Professional Documents
Culture Documents
Abstract
Introduction. Patients in intensive care units have a higher risk of presenting pressure ulcers due to immobility. Frequent
repositioning is a conventional care strategy to prevent its onset but is less applied during night shifts. Objective: To structure
a predictive model of repositioning frequency during sleep in a healthy adult population based on machine learning and
posterior validation in a cohort of adult patients hospitalized in intensive care. Methodology. A retrospective observational
study based on regular polysomnography tests obtained during a night's sleep in 180 adults between June 2021 and June 2022
at the Colombian Pulmonary Foundation. We obtained repositioning type and frequency, body mass index, and age to train a
computer algorithm based on a decision tree (Machine Learning). Then, algorithm validation was performed in a retrospective
sample of 986 adult patients who received repositioning in intensive care. Results. A total of 1,444 hours of sleep and 1,855
position changes were analyzed. The supine position was the most frequently adopted. Individuals between 18-45 years and ≥
46 years with overweight reported a higher frequency of repositioning (11.4 and 10.1 times), respectively, combining the three
types of positions (supine, right lateral, and left lateral). The predictive model applied in patients reported a mean of 1.83
(0.12-10.3) more repositioning times compared with the frequency of repositioning provided by health care providers 0.28
times (IQR 0.0-2.0) [ low correlation of 0.52, p = 0.101]. Conclusion. A machine learning predictive model showed that
critical patients might be mobilized more frequently based on age and BMC.
Keywords: Repositioning, pressure ulcer, machine learning, predictive model, clinical care prevention.
The developing predictive models based on machine learning aimed to improve the frequency of repositioning on critical
patients may prevent hospital pressure ulcers.
NIH category Clinical Research
presented; in this case, a database of 180 records was values determine the precision of the calculation of the
structured where the critical variables were age and BMI mean square error (MSE), which shows the mean square
and, as output variables, the number of mobilizations in a difference between the real values and the inferred values,
shift. (Table 1). and the analysis of the absolute error mean (MAE) which
shows the absolute mean difference between the actual
TABLE 1 values and the assumed values. The value of the depth of
VARIABLE CATEGORY the tree is modified until obtaining the value where this
error does not change to adjust the errors. For this purpose,
# Category Age IMC the age and BMI of the patients included in the study were
used as test data, and the errors between the real value and
Variable the preceded value were obtained.
1 18 a 40 15 a 24.9 [Normal]
years Implementation: Once the model was obtained, it
2 41 a 60 25 a 29.9 should be understood that the development of the web
years [Overweight] application is done through the use of tag languages such
3 > 60 año > 30 [Obesity] as HTML and CSS and programming languages such as
years JavaScript. For this activity, tools proposed by Piotr
Ptonoski (20) were used for the rules extraction in
The variables were grouped by categories according to pseudocode from the decision tree.
the behavior and distribution of each of them, with their
associated category, which is shown in Table 1. In this With the set of rules as text, it was proceeding to
way, the training data set and the test data set of the implement it in JavaScript, meaning that the conditional
decision tree prediction algorithm. sentence of the pseudocode obtained complies with the
necessary syntax so as not to generate any error during its
Method Used: with this structured database, the Python use. This statement should remain, as shown in Figure 1.
programming language and the Scikit-Learn library were For this, the Excel tool and its "search and replace text"
defined, focusing on data science with classification and function were used to make the changes to comply with
regression algorithms (18). An algorithm classification the syntax. Figure 6 shows the replaced text blocks and the
using the tree function is presented within this library. order in which these changes were made.
DecisionTreeRegressor allows training a learning
algorithm and obtaining a classification model based on TABLE 2.
the decision tree method. IF() CONDITIONAL STRUCTURE IN JAVASCRIPT
LANGUAGE.
