EP4706048A2

EP4706048A2 - Wound closure prediction model

Info

Publication number: EP4706048A2
Application number: EP24798094.9A
Authority: EP
Inventors: Felipe LISBOA; Seth A. SCHOBEL-MCHUGH; Eric A. ELSTER; Henry T. ROBERTSON; Michael Rouse; Scott F. GREY
Original assignee: Henry M Jackson Foundation for Advancedment of Military Medicine Inc
Current assignee: Henry M Jackson Foundation for Advancedment of Military Medicine Inc
Priority date: 2023-04-26
Filing date: 2024-04-26
Publication date: 2026-03-11
Also published as: WO2024227036A3; WO2024227036A2; AU2024259919A1

Abstract

The present disclosure describes techniques for generating a wound closure prediction model. The wound closure prediction model may generate a wound closure prediction metric that is indicative of how successful closing a wound during the next debridement of the wound may be. The wound closure prediction model may be generated by using a plurality of variable selection models to select model parameters from reference clinical parameters and reference biomarker parameters. Candidate wound closure prediction models may be generated based on the model parameters, wherein the wound closure prediction model is selected from the candidate models based on a performance metric associated with each candidate model.

Description

WOUND CLOSURE PREDICTION MODEL

STATEMENT REGARDING FEDERALLY SPONSORED RESEARCH OR DEVELOPMENT

[0001] This invention was made with government support under HU00012120029 awarded by the Unifonn Services University of the Health Sciences. The government has certain rights in the invention.

CROSS REFERENCE TO RELATED APPLICATIONS

[0002] This application claims the benefit of U.S. Provisional Patent Application 63/498,513, filed on April 26, 2023, which is hereby incorporated by reference in its entirety.

FIELD OF THE DISCLOSURE

[0003] Described herein are methods, systems, and computational environments for predicting successful wound closure using a wound closure prediction model. Also described are methods, systems, and computational environments for generating the wound closure prediction model.

BACKGROUND

[0004] Trauma wounds, particularly combat-related trauma wounds, can result in complications. Principal amongst these complications is early wound failure (e.g., dehiscence and/or infection) which can occur ranging from 15-30% of those who suffer from trauma wounds. Closing wounds as early as possible can mitigate complications from arising. However, closing a wound too early can similarly result in complications. Therefore, it is important to identify and/or predict as early as possible when a wound can be successfully closed. This will be highly informative not just in a traditional clinical setting, but also in low resource environments and military operations.

SUMMARY OF THE DISCLOSURE

[0005] This Sum man is provided to introduce a selection of concepts in a simplified form that are further described below in the Detailed Description. This Summary is not intended to identify all key features or essential features of the claimed subject matter, nor is it intended to be used alone as an aid in determining the scope of the claimed subject matter.

[0006] Techniques for determining a wound closure prediction metric using the wound closure prediction model and for generating the wound closure prediction model are discussed herein. [0007] In embodiments, there are provided a method for predicting wound closure for a wound of a subject comprising: receiving a first value associated with a clinical parameter associated with the subject and a second value associated with a biomarkcr parameter; executing a wound closure prediction model for the wound using the clinical parameter and the biomarker parameter, wherein the wound closure prediction model is generated by performing operations comprising: generating a data structure storing a plurality of reference clinical parameters associated with a plurality of subjects and a plurality of reference biomarker parameters; executing a plurality of variable selection models on the plurality of reference clinical parameters and the plurality of reference biomarkcr parameters to select a plurality of model parameters, the plurality’ of model parameters comprise a subset of at least one reference clinical parameter of the plurality of reference clinical parameter and at least one reference biomarker parameter of the plurality of reference biomarker parameters; determining a plurality of candidate wound prediction models, each candidate model comprising at least one model parameter of the plurality of model parameters; determining a performance metric for each candidate model of the plurality of candidate models; selecting based on tire performance metric, a candidate model from the plurality of candidate models as the wound closure prediction model for the wound; and outputting, by the wound closure prediction model, a prediction associated with closure of tire wound.

[0008] In embodiments, there are provided a method for generating a wound closure prediction model for a wound of a subject comprising: generating a data structure storing a plurality of clinical parameters associated with a plurality of subjects and plurality of biomarkcr parameters associated with one or more biomarkers; executing a plurality of variable selection models on the plurality of clinical parameters and the plurality of biomarker parameters to select a plurality of model parameters, the plurality of model parameters comprise a subset of at least one clinical parameter of the plurality of clinical parameters and at least one biomarker parameter of the plurality of biomarker parameters; determining a plurality of candidate wound closure prediction models, each candidate model comprising at least one model parameter of the plurality of model parameters; determining a performance metric for each candidate model of the plurality of candidate models; selecting, based on the performance metric, a candidate model from the plurality of candidate models as the wound closure prediction model.

[0009] In embodiments, there are provided a system for predicting wound closure for a wound of a subject comprising: one or more processors; an input component; an output component; one or more computer readable media storing computer-executable instructions that, when executed, causes the one or more processors to perform operations comprising: receiving, from tire input component, a first value associated with a clinical parameter associated with the subject and a second value associated with a biomarker parameter; executing a wound closure prediction model using the clinical parameter and the biomarker parameter to generate a prediction associated with closing the wound, wherein the model is generated by performing operations comprising: generating a data structure storing a plurality of reference clinical parameters associated with a plurality of subjects and a plurality of reference biomarker parameters; executing a plurality of variable selection models on the plurality of reference clinical parameters and the plurality of reference biomarker parameters to select a plurality of model parameters, the plurality of model parameters comprising a subset of at least one reference clinical parameter of the plurality of reference clinical parameters and at least one biomarker parameter of tire plurality of reference biomarker parameters; determining a plurality of candidate wound closure prediction models, each candidate model comprising at least one model parameter of tire plurality of model parameters; determining a performance metric for each candidate model of the plurality of candidate models; selecting, based on the performance metric, a candidate model from the plurality of candidate models as tire wound closure prediction model; and outputting, by the model and at the output component, the prediction associated w ith the closing of the wound.

[0010] In embodiments, there are provided a system for generating a wound closure prediction model for a wound of a subject comprising: one or more processors; one or more non-transitory computer-readable media storing computer-executable instructions that, when executed, cause the one or more processors to perform operations comprising: generating a data structure storing a plurality of clinical parameters associated with a plurality of subjects and a plurality of biomarker parameters; executing a plurality of variable selection models on the plurality of clinical parameters and the plurality of biomarker parameters to select a plurality of model parameters, the plurality of model parameters comprising a subset of at least one clinical parameter of the plurality of clinical parameters and at least one biomarker parameter of the plurality of biomarker parameters; determining a plurality of candidate wound closure prediction models, each candidate model comprise at least one model parameter of the plurality of model parameters; determining a performance metric for each candidate model of the plurality of candidate models; and selecting, based on tire performance metric, a candidate model from the plurality of candidate models as the wound closure prediction model.

[0011] In embodiments, the variable selection models include one or more machine learning algorithms that may include decision trees, random forests, neural networks, logistic regression, support vector machine, Bayesian belief network, LASSO, Relief-based feature selection, Stable Iterative Variable Selection, or Elastic-net. [0012] In embodiments, determining the plurality of candidate models comprises: determining that a first value of at least one biomarker parameter associated with a candidate model of the plurality of candidate models is missing; and estimating, based on the missing first value, the first value using a reference value.

[0013] In embodiments, the plurality of clinical parameters includes wound characteristics. Examples of the wound characteristics include a wound type, a number of wounds, a wound location, a wound length, a wound width, and/or a wound depth.

BRIEF DESCRIPTION OF THE DRAWINGS

[0014] The detailed description is described with reference to the accompanying figures. In the figures, the left-most digit(s) of a reference number identifies the figure in which the reference number first appears. The use of the same reference numbers in different figures indicates similar or identical components or features. The figures are merely exemplary to illustrate certain features that may be used singularly or in combination with other features, and the present disclosure should not be limited to the embodiments shown.

[0015] FIG. 1 illustrates a block diagram of an example system for generating a wound closure prediction model.

[0016] FIG. 2 illustrates a flow diagram of an example process of generating the wound closure prediction model.

[0017] FIG. 3 illustrates a block diagram of an example system for determining a wound closure prediction metric using the wound closure prediction model.

[0018] FIG. 4 illustrates a flow diagram of an example process of using the wound closure prediction model to determine a wound closure prediction metric.

[0019] FIG. 5 illustrates a flow diagram of an example process of detennining a prediction threshold for determining the wound closure prediction metric.

[0020] FIG. 6 illustrates a flow diagram of an example surgical debridement process for a wound.

[0021] FIG. 7 illustrates a flow diagram of an example process of determining wound closure by using a device that includes the wound closure prediction model.

[0022] FIGS. 8A-8C illustrate an example report generated by the wound closure prediction model.

[0023] FIG. 9 illustrates an example process for generating biomarker parameters using the biomarker detection device and an immunoassay. [0024] FIG. 10 illustrates a plate map of a plate used for the immunoassay and a well map for one well of the plate.

