Journal of
Personalized
Medicine
Article
Development and Trends in Artificial Intelligence in Critical
Care Medicine: A Bibliometric Analysis of Related Research
over the Period of 2010–2021
Xiao Cui 1,† , Yundi Chang 1,† , Cui Yang 1 , Zhukai Cong 2 , Baocheng Wang 3,4, * and Yuxin Leng 1, *
Citation: Cui, X.; Chang, Y.; Yang, C.;
Cong, Z.; Wang, B.; Leng, Y.
Development and Trends in Artificial
Intelligence in Critical Care Medicine:
A Bibliometric Analysis of Related
Research over the Period of
2010–2021. J. Pers. Med. 2023, 13, 50.
https://doi.org/10.3390/
jpm13010050
Academic Editor: Keliang Xie
Received: 8 December 2022
Revised: 22 December 2022
Accepted: 22 December 2022
Published: 27 December 2022
Copyright: © 2022 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
J. Pers. Med. 2023, 13, 50. https://doi.org/10.3390/jpm13010050
1 Department of Critical Care Medicine, Peking University Third Hospital, 49 North Garden Road, Haidian
District, Beijing 100191, China
2 Department of Anesthesiology, Peking University Third Hospital, 49 North Garden Road, Haidian District,
Beijing 100191, China
3 National Science Library, Chinese Academy of Sciences, 33 Beisihuan Xilu, Haidian District,
Beijing 100090, China
4 School of Economics and Management, University of Chinese Academy of Sciences, 19A Yuquan Road,
Beijing 100049, China
* Correspondence: wangbc@mail.las.ac.cn (B.W.); lengyx@bjmu.edu.cn (Y.L.)
† These authors contributed equally to this work.
Abstract: Background: The intensive care unit is a center for massive data collection, making it the
best field to embrace big data and artificial intelligence. Objective: This study aimed to provide
a literature overview on the development of artificial intelligence in critical care medicine (CCM)
and tried to give valuable information about further precision medicine. Methods: Relevant studies
published between January 2010 and June 2021 were manually retrieved from the Science Citation
Index Expanded database in Web of Science (Clarivate), using keywords. Results: Research related to
artificial intelligence in CCM has been increasing over the years. The USA published the most articles
and had the top 10 active affiliations. The top ten active journals are bioinformatics journals and are
in JCR Q1. Prediction, diagnosis, and treatment strategy exploration of sepsis, pneumonia, and acute
kidney injury were the most focused topics. Electronic health records (EHRs) were the most widely
used data and the “-omics” data should be integrated further. Conclusions: Artificial intelligence in
CCM has developed over the past decade. With the introduction of constantly growing data volume
and novel data types, more investigation on artificial intelligence ethics and model correctness and
extrapolation should be performed for generalization.
Keywords: artificial intelligence; critical care medicine; intensive care units; precision medicine; big
data; bibliometric analysis
1. Introduction
In recent years, the rapid development of precision medicine has led to significant
improvements in diagnosing and treating many diseases [1]. However, overcoming the
interference of patient heterogeneity, such as complex primary diseases and multiple risk
factors, has become an obstacle to the development of precision medicine in critical care
medicine (CCM) [2,3]. Accordingly, ensuring that the commonalities are identified may be
an effective strategy for addressing this problem [4]. Patients in intensive care units (ICUs)
are continuously monitored and treated, generating various real-time data, including
manually recorded and digitally obtained data, such as temperature, urine volume, oxygen
saturation, etc. These data should be considered in combination to assist in clinical decision-
making. However, too much information increases the difficulty of data interpretation,
especially when data conflicts. Identifying which data is more reliable and reflects the
commonality confuses clinicians [5]. This makes ICUs the best field for embracing Big Data.
https://www.mdpi.com/journal/jpm
J. Pers. Med. 2023, 13, 50 2 of 10

