C

Cf

ATa

b

c

d

e

f

U

a

ARRA

KCCCC

i

B

SA

V((

h0

Resuscitation 92 (2015) 107–114

Contents lists available at ScienceDirect

Resuscitation

j our na l ho me pa g e: www. elsev ier .com/ locate / resusc i ta t ion

linical paper

hest compression release velocity: Association with survival andavorable neurologic outcome after out-of-hospital cardiac arrest�,��

lexander Kovacsa, Tyler F. Vadeboncoeurb, Uwe Stolzc, Daniel W. Spaitec,aro Irisawad , Annemarie Silvere, Bentley J. Bobrowa,c,f,∗

University of Arizona College of Medicine-Phoenix, 550 E Van Buren St., Phoenix, AZ 85004, United StatesDepartment of Emergency Medicine, Mayo Clinic Florida, 4500 San Pablo Road, Jacksonville, FL 32224, United StatesDepartment of Emergency Medicine, University of Arizona, PO Box 245057, 1501 N. Campbell, Tucson, AZ 85724-5057, United StatesDepartment of Traumatology and Acute Critical Care, Osaka University Hospital, 2-15 Yamadaoka, Suita, Osaka 565-0871, JapanZoll Medical Corporation, 269 Mill Rd, Chelmsford, MA 01824, United StatesBureau of Emergency Medical Services and Trauma System, Arizona Department of Health Services, 150 N. 18th Avenue, #540, Phoenix, AZ 85007-3248,nited States

r t i c l e i n f o

rticle history:eceived 17 November 2014eceived in revised form 1 April 2015ccepted 23 April 2015

eywords:ardiac arrestPRPR qualityardiac resuscitation

a b s t r a c t

Purpose: We evaluated the association between chest compression release velocity (CCRV) and outcomesafter out-of-hospital cardiac arrest (OHCA).Materials and methods: CPR quality was measured using a defibrillator with accelerometer-based tech-nology (E Series, ZOLL Medical) during OHCA resuscitations by 2 EMS agencies in Arizona between10/2008 and 06/2013. All non-EMS-witnessed adult (≥18 years) arrests of presumed cardiac etiologywere included. The association between mean CCRV (assessed as an appropriate measure of central ten-dency) and both survival to hospital discharge and neurologic outcome (Cerebral Performance Categoryscore = 1 or 2) was analyzed using multivariable logistic regression to control for known and potentialconfounders and multiple imputation to account for missing data.Results: 981 OHCAs (median age 68 years, 65% male, 11% survival to discharge) were analyzedwith 232 (24%) missing CPR quality data. All-rhythms survival varied significantly with CCRV [fast(≥400 mm/s) = 18/79 (23%); moderate (300–399.9 mm/s) = 50/416 (12%); slow (<300 mm/s) 17/255(7%); p < 0.001], as did favorable neurologic outcome [fast = 14/79 (18%); moderate = 43/415 (10%);slow = 11/255 (4%); p < 0.001]. Fast CCRV was associated with increased survival compared to slow[adjusted odds ratio (aOR) 4.17 (95% CI: 1.61, 10.82) and moderate CCRV [aOR 3.08 (1.39, 6.83)]. Fast

CCRV was also associated with improved favorable neurologic outcome compared to slow [4.51 (1.57,12.98)]. There was a 5.2% increase in the adjusted odds of survival for each 10 mm/s increase in CCRV[aOR 1.052 (1.001, 1.105)].Conclusion: CCRV was independently associated with improved survival and favorable neurologic out-come at hospital discharge after adult OHCA.

© 2015 Published by Elsevier Ireland Ltd.

� A Spanish translated version of the abstract of this article appears as Appendixn the final online version at http://dx.doi.org/10.1016/j.resuscitation.2015.04.026.�� Presented at the European Resuscitation Council Resuscitation Symposium inilbao, Spain, May, 2014.∗ Corresponding author at: Bureau of Emergency Medical Services and Traumaystem, Arizona Department of Health Services, 150 N. 18th Avenue, #540, Phoenix,Z 85007-3248, United States.

E-mail addresses: Akovacs@email.arizona.edu (A. Kovacs),adeboncoeur.tyler@mayo.edu (T.F. Vadeboncoeur), Ustolz@aemrc.arizona.edu

U. Stolz), Dan@aemrc.arizona.edu (D.W. Spaite), Taroirisawa@yahoo.co.jpT. Irisawa), Asilver@zoll.com (A. Silver), Bentley.Bobrow@azdhs.gov (B.J. Bobrow).

ttp://dx.doi.org/10.1016/j.resuscitation.2015.04.026300-9572/© 2015 Published by Elsevier Ireland Ltd.

1. Introduction

Annually, EMS treats over 350,000 out-of-hospital cardiacarrests (OHCA) in the U.S.1 Outcomes vary widely between com-munities with survival rates for ventricular fibrillation (VF) rangingfrom 3.3% to 45%.2–4 The quality of CPR delivered, defined by chestcompression (CC) rate, depth, fraction, and “recoil,” is thought toimpact outcomes and is believed to be one factor contributing to

the large disparities in survival.4–8 Both the European ResuscitationCouncil (ERC) and the American Heart Association (AHA) recom-mend complete chest wall recoil as a component of high-qualityCPR.9,10

1 citation 92 (2015) 107–114

apTlpflfdp

mad

2

2

zaTbtcaeESaaa

2

sdpncrSt

ctoecewotibr

2

dppw

08 A. Kovacs et al. / Resus

While optimal CC rate, depth and fraction have been associ-ted with improved outcomes from cardiac arrest,5–7,11,12 there is aaucity of data describing the release phase of chest compressions.he AHA and ERC recommendations of complete chest recoil/noeaning are based on animal studies which have implicated com-lete chest wall recoil as an important factor in maximizing bloodow during CPR13–15 and clinical data demonstrating that providers

requently do not allow complete recoil.6,16–18 There are no dataemonstrating an association between optimization of the releasehase of chest compressions and clinical outcomes.

