April 1, 2020
- Describes the development of the CAFE study and its role in determining variation in trauma patient care
- Outlines studies that identify significant variation in patient outcomes among trauma centers categorized at the same level
- Highlights the evolution of OOSTT and its role in annotating and tagging data
Trauma is the leading cause of death for individuals younger than 45 years old, and unintentional injury is the third leading cause of death across all age groups in the U.S.1,2 However, millions of patients in this country survive trauma every year and often face lifelong physical, mental, and financial problems. According to the U.S. Centers for Disease Control and Prevention, in 2013, the total lifetime medical and work loss costs of injuries and violence in the U.S. was $671 billion.3
In an effort to gain additional insight into the provision of trauma care in the U.S., this article focuses on the National Institutes of Health (NIH)-funded study titled Comparative Assessment Framework for Environments of Trauma Care (CAFE), which is currently under way. The study is designed to assess the organizational attributes of trauma centers and trauma systems and determine the possible causes of variations in trauma patient care. This article reviews the development of U.S. trauma systems, outlines the purposes and structure of CAFE, and explains how trauma centers can contribute to this important work.
Trauma care has improved significantly over the last 50 years, often under the leadership of the American College of Surgeons (ACS) and its Committee on Trauma (COT).4 The first two trauma centers were established in 1966 at Cook County Hospital, now the John H. Stroger, Jr. Hospital of Cook County, Chicago, IL, and San Francisco General Hospital, CA. In 1969, R. Adams Cowley, MD, FACS, created the Maryland system of trauma care.5
In 1976, the ACS COT published the first edition of Optimal Hospital Resources for Care of the Seriously Injured, now known as Resources for Optimal Care of the Injured Patient.7,8 This manual has served as a guidebook for trauma centers and trauma systems and established the categorization of trauma centers as Level I, II, III, or IV based on minimum standards for each level of care. In the late 1980s, the ACS COT created the Verification Review Committee to conduct onsite consultations and verifications for hospitals and regions seeking to establish or verify trauma centers or systems.4 During the verification review process, a trauma center is assessed on criteria outlined in Resources for Optimal Care of the Injured Patient, including volumes of severely injured patients, 24-hour availability of trauma surgeons and other specialists, surgical capabilities, and availability of specialized equipment. Based on the review, the trauma center is categorized at the appropriate level.
Regional trauma systems apply a coordinated approach to trauma care for patients in a specific geographic area and are predicated on the idea that trauma patients receive better care when processes and resources—including prehospital care and transportation, acute hospital care, and rehabilitation—are organized as a larger system. The ultimate goal is for facilities to have an identifiable role based on resources and community needs and for trauma centers at various levels to cooperate to ensure optimal care and efficient use of resources.8
This goal is achieved through the concept of definitive care facilities. A facility designated at a certain level has the resources to provide definitive care for a certain severity of injury. It is anticipated that quality of care should be similar among all levels of trauma centers; the main distinction is the severity and volume of injured patients because of the availability of resources.
Both trauma systems and individual trauma centers have been shown to improve mortality rates, quality of care, and patient outcomes in general; however, significant variation remains in patient outcomes among trauma centers of the same level, as the following studies demonstrate. Shafi and colleagues classified patients from 47 ACS-verified Level I trauma centers into three injury severity groups and compared their survival rates. They discovered that five centers had significantly worse survival rates for mild injuries and that the percentages of outcome disparities increased as injury severity increased.9 Cudnik and colleagues conducted a similar study to assess differences in mortality rates at the 10 Level I trauma centers in Ohio, adjusting for patient demographics, injury severity, mechanism of injury, and emergency medical services (EMS) and emergency department (ED) procedures. The authors concluded that both the adjusted and unadjusted mortality rates of patients varied greatly.10 Glance and colleagues reached a similar conclusion after reviewing submissions from 125 hospitals to the National Trauma Data Bank in 2006. After adjusting for injury severity, the authors determined that patients admitted to the worst-performing hospitals were at least 50 percent more likely to die than patients admitted to the average hospital.11
It is important to examine what contributing factors account for these differences in patient outcomes among trauma centers, particularly those classified at the same level. We hope to better understand these differences by studying the structures and processes of trauma centers and systems. Little is known about the structures and processes that lead to optimal care for the injured patient, although the role of structure in influencing performance of health care facilities is widely recognized.12 Donabedian’s structure-process-outcome model is a widely accepted theoretical framework for identifying quality measures in health care. In this framework, structural indicators are defined as the staffing patterns and organizational resources associated with providing care (for example, facility operating capacities); process indicators refer to actions that are performed on patients (such as medical or surgical procedures); and outcome indicators are the states that result from care processes (including increases in quality of life and decreases in mortality rates).13-15 Good structures increase the likelihood of sound processes, and sound processes increase the likelihood of favorable outcomes; good structure also can directly improve outcomes.16 Often, no direct relationship is evident between structure, processes, and the eventual outcomes; rather, the “interrelationship of structure and process, as well as individual patient characteristics, dictate the final outcome.”17
CAFE seeks to address the importance of structure in influencing organizational performance and the difficulties in understanding structural differences and similarities across trauma centers and systems. Until now, it could be difficult to acquire knowledge and comparable data about the organizational structures of trauma centers and systems for the purpose of external and internal evaluation or comparative research.