Artificial intelligence (AI), with a better capacity to handle big data than the human
brain, has been of great aid in clinical decision-making [6,7], contributing to the practical
success of precision medicine in CCM. In 1981, logistic regression was implemented to
verify the validity of the APACHE score [8], which was the initial application of AI in
aiding clinical decision-making in CCM. Since then, AI has become widely recognized in
digging clinical evidence [9,10]. Researchers have begun to realize that data-processing
capabilities are essential for CCM [9,11], and many applications of AI have been carried out.
Electronic health records (EHRs), integrated with various types of medical information, are
the basis of evidence-based medicine. This data volume is ideal for AI models, which thrive
on large datasets [9]. The widespread adoption of EHRs brought the data science world
to the patient’s bedside and paved the way for data-based AI models [11]. The MIMIC
database is a perfect example of integrated EHR data. The MIMIC is an extensive, freely
available database comprising health-related data from patients admitted to the critical care
units of the Beth Israel Deaconess Medical Center from 2001 to 2019 [12,13]. Several specific
disease/complication evaluation models have been generated from the MIMIC database to
promote decision-making [14]. The implementation of AI models for infectious diseases
with unknown pathogens initially, such as sepsis and COVID-19, has made significant
progress in the early prediction, diagnosis, and exploration of early treatment strategies [15]
and remains crucial for the foreseeable future.
Bibliometrics is a field of quantitative science that analyzes the scientific literature to
provide an overview of a certain topic [16]. It could reveal the current hotspots and future
directions in a certain area by bibliometric methods. Although there has been an increasing
focus on AI in CCM, a detailed analysis of the development and trends in AI in CCM has
not yet been carried out. Additionally, with the announcement of the research initiative on
precision medicine in 2015, the “-omics” technology has been valued for its identification
of complex biological mechanisms [17]. However, literature that provides what happened
and is happening with a combination of omics and AI in CCM is still lacking. In this study,
we aimed to reveal current trends of AI in CCM by analyzing the hot topics and reflecting
on the current status of the application of omics with AI in CCM. Specifically, AI studies
were retrieved in the field of CCM between January 2010 and June 2021, and the varying
trends in published studies of AI in CCM were analyzed. We hope that our survey will
further deepen the critical care experts’ understanding of AI in CCM and promote the
commonality mining for critical illnesses.

2. Materials and Methods

2.1. Retrieval of AI Studies in Adult CCM
To clarify the current status of AI applied in the research of adult CCM, all relevant studies
published between January 2010 and June 2021 were retrieved from the Science Citation Index
Expanded database in Web of Science (Clarivate). After a preliminary screening by searching
titles, abstracts, and keywords, such as “artificial intelligence,” “machine learning,” “deep
learning,” “neural network,” “expert system,” “data mining,” “electronic hospital record,” and
their variants, all studies were manually checked to ensure compliance with topic suitability
by two different critical care specialists. In the event of discordance between the two reviewers,
a third reviewer independently evaluated the paper.

2.2. Statistics and Analysis

A total of 2301 studies were primarily acquired, of which 1388 confirmed the appli-
cation of AI in adult CCM based on manual selection (Figure 1). Articles and meeting
abstracts were the main document types, accounting for 86.8% of studies. Other document
types included reviews, editorial material, letters, proceedings papers, early access, book
chapters, and news items (Table 1). To reflect the research status, the author countries,
author affiliations, publishing journals, and documents’ citation frequency of 937 original
research studies were subsequently analyzed. Co-occurrence analysis on publications’
keywords was conducted by VOSViewer (v1.6.8) [15].
 chapters, and news items (Table 1). To reflect the research status, the author countries, 97
author affiliations, publishing journals, and documents’ citation frequency of 937 original 98
research studies were subsequently analyzed. Co-occurrence analysis on publications’ 99
keywords was conducted by VOSViewer (v1.6.8) [15]. 100
J. Pers. Med. 2023, 13, 50 3 of 10

Figure 1. Flowchart outlining the search strategies in detail.

101
Table 1. Document types of AI papers in adult critical care based on manual selection.
Figure 1. Flowchart outlining the search strategies in detail. 102
Type of Document Number Percent (%)
Table 1. Document types
Article of AI papers in adult critical
937 care based on manual selection.
67.5% 103
Meeting Abstract 268 19.3%
Type of Document Number Percent (%)
Review 82 5.9%
Article
Editorial Material 93773 67.5%
5.3%
MeetingLetter
Abstract 26826 19.3%
1.9%
Proceeding
Review Paper 8223 1.7%
5.9%
Early Acess 19 1.4%
Editorial Material 73 5.3%
Book Chapter 3 0.2%
LetterItem
News 262 1.9%
0.1%
Proceeding Paper 23 1.7%
Early Acess
3. Results 19 1.4%
Book Chapter 3 0.2%
3.1. Highly Active Countries, Affiliations, and Journals in the Field of AI in CCM
News
During Item
the past decade, the number of2 publications relating to AI0.1% in adult CCM has
continued to escalate yearly, especially between 2018 and 2020, owing to the emergence
3.and
Results
widespread use of deep learning. In 2020, the number of articles nearly tripled that in 104
3.1. Highly
2018, withActive Countries,
an average of 23Affiliations, and Journals
articles published per in the Field
month. Theof number
AI in CCMof publications in 105
the During
first halfthe
of past
2021decade,
exceeded that in 2019 (Figure 2A). This tremendous
the number of publications relating to AI in adult growth
CCMimplies
has 106
that the application of AI in CCM research has become popular and
continued to escalate yearly, especially between 2018 and 2020, owing to the emergence widely accepted. 107
The USA isuse
and widespread in aofleading positionIn
deep learning. in 2020,
the field
the of AI for of
number adult CCM.
articles Although
nearly European
tripled that in 108
countries account for the highest number of publications in the top 10
2018, with an average of 23 articles published per month. The number of publications in active countries, the 109
USA produced the most publications in various journals. The number of publications in
the USA exceeded the sum of the other nine countries, followed by China and England
(Figure 2B). Not surprisingly, the most active affiliations are all from the USA, whose
publications have received significant attention from peer researchers. The average citation
frequency of these studies ranged from 8 to 27 (Figure 2C). Regarding journal selection, the
top 10 active journals were generally of high quality, and most of them were JCR (Journal
Citation Reports) Q1. In addition, most of the top 10 journals were bioinformation journals,
followed by critical care and comprehensive journals. Among these journals, CCM was the
most popular (Figure 2D).
J. Pers. Med. 2022, 12, x FOR PEER REVIEW 4 of 11

