- Email: [email protected]
Association of neighborhood characteristics with incidence of out-of-hospital cardiac arrest and rates of bystander-initiated CPR: Implications for community-based education intervention
Resuscitation 85 (2014) 1512–1517
Contents lists available at ScienceDirect
Resuscitation journal homepage: www.elsevier.com/locate/resuscitation
Clinical Paper
Association of neighborhood characteristics with incidence of out-of-hospital cardiac arrest and rates of bystander-initiated CPR: Implications for community-based education intervention Emil L. Fosbøl a,∗ , Matthew E. Dupre a , Benjamin Strauss b,c , Douglas R. Swanson d , Brent Myers e , Bryan F. McNally f , Monique L. Anderson a , Akshay Bagai g , Lisa Monk a , J. Lee Garvey h , Matthew Bitner i , James G. Jollis a , Christopher B. Granger a a
Duke Clinical Research Institute, Durham, NC, USA School of Natural Resources and Environment, University of Michigan, Ann Arbor, MI, USA c Nicholas School of the Environment, Duke University, Durham, NC, USA d Mecklenburg Emergency Medical Services Agency, Charlotte, NC, USA e Wake County Department of Emergency Medical Services, Raleigh, NC, USA f Emory University School of Medicine, Rollins School of Public Health, Atlanta, GA, USA g St. Michael’s Hospital, University of Toronto, Toronto, Canada h Department of Emergency Medicine, Carolinas Medical Center, Charlotte, NC, USA i Division of Emergency Medicine, Department of Surgery, Duke University Health System, Durham, NC, USA b
a r t i c l e
i n f o
Article history: Received 6 January 2014 Received in revised form 5 August 2014 Accepted 11 August 2014 Keywords: Out-of-hospital cardiac arrest OHCA Bystander cardiac pulmonary resuscitation
a b s t r a c t Objective: A 10-fold regional variation in survival after out-of-hospital cardiac arrest (OHCA) has been reported in the United States, which partly relates to variability in bystander cardiopulmonary resuscitation (CPR) rates. In order for resources to be focused on areas of greatest need, we conducted a geospatial analysis of variation of CPR rates. Methods: Using 2010–2011 data from Durham, Mecklenburg, and Wake counties in North Carolina participating in the Cardiac Arrest Registry to Enhance Survival (CARES) program, we included all patients with OHCA for whom resuscitation was attempted. Geocoded data and logistic regression modeling were used to assess incidence of OHCA and patterns of bystander CPR according to census tracts and factors associated herewith. Results: In total, 1466 patients were included (median age, 65 years [interquartile range 25]; 63.4% men). Bystander CPR by a layperson was initiated in 37.9% of these patients. High-incidence OHCA areas were characterized partly by higher population densities and higher percentages of black race as well as lower levels of education and income. Low rates of bystander CPR were associated with population composition (percent black: OR, 3.73; 95% CI, 2.00–6.97 per 1% increment in black patients; percent elderly: 3.25; 1.41–7.48 per 1% increment in elderly patients; percent living in poverty: 1.77, 1.16–2.71 per 1% increase in patients living in poverty). Conclusions: In 3 counties in North Carolina, areas with low rates of bystander CPR can be identified using geospatial data, and education efforts can be targeted to improve recognition of cardiac arrest and to augment bystander CPR rates. © 2014 Elsevier Ireland Ltd. All rights reserved.
1. Introduction Survival remains poor after out-of-hospital cardiac arrest (OHCA)—a condition affecting more than 300,000 people in the
∗ Correspondence to: Duke Clinical Research Institute, 2400 Pratt Street, Room 7461, Terrace Level, Durham, NC 27705, USA. E-mail address: [email protected] (E.L. Fosbøl). http://dx.doi.org/10.1016/j.resuscitation.2014.08.013 0300-9572/© 2014 Elsevier Ireland Ltd. All rights reserved.
United States annually1 A 10-fold regional variation in survival has been reported in the United States,2 which in part is related to variability in rates of bystander-initiated cardiopulmonary resuscitation (CPR).3–5 A recent study reported an association between income and race and the likelihood of bystander-initiated CPR in the United States: neighborhoods characterized by low income and more black people had a lower probability of bystanders initiating CPR.6 Other studies have demonstrated that high-incidence areas of OHCA are identifiable,7–11 yet fewer studies have examined
E.L. Fosbøl et al. / Resuscitation 85 (2014) 1512–1517
how to identify areas of low bystander-initiated CPR.6,8 Further data regarding granular identification of bystander CPR rates at the neighborhood level are needed. Such identification could provide an opportunity to focus training efforts on areas of greatest need. We hypothesized that low-CPR-rate areas are identifiable and that geographic determinants could be determined. To address these issues, we performed a geospatial analysis of data from the Cardiac Arrest Registry to Enhance Survival (CARES) program. We aimed to (1) characterize geospatial patterns of OHCA in 3 counties in North Carolina; (2) identify high-incidence OHCA areas and assess variability in bystander-initiated CPR rates; and (3) examine patient and community factors related to low rates of bystander-initiated CPR in high-incidence OHCA areas. 2. Methods 2.1. Data sources The CARES program is a collaborative effort of the Centers for Disease Control and Prevention, American Heart Association (AHA), and Emory University Department of Emergency Medicine, Section of Prehospital and Disaster Medicine.12 Since 2012, a private funding model for CARES began that includes the American Red Cross, Medtronic Foundation HeartRescue Project, AHA, and Zoll Medical Corporation. This voluntary quality-improvement effort was initiated in 2004 and now includes numerous geographical sites throughout the United States collecting data on OHCA. Using the Utstein style of statistics for OHCA,13 CARES is capable of identifying and tracking all cases of cardiac arrest in a defined geographic area. CARES incorporates information from dispatch, emergency medical services (EMS), and hospital records to complete the Utstein data elements. The Duke University institutional Review Board approved this study. CARES data are censored for direct personal identifiers and, as a result, the Institutional Review Board waived informed consent and authorization for this study. Using data from the CARES program for three counties in North Carolina (Durham, Mecklenburg, and Wake), we identified all OHCAs from January 1, 2010 to December 31, 2011. We included all OHCAs for which the etiology was thought to be cardiac, as in only those patients for whom resuscitation was initiated or continued by EMS and/or a shock was delivered by an automated external defibrillator (AED). We excluded OHCAs that happened after the arrival of EMS (n = 307) and those for which location was one of the following: jail, nursing home/assisted living facility, or hospital or medical center (n = 249). Geographical- and community-level data were then linked with CARES data through the OHCA incident address. 2.2. Geocoding and geographical data Geocoding of incident addresses was performed using ArcGIS 10.0 software (ESRI, Redlands, CA). Addresses were geocoded to the street level, meaning that each address was matched to a latitude and longitude at the street centerline on which the addresses is located. The geocoding process creates a spatial dataset, which allows for the linkage of disparate data based on common geography. Ninety-five percent of the addresses were successfully geocoded for 2010–2011 Tract-level geographical information was obtained from the 2010 Census Summary Files 1 and 3 and was linked to the geocoded NC CARES data. Several geographic variables were also calculated to quantify and characterize the area in which each incident occurred. For each geocoded incident, a 5-mile buffer was calculated. Common geography was used to link each incident to census tract information, including demographic and economic factors.