[0025] FIG. 11 illustrates an area under the receiver operating characteristic (AUROC) plot associated with the wound closure prediction model and a calibration plot associated with the wound closure prediction model

[0026] FIG. 12 illustrates an example clinical impact plot associated with the wound closure prediction model.

[0027] FIG. 13 illustrates an example table of candidate wound prediction models.

[0028] FIG. 14 illustrates an example table of fit indices for the clinically relevant subgroups for the wound closure prediction model.

[0029] FIG. 15 illustrates an example table of simulated wound closure rates for current practice and estimated wound closure rates from tire wound closure prediction model at the selected prediction threshold.

DETAILED DESCRIPTION

[0030] Tire following detailed description is presented to enable any person skilled in the art to make and use the subject of the application. For purposes of explanation, specific nomenclature is set forth to provide a thorough understanding of the present disclosure. However, it will be apparent to one skilled in the art that these specific details are not required to practice the subject of the application. Descriptions of specific applications are provided only as representative examples. The present application is not intended to be limited to the embodiments shown, but is to be accorded the widest possible scope consistent with the principles and features disclosed herein.

[0031] The present disclosure provides methods for predicting whether a wound may be successfully closed. In embodiments, the wound may be of a

[0032] Technical and scientific terms used herein have the meanings commonly understood by one of ordinary skill in the art to which the present disclosure pertains, unless otherwise defined. [0033] As used herein, tire singular forms “a,” “an," and “tire" designate both the singular and the plural, unless expressly stated to designate the singular only.

[0034] As used herein the terms “marker" and “biomarkers" are used interchangeably to refer to a measurable substance from a biological sample. For example, these can comprise one or more protein data markers, one or more nucleic acid data markers, one or more metabolite data markers, or a combination thereof. [0035] As used herein, the term “pre-debridement” refers to a period of time before the debridement of a wound of a subject.

[0036] As used herein, the term ‘‘post-debridement” refers to a period of time after the debridement of a wound of a subject.

[0037] As used herein, the term “overnight” refers to a period of time of at least 12 hours.

[0038] As used herein, the term “serum” refers to fluid that remains from blood plasma after fibrinogen, prothrombin, and other clotting factors have been removed or fluid from blood plasma without fibrinogen, prothrombin, and other clotting factors.

[0039] As used herein, the term “effluent” refers to fluid and/or other drainages from a wound.

[0040] As used herein, the term “serum biomarker” refers to a biomarker derived, extracted, or otherwise obtained from the serum.

[0041] As used herein, the term “effluent biomarker” refers to a biomarker derived, extracted, or otherwise obtained from the effluent.

[0042] Examples of serum biomarkers may be, but are not limited to, basic fibroblast growth (FGFBASIC), epidermal growth factor (EGF), eotaxin-1 (CCL-11), eotaxin-3 (CCL-26), fms related receptor tyrosine kinase- 1 (FLT-1), granulocyte-colony stimulating factor (GCSF), granulocyte-monocyte colony stimulating factor (GMCSF). hepatocyte growth factor (HGF), interferon alpha-2 (IFNa2a). interferon gamma (IFNy), interleukin- 10 (IL-10), interleukin- 12/interleukin-23p40 (IL-12/IL-23p40), interleukin- 15 (IL-15), interleukin- 16 (IL-16), interleukin- 17A (IL-17A), interleukin-1 receptor antagonist (IL1RA), interleukin-22 (IL22), interleukin-2 receptor subunit alpha (IL-2RA), interleukin-3 (IL-3), interleukin-5 (IL-5), interleukin-6 (IL-6), interleukin-7 (IL-7), interleukin-8 (IL-8), interleukin-9 (IL-9), interferon gamma-inducing protein- 10 (IP- 10), monocyte chemoattractant protein- 1 (MCP-1). monocyte chemoattractant protein-4 (MCP-4), macrophage-derived chemokine (MDC), monokine induced by gamma interferon (MIG, CXCL9), macrophage inflammatory? protein- 1 b (MIP- 1 B), placental growth factor (PLGF), thy mus and activation regulated chemokine (TARC), angiopoientin-1 receptor (TIE-2), tumor necrosis factor alpha (TNFA), vascular endothelial growth factor (VEGF, VEGFA), vascular endothelial growth factor C (VEGFC), and/or vascular endothelial growth factor D (VEGFD).

[0043] Examples of effluent biomarkers include FGFBASIC, FLT-1, GMCSF, IFNy, IL- 10, IL- 12/IL-2pP40, IL- 15, IL- 16, IL- 17 A, interleukin- 1 alpha (IL-1 A), interleukin- 1 beta (IL- IB), IL- 2RA, IL-5, IL-7, MDC, MIG, PLGF, TARC, TIE-2, VEGFA, VEGFC, and/or VEGFD. [0044] As used herein, the term “reference clinical parameter’’ refers to one or more parameters associated with one or more subjects being used to generate tire wound closure prediction model.

[0045] As used herein, the term “reference biomarker parameter" refers to one or more parameters associated with one or more biomarkers being used to generate the wound closure prediction model.

[0046] As used herein, the term “clinical parameter” refers to one or more parameters associated with a subject or patient that are used by the wound closure prediction model to generate a prediction metric.

[0047] As used herein, the term “biomarker parameter” refers to one or more parameters associated with one or more biomarkers that are used by the wound closure prediction model to generate a prediction metric. In embodiments, the biomarker parameter may be associated with one or more biomarkers that arc associated with a wound that the wound closure prediction model is being used for.

[0048] As used herein, the term “successful wound closure” is defined as a wound that has been closed by a direct approximation of wound borders by suture, flaps, skin grafts, or other types of wound closing techniques and does not require re-exploration or a return to an operation room or another clinical environment for additional surgical debridement.

[0049] As used herein, the term “debridement.” “wound debridement,” and “surgical debridement” refers to the medical removal of dead, damaged, and/or infected tissue associated with a wound to improve the healing potential of remaining healthy tissue associated with the wound.

[0050] As used herein, the term “injury severity score (ISS)” refers to a medical score for assessing trauma severity. The ISS may correlate with mortality, morbidity, and hospitalization time after trauma.

[0051] As used herein, the term “wound size” refers to a wound area, a wound depth, a wound width, and a wound depth associated with a wound. A unit of measurement for the wound size may include centimeters, inches, or other units of measurement. For example, the wound area may be measured in squared centimeters.

[0052] As used herein, the term “wound length” refers to a longest longitudinal extension of the wound.

[0053] As used herein, the term “wound width” refers to a longest extension perpendicular to the wound length. [0054] As used herein, the term “wound depth” refers to a distance between a deepest point and a surface of the wound.

[0055] As used herein, the term “wound characteristic” refers features that describe the severity of wounds such as a number of wounds, the wound size, a wound location, and/or a wound type.

[0056] FIG. 1 illustrates a block diagram of an example system 100 for generating a wound closure prediction model. The example system 100 includes a wound specimen collection device 102, a biomarker detection device 104, first computing device(s) 106, a wound measurement device 122. and second computing device(s) 124. The first computing device(s) includes an input component 108. first processors) 110. a memory 112, a clinical parameter component 114, a biomarker parameter component 116, a prediction model generating component 118, and an output component 120. The second computing device(s) 124 includes second processor(s) 126, a memory 128, and a prediction model component 130. In embodiments, a wound closure prediction model may be generated at the first computing device(s) 106 and transferred to the second computing device(s) 124. Alternatively, tire wound closure prediction model may be generated at the second computing device 124.

[0057] Tire wound specimen collection device 102 may be a device that is positioned at or proximate a wound of a subject to collect biological sample(s) from the wound. In embodiments, the biological sample(s) may be fluid or other types of discharge (e.g., blood, effluent, debris, and/or the like). In embodiments, the wound specimen collection device 102 may be a vacuum device such as a vacuum canister that is placed proximately or at the wound to collect the biological sample(s) over time. In embodiments, the wound specimen collection device 102 may be used to collect the biological sample(s) for a first period of time before debridement of the wound and/or for a second period of time after the debridement. For example, the wound specimen collection device 102 may be positioned at or proximate the wound for a first period of time of 12 or more hours and for a second period of time of 2 or more hours.

[0058] In embodiments, the biological sample(s) collected by wound specimen collection device 102 may be transferred to a biomarker detection device 104. The biomarker detection device 104 may detect, identify, and quantify one or more biomarkers from a biological sample of the biological samplc(s) as biomarkcr paramctcr(s). The biomarkcr paramctcr(s) may be detected, identified, and quantified (in other words, determined) by analyzing effluents or serums collected from the wound, and wherein analyzing tire effluents or serums comprises performing an immunoassay, the immunoassay comprising one or more of electrochemiluminescence assay, affinity capture assay, immunometric assay, heterogeneous chemiluminescence immunometric assay, homogeneous chemiluminescence immunometric assay, ELISA, western blotting, radioimmunoassay, magnetic immunoassay, real-time immunoquantitative polymerase chain reaction (iqPCR), and surface-enhanced Raman spectroscopy (SERS) label-free assay.

