Risk Adjustment Methods Research Articles

A Risk-Adjusted Method for Determining OCS Lease Abandonment Liability in the Gulf of Mexico for BOEMRE Supplemental Bonding

Summary In recent years, because of widespread hurricane destruction in the Gulf of Mexico (GOM) and unprecedented losses to the offshore industry, concern has been raised by the federal government that small operators may have difficulty in completing their decommissioning requirements. Hurricane-destroyed infrastructure is significantly more expensive to decommission than normal operations, and small operators are at risk to financial liability and potential bankruptcy. If a company goes bankrupt and there are no capable working-interest owners or previous leaseholders, the financial liability may fall upon the US government. To ensure that the government is adequately protected from incurring costs associated with offshore-lease abandonment, Minerals Management Service (MMS), recently renamed the Bureau of Ocean Energy Management, Regulation and Enforcement (BOEMRE), requires operators to post a supplemental bond if at least one working-interest owner on a lease does not satisfy a minimum financial capacity. A supplemental-bonding formula was developed by MMS in the early 1990s that is based upon the estimated cost of decommissioning. In recognition of the long period of time since the supplemental-bonding formula was developed, an updated risk-adjusted bonding mechanism is presented in this paper. Our purpose is to update the 1990-legacy formula and present risk-adjusted alternatives to incorporate the uncertainty inherent in the implementation of formula mechanisms. The philosophy of making rational tradeoffs between cost and risks is common throughout the offshore industry, and this paper describes the selection of bonding levels that balance the need of the government to minimize its decommissioning exposure at an acceptable level of risk.

Using the Johns Hopkins Aggregated Diagnosis Groups (ADGs) to Predict Mortality in a General Adult Population Cohort in Ontario, Canada

Administrative healthcare databases are increasingly used for health services and comparative effectiveness research. When comparing outcomes between different treatments, interventions, or exposures, the ability to adjust for differences in the risk of the outcome occurring between treatment groups is important. Similarly, when conducting healthcare provider profiling, adequate risk-adjustment is necessary for conclusions about provider performance to be valid. There are limited validated methods for risk adjustment in ambulatory populations using administrative healthcare databases. To examine the ability of the Johns Hopkins' Aggregated Diagnosis Groups (ADGs) to predict mortality in a general ambulatory population cohort. Retrospective cohort constructed using population-based administrative data. All 10,498,413 residents of Ontario, Canada between the ages of 20 and 100 years who were alive on their birthday in 2007. Subjects were randomly divided into derivation and validation samples. Death within 1 year of the subject's birthday in 2007. A logistic regression model consisting of age, sex, and indicator variables for 28 of the 32 ADG categories had excellent discrimination: the c-statistic (equivalent to the area under the receiver operating characteristic curve) was 0.917 in both derivation and validation samples. Furthermore, the model showed very good calibration. In comparison, the use of the Charlson comorbidity index or the Elixhauser comorbidities resulted in a minor decrease in discrimination compared with the use of the ADGs. Logistic regression models using age, sex, and the John Hopkins ADGs were able to accurately predict 1-year mortality in a general ambulatory population of subjects.

Risikoadjustierte Qualitätsbeurteilung neonatologischer Einrichtungen an Hand des neuen Qualitätskriteriums „Verlegerate von reifen Neugeborenen“

1. The transfer rate of mature newborns will be presented as a new quality indicator. 2. Another objective of this study was to adjust the transfer rate of mature newborns of different hospitals according to their "risk" profile of patients by multivariate analysis. The perinatal database of 118,416 newborns of the Saxonian quality surveillance from 2001 to 2004 was analysed. Based on 17 clinical and 3 structural factors, a logistic regression model was used to develop a specific "risk" predictor for the quality indicator "transfer rate". For care level III (basic care) a "risk" predictor for the transfer rate was developed, which consists of 15 factors. The AUC(ROC)-value of this quality indicator was 78.6%, which is sufficient. The hospital ranking based on the adjusted risk assessment was different from the hospital ranking prior to this adjustment. The average correction of ranking position was 10.4 for 43 clinics. 1. The new quality indicator "transfer rate of mature newborns" can be recommended. 2. The application of the risk adjustment method proposed here allows for a more objective comparison of the quality indicator "transfer rate" among different hospitals.

