Anthony R BurrellMBBS, FANZCA, FCICM1

Mary-Louise McLawsDipTropPubHlth, MPH, PhD2

Mary FullickRN, BHlthSci(Nurs)1

Rosemary B SullivanRN, BHlthSci(Admin),

MBus(HRM)1

DoungkamolSindhusakePhD, MPH, MA1

1 Clinical ExcellenceCommission, Sydney, NSW.

2 University of New SouthWales, Sydney, NSW.

m.mclaws@unsw.edu.aurosemary.sullivan@health.nsw.gov.au

doi: 10.5694/mja15.00657

Online first 25/01/16

Research

SEPSIS KILLS: early intervention saves lives

Abstract

he increasing incidence ofsepsis is well recognised, and

Objective: To implement a statewide program for the early recognitionand treatment of sepsis in New South Wales, Australia.

Setting: Ninety-seven emergency departments in NSW hospitals.

Intervention: A quality improvement program (SEPSIS KILLS) thatpromoted intervention within 60 minutes of recognition, including taking ofblood cultures, measuring serum lactate levels, administration ofintravenous antibiotics, and fluid resuscitation.

Main outcome measures: Time to antibiotics and fluid resuscitation;mortality rates and length of stay.

Results: Data for 13567 patients were entered into the database. Theproportion of patients receiving intravenous antibiotics within 60 minutes oftriage increased from 29.3% in 2009e2011 to 52.2% in 2013. Thepercentage for whom a second litre of fluid was started within 60 minutesrose from 10.6% to 27.5% (each P<0.001). The proportion of patientsclassed as Australasian Triage Scale (ATS) 1 increased from 2.3% in2009e2011 to 4.2% in 2013, and the proportion classed as ATS 2 rose from40.7% in 2009e2011 to 60.7% in 2013 (P<0.001). There was a lineardecrease in mortality from 19.3% in 2009e2011 to 14.1% in 2013; there wasalso a significant decline in time in intensive care and total length of stay(each P<0.0001). The mortality rate for patients with severe sepsis (serumlactate �4mmol/L or systolic blood pressure [SBP] <90mmHg) was19.7%. The mortality rates for patients with severe sepsis admitted tointensive care and for those admitted to a ward did not change significantlyover time. The proportion of patients with uncomplicated sepsis(SBP �90mmHg, serum lactate <4mmol/L) transferred to a wardincreased, and the mortality rate after transfer increased from 3.2% in2009e2011 to 6.2% in 2013 (P<0.05). The survival benefit was greatest forpatients with evidence of haemodynamic instability (SBP <90mmHg) butnormal lactate levels (P¼0.03).

Conclusions: The SEPSIS KILLS program has improved the process ofcare for patients with sepsis in NSW hospitals. The program has focusedattention on sepsis management in the wards.

T is generally attributed to thegrowing prevalence of chronic con-ditions in ageing populations.1-3 InNew South Wales, the number ofpatients with a diagnosis of sepsis inthe Admitted Patient Data Collection(APDC) has increased, and sepsiswas involved in 17.5% of in-hospitaldeaths in 2009, compared with amortality of 1.5% for all hospitalseparations (unpublished data).

The clinical presentation of sepsismay be subtle; fever is not alwayspresent.4,5 In NSW, failure to recog-nise and respond to sepsis has beenregularly reported. In 2009, 167 in-cidents were highlighted in a clinicalfocus report published by theClinicalExcellence Commission (CEC).6 AQuality Systems Assessment in 2011,completed by over 1500 respondentsacross the NSW hospital system, re-ported that 34% of clinical units didnot have guidelines or protocols formanaging sepsis.7

This article reports on the SEPSISKILLS program of the CEC, whichaims to promote the skills andknowledge needed for recognisingand managing patients withsepsis in NSW hospital emergencydepartments.

