Data were obtained from the ROC Cardiac Arrest Epidemiological Registry or “Epistry” which has been described in detail previously [12
]. This registry is a prospective database of all persons evaluated for OOHCA by participating EMS agencies. Over 250 EMS agencies at sites in Canada and the US contribute data on an ongoing basis. Multiple mechanisms for surveillance are used at local sites to assure investigators of capture of all appropriate cases. Data are abstracted from EMS records and hospital records by data coordinators who use common data definitions. Data are transmitted to a central coordinating center by web-entry into an electronic database or batch upload of multiple cases grouped together. Within-form error checks are used to decrease data entry errors.
Subjects included were OOHCA patients of any age who received hospital care. “Cardiac arrest” was defined as receiving either (1) chest compressions by professional responder (paramedic, first-responder or other health professional arriving as part of the organized EMS response), or (2) rescue shock by professional responder or a bystander using an automated external defibrillator (AED). Cardiac arrest cases that were associated with trauma were excluded. “Received hospital care” was defined as either (1) arrived at the hospital with pulses, or (2) arrived at the hospital without documented pulses, but was discharged or died > 1 day later (i.e. was not pronounced dead on arrival at the hospital). When the subject was delivered to one hospital, and transferred to a second hospital on the same day, the second hospital was considered to be the treating hospital. These definitions were selected in order to exclude cases of EMS-treated OOHCA who never had restoration of pulses, and therefore who were not exposed to in-hospital post-resuscitation care. Subjects were entered into the database between December 1, 2005 and July 1, 2007, a time-period that excludes any interventional ROC trials.
Hospital characteristics were obtained from three sources. First, the EMS Structures Survey collected standardized data from each ROC site completed by local investigators who described the local hospitals that receive patients from their EMS systems [13
]. Second, the American Hospital Directory is a publicly available database on US hospitals that is collated from a variety of public and payor databases (Medicare, licensed beds) (www.ahd.com
). Third, the 2006–2007 Guide to Canadian Healthcare Facilities is a collation of data on Canadian hospitals provided by member hospitals (Volume 14, 2006–2007, www.cha.ca
). Investigators examined the data from all three sources for each hospital that received subjects in the Epistry database in order to collate hospital characteristics. In the event of discrepancies, the report of local investigators who were familiar with the local hospital was used.
Hospitals were categorized based on several characteristics. Number of beds was used to categorize hospitals as large (>400 beds), medium (251–400 beds), and small (<250 beds). Capacity to perform acute cardiovascular interventions was defined as the presence of a cardiac catheterization laboratory, but the databases did not distinguish between capacity to perform emergent and elective catheterization or PCI. Trauma center designation was categorized as Level 1, Level 2, or non-trauma center. A teaching hospital was defined as one that listed active residency programs based at that hospital.
Patient and arrest characteristics were tabulated with descriptive statistics and compared between the number of hospital beds (size category) and cardiac catheterization capabilities of hospitals. Sample sizes within cardiac catheterization designation were similar across the different hospital size categories. Chi-Square tests examined the independent association of hospital size category and catheterization capability with the primary outcome, survival to hospital discharge. Hospital length of stay was considered as a secondary outcome.
To account for known potential confounders, we used multiple logistic regression to examine the association of hospital type and survival to hospital discharge. We adjusted for variables previously associated with outcome: site; hospital characteristics including trauma level designation (1,2, non-trauma), teaching institution; patient characteristics including gender, age (<1, 1–11, 12–19, 20–39, 40–60, 61–75, >75, unknown), witnessed collapse (witnessed, unwitnessed, or unknown), initial ECG rhythm (VT/VF, PEA, asystole, AED no-shock, or cannot determine), bystander attempts at cardiopulmonary resuscitation (CPR), and EMS process characteristics including response time (time from call to dispatch to first EMS vehicle arrival at scene). Variables were considered significantly associated with the outcome if p < 0.05.
In order to describe the relationship between hospital volume of OOHCA patients and outcome, we plotted survival versus annualized number of cardiac arrests in the database for all hospitals. To avoid bias against hospitals with small volumes of cardiac arrest patients (<10 post-cardiac arrest patients a year), this analysis included all hospitals, regardless of size.
Analyses of data were performed in S-Plus, v 6.2 (TIBCO Software, Palo Alto CA) or STATA v9.1 (StataCorp LP, College Station TX).