1513
2.3. Study covariates The CARES registry included data on patient age (years), sex, and race/ethnicity. Bystander CPR was defined as the initiation of resuscitation by any person not identified as medical personnel or part of the 911 response team (i.e., EMS or first responder). Information regarding bystander CPR was ascertained from EMS records through the data submission system developed by CARES. Additional patient-level data included whether the cardiac arrest was witnessed (y/n), whether it occurred at home or at a private residence (vs. public), whether there was shockable rhythm at presentation (y/n), whether there was return of spontaneous circulation (ROSC, y/n), time from 911 call to EMS arrival (minutes), and subsequent discharge status as alive (y/n) and alive in a good cerebral performance category (y/n) according to Utstein criteria.13 Overall, missing data on patientlevel characteristics in CARES were minimal (<0.5%); however, approximately 17% of patients had missing data on 911-toarrival time (primarily in Wake county) and were imputed to the county-specific average response time. Sensitivity analyses for this measure showed that alternative approaches to handling missing data (i.e., imputation or omission) had little impact on the findings. Tract-level characteristics identified from linked U.S. Census data included population density (persons per square mile), race/ethnic composition, age composition, educational level, and household income. All variables were continuous and analyzed as such. 2.3.1. Variables of interest and study outcome For this study, we examined two main variables: incidence of cardiac arrest by tract and bystander CPR rate by tract. These two variables were then combined in a 2 × 2 fashion, and the study outcome was then the combination of high incidence of cardiac arrest and low rate of bystander CPR. Low cardiac arrest rate was defined as ≤0.06% (median rate) of the population with cardiac arrest. High cardiac arrest rate was >0.06%. Likewise, low bystander CPR rate was defined as ≤33.33%, and high bystander rate was defined as >33.33%. 2.4. Statistical analysis We used the Kruskal–Wallis test for continuous variables and 2 tests for categorical variables to test for patient and tract-level differences across groups. Next, to identify factors associated with low rates of bystander CPR in areas with high rates of OHCA, we estimated logistic regression models with robust variance estimators to account for clustering within census tracts. A P value of <0.05 was considered statistically significant. All analyses were performed using SAS versions 9.2 and 9.3 (SAS Institute, Cary, NC) and Stata version 12.1 (StataCorp, College Station, TX) software. 3. Results 3.1. OHCA characteristics From January 2010 to December 2011, a total of 1466 incidences of OHCAs from three counties in North Carolina met our selection criteria and were included. High/low incidence of OHCA at the tract level is shown by county in Fig. 1; high/low bystander CPR rate at the tract level is shown in Fig. 2; and the study outcome (tracts of high OHCA incidence and low bystander CPR rate) is shown by tract in Fig. 3. Table 1 shows the patient-level and tract-level baseline characteristics of the cohort according to the variables of interest (the combination of high vs. low incidence of OHCA and high vs. low
1514
E.L. Fosbøl et al. / Resuscitation 85 (2014) 1512–1517
Fig. 1. Cardiac arrest rate as percent of population in three counties in North Carolina, USA.
rate of bystander CPR). For patient-level characteristics, the overall age of the cohort was 65 years, and 63.4% were men. A little less than half (47%) of the OHCAs were witnessed by a bystander, and 38% of all OHCAs received bystander-initiated CPR. The high OHCA incidence and low bystander CPR group was overall characterized by a higher rate of black race at the individual level (61% vs. 24% to 39% in other groups) and also at the tract level. In addition, a higher percentage of the population at the tract level was living in poverty.
3.2. Factors associated with low rates of bystander-initiated CPR in high-incidence areas Table 2 shows the adjusted estimated odds ratios (ORs) for factors associated with low rates of bystander-initiated CPR in high-incidence areas of OHCA (all others as reference). An increment in tract-level percentage of black race was also associated
with lower rates of bystander-initiated CPR, as were increments in the tract-level percentage of people classified as poor and elderly. 4. Discussion This study used a novel geospatial approach for evaluating geographical differences in OHCA incidence and care. Our study resulted in two main findings. First, by using geospatial data compiled with the CARES registry, we were able to identify areas of high OHCA incidence and, in turn, areas with low rates of bystander-initiated CPR. Second, census tracts characterized by a high percentage of people who were black, were elderly, and had low income were associated with lower rates of bystander-initiated CPR. Our study demonstrates that identifying high-incidence areas of OHCA is possible. This could be employed on a national scale according to coverage and dissemination of the CARES registry.
Fig. 2. Bystander CPR rates in three counties in North Carolina, USA. Abbreviations: CPR; cardiopulmonary resuscitation.
E.L. Fosbøl et al. / Resuscitation 85 (2014) 1512–1517
1515
Fig. 3. Areas with high rates of cardiac arrest and low rates of bystander CPR. Abbreviations: CPR; cardiopulmonary resuscitation.