The function tree: decisionTreeRegressor has different
arguments that are parameterized to adjust the precision
with which the model makes its inferences. Within these
arguments is the depth of the decision tree, that is, the Complies with
maximum number of nodes generated until reaching a opening the “(“
1 If( → If ((
response. It must be considered that at some value of this for the section of
parameter, the precision will not be more significant; there conditions
is a value for the parameter in which the accuracy will be The
the same no matter if it increases (19). conditional AND
2 and → &&
in JavaScript is
After defining the depth of the tree, the training of the replaced by &&
learning algorithm was carried out with the Scikit-Learn Closes the “)”
library, which it receives as parameters: age and BMI. from the section
This function delivers as a response a classification model of conditions,
based on the training data, with which inferences can now opens the “{“ for
be made about new cases that are delivered to the model, then ) the actions
taking into account that the input data for these models 3 response : → {result executed when
must be coherent with the training variables since 3.0 = the if() is met and
otherwise, the model could infer erroneous responses; defines the
however, this was adjusted with traditional logic in the variable results
application inputs. and gives it a
value
Evaluation metrics: With the classification model, 4 Concaten → Closes the “}” of the
ate function actions executed if
“}” conditions are met from if()
Of 1,444 hours of tests (mean 7.78 h/ patient-test), 1,130 h
Table 2. Step by step to get proper syntax in JavaScript. (67,815 min) were hours of sleep (mean 6.27 h, SD 1.35).
The time spent in the supine position was the longest,
followed by time in the right and left lateral positions
(Table 4).
I. Validity of the model [calculator]
We evaluated the concordance between the usual Table 4. Overall description of times during the
mobilization rate (observed) performed by nurses on adult polysomnography.
patients hospitalized for one night in the ICU compared
with the frequency of mobilization obtained by applying Total, in minutes(min) Total, in Hours (h)
the calculator expected repositioning (for one night). For Variable
Total Mean (SD) Total Mean (SD)
this comparison, data were selected from a group of 986
Time in 84,04
patients (included in the PENFUP-2 study) (17) aged ≥18 466.88 ± 53.31 1,444 7.78 ± 0.88
bed/test 0
years (similar to the polysomnography model), Time 11,80
65.55 ± 63.02 197 1.09 ± 1.05
hospitalized in the ICU, exposed to sedation, in a awake 0
Time of 67,81
vulnerable state, with mobilization requirement by sleep 5
376.75 ± 81.05 1,130 6.27 ± 1.35
healthcare personnel. The information was obtained Time 31,49
175.97 ±118.10 525 2.91 ± 1.97
retrospectively from the information reported in the Supine 9
Time 18,74
PENFUP 2 study following the published methods of the Right 6
104.72 ±77.88 312 1.73 ± 1.3
Penfup-phase 2 study [patients signed the consent form]. 16,39
Time Left 91.58 ± 89.90 273 1.51 ± 1.49
4
II. RESULTS
A. Description de demographic data B. Normal repositioning changes during sleep
The study included 180 patients' clinical histories who The study evaluated 1,855 bodily position changes during
complied with the eligibility criteria. From this total, 63% sleep (mean 10.3, SD 5.65) obtained from the movement
were women, with a median age of 43 (RIQ 18-44). Most plots for each patient. The evaluation of the inter-observer
of the patients classified by age were aged 18 to 40 years correlation for each movement was very good (between
(43%). Although the average population was included, the primary observer and observer No. 2, Pearson
according to the evaluated body mass index, 43.9% were coefficient 96.5%, mean 0.12, SD 0.64). The supine
identified as overweight and 15.6% as obesity (Table 3). position had the highest observed occurrence (Table 5).
The Number of prone and undefined position changes
Table 3. Demographic description of patients represented only 3.3% of the types of movements .
