Peter L. Almenoff

Hyperglycemia–related mortality in critically ill patients varies with admission diagnosis*

Objectives: Hyperglycemia during critical illness is common and is associated with increased mortality. Intensive insulin therapy has improved outcomes in some, but not all, intervention trials. It is unclear whether the benefits of treatment differ among specific patient populations. The purpose of the study was to determine the association between hyperglycemia and risk– adjusted mortality in critically ill patients and in separate groups stratified by admission diagnosis. A secondary purpose was to determine whether mortality risk from hyperglycemia varies with intensive care unit type, length of stay, or diagnosed diabetes. Design: Retrospective cohort study. Setting: One hundred seventy-three U.S. medical, surgical, and cardiac intensive care units. Patients: Two hundred fifty-nine thousand and forty admissions from October 2002 to September 2005; unadjusted mortality rate, 11.2%. Interventions: None. Measurements and Main Results: A two–level logistic regression model determined the relationship between glycemia and mortality. Age, diagnosis, comorbidities, and laboratory variables were used to calculate a predicted mortality rate, which was then analyzed with mean glucose to determine the association of hyperglycemia with hospital mortality. Hyperglycemia was associated with increased mortality independent of illness severity. Compared with normoglycemic individuals (70–110 mg/dL), adjusted odds of mortality (odds ratio, [95% confidence interval]) for mean glucose 111–145, 146–199, 200–300, and >300 mg/dL was 1.31 (1.26–1.36), 1.82 (1.74–1.90), 2.13 (2.03–2.25), and 2.85 (2.58–3.14), respectively. Furthermore, the adjusted odds of mortality related to hyperglycemia varied with admission diagnosis, demonstrating a clear association in some patients (acute myocardial infarction, arrhythmia, unstable angina, pulmonary embolism) and little or no association in others. Hyperglycemia was associated with increased mortality independent of intensive care unit type, length of stay, and diabetes. Conclusions: The association between hyperglycemia and mortality implicates hyperglycemia as a potentially harmful and correctable abnormality in critically ill patients. The finding that hyperglycemia–related risk varied with admission diagnosis suggests differences in the interaction between specific medical conditions and injury from hyperglycemia. The design and interpretation of future trials should consider the primary disease states of patients and the balance of medical conditions in the intensive care unit studied.

Incidence and outcomes of acute kidney injury in intensive care units: a Veterans Administration study.

Objectives:To examine the effect of severity of acute kidney injury or renal recovery on risk-adjusted mortality across different intensive care unit settings. Acute kidney injury in intensive care unit patients is associated with significant mortality. Design:Retrospective observational study. Setting:There were 325,395 of 617,927 consecutive admissions to all 191 Veterans Affairs ICUs across the country. Patients:Large national cohort of patients admitted to Veterans Affairs ICUs and who developed acute kidney injury during their intensive care unit stay. Measurements and Main Results:Outcome measures were hospital mortality, and length of stay. Acute kidney injury was defined as a 0.3-mg/dL increase in creatinine relative to intensive care unit admission and categorized into Stage I (0.3 mg/dL to <2 times increase), Stage II (≥2 and <3 times increase), and Stage III (≥3 times increase or dialysis requirement). Association of mortality and length of stay with acute kidney injury stages and renal recovery was examined. Overall, 22% (n = 71,486) of patients developed acute kidney injury (Stage I: 17.5%; Stage II: 2.4%; Stage III: 2%); 16.3% patients met acute kidney injury criteria within 48 hrs, with an additional 5.7% after 48 hrs of intensive care unit admission. Acute kidney injury frequency varied between 9% and 30% across intensive care unit admission diagnoses. After adjusting for severity of illness in a model that included urea and creatinine on admission, odds of death increased with increasing severity of acute kidney injury. Stage I odds ratio = 2.2 (95% confidence interval, 2.17–2.30); Stage II odds ratio = 6.1 (95% confidence interval, 5.74, 6.44); and Stage III odds ratio = 8.6 (95% confidence interval, 8.07–9.15). Acute kidney injury patients with sustained elevation of creatinine experienced higher mortality risk than those who recovered. Interventions:None. Conclusions:Admission diagnosis and severity of illness influence frequency and severity of acute kidney injury. Small elevations in creatinine in the intensive care unit are associated with increased risk-adjusted mortality across all intensive care unit settings, whereas renal recovery was associated with a protective effect. Strategies to prevent even mild acute kidney injury or promote renal recovery may improve survival.