Methods

The focus of the program is toRECOGNISE risk factors, signs andsymptoms of sepsis; RESUSCITATEwith rapid intravenous fluids andantibiotics; and REFER to senior cli-nicians and teams. Standardisedsepsis tools were developed inconsultation with NSW emergencyphysicians, and included adult andpaediatric pathways that built on theNSW deteriorating patient system,Between the Flags (BTF).8 The vitalsigns assessed in the sepsis path-ways were consistent with BTF, andvaried marginally from acceptedsystemic inflammatory responsecriteria (Box 1).

The SEPSIS KILLS pathways pro-mote bundles of care, with the

emphasis on early intervention. Theadult bundle includes taking bloodcultures, measuring serum lactatelevels, administering intravenousantibiotics within an hour of triageand recognition, and administeringa fluid bolus of 20mL/kg, followedby another bolus of 20mL/kg (ifnecessary) and inotropic drugs ifthe patient’s condition is not fluid-responsive. If no improvement isobserved, senior medical review andadmission to intensive care orretrieval to a major centre should beconsidered (Box 2). The paediatricbundle emphasises the importanceof early senior clinical review anddecision making. In addition, anempiric antibiotic guideline wasdeveloped with advice fromexpert infectious disease physicians.Emphasis was placed on the firstdose of antibiotics, thereby allowing

MJA 204 (2)

time for further assessment anddiagnosis. Because ofwide variationsin practice, the guideline alsocontained details on how antibi-otics could be administered mostexpeditiously.

The program was implemented in2011 with a top-down, bottom-upapproach, with strong leadershipfrom medical and nursing clinicalleads, and supported by the localhealth district Clinical GovernanceUnits. Participation was not manda-tory, and no extra resources wereprovided to participating emergencydepartments who implemented theprogram. The CEC team supportedclinicians by holding a prelimi-nary launch, monthly CECehospitalteleconferences, executive reports,newsletters, site visits and work-shops. A range of online resources

j 1 February 2016 1.e1

1 The SEPSIS KILLS pathway for adult patients in hospital emergency departments, page 1

1.e2

Research

was provided, including a SepsisToolkit (implementation guide) andvarious education tools.

Emergency departments wereencouraged to collect prospectivedata on paediatric and adult patients

MJA 204 (2) j 1 February 2016

with a provisional diagnosis of sepsiswho had received intravenous anti-biotics. An online sepsis database(from August 2011) facilitatedcollection of a minimum datasetfor each patient that included theirdate of birth, triage time and date,

triage category, clinical observations(including systolic blood pressure[SBP] and serum lactate levels),time and date of initial intrave-nous antibiotic treatment and ofcommencement of the second litre ofintravenous fluid, the presumptivesource of sepsis, and the dispositionof patients following emergencydepartment treatment. Data werecollected either prospectively or byretrospective chart review. Thedatabase allowed emergency de-partments to monitor time to antibi-otics and fluids in real time, and tocompare this with the correspondinglocal health district and NSW data.

Ethics approval was obtained fromthe NSW Sepsis Register, which wasdeveloped as a public health anddisease register under s98 of thePublic Health Act 2011. The SepsisRegister is maintained by the CEC.

Data analysisAnalysis of processmeasures (time toantibiotic, time to intravenous fluid)was based on data from the SEPSISKILLS database. A total of 13 567SEPSIS KILLS records were submit-ted for linkage to the APDC to assessassociations between in-hospitalmortality and sepsis severity andpatient disposition. Patients wereclassified by emergency departmentstaff according to the AustralasianTriage Scale (ATS).9 The cases werefurther classified as being severe oruncomplicated sepsis according tothe serum lactate levels and present-ing SBP of the patient.

To assess the population-level impactof the SEPSIS KILLS program, weanalysed health outcomes (in-hospi-tal mortality, hours in intensive care,length of stay) for paediatric andadult patients separated from NSWhospitals with ICD-10-AM (Interna-tional Classification of Diseases,10th revision, Australian modifica-tion) discharge diagnosis codesconsistent with sepsis10 recorded inthe Admitted Patient, EmergencyDepartment Attendance and DeathsRegister. This register was accessedthrough the NSWMinistry of HealthSecure Analytics for PopulationHealth Research and Intelligence(SAPHaRI) system. Linkage was

2 The SEPSIS KILLS pathway for adult patients in hospital emergency departments, page 2

Research

undertaken by the Centre for HealthRecord Linkage (CHeReL). Only pa-tients admitted to public hospitalswith emergency departments wereincluded in the analysis. Trend ana-lysis was performed for the run-inperiod, August 2009 e December2011, and for the two following years,2012 and 2013. Outcomes by sepsisseverity could not be analysed at thepopulation level because of the lackof consensus aboutusing ICD-10-AMcodes to differentiate between severeand uncomplicated sepsis.

