BACKGROUND/RATIONALE:
Care provided in the intensive care unit (ICU) is among the most costly to hospitals and risky to patients. A system that could both monitor performance and provide feedback is lacking although measurement is fundamental to improving any process or system. In this study we examine how ICU factors affect risk adjusted ICU outcomes.

OBJECTIVE(S):
This project used a computer based risk adjustment method to identify important contributors to indicators of quality (standardized mortality ratio, and observed minus expected length of say), compare ICUs, examine the effect if any of poverty, race, nursing staffing and unit characteristics on outcome.

METHODS:
Setting/cases: All first ICU admissions (29,377 cases) to 34 ICUs in 17 VA hospitals (Feb 96 thru July 97), and all ICU admissions to 35 ICUs in 15 VA hospital systems (June 99 through Dec 2000). Severity adjustment: The severity measure (SOI) used binary variables to represent ICU diagnosis from index hospitalization ICD-9-CM coding categorized into 38 mutually exclusive groups, comorbid disease burden determined from the index and past hospitalizations classified into 31 groups (adapting a validated method) [14] [15], admission source based on 2 mutually exclusive categories (direct admissions - emergency room or operating room, and transfer admission those from the floor or from other hospitals), and cubic splines for continuous variables including age, and the most abnormal laboratory value for 10 specific variables from 24 hours surrounding ICU admission (sodium, blood urea nitrogen, creatinine, glucose, bilirubin, hematocrit, white blood count, partial pressure of oxygen and a combined variable partial pressure of carbon dioxide and pH). The (AUROC on 11,646 validation dataset = 0.885). Other variables: We created a mean glucose value summing all measured glucose for each patient and dividing the number of measurements after excluding glucose values within 48 hours of death or discharge from the ICU. From the Nursing File, we extracted the number of staffing hours per shift divided into 3 groups (RN, LPN, AIDE) over the dataset (99 – 00). Cases from the 99 – 00 database were matched to the NSQIP file and to cases with APACHE II scoring to benchmark the system.

FINDINGS/RESULTS:
1.Hyperglycemia is common in VA ICUS, (31% cases had a mean glucose value >200 mg/dl) and 46% of cases did not carry ICD-9-CM codes for diabetes. A significant association with mortality was seen in patients with mean glucose values > 145 mg/dl after severity adjustment. The relationship with mortality was greater in those patients without diagnoses of diabetes. (manuscript in preparation)
2. In preliminary analysis, after adjustment for severity, lower numbers of nursing hours per patient were not independently associated with increase in morbid conditions or mortality.
3. Race was not associated with risk adjusted increased risk for mortality in the VA ICUs (96 – 97 sample) but reported income less than 50% of the poverty level (e.g., < $4100/ yr) was associated with increased risk adjusted mortality risk.
4. In the cohort, of 7411 patients, 81% cases had a predicted mortality less than or equal to 10%, and only 6% of cases had predicted mortality greater than 30%. We matched mortality prediction using the VA ICU risk adjustment and national Surgical Quality Improvement Project of 7411 cases in a 10 x 10 table by ascending decile of mortally risk. 80% of predictions matched within ± 1 decile, with the best agreement in cases with less than 30% predicted mortality.
5. In a study with Kaiser Permanente Northern California (Gabriel Escobar, PI), the VA ICU risk adjustment measure exhibited excellent discrimination and calibration in a cohort of 38,333 cases (c statistic for mortality prediction 0.86, R2 for length of stay prediction .21).

IMPACT:
1)Using this risk adjustment method to measure outcomes, we are collaborating with Peter Almenoff, Program Director, Pulmonary Critical Care and VISN Directors from 6 VISN (1,10,15,16,20,23) to to improve quality in their ICUs implementing evidenced based practices.

2) Gabriel Escobar, M.D. Kaiser- Permanente Northern California is using this model to understand utilization of ICU beds and admission and discharge practices. Together we have submitted a TRIP to benchmark outcomes and implement best practices in the VA and in Kaiser Permanente California.

---

External Links for this Project

---

PUBLICATIONS:

Journal Articles

1. Render ML, Kim HM, Welsh DE, Timmons S, Johnston J, Hui S, Connors AF, Wagner D, Daley J, Hofer TP. Automated intensive care unit risk adjustment: results from a National Veterans Affairs study. Critical care medicine. 2003 Jun 1; 31(6):1638-46. [view]
2. Render ML, Kim HM, Deddens J, Sivaganesin S, Welsh DE, Bickel K, Freyberg R, Timmons S, Johnston J, Connors AF, Wagner D, Hofer TP. Variation in outcomes in Veterans Affairs intensive care units with a computerized severity measure. Critical care medicine. 2005 May 1; 33(5):930-9. [view]

DRA: Health Systems, Acute and Combat-Related Injury
DRE: Etiology, Diagnosis
Keywords: Adverse events, Organizational issues, Outcomes, Risk factors
MeSH Terms: Intensive Care