the first half of 2021 exceeded that in 2019 (Figure 2A). This tremendous growth implies 110
J. Pers. Med. 2023, 13, 50 that the application of AI in CCM research has become popular and widely accepted. 4 111
of 10

112
Figure 2. The publication trends, active countries, journals, and affiliations in the research of AI in
Figure
CCM.2.(A).
TheNumber
publication trends, active
of published countries,
papers, journals,
*, 2021 number and
is affiliations in the
partial to Jun research
2021. of AI in
(B). Active 113
countries.
CCM. A. Number of published papers, *, 2021 number is partial to Jun 2021. B. Active countries. C. 114
(C). Active affiliations. (D). Active journals.
Active affiliations. D. Active journals. 115

3.2. Variation of Author’s Keywords

The USA is in a leading position in the field of AI for adult CCM. Although European 116
Regarding
countries keywords,
account for 2038
the highest authors
number provided keywords
of publications in the top that appeared
10 active in 937
countries, thearticles
117
with 4111 frequencies. The top ten keywords accounted for 25% of the
USA produced the most publications in various journals. The number of publications in 118 total frequency.
They
the USAare “machine
exceeded thelearning”
sum of the (256),
other“intensive care unit”
nine countries, (136),by“critical
followed care”England
China and (126), “EHRs”
119
(94), “prediction” (85), “sepsis” (85), “clinical decision support systems” (70),
(Figure 2B). Not surprisingly, the most active affiliations are all from the USA, whose pub- “artificial
120
intelligence”
lications (62), “deep
have received learning”
significant (58), and
attention “mortality”
from (55; Figure
peer researchers. 3A).
The The variation
average citation trends
121
of these high-frequency keywords are shown in Figure 3B. Studies with deep learning as
a keyword began to appear in 2017. Through keyword co-occurrence analysis, this study
found that the mortality of sepsis may be representative of the AI applied in the field of
CCM (Figure 3C).
 They are “machine learning” (256), “intensive care unit” (136), “critical care” (126), 130
“EHRs” (94), “prediction” (85), “sepsis” (85), “clinical decision support systems” (70), “ar- 131
tificial intelligence” (62), “deep learning” (58), and “mortality” (55; Figure 3A). The varia- 132
tion trends of these high-frequency keywords are shown in Figure 3B. Studies with deep 133
learning as a keyword began to appear in 2017. Through keyword co-occurrence analysis, 134
J. Pers. Med. 2023, 13, 50
this study found that the mortality of sepsis may be representative of the AI applied5in of 10
135
the field of CCM (Figure 3C). 136

137
Figure 3. Variations of author-provided keywords. (A). Distribution of keywords frequency.
(B). Variation trend of the Top 10 keywords. (C). Co-occurrence analysis.