Areas with clusters of high incidence could be educated through community CPR training or AED placement, and such efforts could be tailored specifically to neighborhoods or on a census tract level. Intervention could be targeted via training and educational efforts, especially in places where the proportion of OHCAs happening in public is high. Another approach could be—as suggested by the AHA and prior studies [14]—to deploy training more broadly to high-risk areas where the incidence of OHCA is high but the corresponding
rate of bystander-initiated CPR is low. Our study suggests that these areas may be identifiable using a few simple characteristics (age, race, mean income, and location) of the local community and the people living in it. A recent study reported an association between income and race and probability of bystander-initiated CPR in the United States. Areas characterized by a higher proportion of lowincome black people had a lower probability of bystander-initiated CPR than neighborhoods of white, non-poor people.6 Other
Table 1 Individual and tract-level characteristics by cardiac arrest rate and bystander CPR rate: CARES, 2010–2011. Overall (n = 1466)
Individual level Age, median (IQR) Male, No. (%) Race/ethnicity, no. (%) White Black Other Cardiac arrest witnessed, no. (%) Cardiac arrest location, no. (%) Private (home/residence) Public Initiated CPR, no. (%) Bystander EMS or first responder Initial shockable rhythm, no. (%) ROSC, no. (%) 911 call to EMS arrival (min.), median (IQR) Discharge status, no. (%) Alive Alive with good cerebral performance Tract Level Population density in thousands, median (IQR) Population composition, median (IQR) Percent elderly Percent white Percent black Percent other race/ethnicity Percent in poverty, median (IQR)
Low cardiac arrest rate
High cardiac arrest rate
Bystander CPR rate
Bystander CPR rate
P value†
High (n = 374)
Low (n = 357)
High (n = 415)
Low (n = 320)
65 (24) 929 (63.4)
64 (25) 251 (67.1)
66 (23) 222 (62.2)
64 (23) 263 (63.4)
64 (23) 193 (60.3)
0.131 0.199
703 (48.0) 569 (38.8) 194 (13.2) 683 (46.6)
219 (58.6) 91 (24.3) 64 (17.1) 186 (49.7)
179 (50.1) 118 (33.1) 60 (16.8) 163 (45.7)
212 (51.1) 164 (39.5) 39 (9.40) 207 (49.9)
93 (29.1) 196 (61.3) 31 (9.69) 127 (39.7)
<0.001 <0.001 0.034 0.005
1081 (73.7) 385 (26.3)
312 (83.4) 62 (16.6)
293 (82.1) 64 (17.9)
222 (53.5) 193 (46.5)
254 (79.4) 66 (20.6)
0.010 0.010
555 (37.9) 911 (62.1) 382 (26.1) 244 (16.6) 7.9 (3.4)
218 (58.3) 114 (30.5) 104 (27.8) 65 (17.4) 9.0 (4.0)
38 (10.6) 236 (66.1) 92 (25.8) 46 (12.9) 8.7 (2.6)
241 (58.1) 119 (28.7) 108 (26.0) 83 (20.0) 7.7 (3.2)
58 (18.1) 198 (61.9) 78 (24.4) 50 (15.6) 7.7 (2.8)
<0.001 <0.001 0.438 0.580 <0.001
159 (10.9) 111 (7.6)
36 (9.6) 26 (7.0)
36 (10.1) 23 (6.4)
52 (12.5) 39 (9.4)
35 (10.9) 23 (7.2)
0.952 0.769
2.21 (1.94)
2.07 (1.97)
2.01 (2.16)
2.39 (1.92)
2.40 (2.05)
0.023
9.00 (6.00) 47.0 (54.0) 30.0 (41.0) 15.0 (15.0) 10.8 (15.2)
8.0 (5.00) 61.0 (39.0) 21.0 (26.0) 16.0 (12.0) 6.30 (10.8)
8.0 (6.00) 51.0 (49.0) 28.0 (35.0) 17.0 (14.0) 7.60 (11.6)
10.0 (7.00) 52.0 (61.0) 25.0 (41.0) 14.0 (17.0) 11.8 (15.8)
10.0 (5.00) 18.0 (26.5) 64.5 (40.0) 14.0 (13.0) 18.9 (20.8)
<0.001 <0.001 <0.001 0.024 <0.001
Abbreviations: CPR, cardiopulmonary resuscitation; EMS, emergency medical services; IQR, interquartile range; ROSC, return of spontaneous circulation. † P values indicate differences between (i) areas with high rates of cardiac arrest and low rates of bystander CPR vs. (ii) other areas.
1516
E.L. Fosbøl et al. / Resuscitation 85 (2014) 1512–1517
Table 2 Individual- and tract-level factors associated with areas that have high rates of cardiac arrest and low rates of bystander CPR (n = 1466).
Individual Level Age (years) Male (vs. female) Race/ethnicity Black (vs. white) Other (vs. white) Cardiac arrest in home/residence (vs. public) Time from 911 call to EMS arrival (in min) Tract Level Population density in thousands Population composition Percent elderly Percent black Percent other race/ethnicity Percent in poverty
OR
(95% CI)
1.21 0.97
(0.81–1.81) (0.73–1.28)
0.360 0.814
1.11 1.06 1.43
(0.74–1.66) (0.55–2.03) (0.88–2.32)
0.628 0.864 0.151
0.87
(0.59–1.28)
0.486
0.76
(0.47–1.22)
0.248
3.25 3.73 0.70
(1.41–7.48) (2.00–6.97) (0.35–1.42)
0.005 <0.001 0.325
1.77
(1.16–2.71)
0.008
greatest need) in the United States for training laypersons in CPR and increasing the bystander-initiated CPR rates throughout the country.
P Value
4.1. Limitations
Abbreviations: OR, odds ratio; CI, confidence interval; EMS, emergency medical services. Note: Continuous variables are in logarithmic scale.