Descriptive and inferential ana-lyses included the calculation of fre-quencies, odds ratios (ORs) and95% confidence intervals, and c2 testsfor trends. Trends over time for pro-cess and outcome measures wereassessed in regression models. Logis-tic regression was used to analysein-hospital deaths, while linearregression models were used for timein intensive care and lengths of stay.Models were adjusted as appropriatefor covariates (age, year, triage cate-gory and severity of sepsis). Statisticalsignificance was defined as P<0.05.

Results

The SEPSIS KILLS programwas implemented as individualemergency departments becameready during 2011. Both retrospec-tive and prospective data werecollected by 97 hospitals to 31December 2013 and entered into thesepsis database.Datawere submittedby 13 tertiary, 19metropolitan and 65rural hospitals. Because of the lownumber of paediatric patients, ana-lysis was restricted to data for adultpatients.

The provisional sources of sepsisincluded the lungs (5216 patients,40.5%), urinary tract (2998, 23.2%),abdomen (1077, 8.4%), skin or softtissue (975, 7.6%), musculoskeletalsystem (98, 0.8%), central nervoussystem (96, 0.7%), vascular device(82, 0.6%), and other systems (973,7.6%). For 1238 patients (9.6%) thesource was unidentified, for 133(1.0%) no source was recorded.

There were age data in 12 879 re-cords. There was a significant

reduction in the mean age of patientsbetween 2009 and 2013, from 67.3years in 2009e2011 to 64.8 years in2013 (P< 0.001 for trend; Box 3).

Data for the process indicators fromthe CEC sepsis database are sum-marised in Box 3. The proportion of

patients who were categorised attriage as ATS 1 (“see immediately”)rose from 2.3% in 2009e2011 to4.2% in 2013. Those categorised asATS 2 (“see within 10 minutes”)increased from 40.7% in 2009e2011to 60.7% in 2013 (P< 0.001). Therewere small reductions in the

MJA 204 (2) j 1 February 2016 1.e3

3 Characteristics, and process and outcome indicators of sepsis-related hospital separations before and afterthe launch of the SEPSIS KILLS program

Characteristics

Run-in period SEPSIS KILLS program in operationP fortrendAug 2009 e Dec 2011 2012 2013

Number of separations 1585 5396 5905

Mean age � SEM, years 67.3 � 0.5 67.6 � 0.3 64.8 � 0.3 <0.001

Triage category* <0.001

1 37 (2.3%) 176 (3.3%) 242 (4.2%)

2 463 (40.7%) 2765 (51.5%) 3532 (60.7%)

3 683 (43.2%) 2034 (37.9%) 1767 (30.4%)

4 207 (13.1%) 378 (7.0%) 267 (4.6%)

5 10 (0.6%) 16 (0.3%) 12 (0.2%)

Missing data 5 22 83

Antibiotic received within 60min 464 (29.3%) 2165 (40.2%) 3083 (52.2%) <0.001

Second litre of intravenous fluidwithin 60min

135 of 1272 patients(10.6%)

521 of 3631 patients(14.3%)

991 of 3609 patients(27.5%)

<0.001

SEM ¼ standard error of the mean. *Percentages exclude patients for whom data was not available (missing data). u

1.e4

Research

proportions of patients classed asATS 3, 4 or 5.

The proportion of patients whoreceived antibiotics within 60 mi-nutes of triage or recognitionincreased from 29.3% in 2009e2011to 52.2% in 2013 (linear trend test,P< 0.001). Similarly, the number ofpatients who started a second litre ofintravenous fluid within one hourrose from 10.6% to 27.5% (lineartrend test, P< 0.001).