3.3. Features of AI in Adult CCM

To clarify the features of AI in adult CCM, the diseases, data types, and clinical
goals most involved in the 937 publications were analyzed. Several areas were open
for examination (Figure 4A). Sepsis, pneumonia, acute kidney injury, hospital-acquired
infection, and brain injuries were the most targeted study topics. Due to COVID-19, the
number of studies on pneumonia increased significantly, as 71 of the 79 pneumonia-related
articles were related to COVID-19 (Figure 4A). Regarding data type, most articles used
EHRs, possibly due to accessibility (exceeding 70%). Waveforms are also a popular subject,
possibly due to ECG and EEG often being used to monitor patients’ vital signs. In addition,
laboratory studies may be a common tool for studying the mechanisms underlying most
diseases (Figure 4B). We also found that computed tomography (CT) data were most
frequently applied (Figure 4D) in pneumonia studies, where all 13 articles focused on
COVID-19. Most publications have applied CT data to construct a deep learning system
to extract the imaging features of COVID-19 and provide a diagnostic system. Due to
their ability to detect and analyze volatile organic compounds, data from electronic noses
(e-noses) offer the possibility of predicting and diagnosing ventilator-associated pneumonia
 laboratory studies may be a common tool for studying the mechanisms underlying mos
diseases (Figure 4B). We also found that computed tomography (CT) data were most fre
quently applied (Figure 4D) in pneumonia studies, where all 13 articles focused o
COVID-19. Most publications have applied CT data to construct a deep learning system
J. Pers. Med. 2023, 13, 50 to extract the imaging features of COVID-19 and provide a diagnostic system. Due to thei
6 of 10
ability to detect and analyze volatile organic compounds, data from electronic noses (e
noses) offer the possibility of predicting and diagnosing ventilator-associated pneumoni
(Figure 4C). Regarding the representative data of precision medicine and personalize
(Figure 4C). Regarding the representative
medicine, only 1.4% of studies data
used of
theprecision
omic datamedicine and personalized
(Figure 4B).
medicine, only 1.4% of studies used the omic data (Figure 4B).
A D isease B D ata typ e C C linical g o als

D E
D isease b y d ata typ e D isease b y clinical g o als

Figure 4. Features of AI in adult CCM. (A). Type of diseases. (Sepsis includes sepsis and septic
Figure 4. Features of AI in adult CCM. A. Type of diseases. (Sepsis includes sepsis and septic shock
shock; pneumonia includes community-acquired pneumonia and COVID-19 pneumonia, but does
pneumonia includes community-acquired pneumonia and COVID-19 pneumonia, but does not in
not include hospital-acquired pneumonia
clude hospital-acquired and ventilator-acquired
pneumonia pneumonia;
and ventilator-acquired acute kidney
pneumonia; injury injury in
acute kidney
includes acute kidney injury and acute renal failure; HAI includes ventilator-acquired pneumonia,
cludes acute kidney injury and acute renal failure; HAI includes ventilator-acquired pneumonia
catheter-associated
catheter-associated infection
infection and other and other hospital-acquired
hospital-acquired infection)
infection) (B). B. Data
Data type. type. (Where
Where studiesstudies ha
more than one data types, all were recorded. (Waveform data includes
had more than one data types, all were recorded. (Waveform data includes ECG, EEG, arterial ECG, EEG, arterial puls
pulse waveforms, and other waveform data; Images data includes endoscopy and facial images.) C. Clin
waveforms, and other waveform data; Images data includes endoscopy and facial images.)
ical goals. (Prediction includes prediction of disease onset/occurrence, progression, and prognosi
(C). Clinical goals. (Prediction includes prediction of disease onset/occurrence, progression, and
segmentation includes spatial segmentation by radiographic features) D. Disease by data type. E
prognosis; segmentation includes spatial segmentation by radiographic features) (D). Disease by
data type. (E). Disease by clinical goals. HAI = hospital-acquired infection; MV = mechanical
ventilation; ARDS = acute respiratory distress syndrome; COPD = chronic obstructive pulmonary
disease; EHR = electronic health record; CT = computed tomography; MRI = magnetic resonance
imaging; ECG = electrocardiogram; EEG = electroencephalogram.

Regarding clinical goals, approximately three-quarters of the studies focused on

prediction, diagnosis, and treatment strategy explorations (Figure 4C). When considering
data type or clinical goals with diseases, we found that besides the EHR data, waveform,
labs, omics, and CT parameters were the most investigated to achieve AI for hot spot-
disease (Figure 4D). Almost all aspects of the clinical goals were concerned with the AI for
sepsis or pneumonia (Figure 4E).