studies have shown similar results for identifying high-incidence OHCA areas,7–11 and our results are similar to those reported for bystander CPR as well. Our results correspond well with these prior findings, although our study adds a more granular perspective and underlines the importance of taking the local community into perspective before deploying a training strategy. The CARES registry provides potential for tailoring AED placement more effectively nationally as well as assessing the prospective effect of targeted education and AED placement. Our study is a proof-of-concept for OHCA surveillance in the United States and underscores the need for improving bystander CPR rates in some neighborhoods. The variation in bystander CPR rates according to geography is significant in these three similar counties in the same state. From prior studies, we know that this variation is substantial between states as well.3,4 Our study also points out an important discrepancy in terms of bystander CPR. The rate of bystander CPR was substantially lower if the OHCA happened in the patient’s home vs. in public (more than two-thirds of cardiac arrests occur in the home). The lower rate of bystander CPR at home might in some degree be explained by patients living alone, as shown previously,15,16 but could also be due to the fact that spouses less often initiate CPR.15,16 Compared with younger people, older adults are generally not as well trained in CPR, which could also help explain this important finding. We believe that emergency dispatchers have a key role here in that they could instruct spouses in performing CPR and thereby improve the likelihood of survival in OHCA. Finally, this finding could also be due to the fact that patients are more likely to be found at home if OHCA occurred during the night. Our study showed that bystander CPR is initiated in a minority of witnessed OHCAs. Other countries have tried to address this issue by introducing mandatory CPR training at various occasions. For example, some European countries require completion of a life support course in order to acquire a driver’s license. This has been shown to be associated with a substantial increase in bystander CPR.17 Other studies have shown that 911 dispatch instructions to perform CPR can improve rates. We believe that our study, together with previous studies, warrants a multifaceted approach (including particular focus on those neighborhoods with
Our study had several limitations. First, we used data from three specific counties in North Carolina, and our results may not be generalizable to communities in different parts of the United States as well as in the rest of the world. Second, we included OHCAs with an assumed cardiac etiology. Third, we had to exclude some cases in our study due to unsuccessful linkage between the CARES registry and the census data because the incident address was invalid, yet this was a small percentage. Fourth, we had only limited information about the medical history of these patients, and follow-up was limited to discharge status. 5. Conclusions In OHCA, areas with low rates of bystander-initiated CPR (e.g., communities with a high proportion of blacks, elderly people, and poverty) can be identified using geospatial data, and education efforts can be targeted to such areas to improve recognition of cardiac arrest and augment bystander CPR rates. Conflict of interest statement E.L. Fosbøl: None. B. Strauss: None. D.R. Swanson: None. B. Myers: None. M.E. Dupre: None. B.F. McNally: Funding for CARES (Cardiac Arrest Registry to Enhance Survival) is provided by: The American Red Cross, Medtronic Foundation Heart Rescue Program, American Heart Association, Zoll Corporation. M.L. Anderson: None. A. Bagai: None. L. Monk: None. L. Garvey: Advisory board member for Philips Healthcare (minor, <$5,000). M. Bitner: None. J.G. Jollis: Medtronic Foundation, Philips Healthcare, Abiomed Inc., The Medicines Company, Astra Zeneca C.B. Granger: All disclosures available at https://dcri.org/ about-us/conflict-of-interest. Sources of funding This work was supported by an award from the American Heart Association Pharmaceutical Roundtable and by David and Stevie Spina. Acknowledgments The authors would like to thank Morgan deBlecourt for her editorial contributions to this manuscript. Ms. deBlecourt did not receive compensation for her contributions, apart from her employment at the institution where this study was conducted. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.resuscitation. 2014.08.013.
E.L. Fosbøl et al. / Resuscitation 85 (2014) 1512–1517
References 1. Roger VL, Go AS, Lloyd-Jones DM, et al. Heart disease and stroke statistics2012 update: a report from the American Heart Association. Circulation 2012;125:e2–220. 2. Nichol G, Thomas E, Callaway CW, et al. Regional variation in out-ofhospital cardiac arrest incidence and outcome. J Am Med Assoc 2008;300: 1423–31. 3. Cummins RO. Emergency medical services and sudden cardiac arrest: the chain of survival concept. Annu Rev Public Health 1993;14:313–33. 4. ECC Committee, Subcommittees and Task Forces of the American Heart Association. 2005 American Heart Association Guidelines for Cardiopulmonary Resuscitation and Emergency Cardiovascular Care. Circulation 2005;112:IV1–203. 5. Sasson C, Rogers MA, Dahl J, Kellermann AL. Predictors of survival from out-ofhospital cardiac arrest: a systematic review and meta-analysis. Circ Cardiovasc Qual Outcomes 2010;3:63–81. 6. Sasson C, Magid DJ, Chan P, et al. Association of neighborhood characteristics with bystander-initiated CPR. N Engl J Med 2012;367:1607–15. 7. Sasson C, Cudnik MT, Nassel A, et al. Identifying high-risk geographic areas for cardiac arrest using three methods for cluster analysis. Acad Emerg Med 2012;19:139–46. 8. Sasson C, Keirns CC, Smith D, et al. Small area variations in out-of-hospital cardiac arrest: does the neighborhood matter? Ann Intern Med 2010;153:19–22. 9. Semple HM, Cudnik MT, Sayre M, et al. Identification of high-risk communities for unattended out-of-hospital cardiac arrests using GIS. J Community Health 2013;38:277–84. 10. Warden C, Cudnik MT, Sasson C, Schwartz G, Semple H. Poisson cluster analysis of cardiac arrest incidence in Columbus, Ohio. Prehosp Emerg Care 2012;16:338–46.
1517
11. Warden CR, Daya M, LeGrady LA. Using geographic information systems to evaluate cardiac arrest survival. Prehosp Emerg Care 2007;11:19–24. 12. McNally B, Robb R, Mehta M, et al. Out-of-hospital cardiac arrest surveillance—Cardiac Arrest Registry to Enhance Survival (CARES), United States, October 1, 2005–December 31, 2010. MMWR Surveill Summ 2011;60:1–19. 13. Jacobs I, Nadkarni V, Bahr J, et al. Cardiac arrest and cardiopulmonary resuscitation outcome reports: update and simplification of the Utstein templates for resuscitation registries: a statement for healthcare professionals from a task force of the International Liaison Committee on Resuscitation (American Heart Association, European Resuscitation Council, Australian Resuscitation Council, New Zealand Resuscitation Council, Heart and Stroke Foundation of Canada, InterAmerican Heart Foundation, Resuscitation Councils of Southern Africa). Circulation 2004;110:3385–97. 14. Sasson C, Meischke H, Abella BS, et al. Increasing cardiopulmonary resuscitation provision in communities with low bystander cardiopulmonary resuscitation rates: a science advisory from the American Heart Association for healthcare providers, policymakers, public health departments, and community leaders. Circulation 2013;127:1342–50. 15. Sasson C, Ramirez L, Medrano M, et al. Barriers to calling 911 and performing cardiopulmonary resuscitation (CPR) in High-risk neighborhoods in Denver. Poster presented at: the American Heart Association Scientific Sessions; November 2012; Los Angeles, CA. Abstract 283. 16. Swor RA, Jackson RE, Compton S, et al. Cardiac arrest in private locations: different strategies are needed to improve outcome. Resuscitation 2003;58:171–6. 17. Wissenberg M, Lippert FK, Folke F, et al. Association of national initiatives to improve cardiac arrest management with rates of bystander intervention and patient survival after out-of-hospital cardiac arrest. J Am Med Assoc 2013;310:1377–84.