The analysis of population-basedAPDC data, which included all

4 Hospital outcomes prior to and afDecember 2013

Outcomes

Deaths, numbers (percentage)

2009e2011

2012

2013

Mean time in intensive care (SEM), hours

2009e2011

2012

2013

Mean length of stay (SEM), days

2009e2011

2012

2013

SEM ¼ standard error of the mean. Source: Admittand Intelligence (SAPHaRI). Data extracted on 1 Ju

MJA 204 (2) j 1 February 2016

separations with emergency depart-ment involvement from public hos-pitals in NSW between January 2009and December 2013, is presented inBox 4. There were 15 801 sepsis hos-pital separations during the run-inperiod of 2009e2011, with a mortal-ity of 19.3%. This rate declined to17.2% in 2012 and 14.1% in 2013.There was a significant lineardecrease over time (P< 0.0001); theOR for death (compared with therun-in period) was 0.87 (95% CI,0.80e0.94) in 2012, and 0.69 (95% CI,0.63e0.74) in 2013. Significant linear

ter the launch of the SEPSIS KILLS pro

Descriptive statistics Od

3053/15 801 (19.3%)

979/5683 (17.2%) 0

870/6167 (14.1%) 0

32.7 (1.0)

26.6 (1.4)

25.8 (1.3)

13.5 (0.1)

12.2 (0.2)

11.5 (0.2)

ed Patient Data Collection, NSW Ministry of Health Sne 2015. u

declines were also observed for timein intensive care and length of stay(for each trend: P< 0.0001).

Linkage of the APDC and sepsis da-tabases showed that the mortalityrate for the 1616 patients with severesepsis (serum lactate � 4mmol/L orSBP < 90mmHg) was 19.7%; thesepatients were significantly morelikely to die than patients withuncomplicated sepsis (serumlactate < 4mmol/L and SBP� 90mmHg) (OR, 3.7; 95% CI,3.2e4.4; P< 0.0001). The mortality

gram, NSW, January 2009 to

ds ratio (95% CI) P for trend

<0.0001

1

.87 (0.80e0.94)

.69 (0.63e0.74)

<0.0001

<0.0001

ecure Analytics for Population Health Research

Research

rate for the 893 patients with hyper-lactataemia (a lactate level of4mmol/L or more; reference inter-val, 0.5e2.0mmol/L) was 24.9%,while that for 637 patients presentingwith cryptic shock — hyper-lactataemia together with normo-tension (SBP � 90mmHg) — was21.2%. There was no change in mor-tality for either group over time.For 734 patients who presentedwith SBP < 90mmHg and lactatelevels < 4mmol/L, mortality was13.5%, which declined significantlyacross the study period (2009e2011,16.5%; 2012, 16.0%; 2013, 9.8;P¼ 0.03). The overall mortality ratefor uncomplicated sepsis patientsincreased significantly over time:3.7% (21/567) in 2009e2011,6.2% (145/2336) in 2012, and6.7% (145/2164) in 2013 (P¼ 0.02).

The mortality rate for the 268 ATS 1patients was 28.8%. The risk of deathfor patients over 65 years of age was3.3 times higher (95% CI, 2.6e4.1)than for patients under 65 years ofage (P¼ 0.001).

The mortality rate for 543 severesepsis patients admitted to inten-sive care did not change signifi-cantly over time — 23.4% (2009e2011), 19.5% (2012) and 16.0% (2013)(P¼ 0.145) — nor did the proportionof the 1073 patients with severesepsis who were admitted to theward and died— 21.4% (2009e2011),21.5% (2012) and 18.4% (2013)(P¼ 0.263). In contrast, the risk ofdeath for 4225 patients withuncomplicated sepsis admitted tothe ward increased significantly:3.2% (15/466) during 2009e2011,6.0% (115/1914) during 2012, and6.2% (115/1845) during 2013(P¼ 0.047).

