Quality and Safety of Initial Management for ST-segment Elevation Myocardial Infarction

Objectives

The broad objectives of this multicenter and multidisciplinary observational cohort study are to investigate patient, provider, and system-related factors that are associated with the quality and safety of initial management for suspected ST-segment elevation myocardial infarction (STEMI) evolving for less than 12 hours in daily practice. The findings of this study should help identifying areas with room for improvement in clinical pathway for STEMI patients. The specific aims of this project are:

- to investigate the independent associations between the delivery of timely acute reperfusion therapy and patient, provider, and system-related factors, according to current guidelines

- to assess the independent associations between the delivery of timely acute reperfusion therapy and in-hospital clinical outcomes (including all-cause mortality and major adverse cardiovascular events)

- to determine the independent risk factors for major bleeding (excluding coronary-artery bypass graft-related bleeding events) among baseline patient and provider characteristics, processes of care, and system-related factors, respectively

- to elucidate the prevalence, risk factors, and in-hospital clinical outcomes independently associated with false-positive cardiac catheterization laboratory activation or inadvertent fibrinolytic therapy

- to determine how hospitals are interconnected by transferring patients with suspected STEMI

Hypotheses The primary hypotheses guiding this project are that delayed reperfusion therapy for suspected STEMI independently relates to provider practice patterns and system barriers. It is further postulated that non-compliance with target delays in implementing primary PCI or fibrinolytic therapy is associated with worse in-hospital clinical outcomes.

The secondary hypotheses are that the delivery of acute reperfusion therapy within target delays is associated with increased rates of false-positive cardiac catheterization laboratory activation, inadvertent fibrinolytic therapy, and bleeding events.

Participating study centers and setting The project will be conducted at 3 emergency medical services (EMS/SAMU) and 23 public and private (for-profit and non-for-profit) acute care hospitals in Northern Alps in France. Northern Alps are a predominantly mountainous area covering 15,000 km² with an estimated population of 1,860,000 inhabitants with large seasonal variations due to tourism. The median distance and driving time from each community hospital to the closest PCI center are 63 km (range, 4.5-132 km) and 43 min (range, 10-88 min), respectively.

As part of a coordinated regional system of reperfusion therapy for STEMI, all participating centers have a long collaborative tradition in STEMI management research.The rationale, data collection methods, verification procedures, and primary outcomes of this registry have been reported in detail elsewhere.

Baseline data collection methods

1. Prospective data collection As part of the ongoing registry, attending physicians or clinical research technicians routinely abstract summary information on demographics, presenting characteristics (including electrocardiographic findings), coronary angiography findings, and delivery and timing of acute reperfusion therapy in both prehospital and hospital settings.

Based on patient home address, travel time and travel distance to the nearest community hospital and PCI-capable hospital will be obtained through calls to commercial Web page (MapQuest or Via Michelin). ZIP code will replace patient home address when unavailable. Previous studies have shown that this method provides reasonable estimates. To account for peak travel time, slowdowns of 20% will be factored in during the hours of 7 to 9 AM and 4 to 7 PM, Monday through Friday.

2. Retrospective data collection Two clinical research nurses will perform retrospective chart review using a computerized data collection instrument. The following variables will be recorded: baseline patient characteristics, clinical variables, comorbid conditions, laboratory variables, procedural interventions (cardiac catheterization, PCI [not primary], and coronary artery bypass grafting) at any time during index hospitalization, medications (on admission, during the first 24 hours, and at discharge). Index hospitalization includes both admitting and transfer hospital stays.

Characteristics will also be collected for attending physicians (gender, age, discipline) as well as admitting and transfer hospitals (community vs PCI-capable facility, size, public or non for profit versus for profit). Consistent with a previous study, PCI centers are defined as hospitals that offer emergency PCI 24 h a day, 7 days a week.

Data management

1. Prospective data collection As part of the ongoing registry, data are prospectively recorded in duplicate on standardized forms designed to facilitate data collection and entry. All completed forms are forwarded to the core coordinating facility of the regional emergency system (RENAU=Réseau Nord-Alpin des Urgences ; www.renau.org), where the data are entered electronically (single-data entry) and maintained on a local computer network. The data entry screens are similar to the paper forms. Original data collection forms are kept in locked file cabinets.

Data integrity is enforced through two complementary approaches:

- Routine edit checks (valid values, range checks, and consistency checks) are performed at the time of data entry by a clinical research technician at the core coordinating facility.

- The registry statistician periodically screens data for additional errors using programs designed to detect missing values or inconsistencies.

