ICU Communication Representation: Clinician-Clinician and Clinician-Computer Interactions

Objective: The aim of this study was to analyze clinical communication factors and interruptions and to develop clinician-clinician and clinician-computer knowledge representation models. Methods: An ICU observational study was combined with medical error reported cases to address the above questions. Researchers shadowed the ICU team, for 55 hours during patient rounds, to capture 6 main communication factors. Simultaneously, a systematic literature search was conducted to identify and extract reported medical error cases caused by clinical communication problem. The search included patient safety data banks, literature databases, newspaper, and reported lawsuits. Results: Out of 242 reported communication errors, 100 cases resulted in active errors while only 13 cases resulted in13 near misses; most of those errors were reported in journal articles (n = 302). As to the observation data, the most frequent communicator during ICU patient rounds was the Attending Physicians. The ratio of interruptions caused by clinicians to technology-aided devices was 3:1 per patient visit. The mean frequency of an Attending Physician interacting with a computer was once per patient visit. Analyzing data from both sources, two communication models representing the clinical communication framework were developed. Conclusion: Clinical communication is essential for effective health care delivery and for improved care outcomes. To further understand clinical communication, primary and secondary data were collected and analyzed and as a result, clinician-clinician and clinician-computer interaction models were developed to put into perspective the various factors involved in the communication process among care givers.


Current communication models
Over the course of many centuries, humans have realized the necessity of communication in order to accomplish most affairs, if not all. In the past, the nature of communication was relatively straight forward; it included direct communication between people. This can be shown in most communication models such as the models of Shannon, Berlo, and Shamm. Those well-constructed models represent the basic methods of communication and do not include new communication mediums. We studied the strengths and limitations of each of the general models then, we measured the relevance of those models against our first clinical communication model [13] . Results showed that there is a gap between the design of previously proposed non-clinical models and typical clinical communication scenarios. Therefore, it is time for ICUs to have their own communication model that represents the framework in which clinicians interact and collaborate.

Methods
In order to develop an ICU communication model that can accurately represent communication events, there is a need to build a knowledge base that encompasses most communication patterns and habits. For that reason, we believe that there is a necessity to collect communication events from two main sources: (1) literature, and (2) observation. Our data collection plan combined previously reported communication failure cases and data generated from our observational study.

Reported cases
The Veterans Health Administration and the Department of Defense are the only two organizations that mandate error reporting [22]. So, error cases were collected from the Agency for Healthcare Research and Quality (AHRQ), literature databases, such as Ovid, newspaper and magazine articles, and lawsuits. We utilized the wellstructured and organized feature of AHRQ called Patient Safety Net (PSNet). PSNet is a national web-based resource that provides the latest statistics and essential resources on patient safety. In order to retrieve all communication cases reported in PSNet, we searched for "medical error cases", which in return, provided 723 reported cases. Then, we refined our research to only include all cases labeled with "communication improvement", the returned number of cases where 668 of which 414 were labeled "communication between providers". Those cases were gathered from various sources and some of the cases were labeled by error types; Table 1 shows the distribution of cases.
Among the reported error incidents, a case reported at the Pennsylvania Patient Safety Advisory report, which discusses how to improve telephone and verbal orders. A nurse received a verbal order from the physician for Zosyn. The patient had documented that they are allergic to penicillin; however, both clinicians were unaware that Zosyn is derived from penicillin. The pharmacy staff was out of the office for the day and the medication was obtained from the night cabinet; two doses were administered to the patient. The next day, the pharmacy notified nursing that Zosyn is prohibited for this patient and the medication was discontinued [19]. This incident shows the importance of feedback, Decision Support Systems (DSS), education, and other communication factors in order to avoid such medication errors. It is labor intense to locate and analyze each error case because of the absence of a unified structure to all the cases. In other words, the ability to utilize Natural Language Processing (NLP) algorithms remains modest, until the cases are formatted in a computer-readable structure.

Observational study
In order to develop a comprehensive ICU communication model, the analysis of reported error cases is not sufficient. We believe that conducting an observational study at the ICU will further strengthen our findings that will be included in the model. Therefore, we conducted a research study in critical units at the307-bed University of Missouri Hospital; which in 2011, according to American Hospital Association, had 35,671 emergency room visits and performed 6,284 inpatient surgeries [1] . This study does not aim at capturing ICU medical errors; rather, we aim at capturing communication patterns and behaviors that can potentially be the cause of medical errors.
The study was conducted by two researchers/observers, who are under health informatics training, in order to increase the validity of data captured. The study design included shadowing 3 attending physicians and their clinical team during patient rounds for two weeks per attending physician. The focus of the study is to observe and understand the communication between the attending physician and their team. During the two-week period, the researchers shadowed the team 6 times, as a repeated measure, with the following distribution: This distribution was suggested by ICU domain experts as well as by our research biostatistician. From a clinical point of view, the first and last day of the two-week period will capture the communication patterns during a chaotic first day and then, the more organized Furthermore, we captured the frequency of interruptions that occurred during rounds. We categorize the types of interruption into two categories: (1) clinician-related interruptions, and (2) technologyaided devices interruptions. We found that both types of interruptions occurred in almost all patient visits however, interruptions caused by clinicians occurred approximately 3 times during a patient visit while interruptions from technological devices occurred only once.