4. Discussion
This study found that the number of articles related to AI in CCM has been growing
over the years, particularly with a fold increase from 2018 to 2020. The quantity as well
as quality of the articles is overwhelming. The top 10 active journals were JCR Q1. This
suggests that AI is gaining increasing attention in CCM. In addition, the USA is in a
dominant position in this field, with the most published articles and the top 10 active
affiliations. China’s publication of AI-related studies in CCM ranks second, suggesting
that current research in this field is in its infancy in China. Moreover, it was observed that
research often focused on common diseases in the ICU, such as sepsis, pneumonia, and
J. Pers. Med. 2023, 13, 50 7 of 10

acute kidney injury. Additionally, more than 90% of articles depended on traditional data
types, such as EHRs and waveforms (including ECG, EEG, arterial pulse waveforms, and
other waveform data), which was possibly due to data accessibility.
Unfortunately, the importance of “-omics” technologies in precision medicine has not
been fully recognized by critical care physicians and is not widely applied in AI modeling
(Figure 4B). High-throughput sequencing coordinated with in-depth biological informa-
tion can reveal detailed differences between individual health and disease statuses [18].
Different omics data provide different but complementary biological information from
different biological layers, and these can be integrated by multi-omics studies to offer a
more comprehensive view of complex diseases which are common in ICU. Publications
based on “multi-omics and machine learning or deep learning” keywords have started
to emerge and become popular in the recent 5 years [19]. Most of the studies focus on
cancer [19], and COVID-19 is a hot topic of critical care medicine in this area with a focus
on the prediction of the severity and exploration of the mechanism of COVID-19 [20,21].
However, the number of related publications is smaller compared with other data types
and most of them only use individual omics according to our study. One possible reason
could be the belief that “-omics” technologies are quite immature to provide reliable results.
Additionally, this field lacks standard operating procedures for data acquisition, integra-
tion, and analysis. Despite the limitations of omics itself, there are still some challenges
during the application of omics in AI models. For example, the imbalanced omics dataset
caused by rare disease classes may lead to an overfitted model [22,23]. In addition, the
classical “curse of dimensionality” problem is inevitable [24]. Omics technologies provide
large amounts of raw data, resulting in computationally intensive methods and likely
misleading algorithm training [19]. Reduced dimensional data could show the interaction
among different omics but may also lead to the ignorance of weak signals and missing
information [22,25,26]. These factors impede the generalization of “-omics” technologies in
intensive care medicine, particularly at community hospitals. This suggests that there is a
lot of work to be done before achieving precision medicine in intensive care medicine.
However, such concerns should not obscure the progress of AI using traditional data
types, which performed superior accuracy and earlier prediction compared with empirical
clinical decisions. In 2015, Pirracchio et al. provided a new mortality prediction algorithm
for ICU patients using the implementation of a super learner [27]. They found that two
super learner prediction models (SL1 and SL2) offered better performance for mortality
prediction (cv-AUROC = 0.85, 0.88, respectively) than the SOFA score (cv-AUROC = 0.71)
and SAPS II score (cv-AUROC = 0.78). AI also performs better in early prediction. Ac-
cording to Wickramaratne et al.’s work, sepsis can be predicted 6 h in advance with their
model (AUROC = 0.97) [28]. In 2018, Meyer et al. used deep machine learning methods
(recurrent neural networks) to predict real-time severe complications [29]. It was shown to
be more accurate than conventional clinical reference tools, enhancing the absolute compli-
cation prediction AUC by 0.29 for bleeding, 0.24 for mortality, and 0.24 for renal failure.
These studies suggest that with the development of new technology and the iteration of
the algorithm, AI with more training data will more accurately assess the prognosis than
today’s various scores in CCM [30]. AI also plays an irreplaceable role in diseases with
unknown pathogens, such as COVID-19. The COVID-19 outbreak has spread globally
and placed tremendous pressure on healthcare resources. AI provides an effective and
efficient strategy to combat the COVID-19 pandemic [15,31,32]. Jiao et al. developed an AI
system to predict the prognosis of patients with COVID-19 based on chest X-rays [33]. The
model showed a significantly better prognostic performance than the severity scores on
both internal (C-index 0.805 vs. 0.781) and external testing (C-index 0.752 vs. 0.715).
Along with the progress of AI in medicine, more and more AI products/devices are
approved by the U.S. Food and Drug Administration (FDA). Some products may be helpful
for healthcare, such as some image analysis software (for example, DeepRhythmAI by
Medicalgorithmics SA) and basic cardiopulmonary function monitoring software (such as
IRNF App by Apple lnc and Air Next by NuvoAir AB), etc. [34]. However, we still lack
J. Pers. Med. 2023, 13, 50 8 of 10