[0059] In embodiments, the quantifying of the one or more biomarkers may include determining a concentration of that biomarker. In embodiments, the biomarker detection device 104 may detect, identify, and quantify (e.g., determine a concentration of) the biomarker during immunoassay. For example, the biological sample(s) may be added to an immunoassay plate that is pre-coated with a capture antibody designed to detect the biomarker, where the plate is then incubated. A detection antibody may be added to the incubated plate, and the plate may be subject to additional incubation. A read buffer may then be added to the plate, and the plate may be analyzed by the biomarker detection device 104 to detect, identify, and quantify the biomarker. In embodiments, when performing the detection, identification, and quantification, the biomarker detection device 104 may further determine or receive additional biomarker parameter(s) such as, but are not limited to a biomarker type (e.g., serum, effluent, or the like), a biomarker time period (e.g., collected from the first period of time pre-debridement or from the second period of time post-debridement), and/or the like.

[0060] In embodiments, as each biomarker is detected, identified, and quantified by the biomarker detection device 104, biomarker parameter(s) associated with the biomarker (e.g., biomarker name, the quantity, the biomarker type, and the biomarker time period) may be transmitted or transferred to the first computing dcvicc(s) 106 and stored at the biomarkcr parameter component 116. In embodiments, the biomarker detection device may include a communication component that can transmit the biomarker parameter(s) to the first computing device(s) 106 using a wired connection or a wireless connection. The first computing device(s) 106 may receive the biomarker parameters) using its communication component. Alternatively, biomarker parameter(s) may be transmitted or transferred to the first computing device(s) 106 after all biomarkcr parameters associated with the biological samplc(s) collected from the wound arc determined. In embodiments, the biomarker parameter(s) received by the first computing device(s) 106 from the biomarker detection device 104 may be used as reference biomarker(s) for generating the wound closure prediction model.

[0061] In embodiments, the first computing device(s) 106 is where the wound closing prediction module is being generated. As illustrated, the first computing device(s) 106 generates the wound closure prediction model, and transmits or transfers the wound closure prediction model to the second computing device(s) 124 where the wound closure model is used to generate a prediction metric for successful wound closure. Alternatively, the first computing device(s) 106 may generate and utilize the wound closure prediction model. The wound closure prediction model may be generated by a single first computing device 106 or multiple computing devices 106 (e.g., by a server cluster).

[0062] In addition to the reference biomarker parameter(s) being received from the biomarker detection device 104 and stored at the biomarker parameter component 116 of the memory 112, the first computing device(s) 106 may receive or determine reference clinical parameter(s) which may be stored at the clinical parameter component 114 of the memory 112. Reference clinical parameters may be received using an input component 108 or multiple input components 108 (e.g., a mouse and/or a keyboard). For example, clinical parameters such as subject name, subject gender, a number of wounds, a wound type, a wound location, and/or the like, may be received using the input component 108.

[0063] In embodiments, the first computing device(s) 106 may receive wound characteristic data from a wound measurement device 122. The wound measurement device 122 may be a physical measurement device (e.g. using a ruler) or an image based measurement device (e.g., determining a size of the wound by taking and using an image of the wound using an imaging component such as a camera). The wound measurement device may be used to measure to determine the wound characteristic data such as a wound length, a wound width, a wound depth, these wound characteristic data may be transmitted or transferred to the first computing device(s) 106 via a wired or wireless connection such as a wired or wireless connection between the communication component of the first computing dcvicc(s) 106 and a communication component of the wound measurement device 122 as reference clinical parameters. In embodiments, the wound measurement device 122 may further determine the wound location and/or the number of wounds (e.g., by using the imaging component) and transmit or transfer to the first computing device(s) 106 the wound location and the number of wounds as reference clinical parameters.

[0064] In embodiments, the reference clinical parameter(s) may be stored at the clinical parameter component 114 in a first data structure (e.g., a database) and the reference biomarker parameter(s) may be stored at the biomarker parameter component 116 in a second data structure. Alternatively, the reference clinical parameter(s) and the reference biomarker parameter(s) may be stored at the memory 112 in a single data structure. In embodiments, the processor(s) 110 may be used to generate the wound closure prediction model at the prediction model generating component using the reference clinical parameter(s) and the reference biomarker parameter(s). Generating the wound closure prediction model is described in additional detail at FIG. 2, as well as throughout this disclosure. In embodiments, the generated wound closure prediction model may be stored at the prediction model generating component 118 or at another location within the memory 112. The wound closure prediction model being stored within the memory 112 may be transferred or transmitted to the prediction model component 130 within the memory 128 of the second computing device(s) 124. The processor(s) 126 may use the wound closure prediction model to determine a prediction metric on a successful wound closure of a wound of the subject. Using the prediction model to determine tire prediction metric on the successful wound closure of a wound of the subject is discussed in additional detail in FIGS. 3 and 4, as well as throughout this disclosure. [0065] FIG. 2 illustrates a flow diagram of an example process 200 of generating a wound closure prediction model. In embodiments, some or all of process 200 can be performed by one or more components described in association with FIG. 1, as described herein. Additionally, some portions of process 500 can be omitted, replaced, and/or reordered. For example, the example process 200 may be performed by tire processor(s) 110 of the first computing device(s) 106. and the generated wound closure prediction model may correspond to the generated wound prediction model detailed in association with FIG. 1.

[0066] At operation 202, the process can include receiving reference clinical parameters. In embodiments, the reference clinical parameters may correspond to the reference clinical parameters described in association with FIG. 1, as well as throughout this disclosure. In embodiments, the reference clinical parameters may include patient parameters and wound parameters. Examples of the patient parameters may include, but are not limited to, patient name, patient age, patient gender, and/or patient physical characteristics (e.g., height, weight, blood pressure, blood cell count, platelet count, sodium levels, potassium levels, glucose levels, and/or the like). Examples of the wound parameters may include, but are not limited to wound type, wound location, wound length, wound width, wound depth, and/or the like. The reference clinical parameters may be stored within a data structure within the memory 112.

[0067] At operation 204, tire process can include receiving reference biomarker parameters. In embodiments, the reference biomarker parameters may correspond to the reference biomarker parameters described in association with FIG. 1, as well as throughout this disclosure. Examples of the reference biomarker parameters are discussed in association with the Example section, as well as throughout this disclosure. The reference biomarker parameters may be stored within the same data structure as the reference clinical parameters or within a separate data structure at the memory 112.

[0068] In embodiments, the operation 204 may further include calculating a coefficient of variation (CV) for each biomarker for a first quantity or a first concentration of the biomarker collected at the first period of time before the debridement of the wound and a second quantity or a second concentration of the biomarker collected at the second period of time after the debridement of the wound as a reference biomarker parameter. The CV may be calculated with the following equation:

^Btomarker being the standard deviation of the concentration of the biomarker measured over the first period of time and the second period of time and I^Biomarker being tire mean of the concentration of the biomarker measured over the first period of time and the second period of time.

[0069] In embodiments, the operations 204 may further include calculating a difference (DIFF) between the first quantity or first concentration and the second quantity and the second concentration for each biomarker.

[0070] In embodiments, at the operations 202 and/or 204, at least some of the reference clinical parameters and/or the reference biomarker parameters may be preconditioned prior to operations 206 and 208. For example, the reference clinical parameters and/or the reference biomarker parameters may be log-transformed and standardized. In embodiments, at least one of the reference clinical parameters may be a dichotomous parameter. A dichotomous parameter may be preconditioned by numerically coding to 0 or 1. Preconditioning the parameters is further discussed in association with the Examples section, as well as throughout this disclosure.

[0071] At operation 206, the process can include determining a first candidate wound closure prediction model. In embodiments, the first candidate model may include first model parameters. In embodiments, the first model parameters may be a first subset that includes at least one of the reference clinical parameters and/or at least one of the reference biomarker parameters.

[0072] In embodiments, the operation 206 may further include executing one or more variable selection models on the reference clinical parameters and/or the reference biomarker parameters to determine the first model parameters. Executing the variable selection models to determine the first model parameters is further discussed in association with the Examples section, as well as throughout this disclosure. In embodiments, the variable selection model may select at least one parameter to be included within the first model parameters that are different from at least one parameter selected to be included within a second model parameter associated with a second candidate wound closure prediction model.

[0073] In embodiments, the first candidate model may calculate a first probability of successful wound closure based on the first model parameters. For example, the first candidate model may calculate the first probability of successful wound closure using the elastic-net algorithm. In embodiments, the first candidate prediction model may include a first logistic regression model.

[0074] At operation 208, the process can include determining the second candidate wound closure prediction model. In embodiments, the second candidate model may include second model parameters. In embodiments, the second model parameters may be a second subset that includes at least one of the reference clinical parameters and/or at least one of the reference biomarker parameters.

[0075] In embodiments, one or more variable selection models may be executed on the reference clinical parameters and/or the reference biomarker parameters to determine the second model parameters. Executing the variable selection models to determine tire second model parameters is further discussed in association with the Examples section, as well as throughout this disclosure.

[0076] In embodiments, the second candidate model may calculate a probability of successful wound closure based on the first model parameters. For example, the second candidate model may calculate the probability of successful wound closure using the elastic-net algorithm. In embodiments, the second candidate prediction model may include a second logistic regression model.

