4.1 General strategy
The basis of our review will be Arksey and O'Malley framework [10] that consists of the next six stages: 1) identifying research questions, 2) identifying relevant studies, 3) study selection, 4) charting the data, 5) collating, summarising and reporting the results, and 6) consultation with stakeholders. We incorporate posterior refinements suggested by Levac et al. [11] and Peters et al. [12, 13]. Specifically, we include: a) a rationale (Section 2), b) the population, context, and concept (PCC) elements (Table 1), c) an ensemble of a multidisciplinary review team that consists of roboticist engineers and a physician with extensive knowledge in telepresence robots, and d) a flowchart that describes the search decision process based on the PRISMA-ScR [14] statement.
Table 1
Definition of our target population, concept, and context.
Criteria
|
Description
|
Population
|
Patients (up to 79 years old) hospitalized in an ICU, and health professionals and care-givers that use telepresence robot systems to attend patients.
|
Concept
|
Design decisions in MRTS that affects medical care to patients entered into ICU.
|
Context
|
Studies written in English and Spanish within all geographical and sociocultural conditions are considered.
|
4.2 Identifying research questions
After discussion by e-mail and videoconferencing, our team identified the following research questions:
-
RQ1: Which are the design features that influence robotic telepresence-based processes in Intensive Care Units?
-
RQ2: Which are the factors and actors involved in robotic telepresence-based processes in Intensive Care Units?
-
RQ3: Which information about the patient and its environment is required for an adequate robotic telepresence-based intensive care?
-
RQ4: Which are the possible interactions between the robot and the different actors of an robotic telepresence-based ICU?
All authors approve this manuscript for submission to
4.3 Identifying relevant studies
Our search will focus primarily on formally published papers. For documents in English, we will search in the next electronic databases: ACM, BMC-Springer Nature, BMJ, BMJ Open, Cochrane Library, Ebscohost, IEEE Xplore, MEDLINE, Science Direct, Scopus, SpringerLink, PubMed, Web of Science, Hindawi, and Wiley Online Library. In Spanish, we will search SciELO, Redalyc, and Dialnet. We will search for grey literature as suggested by [15]. Our search in English will consider OpenGrey and Grey Literature Report, and Google Scholar, Google Dataset Search, and Google Web Search for searches in both English and Spanish.
The basic structure of our search string is described in Table 2. It is composed of three main concepts. For each concept we selected search words with the wildcard “*” to substitute zero or more characters whenever necessary. Only the asterisk wildcard is considered as it is common to many search engines. In English, we tested words with the wildcard using the online version of the Chambers Crosswords Dictionary and Thesaurus [16, 17]. In Spanish, we used the Diccionario de la Lengua Española of the Real Academia Española [18]. In addition, we carefully chose synonyms for some of the words. The form of our search string will be a conjunction of disjunctions as follows: we will associate search words of a main concept with OR connectors, and then the three groups will be joined together with AND connectors. Whenever possible, our search string will be restricted to titles, abstracts, and keywords.
4.4 Study selection
4.4.1 Inclusion criteria
We will consider any type of papers devoted to MRTS with wheeled robots applied in ICU. Papers must have been published in academic peer-reviewed journals or conferences, from 2000 to 2021, in either English or Spanish languages, and in medical or engineering disciplines. As grey literature, we will incorporate book chapters, official government documentation and technical reports not published in traditional scientific journals, but endorsed by international health associations or academic groups or centres.
Table 2
Main concepts and search words with wildcards that compose our search string.