The purpose of this analysis was to assess whether CCRV, theaximal velocity of chest compression release in the posterior to

nterior direction, is associated with improved survival to hospitalischarge and a favorable neurologic outcome after adult OHCA.

. Methods

.1. Study setting

Data for this study were collected from 2 EMS agencies in Ari-ona. The Mesa Fire and Medical Department is a suburban-basedgency that responds to 70,000 calls per year (population 439,000).here are 373 emergency medical technicians (EMTs) employedy the department, which deploys two EMT Basics (EMT-B) andwo EMT Paramedics (EMT-P) as a typical crew. Guardian Medi-al Transport (GMT) responds to a suburban and rural area withpproximately 14,000 annual 9-1-1 calls (population 80,000). GMTmploys 80 EMTs and dispatches a crew of at least one EMT-P andMT-B to emergency calls. Both EMS agencies participate in theave Hearts in Arizona Registry and Education (SHARE) Program,

statewide cardiac resuscitation quality improvement program,nd use minimally interrupted cardiac resuscitation (MICR) as theirdult resuscitation protocol.19,20

.2. Study design and population

We describe a prospective, before-after, observational cohorttudy of consecutive adult (aged ≥18 years) OHCA of presumed car-iac etiology between 10/2008 and 6/2013. Cases were excluded ifrehospital resuscitation was not initiated, the patient had a do-ot-resuscitate order, the arrest was witnessed by EMS, or if theause of the arrest was presumed to be non-cardiac (known respi-atory arrest, suicide, trauma, drowning, or drug overdose). TheHARE Program also tracks hospital data and if it becomes clear thathe cause of the arrest was non-cardiac, the database is updated.

Baseline CPR quality and outcome data (10/2/08–2/27/10) wereollected during Phase 1 (P1) before a dedicated educational ini-iative (described in detail previously)5,21 and without the usef real-time audiovisual feedback (RTAVF) for CC quality. Theducational intervention included 2 h of didactics and 2 h of psy-homotor practice, termed “Scenario-Based Training” (SBT), whichmphasizes a team approach to resuscitation and close complianceith the parameters of high-quality CPR.9,22 Phase 2 (P2) began

n 5/19/10 after the completion of training and the enabling ofhe monitor-defibrillator’s RTAVF software (E-series, ZOLL Med-cal), with data collection ending 6/27/13. This technology haseen described in detail previously and does not include feedbackegarding CCRV.21

.3. Data collection

CCRV was assessed via the accelerometer present in the

efibrillators.21 Recorded CC quality data included overall com-ression fraction, pre-shock pause, and minute-by-minute com-ression rate, depth, and release velocity data. CPR quality dataere matched with patient first care reports which were collected

Fig. 1. Study population inclusion and exclusion flowchart.

as part of the SHARE Program. SHARE employs an Utstein-styledatabase linked to hospital outcomes and has been previouslydescribed.19

As an Arizona Department of Health Services (ADHS)-sponsoredpublic health initiative, Arizona’s Attorney General has determinedthat the SHARE Program is exempt from the requirements of theHealth Insurance Portability and Accountability Act (HIPAA) allow-ing linkage of EMS and hospital data, tracking of OHCA events,and evaluation of efforts to improve resuscitation care. By virtue ofbeing a public health initiative, the ADHS Human Subjects ReviewBoard and the University of Arizona Institutional Review Boardhave determined that neither the interventions nor their evalu-ation constitute Human Subjects Research and have approved thepublication of de-identified data. The project is registered at clinicalTrials.gov#NCT01999036.

2.4. Statistical analysis

The analytical approach is similar to Vadeboncoeur et al.5

The primary outcomes for this study were survival to hospi-tal discharge and survival with favorable neurologic outcome[Cerebral Performance Category (CPC) Score = 1 or 2]. CCRV wasassessed as a categorical variable [fast (≥400 mm/s), moderate(300–399.9 mm/s), or slow (<300 mm/s)] as well as a continu-ous variable (mm/s).23 Crude proportions were compared using

Fisher’s exact test. We analyzed the distribution of release veloci-ties for chest compressions of individual OHCA cases to determinethe appropriateness of using mean CCRV as a measure of CPRquality. Multivariable logistic regression was used to assess the

A. Kovacs et al. / Resuscitation 92 (2015) 107–114 109

Table 1Characteristics of Study Population.

Characteristic Unimputed Data (complete cases only) Imputed Dataset(complete + imputed cases)

n/N (missing) Percent (95% CI)* Percent (95% CI)*

Age, years-median (95% CI)** 981 (0) 68 (66.6, 69.4) 68 (66.6, 69.4)Male Sex 638/980 (1) 65.1 (62.0, 68.1) 65.1 (62.2, 68.1)Witnessed Arrest 401/973 (8) 41.2 (38.1, 44.4) 41.2 (38.1, 44.3)Initial Shockable Rhythm (V-fib/V-tach) upon EMS arrival 251/980 (1) 25.6 (22.9, 28.5) 25.7 (23.0, 28.5)Initial Shockable Rhythms Defibrillated by EMS 239/251 (0) 95.2 (91.8, 97.5) ***

Location of Arrest 980 (1)Residential 476/980 48.5 (45.3, 51.7) 48.6 (45.4, 51.7)Medical Facility 109/980 11.1 (9.2, 13.2) 11.1 (9.1, 13.1)Public 395/980 40.3 (37.2, 43.5) 40.3 (37.3, 43.4)

Provision of Bystander CPR 464/973 (8) 47.7 (44.5, 50.9) 47.8 (44.6, 50.9)High Quality MICR 808/964 (17) 83.8 (81.3, 86.1) 83.9 (81.6, 86.2)Prehospital ROSC 249/980 (1) 25.4 (22.7, 28.3) 25.5 (22.7, 28.2)Prehospital Intubation 408/981 (0) 41.6 (38.5, 44.7) ***

EMS Dispatch to Arrival Interval, minutes – median (95% CI)** 944 (37) 5 (5, 6) 5 (5, 6)Use of Therapeutic Hypothermia 61/981 (0) 6.2 (4.9, 8.0) ***