Veterans Affairs intensive care unit risk adjustment model: validation, updating, recalibration.

Background:A valid metric is critical to measure and report intensive care unit (ICU) outcomes and drive innovation in a national system. Objectives:To update and validate the Veterans Affairs (VA) ICU severity measure (VA ICU). Research Design:A validated logistic regression model was applied to two VA hospital data sets: 36,240 consecutive ICU admissions to a stratified random sample of moderate and large hospitals in 1999–2000 (cohort 1) and 81,964 cases from 42 VA Medical Centers in fiscal years 2002–2004 (cohort 2). The model was updated by adding diagnostic groups and expanding the source of admission variables. Measures:C statistic, Hosmer-Lemeshow goodness-of-fit statistic, and Briers score measured predictive validity. Coefficients from the 1997 model were applied to predictors (fixed) in a logistic regression model. A 10 × 10 table compared cases with both VA ICU and National Surgical Quality Improvement Performance metrics. The standardized mortality ratios divided observed deaths by the sum of predicted mortality. Results:The fixed model in both cohorts had predictive validity (cohort 1: C statistic = 0.874, Hosmer-Lemeshow goodness-of-fit C statistic chi-square = 72.5; cohort 2: 0.876, 307), as did the updated model (cohort 2: C statistic = 0.887, Hosmer-Lemeshow goodness-of-fit C statistic chi-square = 39). In 7,411 cases with predictions in both systems, the standardized mortality ratio was similar (1.04 for VA ICU, 1.15 for National Surgical Quality Improvement Performance), and 92% of cases matched (±1 decile) when ordered by deciles of mortality. The VA ICU standardized mortality ratio correlates with the National Surgical Quality Improvement Performance standardized mortality ratio (r2 = .74). Variation in discharge and laboratory practices may affect performance measurement. Conclusion:The VA ICU severity model has face, construct, and predictive validity.

Prolongation of the half-life of lactate after maximal exercise in patients with hepatic dysfunction

Decreased hepatic clearance of exogenous sodium lactate has previously been demonstrated in patients with hepatic dysfunction. The purpose of this study was to obtain a more precise understanding of the rate of metabolic normalization or decrease of endogenously produced lactate in patients with hepatic cirrhosis. The differential kinetics of lactate metabolism are of clinical interest. Male volunteer patients with hepatic cirrhosis (n = 7), who had survived acute hospitalization, were compared to healthy age-matched males with normal liver function (n = 7). After arterial cannulation, bicycle ergometry was performed at a workload of 25 watts (W); the load was increased by increments of 25 W at 2-min intervals to maximum aerobic capacity. Lactate was measured in arterial blood before, at 4-min intervals during, and on a minimum of 11 occasions in the 30 to 70 min after exercise. The time interval during which lactate declined linearly to half its maximal concentration (Lt50) was graphically computed. The Lt50 was 34.8 +/- 4.5 min (mean +/- SEM) in the experimental group and 14.1 +/- 1.3 min in the control subjects (p less than .005). Lactate disappears from the bloodstream almost three times more slowly in patients with hepatic cirrhosis. The implication for interpretation of changes in lactate during circulatory shock in the presence of liver dysfunction is addressed.

Reduction of central line infections in Veterans Administration intensive care units: an observational cohort using a central infrastructure to support learning and improvement

Background Elimination of hospital-acquired infections is an important patient safety goal. Setting All 174 medical, cardiac, surgical and mixed Veterans Administration (VA) intensive care units (ICUs). Intervention A centralised infrastructure (Inpatient Evaluation Center (IPEC)) supported the practice bundle implementation (handwashing, maximal barriers, chlorhexidinegluconate site disinfection, avoidance of femoral catheterisation and timely removal) to reduce central line-associated bloodstream infections (CLABSI). Support included recruiting leadership, benchmarked feedback, learning tools and selective mentoring. Data collection Sites recorded the number of CLABSI, line days and audit results of bundle compliance on a secure website. Analysis CLABSI rates between years were compared with incidence rate ratios (IRRs) from a Poisson regression and with National Healthcare Safety Network referent rates (standardised infection ratio (SIR)). Pearsons correlation coefficient compared bundle adherence with CLABSI rates. Semi-structured interviews with teams struggling to reduce CLABSI identified common themes. Results From 2006 to 2009, CLABSI rates fell (3.8–1.8/1000 line days; p<0.01); as did IRR (2007; 0.83 (95% CI 0.73 to 0.94), 2008; 0.65 (95% CI 0.56 to 0.76), 2009; 0.47 (95% CI 0.40 to 0.55)). Bundle adherence and CLABSI rates showed strong correlation (r=0.81). VA CLABSI SIR, January to June 2009, was 0.76 (95% CI 0.69 to 0.90), and for all FY2009 0.88 (95% CI 0.80 to 0.97). Struggling sites lacked a functional team, forcing functions and feedback systems. Conclusion Capitalising on a large healthcare system, VA IPEC used strategies applicable to non-federal healthcare systems and communities. Such tactics included measurement through information technology, leadership, learning tools and mentoring.