[0077] At operation 210, the process can include determining first performance metric(s) associated with the first candidate prediction model. In embodiments, the first perfonnance metric(s) may be indicative of the performance of the first candidate model at predicting a successful wound closure using the first model parameters. Examples of the first performance metric(s) may include, but are not limited to, a first discrimination associated with the first candidate model, first calibration values associated with the first candidate model, and a first overall accuracy score associated with the first candidate model. In embodiments, the first discrimination values may include a first area under the receiver operator curve (AUROC), tire first calibration values may include a first intercept and a first slope and the first overall accuracy score may include a first Scaled Brier Score.

[0078] At operation 212, the process can include determining second performance metric(s) associated with the second candidate prediction model. In embodiments, the second performance mctric(s) may be indicative of the performance of the second candidate model in predicting a successful wound closure using the second model parameters. Examples of the second perfonnance metric(s) may include, but are not limited to a second discrimination associated with the second candidate model, second calibration values associated with the second candidate model, and a second overall accuracy score associated with the second candidate model. In embodiments, the second discrimination values may include a second AUROC, the second calibration values may include a second intercept and a second slope and the second overall accuracy score may include a Scaled Brier Score

[0079] At operation 214, the process can include detennining whether the first perfonnance metric is indicative of a better fit than the second performance metric. For example, the process can include determining whether the first AUROC value is greater than the second AUROC value. If the first AUROC value is greater than the second AUROC value, then the first candidate prediction model may be determined as being better at discriminating between wounds ready for closure from those that arc not than tire second candidate prediction model. In such a scenario, the process may proceed to operation 218 to select the first candidate model as the wound closure prediction model. If the first AUROC is less than the second AUROC, then tire process may proceed to operation 216 to select the second candidate prediction model as the wound closure prediction model. In embodiments, the calibration values may be considered with the AUROC values to determine which candidate model is more likely to accurately predict a successful wound closure across different wounds.

[0080] At operation 220, the process can include using the wound closure prediction model to determine a prediction metric associated with the successful closure of the wound. In embodiments, the prediction metric may be a probability or a probability score associated with successfully closing the wound.

[0081] While FIG. 2 illustrates that tire first candidate wound closure prediction model and the second candidate wound prediction model are determined, additional candidate wound closure prediction models beyond the first and second candidate prediction models may be determined (e.g., Nth candidate model, N+l candidate model, N+2 candidate model, and so forth), each with their own model parameters that differ from the first candidate prediction model and the second candidate prediction model. Additional perfonnance metrics may be determined for each additional candidate model, and these additional perfonnance metrics may be used to select the wound closure prediction model from all candidate prediction models.

[0082] In embodiments, the wound closure prediction model may be evaluated within a plurality of clinically important subgroups. Examples of the clinically important subgroups include single wound versus multiple wound, injury severity score above a severity threshold versus injury' severity score below a severity threshold, wound type, wound area above a size threshold versus wound area below a size threshold (e.g., above versus below 250 cm³), wound location (e.g.. upper versus lower extremity), and/or injury mechanism (e.g., blast, gunshot, and/or the like). Performance metric(s) may be determined for the wound closure prediction model for each subgroup. These performance metric(s) may correspond to the same types of metrics from the operation 214. In embodiments, these performance metric(s) may be indicative of scenarios where the wound closure prediction model can accurately predict successful wound closure. Evaluating the wound closure prediction model using the clinically important subgroups is further discussed in association with the Example section, as well as throughout this disclosure.

[0083] In embodiments, each first computing device of the computing devices 106 may select a wound closure prediction model. Each wound prediction model from each computer device 106 may be evaluated for prediction accuracy based on their respective performance metric.

[0084] In embodiments, the wound closure prediction model may be evaluated using simulations. The simulation may use simulated patient data, wound data, and biomarker data to generate prediction metrics for simulated wounds. The prediction metrics may be compared to expected results to evaluate the performance of the wound closure performance model. For the simulations, the wound closure prediction model is discussed in association with FIGS. 3 and 4 may be utilized.

[0085] FIG. 3 illustrates a block diagram of an example system 300 for using a wound closure prediction model to determine a wound closure prediction metric for a wound of a subject. In embodiments, the wound closure prediction model may correspond to the wound closure prediction model generated as described in association with FIGS. 1 and 2. The example system 100 includes a wound specimen collection device 302, a biomarker detection device 304, a computing device 306, and a wound measurement device 322. In embodiments, the wound specimen collection device 302 may correspond to the wound specimen collection device 102 of FIG. 1, the biomarker detection device 304 may correspond to the biomarker detection device 104 of FIG. 1. the computing device may correspond to a computing device of the second computing device(s) 124 of FIG. 1, and the wound measurement device may correspond to the wound measurement device 122 of FIG. 1. The computing device 306 includes an input component 308, processor(s), memory 312, which further includes a clinical parameter component 314, a biomarker parameter component 316. and a prediction model component 318, and an output component 320.

[0086] In embodiments, the wound specimen collection device 302 may collect biological samples, such as effluents, from the wound of the subject as described in association with FIG. 1 . In embodiments tire wound specimen collection device may be utilized to collect first biological samples for a first period of time before debridement of tire wound and second biological samples for a second period of time after the debridement of the wound. The biomarker detection device 304 may determine biomarker parameters associated with the collected biological sample as described in association with FIG. 1. [0087] In embodiments, the biomarker detection device may only determine biomarker parameters that are relevant to model parameters associated with the wound closure prediction model. For example, the model parameters may be associated with concentrations or quantities of GM-CSF, IL-5, IL-7, IL-12/IL-23p40, IL-15, and IL-17A within effluents collected from the first period of time and the second period of time, and tire biomarker detection device 304 may determine just the concentration or quantity of GM-CSF, IL-5, IL-7, IL-12/IL-23p40, IL-15, and IL-17A within effluents collected from the first period of time and the second period of time from the biological samples.

[0088] In embodiments, the wound measurement device may measure and/or determine wound parameters as described in association with FIG. 1. In embodiments, similarly, to those described in association with the biomarker detection device 304. the wound measurement device may only measure and/or determine wound parameters associated with the model parameters. For example, if the relevant model parameters are wound length and wound width, then the wound measurement device may measure and/or determine just the wound length and wound width.

[0089] In embodiments, the computing device 306 may receive the wound parameters from the wound measurement device 322 and the biomarker parameters from the biomarker detection device 304. The computing device 306 may receive these parameters using a communication component or the input component 308. The wound measurement device and the biomarker detection device may transmit their respective parameters using their own respective communication components in communication with the communication component of the computing device 306. The wound parameters may be stored at the clinical parameter component 314 and the biomarker parameters may be stored at the biomarker parameter component 316. Alternatively, the wound parameters and the biomarker parameters may be stored together at a model parameter component. In embodiments, the wound parameters may be a subset of clinical parameters.

[0090] In embodiments, the input component may include a mouse, a keyboard, a touch screen, and/or the like. In embodiments, a user may utilize the input component to supplement clinical parameter(s) and/or biomarker parameter(s) that are model parameters, but were not determined by the biomarker detection device 304 and/or the wound measurement device 322. For example, the user may input a wound type and/or a wound location as clinical parameters.

[0091] In embodiments, the processor(s) 310 may execute the wound closure prediction model stored at tire prediction model component 318 to determine the wound closure prediction metric. Determining the wound closure prediction metric(s) is described in association with FIG. 4, as well as throughout this disclosure. [0092] In embodiments, the output component 320 may be a display used to output a report or a recommendation associated with the wound closure prediction metric and/or the wound closure prediction metric to a user such as a clinician and/or a surgeon performing tire debridement. The report, recommendation, and/or the wound closure prediction metric may further be transmitted to a user device of the clinician and/or the surgeon (e.g., to a mobile phone, a tablet, a personal computer, and/or the like) using the communication component of the computing device 306.

[0093] FIG. 4 illustrates a flow diagram of an example process 400 associated with using a wound closure prediction model to determine a wound closure prediction metric. In embodiments, some or all of the process 400 can be performed by one or more components described in association with FIG. 3, as described herein. Additionally, some portions of process 400 can be omitted, replaced, and/or reordered. In embodiments, the wound closure prediction model may be the wound closure prediction model described in association with FIGS. 1-3. In embodiments, the wound closure prediction metric may be indicative of how likely a successful closure of the wound is. In addition to what is discussed below in association with FIG. 4, using the wound closure prediction model to determine tire wound closure prediction metric is further discussed in association with the Examples section, as well as throughout this disclosure.

[0094] At operation 402, the process can include receiving clinical parameter(s). In embodiments, the clinical parameter(s) may be clinical parameters that are associated with model parameter(s) of the wound closure prediction model. For example, the model parameters may include wound length and wound width, and therefore, the clinical parameters received wound include wound length and wound width. The clinical parameter(s) may be received using the input component 308 and/or the wound measurement device 322 of FIG. 3.

[0095] At operation 404, the process can include receiving biomarker param eter(s). In embodiments, the biomarker parameter(s) may be biomarker parameters that are associated with the model parameter(s) of the wound closure prediction model. The biomarker parameter(s) may be received using the input component 308 and/or the biomarker detection device 304 of FIG. 3.