Post CPR Quality Improvement Intervention Period (vs. Pre) 700/981 (0) 71.4 (68.4, 74.2) ***

Survival to Hospital Discharge 110/981 (0) 11.2 (9.3, 13.4) ***

Favorable Neurologic Outcome (CPC Score = 1 or 2) 88/978 (3) 9.0 (7.3, 11.0) ***

Witnessed Arrests & Shockable RhythmsSurvival to Hospital Discharge 65/160 (0) 40.6 (32.9, 48.7) ***

Favorable Neurological Outcome 53/157 (3) 33.8 (26.4, 41.7) ***

Mean CC Release Velocity, mm/s – mean (95% CI) 750 (232) 324.7 (320.1, 329.3) 321.3 (316.7, 325.8)Mean CC Release Velocity

<300 mm/s 255/750 34.0 (30.6, 37.4) 36.2 (32.9, 39.6)300–399.9 mm/s 416/750 55.5 (51.9, 59.0) 53.3 (49.8, 56.8)≥400 mm/s 79/750 10.5 (8.3, 12.7) 10.5 (8.4, 12.6)

Mean CC Rate, CCs/minute-median (95% CI)** 750 (232) 102.7 (102.1, 103.3) 104.0 (103.1, 104.8)Mean CC Fraction, percent-median (95% CI)** 750 (232) 83.3 (82.4, 84.2) 82.2 (81.3, 83.2)Mean CC Depth, mm – mean (95% CI) 750 (232) 52.7 (51.9, 53.5) 51.9 (51.2, 52.7)Percent of CCs 51 mm or deeper-median (95% CI)** 750 (232) 62.5 (58.7, 66.3) 57.1 (53.7, 60.6)

CI, confidence Interval; CCRV, Chest Compression Release Velocity; CCs, Chest Compressions; CPC, Cerebral Performance Category; CPR, Cardiopulmonary Resuscitation;EMS, Emergency Medical Services; MICR, Minimally Interrupted Cardiac Resuscitation; mm/s, millimeters per second; ROSC, Return of Spontaneous Circulation; V-fib/V-tach,Ventricular fibrillation/Ventricular tachycardia.

data n

arpOCapqpiwioCbai

vatwAa

iwvt

* Percent (95% CI) unless otherwise noted.** Median and 95% CI calculated using quantile (median) regression.

*** Same as unimputed data because either no missing data or analysis of imputed

ssociation between CCRV and outcomes, calculating adjusted oddsatios (aORs) for each outcome. The analyses included known inde-endent risk factors for survival and neurologic outcome followingHCA and potential confounders associated with the outcomes andCRV [chest compression (CC) rate, CC fraction, age, sex, witnessedrrest, initial rhythm upon EMS arrival, location of OHCA, EMS dis-atch to arrival interval, bystander CPR, MICR by EMS (high vs. lowuality)],5,22 provision of therapeutic hypothermia during prehos-ital or hospital care, and pre vs. post CPR quality improvement

ntervention.21 To account for potential secular trends in outcomese also included year of arrest in our final models. OHCAs with an

nitial shockable rhythm upon EMS arrival were identified a pri-ri as a clinically important subgroup and the association betweenCRV and outcomes was assessed separately. An interaction termetween CCRV and shockable rhythm is presented as a sensitivitynalysis to determine if the impact of CCRV on outcomes varied bynitial rhythm.

We used mixed model logistic regression (“xtmelogit”, Stataersion 12.1; StataCorp, College Station, TX) with both EMS agencynd hospital considered as random effects. Hospitals with fewerhan 10 cases were combined into one cluster and all cases thatere not transported to a hospital were combined into one cluster.

likelihood ratio test for the inclusion of hospital or EMS agency as random effect was considered significant with a p-value of ≤0.05.

Fractional polynomial regression was used to assess the linear-

ty of all continuous variables in the logit scale. Multiple imputation

as used to account for missing data.24 All non-CPR quality metricsariables in Table 1, along with year of arrest, EMS agency, andreating hospital, were used to impute missing CPR quality data

ot possible because imputed datasets varied for subsets of imputed variables.

using linear regression, with compression fraction truncated to0 or 100. Missing CPR quality data followed a monotonic miss-ing pattern (if one CPR quality metric was missing for a case, allwere missing) and were imputed separately. We used multiplechained equations using all variables from the primary analysisalong with EMS agency, hospital, and year of arrest to impute miss-ing values for all remaining confounders and prognostic variables.Missing binary variables were imputed using logistic regressionand dispatch to arrival interval was imputed using negative bino-mial regression.25,26 Twenty imputed data sets were generated.Model fit was assessed using the Hosmer–Lemeshow (H-L) good-ness of fit test. Model discrimination was assessed using the areaunder the receiver operating characteristics curve (AUC) and diag-nostics were evaluated using the Pearson, deviance, and Anscomberesiduals as described previously.21 All analyses were conductedusing Stata 12.1 (StataCorp, College Station, TX).

3. Results

There were a total of 1493 consecutive OHCA treated by EMSduring the study period and following exclusions there were 981adult OHCAs with presumed cardiac etiology not witnessed byEMS (Fig. 1). A total of 272 cases had at least one missing studyvariable; however, CPR quality metrics data were missing in 232cases (23.6%), comprising the vast majority of missing data. Table 1

shows the study population characteristics for both complete cases(excluding missing data) and the entire population (missing valuesimputed using multiple imputation). In general, characteristics didnot differ greatly between complete cases and the imputed data

110 A. Kovacs et al. / Resuscitation 92 (2015) 107–114

F ion in

3 esent

sTotrOoifC≥

wmsss

ap(n(iawma(ca

ig. 2. Distribution of individual maximum recoil velocities for each chest compress00–399.9, 400+ mm/s). Solid lines represent the mean CCRV and dashed lines repr

et. The median age was 68 and the proportion of males was 65%.he median EMS dispatch to arrival interval was 5 min. A totalf 41% of OHCA were witnessed by a bystander, 26% had an ini-ial shockable rhythm upon EMS arrival, and 25.4% of OHCAs hadeturn of spontaneous circulation before arrival at the hospital.verall survival to discharge was 11.2% and favorable neurologicutcome upon discharge was 9.0%. For witnessed OHCA with annitial shockable rhythm upon EMS arrival, survival was 40.6% andavorable neurologic outcome was 33.8%. The average of all meanCRVs was 324.7 mm/s and 10.5% of OHCA cases had a mean CCRV400 mm/s.