Clinician-clinician communication
While interacting, clinicians constantly exchange information and knowledge; this rather complicated process involves two or more communicators, communication factors, and communication events. A communicator is a clinician who interacts with other clinicians by sending or receiving clinical information. Communication factors refer to tacit and explicit knowledge that affect the way a communicator formulates or interprets a message. Communication activities represent clinician's behaviors and technological instances that have an impact on communication. Figure 1 represents the clinician-clinician communication model where a clinician communicates with one or more fellow clinicians. The model shows communication factors and activities that were reported in literature and were observed during the observational study.
The communication factors form a knowledge base that shapes the way a clinician interacts with peers. Continuing our work towards further understanding factors that impact communication [13], there are six communication factors that are responsible for how a clinician creates and perceives a message. The level of education and training, and the years of experience a clinician acquired are considered essential factors in the way clinicians exchange information. The previously mentioned factors, when at similar levels, facilitate inter and intracommunication among clinicians since with more education and experience clinicians self-learn how to effectively communicate with their peers.
Considering a clinician's culture is key to reaching high levels of communication effectiveness. Initially, we referred to culture as the background and tradition of a clinician. For example, in some international cultures to agree to a statement the person shakes their head sideways; while in the U.S. A person nods their head to indicate acceptance or agreement. Based on experience, we added a new dimension to the definition of culture, which is the variance in clinical backgrounds and practices among clinical specialties. Finally, we have observed that the overwhelming flow of information and knowledge combined with long working shifts can cause an incorrect recall of events or information. For that reason, cognitive psychology plays an important role during clinician-clinician interaction.
Moving on to communication activities that occur during patient visits at the ICU, we have observed the occurrence of 6 main activities. Even though it is not frequent, multitasking by a receiver usually leads to the sender repeating the message, mostly when the sender requests feedback from the receiver. The other activities are considered as interruptions, which we define as the interference to an ongoing conversation resulting in a pause or an end to the conversation. Clinician interruptions occur when a question interrupts the conversation, or if team members who are not participating in the conversation engage in a side conversation. External interruptions refer to questions or statements from non-clinicians, such as a patient's family. During patient rounds, pagers and system alerts constituted most interruptions in the technology-aided devices category; nevertheless, their benefits cannot be understated towards patient safety and clinical workflow.

Clinician-computer interaction
Computer interactions are an undivided part of the overall communication process. The utilization of computer systems is clearly present in each patient visitation, figure 2 represents the ICU Clinician-Computer Interaction; the model discussed clinical communication from two perspectives, the clinician and computer. During our observation, we summarized the usability of clinicians to computer systems into storing and retrieving patient information, request new tasks, viewing various types of imaging. In order for a clinician to effectively interact with a computer system, the clinician must receive adequate system training. We define adequate training as  the level of which a clinician can fully and correctly use the available system functionalities. Clinicians should be able to navigate the system efficiently by finding the intended information in the shortest and quickest route. Moreover, different clinical systems provide different data interpretations, and clinicians should be able to correctly interpret those formats in a way that is least confusing. Through an effective training, the cognitive gap between system capability and users' expectation should be minimized. Furthermore, in the case of unusual system behavior, it is necessary that clinicians provide problem solving skills. For instance, in the case of error alerts to users, clinicians avoid delays by working around minor or temporary errors either through the system or in person. Also, in the rare instance of fatal errors, such as computer crashing, clinicians should have problem solving skills that would provide an alternate solution to the situation until the system is up again.
On the other hand, implementing a computer system in the ICU should focus on three main areas: (1) graphical user interface, (2) efficiency, (3) data representation. A user friendly interface can be reached through simple design and consistent navigation. Clinician's opinion on the interface must be fully considered when developing the system, their opinion on the design and navigation is important; however, their similarly important opinion on the functionality and effectiveness must not be left out. Furthermore, an efficient data storing and retrieving process that is error-free and easy to use is evident in critical units. Due to the labor intensive environment, clinicians are more susceptible to typos while requesting a patient's medical record; hence, system feedback is an essential error-checking procedure prior to executing a command. Also, standard terminology throughout the system is necessary to avoid inconsistencies and confusion. Finally, meaningful and reliable data representation is necessary for effective clinician-computer interaction. Visual representations are easier to read and interpret than many numeric values. Also, it is necessary for the system to represent real-time, accurate patient data, and high imaging quality.

Discussion
The primary purpose of this study research is to use qualitative approach to investigate communication events, interruptions, and interactions that the clinical team experienced within the context of critical care. The data collected through reported cases and observation resulted in 3 proximate outcomes that assist further understanding of the problem: 1) frequency of occurrence of instances, 2) representation of clinician-clinician interaction, 3) representation of cliniciancomputer interaction.