[0096] In embodiments, upon receiving the clinical parameter(s) and the biomarker parameter(s), the process can further include determining whether one or more values of the clinical parameter(s) and/or the biomarker parameter(s) are missing. Upon detecting any missing values, the process can further include imputing any missing values. In embodiments, imputing the missing values may include estimating the first value using a reference value. In embodiments, while this imputation method is discussed as being used while using the wound closure prediction model, this imputation method or a similar imputation method may also be used while generating the model parameters for the candidate models described in association with FIGS. 1 and 2 to estimate missing values within the reference clinical parameters, the reference biomarker parameters, and/or the model parameters.

[0097] At operation 406, the process can include determining tire wound closure prediction metric. In embodiments, the wound closure prediction metric may include a probability of successful wound closure. Values of the clinical parameter(s) and the biomarker parameter! s) maybe applied to the wound closure prediction model to generate the wound closure prediction metric. In embodiments, prior to using the wound closure prediction metric, the clinical parameter(s) and the biomarker parameter(s) may be preconditioned as described in association with FIG. 2, as well as throughout this disclosure. In embodiments, tire wound closure prediction metric comprises a probability or a probability score for the successful closure of the wound.

[0098] At operation 408, the process can include determining whether the wound closure prediction metric is at or above a prediction threshold. In embodiments, the prediction threshold is a threshold indicative of whether a recommendation to close tire wound at the next debridement of the wound should be provided. Therefore, in a scenario where the prediction metric is at or above the prediction threshold (e.g., tire probability of successful wound closure is at or above a threshold probability; the probability threshold is also referred to as a close threshold), then the process may continue to operation 410. However, in a scenario where the prediction metric is below the prediction threshold, then the process may continue to operation 412.

[0099] At the operation 410, the process can include generating a first report recommending closing the wound at the next debridement of tire wound. In embodiments, the first report may include the wound closure prediction metric and a first recommendation that tire wound be closed. The operation 410 may further include determining a status of the closed wound (e.g.. whether the wound was successfully closed). The clinical parameter(s), the biomarker parameter(s), and the status of tire closed wound may be used to update the model parameter(s) of the prediction model to increase an accuracy of tire prediction model.

[00100] At the operation 412, the process can include generating a second report recommending to keep the wound open at the next debridement of the wound. In embodiments, the second report may include the wound closure prediction metric and a second recommendation that the wound be kept open.

[00101] In embodiments, rather than generating a first report or a second report, the process can alternatively include generating a single report where the report includes the wound closure prediction metric and both recommendations.

[00102] In embodiments, the first report or the second report may be provided to a user such as a clinician and/or a surgeon performing the debridement to assist the clinician and/or the surgeon in determining whether to close the wound at the next debridement as described in association with FIG. 3. In embodiments, the prediction threshold may be determined by the processor(s) 310 of FIG. 3 and stored at the memory' 312 of FIG. 3.

[00103] FIG. 5 illustrates a flow diagram of an example process 500 of determining a prediction threshold for determining the wound closure prediction metric. In embodiments, the prediction threshold may correspond to the prediction threshold described in association with FIG. 4. Determining the prediction threshold is further discussed in association with the Examples section, as well as throughout this disclosure.

[00104] At operation 502, the process can include receiving first wound data. In embodiments, the first wound data may represent data from 100 patients, though other numbers of patients are also contemplated. The first wound data may include real data and/or simulated data. The first wound data may include all model parameters of the wound closure prediction model.

[00105] At operation 504, the process can include instantiating a first simulation using the first wound data and the prediction models at different candidate prediction probability thresholds. In embodiments, the candidate prediction probability thresholds may include, but are not limited to a range between 0.6 and 1. In embodiments, the simulation may include generating wound closure prediction metrics for all first wound data at each prediction threshold. For example, when instantiating the first simulation at a prediction threshold of 0.6, when the wound closure prediction model outputs a wound prediction closure metric that is at or greater than 0.6, then the first simulation would simulate the closing of the wound. The first simulation would then simulate whether the wound was successfully closed and log the result.

[00106] At operation 506, the process can include selecting the prediction threshold based on the first simulation. In embodiments, selecting the prediction threshold can include determining a ratio or percentage of successfully closed wounds to all closed wounds, w herein all closed w ounds include wounds that were successfully closed and wounds that were closed but later faced complications such as dehiscence or infection, where the prediction threshold may be selected as a value that meets or exceeds a ratio or percentage threshold. The selected prediction threshold may be selected based on its value exceeding the ratio or percentage threshold, but is the closest in value to the ratio or percentage threshold.

[00107] At operation 508, the process can include receiving second wound data and reference wound closure data. In embodiments, second wound data may include a similar type of data as the first wound data, but may include more values than the first wound data. In embodiments, the reference wound closure data may be real-world wound closure data associated with the second wound data (e g., real-world data of wounds that were closed and whether the closed wounds were successfully closed). The reference wound closure data may include, but are not limited to, an amount of closed wounds, an amount of successfully closed wounds, and a percentage or ratio of successful wound closures.

[00108] At operation 510, the process can include instantiating a second simulation using second wound data and the wound closure prediction model at the determined or selected prediction threshold. In embodiments, the second simulation can simulate, log, and aggregate a second number of wounds closed, a second amount of successful wound closures, and a second percentage or ratio of successful wound closures. In embodiments, the first number of wounds closed and the second number of wounds closed may correspond to a same value while the first amount of successful wound closures and the second amount of successful wound closure may correspond to different values, and as a consequence, the first percentage or ratio of successful wound closures and the second percentage or ratio of wound closures may be different.

[00109] At operation 512, the process can include verifying the prediction threshold based on the second simulation and the reference wound closure data. In embodiments, verifying the prediction threshold can include determining whether the second amount of successful wound closure is greater than the first amount of successful wound closure and/or the second percentage or ratio is greater than the first percentage or ratio. For example, the prediction threshold may be successfully verified if the second amount is greater than the first amount or the second percentage is greater than the first percentage.

[00110] While one or more examples of the techniques described herein have been described, various alterations, additions, permutations, and equivalents thereof are included within the scope of the techniques described herein.

[00111] In the description of examples, reference is made to the accompanying drawings that form a part hereof, which show by way of illustration specific examples of the claimed subject matter. It is to be understood that other examples may be used and that changes or alterations, such as structural changes, may be made. Such examples, changes or alterations are not necessarily departures from the scope with respect to the intended claimed subject matter. While the steps herein may be presented in a certain order, in some cases the ordering may be changed so that certain inputs are provided at different times or in a different order without changing the function of the systems and methods described. The disclosed procedures could also be executed in different orders. Additionally, various computations that are herein need not be performed in the order disclosed, and other examples using alternative orderings of tire computations could be readily implemented. In addition to being reordered, the computations could also be decomposed into subcomputations with the same results. [00112] Although the subject matter has been described in language specific to structural features and/or methodological acts, it is to be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or acts described. Rather, the specific features and acts are disclosed as example forms of implementing the claims.

[0113] The components described herein represent instructions that may be stored in any type of computer-readable medium (also referred to as a computer-readable storage medium or computer-readable storage medium) and may be implemented in software and/or hardware. All of the methods and processes described above may be embodied in, and fully automated via, software code modules and/or computer-executable instructions executed by one or more computers or processors, hardware, or some combination thereof. Some or all of the methods may alternatively be embodied in specialized computer hardware. A computer-readable medium may be. for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of tire computer-readable medium would include the following: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory' (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the context of this document, a computer readable storage medium may be any tangible medium that can contain or store a program for use by or in connection with an instruction execution system, apparatus, or device. In embodiments, the memories 112, 128. and 312 of FIGS. 1 and 3 respectively may be computer-readable mediums.

[0114] Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing. Computer program code for carrying out operations for aspects of the present disclosure may be written in any combination of one or more programming languages, including an object-oriented programming language such as Java, Smalltalk, C++, or the like, and conventional procedural programming languages, such as the “C” programming language or similar programming languages. The program code may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer, and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection may be made to an external computer (for example, through the Internet using an Internet Service Provider).

[00115] Conditional language such as, among others, “may,” “could,” “may” or “might,” unless specifically stated otherwise, are understood within the context to present that certain examples include, while other examples do not include, certain features, elements and/or steps. Thus, such conditional language is not generally intended to imply that certain features, elements, and/or steps are in any way required for one or more examples or that one or more examples necessarily include logic for deciding, with or without user input or prompting, whether certain features, elements and/or steps arc included or arc to be performed in any particular example.

[00116] Conjunctive language such as the phrase “at least one ofX, Y or Z,” unless specifically stated otherwise, is to be understood to present that an item. term. etc. may be either X, Y. or Z. or any combination thereof, including multiples of each element. Unless explicitly described as singular, “a” means singular and plural.

[00117] Any routine descriptions, elements, or blocks in the flow diagrams described herein and/or depicted in the attached figures should be understood as potentially representing modules, segments, or portions of code that include one or more computer-executable instructions for implementing specific logical functions or elements in the routine. Alternate implementations are included within the scope of the examples described herein in which elements or functions may be deleted, or executed out of order from that shown or discussed, including substantially synchronously, in reverse order, with additional operations, or omitting operations, depending on the functionality involved as would be understood by those skilled in the art.