included in the study
Table 5. Description of repositioning frequency
Variable Total (%) during sleep
Age category, n (%)
Variable n (%) Mean ± SD
18 to 40 years 78 (43.3) 802
Supine event 4.46 ± 2.66
41 to 60 years 75 (41.7) (43.23%)
527 (28.40
> 60 years 27 (15) Right lateral event 2.93 ± 2.11
%)
464 (25.01
BMI category, n (%) Left lateral event 2.58 ± 2.1
%)
15 to 24.9 [Normal] 73 (40.6) Prone event 34 (1.8 %) 0.19 ± 0.69
25 to 29.9 [Overweight] 79 (43.9) Undefined event 28 (1.5 %) 0.15 ± 0.63
Total 1855
> 30 [Obesity] 28 (15.6)
PDF version of this document, please download the including figures and tables. You must also send your final
electronic file from the IEEE Web site at manuscript on a disk, via e-mail, or through a Web
http://www.ieee.org/web/publications/authors/jtehm/index.html manuscript submission system as directed by the society
so you can use it to prepare your manuscript. contact. You may use Zip or CD-ROM disks for large
TABLE I
UNITS FOR MAGNETIC PROPERTIES
Conversion from Gaussian and
Symbol Quantity
CGS EMU to SI a
magnetic flux 1 Mx 108 Wb = 108 V·s
B magnetic flux density, 1 G 104 T = 104 Wb/m2
magnetic induction
H magnetic field strength 1 Oe 103/(4) A/m
m magnetic moment 1 erg/G = 1 emu
103 A·m2 = 103 J/T
M magnetization 1 erg/(G·cm3) = 1 emu/cm3
103 A/m
4M magnetization 1 G 103/(4) A/m
specific magnetization 1 erg/(G·g) = 1 emu/g 1 A·m2/kg
j magnetic dipole 1 erg/G = 1 emu
moment 4 1010 Wb·m
J magnetic polarization 1 erg/(G·cm3) = 1 emu/cm3
4 104 T
, susceptibility 1 4
mass susceptibility 1 cm3/g 4 103 m3/kg
permeability 1 4 107 H/m
Fig. 1. Magnetization as a function of applied field. Note that “Fig.” is
= 4 107 Wb/(A·m)
abbreviated. There is a period after the figure number, followed by two
r relative permeability r
spaces. It is good practice to explain the significance of the figure in the
caption. w, W energy density 1 erg/cm3 101 J/m3
N, D demagnetizing factor 1 1/(4)
E. Copyright Form Vertical lines are optional in tables. Statements that serve as captions for
the entire table do not need footnote letters.
An IEEE copyright form should accompany your final a
Gaussian units are the same as cgs emu for magnetostatics; Mx =
submission. You can get a .pdf, .html, or .doc version at maxwell, G = gauss, Oe = oersted; Wb = weber, V = volt, s = second, T =
http://www.ieee.org/copyright. Authors are responsible for tesla, m = meter, A = ampere, J = joule, kg = kilogram, H = henry.
obtaining any security clearances.
compositions are not specified, separate chemical symbols expositions and critical reviews of classical subjects and
by en-dashes; for example, “NiMn” indicates the topics of current interest.
intermetallic compound Ni 0.5Mn0.5 whereas “Ni–Mn” Authors should consider the following points:
indicates an alloy of some composition Ni xMn1-x. 1) Technical papers submitted for publication must
Be aware of the different meanings of the homophones advance the state of knowledge and must cite relevant
“affect” (usually a verb) and “effect” (usually a noun), prior work.
“complement” and “compliment,” “discreet” and 2) The length of a submitted paper should be
“discrete,” “principal” (e.g., “principal investigator”) and commensurate with the importance, or appropriate to
“principle” (e.g., “principle of measurement”). Do not the complexity, of the work. For example, an obvious
confuse “imply” and “infer.” extension of previously published work might not be
Prefixes such as “non,” “sub,” “micro,” “multi,” and appropriate for publication or might be adequately
“ultra” are not independent words; they should be joined treated in just a few pages.
to the words they modify, usually without a hyphen. There 3) Authors must convince both peer reviewers and the
is no period after the “et” in the Latin abbreviation “et al.” editors of the scientific and technical merit of a paper;
(it is also italicized). The abbreviation “i.e.,” means “that the standards of proof are higher when extraordinary
is,” and the abbreviation “e.g.,” means “for example” or unexpected results are reported.
(these abbreviations are not italicized). 4) Because replication is required for scientific progress,
An excellent style manual and source of information for papers submitted for publication must provide
science writers is [9]. A general IEEE style guide and an sufficient information to allow readers to perform
Information for Authors are both available at similar experiments or calculations and use the
http://www.ieee.org/web/publications/authors/transjnl/index.htm reported results. Although not everything need be
l disclosed, a paper must contain new, useable, and
fully described information. For example, a
specimen’s chemical composition need not be
VI. EDITORIAL POLICY reported if the main purpose of a paper is to introduce
Submission of a manuscript is not required for a new measurement technique. Authors should expect
participation in a conference. Do not submit a reworked to be challenged by reviewers if the results are not
version of a paper you have submitted or published supported by adequate data and critical details.
elsewhere. Do not publish “preliminary” data or results. 5) Papers that describe ongoing work or announce the
The submitting author is responsible for obtaining latest technical achievement, which are suitable for
agreement of all coauthors and any consent required from presentation at a professional conference, may not be
sponsors before submitting a paper. IEEE T RANSACTIONS appropriate for publication in a T RANSACTIONS or
and JOURNALS strongly discourage courtesy authorship. It JOURNAL.
is the obligation of the authors to cite relevant prior work.
