over time. An urgent procedure (odds ratio 3.3), older age (9% per year) and female sex (odds ratio 1.5) were associated with increased risk for 30 day mortality, while age (7% per year) and a recent myocardial infarction (odds ratio 1.16) influenced long term survival. Internal mammary artery grafts were followed by better short and long term survival, though there was an obvious selection bias in favour of younger male patients.

CONCLUSIONS

This study shows worsening crude mortality at 30 days after CABG from the mid 1980s, associated with the inclusion of higher risk patients. Older age, an acute myocardial infarction in the year before surgery, and the use of sephenous vein grafts only were associated with poorer long term survival and greater risk of a recurrent cardiac event. Female sex predicted recurrent events but not long term survival.

Management of hypoplastic left heart syndrome in the 1990s

We used the discharge database of the University Hospital Consortium to determine the management and early outcome of neonates with hypoplastic left heart syndrome admitted to member institutions from 1990 to 1999. Of the 2,264 patients, 1,203 underwent a Norwood procedure, with 42% mortality. Cardiac transplantation was performed in 72, with 38% mortality, and 217 (10%) were discharged without any surgical procedure. The proportion of patients managed by the Norwood procedure increased from 43% during the first half of the decade to 59% in the second half, with corresponding decreases in the proportion managed by transplantation or nonintervention. A mortality rate of < or =40% was achieved in all 5 institutions performing >50 Norwood procedures, and by 9 of 40 institutions performing <50. Performance of a Norwood procedure has become the most frequent management for neonates with hypoplastic left heart syndrome. Lower operative mortality rates are generally, but not exclusively, achieved by institutions with high surgical volume.