Contents lists available at ScienceDirect
Resuscitation journal homepage: www.elsevier.com/locate/resuscitation
Clinical Paper
Association of neighborhood characteristics with incidence of out-of-hospital cardiac arrest and rates of bystander-initiated CPR: Implications for community-based education intervention Emil L. Fosbøl a,∗ , Matthew E. Dupre a , Benjamin Strauss b,c , Douglas R. Swanson d , Brent Myers e , Bryan F. McNally f , Monique L. Anderson a , Akshay Bagai g , Lisa Monk a , J. Lee Garvey h , Matthew Bitner i , James G. Jollis a , Christopher B. Granger a a
Duke Clinical Research Institute, Durham, NC, USA School of Natural Resources and Environment, University of Michigan, Ann Arbor, MI, USA c Nicholas School of the Environment, Duke University, Durham, NC, USA d Mecklenburg Emergency Medical Services Agency, Charlotte, NC, USA e Wake County Department of Emergency Medical Services, Raleigh, NC, USA f Emory University School of Medicine, Rollins School of Public Health, Atlanta, GA, USA g St. Michael’s Hospital, University of Toronto, Toronto, Canada h Department of Emergency Medicine, Carolinas Medical Center, Charlotte, NC, USA i Division of Emergency Medicine, Department of Surgery, Duke University Health System, Durham, NC, USA b
a r t i c l e
i n f o
Article history: Received 6 January 2014 Received in revised form 5 August 2014 Accepted 11 August 2014 Keywords: Out-of-hospital cardiac arrest OHCA Bystander cardiac pulmonary resuscitation
a b s t r a c t Objective: A 10-fold regional variation in survival after out-of-hospital cardiac arrest (OHCA) has been reported in the United States, which partly relates to variability in bystander cardiopulmonary resuscitation (CPR) rates. In order for resources to be focused on areas of greatest need, we conducted a geospatial analysis of variation of CPR rates. Methods: Using 2010–2011 data from Durham, Mecklenburg, and Wake counties in North Carolina participating in the Cardiac Arrest Registry to Enhance Survival (CARES) program, we included all patients with OHCA for whom resuscitation was attempted. Geocoded data and logistic regression modeling were used to assess incidence of OHCA and patterns of bystander CPR according to census tracts and factors associated herewith. Results: In total, 1466 patients were included (median age, 65 years [interquartile range 25]; 63.4% men). Bystander CPR by a layperson was initiated in 37.9% of these patients. High-incidence OHCA areas were characterized partly by higher population densities and higher percentages of black race as well as lower levels of education and income. Low rates of bystander CPR were associated with population composition (percent black: OR, 3.73; 95% CI, 2.00–6.97 per 1% increment in black patients; percent elderly: 3.25; 1.41–7.48 per 1% increment in elderly patients; percent living in poverty: 1.77, 1.16–2.71 per 1% increase in patients living in poverty). Conclusions: In 3 counties in North Carolina, areas with low rates of bystander CPR can be identified using geospatial data, and education efforts can be targeted to improve recognition of cardiac arrest and to augment bystander CPR rates. © 2014 Elsevier Ireland Ltd. All rights reserved.
1. Introduction Survival remains poor after out-of-hospital cardiac arrest (OHCA)—a condition affecting more than 300,000 people in the
∗ Correspondence to: Duke Clinical Research Institute, 2400 Pratt Street, Room 7461, Terrace Level, Durham, NC 27705, USA. E-mail address: [email protected] (E.L. Fosbøl). http://dx.doi.org/10.1016/j.resuscitation.2014.08.013 0300-9572/© 2014 Elsevier Ireland Ltd. All rights reserved.
United States annually1 A 10-fold regional variation in survival has been reported in the United States,2 which in part is related to variability in rates of bystander-initiated cardiopulmonary resuscitation (CPR).3–5 A recent study reported an association between income and race and the likelihood of bystander-initiated CPR in the United States: neighborhoods characterized by low income and more black people had a lower probability of bystanders initiating CPR.6 Other studies have demonstrated that high-incidence areas of OHCA are identifiable,7–11 yet fewer studies have examined
E.L. Fosbøl et al. / Resuscitation 85 (2014) 1512–1517
how to identify areas of low bystander-initiated CPR.6,8 Further data regarding granular identification of bystander CPR rates at the neighborhood level are needed. Such identification could provide an opportunity to focus training efforts on areas of greatest need. We hypothesized that low-CPR-rate areas are identifiable and that geographic determinants could be determined. To address these issues, we performed a geospatial analysis of data from the Cardiac Arrest Registry to Enhance Survival (CARES) program. We aimed to (1) characterize geospatial patterns of OHCA in 3 counties in North Carolina; (2) identify high-incidence OHCA areas and assess variability in bystander-initiated CPR rates; and (3) examine patient and community factors related to low rates of bystander-initiated CPR in high-incidence OHCA areas. 2. Methods 2.1. Data sources The CARES program is a collaborative effort of the Centers for Disease Control and Prevention, American Heart Association (AHA), and Emory University Department of Emergency Medicine, Section of Prehospital and Disaster Medicine.12 Since 2012, a private funding model for CARES began that includes the American Red Cross, Medtronic Foundation HeartRescue Project, AHA, and Zoll Medical Corporation. This voluntary quality-improvement effort was initiated in 2004 and now includes numerous geographical sites throughout the United States collecting data on OHCA. Using the Utstein style of statistics for OHCA,13 CARES is capable of identifying and tracking all cases of cardiac arrest in a defined geographic area. CARES incorporates information from dispatch, emergency medical services (EMS), and hospital records to complete the Utstein data elements. The Duke University institutional Review Board approved this study. CARES data are censored for direct personal identifiers and, as a result, the Institutional Review Board waived informed consent and authorization for this study. Using data from the CARES program for three counties in North Carolina (Durham, Mecklenburg, and Wake), we identified all OHCAs from January 1, 2010 to December 31, 2011. We included all OHCAs for which the etiology was thought to be cardiac, as in only those patients for whom resuscitation was initiated or continued by EMS and/or a shock was delivered by an automated external defibrillator (AED). We excluded OHCAs that happened after the arrival of EMS (n = 307) and those for which location was one of the following: jail, nursing home/assisted living facility, or hospital or medical center (n = 249). Geographical- and community-level data were then linked with CARES data through the OHCA incident address. 2.2. Geocoding and geographical data Geocoding of incident addresses was performed using ArcGIS 10.0 software (ESRI, Redlands, CA). Addresses were geocoded to the street level, meaning that each address was matched to a latitude and longitude at the street centerline on which the addresses is located. The geocoding process creates a spatial dataset, which allows for the linkage of disparate data based on common geography. Ninety-five percent of the addresses were successfully geocoded for 2010–2011 Tract-level geographical information was obtained from the 2010 Census Summary Files 1 and 3 and was linked to the geocoded NC CARES data. Several geographic variables were also calculated to quantify and characterize the area in which each incident occurred. For each geocoded incident, a 5-mile buffer was calculated. Common geography was used to link each incident to census tract information, including demographic and economic factors.