Concept
|
Search words (English)
|
Search words (Spanish)
|
Robot
|
robot*, robot* system, clinical robot*, robot* design, robot* avatar*, mobile robot*, robot* platform*
|
robot*, sistema* robótico*, robot* clínico*, diseño de robot*, avatar*, robot* móvil*, plataforma* robótica*
|
Telepresence
|
telepresence, telemedicine, telehealth, remote presence
|
telepresencia*, telemedicina, telesalud, presencia remota
|
ICU
|
intensive care unit, ICU, hospital*, medical center*, health* facilit*
|
unidad de cuidados intensivos, UCI, hospital*, centro* médico*, establecimiento* sanitario*, centro* de salud
|
Information from private institutions will be taken into account if it is publicly available on the Web. Studies of all types of basic and clinical research will be included. We understand basic research as focused on experimental studies in controlled environments and clinical research as studies performed in real world conditions [19, 20]. Our attention will focus on primary sources. Papers cited in the documents spotted in our original search will also be considered.
4.4.2 Exclusion criteria
We will omit position statements, editorials, letters, interviews, speeches, research protocols, dissertations, internet articles or videos, and papers in progress. Moreover, animal research and open access encyclopedias will not be considered.
4.4.3 Selection process
Paper selection will follow the steps depicted in the flow diagram of Figure 2. The authors will carry out an initial search in databases and gray literature sources. The bibliographic references of the papers identified along with their keywords and abstracts will be stored in Zotero software [21]. Two authors will remove duplicate records. Paper screening will be executed by two authors using titles and abstracts. Each paper will be assessed for eligibility by two authors independently by scanning the entire paper. One author will examine the assessment process once 10% of the papers have been inspected. Selection criteria will be revisited periodically and adjustments to our inclusion and exclusion criteria will be discussed and agreed by the team. Authors involved on each step will cross-check their results, and any discrepancy will be resolved on consensus. If there is no agreement, a decision will be taken by all team members and an external stakeholder, if it is needed.
4.5 Charting the data
Table 3 shows the arrangement of our preliminary set of attributes chosen to fulfill our research objective. Paper information is useful for ease of reference and for classifying studies by their goals and type of research. The rest of the attributes are organized into a two-level hierarchy of categories in which the main level corresponds to the Donabedian’s structure-process-outcome model [22]. This description is similar to other models found in engineering where outcome measurements are required [23], and hence becomes familiar to various of the team members. The Structure category is divided into the 6 classes of elements of a telepresence robot system proposed in [3], that is: i) the physical robot, ii) the robot’s user, iii) the interfaces, iv) the user’s environment and v) the remote environment, and vi) the people interacting with the robot. In addition, the table comprises the relation cost-benefit of the robots, along with legal, ethics and moral implications, and the technical training required by the health professionals to maintain, assist, operate or evaluate the robot. Attributes grouped into the Process category describe the type of patient care that is delivered and the actions performed on purpose. The Output category consists of attributes related to changes in the health situation of the patient, the attitude and perception of physicians, patients and nurses, and the overall acceptance for the robotic system. Each set of attributes are related to a question described in Section 4.2.
As a pilot procedure, two authors will select at random 10% of the articles included in the study, and will collect data using the extraction table described above. The goal is to identify any differences in criteria for judging data, and to evaluate our initial dataform. A third author will review the consistency of the extraction. Any differences will be discussed first by the three authors, or with all the team members in one of the meetings that will be scheduled regularly. Lessons learned will be shared with the team members, and each one of the four authors will extract data from approximately 20% of the remaining 80% of the papers. Observations to the set of attributes will be accepted for discussion during the whole extraction process and the modifications approved by the team will be implemented accordingly.
Table 3
Category
|
Sub-category
|
Attribute
|
Related questions
|
Paper information
|
Bibliographic data
|
Title, First Author, Year, Abstract, Type of publication, Publisher, Country
|
Only for statistical information and reference
|
|
General methodology
|
Research objective and questions, Type of study design
|
RQ4, RQ2
|
Structure
|
The physical robot
|
Locomotion system, cover material, physical appearance, size, shape, ergonomy, actuators, sensors
|
RQ1
|
|
The physician operator
|
Specialty, years of experience with robots
|
RQ2, RQ4
|
|
The interfaces
|
Communication interfaces, communication network, bluetooth, LAN
|
RQ1
|
|
The operational environment and remote control station
|
Furniture and lighting conditions in the operator and patient environments
|
RQ2, RQ4
|
Table 3: Data extraction table (Cont.).