Risikoadjustierte Qualitätsbeurteilung am Beispiel der neonatalen Spätinfektion

The weak point of the countrywide perinatal/neonatal quality surveillance is the ignorance of interhospital differences in the case mix of patients. As a result, this approach does not produce reliable benchmarking. The objective of this study was to adjust the result of the late-onset infection incidence of different hospitals according to their risk profile of patients by multivariate analysis. The perinatal/neonatal database of 41,055 newborns of the Saxonian quality surveillance from 1998 to 2004 was analysed. Based on 18 possible risk factors, a logistic regression model was used to develop a specific risk predictor for the quality indicator "late-onset infection". The developed risk predictor for the incidence of late-onset infection could be described by 4 of the 18 analysed risk factors, namely gestational age, admission from home, hypoxic ischemic encephalopathy and B-streptococcal infection. The AUC(ROC) value of this quality indicator was 83.3%, which demonstrates its reliability. The hospital ranking based on the adjusted risk assessment was very different from hospital rankings before this adjustment. The average correction of ranking position was 4.96 for 35 clinics. The application of the risk adjustment method proposed here allows for a more objective comparison of the incidence of the quality indicator "late onset infection" among different hospitals.

A Timing Strategy for Investments in U.S. Hospitality Stocks

Historical data for the period from January 2, 1973 to May 30, 2008 show that U.S. airline, hotel, restaurant, and travel and leisure sector indices underperformed the market portfolio (S&P 500) in terms of the Sharpe ratio. This result suggests that simply buying and holding hospitality sector stocks is a poor investment strategy. It is necessary for hospitality stock investors to find a method of improving their investments in U.S. hospitality sector stocks. This study offers a timing strategy for investing in U.S. hospitality sector stocks. Directional changes in the Fed discount rate signal when to buy or sell stocks of different hospitality sectors. Accordingly, a timing strategy is formulated and its performance relative to a passive buy-and-hold (market portfolio) strategy is examined with five risk-adjusted performance measures. Empirical evidence derived from the daily trading data over a 35-year period generally supports the superiority of the proposed timing strategy for investments in hotel, restaurant, and travel and leisure sector stocks over the buy-and-hold strategy based on various risk-adjusted performance evaluation methods.

Simplified Alternative to the TRISS Method for Resource‐Constrained Settings

We developed simple methods of risk adjustment for evaluating the quality of injury care (predicting survival probabilities of the injured) by fully utilizing routinely collected data in injury surveillance and clinical practices. Widely used methods of risk adjustment require additional data that are difficult to collect in resource-constrained settings. We developed logistic regression models that predict survival using data obtained from 9,840 victims aged 15 years or older with blunt traumatic injuries who were registered in the Japan Trauma Data Bank, Japan's national trauma registry, between January 2004 and December 2007. The models included three predictors: age, an anatomical injury severity parameter such as a simplified severity categorization (minor, moderate, and severe) described in the Injury Surveillance Guidelines, and a physiological status parameter. The models' abilities to predict survival probabilities were evaluated using the area under the receiver-operating characteristic curve (AUROCC). The simplified three-predictor models showed good performance with the AUROCC ranging from 0.86 to 0.94. In particular, the models with a consciousness level indicator as a physiological parameter showed a high AUROCC, ranging from 0.93 to 0.94, which was not much different from the performance of the widely used method that shows an AUROCC of 0.96. Simplified methods of risk adjustment that require only routinely collected data will facilitate evaluation and improvement in the quality of injury care in resource-constrained low- and middle-income countries, where injuries are a growing public health concern.

The impact of adjusting for reliability on hospital quality rankings in vascular surgery

Hospital quality in vascular surgery is often measured using mortality. We sought to determine whether adjusting mortality for statistical reliability changes hospital quality rankings for vascular surgery. Patients undergoing five common vascular surgery procedures (open and endovascular abdominal aortic aneurysm repair, carotid endarterectomy, lower extremity bypass, and aorto-femoral bypass) in the National Surgical Quality Improvement Project (NSQIP) in 2007 were identified (n = 14,559). For each hospital, we first calculated a ratio of observed to expected mortality (O-E ratio) using standard NSQIP techniques. We then adjusted these estimates for statistical noise using empirical Bayes methods, a technique known as reliability adjustment. We then compared rankings based on the standard O-E ratio to the rankings after reliability adjustment. A total of 172 hospitals reported an average adjusted mortality rate of 2.4% for the five procedures, varying from 0% to 17%. After adjusting for statistical noise using reliability adjustment, hospital mortality was greatly diminished, varying only from 1.7% to 4.1%. This adjustment for reliability had a dramatic effect on hospital rankings. Overall, 43% of hospitals were reclassified into either a higher or lower quartile of performance using traditional methods of risk-adjustment. Fifty-one percent all hospitals in the "best" quartile of performance according to traditional O-E ratios are not classified in the "best" quartile after adjusting for statistical noise. Twenty-six percent of hospitals in the "worst" quartile were no longer classified as such after adjusting for noise. Adjusting mortality for reliability reduces statistical noise and provides more stable estimates of hospital quality. Reliability adjustment should be standard for comparing hospital quality.