Individual CCRVs of each chest compression for each patientere approximately normally distributed with similar mean andedian CCRV. Fig. 2 shows the distribution of CCRVs for 9 randomly

elected OHCA cases (3 cases from each of the CCRV categoriestudied; >400, 300–399.9, <300 mm/s) along with means, medians,tandard deviations.

Tables 2 and 3 show the logistic regression results for crudend adjusted analyses as well as adjusted analyses using multi-le imputation to account for missing data. Clustering by hospitaln = 9) was statistically significant for both survival and favorableeurologic outcome (p < 0.001), while clustering by EMS agencyn = 2) was not (p = 0.5 and 1.0, respectively). All models thus onlyncluded hospital as a random effect. Survival to hospital dischargend favorable neurologic outcome were significantly associatedith CCRV in crude analyses [Survival: fast CCRV = 18/79 (22.8%),oderate = 50/416 (12.0%), slow = 17/255 (6.7%), p < 0.001; Favor-

ble Neurologic Outcome: fast = 14/79 (17.7%), moderate = 43/41510.4%), slow = 11/255 (4.3%); p < 0.001]. While aORs from theomplete-case (unimputed) analyses did not differ greatly fromORs using the multiple imputation data, estimates did appear

nine randomly selected OHCAs stratified by overall mean CCRV categories (0–299.9, the median CCRV for each distribution.

more conservative for the imputation analysis and we highlightestimates from the multiple imputation analyses. Both survival andfavorable neurologic outcome were significantly higher for caseswith fast compared to slow CCRV [aOR 4.17 (95% CI: 1.61, 10.82)and 4.51 (1.57, 12.98), respectively]. Only survival was signifi-cantly higher for fast compared to moderate CCRV [aOR 3.08 (1.39,6.83)]. When assessed as a continuous variable, CCRV remainedan independent predictor of survival, with adjusted odds of sur-vival increasing 5.2% for each 10 mm/s increase in CCRV [aOR 1.052(1.001, 1.105)]. Fig. 3 shows the adjusted probability of survival(predicted marginal probabilities) for OHCAs across the range ofCCRVs from the final logistic regression model for survival, con-trolling for all other variables in the model. The probability ofsurvival increases steadily along with CCRV. Fig. 3 also shows thedistribution of predicted survivorship for individual OHCA casesbased on the values from the final multivariable logistic regres-sion model. CCRV as a continuous variable was not significantlyassociated with favorable neurologic outcome [aOR 1.047 (0.994,1.104)].

Final logistic regression models for survival and favorable neu-rologic outcome had adequate fit (Tables 2 and 3) based on theH-L goodness of fit test (p > 0.2) and had excellent discrimination(AUC >0.9). Diagnostic analyses showed no covariate patterns wereoverly influential on final model results and exclusion of the topfive covariate patterns with the most extreme values for diagnosticstatistics did not change regression coefficients for CCRV in eithermodel by >10%. In addition, no fractional polynomial transfor-

mation for any continuous variable (including CCRV) significantlyimproved model fit for either outcome. There was no significantinteraction between mean CCRV and shockable rhythms (p = 0.16for survival, p = 0.08 for neurologic outcome).

A. Kovacs et al. / Resuscitation 92 (2015) 107–114 111

Table 2Logistic regression analyses for association between chest compression release velocity (CCRV) and survival to hospital discharge.

Survival to Hospital Discharge Non-Imputed (N = 711) Imputed (N = 981)

n/N (%) Crude OR (95% CI) Adjusted OR*** (95% CI) Adjusted OR (95% CI)

Mean CCRV*

<300 mm/s 17/255 (6.7) 1 [Ref.] 1 [Ref.] 1 [Ref.]300–399.9 mm/s 50/416 (12.0) 1.753 (0.969, 3.170) 1.223 (0.574, 2.605) 1.355 (0.646, 2.842)≥400 mm/s 18/79 (22.8) 3.952 (1.844, 8.473) 5.193 (1.949, 13.838) 4.173 (1.609, 10.821)

Mean CCRV (per 10 mm/s)* – 1.058 (1.018, 1.100) 1.056 (1.005, 1.111) 1.052 (1.001, 1.105)CC Rate (per 10 CCs/minute)** – 1.089 (0.930, 1.275) 1.037 (0.784, 1.03) 0.967 (0.738, 1.269)CC Fraction (per 10% increment)** – 0.984 (0.827, 1.171) 0.909 (0.681, 1.214) 0.943 (0.710, 1.252)Age (per year)** – 0.965 (0.952, 0.978) 0.964 (0.944, 0.984) 0.968 (0.951, 0.984)Witnessed Arrest**

No 27/572 (4.7) 1 [Ref.] 1 [Ref.] 1 [Ref.]Yes 83/401 (20.7) 3.797 (2.362, 6.106) 3.789 (1.900, 7.555) 3.793 (2.113, 6.810)

Initial Rhythm Upon EMS Arrival**

Non-shockable 30/729 (4.1) 1 [Ref.] 1 [Ref.] 1 [Ref.]Shockable (V-fib/V-tach) 80/251 (31.9) 7.704 (4.810, 12.338) 8.020 (4.125, 15.593) 6.329 (3.624, 11.055)

Sex**

Male 70/638 (11.0) 1 [Ref.] 1 [Ref.] 1 [Ref.]Female 39/342 (11.4) 1.156 (0.747, 1.788) 2.022 (1.053, 3.881) 2.095 (1.204, 3.645)