Discussion

SEPSIS KILLS is a quality improve-ment program that aims to reducepreventable harm to patients withsepsis by recognising the conditionearly and managing it promptly. It isbased on the principle that earlyrecognition and aggressive man-agement with antibiotics andfluids will improve outcomes.11-13

It is consistent with the 3-hour

component of the resuscitationbundle outlined in the internationalguidelines of the Surviving SepsisCampaign.3

The program was not planned as abefore-and-after project, but wasindependently implemented by in-dividual emergency departmentsduring 2011. More than 80% of NSWemergency departments (175 of 220)used the sepsis pathways, and 97emergency departments submittedover 13 000 records. The resultingincrease in the number of patients forwhom antibiotics were initiatedwithin 60 minutes of recognition andthe increased likelihood of the secondlitre of fluids being started in the firsthour indicate that the program wassuccessful. Greater urgency is alsoapparent from themarked increase inthe number of patients classified attriage as ATS 2.We cannot, however,explain the significant age differencebetween the patients seen in 2012and 2013.

Reviewing the population-basedAPDC hospitalisations with an ICD-10-AM code for sepsis showed thatthere was a steady reduction inmortality over time. Contrary towhat we expected, the survivalbenefit in our patients appears tohave been greatest for thosewith evidence of haemodynamicinstability (SBP < 90mmHg) butnormal lactate levels.

The mortality rate of 15%e20% forpatients admitted to intensive carewith severe sepsis (one-third of theoverall sample) is consistent with theoverall mortality rate in Australianand New Zealand intensive careunits.14,15 In 2013, the crudemortalityrate for the patients admitted towards was higher than that for theintensive care group. We believe therelatively high proportion of wardpatients may be the result of anunderappreciation of the potentialmortality of sepsis, of the significanceof elevated lactate levels, and of thetime course of the septic process, aswell as of failure to recognise crypticshock16 and the obvious andpracticalproblem of intensive care unit bedavailability. We did not assess howmany had end-of-life treatment limi-tations in place.

Managing large numbers of patientswith sepsis on the wards has beendescribed elsewhere.17,18 These pa-tients have not been well studied,although a number of studies haveidentified deficiencies in care.19-21

The significant increase in mortalityamong patients with uncomplicatedsepsis admitted to the ward causesconcerns that some of our wardpatients may have qualified forintensive therapy. An increase inmortality in less severe sepsis has alsobeen documented by other authors.22

The major challenge was managingthe prescribing of antibiotics. Despitegeneral acceptance of expert guide-lines for prescribing antibiotics, dif-ferences in their prescription andadministration were observed. Thisevidenceepractice gap is well recog-nised,23 and the empiric antibioticguideline was developed to promoteappropriate antibiotic prescribingpractices and optimal outcomes.24

The empiric guideline was consis-tent with the principles of antimi-crobial stewardship, and, while eachsite was allowed to modify it ac-cording to local opinion and antibi-otic resistance patterns, changeswereinfrequent. Particular anxieties wereexpressed about prescribing andadministering gentamicin. Theadministration component of theguideline was developed to promotethe most expeditious method ofadministration rather than favouringthe slow infusion that had becomenormal practice. Despite theemphasis on the first dose and timelyreview, antibiotics were oftencontinued long after they shouldhave been reviewed, following con-sideration of the results of pathologyinvestigations.

Other challenges beyond our controlincluded educating a high turn-over workforce in emergency de-partments, as well as medicalengagement, particularly in rural fa-cilities where governance is difficultand there is no doctor on site, orlocum medical staff are more com-mon. There were wide variations inthe methods of blood culture collec-tion, and a standardised guideline forblood cultures was subsequentlyadded to the Sepsis Toolkit.

MJA 204 (2) j 1 February 2016 1.e5

1.e6

Research

LimitationsThis work is subject to the limitationsof any quality improvement projectat multiple sites. The prolonged run-in period was not ideal. Ourapproach was not to measurecompliance with the care bundle, asundertaken elsewhere, but to usetime as a measure for promotingbehavioural change among emer-gency department clinicians. Assess-ing patient outcomes was the majordifficulty. The voluntary nature ofdata collection resulted in its incon-sistent submission, and the lack ofstrict diagnostic criteria for sepsisresulted in patients with conditionsfrom across the inflammatory con-ditionesepsis spectrum beingincluded in the SEPSIS KILLS data-base. Resource limitations alsomeantthat some sites implemented thepathways but did not submit data.