Risk Stratification, Risk Adjustment, and Other Risks

This article has been selected for the ANESTHESIOLOGY CME Program. Learning objectives and disclosure and ordering information can be found in the CME section at the front of this issue.

Broadly Applicable Risk Stratification System for Predicting Duration of Hospitalization and Mortality

Hospitals are increasingly required to publicly report outcomes, yet performance is best interpreted in the context of population and procedural risk. We sought to develop a risk-adjustment method using administrative claims data to assess both national-level and hospital-specific performance. A total of 35,179,507 patient stay records from 2001-2006 Medicare Provider Analysis and Review (MEDPAR) files were randomly divided into development and validation sets. Risk stratification indices (RSIs) for length of stay and mortality endpoints were derived from aggregate risk associated with individual diagnostic and procedure codes. Performance of RSIs were tested prospectively on the validation database, as well as a single institution registry of 103,324 adult surgical patients, and compared with the Charlson comorbidity index, which was designed to predict 1-yr mortality. The primary outcome was the C statistic indicating the discriminatory power of alternative risk-adjustment methods for prediction of outcome measures. A single risk-stratification model predicted 30-day and 1-yr postdischarge mortality; separate risk-stratification models predicted length of stay and in-hospital mortality. The RSIs performed well on the national dataset (C statistics for median length of stay and 30-day mortality were 0.86 and 0.84). They performed significantly better than the Charlson comorbidity index on the Cleveland Clinic registry for all outcomes. The C statistics for the RSIs and Charlson comorbidity index were 0.89 versus 0.60 for median length of stay, 0.98 versus 0.65 for in-hospital mortality, 0.85 versus 0.76 for 30-day mortality, and 0.83 versus 0.77 for 1-yr mortality. Addition of demographic information only slightly improved performance of the RSI. RSI is a broadly applicable and robust system for assessing hospital length of stay and mortality for groups of surgical patients based solely on administrative data.

Is Risk-Adjusted Mortality an Indicator of Quality of Care in General Surgery?

Profiling of hospitals using risk-adjusted mortality rates as a measure of quality is becoming increasingly frequent. We sought to determine the validity of this approach by comparing the risk-adjusted predicted mortality to the findings of concurrent peer review and retrospective chart review of deaths that occur on a general surgery service. Consecutive patients admitted to a busy general surgery service from January 2000 to January 2006 were prospectively entered into the Surgical Activity Tracking System. Rigorous, systematic peer review was performed concurrently by service members on all deaths. Adjudication was later validated by an independent senior surgeon. Three methodologies of risk adjustment (University Health Consortium, Physiological and Operative Severity Score for the enUmeration of Mortality, and the Charlson index) were used and compared the "excess mortality" predicted by each to the number of potentially preventable deaths determined by peer review. A total of 9623 patients were admitted and 75 died (0.7%). University Health Consortium and Physiological and Operative Severity Score predicted an excess mortality of 62 and 65 deaths, respectively; Charlson predicted that 73% of the cohort would be dead in 1 year. Concurrent and retrospective peer review found that death was potentially preventable in only 22 and 21 patients, respectively. Peer adjudication and extensive clinical review adds much to the analysis of an adverse outcome, similar to the "black box" in an airplane crash. Although methods of risk adjustment may be helpful in identifying patients for peer review, they should be used for internal process improvement and not published as metrics of hospital or provider performance.

Comparison of the performance of the CMS Hierarchical Condition Category (CMS-HCC) risk adjuster with the charlson and elixhauser comorbidity measures in predicting mortality