products specifically for critically ill patients. More targeted AI devices for patients in ICU
are expected.
Meanwhile, the challenges of implementing AI in the ICU have captured public
attention [11]. AI application is generally associated with barriers concerning data collec-
tion/management and the development/generalization of models. The most important
aspect that should be considered is data sharing [35,36]. ICU data were shared among
hospitals, however, using such sensitive information and sharing this data experiences
several difficulties, such as privacy concerns, ethical considerations, attribution issues,
laws, and regulations. Ensuring data security hampers the progress of AI from this angle.
Therefore, an international consensus is urgently needed. Furthermore, it should be noted
that AI is not always accurate. Some researchers have raised the technical challenges of
machine learning in CCM, such as the difficulty of manually calibrating and adjusting
models calculated using AI and whether this will affect their applicability [37]. In addition,
inaccurate data re-entry irreparably damages the continuous learning model and conse-
quently affects its localization [38,39]. This may have led to incorrect clinical decisions.
Thus, physicians need to make decisions with personal experience and should not rely
excessively on AI; be a good master, but not a slave to AI.
Admittedly, it is believed that a continuous learning algorithm can enhance its effective-
ness [40]. It would be convenient for real-time updates regarding localization and to improve
accuracy. However, the principle of machine learning methods is not yet fully understood;
therefore, we can use the fundamental computing model as the first auxiliary clinical judg-
ment. Data can be periodically input in the backup after professional supervision [38]. If the
supervision team detects abnormal data, it can select and remove them in time.
It is believed that AI models should serve as an aid but not as a replacement for clinical
judgment. However, scientists who use AI may obtain more valuable information than
those who do not. Although AI may not be fully accurate owing to the live streaming data
and different models, it can be used as a reference, similar to clinical scores, to guide clinical
decision-making. The iteration of the algorithm improves the credibility and reliability of
AI models and resolves the aforementioned problems [41].
In conclusion, our study has provided an overview of the AI field in CCM and has
revealed the development status and main research topics in this area. Through the anal-
ysis, we also identify new perspectives for future research, for instance, the combination
of “-omics” technology and AI models. We hope our study could provide some valuable
information to researchers, considering the possibility of using existing data for AI, orga-
nizing disciplinary teams, and getting a better design of clinical trials. We believe that with
a combination of personal experience and proper AI models, AI will show its great power
in achieving precision medicine in CCM.

Author Contributions: Conceptualization, Y.L.; Data curation, X.C., Y.C., C.Y. and Z.C.; Formal
analysis, X.C., Y.C. and B.W.; Investigation, X.C. and Y.L.; Methodology, X.C. and Y.C.; Resources,
Y.L.; Supervision, B.W. and Y.L.; Validation, C.Y. and Z.C.; Visualization, B.W.; Writing — original
draft, X.C. and Y.C.; Writing — review & editing, B.W. and Y.L. All authors have read and agreed to
the published version of the manuscript.
Funding: This research was funded by the National Natural Science Foundation of China, grant
number 82172126.
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: Not applicable.
Conflicts of Interest: The authors declare no conflict of interest.
J. Pers. Med. 2023, 13, 50 9 of 10