Reviewing the outcomes of patientswith an ICD-10-AMcode for sepsis inthe population-based administrativeAPDC is an accepted approach. This,however, entails the risk of reviewingthe outcomes of a different group ofpatients, a group for whom the final

MJA 204 (2) j 1 February 2016

diagnosis might not be directlyrelated to sepsis or its severity. This isa potential problemwhen comparingthe final diagnosis in the APDCdatabase with the provisional diag-nosis in the SEPSIS KILLS data.

Finally, the improved outcomesdescribed in our article may be theresult of the SEPSIS KILLS pro-gram, butmayalso be related to otherinitiatives for improving quality ofcare.

Implications for clinicians,researchers and policy makersThe observation that patients withsevere sepsis are being managed onthe wards highlights the need for ashift in the focus of both practiceimprovement and research fromintensive care to ward management.It also raises the problem of sepsisand the deteriorating patient. Weinformally estimated that around30%of patientswho required aRapidResponse call had sepsis, but thismay be an underestimate, with ratespossibly as high as 50%e60%.25

Finally, our work confirms the needfor continued research into risk

stratification tools for sepsis in theemergency department. In themeantime, all patients with lactatelevels of 4mmol/L or greater requireintensive care unit review andadmission.

The SEPSIS KILLS program pro-motes early recognition and man-agement of sepsis during thefirst fewhours in NSW emergency de-partments. By focusing on the prin-ciple of “Recognise, Resuscitate,Refer” it is possible to reduce the timebefore antibiotics are administeredand fluid resuscitation initiated. Thisprogram could be applied in otherjurisdictions and its integrationwith antimicrobial stewardship re-quirements should be considered.

Acknowledgements: We especially thank the SEPSISKILLS team members Lisa Coombs and Paul Hunstead.We also acknowledge the contributions to the programbyMaureen Edgtton-Winn, Ann Callaway and SarahPatterson. Jun Bai and the CEC data team developed thedatabase.

Competing interests: No relevant disclosures.

Received 5 Jun 2015, accepted 2 Oct 2015.n

ª 2016 AMPCo Pty Ltd. Produced with Elsevier B.V. All rightsreserved.

Research

1 Mayr FB, Yende S, Angus DC. a0e7-3d74cc9b733f/Policy-on-the- 18 Esteban A, Frutos-Vivar F, Ferguson ND,

Epidemiology of severe sepsis.Virulence 2014; 5: 4-11.

2 Wang HE, Shapiro NI, Griffin R, et al.Chronic medical conditions and risk ofsepsis. PLoS One 2012; 7: e48307.

3 Dellinger RP, Levy MM, Rhodes A, et al.Surviving sepsis campaign:international guidelines formanagement of severe sepsis andseptic shock: 2012. Crit Care Med 2013;41: 580-637.

4 Kaukonen K-M, Bailey M, Pilcher D,et al. Systemic inflammatory responsesyndrome criteria in defining severesepsis. N Engl J Med 2015; 372:1629-1638.

5 Gray A, Ward K, Lees F, et al. Theepidemiology of adults with severesepsis and septic shock in Scottishemergency departments. Emerg Med J2013; 30: 397-401.

6 Clinical Excellence Commission.Recognition and management ofsepsis (clinical focus report). Sydney:CEC, 2012. http://www.cec.health.nsw.gov.au/__data/assets/pdf_file/0004/259375/patient-safety-report-sepsis-2012.pdf (accessed Nov 2015).

7 Clinical Excellence Commission. 2011Quality Systems Assessment selfassessment. Supplementary report esepsis. Sydney: CEC, 2012. http://www.cec.health.nsw.gov.au/__data/assets/pdf_file/0009/259308/2011-qsa-thematic-report-sepsis_qsa_final.pdf(accessed Nov 2015).