BackgroundThe Centers for Medicare and Medicaid Services (CMS) has implemented the CMS-Hierarchical Condition Category (CMS-HCC) model to risk adjust Medicare capitation payments. This study intends to assess the performance of the CMS-HCC risk adjustment method and to compare it to the Charlson and Elixhauser comorbidity measures in predicting in-hospital and six-month mortality in Medicare beneficiaries.MethodsThe study used the 2005-2006 Chronic Condition Data Warehouse (CCW) 5% Medicare files. The primary study sample included all community-dwelling fee-for-service Medicare beneficiaries with a hospital admission between January 1st, 2006 and June 30th, 2006. Additionally, four disease-specific samples consisting of subgroups of patients with principal diagnoses of congestive heart failure (CHF), stroke, diabetes mellitus (DM), and acute myocardial infarction (AMI) were also selected. Four analytic files were generated for each sample by extracting inpatient and/or outpatient claims for each patient. Logistic regressions were used to compare the methods. Model performance was assessed using the c-statistic, the Akaike's information criterion (AIC), the Bayesian information criterion (BIC) and their 95% confidence intervals estimated using bootstrapping.ResultsThe CMS-HCC had statistically significant higher c-statistic and lower AIC and BIC values than the Charlson and Elixhauser methods in predicting in-hospital and six-month mortality across all samples in analytic files that included claims from the index hospitalization. Exclusion of claims for the index hospitalization generally led to drops in model performance across all methods with the highest drops for the CMS-HCC method. However, the CMS-HCC still performed as well or better than the other two methods.ConclusionsThe CMS-HCC method demonstrated better performance relative to the Charlson and Elixhauser methods in predicting in-hospital and six-month mortality. The CMS-HCC model is preferred over the Charlson and Elixhauser methods if information about the patient's diagnoses prior to the index hospitalization is available and used to code the risk adjusters. However, caution should be exercised in studies evaluating inpatient processes of care and where data on pre-index admission diagnoses are unavailable.

Risk-Adjusted Indices for Measuring the Quality of Inpatient Care

This article describes a risk-adjustment method for profiling hospitals and physicians on key measures of clinical quality using readily available administrative data. By comparing actual and expected rates of mortality, complications, readmissions, and patient safety events, this method enables providers to identify both favorable and adverse outcomes performance.

Learning Curves for Bag-and-mask Ventilation and Orotracheal Intubation

In this study, the authors determined the success and failure rates for interns learning bag-and-mask ventilation and orotracheal intubation. Their goal was to determine the amount of experience needed to perform these procedures correctly. The authors recorded 695 bag-and-mask ventilations and 679 orotracheal intubations performed by 15 inexperienced interns during their 3 month-long anesthesia rotations. Learning curves for each procedure for each intern were constructed with both the standard and risk-adjusted cumulative sum methods. The average number of procedures required to attain a failure rate of 20% was estimated for each technique. Fourteen of 15 interns attained acceptable failure rates at bag-and-mask ventilation after 27 +/- 13 procedures, with a median (95% confidence interval) of 25 (15-32) procedures to cross the decision limit when considering all 15 interns. Nine of 15 interns attained acceptable failure rates at orotracheal intubation after 26 +/- 8 procedures, with a median of 29 (22-not estimable) procedures to cross the limit when considering all interns. The proportion of interns who attained acceptable failure rates for mask ventilation was greater than for tracheal intubation (93% vs. 60%, P = 0.025). Overall, our interns achieved a bag-and-mask ventilation failure rate of 20% or better after a median of 25 procedures; approximately 80% of interns achieved the goal after 35 procedures or less. Participating interns developed mask ventilation skills faster than orotracheal intubation skills, and there was more variability in the rate at which intubation skills developed. A median of 29 procedures was required to achieve an 80% orotracheal intubation success rate.

Regional Variations in Diagnostic Practices

Current methods of risk adjustment rely on diagnoses recorded in clinical and administrative records. Differences among providers in diagnostic practices could lead to bias. We used Medicare claims data from 1999 through 2006 to measure trends in diagnostic practices for Medicare beneficiaries. Regions were grouped into five quintiles according to the intensity of hospital and physician services that beneficiaries in the region received. We compared trends with respect to diagnoses, laboratory testing, imaging, and the assignment of Hierarchical Condition Categories (HCCs) among beneficiaries who moved to regions with a higher or lower intensity of practice. Beneficiaries within each quintile who moved during the study period to regions with a higher or lower intensity of practice had similar numbers of diagnoses and similar HCC risk scores (as derived from HCC coding algorithms) before their move. The number of diagnoses and the HCC measures increased as the cohort aged, but they increased to a greater extent among beneficiaries who moved to regions with a higher intensity of practice than among those who moved to regions with the same or lower intensity of practice. For example, among beneficiaries who lived initially in regions in the lowest quintile, there was a greater increase in the average number of diagnoses among those who moved to regions in a higher quintile than among those who moved to regions within the lowest quintile (increase of 100.8%; 95% confidence interval [CI], 89.6 to 112.1; vs. increase of 61.7%; 95% CI, 55.8 to 67.4). Moving to each higher quintile of intensity was associated with an additional 5.9% increase (95% CI, 5.2 to 6.7) in HCC scores, and results were similar with respect to laboratory testing and imaging. Substantial differences in diagnostic practices that are unlikely to be related to patient characteristics are observed across U.S. regions. The use of clinical or claims-based diagnoses in risk adjustment may introduce important biases in comparative-effectiveness studies, public reporting, and payment reforms.