[00118] Many variations and modifications may be made to the above-described examples, the elements of which are to be understood as being among other acceptable examples. All such modifications and variations are intended to be included herein within the scope of this disclosure and protected by the following claims.

[00119] Aspects of the present disclosure are described herein with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the disclosure. It will be understood that each block of the flowchart illustrations and/or block diagrams, and combinations of blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general-purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions/acts specified in the flowchart and/or block diagram block or blocks.

[00120] These computer program instructions may also be stored in a computer readable medium that can direct a computer, other programmable data processing apparatus, or other devices to function in a particular manner, such that the instructions stored in the computer readable medium produce an article of manufacture including instructions which implement the ftmction/act specified in the flowchart and/or block diagram block or blocks. The computer program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other devices to cause a scries of operational steps to be performed on the computer, other programmable apparatus, or other devices to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide processes for implementing the functions/acts specified in the flowchart and/or block diagram block or blocks. [00121] The flowchart and block diagrams in tire Figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to embodiments of the present disclosure. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the blocks may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems that perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.

[00122] Although the figures show a specific order of method steps, the order of the steps may differ from what is depicted. Also, two or more steps may be performed concurrently or with partial concurrence. Such variation will depend on the software and hardware systems chosen and on the designer's choice. All such variations are within the scope of the disclosure. Likewise, software implementations could be accomplished with standard programming techniques with rule-based logic and other logic to accomplish the various connection steps, processing steps, comparison steps, and decision steps.

[00123] As will be understood by one of ordinary skill in the art. each embodiment disclosed herein can comprise, consist essentially of, or consist of its particular stated element, step. ingredient, or component. Thus, the terms '‘include” or ‘'including” should be interpreted to recite: “comprise, consist of, or consist essentially of.” Tire transition term “comprise” or '‘comprises” means includes, but is not limited to, and allows for the inclusion of unspecified elements, steps, ingredients, or components, even in major amounts. The transitional phrase “consisting of’ excludes any element, step, ingredient, or component not specified. The transition phrase “consisting essentially of’ limits the scope of the embodiment to the specified elements, steps, ingredients, or components and to those that do not materially affect the embodiment.

EXEMPLARY EMBODIMENTS

1: A method of generating a wound closure prediction model for a wound of a subject comprising: generating a data structure storing a plurality of reference clinical parameters associated with a plurality of subjects and a plurality of reference biomarker parameters associated with one or more biomarkers; executing a plurality of variable selection models on the plurality of reference clinical parameters and the plurality of reference biomarker parameters to select a plurality of model parameters, the plurality of model parameters comprise a subset of at least one reference clinical parameter of the plurality of reference clinical parameters and at least one reference biomarker parameter of the plurality of reference biomarker parameters; determining a plurality of candidate wound closure prediction models, each candidate model comprising at least one model parameter of the plurality of model parameters; determining a performance metric for each candidate model of the plurality of candidate models; and selecting, based on the performance metric, a candidate model from the plurality of candidate models as the wound closure prediction model, the wound closure prediction model configured to determine a prediction metric associated with a successful closure of the wound.

2: The method of embodiment 1. wherein determining the plurality of candidate models comprises: determining that a first value of at least one reference biomarker parameter associated with a candidate model of the plurality of candidate models is missing; and estimating, based on the missing first value, the first value using a reference value.

3: The method of embodiments 1 or 2, wherein the performance metric comprises a magnitude of an area under a receiver operator curve associated with each candidate model.

4: The method of any of embodiments 1-3, wherein selecting the plurality of model parameters comprises: executing the plurality of variable selection models on the plurality of reference clinical parameters and the plurality of reference biomarker parameters to select first model parameters and second model parameters, the first model parameters comprising at least one parameter different from the second model parameters; and determining the plurality of candidate wound closure prediction models comprises determining a first candidate wound closure prediction model and a second candidate wound closure prediction model, the first candidate wound closure prediction model comprising the first model parameters and the second candidate wound closure prediction model comprising the second model parameters.

5: Hie method of any of embodiments 1-4, wherein the reference biomarker parameters are determined by analyzing effluents or serums collected from the wound, and wherein analyzing the effluents or serums comprises performing an immunoassay, the immunoassay comprising one or more of electrochemiluminescence assay, affinity capture assay, immunometric assay, heterogeneous chemiluminescence immunometric assay, homogeneous chemiluminescence immunometric assay, ELISA, western blotting, radioimmunoassay, magnetic immunoassay, realtime immunoquantitative polymerase chain reaction (iqPCR), and surface-enhanced Raman spectroscopy (SERS) label-free assay.

6: The method of any of embodiments 1-5, wherein the plurality of reference clinical parameters comprise a plurality of wound characteristics, wherein a wound characteristic of the plurality of wound characteristics comprises: a wound type; a number of wounds; a wound location; a wound length; a wound width; or a wound depth.

7: The method of any of embodiments 1-6, wherein a reference biomarker parameter of the plurality of reference biomarker parameters comprises: a biomarker type of a biomarker; a concentration of the biomarker; a first period of time associated with collecting a first biological sample associated with the biomarker from the wound before debridement of the wound; a second period of time associated with collecting a second biological sample associated with the biomarker from the wound after the debridement of the wound; a coefficient of variation associated with a standard deviation of a first concentration of the biomarker associated with the first period of time and a second concentration of tire biomarker associated with the second period of the time and a mean of the first concentration and the second concentration; or a difference between the first concentration and the second concentration.

8: The method of embodiment 7, wherein the first period of time comprises a period of time of at least 12 hours prior to the debridement of tire wound and the second period of time comprises a period of time of at least 2 hours after the debridement of the wound.

9: The method of any of embodiments 1-8. wherein the plurality of variable selection models comprises one or more machine learning algorithms comprising decision trees, random forests, neural networks, logistic regression, support vector machine, Bayesian belief network, LASSO, Relief-based feature selection, Stable Iterative Variable Selection, or Elastic-net. 10: The method of any of embodiments 1-9, wherein the plurality of reference biomarker parameters is associated with at least one serum biomarker comprising: basic fibroblast growth (FGFBASIC), epidermal growth factor (EGF), eotaxin-1 (CCL-11), eotaxin-3 (CCL-26), fms related receptor tyrosine kinase- 1 (FLT-1), granulocyte-colony stimulating factor (GCSF), granulocyte-monocyte colony stimulating factor (GMCSF). hepatocyte growth factor (HGF), interferon alpha-2 (IFNa2a). interferon gamma (IFNg). interleukin- 10 (IL- 10), interleukin- 12/interleukin-23p40 (IL-12/IL-23p40), interleukin- 15 (IL-15), interleukin- 16 (IL-16), interleukin- 17A (IL-17A), interleukin- 1 receptor antagonist (IL IRA), interleukin-22 (IL22), interleukin-2 receptor subunit alpha (IL-2RA), interleukin-3 (IL-3), interleukin-5 (IL-5), interleukin-6 (IL-6), interleukin-7 (IL-7), interleukin-8 (IL-8), interleukin-9 (IL-9), interferon gamma-inducing protein- 10 (IP- 10), monocyte chemoattractant protein- 1 (MCP-1). monocyte chemoattractant protein-4 (MCP-4), macrophage-derived chemokine (MDC), monokine induced by gamma interferon (MIG), macrophage inflammatory protein-lb (MIP-1B), placental growth factor (PLGF), thymus and activation regulated chemokine (TARC), angiopoientin- 1 receptor (TIE-2), tumor necrosis factor alpha (TNFA), vascular endothelial growth factor (VEGFA), vascular endothelial growth factor C (VEGFC). or vascular endothelial growth factor D (VEGFD).

11 : The method of any of embodiments 1-10, wherein the plurality of reference biomarker parameters is associated with at least one effluent biomarker comprising: FGFBASIC. FLT-1, GMCSF, IFNg, IL- 10, IL-12/IL-2pP40, IL- 15, IL- 16, IL-17A, interleukin- 1 alpha (IL-1A), interleukin-1 beta (IL-1B), IL-2RA, IL-5, IL-7, MDC, MIG, PLGF, TARC, TIE-2, VEGFA, VEGFC, or VEGFD.

12: The method of any of embodiments 1-11, further comprising: detennining fit indices associated with using the wound closure prediction model, wherein the fit indices are detennined based on at least one of: an injury mechanism; a number of injuries; a wound location; a wound type; a wound size; or a current number of debridement.

13: The method of embodiment 12, further comprising: generating, based on clinical wound closure data and estimated wound closure data from the wound closure prediction model, a simulation to determine a probability threshold for wound closure.

