icon-    folder.gif   Conference Reports for NATAP  
  International AIDS Conference
Durban, South Africa
July 18-22 2016
Back grey_arrow_rt.gif
Two Methods Offer Good Mortality Prediction in Critically Ill HIV Patients
  21st International AIDS Conference (AIDS 2016), July 18-22, 2016, Durban, South Africa
Mark Mascolini
Two well-known algorithms provided "moderate to good" predictions of mortality in HIV-positive people admitted to a US intensive care unit (ICU) in 2007-2013 [1]. Among 4914 HIV patients studied, 8.5% needed ICU care.
University of Washington researchers who conducted this study noted that critical illness and ICU admission remain frequent among people with HIV despite an overall drop in mortality in recent years. They conducted this study to compare predictions of mortality in the hospital, after 30 days, and after 1 year with two tools: APACHE II (Acute Physiology and Chronic Health Evaluation) and VACS (Veterans Aging Cohort Study risk index).
The researchers used electronic medical records, administrative data, and state death data to identify HIV-positive people who used the ICU at the University of Washington or Harborview Medical Center, to figure APACHE and VACS scores for each person, and to determine who died. They used logistic regression to assess mortality predictions with APACHE and VACS in the hospital, after 30 days, and after 1 year. They compared predictive value of the two systems by determining area under the receiver operating characteristic curve (ROC AUC) with the DeLong method.
From October 2007 through December 2013, University of Washington units cared for 4914 people with HIV, 419 (8.5%) of whom became critically ill and required ICU admission. The ICU group had a median age of 47 (range 18 to 77), 17% were women, 59% white, and 22% black. Almost half of these people (49%) had a viral load below 500 copies, 38% had a CD4 count below 200, and 43% had HCV coinfection. The largest proportions got admitted to the ICU with neurologic diagnoses (28%), respiratory conditions (15%), or cardiovascular problems (11%). While 7% of people admitted to the ICU died in the hospital, 8% died in 30 days and 20% died in 1 year.
Median APACHE score stood at 19 (interquartile range 13 to 26) and median VACS score at 57 (interquartile range 34 to 80). Every 5 point higher APACHE or VACS score was significantly associated with hospital mortality (odds ratio [OR] 1.722 for APACHE and 1.114 for VACS), 30-day mortality (OR 1.652 for APACHE and 1.144 for VACS), and 1-year mortality (OR 1.323 for APACHE and 1.132 for VACS) (P < 0.001 for all associations). ROC AUC analysis discerned little difference in predictive value with APACHE versus VACS for in-hospital mortality or 30-day mortality. But there was a trend toward greater discrimination with VACS than APACHE for 1-year mortality (ROC AUC 0.702, 95% confidence interval 0.637 to 0.766, for VACS and 0.634, 95% confidence interval 0.568 to 0.701, for APACHE, P = 0.0661).
The investigators underlined the frequency of ICU admission in this contemporary US HIV population. They rated APACHE and VACS "moderate to good" in predicting short- and long-term mortality. And they suggested that "these metrics can be used by clinicians, researchers and health administrators to predict mortality or standardize admission disease severity in HIV-positive patients with critical illness."
1. Goldman J, Crothers K, Fong C, et al. Mortality prediction in HIV-infected persons with critical illness. 21st International AIDS Conference (AIDS 2016). July 18-22, 2016. Durban, South Africa. Abstract TUPEB032.