Provision of Therapeutic Hypothermia**

No 76/920 (8.3) 1 [Ref.] 1 [Ref.] 1 [Ref.]Yes 34/61 (55.7) 9.141 (5.109, 16.355) 6.735 (2.856, 15.884) 8.989 (4.261, 18.962)

EMS dispatch to on scene arrival (per minute)** – 0.974 (0.904, 1.049) 0.924 (0.803, 1.064) 0.933 (0.842, 1.033)Public Location of OHCA

No 50/585 (8.6) 1 [Ref.] 1 [Ref.] 1 [Ref.]Yes 60/395 (15.2) 1.709 (1.122, 2.605) 2.117 (1.047, 4.279) 1.933 (1.060, 3.526)

Provision of Bystander CPR**

No 50/509 (9.8) 1 [Ref.] 1 [Ref.] 1 [Ref.]Yes 59/464 (12.7) 1.252 (0.823, 1.904) 1.519 (0.821, 2.809) 1.125 (0.667, 1.897)

Use of high quality MICR**

No 17/156 (10.9) 1 [Ref.] 1 [Ref.] 1 [Ref.]Yes 92/808 (11.4) 1.444(0.818, 2.549) 0.743 (0.301, 1.838) 0.963 (0.460, 2.018)

Post vs. Pre CPR quality improvement intervention**

Pre 26/281 (9.3) 1 [Ref.] 1 [Ref.] 1 [Ref.]Post 84/700 (12.0) 1.256 (0.774, 2.039) 5.552 (1.488, 20.72) 2.931 (0.955, 8.996)

Year (per year) – 0.971 (0.838, 1.126) 0.814 (0.560, 1.167) 0.887 (0.646, 1.219)

CI, Confidence Interval; CCRV, Chest Compression Release Velocity; CCs, Chest Compressions; CPR, Cardiopulmonary Resuscitation; EMS, Emergency Medical Services;MICR, Minimally Interrupted Cardiac Resuscitation; mm/s, millimeters per second; OR, Odds Ratio; ROSC, Return of Spontaneous Circulation; V-fib/V-tach, Ventricularfibrillation/Ventricular tachycardia.

* CCRV as a continuous variable (per 10 mm/s) and categorical variable (<300, 300–399.9, ≥400 mm/s) included separately in final models adjusted for all other variableslisted with **.

** Adjusted ORs for non-CCRV variables from final model with CCRV as continuous varia*** Hosmer–Lemeshow goodness-of-fit p-value = 0.626; area under the receiver operatin

Fig. 3. Probability of survival (marginal predictions from final logistic regressionmodel) for values of chest compression release velocity (black solid line) alongwith 95% confidence bands (black dashed lines) for out-of-hospital cardiac arrests,controlling for all other risk factors and confounders (age, sex, shockable rhythm,witnessed arrest, location of OHCA, EMS dispatch to scene arrival interval, use of highquality MICR by EMS, pre vs. post quality improvement intervention, provision oftherapeutic hypothermia). Hollow circles (survivors) and X’s (deaths) represent theptm

redicted probability of survival for individual OHCA cases based on the characteris-

ics of individual cases for all relevant risk factors and confounders above, including

ean chest compression release velocity, from the final logistic regression models.

ble.g characteristics curve = 0.914.

For OHCAs with an initial shockable rhythm upon EMSarrival, both outcomes were also significantly associated withCCRV in unadjusted analyses [Survival: fast CCRV = 14/23 (60.9%),moderate = 38/125 (30.4%), slow = 13/52 (25.0%), p = 0.009; Favor-able Neurologic Outcome: fast = 12/23 (52.2%), moderate = 33/124(26.6%), slow = 9/52 (17.3%); p = 0.01]. For OHCAs with an initialshockable rhythm, fast CCRV was also independently associatedwith both increased survival and favorable neurologic outcomecompared to slow CCRV [aOR 5.03 (95% CI: 1.36, 18.57) and 6.04(1.58, 23.15), respectively] and moderate CCRV [aOR 4.64 (1.52,14.20) and 4.03 (1.27, 12.82), respectively], after controlling forall previous confounders (all variables in Tables 2 and 3 exceptfor initial rhythm), adjusting for clustering by hospital, and usingmultiple imputation to account for missing data. CCRV as a con-tinuous variable was significantly associated with both survival[aOR per 10 mm/s = 1.075 (1.001, 1.154)] and favorable neurologicoutcome [aOR per 10 mm/s = 1.079 (1.002, 1,161)] for OHCAs withan initial shockable rhythm. The final models for the subset ofOHCAs with an initial shockable rhythm for both survival andfavorable neurologic outcome had adequate fit (H-L goodness offit p-values >0.3) and discrimination (AUC >0.8). Model diagnos-tics did not identify any potentially miscoded or overly influential

cases.

112 A. Kovacs et al. / Resuscitation 92 (2015) 107–114

Table 3Logistic regression analyses for association between chest compression release velocity (CCRV) and favorable neurologic outcome at hospital discharge.

Favorable Neurologic Outcome (CPC Score = 1 or 2) Non-Imputed Data (N = 710) Imputed Data (N = 978)

n/N (%) Crude OR (95% CI) Adjusted OR*** (95% CI) Adjusted OR (95% CI)

Mean CCRV*

<300 mm/s 11/255 (4.3) 1 [Ref.] 1 [Ref.] 1 [Ref.]300–399.9 mm/s 43/415 (10.4) 2.382 (1.188, 4.778) 1.972 (0.807, 4.822) 1.900 (0.814, 4.433)≥400 mm/s 14/79 (17.7) 4.565 (1.915, 10.886) 5.774 (1.907, 17.477) 4.512 (1.569, 12.976)

Mean CCRV (per 10 mm/s)* – 1.056 (1.012, 1.101) 1.052 (0.996, 1.110) 1.047 (0.994, 1.104)CC Rate (per 10 CC/minute)** – 1.104 (0.932, 1.309) 1.006 (0.750, 1.348) 0.977 (0.738, 1.293)CC Fraction (per 10% increment)** – 0.898 (0.750, 1.075) 0.756 (0.561, 1.019) 0.772 (0.574, 1.039)Age (per year)** – 0.961 (0.947, 0.976) 0.959 (0.936, 0.982) 0.962 (0.944, 0.981)Witnessed Arrest**