14: A method of predicting successful closure of a wound of a subject comprising: receiving a first value associated with a clinical parameter associated with the subject and a second value associated with a biomarker parameter; executing a wound closure prediction model using the clinical parameter and the biomarker parameter to generate a prediction associated with closing the wound, wherein the wound closure prediction model is generated by performing operations comprising: generating a data structure storing a plurality of reference clinical parameters associated with a plurality of subjects and a plurality of reference biomarker parameters; executing a plurality of variable selection models on the plurality of reference clinical parameters and the plurality of reference biomarker parameters to select a plurality' of model parameters, the plurality of model parameters comprising a subset of at least one reference clinical parameter of the plurality of reference clinical parameters and at least one reference biomarker parameter of the plurality of reference biomarker parameters; determining a plurality candidate wound closure prediction models, each candidate model comprising at least one model parameter of the plurality of model parameters; determining a performance metric for each candidate model of the plurality of candidate models; and selecting, based on the performance metric, a candidate model from the plurality of candidate models as the wound closure prediction model; and outputting, by the wound closure prediction model, a prediction metric associated with the successful closure of the wound.

15: The method of embodiment 14, wherein tire prediction metric comprises a probability associated with successful wound closure and a recommendation associated with the probability'.

16: The method of embodiment 15, wherein the recommendation is determined based on the probability and a probability threshold.

17: The method of any of embodiments 14-16, wherein the clinical parameter corresponds to a reference clinical parameter of the model parameters and the biomarker parameter corresponds to a reference biomarker parameter of the model parameters.

18: The method of any of embodiments 14-17, wherein, the prediction metric indicates closing the wound, the method further comprising: executing the closure of the wound; determining a status associated with the closed wound; and updating, based on the first value, the second value, and tire status associated with the closed wound, the wound closure prediction model.

19: A system for generating a wound closure prediction model for a wound of a subject comprising: one or more processors; and one or more non-transitory computer-readable media storing computer-executable instructions that, when executed, cause the one or more processors to perform operations comprising: generating a data structure storing a plurality of reference clinical parameters associated with a plurality of subjects and a plurality of reference biomarker parameters; executing a plurality of variable selection models on the plurality of reference clinical parameters and the plurality of reference biomarker parameters to select a plurality of model parameters, the plurality of model parameters comprising a subset of at least one reference clinical parameter of the plurality of reference clinical parameters and at least one reference biomarker parameter of the plurality of reference biomarker parameters; determining a plurality of candidate wound closure prediction models, each candidate model comprises at least one model parameter of the plurality of model parameters; determining a performance metric for each candidate model of the plurality of candidate models; and selecting, based on the performance metric, a candidate model from the plurality of candidate models as the wound closure prediction model, the wound closure prediction model configured to determine a prediction metric associated with a successful closure of the wound.

[00124] 20: A system for predicting a successfill closure for a wound of a subject comprising: one or more processors; an input component; an output component; and one or more computer- readable media storing computer-executable instructions that, when executed, cause the one or more processors to perform operations comprising: receiving, from the input component, a first value associated with a clinical parameter associated with the subject and a second value associated with a biomarker parameter; executing a wound closure prediction model for predicting wound closure for the wound using the clinical parameter and the biomarker parameter to generate a prediction associated with closing the wound, wherein the wound closure prediction model is generated by performing operations comprising: generating a data structure storing a plurality of reference clinical parameters associated with a plurality of subjects and a plurality of reference biomarker parameters; executing a plurality of variable selection models on the plurality of reference clinical parameters and the plurality of reference biomarker parameters to select a plurality of model parameters, the plurality of model parameters comprising a subset of at least one reference clinical parameter of the plurality of reference clinical parameters and at least one reference biomarker parameter of the plurality of reference biomarker parameters; determining a plurality of candidate wound closure prediction models, each candidate model comprising at least one model parameter of the plurality of model parameters; detennining a performance metric for each candidate model of the plurality of candidate models; selecting, based on the perfonnance metric, a candidate model from the plurality of candidate models as the wound closure prediction model; and outputting, by the wound closure prediction model and at the output component, a prediction metric associated with a successfill closure of the wound.

[00125] While the example clauses described above arc described with respect to one particular implementation, it should be understood that, in the context of this document, the content of the example clauses can also be implemented via a method, device, system, computer-readable medium, and/or another implementation. Additionally, any of examples 1-## may be implemented alone or in combination with any other one or more of the examples 1-##.

EXAMPLES

[00126] Example 1: FIG. 6 illustrates an example process 600 associated with collecting biological samples before and after surgical debridement of the wound. Prior to debridement, first wound VAC canister(s) are placed on the patient's wound(s) for a minimum 12 hours . Effluent from the first canisters is collected to derive biomarkers (labeled Eon 12 hrs.). During debridement, 10 to 30 mL of blood is taken to derive serum biomarkers. Post debridement, second VAC canister(s) are placed on the patient's wound(s) for two hours, where again effluent is collected (labeled E2 2 hrs.). Patients then receive standard wound therapy until tire next debridement or until the wound is closed, usually within 48 to 72 hours.

[00127] Example 2: FIG. 7 illustrates an example process 700 of determining wound closure by using a device that includes the wound closure prediction model. The steps shown in the example process 600 correspond with the steps described in association with FIGS. 4 and 5.

[00128] At Day N, effluent samples from the wound may be collected overnight (e.g., for a minimum of 12 hours and/or up to 16 hours prior to debridement surgery) and for two hours after the debridement surgery. The effluent samples may be processed, frozen, and stored until it is ready for biomarker detection and wound closure prediction. If necessary; the effluent samples may be transported to a location where biomarker parameters may be detennined.

[00129] At Day N+l, the effluent samples may be thawed for biomarker parameter detection and determination. A Meso Scale Discovery (MSD) immunoassay may be performed on the effluent samples to determine the biomarker parameters. A clinical decision support tool (CDST) may be performed using the biomarker parameters and clinical parameters to determine a wound closure prediction metric. The CDST may correspond to the computing device 306 of FIG. 3 and may include the wound closure prediction model. The CDST may output a report that includes tire wound closure prediction metric and a recommendation on whether to close the wound (e.g., possible estimated outcomes). The recommendation may be determined based on whether the wound closure prediction metric is above a prediction threshold. The possible estimated outcomes may be that the wound is ready for surgical closure or the wound is not ready for surgical closure. For example, if the prediction threshold is 85%, then when the prediction metric is under 85%, the recommendation may be that the wound is not ready for surgical closure, but if the prediction metric is at or above 85%, then the recommendation may be that the wound is ready for surgical closure. In example 2, the MSD and the CDST may constitute a single wound closure prediction device.

[00130] At Day N+2, the next surgical treatment may be scheduled to treat the wound. The wound closure prediction device may communicate at least the recommendation for the wound to the surgeon. The surgeon may decide on whether to close the w ound based on the recommendation. In a scenario where the recommendation is to not close the wound, the surgical treatment may be to perform the debridement surgery as described in association with Day N, and the process may restart at Day N until the surgeon decides to close the wound. [00131] Example 3: FIGS. 8A-8C illustrates an example report from a computing device associated with predicting wound closure. This example report may be generated by, for example, the computing device 306 of FIG. 3.

[00132] FIG. 8A illustrates a first page 802 of the report. The first page may include information such as the patient’s name, date of birth, the patient’s sex or gender, a report date, the physician who ordered the wound closure prediction, a referring laboratory, a time of pre-surgery wound effluent collection, a time of post-surgery wound effluent collection, a date of specimen received, a time of report, a description of the assay that is used to determine the biomarker parameters and the CDST, and the assay results. Tire assay results include the names of the biomarkers and the concentrations (in picograms per milliliter) of the biomarkers collected before debridement of the wound and after debridement of the wound.

[00133] FIG. 8B illustrates a second page 804 of the report. The second page may include the wound closure prediction metric, a first recommendation associated with tire wound closure prediction metric being at or above the prediction threshold, and a second recommendation associated with the wound closure prediction metric being lower than the prediction threshold. On second page 804, the prediction threshold is 0.85, and therefore, the first recommendation indicates that for a first scenario where the wound closure prediction metric (e.g., labeled on the second page 804 as a Successful Wound Closure Probability score (PS)) that is at or exceed 0.85, the first recommendation would apply, and for a second scenario where the wound closure prediction metric is below 0.85, then the second recommendation would apply. Tire second page 804 further includes an intended use section and an assay limitation section, which continues and finishes at a third page 806 of the report as illustrated by FIG. 8C.