The Transactions and Journals Department does not
publish conference records or proceedings. The VIII.CONCLUSION
TRANSACTIONS does publish papers related to conferences Please include a brief summary of the possible clinical
that have been recommended for publication on the basis implications of your work in the conclusion section.
of peer review. As a matter of convenience and service to Although a conclusion may review the main points of the
the technical community, these topical papers are paper, do not replicate the abstract as the conclusion.
collected and published in one issue of the TRANSACTIONS. Consider elaborating on the translational importance of the
At least two reviews are required for every paper work or suggest applications and extensions.
submitted. For conference-related papers, the decision to
accept or reject a paper is made by the conference editors APPENDIX
and publications committee; the recommendations of the
Appendixes, if needed, appear before the
referees are advisory only. Undecipherable English is a
acknowledgment.
valid reason for rejection. Authors of rejected papers may
revise and resubmit them to the T RANSACTIONS as regular
ACKNOWLEDGMENT
papers, whereupon they will be reviewed by two new
referees. The preferred spelling of the word “acknowledgment”
in American English is without an “e” after the “g.” Use
the singular heading even if you have many
VII. PUBLICATION PRINCIPLES acknowledgments. Avoid expressions such as “One of us
(S.B.A.) would like to thank ... .” Instead, write “F. A.
The contents of IEEE T RANSACTIONS and JOURNALS are
Author thanks ... .” Sponsor and financial support
peer-reviewed and archival. The T RANSACTIONS publishes
acknowledgments are placed in the unnumbered
scholarly articles of archival value as well as tutorial
footnote on the first page, not here.
REFERENCES [29] (Journal Online Sources style) K. Author. (year, month). Title.
Journal [Type of medium]. Volume(issue), paging if given.
[1] G. O. Young, “Synthetic structure of industrial plastics (Book style Available: http://www.(URL)
with paper title and editor),” in Plastics, 2nd ed. vol. 3, J. Peters, Ed. [30] R. J. Vidmar. (1992, August). On the use of atmospheric plasmas as
New York: McGraw-Hill, 1964, pp. 15–64. electromagnetic reflectors. IEEE Trans. Plasma Sci. [Online]. 21(3).
[2] W.-K. Chen, Linear Networks and Systems (Book style). Belmont, pp. 876–880. Available:
CA: Wadsworth, 1993, pp. 123–135. http://www.halcyon.com/pub/journals/21ps03-vidmar
[3] H. Poor, An Introduction to Signal Detection and Estimation. New
York: Springer-Verlag, 1985, ch. 4.
[4] B. Smith, “An approach to graphs of linear forms (Unpublished work
style),” unpublished.
[5] E. H. Miller, “A note on reflector arrays (Periodical style—Accepted
for publication),” IEEE Trans. Antennas Propagat., to be published. First A. Author (M’76–SM’81–F’87) and the other authors may include
[6] J. Wang, “Fundamentals of erbium-doped fiber amplifiers arrays biographies at the end of regular papers. Biographies are often not
(Periodical style—Submitted for publication),” IEEE J. Quantum included in conference-related papers. This author became a Member (M)
Electron., submitted for publication. of IEEE in 1976, a Senior Member (SM) in 1981, and a Fellow (F) in 1987.
[7] C. J. Kaufman, Rocky Mountain Research Lab., Boulder, CO, private The first paragraph may contain a place and/or date of birth (list place,
communication, May 1995. then date). Next, the author’s educational background is listed. The
[8] Y. Yorozu, M. Hirano, K. Oka, and Y. Tagawa, “Electron degrees should be listed with type of degree in what field, which
spectroscopy studies on magneto-optical media and plastic substrate institution, city, state, and country, and year degree was earned. The
interfaces (Translation Journals style),” IEEE Transl. J. Magn.Jpn., author’s major field of study should be lower-cased.
vol. 2, Aug. 1987, pp. 740–741 [Dig. 9th Annu. Conf. Magnetics The second paragraph uses the pronoun of the person (he or she) and
Japan, 1982, p. 301]. not the author’s last name. It lists military and work experience, including
[9] M. Young, The Techincal Writers Handbook. Mill Valley, CA: summer and fellowship jobs. Job titles are capitalized. The current job
University Science, 1989. must have a location; previous positions may be listed without one.