A Risk Adjustment Method for Newborns Undergoing Noncardiac Surgery

To develop a risk adjustment method for in-hospital mortality in newborns undergoing noncardiac surgery. Understanding variation in outcomes is critical to guide quality improvement. Reliable outcome assessments need risk adjustment to allow comparisons. Infants <or=30-days-old undergoing noncardiac surgical procedures were identified using the Kids' Inpatient Database (KID); year 2000. Premature infants were excluded. Procedures identified by ICD-9-CM codes with >or=20 cases in the data set were placed into 4 risk categories by in-hospital mortality rates. Clinical variables were added to the model to better predict mortality; areas under the receiver-operator characteristic (ROC) curves were compared. The final model was validated in the KID 2003 database. Among 6103 eligible cases in the KID 2000, 5117 (83.8%) could be assigned to a risk category. Mortality rates were 0.2% in risk category 1, 2.5% in category 2, 6.4% in 3, and 18.4% in 4. The odds of mortality increased in each risk category relative to category 1 (P < 0.001 for each). In multivariable models adjusting for risk category, the clinical variables most predictive of in-hospital death were serious respiratory conditions and necrotizing enterocolitis. The area under the ROC curve for the full model including clinical risk factors was 0.92 in the KID 2000. The model was validated using data for KID 2003 and showed excellent discrimination (ROC = 0.90). This validated method provides a means of risk adjustment in groups of newborns undergoing noncardiac surgery, and should allow for comparative analyses of in-hospital mortality.

Cancer's Next Frontier

Cancer's Next Frontier

Risk Adjustment Using Administrative Data-Based and Survey-Derived Methods for Explaining Physician Utilization

The objective of this study was to evaluate an administrative data-based risk adjustment method for predicting physician utilization and the contribution of survey-derived indicators of health status. The results of this study will support the use of administrative data for planning, reimbursement, and assessing equity of physician utilization. The Ontario portion of the 2000-2001 Canadian Community Health Survey was linked with administrative physician claims data from 2002-2003 and 2003-2004. Explanatory models of family physician (FP) and specialist physician (SP) utilization were run using demographic information and The Johns Hopkins University Adjusted Clinical Groups (ACG) Case-mix System. Survey-based measures of health status were then added to the models. The coefficient of determination, R, indicated the models' explanatory power. The study sample consisted of 25,558 individuals aged 20 to 79 years representing approximately 7.8 million people. Over the 2 years of study period, 82.5% of the study population had a FP visit with a median of 6 visits and 53.2% had a SP visit with a median of 1 visit. The R values based on administrative data alone were 33% and 21% for the frequency of FP and SP visits and 16% and 35% for having one or more visit to an FPs and SPs, respectively. The addition of the survey-based measures to the administrative data-based models produced less than a 2% increase in explanatory power for any outcome. Administrative data-based measures of morbidity burden are valid and useful indicators of future physician utilization. The survey-derived measures used in this study did not contribute significantly to models on the basis of administrative data-based measures. These findings support the future use of administrative data-based data and Adjusted Clinical Groups for planning, reimbursement, and research.

Adverse event rates in congenital cardiac catheterization - a multi-center experience.

To describe case mix variation among institutions, and report adverse event rates in congenital cardiac catheterization by case type. Reported adverse event rates for patients with congenital heart disease undergoing cardiac catheterization vary considerably, due to non-comparable standards of data inclusion, and highly variable case mix. The Congenital Cardiac Catheterization Outcomes Project (C3PO) has been capturing case characteristics and adverse events (AE) for all cardiac catheterizations performed at six pediatric institutions. Validity and completeness of data were independently audited. Between 2/1/07 and 4/30/08, 3855 cases (670 biopsy, 1037 diagnostic, and 2148 interventional) were recorded, median number of cases per site 480 (308 to 1526). General anesthesia was used in 70% of cases (28 to 99%), and 22% of cases (15 to 26%) were non-electively or emergently performed. Three institutions performed a higher proportion of interventions during a case, 72 to 77% compared to 56 to 58%. The median rate of AE reported per institution was 16%, ranging from 5 to 18%. For interventional cases the median rate of AE reported per institution was 19% (7 to 25%) compared to 10% for diagnostic cases (6 to 16%). The incidence of AE was significantly higher for interventional compared to diagnostic cases (20% vs 10%, p<0.001), as was the incidence of higher severity AE (9% vs 5%, p<0.001). Adverse events in biopsy cases were uncommon. In this multi-institutional cohort, the incidence of AE is higher among interventional compared to diagnostic cases, and is very low among biopsy cases. Equitable comparisons among institutions will require the development and application of risk adjustment methods.