8 Hughes C, Pain C, Braithwaite J,Hillman K. “Between the flags”:implementing a rapid response systemat scale. BMJ Qual Saf 2014; 23: 714-717.

9 Australasian College for EmergencyMedicine. Policy on the AustralasianTriage Scale. (Policy P06.). July 2013.https://www.acem.org.au/getattachment/693998d7-94be-4ca7-

Australasian-Triage-Scale.aspx(accessed Nov 2015),

10 Sundararajan V, Macisaac CM, PresneillJJ, et al. Epidemiology of sepsis inVictoria, Australia. Crit Care Med 2005;33: 71-80.

11 Kumar A, Roberts D, Wood KE, et al.Duration of hypotension beforeinitiation of effective antimicrobialtherapy is the critical determinant ofsurvival in human septic shock. CritCare Med 2006; 34: 1589-1596.

12 Puskarich MA, Trzeciak S, Shapiro NI,et al. Association between timing ofantibiotic administration and mortalityfrom septic shock in patients treatedwith a quantitative resuscitationprotocol. Crit Care Med 2011; 39:2066-2071.

13 Ferrer R, Martin-Loeches I, Phillips G,et al. Empiric antibiotic treatmentreduces mortality in severe sepsis andseptic shock from the first hour: resultsfrom a guideline-based performanceimprovement program. Crit Care Med2014; 42: 1749-1755.

14 Kaukonen KM, Bailey M, Suzuki S, et al.Mortality related to severe sepsis andseptic shock among critically illpatients in Australia and New Zealand,2000e2012. JAMA 2014; 311: 1308-1316.

15 Peake SL, Delaney A, Bailey M, et al.Goal-directed resuscitation for patientswith early septic shock. N Engl J Med2014; 371: 1496-1506.

16 Mikkelsen ME, Miltiades AN, Gaieski DF,et al. Serum lactate is associated withmortality in severe sepsis independentof organ failure and shock. Crit CareMed 2009; 37: 1670-1677.

17 Whittaker SA, Fuchs BD, Gaieski DF,et al. Epidemiology and outcomes inpatients with severe sepsis admitted tothe hospital wards. J Crit Care 2015; 30:78-84.

et al. Sepsis incidence and outcome:contrasting the intensive care unit withthe hospital ward. Crit Care Med 2007;35: 1284-1289.

19 Gao F, Melody T, Daniels DF, et al. Theimpact of compliance with 6-hour and24-hour sepsis bundles on hospitalmortality in patients with severe sepsis:a prospective observational study. CritCare 2005; 9: R764-R770.

20 Marwick CA, Guthrie B, Pringle JE, et al.A multifaceted intervention to improvesepsis management in generalhospital wards with evaluationusing segmented regression ofinterrupted time series. BMJ Qual Saf2014; 23: e2.

21 Jaderling G, Bell M, Martling CR, et al.ICU admittance by a rapidresponse team versus conventionaladmittance, characteristics, andoutcome. Crit Care Med 2013; 41:725-731.

22 Liu V, Escobar GJ, Greene JD, et al.Hospital deaths in patients with sepsisfrom 2 independent cohorts. JAMA2014; 312: 90-92.

23 Runciman WB, Hunt TD, Hannaford NA,et al. CareTrack: assessing theappropriateness of health care deliveryin Australia. Med J Aust 2012; 197:100-105. https://www.mja.com.au/journal/2012/197/2/caretrack-assessing-appropriateness-health-care-delivery-australia

24 Kumar A, Ellis P, Arabi Y, et al.Initiation of inappropriate antimicrobialtherapy results in a fivefold reductionof survival in human septic shock.Chest 2009; 136: 1237-1248.

25 Cross G, Bilgrami I, Eastwood G, et al.The epidemiology of sepsis duringrapid response team reviews in ateaching hospital. Anaesth IntensiveCare 2015; 43: 193-198.-

MJA 204 (2) j 1 February 2016 1.e7