2. Retrospective data collection To ensure optimal quality, all data collected retrospectively by chart review will be entered electronically by clinical research nurses using a personal identification code and password-protected web-based data collection system. The clinical research nurses will receive formal training in the methods of data abstraction and recording. An operation manual that includes definitions and acceptable data sources for all variables will be distributed. Reliability of data abstraction will be assessed by randomly selecting 30 cases for independent collection by the two study nurses.

The data manager will screen data for additional errors using programs designed to detect missing values or inconsistencies.

Statistical analysis

1. Overview Standard descriptive summary statistics will be used for reporting continuous and categorical variables. In univariate analysis, inferential comparisons will be performed using analysis of variance or the Kruskal-Wallis rank sum test. Logarithm transformation may be used for highly skewed variables prior to inferential comparisons. Categorical variables will be analyzed with the use of Khi² or Fisher exact tests. Trends across ordered groups will be analyzed for statistical significance using the Khi² test for trends for categorical variables and the non-parametric Wilcoxon-type test for trends for continuous variables, respectively. Two-sided P values of less than 0.05 will be considered statistically significant.

2. Primary effectiveness outcome A non-parsimonious multivariable logistic regression model will be developed to identify baseline characteristics that are independently associated with timely acute reperfusion therapy. Covariates entered into the multivariable model will be prespecified before modelling, without investigating their association with the primary study outcome.

Secondary analysis restricted to the subset of patients with prehospital management will use multilevel logistic regression, with the 3 levels defined by patients, mobile intensive care units, and hospitals. The Metropolis-Hastings algorithm (a Markov Chain Monte Carlo method) will be used to account for the cross-classification of mobile intensive care units that transported patients at more than one study site.

3. Secondary medical outcomes Because timely acute reperfusion therapy is not randomly assigned in this observational study, unadjusted comparisons of medical outcomes between study groups might be confounded by imbalances in baseline characteristics.For this purpose, propensity score analysis that compensate for differences in measured baseline characteristics will be performed between timely acute reperfusion therapy recipients and non-recipients.

Practically, a propensity score will be derived for the receipt of timely acute reperfusion therapy using a full non-parsimonious logistic regression model that includes baseline characteristics as covariates.

Each patient will be assigned a propensity score, which may range from 0.00 to 1.00 and reflects the conditional probability of timely acute reperfusion receipt given his baseline characteristics.

To compare secondary medical outcomes among patients with similar conditional probability of timely acute reperfusion therapy, a cohort of timely acute reperfusion therapy recipients and non-recipients matched by propensity score will be also defined. For this purpose, an algorithm will be used to match each timely acute reperfusion therapy recipient to a single non-recipient who have the nearest propensity score within one digit. If this cannot not be done, that timely acute reperfusion therapy recipient will be excluded from the propensity-matched analysis. The propensity-matched cohort will be evaluated for significant residual imbalances in baseline characteristics. Then logistic regression will be used to estimate the odds ratio for each secondary medical outcomes associated with the receipt of timely acute reperfusion therapy among propensity score-matched patients.

4. Secondary effectiveness outcomes Using the same modelling approach as previously described (i.e., b. primary effectiveness outcome), multivariable logistic regression will be performed to identify baseline characteristics that are independently associated with false-positive cardiac catheterization laboratory activation among patients who are candidate for primary PCI. A first-order interaction term involving prehospital cardiac catheterization laboratory activation will be tested for statistical significance. If appropriate, analysis will be stratified according to prehospital versus hospital cardiac catheterization laboratory activation. Then medical outcomes will be compared between false positive and true positive cardiac catheterization laboratory activation after adjusting for quintile of propensity score.

In a separate analysis, multivariate logistic regression will be performed to identify baseline characteristics that are independently associated with inadvertent fibrinolytic therapy among fibrinolysis recipients. A first-order interaction term involving prehospital fibrinolysis will be tested for statistical significance and, if appropriate, analysis will be stratified according to the receipt of prehospital versus hospital fibrinolysis. Then odds ratio will be estimated for each medical outcome associated with inadvertent fibrinolysis after adjusting for quintile of propensity score.

Harms and safety In this retrospective observational study, severe adverse events occurring during hospitalization will be specified as secondary medical study outcomes (mortality, major adverse cardiac events [reinfarction, cardiogenic shock, cardiac arrest, heart failure, stroke], and major bleeding).

Other prespecified adverse events will be recorded, including atrial fibrillation or flutter, sustained ventricular tachycardia, atrioventricular block (Mobitz II or 3rd degree), mechanical complications (ventricular septal defect, mitral regurgitation, free wall rupture), venous thromboembolism (deep vein thrombosis and/or pulmonary embolism), acute renal failure, thrombocytopenia and heparin-induced thrombocytopenia, and non-major bleeding events.

Adverse events will be collected through a structured chart review. There will be no attempt to determine potential causality.