CAFE, which started in 2015, is designed to collect detailed information on the organizational attributes of trauma centers and trauma systems using an electronic system. The project uses a web-based service that permits trauma center representatives to submit data about their trauma center or trauma system and to conduct an anonymous self-assessment of their institutions.
The first step we took to understand the differences in organizational structures of trauma centers and systems was to review the language of the domain. Though the development of trauma center standards has led to a consistent and reusable terminology, a formal or controlled vocabulary to describe the structure of trauma centers or systems had yet to be developed, hindering the acquisition of structured data and the comparison of organizational structures and procedures across institutions.
We began with a domain analysis, which relied on a group of subject matter experts (SMEs) with substantial experience in trauma care and management and assessment of trauma programs, trauma centers, and trauma systems. A domain analysis for the purpose of developing a controlled vocabulary involves identifying the relevant terms to represent the domain of interest and providing definitions for those terms. Including the ACS COT in this process helped ensure that the definitions and web service used to complete the assessment met the needs of the trauma center and trauma system management community.
We initially used natural language processing to extract lists of relevant terms from published abstracts pertaining to trauma care. We then used software to analyze the relationships among the terms and began holding regular teleconferences with the CAFE SMEs to discuss definitions. The SMEs used a multistep process to review material from the ACS COT, particularly the prereview questionnaires (PRQs), and added information from their own experience to develop a set of 142 terms and provide user-centered definitions. The investigators then created formal definitions, which we assessed via a survey of 75 members of the ACS COT who were not participants in the expert group. Each survey respondent was asked to provide feedback on 20 terms. We used the responses and comments to correct the issues raised by the SMEs.18
These terms and their corresponding definitions became the basis for the Ontology of Organizational Structures of Trauma systems and Trauma centers (OOSTT).19 An ontology is a standardized representational framework that provides a set of terms for the consistent description (annotation or tagging) of data and information.20 Having a standardized ontology promotes greater consistency in data description; as such, all terms have both textual (for use by humans) and logical (for use by computers) definitions. Ontologies can be represented on a graph where terms are the nodes and the ontological relations between the terms are edges.20 Ultimately, an ontology can help limit complexity and turn data into information and knowledge.
Our ontology development was guided by and validated with competency questions. The SMEs, who represent potential users, created the competency questions based on user scenarios, which took the form of queries, such as the following:
After we released the initial ontology, we populated it with virtual data and reformulated the competency questions as queries in description logic—a family of formal knowledge representation languages. We ran these queries against the populated ontology as a means of validation. The recall and precision metrics of each query showed that we succeeded in formally representing the entities relevant to the use cases.
OOSTT is the backbone of the open-source, online questionnaire used to gather anonymized data about the organizational structures and processes of trauma centers and systems. The questions are based on a hospital PRQ for Level I and Level II trauma centers developed by the Arkansas Department of Health. To ensure that respondents are using the domain terms consistently with our definitions, the OOSTT definition is displayed whenever users hover over a term.
Annotating the collected data with the ontology enables the use of automatic inference. A major goal of the project is to infer which institutions follow the ACS COT recommendations outlined in Resources for Optimal Care of the Injured Patient. For example, the recommended requirements for the TMD differ according to the trauma center level. The OOSTT implementation can represent the role of a TMD compliant with ACS COT recommendations and then use reasoning to infer which institutions satisfactorily meet those recommendations. Though the data are deidentified, it will still be possible to retrieve a count of compliant institutions.
A key function of the ontology is to provide trauma centers and systems with the ability to complete self-assessments. As respondents complete the questionnaire, their answers generate a comparative graphic representation of organizational structures that quickly displays the differences between a respondent’s organization and the other organizations where data has already been gathered. Such visualization enables users to quickly assess which components and relations are missing in their institutions. It also can highlight innovative areas of their trauma centers.
We anticipate that these visualizations will be of great value to representatives of trauma centers and systems, especially those that lack access to the ACS COT or similar consultation and verification bodies because of limited resources or other barriers. For the first time, these institutions will have an easy way to optimize organizational structures consistent with those approved by the ACS COT and minimize unapproved structures without undergoing a formal consultation.
Our final efforts on this project will be to conduct a usability study of the online questionnaire. In the coming months, we will solicit feedback from the ACS COT and potential users working in trauma centers and systems through a quantitative assessment of the user interface. We will evaluate effectiveness, efficiency, error tolerance, ease of learning, and engagement.
Another goal of this project is to provide tools to help other researchers conduct comparative research on the organizational structures at trauma care centers and systems. We are working to develop a suite of query tools that will enable users to find and export data from the system. To facilitate data analysis, we are creating web-based tools that will allow researchers to develop and run ad hoc queries and will return results as either a graph or a table. The web service will provide the option for users to save data output to a personal history, allowing them to run a series of comparisons and retain the results. Users who prefer a visual style will have a graph-and-edge query tool that uses mouse clicks to choose relevant terms and relations and that can add string-based filtering where desired. For users who prefer a text-based interface, we will develop a second query tool that will use standard elements, such as drop-down fields, check boxes, and buttons, and that will enable complex textual queries to be built from standard logical query functions; that is, some intersection or union. These tools also will undergo usability testing.
We anticipate that at this juncture, CAFE provides an infrastructure for assessing organizational structures in trauma systems and trauma centers that contribute to better, more consistent patient care. For more information on how your trauma center can participate in the CAFE study, contact Dr. Brochhausen at email@example.com.