No 22/572 (3.9) 1 [Ref.] 1 [Ref.] 1 [Ref.]Yes 66/398 (16.6) 3.535 (2.102, 5.944) 3.012 (1.442, 6.292) 3.511 (1.849, 6.665)

Initial Rhythm Upon EMS Arrival**

Non-shockable 22/729 (3.0) 1 [Ref.] 1 [Ref.] 1 [Ref.]Shockable (V-fib/V-tach) 66/248 (26.6) 8.535 (5.023, 14.501) 9.448 (4.469, 19.974) 6.507 (3.515, 12.046)

Sex**

Male 60/637 (9.4) 1 [Ref.] 1 [Ref.] 1 [Ref.]Female 27/340 (7.9) 0.901 (0.551, 1.473) 1.274 (0.619, 2.621) 1.496 (0.810, 2.762)

Provision of Therapeutic Hypothermia**

No 60/917 (6.5) 1 [Ref.] 1 [Ref.] 1 [Ref.]Yes 28/61 (45.9) 8.159 (4.501, 14.789) 8.065 (3.237, 20.094) 9.563 (4.339, 21.078)

EMS dispatch to on-scene arrival (per minute)** – 0.953 (0.871, 1.044) 0.949 (0.825, 1.093) 0.916 (0.817, 1.028)Public Location of OHCA**

No 35/583 (6.0) 1 [Ref.] 1 [Ref.] 1 [Ref.]Yes 53/394 (13.5) 2.221 (1.390, 3.550) 2.298 (1.063, 4.969) 2.250 (1.163, 4.352)

Provision of Bystander CPR**

No 41/509 (8.1) 1 [Ref.] 1 [Ref.] 1 [Ref.]Yes 46/461 (10.0) 1.167 (0.737, 1.846) 1.505 (0.771, 2.938) 1.092 (0.616, 1.936)

Use of high quality MICR**

No 16/155 (10.3) 1 [Ref.] 1 [Ref.] 1 [Ref.]Yes 72/806 (8.9) 1.120 (0.621, 2.023) 0.495 (0.192, 1.276) 0.710 (0.325, 1.551)

Post vs. Pre Intervention**

Pre 19/280 (6.8) 1 [Ref.] 1 [Ref.] 1 [Ref.]Post 69/699 (9.9) 1.351(0.821, 2.222) 5.919 (1.424, 24.598) 3.384 (1.001, 11.351)

Year (per year)** – 0.995 (0.857, 1.155) 0.973 (0.661, 1.434) 1.048 (0.742, 1.479)

CI, Confidence Interval; CCRV, Chest Compression Release Velocity; CCs, Chest Compressions; CPR, Cardiopulmonary Resuscitation; CPC, Cerebral Performance Category; EMS,Emergency Medical Services; MICR, Minimally Interrupted Cardiac Resuscitation; mm/s, millimeters per second; OR, Odds Ratio; ROSC, Return of Spontaneous Circulation;V-fib/V-tach, Ventricular fibrillation/Ventricular tachycardia.

* CCRV as a continuous variable (per 10 mm/s) and categorical variable (<300, 300–399.9, ≥400 mm/s) included separately in final models adjusted for all other variableslisted with **.

iable.ing ch

4

paotctillt

TacccTasmac

** Adjusted ORs for non-CCRV variables from model with CCRV as continuous var*** Hosmer–Lemeshow goodness-of-fit p-value = 0.294; area under receiver operat

. Discussion

High-quality CPR is an important determinant of outcomes inatients who suffer cardiac arrest, yet the quality of prehospitalnd in-hospital CPR is highly variable.6,16–18,27–31 Compressionsf insufficient depth and rate, frequent and prolonged interrup-ions to chest compressions, and incomplete chest recoil areommon.6,12,17,32 Encouragingly, CPR quality is a modifiable fac-or with previous studies demonstrating that CPR quality can bemproved.5,21 Differences in CPR quality likely play a role in thearge reported variations in outcomes for OHCAs.6,21,33–35 Previousiterature demonstrates that CC depth, rate, and fraction as well ashe preshock pause impact survival from OHCA.5–7,11,12,21,27,33,36

The release phase of chest compressions is poorly described.wo human studies have implicated incomplete chest wall recoils negatively impacting hemodynamics, but were not performed onardiac arrest patients.37,38 Observational studies describe a lack ofomplete chest wall recoil, also referred to as leaning, during resus-itation from both OHCA and in-hospital cardiac arrest.6,17,29,30

o our knowledge, there are no clinical data demonstrating anssociation between complete chest recoil and outcomes. A recent

tudy by Cheskes et al. assessed the association between CCRV, theaximal velocity of chest compression release in the posterior to

nterior direction, and clinical outcomes.23 They found an asso-iation between CCRV and survival to hospital discharge in their

aracteristics curve = 0.911.

univariate and categorical analyses, however there was no inde-pendent statistical association in their multivariable model.

Importantly, CCRV and complete chest recoil are not synony-mous. CCRV measures the velocity of decompression rather thanthe completeness of decompression. Theoretically, achieving chestrecoil faster would augment the negative intrathoracic suction andthus improve venous return to the heart.39 We believe that ourstudy is the first to show an independent association between anymeasure of chest wall recoil and clinical outcomes. In the cur-rent analysis, fast CCRV (≥400 ms) was independently associatedwith higher survival to hospital discharge and favorable neuro-logic outcome. The adjusted odds of survival increased more thanfour-fold by increasing the CCRV from slow (<300 mm/s) to fast(≥400 mm/s) and more than tripled by increasing CCRV from mod-erate (300–399.9 mm/s) to fast (≥400 mm/s). Results for favorableneurologic outcome were similar; however, the only significantimprovement was from slow to fast. Importantly, there was a 5.2%increase in the odds of survival for every 10 mm/s increase in CCRV,suggesting that faster CCRV may generate better blood flow com-pared with slower CCRV, across a wide range of CCRV. The wideconfidence bands when CCRV is above 400 mm/s (Fig. 3) limit con-

clusions about the peak or optimal CCRV.