Category
|
Sub-category
|
Attribute
|
Related questions
|
Structure
|
The interactions with the robot
|
Number of nurses, family, caregivers, and patients, input and output modalities, devices used
|
RQ2, RQ4
|
|
Autonomy
|
Perceptual and cognitive capabilities
|
RQ1, RQ4
|
|
Technical support
|
Installation, preventive and corrective maintenance, upgrades, type of technicians
|
RQ1, RQ4
|
|
Cost-benefit relationship
|
Type of analysis, health care costs and non health care costs [25].
|
RQ4, RQ2
|
|
Legal, ethics and moral regulations
|
Liability, privacy invasion, non-verbal language, non appropriate contact, informed consent, computer fraud, cyber security, potential torts,
|
RQ3, RQ4
|
Table 3: Data extraction table (Cont.).
Category
|
Sub-category
|
Attribute
|
Related questions
|
Structure
|
Training of medical personnel
|
Manuals, books, on line or in situ courses, simulators
|
RQ4, RQ1
|
|
Health care databases
|
Type of databases, type of information stored
|
RQ1, RQ2, RQ3, RQ4
|
|
Communication infrastructure
|
Type of technology, network connections, speed of transmission
|
|
Process
|
Robotic interaction
|
Type of interactions with the patient, with the environment, and feedback to the operator
|
RQ4
|
|
Medical attention
|
Type of health condition of the patient, type of medical attention to the patient
|
RQ2
|
|
Robot usability
|
Cognitive load, learning curve in both operator and interactants, error identification, recovery and prevention capabilities
|
RQ1, RQ2
|
|
Access to health care databases
|
Type of retrieved data, display options
|
RQ1, RQ2, RQ3, RQ4
|
Table 3: Data extraction table (Cont.).
Category
|
Sub-category
|
Attribute
|
Related questions
|
Outcomes
|
Perception and acceptance
|
User satisfaction and technology acceptance from physicians, patients, health workers
|
RQ2, RQ4
|
|
Healthcare effects
|
Patient mortality, overall health status of the patient during the usage of the robot
|
RQ1, RQ2, RQ4
|
|
Robot usage time
|
|
RQ2, RQ4
|
4.6 Collating, summarizing and reporting the results
We propose both a basic quantitative analysis and a thematic synthesis [26] to summarize and report our findings. Quantitative analysis will be descriptive statistics representing data such as the studies found by year, country distribution, number of studies in which each attribute is considered, types of study design, satisfaction and acceptance of the stakeholders, medical conditions treated, patient mortality, and average robot usage. Table 4 is an example of a matrix that describes the number of research papers that study each attribute. This table will be useful to spot those robots features affecting one or more of the ICU related Donabedian categories.
Thematic synthesis will be performed in 3 steps: 1) coding the text of the findings of each study, 2) organizing codes into a tree structure composed of descriptive themes and, at the same time, by establishing conceptual relationships (or translations) between studies; finally, 3) generating analytic themes that allow to transcend the synthesis of the original findings into new interpretations or concepts. This latter step will be carried out through periodical discussions to iteratively refine the analytic themes that cover and explain the descriptive themes. To do so we will use EPPI-Reviewer software [27]. Three of the authors will perform steps 1 to 3. The rest of the team will track the progress and help to resolve any controversy whenever necessary. We propose thematic synthesis because it is an adequate approach to arise to new hypotheses [28], one of our most fundamental goals.
Table 4: Example of a summary table.
4.7 Consultation with stakeholders
During the entire process we are going to work closely with a group of two health specialists and one decision maker that is also a health professional, all of them experienced on either telepresence robots or robotics systems in medicine. We will keep them informed about our progress in order to receive continuous feedback.