1513
2.3. Study covariates The CARES registry included data on patient age (years), sex, and race/ethnicity. Bystander CPR was defined as the initiation of resuscitation by any person not identified as medical personnel or part of the 911 response team (i.e., EMS or first responder). Information regarding bystander CPR was ascertained from EMS records through the data submission system developed by CARES. Additional patient-level data included whether the cardiac arrest was witnessed (y/n), whether it occurred at home or at a private residence (vs. public), whether there was shockable rhythm at presentation (y/n), whether there was return of spontaneous circulation (ROSC, y/n), time from 911 call to EMS arrival (minutes), and subsequent discharge status as alive (y/n) and alive in a good cerebral performance category (y/n) according to Utstein criteria.13 Overall, missing data on patientlevel characteristics in CARES were minimal (<0.5%); however, approximately 17% of patients had missing data on 911-toarrival time (primarily in Wake county) and were imputed to the county-specific average response time. Sensitivity analyses for this measure showed that alternative approaches to handling missing data (i.e., imputation or omission) had little impact on the findings. Tract-level characteristics identified from linked U.S. Census data included population density (persons per square mile), race/ethnic composition, age composition, educational level, and household income. All variables were continuous and analyzed as such. 2.3.1. Variables of interest and study outcome For this study, we examined two main variables: incidence of cardiac arrest by tract and bystander CPR rate by tract. These two variables were then combined in a 2 × 2 fashion, and the study outcome was then the combination of high incidence of cardiac arrest and low rate of bystander CPR. Low cardiac arrest rate was defined as ≤0.06% (median rate) of the population with cardiac arrest. High cardiac arrest rate was >0.06%. Likewise, low bystander CPR rate was defined as ≤33.33%, and high bystander rate was defined as >33.33%. 2.4. Statistical analysis We used the Kruskal–Wallis test for continuous variables and 2 tests for categorical variables to test for patient and tract-level differences across groups. Next, to identify factors associated with low rates of bystander CPR in areas with high rates of OHCA, we estimated logistic regression models with robust variance estimators to account for clustering within census tracts. A P value of <0.05 was considered statistically significant. All analyses were performed using SAS versions 9.2 and 9.3 (SAS Institute, Cary, NC) and Stata version 12.1 (StataCorp, College Station, TX) software. 3. Results 3.1. OHCA characteristics From January 2010 to December 2011, a total of 1466 incidences of OHCAs from three counties in North Carolina met our selection criteria and were included. High/low incidence of OHCA at the tract level is shown by county in Fig. 1; high/low bystander CPR rate at the tract level is shown in Fig. 2; and the study outcome (tracts of high OHCA incidence and low bystander CPR rate) is shown by tract in Fig. 3. Table 1 shows the patient-level and tract-level baseline characteristics of the cohort according to the variables of interest (the combination of high vs. low incidence of OHCA and high vs. low
1514
E.L. Fosbøl et al. / Resuscitation 85 (2014) 1512–1517
Fig. 1. Cardiac arrest rate as percent of population in three counties in North Carolina, USA.
rate of bystander CPR). For patient-level characteristics, the overall age of the cohort was 65 years, and 63.4% were men. A little less than half (47%) of the OHCAs were witnessed by a bystander, and 38% of all OHCAs received bystander-initiated CPR. The high OHCA incidence and low bystander CPR group was overall characterized by a higher rate of black race at the individual level (61% vs. 24% to 39% in other groups) and also at the tract level. In addition, a higher percentage of the population at the tract level was living in poverty.
3.2. Factors associated with low rates of bystander-initiated CPR in high-incidence areas Table 2 shows the adjusted estimated odds ratios (ORs) for factors associated with low rates of bystander-initiated CPR in high-incidence areas of OHCA (all others as reference). An increment in tract-level percentage of black race was also associated
with lower rates of bystander-initiated CPR, as were increments in the tract-level percentage of people classified as poor and elderly. 4. Discussion This study used a novel geospatial approach for evaluating geographical differences in OHCA incidence and care. Our study resulted in two main findings. First, by using geospatial data compiled with the CARES registry, we were able to identify areas of high OHCA incidence and, in turn, areas with low rates of bystander-initiated CPR. Second, census tracts characterized by a high percentage of people who were black, were elderly, and had low income were associated with lower rates of bystander-initiated CPR. Our study demonstrates that identifying high-incidence areas of OHCA is possible. This could be employed on a national scale according to coverage and dissemination of the CARES registry.
Fig. 2. Bystander CPR rates in three counties in North Carolina, USA. Abbreviations: CPR; cardiopulmonary resuscitation.
E.L. Fosbøl et al. / Resuscitation 85 (2014) 1512–1517
1515
Fig. 3. Areas with high rates of cardiac arrest and low rates of bystander CPR. Abbreviations: CPR; cardiopulmonary resuscitation.