The available animal data supports our findings. Incom-plete chest recoil has been shown to have an impact on thehemodynamics of cardiac arrest in porcine models.13–15 Niles

citatio

epdevp

tfBiprgjitti

4

bdtintttottgsedtfcotec

5

rifat

C

t

zc

1

1

1

1

1

1

1

1

1

1

2

2

2

2pression release velocity and outcomes from out-of-hospital cardiac arrest.Resuscitation 2015;86:38–43.

A. Kovacs et al. / Resus

t al. found that chest wall lean led to decreased cardiac out-ut and coronary perfusion pressures.15 Other porcine modelsemonstrate that incomplete chest wall recoil leads to increasedndotracheal pressure, decreased mean arterial pressure, impededenous return, and decreased coronary and cerebral perfusionressures.13,14

The results of our study, taken in context with the prior litera-ure, suggest that faster CCRV has a positive impact on outcomesrom OHCA. While CCRV is impacted by chest wall physiology,obrow et al. demonstrated that it can be improved through ded-

cated efforts.21 Even if complete chest recoil is achieved, therovider could still lean while releasing. This leaning during theelease slows the CCRV and theoretically has adverse effects on theeneration of negative pressure and subsequent venous return. Asust one CC metric, CCRV likely does not work in isolation, making itmportant to identify the role that CCRV plays in conjunction withhe other compression variables, rate and depth. It is also impor-ant to determine the impact of the variation of these metrics withinndividuals.

.1. Limitations

While this study demonstrated an independent associationetween fast CCRV and improved outcomes, its observationalesign does not allow a specific determination of causality. Addi-ional larger prospective studies are needed to determine thedeal CCRV, the impact of CCRV on outcomes in patients withon-shockable rhythms, and to assess the impact of dedicatedraining and improved RTAVF on CCRV. The EMS agencies inhis study all use the MICR protocol (as opposed to ACLS) andhe differences in protocol could affect the influence of CCRV onutcomes compared to systems using standard ACLS. The facthat these EMS systems are part of a large statewide resuscita-ion quality improvement program could make the results lesseneralizable. Additionally, we only used one type of CPR mea-urement and feedback device. Finally, as is common in EMSvaluations, there was a significant amount of missing CPR qualityata. In order to mitigate the potential impact, we used mul-iple imputation to reduce the chance of bias that can occurrom omitting cases with missing data. Our statistical modelsontained more covariates than the general rule of thumb ofne covariate for each ten outcomes; however, this “rule ofen” has been shown to be conservative and can be relaxed,specially for risk factor modeling to adequately control foronfounding.40

. Conclusion

In this study, chest compression release velocity during EMSesuscitation was independently associated with a significantmprovement in both survival and favorable neurologic outcomeollowing adult OHCA. Further study is required to confirm the link-ge between CCRV and improved outcomes and to determine ifhere is a threshold or optimal chest compression release velocity.

onflicts of interest statement

Annemarie Silver, PhD, is an employee of Zoll Medical Corpora-

ion.

Drs. Bobrow and Spaite disclose that the University of Ari-ona receives support from the Medtronic Foundation involvingommunity-based translation of resuscitation science.

2

2

n 92 (2015) 107–114 113

Acknowledgments

The authors thank Guardian Medical Transport and the MesaFire and Medical Department for participating in the SHARE Pro-gram and for their efforts to improve survival from OHCA. We thankZOLL Medical which assisted in the data collection for this study.We also thank Jose Guillen-Rodriquez for statistical programminghelp.

References

1. Roger VL, Go AS, Lloyd-Jones DM, et al. Heart disease and stroke statistics – 2012update: a report from the American Heart Association. Circulation 2012:125.

2. Nichol G, Thomas E, Callaway CW, et al. Regional variation in out-of-hospitalcardiac arrest incidence and outcome. JAMA 2008;300:1423–31.

3. Rea TD, Eisenberg MS, Sinibaldi G, White RD. Incidence of EMS-treated out-of-hospital cardiac arrest in the United States. Resuscitation 2004;63:17–24.

4. Sasson C, Rogers MA, Dahl J, Kellermann AL. Predictors of survival from out-of-hospital cardiac arrest a systematic review and meta-analysis. Circ CardiovascQual Outcomes 2010;3:63–81.

5. Vadeboncoeur T, Stolz U, Panchal A, et al. Chest compression depth and survivalin out-of-hospital cardiac arrest. Resuscitation 2014;85:182–8.

6. Wik L, Kramer-Johansen J, Myklebust H, et al. Quality of cardiopulmonary resus-citation during out-of-hospital cardiac arrest. JAMA 2005;293:299–304.

7. Abella BS, Sandbo N, Vassilatos P, et al. Chest compression rates duringcardiopulmonary resuscitation are suboptimal: a prospective study during in-hospital cardiac arrest. Circulation 2005;111:428–34.

8. Valenzuela TD, Kern KB, Clark LL, et al. Interruptions of chest compressions dur-ing emergency medical systems resuscitation. Circulation 2005;112:1259–65.

9. Travers AH, Rea TD, Bobrow BJ, et al. Part 4: CPR overview: 2010 AmericanHeart Association Guidelines for cardiopulmonary resuscitation and emergencycardiovascular care. Circulation 2010;122:S676–84.

0. Koster RW, Baubin MA, Bossaert LL, et al. European Resuscitation Council Guide-lines for Resuscitation 2010 Section 2. Adult basic life support and use ofautomated external defibrillators. Resuscitation 2010;81:1277–92.

1. Edelson DP, Abella BS, Kramer-Johansen J, et al. Effects of compression depthand pre-shock pauses predict defibrillation failure during cardiac arrest. Resus-citation 2006;71:137–45.

2. Stiell IG, Brown SP, Christenson J, et al. What is the role of chest compres-sion depth during out-of-hospital cardiac arrest resuscitation? Crit Care Med2012;40:1192.

