Virtual Journal ClubACMQPranavi Sreeramoju, MD, MPHNovember 17, 2015

Sepsis• Systemic Inflammatory Response Syndrome (SIRS) requires the

presence of two of the following four factors:• Temperature < 36.0 C or > 38.0 C• Heart rate > 90 beats per minute• Respiratory rate > 24 breaths per minute or PaCO2 <32• WBC <4K or > 12K or bandemia >10%

• Sepsis is SIRS plus the suspected or known presence of infection.

• Severe Sepsis is sepsis plus evidence of organ dysfunction

• Septic Shock is sepsis plus hypotension (systolic blood pressure (SBP) <90 mmHg or mean arterial pressure (MAP) <70 mmHg or a SBP decrease >40 mmHg from baseline) that is unresponsive to adequate resuscitation.

Sepsis• Sepsis is the 11th cause of mortality in the US causing >38k

deaths (more if you add pneumonia)• Many deaths preventable with proper care in the first six

hours of detection

National Vital Statistics Report 2013

Sepsis BundleTO BE COMPLETED WITHIN 3 HOURS OF TIME OF PRESENTATION*: 1. Measure lactate level 2. Obtain blood cultures prior to administration of antibiotics 3. Administer broad spectrum antibiotics 4. Administer 30ml/kg crystalloid for hypotension or lactate

≥4mmol/L ( included in 3-hr bundle per Surviving Sepsis Campaign)

• * “Time of presentation” is defined as the time of triage in the emergency department or, if presenting from another care venue, from the earliest chart annotation consistent with all elements of severe sepsis or septic shock ascertained through chart review.

Sepsis Bundle

TO BE COMPLETED WITHIN 6 HOURS OF TIME OF PRESENTATION: 5. Administer 30ml/kg crystalloid for hypotension or lactate

≥4mmol/L (NQF counts this in the 6-hr bundle)6. Apply vasopressors (for hypotension that does not respond

to initial fluid resuscitation) to maintain a mean arterial pressure (MAP) ≥65mmHg

7. In the event of persistent hypotension after initial fluid administration (MAP < 65 mm Hg) or if initial lactate was ≥4 mmol/L, re-assess volume status and tissue perfusion and document findings

8. Re-measure lactate if initial lactate elevated.

Sepsis: Early Goal-Directed Therapy• Rivers 2001: Severe sepsis and septic shock: Single center -

130 receiving EGDT vs. 133 receiving standard therapy – mortality was 30.5% vs. 46.5%.

• PROCESS Trial 2014: Septic shock – 31 EDs – three groups - protocol-based EGDT (n=439), protocol-based standard therapy (n=446), or usual care (n=456). Mortality was 21%, 18.2% and 18.9% respectively.

• ARISE Trial 2014: Early septic shock – 51 EDs – EGDT (n=796) vs usual care (n=804). Mortality was 18.6% and 18.8% respectively.

• Mouncey et al 2015: Septic shock – 56 hospitals – EGDT (n=630) vs usual care (n=630). Mortality was 29.5% vs 29.2% respectively.

Sepsis• Surviving Sepsis Campaign collaborative: launched in 2002 –

7.5 year experience in 218 hospitals – severe sepsis and septic shock - hospitals with high compliance had 29.0% mortality vs. 38.6% in hospitals with low compliance with bundle (Levy MM et al. Crit Care Med. 2015 Jan;43(1))

• The UTSW/ Parkland experience• CMS initiative• University HealthSystem Consortium (UHC) data

THE IMPORTANCE OF ACCURATE MEASUREMENT!!

Stakes are getting higher and higher!

ARTICLEToday’s Journal Club

How much does sepsis mortality vary between hospitals?

Hospital Variations in Severe Sepsis MortalityHenry E. Wang, MD, MS1, John P. Donnelly, MSPH1,Nathan I. Shapiro, MD, MPH2, Samuel F. Hohmann, PhD, MS-HSM3,4 and Emily B. Levitan, ScD5

American Journal of Medical Quality 2015, Vol. 30(4) 328–336

1.University of Alabama School of Medicine, Birmingham, AL2.Beth Israel Deaconess Medical Center, Boston, MA3.University HealthSystem Consortium, Chicago, IL4.Rush University, Chicago, IL5.University of Alabama at Birmingham, Birmingham, AL

Setting and Context• UHC is a cooperative with 120 not-for-profit academic medical

centers and 300 affiliated hospitals in 42 states• ‘Collaboration that drives improvement’• Hospital discharge data set• Sepsis mortality and sepsis care are gaining increased

attention• ‘How much does sepsis mortality vary between

hospitals?’

UHC Sepsis Mortality Data• Observed to Expected ratio to calculate mortality index• Aggregate of POA and non-POA (although UHC does sub-

divide into each category)• 45-60 day delay• Comparison with other UHC hospitals• Ranking within UHC drives friendly competition

UHC Sepsis Mortality Data• Inclusions for UHC sepsis mortality data in general:• Includes Postoperative sepsis• Includes Septicemia of Newborn

• Exclusions for UHC sepsis mortality data in general:• Meningococcemia• “Bad Data”• Nonviable Neonate• Hospice

• Additional exclusions for this study:• <18 years of age• Prisoners• Admitted to a psychiatric unit• Transferred from another acute care hospital• <5 expected deaths

Study Design• Retrospective Observational• January 2012- December 2012• Measures: • Hospital death in a patient with severe sepsis per

coding/administrative data• Expected mortality calculated using UHC risk adjustment models• Mortality index calculated

• Analysis – comparison of mortality index in different hospitals

UHC Risk Adjustment Model• Separate logistic regression models are developed for each

base diagnosis group, leading to an estimated probability of hospital death.

• Each base model uses inpatient death as the dependent variable, and patient demographics and comorbid conditions as the independent variables.

• UHC incorporates additional variables into the model on an annual basis. The models are updated annually using at least 2 years of data from consortium medical centers.

• Hospital characteristics – census, teaching status, etc.

ResultsSevere sepsis observed mortality (median = 8.6%; IQR = 6.8%-10.3%; Range = 0.9%-18.2%)

ResultsSevere sepsis observed-to-expected (O:E) mortality ratios (median = 0.91; IQR = 0.77-1.05; Range = 0.16-1.95)

Conclusion

Variations in severe sepsis mortality were observed between UHC participating institutions

CRITIQUE

Strengths of Study• Robust data set• Established methodology for risk adjustment

Limitations of Study• None per se

• General limitation• Coding data and consistency among institutions

What does the variation mean?• Is the variation a reflection of true differences in sepsis care at

institutions?• What about other service delivery factors? E.g., availability of

hospice service• What about other infrastructure characteristics? E.g., coding

team • Need to explore this variation in follow-up studies• How would different hospitals at different points along the

spectrum respond to QI interventions to improve sepsis mortality?

Thank You!Comment? Question?