Areas with clusters of high incidence could be educated through community CPR training or AED placement, and such efforts could be tailored specifically to neighborhoods or on a census tract level. Intervention could be targeted via training and educational efforts, especially in places where the proportion of OHCAs happening in public is high. Another approach could be—as suggested by the AHA and prior studies [14]—to deploy training more broadly to high-risk areas where the incidence of OHCA is high but the corresponding
rate of bystander-initiated CPR is low. Our study suggests that these areas may be identifiable using a few simple characteristics (age, race, mean income, and location) of the local community and the people living in it. A recent study reported an association between income and race and probability of bystander-initiated CPR in the United States. Areas characterized by a higher proportion of lowincome black people had a lower probability of bystander-initiated CPR than neighborhoods of white, non-poor people.6 Other
Table 1 Individual and tract-level characteristics by cardiac arrest rate and bystander CPR rate: CARES, 2010–2011. Overall (n = 1466)
Individual level Age, median (IQR) Male, No. (%) Race/ethnicity, no. (%) White Black Other Cardiac arrest witnessed, no. (%) Cardiac arrest location, no. (%) Private (home/residence) Public Initiated CPR, no. (%) Bystander EMS or first responder Initial shockable rhythm, no. (%) ROSC, no. (%) 911 call to EMS arrival (min.), median (IQR) Discharge status, no. (%) Alive Alive with good cerebral performance Tract Level Population density in thousands, median (IQR) Population composition, median (IQR) Percent elderly Percent white Percent black Percent other race/ethnicity Percent in poverty, median (IQR)
Low cardiac arrest rate
High cardiac arrest rate
Bystander CPR rate
Bystander CPR rate
P value†
High (n = 374)
Low (n = 357)
High (n = 415)
Low (n = 320)
65 (24) 929 (63.4)
64 (25) 251 (67.1)
66 (23) 222 (62.2)
64 (23) 263 (63.4)
64 (23) 193 (60.3)
0.131 0.199
703 (48.0) 569 (38.8) 194 (13.2) 683 (46.6)
219 (58.6) 91 (24.3) 64 (17.1) 186 (49.7)
179 (50.1) 118 (33.1) 60 (16.8) 163 (45.7)
212 (51.1) 164 (39.5) 39 (9.40) 207 (49.9)
93 (29.1) 196 (61.3) 31 (9.69) 127 (39.7)
<0.001 <0.001 0.034 0.005
1081 (73.7) 385 (26.3)
312 (83.4) 62 (16.6)
293 (82.1) 64 (17.9)
222 (53.5) 193 (46.5)
254 (79.4) 66 (20.6)
0.010 0.010
555 (37.9) 911 (62.1) 382 (26.1) 244 (16.6) 7.9 (3.4)
218 (58.3) 114 (30.5) 104 (27.8) 65 (17.4) 9.0 (4.0)
38 (10.6) 236 (66.1) 92 (25.8) 46 (12.9) 8.7 (2.6)
241 (58.1) 119 (28.7) 108 (26.0) 83 (20.0) 7.7 (3.2)
58 (18.1) 198 (61.9) 78 (24.4) 50 (15.6) 7.7 (2.8)
<0.001 <0.001 0.438 0.580 <0.001
159 (10.9) 111 (7.6)
36 (9.6) 26 (7.0)
36 (10.1) 23 (6.4)
52 (12.5) 39 (9.4)
35 (10.9) 23 (7.2)
0.952 0.769
2.21 (1.94)
2.07 (1.97)
2.01 (2.16)
2.39 (1.92)
2.40 (2.05)
0.023
9.00 (6.00) 47.0 (54.0) 30.0 (41.0) 15.0 (15.0) 10.8 (15.2)
8.0 (5.00) 61.0 (39.0) 21.0 (26.0) 16.0 (12.0) 6.30 (10.8)
8.0 (6.00) 51.0 (49.0) 28.0 (35.0) 17.0 (14.0) 7.60 (11.6)
10.0 (7.00) 52.0 (61.0) 25.0 (41.0) 14.0 (17.0) 11.8 (15.8)
10.0 (5.00) 18.0 (26.5) 64.5 (40.0) 14.0 (13.0) 18.9 (20.8)
<0.001 <0.001 <0.001 0.024 <0.001
Abbreviations: CPR, cardiopulmonary resuscitation; EMS, emergency medical services; IQR, interquartile range; ROSC, return of spontaneous circulation. † P values indicate differences between (i) areas with high rates of cardiac arrest and low rates of bystander CPR vs. (ii) other areas.
1516
E.L. Fosbøl et al. / Resuscitation 85 (2014) 1512–1517
Table 2 Individual- and tract-level factors associated with areas that have high rates of cardiac arrest and low rates of bystander CPR (n = 1466).
Individual Level Age (years) Male (vs. female) Race/ethnicity Black (vs. white) Other (vs. white) Cardiac arrest in home/residence (vs. public) Time from 911 call to EMS arrival (in min) Tract Level Population density in thousands Population composition Percent elderly Percent black Percent other race/ethnicity Percent in poverty
OR
(95% CI)
1.21 0.97
(0.81–1.81) (0.73–1.28)
0.360 0.814
1.11 1.06 1.43
(0.74–1.66) (0.55–2.03) (0.88–2.32)
0.628 0.864 0.151
0.87
(0.59–1.28)
0.486
0.76
(0.47–1.22)
0.248
3.25 3.73 0.70
(1.41–7.48) (2.00–6.97) (0.35–1.42)
0.005 <0.001 0.325
1.77
(1.16–2.71)
0.008
greatest need) in the United States for training laypersons in CPR and increasing the bystander-initiated CPR rates throughout the country.
P Value
4.1. Limitations
Abbreviations: OR, odds ratio; CI, confidence interval; EMS, emergency medical services. Note: Continuous variables are in logarithmic scale.