Infrastructure for quality transformation: measurement and reporting in veterans administration intensive care units

Background Veterans Health Administration (VA) intensive care units (ICUs) develop an infrastructure for quality improvement using information technology and recruiting leadership. Methods Setting Participation by the 183 ICUs in the quality improvement program is required. Infrastructure includes measurement (electronic data extraction, analysis), quarterly web-based reporting and implementation support of evidence-based practices. Leaders prioritise measures based on quality improvement objectives. The electronic extraction is validated manually against the medical record, selecting hospitals whose data elements and measures fall at the extremes (10th, 90th percentile). Results are depicted in graphic, narrative and tabular reports benchmarked by type and complexity of ICU. Results The VA admits 103u2008689±1156 ICU patients/year. Variation in electronic business practices, data location and normal range of some laboratory tests affects data quality. A data management website captures data elements important to ICU performance and not available electronically. A dashboard manages the data overload (quarterly reports ranged 106—299 pages). More than 85% of ICU directors and nurse managers review their reports. Leadership interest is sustained by including ICU targets in executive performance contracts, identification of local improvement opportunities with analytic software, and focused reviews. Conclusion Lessons relevant to non-VA institutions include the: (1) need for ongoing data validation, (2) essential involvement of leadership at multiple levels, (3) supplementation of electronic data when key elements are absent, (4) utility of a good but not perfect electronic indicator to move practice while improving data elements and (5) value of a dashboard.

Despite variation in volume, Veterans Affairs hospitals show consistent outcomes among patients with non-postoperative mechanical ventilation

Objective:To assess the relationship between volume of nonoperative mechanically ventilated patients receiving care in a specific Veterans Health Administration hospital and their mortality. Design:Retrospective cohort study. Setting:One-hundred nineteen Veterans Health Administration medical centers. Patients:We identified 5,131 hospitalizations involving mechanically ventilated patients in an intensive care unit during 2009, who did not receive surgery. Interventions:None. Measurements and Main Results:We extracted demographic and clinical data from the VA Inpatient Evaluation Center. For each hospital, we defined volume as the total number of nonsurgical admissions receiving mechanical ventilation in an intensive care unit during 2009. We examined the hospital contribution to 30-day mortality using multilevel logistic regression models with a random intercept for each hospital. We quantified the extent of interhospital variation in 30-day mortality using the intraclass correlation coefficient and median odds ratio. We used generalized estimating equations to examine the relationship between volume and 30-day mortality and risk-adjusted all models using a patient-level prognostic score derived from clinical data representing the risk of death conditional on treatment at a high-volume hospital. Mean age for the sample was 65 (SD 11) yrs, 97% were men, and 60% were white. The median VA hospital cared for 40 (interquartile range 19–62) mechanically ventilated patients in 2009. Crude 30-day mortality for these patients was 36.9%. After reliability and risk adjustment to the median patient, adjusted hospital-level mortality varied from 33.5% to 40.6%. The intraclass correlation coefficient for the hospital-level variation was 0.6% (95% confidence interval 0.1, 3.4%), with a median odds ratio of 1.15 (95% confidence interval 1.06, 1.38). The relationship between hospital volume of mechanically ventilated and 30-day mortality was not statistically significant: each 50-patient increase in volume was associated with a nonsignificant 2% decrease in the odds of death within 30 days (odds ratio 0.98, 95% confidence interval 0.87–1.10). Conclusions:Veterans Health Administration hospitals caring for lower volumes of mechanically ventilated patients do not have worse mortality. Mechanisms underlying this finding are unclear, but, if elucidated, may offer other integrated health systems ways to overcome the disadvantages of small-volume centers in achieving good outcomes.