[10] J. U. Duncombe, “Infrared navigation—Part I: An assessment of Information concerning previous publications may be included. Try not to
feasibility (Periodical style),” IEEE Trans. Electron Devices, vol. list more than three books or published articles. The format for listing
ED-11, pp. 34–39, Jan. 1959. publishers of a book within the biography is: title of book (city, state:
[11] S. Chen, B. Mulgrew, and P. M. Grant, “A clustering technique for publisher name, year) similar to a reference. Current and previous research
digital communications channel equalization using radial basis interests end the paragraph.
function networks,” IEEE Trans. Neural Networks, vol. 4, pp. 570– The third paragraph begins with the author’s title and last name (e.g.,
578, Jul. 1993. Dr. Smith, Prof. Jones, Mr. Kajor, Ms. Hunter). List any memberships in
[12] R. W. Lucky, “Automatic equalization for digital communication,” professional societies other than the IEEE. Finally, list any awards and
Bell Syst. Tech. J., vol. 44, no. 4, pp. 547–588, Apr. 1965. work for IEEE committees and publications. If a photograph is provided,
[13] S. P. Bingulac, “On the compatibility of adaptive controllers the biography will be indented around it. The photograph is placed at the
(Published Conference Proceedings style),” in Proc. 4th Annu. top left of the biography. Personal hobbies will be deleted from the
Allerton Conf. Circuits and Systems Theory, New York, 1994, pp. 8– biography.
16.
[14] G. R. Faulhaber, “Design of service systems with priority
reservation,” in Conf. Rec. 1995 IEEE Int. Conf. Communications,
pp. 3–8.
[15] W. D. Doyle, “Magnetization reversal in films with biaxial
anisotropy,” in 1987 Proc. INTERMAG Conf., pp. 2.2-1–2.2-6.
[16] G. W. Juette and L. E. Zeffanella, “Radio noise currents n short
sections on bundle conductors (Presented Conference Paper style),”
presented at the IEEE Summer power Meeting, Dallas, TX, Jun. 22–
27, 1990, Paper 90 SM 690-0 PWRS.
[17] J. G. Kreifeldt, “An analysis of surface-detected EMG as an
amplitude-modulated noise,” presented at the 1989 Int. Conf.
Medicine and Biological Engineering, Chicago, IL.
[18] J. Williams, “Narrow-band analyzer (Thesis or Dissertation style),”
Ph.D. dissertation, Dept. Elect. Eng., Harvard Univ., Cambridge,
MA, 1993.
[19] N. Kawasaki, “Parametric study of thermal and chemical
nonequilibrium nozzle flow,” M.S. thesis, Dept. Electron. Eng.,
Osaka Univ., Osaka, Japan, 1993.
[20] J. P. Wilkinson, “Nonlinear resonant circuit devices (Patent style),”
U.S. Patent 3 624 12, July 16, 1990.
[21] IEEE Criteria for Class IE Electric Systems (Standards style), IEEE
Standard 308, 1969.
[22] Letter Symbols for Quantities, ANSI Standard Y10.5-1968.
[23] R. E. Haskell and C. T. Case, “Transient signal propagation in
lossless isotropic plasmas (Report style),” USAF Cambridge Res.
Lab., Cambridge, MA Rep. ARCRL-66-234 (II), 1994, vol. 2.
[24] E. E. Reber, R. L. Michell, and C. J. Carter, “Oxygen absorption in
the Earth’s atmosphere,” Aerospace Corp., Los Angeles, CA, Tech.
Rep. TR-0200 (420-46)-3, Nov. 1988.
[25] (Handbook style) Transmission Systems for Communications, 3rd
ed., Western Electric Co., Winston-Salem, NC, 1985, pp. 44–60.
[26] Motorola Semiconductor Data Manual, Motorola Semiconductor
Products Inc., Phoenix, AZ, 1989.
[27] (Basic Book/Monograph Online Sources) J. K. Author. (year, month,
day). Title (edition) [Type of medium]. Volume (issue). Available:
http://www.(URL)
[28] J. Jones. (1991, May 10). Networks (2nd ed.) [Online]. Available:
http://www.atm.com