Communication instances frequency
We identified 6 main communication factors that repeatedly occurred during team communication with the Attending Physician: 1) patient information conveyed to the Attending, 2) feedback provided to the Attending, 3) frequency of communication done by the Attending, 4) interruptions by the team, 5) interruptions by technological devices, 6) Attending's interaction with computers. Table 2 shows that the majority of communication done in the ICU is done for the purpose of either giving or receiving information, which agrees with the findings of Coiera and Moss [7,17] with regards to the operating room and emergency departments. Communication done by the Attending to the team members was more than double the communication done by the team members to the Attending, which shows that more information is being transferred from the Attending which is logical due to the role and responsibilities of the Attending towards the patients and the team. Moreover, the Attending provided feedback, verbal and nonverbal, 1.5 more times than the communication instances they received from the team. This shows the persistency of the Attending to acknowledge and confirm the messages they received and also, an indication to the team to follow the same habit of clearly showing their understanding of a given statement. A significant number of interruptions were observed during communication. The frequency of an interruption caused by clinician  was approximately triple the frequency of an interruption caused by technology-aided devices.This ratio agrees with the findings of Patel at al., which suggests that human interruptions are usually twice as frequent to technological interruptions [4] .From observation, we identify that human interruptions, even though more frequent, can be controlled by increase in awareness and training, the reason for this belief is that all interruption variables, such as the interrupter, are present in the room. However, interruptions caused by technologies, such as telephones and pagers, might be harder to control since the interrupter is not on-site; nevertheless, options such as putting personal cellphones on silent could minimize this frequency.

Clinicians interaction
When clinicians communicate numerous factors are taken into consideration which makes the communication process complicated. We identified that communication among clinicians is affected tacit knowledge and external activity. The way a clinician was trained, the level of education, and the years of experience shape how they formulate or perceive a message. During the study we observed that clinicians with more of the previously mentioned factors can more accurately articulate their messages. Similarly, there is a language factor which represents two aspects: The first is good use of English for non-native speakers, and secondly, the use of standardized clinical terminology during communication. This agrees with the findings of a study surveying 64 members of the National Association of School Nurses, which suggested that the use of standardized terminology among nurses reduces symptoms after intervention, and enhances patient safety [5].
External factors seem to limit the communication process rather than facilitate. The frequent occurrence of side conversations, pager and computer alerts, and multitasking presents a disruption to the ongoing conversation and the result was a request to repeat, or a question aimed at continuing the conversation. We also observed team members multitasking during communication, while the justification is understood, the consequence of multitasking can range from mishearing to executing the wrong order and hence, there is a higher chance for medical errors.

Clinicians and computers
While shadowing the clinical team, several rare instances of human-computer interactions occurred. When reviewing the latest X-Ray for an ICU patient, the image was hard to read and interpret and the Attending reported that the quality of imaging was of fair quality and better representation and quality is needed. Another instance, during patient rounds the Attending requested the medical record of the patient to be retrieved from the system, upon retrieval, the resident notified the Attending of the patient information; however, the Attending realized that the information is incorrect. The resident incorrectly typed the wrong information and the system retrieved the incorrect medical record.
When representing the communication process between clinicians and computers, there are two dimensions to highlight: 1) the user, 2) the computer system. Users must have comprehensive understanding of the system, including usability, and correctly storing and retrieving information, and problem solving skills. As for computer systems, the most important feature is to design a clinician-centered system that will provide convenient design and functionality that suits the needs of clinicians.

Future direction and limitations
Studying ICU communication is a tedious and labor intense activity, and affording adequate human resources to study this important phenomenon is a major limitation to this study. In this study we focused on the Attending physician, since they are at the top of the hierarchy; however, in future work we aim to study other clinical roles such as Fellows, Residents, and Registered Nurses etc. Another limitation of the study is that it was carried out during morning rounds, which are communication intense and interaction diverse;however, there is a need to broaden this research to include other sessions such as handoff sessions and morning meetings, which will be included in future studies.

Conclusion
Without a doubt, the significance of communication in health care is pivotal with regards to better care and enhanced patient safety. In this article, we further studied clinical communication by analyzingdata from literature and by observation. The focus of the study was to further understand key communication factors and activities that occur during conversation. Based on reported error cases and observation data, we proposed two ICU clinical communication models with a focus on Clinician-Clinician and Clinician-Computer interactions. This initial attempt, to our best knowledge, to represent ICU clinical communication is a significant step towards an ultimate goal of this research of building an exhaustive clinical communication ontology, which is consistent with our early work [9] . Along with the communication models, the ontology will serve as an educational tool for clinicians, and we aim to utilize it in medical error reporting system in order to increase the quality of reported error cases. Our efforts aim at reaching better understanding of the clinical communication framework in order to decrease medical errors, enhance healthcare quality, and thus, improving patient safety.