References
1. Ramaswami, R.; Bayer, R.; Galea, S. Precision Medicine from a Public Health Perspective. Annu. Rev. Public Health 2018,
39, 153–168. [CrossRef] [PubMed]
2. Maslove, D.M.; Lamontagne, F.; Marshall, J.C.; Heyland, D.K. A path to precision in the ICU. Crit. Care 2017, 21, 79. [CrossRef]
[PubMed]
3. Lazăr, A.; Georgescu, A.M.; Vitin, A.; Azamfirei, L. Precision Medicine and its Role in the Treatment of Sepsis: A Personalised
View. J. Crit. Care Med. 2019, 5, 90–96. [CrossRef]
4. Prescott, H.C.; Carmichael, A.G.; Langa, K.M.; Gonzalez, R.; Iwashyna, T.J. Paths into Sepsis: Trajectories of Presepsis Healthcare
Use. Ann. Am. Thorac. Soc. 2019, 16, 116–123. [CrossRef]
5. Chaudhry, F.; Hunt, R.J.; Hariharan, P.; Anand, S.K.; Sanjay, S.; Kjoller, E.E.; Bartlett, C.M.; Johnson, K.W.; Levy, P.D.;
Noushmehr, H.; et al. Machine Learning Applications in the Neuro ICU: A Solution to Big Data Mayhem? Front. Neurol.
2020, 11, 554633. [CrossRef]
6. Shortliffe, E.H.; Sepúlveda, M.J. Clinical Decision Support in the Era of Artificial Intelligence. JAMA 2018, 320, 2199–2200.
[CrossRef]
7. Obermeyer, Z.; Lee, T.H. Lost in Thought—The Limits of the Human Mind and the Future of Medicine. N. Engl. J. Med. 2017,
377, 1209–1211. [CrossRef] [PubMed]
8. Knaus, W.A.; Draper, E.A.; Wagner, D.P.; Zimmerman, J.E. APACHE II: A severity of disease classification system. Crit. Care Med.
1985, 13, 818–829. [CrossRef]
9. Sanchez-Pinto, L.N.; Luo, Y.; Churpek, M.M. Big Data and Data Science in Critical Care. Chest 2018, 154, 1239–1248. [CrossRef]
10. Hinton, G. Deep Learning—A Technology with the Potential to Transform Health Care. JAMA 2018, 320, 1101–1102. [CrossRef]
11. van de Sande, D.; van Genderen, M.E.; Huiskens, J.; Gommers, D.; van Bommel, J. Moving from bytes to bedside: A systematic
review on the use of artificial intelligence in the intensive care unit. Intensive Care Med. 2021, 47, 750–760. [CrossRef] [PubMed]
12. Saeed, M.; Lieu, C.; Raber, G.; Mark, R.G. MIMIC II: A massive temporal ICU patient database to support research in intelligent
patient monitoring. Comput. Cardiol. 2002, 29, 641–644. [PubMed]
13. Johnson, A.E.; Pollard, T.J.; Shen, L.; Lehman, L.W.; Feng, M.; Ghassemi, M.; Moody, B.; Szolovits, P.; Celi, L.A.; Mark, R.G.
MIMIC-III, a freely accessible critical care database. Sci. Data 2016, 3, 160035. [CrossRef] [PubMed]
14. Data, M.I.T.C.; Mark, R. The Story of MIMIC. In Secondary Analysis of Electronic Health Records; Springer: Cham, Switzerland, 2016;
pp. 43–49.
15. Tran, N.K.; Albahra, S.; May, L.; Waldman, S.; Crabtree, S.; Bainbridge, S.; Rashidi, H. Evolving Applications of Artificial
Intelligence and Machine Learning in Infectious Diseases Testing. Clin. Chem. 2021, 68, 125–133. [CrossRef] [PubMed]
16. Bias, R.G. Research Methods for Human-Computer Interaction. J. Am. Soc. Inf. Sci. Technol. 2010, 61, 204–205. [CrossRef]
17. Collins, F.S.; Varmus, H. A new initiative on precision medicine. N. Engl. J. Med. 2015, 372, 793–795. [CrossRef]
18. Chen, R.; Mias, G.I.; Li-Pook-Than, J.; Jiang, L.; Lam, H.Y.; Chen, R.; Miriami, E.; Karczewski, K.J.; Hariharan, M.; Dewey, F.E.; et al.
Personal omics profiling reveals dynamic molecular and medical phenotypes. Cell 2012, 148, 1293–1307. [CrossRef]
19. Reel, P.S.; Reel, S.; Pearson, E.; Trucco, E.; Jefferson, E. Using machine learning approaches for multi-omics data analysis: A
review. Biotechnol. Adv. 2021, 49, 107739. [CrossRef]
20. Thomas, T.; Stefanoni, D.; Dzieciatkowska, M.; Issaian, A.; Nemkov, T.; Hill, R.C.; Francis, R.O.; Hudson, K.E.; Buehler, P.W.;
Zimring, J.C.; et al. Evidence of Structural Protein Damage and Membrane Lipid Remodeling in Red Blood Cells from COVID-19
Patients. J. Proteome Res. 2020, 19, 4455–4469. [CrossRef]
21. Overmyer, K.A.; Shishkova, E.; Miller, I.J.; Balnis, J.; Bernstein, M.N.; Peters-Clarke, T.M.; Meyer, J.G.; Quan, Q.; Muehlbauer, L.K.;
Trujillo, E.A.; et al. Large-Scale Multi-omic Analysis of COVID-19 Severity. Cell Syst. 2021, 12, 23–40.e27. [CrossRef]
22. Haas, R.; Zelezniak, A.; Iacovacci, J.; Kamrad, S.; Townsend, S.; Ralser, M. Designing and interpreting “multi-omic” experiments
that may change our understanding of biology. Curr. Opin. Syst. Biol. 2017, 6, 37–45. [CrossRef] [PubMed]
23. Rimoldi, S.F.; Scherrer, U.; Messerli, F.H. Secondary arterial hypertension: When, who, and how to screen? Eur. Heart J. 2014,
35, 1245–1254. [CrossRef] [PubMed]
24. Tozzi, A.; Peters, J.F. The Borsuk-Ulam theorem solves the curse of dimensionality: Comment on “The unreasonable effectiveness
of small neural ensembles in high-dimensional brain” by Alexander, N. Gorban et al. Phys. Life Rev. 2019, 29, 89–92. [CrossRef]
[PubMed]
25. Meng, C.; Zeleznik, O.A.; Thallinger, G.G.; Kuster, B.; Gholami, A.M.; Culhane, A.C. Dimension reduction techniques for the
integrative analysis of multi-omics data. Brief. Bioinform. 2016, 17, 628–641. [CrossRef] [PubMed]
26. Bush, W.S.; Dudek, S.M.; Ritchie, M.D. Biofilter: A knowledge-integration system for the multi-locus analysis of genome-wide
association studies. In Proceedings of the 2009 Pacific Symposium on Biocomputing, Kohala Coast, HI, USA, 5–9 January 2009;
pp. 368–379.
27. Pirracchio, R.; Petersen, M.L.; Carone, M.; Rigon, M.R.; Chevret, S.; van der Laan, M.J. Mortality prediction in intensive care units
with the Super ICU Learner Algorithm (SICULA): A population-based study. Lancet Respir. Med. 2015, 3, 42–52. [CrossRef]
28. Wickramaratne, S.D.; Shaad Mahmud, M.D. Bi-Directional Gated Recurrent Unit Based Ensemble Model for the Early Detection
of Sepsis. In Proceedings of the 2020 42nd Annual International Conference of the IEEE Engineering in Medicine and Biology
Society (EMBC), Montreal, QC, Canada, 20–24 July 2020; Volume 2020, pp. 70–73. [CrossRef]
J. Pers. Med. 2023, 13, 50 10 of 10