3. Yannopoulos D, McKnite S, Aufderheide TP, et al. Effects of incomplete chest walldecompression during cardiopulmonary resuscitation on coronary and cere-bral perfusion pressures in a porcine model of cardiac arrest. Resuscitation2005;64:363–72.

4. Zuercher M, Hilwig RW, Ranger-Moore J, et al. Leaning during chest compres-sions impairs cardiac output and left ventricular myocardial blood flow in pigletcardiac arrest. Crit Care Med 2010;38:1141–6.

5. Niles DE, Sutton RM, Nadkarni VM, et al. Prevalence and hemodynamic effectsof leaning during CPR. Resuscitation 2011;82:S23–6.

6. Abella BS, Alvarado JP, Myklebust H, et al. Quality of cardiopulmonary resusci-tation during in-hospital cardiac arrest. JAMA 2005;293:305–10.

7. Fried DA, Leary M, Smith DA, et al. The prevalence of chest compres-sion leaning during in-hospital cardiopulmonary resuscitation. Resuscitation2011;82:1019–24.

8. Aufderheide TP, Pirrallo RG, Yannopoulos D, et al. Incomplete chest wall decom-pression: a clinical evaluation of CPR performance by EMS personnel andassessment of alternative manual chest compression–decompression tech-niques. Resuscitation 2005;64:353–62.

9. Bobrow BJ, Vadeboncoeur TF, Clark L, Chikani V. Establishing Arizona’s statewidecardiac arrest reporting and educational network. Prehospital Emerg Care Off JNatl Assoc EMS Physicians Natl Assoc State EMS Dir 2008;12:381–7.

0. Bobrow BJ, Clark LL, Ewy GA, et al. Minimally interrupted cardiac resuscita-tion by emergency medical services for out-of-hospital cardiac arrest. JAMA2008;299:1158–65.

1. Bobrow BJ, Vadeboncoeur TF, Stolz U, et al. The influence of scenario-based train-ing and real-time audiovisual feedback on out-of-hospital cardiopulmonaryresuscitation quality and survival from out-of-hospital cardiac arrest. Ann EmergMed 2013:62.

2. Meaney PA, Bobrow BJ, Mancini ME, et al. Cardiopulmonary resuscitation qual-ity: improving cardiac resuscitation outcomes both inside and outside thehospital a consensus statement from the American Heart Association. Circu-lation 2013;128:417–35.

3. Cheskes S, Common MR, Byers AP, et al. The association between chest com-

4. Newgard CD, Haukoos JS. Advanced statistics: missing data in clinical research– part 2: multiple imputation. Acad Emerg Med Off J Soc Acad Emerg Med2007;14:669–78.

5. Royston P, Carlin JB, White IR. Multiple imputation of missing values: new fea-tures for mim. Stata J 2009;9:252.

1 citatio

2

2

2

2

3

3

3

3

3

3

3

3

3

3

monary resuscitation with augmentation of negative intrathoracic pressure

14 A. Kovacs et al. / Resus

6. White IR, Royston P, Wood AM. Multiple imputation using chained equations:issues and guidance for practice. Stat Med 2011;30:377–99.

7. Edelson DP, Litzinger B, Arora V, et al. Improving in-hospital cardiacarrest process and outcomes with performance debriefing. Arch Intern Med2008;168:1063–9.

8. Silver A, Vadeboncoeur T, Venuti M, et al. Chest compression quality declines inthe minutes preceding scene departure in out-of-hospital cardiac arrest. Resus-citation 2013;84:S27.

9. Sutton RM, Niles D, Nysaether J, et al. Quantitative analysis of CPR qualityduring in-hospital resuscitation of older children and adolescents. Pediatrics2009;124:494–9.

0. Niles D, Nysaether J, Sutton R, et al. Leaning is common during in-hospital pedi-atric CPR, and decreased with automated corrective feedback. Resuscitation2009;80:553–7.

1. Tomlinson A, Nysaether J, Kramer-Johansen J, Steen P, Dorph E. Compressionforce–depth relationship during out-of-hospital cardiopulmonary resuscitation.Resuscitation 2007;72:364–70.

2. Olasveengen TM, Tomlinson A-E, Wik L, et al. A failed attempt to improve quality

of out-of-hospital CPR through performance evaluation. Prehosp Emerg Care2007;11:427–33.

3. Kramer-Johansen J, Myklebust H, Wik L, et al. Quality of out-of-hospital car-diopulmonary resuscitation with real time automated feedback: a prospectiveinterventional study. Resuscitation 2006;71:283–92.

4

n 92 (2015) 107–114

4. Mirza M, Brown TB, Saini D, et al. Instructions to “push as hard as you can”improve average chest compression depth in dispatcher-assisted cardiopul-monary resuscitation. Resuscitation 2008;79:97–102.

5. Sutton RM, Nadkarni V, Abella BS. “Putting it all together” to improve resuscita-tion quality. Emerg Med Clin North Am 2012;30:105–22.

6. Christenson J, Andrusiek D, Everson-Stewart S, et al. Chest compression frac-tion determines survival in patients with out-of-hospital ventricular fibrillation.Circulation 2009;120:1241–7.

7. Sutton RM, Niles D, Nysaether J, et al. Effect of residual leaning force onintrathoracic pressure during mechanical ventilation in children. Resuscitation2010;81:857–60.

8. Glatz AC, Nishisaki A, Niles DE, et al. Sternal wall pressure comparable toleaning during CPR impacts intrathoracic pressure and haemodynamics inanaesthetized children during cardiac catheterization. Resuscitation 2013;84:1674–9.

9. Aufderheide TP, Frascone RJ, Wayne MA, et al. Standard cardiopul-monary resuscitation versus active compression-decompression cardiopul-

for out-of-hospital cardiac arrest: a randomised trial. Lancet 2011;377:301–11.

0. Vittinghoff E, McCulloch CE. Relaxing the rule of ten events per variable in logisticand Cox regression. Am J Epidemiol 2007;165:710–8.