Variations In Efficiency And The Relationship To Quality Of Care In The Veterans Health System

There is widespread belief that the US health care system could realize significant improvements in efficiency, savings, and patient outcomes if care were provided in a more integrated and accountable way. We examined efficiency and its relationship to quality of care for medical centers run by the Veterans Health Administration of the Department of Veterans Affairs (VA), a national, vertically integrated health care system that is accountable for a large patient population. After devising a statistical model to indicate efficiency, we found that VA medical centers were highly efficient. We also found only modest variation in the level of efficiency and cost across VA medical centers, and a positive correlation overall between greater efficiency and higher inpatient quality. These findings for VA medical centers suggest that efforts to drive integration and accountability in other parts of the US health care system might have important payoffs in reducing variations in cost without sacrificing quality. Policy makers should focus on what aspects of certain VA medical centers allow them to provide better care at lower costs and consider policies that incentivize other providers, both within and outside the VA, to adopt these practices.

Predicting potentially avoidable hospitalizations.

Background:Hospitalizations due to ambulatory care sensitive conditions (ACSCs) are widely accepted as an indicator of primary care access and effectiveness. However, broad early intervention to all patients in a health care system may be deemed infeasible due to limited resources. Objective:To develop a predictive model to identify high-risk patients for early intervention to reduce ACSC hospitalizations, and to explore the predictive power of different variables. Methods:The study population included all patients treated for ACSCs in the VA system in fiscal years (FY) 2011 and 2012 (n=2,987,052). With all predictors from FY2011, we developed a statistical model using hierarchical logistic regression with a random intercept to predict the risk of ACSC hospitalizations in the first 90 days and the full year of FY2012. In addition, we configured separate models to assess the predictive power of different variables. We used a random split-sample method to prevent overfitting. Results:For hospitalizations within the first 90 days of FY2012, the full model reached c-statistics of 0.856 (95% CI, 0.853–0.860) and 0.856 (95% CI, 0.852–0.860) for the development and validation samples, respectively. For predictive power of the variables, the model with only a random intercept yielded c-statistics of 0.587 (95% CI, 0.582–0.593) and 0.578 (95% CI, 0.573–0.583), respectively; with patient demographic and socioeconomic variables added, the c-statistics improved to 0.725 (95% CI, 0.720–0.729) and 0.721 (95% CI, 0.717–0.726), respectively; adding prior year utilization and cost raised the c-statistics to 0.826 (95% CI, 0.822–0.830) and 0.826 (95% CI,0.822–0.830), respectively; the full model was reached with HCCs added. For the 1-year hospitalizations, only the full model was fitted, which yielded c-statistics of 0.835 (95% CI, 0.831–0.837) and 0.833 (95% CI, 0.830–0.837), respectively, for development and validation samples. Conclusions:Our analyses demonstrate that administrative data can be effective in predicting ACSC hospitalizations. With high predictive ability, the model can assist primary care providers to identify high-risk patients for early intervention to reduce ACSC hospitalizations.

Risk Adjustment Tools for Learning Health Systems: A Comparison of DxCG and CMS‐HCC V21

OBJECTIVEnTo compare risk scores computed by DxCG (Verisk) and Centers for Medicare and Medicaid Services (CMS) V21.nnnRESEARCH DESIGNnAnalysis of administrative data from the Department of Veterans Affairs (VA) for fiscal years 2010 and 2011.nnnSTUDY DESIGNnWe regressed total annual VA costs on predicted risk scores. Model fit was judged by R-squared, root mean squared error, mean absolute error, and Hosmer-Lemeshow goodness-of-fit tests. Recalibrated models were tested using split samples with pharmacy data.nnnDATA COLLECTIONnWe created six analytical files: a random sample (nxa0=xa02xa0million), high cost users (nxa0=xa0261,487), users over age 75 (nxa0=xa0644,524), mental health and substance use users (nxa0=xa0830,832), multimorbid users (nxa0=xa0817,951), and low-risk users (nxa0=xa078,032).nnnPRINCIPAL FINDINGSnThe DxCG Medicaid with pharmacy risk score yielded substantial gains in fit over the V21 model. Recalibrating the V21 model using VA pharmacy data-generated risk scores with similar fit statistics to the DxCG risk scores.nnnCONCLUSIONSnAlthough the CMS V21 and DxCG prospective risk scores were similar, the DxCG model with pharmacy data offered improved fit over V21. However, health care systems, such as the VA, can recalibrate the V21 model with additional variables to develop a tailored risk score that compares favorably to the DxCG models.