29. Meyer, A.; Zverinski, D.; Pfahringer, B.; Kempfert, J.; Kuehne, T.; Sündermann, S.H.; Stamm, C.; Hofmann, T.; Falk, V.; Eickhoff,
C. Machine learning for real-time prediction of complications in critical care: A retrospective study. Lancet Respir. Med. 2018,
6, 905–914. [CrossRef]
30. Gutierrez, G. Artificial Intelligence in the Intensive Care Unit. Crit. Care 2020, 24, 101. [CrossRef]
31. Zhou, Y.; Wang, F.; Tang, J.; Nussinov, R.; Cheng, F. Artificial intelligence in COVID-19 drug repurposing. Lancet Digit. Health
2020, 2, e667–e676. [CrossRef]
32. Huang, S.; Yang, J.; Fong, S.; Zhao, Q. Artificial intelligence in the diagnosis of COVID-19: Challenges and perspectives. Int. J.
Biol. Sci. 2021, 17, 1581–1587. [CrossRef]
33. Jiao, Z.; Choi, J.W.; Halsey, K.; Tran, T.M.L.; Hsieh, B.; Wang, D.; Eweje, F.; Wang, R.; Chang, K.; Wu, J.; et al. Prognostication of
patients with COVID-19 using artificial intelligence based on chest x-rays and clinical data: A retrospective study. Lancet Digit.
Health 2021, 3, e286–e294. [CrossRef]
34. FDA. Artificial Intelligence and Machine Learning (AI/ML)-Enabled Medical Devices. Available online: https://www.fda.gov/
medical-devices/software-medical-device-samd/artificial-intelligence-and-machine-learning-aiml-enabled-medical-devices (ac-
cessed on 21 December 2022).
35. Bailly, S.; Meyfroidt, G.; Timsit, J.F. What’s new in ICU in 2050: Big data and machine learning. Intensive Care Med. 2018,
44, 1524–1527. [CrossRef] [PubMed]
36. He, J.; Baxter, S.L.; Xu, J.; Xu, J.; Zhou, X.; Zhang, K. The practical implementation of artificial intelligence technologies in
medicine. Nat. Med. 2019, 25, 30–36. [CrossRef] [PubMed]
37. Mullainathan, S.; Obermeyer, Z. Does Machine Learning Automate Moral Hazard and Error? Am. Econ. Rev. 2017, 107, 476–480.
[CrossRef] [PubMed]
38. Cabitza, F.; Rasoini, R.; Gensini, G.F. Unintended Consequences of Machine Learning in Medicine. JAMA 2017, 318, 517–518.
[CrossRef] [PubMed]
39. Wilbanks, B.A.; Berner, E.S.; Alexander, G.L.; Azuero, A.; Patrician, P.A.; Moss, J.A. The effect of data-entry template design and
anesthesia provider workload on documentation accuracy, documentation efficiency, and user-satisfaction. Int. J. Med. Inform.
2018, 118, 29–35. [CrossRef]
40. Curran, G.M.; Bauer, M.; Mittman, B.; Pyne, J.M.; Stetler, C. Effectiveness-implementation hybrid designs: Combining elements of
clinical effectiveness and implementation research to enhance public health impact. Med. Care 2012, 50, 217–226. [CrossRef]
41. The Lancet. Artificial intelligence in health care: Within touching distance. Lancet 2017, 390, 2739. [CrossRef]

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.