Improved hospital-level risk adjustment for surveillance of healthcare-associated bloodstream infections: a retrospective cohort study

BackgroundTo allow direct comparison of bloodstream infection (BSI) rates between hospitals for performance measurement, observed rates need to be risk adjusted according to the types of patients cared for by the hospital. However, attribute data on all individual patients are often unavailable and hospital-level risk adjustment needs to be done using indirect indicator variables of patient case mix, such as hospital level. We aimed to identify medical services associated with high or low BSI rates, and to evaluate the services provided by the hospital as indicators that can be used for more objective hospital-level risk adjustment.MethodsFrom February 2001-December 2007, 1719 monthly BSI counts were available from 18 hospitals in Queensland, Australia. BSI outcomes were stratified into four groups: overall BSI (OBSI), Staphylococcus aureus BSI (STAPH), intravascular device-related S. aureus BSI (IVD-STAPH) and methicillin-resistant S. aureus BSI (MRSA). Twelve services were considered as candidate risk-adjustment variables. For OBSI, STAPH and IVD-STAPH, we developed generalized estimating equation Poisson regression models that accounted for autocorrelation in longitudinal counts. Due to a lack of autocorrelation, a standard logistic regression model was specified for MRSA.ResultsFour risk services were identified for OBSI: AIDS (IRR 2.14, 95% CI 1.20 to 3.82), infectious diseases (IRR 2.72, 95% CI 1.97 to 3.76), oncology (IRR 1.60, 95% CI 1.29 to 1.98) and bone marrow transplants (IRR 1.52, 95% CI 1.14 to 2.03). Four protective services were also found. A similar but smaller group of risk and protective services were found for the other outcomes. Acceptable agreement between observed and fitted values was found for the OBSI and STAPH models but not for the IVD-STAPH and MRSA models. However, the IVD-STAPH and MRSA models successfully discriminated between hospitals with higher and lower BSI rates.ConclusionThe high model goodness-of-fit and the higher frequency of OBSI and STAPH outcomes indicated that hospital-specific risk adjustment based on medical services provided would be useful for these outcomes in Queensland. The low frequency of IVD-STAPH and MRSA outcomes indicated that development of a hospital-level risk score was a more valid method of risk adjustment for these outcomes.

A comparison between the APACHE II and Charlson Index Score for predicting hospital mortality in critically ill patients

BackgroundRisk adjustment and mortality prediction in studies of critical care are usually performed using acuity of illness scores, such as Acute Physiology and Chronic Health Evaluation II (APACHE II), which emphasize physiological derangement. Common risk adjustment systems used in administrative datasets, like the Charlson index, are entirely based on the presence of co-morbid illnesses. The purpose of this study was to compare the discriminative ability of the Charlson index to the APACHE II in predicting hospital mortality in adult multisystem ICU patients.MethodsThis was a population-based cohort design. The study sample consisted of adult (>17 years of age) residents of the Calgary Health Region admitted to a multisystem ICU between April 2002 and March 2004. Clinical data were collected prospectively and linked to hospital outcome data. Multiple regression analyses were used to compare the performance of APACHE II and the Charlson index.ResultsThe Charlson index was a poor predictor of mortality (C = 0.626). There was minimal difference between a baseline model containing age, sex and acute physiology score (C = 0.74) and models containing either chronic health points (C = 0.76) or Charlson index variations (C = 0.75, 0.76, 0.77). No important improvement in prediction occurred when the Charlson index was added to the full APACHE II model (C = 0.808 to C = 0.813).ConclusionThe Charlson index does not perform as well as the APACHE II in predicting hospital mortality in ICU patients. However, when acuity of illness scores are unavailable or are not recorded in a standard way, the Charlson index might be considered as an alternative method of risk adjustment and therefore facilitate comparisons between intensive care units.