studies have shown similar results for identifying high-incidence OHCA areas,7–11 and our results are similar to those reported for bystander CPR as well. Our results correspond well with these prior findings, although our study adds a more granular perspective and underlines the importance of taking the local community into perspective before deploying a training strategy. The CARES registry provides potential for tailoring AED placement more effectively nationally as well as assessing the prospective effect of targeted education and AED placement. Our study is a proof-of-concept for OHCA surveillance in the United States and underscores the need for improving bystander CPR rates in some neighborhoods. The variation in bystander CPR rates according to geography is significant in these three similar counties in the same state. From prior studies, we know that this variation is substantial between states as well.3,4 Our study also points out an important discrepancy in terms of bystander CPR. The rate of bystander CPR was substantially lower if the OHCA happened in the patient’s home vs. in public (more than two-thirds of cardiac arrests occur in the home). The lower rate of bystander CPR at home might in some degree be explained by patients living alone, as shown previously,15,16 but could also be due to the fact that spouses less often initiate CPR.15,16 Compared with younger people, older adults are generally not as well trained in CPR, which could also help explain this important finding. We believe that emergency dispatchers have a key role here in that they could instruct spouses in performing CPR and thereby improve the likelihood of survival in OHCA. Finally, this finding could also be due to the fact that patients are more likely to be found at home if OHCA occurred during the night. Our study showed that bystander CPR is initiated in a minority of witnessed OHCAs. Other countries have tried to address this issue by introducing mandatory CPR training at various occasions. For example, some European countries require completion of a life support course in order to acquire a driver’s license. This has been shown to be associated with a substantial increase in bystander CPR.17 Other studies have shown that 911 dispatch instructions to perform CPR can improve rates. We believe that our study, together with previous studies, warrants a multifaceted approach (including particular focus on those neighborhoods with
Our study had several limitations. First, we used data from three specific counties in North Carolina, and our results may not be generalizable to communities in different parts of the United States as well as in the rest of the world. Second, we included OHCAs with an assumed cardiac etiology. Third, we had to exclude some cases in our study due to unsuccessful linkage between the CARES registry and the census data because the incident address was invalid, yet this was a small percentage. Fourth, we had only limited information about the medical history of these patients, and follow-up was limited to discharge status. 5. Conclusions In OHCA, areas with low rates of bystander-initiated CPR (e.g., communities with a high proportion of blacks, elderly people, and poverty) can be identified using geospatial data, and education efforts can be targeted to such areas to improve recognition of cardiac arrest and augment bystander CPR rates. Conflict of interest statement E.L. Fosbøl: None. B. Strauss: None. D.R. Swanson: None. B. Myers: None. M.E. Dupre: None. B.F. McNally: Funding for CARES (Cardiac Arrest Registry to Enhance Survival) is provided by: The American Red Cross, Medtronic Foundation Heart Rescue Program, American Heart Association, Zoll Corporation. M.L. Anderson: None. A. Bagai: None. L. Monk: None. L. Garvey: Advisory board member for Philips Healthcare (minor, <$5,000). M. Bitner: None. J.G. Jollis: Medtronic Foundation, Philips Healthcare, Abiomed Inc., The Medicines Company, Astra Zeneca C.B. Granger: All disclosures available at https://dcri.org/ about-us/conflict-of-interest. Sources of funding This work was supported by an award from the American Heart Association Pharmaceutical Roundtable and by David and Stevie Spina. Acknowledgments The authors would like to thank Morgan deBlecourt for her editorial contributions to this manuscript. Ms. deBlecourt did not receive compensation for her contributions, apart from her employment at the institution where this study was conducted. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.resuscitation. 2014.08.013.
E.L. Fosbøl et al. / Resuscitation 85 (2014) 1512–1517
References 1. Roger VL, Go AS, Lloyd-Jones DM, et al. Heart disease and stroke statistics2012 update: a report from the American Heart Association. Circulation 2012;125:e2–220. 2. Nichol G, Thomas E, Callaway CW, et al. Regional variation in out-ofhospital cardiac arrest incidence and outcome. J Am Med Assoc 2008;300: 1423–31. 3. Cummins RO. Emergency medical services and sudden cardiac arrest: the chain of survival concept. Annu Rev Public Health 1993;14:313–33. 4. ECC Committee, Subcommittees and Task Forces of the American Heart Association. 2005 American Heart Association Guidelines for Cardiopulmonary Resuscitation and Emergency Cardiovascular Care. Circulation 2005;112:IV1–203. 5. Sasson C, Rogers MA, Dahl J, Kellermann AL. Predictors of survival from out-ofhospital cardiac arrest: a systematic review and meta-analysis. Circ Cardiovasc Qual Outcomes 2010;3:63–81. 6. Sasson C, Magid DJ, Chan P, et al. Association of neighborhood characteristics with bystander-initiated CPR. N Engl J Med 2012;367:1607–15. 7. Sasson C, Cudnik MT, Nassel A, et al. Identifying high-risk geographic areas for cardiac arrest using three methods for cluster analysis. Acad Emerg Med 2012;19:139–46. 8. Sasson C, Keirns CC, Smith D, et al. Small area variations in out-of-hospital cardiac arrest: does the neighborhood matter? Ann Intern Med 2010;153:19–22. 9. Semple HM, Cudnik MT, Sayre M, et al. Identification of high-risk communities for unattended out-of-hospital cardiac arrests using GIS. J Community Health 2013;38:277–84. 10. Warden C, Cudnik MT, Sasson C, Schwartz G, Semple H. Poisson cluster analysis of cardiac arrest incidence in Columbus, Ohio. Prehosp Emerg Care 2012;16:338–46.
1517
11. Warden CR, Daya M, LeGrady LA. Using geographic information systems to evaluate cardiac arrest survival. Prehosp Emerg Care 2007;11:19–24. 12. McNally B, Robb R, Mehta M, et al. Out-of-hospital cardiac arrest surveillance—Cardiac Arrest Registry to Enhance Survival (CARES), United States, October 1, 2005–December 31, 2010. MMWR Surveill Summ 2011;60:1–19. 13. Jacobs I, Nadkarni V, Bahr J, et al. Cardiac arrest and cardiopulmonary resuscitation outcome reports: update and simplification of the Utstein templates for resuscitation registries: a statement for healthcare professionals from a task force of the International Liaison Committee on Resuscitation (American Heart Association, European Resuscitation Council, Australian Resuscitation Council, New Zealand Resuscitation Council, Heart and Stroke Foundation of Canada, InterAmerican Heart Foundation, Resuscitation Councils of Southern Africa). Circulation 2004;110:3385–97. 14. Sasson C, Meischke H, Abella BS, et al. Increasing cardiopulmonary resuscitation provision in communities with low bystander cardiopulmonary resuscitation rates: a science advisory from the American Heart Association for healthcare providers, policymakers, public health departments, and community leaders. Circulation 2013;127:1342–50. 15. Sasson C, Ramirez L, Medrano M, et al. Barriers to calling 911 and performing cardiopulmonary resuscitation (CPR) in High-risk neighborhoods in Denver. Poster presented at: the American Heart Association Scientific Sessions; November 2012; Los Angeles, CA. Abstract 283. 16. Swor RA, Jackson RE, Compton S, et al. Cardiac arrest in private locations: different strategies are needed to improve outcome. Resuscitation 2003;58:171–6. 17. Wissenberg M, Lippert FK, Folke F, et al. Association of national initiatives to improve cardiac arrest management with rates of bystander intervention and patient survival after out-of-hospital cardiac arrest. J Am Med Assoc 2013;310:1377–84.