[00134] Example 4: FIG. 9 illustrates an example process 900 for determining biomarker parameters. Step 1 of the process may include preparing samples and reagents. In embodiments, the samples may include the collected biological samples described in association with FIGS. 1-4. Step 1 may further include thawing the samples and a diluent in a water bath, aliquoting the samples, spinning the aliquoted samples in a centrifuge at 2000g for 3 minutes, and making a wash buffer and read buffer. Step 2 may include reconstituting a calibrator and controls and allowing the reconstituted calibrator and controls to sit for 15 minutes. Step 3 may include adding the diluent (diluent 43) to standard tubes 2-8 and control tubes 1-3. Step 4 may include creating a 1:2 dilution of diluent 43 and the samples in each cluster tube. Step 5 may include creating a standard curve using a 1 :4 serial dilution 8 times to create 8 standards. Hie 8 standards may be for effluent samples. Because the 8 standards are created, the standard curve is an 8-point standard curve. Step 5 may further include diluting the control tubes at a 1 :2 dilution. Step 6 may include washing a plate three times with 150 microliters of wash buffer. In embodiments, the plate may be a 96-well 10-spot plate. Each spot per well may be pre-coated with an antibody specific to biomarkers of interest as illustrated in FIG. 10. Step 7 may include loading a plate with 50 microliters of the samples, the standards, and the controls according to FIG. Step 7 may further includes covering the plate and perfonning a first incubation of tire plate for 2 hours in a shaker incubator. Step 8 may include creating a detection antibody solution before the first incubation is complete. In embodiments, the detection antibody may be SULFO-Tag. Step 9 may include washing the plate 3 times with 150 microliters of wash buffer after step 7 is completed. Step 10 may include adding 25 microliters of the detection antibody solution created in Step 8 to each well of the plate. Step 10 may further include covering the plate and performing a second incubation of the plate for 2 hours in the shaker incubator. Step 11 may include washing the plate 3 times with 150 microliters of wash buffer after step 10 is completed. Step 12 may include adding 150 microliters of 2x read buffer to tire plate. Step 13 may include analyzing the plate using a biomarker detection device. In embodiments, the biomarker detection device may be an MSD Sector S600. Step 13 may further include determining a concentration for each biomarker of interest by fitting the results from the biomarker detection device to the standard curve using a 4-parameter logistics software. The concentrations for each biomarker of interest may be used as biomarker parameters to predict successful wound closure by a wound closure prediction model.

[00135] FIG 10 illustrates a plate map 1000 used to determine the biomarker parameters according to Example 5 and a well map 1002 illustrating the cells of each well where the biomarkers are being detected. With respect to the plate map 1000, STD indicates the wells where the standard is added to the plate; CTRL indicates the wells where the control is added to the plate; and Sample indicates the wells where the samples are added to the plate. With respect to the well map 1002, the well map 1002 indicates that cell 1 may be pre-treated with antibodies designed to assist in detecting GM-CSF, cell 3 may be pre-treated with antibodies designed to assist in detecting IL-5, cell 4 may be pre-treated with antibodies designed to assist in detecting IL-7, cell 5 may be pretreated with antibodies designed to assist in detecting IL-12/IL-23p40, cell 6 may be pre-treated with antibodies designed to assist in detecting IL- 15, and cell 8 may be pre-treated with antibodies designed to assist in detecting IL-17A. In embodiments, all wells of the plate may be pre-treated with antibodies as illustrated by the well map 1002.

[00136] Example 5: The population used for generating the wound closure prediction model may be from two observational cohort studies of acute traumatic wound healing. Tire first is “The Use of Vacuum Assisted Wound Closure Device in treating Extremity Wounds” (WoundVAC: USU IRB study number 352334) that examined the use of negative pressure therapy with vacuum assisted closure (VAC) on wound management in combat casualties treated at Walter Reed Military Medical center (WRNMMC). The second study was observational cohort of civilian acute trauma patients at Grady Memorial Hospital in Atlanta Georgia, “The effect of inflammatory biomarker expression on decisions to close traumatic wounds” (WoundDX: Grady IRB study number 00058229) that examined the effect of standardized wound management on wound profdes. Subjects and wounds were selected from these two cohorts if they met the following eligibility criteria.

[00137] All patients met the following inclusion criteria: ISS > 9, wound surface area > 72 cm², extremity injury (including shoulder and buttock - without visceral communication), wound receiving wound VAC treatment, and a minimal age of 18 years.

[00138] The following co-morbidities were used as exclusion criteria: patient with wounds closed immediately after enrollment in the WoundVAC or WoundDX studies without debridement, coronary artery disease, diabetes mellitus (insulin-dependent type I or type II diabetes), peripheral vascular disease, age > 65 years, connective tissue disorders, preexisting immunosuppressing conditions or immunosuppression therapy, and/or pregnancy.

[00139] The analysis population only used patients in military and civilian studies whose treatment received negative pressure therapy with VAC on wound management. These patients underwent strict wound management where wound examination and debridement occurred every' 24 or 48 hours in tire operating room with patients remaining in the hospital until wounds were successfully closed as described in association with FIG. 6, as well as throughout this disclosure. The analysis population further provided enough serum and effluent samples to obtain MSD data. This resulted in 100 wounds from the WoundVAC study (23 dehisced) and 22 wounds from the WoundDX study (6 dehisced).

[00140] Example 6: Data Curation.

[00141] The reference clinical parameters and the reference biomarker parameters used to generate the wound closure prediction model may be collected from serum samples during debridement of the wound and effluent samples before and after the debridement. The reference clinical variables used to generate the wound closure prediction model are listed in Table 1 below:

[00142] Table 1: Clinical variables used to generate the wound closure prediction model

[00143] The biomarkers collected from the serum and effluent samples are listed in Table 2 below.

[00144] Table 2: Biomarkers from serum collected during debridement and effluent samples collected pre- and post-debridement.

[00145] Prior to generating the wound closure prediction model, all continuous parameters were appropriately transformed (e.g., biomarker parameters were log-transformed to approximate normality) and then standardized to facilitate modeling. All dichotomous parameters were numerically coded as 0 and 1 as is most appropriate for the parameter.

[00146] Example: 7: Missing Data.

[00147] There are two types of missingness in the data: standard missing data found in the clinical variables that are assumed to be missing at random (MAR), and missing data in the MSD biomarker assays where missingness occurs due to the restricted sensitivity and precision of the applied assay.

[00148] Most biomarkers display a lower limit of detection (LLOD), the lowest concentration that is identifiably different from background noise. Biomarkers also display a lower limit of quantification (LLOQ) and upper limit of quantification (ULOQ) where biomarker values (e.g., concentration) are known to follow approximately normal distributions when log-transformed, which permits the use of imputation methods that assume the missing values are from recognized data generation distribution.

[00149] Using the above assumptions, the following methods were used to handle censored and missing data.

[00150] Clinical parameters: continuous and categorical clinical parameters were imputed using a random forest model to the observed values of each parameter, and the missing values were calculated based on that model, assuming missingness is MAR.

[00151] Biomarker parameters: If the missing biomarker is below the LLOD or LLOQ and tire per plate LLOQ value is less than 2.5% quantile of known values, the per plate LLOQ value was treated as the estimated value.

[00152] If the missing biomarker is above the ULOQ, and the per plate ULOQ value is less than 97.5% quantile of known values, the per plate ULOQ value was treated as the estimated value. [00153] Missing values in tire middle of the distribution were treated as MAR and imputed using the random forest model employed for imputing the clinical variables.

[00154] All imputed values were compared to the limits of quantification and if within the range of the limits of quantification, the imputed value was retained, and if outside tire range of the limits of quantification, the limit of detection value was treated as a point estimate of a mean in a normal distribution and tire imputed value is computed as a conditional expectation using a standard deviation of the known values.

[00155] Example 8: Imputation for the generated wound closure prediction model.

[00156] Tire imputation method described in association with Example 7 uses all available data in order to generate the wound closure prediction model. However, this is not feasible when using the generated wound closure prediction model as the entry of all data used in the imputation method is not currently possible. As a result, simpler imputation algorithms were developed and tested following the criteria.

[00157] Only parameters selected for the generated wound closure prediction model (e.g., the model parameters as described in association with FIGS. 1-4) were used for imputation.

[00158] The clinical parameters listed in Table 1 will not be imputed, but were required if they were used as the model parameters, as these parameters should be available when wound debridement surgery takes place. These parameters may be used in the imputation algorithm as predictors.

[00159] Biomarker parameters listed in Table 2 were imputed and may be used as predictors in the imputation algorithm.

[00160] Tire selection of the wound closure prediction model was based on how well the selected model imputes missing values in simulations and their impact on the algorithm’s performance in terms of the discrimination criteria set forth in this disclosure.

[00161] In terms of the effluent biomarker values that the wound closure prediction model is focused on, missingness was relatively rare, occurring most frequently in IL-17A overnight predebridement effluent, followed by IL-5 measured two hours post-debridement. All effluent biomarkers had a missingness of less than 3%. However, 27% of wounds had one or more missing parameters. To address this, each missing effluent biomarker was imputed using coefficients from a linear regression model that uses the rest of the biomarkers in tire wound characteristics to impute the missing effluent values, assuming a lognormal distribution for the biomarker parameters.

[00162] Discussion with laboratory personnel indicates that the most likely reason for the missingness of biomarker parameters is the plate failed to run in its entirety. In embodiments, if four or more of tire biomarkers from either tire pre-debridement sample or the post-debridement sample are missing, no imputation would be attempted, or wound closure prediction would be attempted.

[00163] Example 9: Wound closure prediction algorithm development.

[00164] Tire large number of analysis variables compared to the relatively limited number of wounds limits the usage of standard variable selection and model comparison methods. This is particularly true for biomarker data that is highly correlated. As a result, a more practical approach that focuses on generating a wound closure prediction model that can be utilized in a hospital setting with limited disruption was developed. Rational content groups of predictive parameters were employed across candidate wound closure prediction models to determine if certain groups of parameters were more predictive than other groups of parameters. The content groups are shown below in Table 3.

[00165] Table 3: Content groups of predictor variables to predict wound closure.

[00166] Feature or variable selection was performed on the parameters shown in Table 3 where parameters that are most informative in relation to predicting successful wound closure were selected while irrelevant or redundant parameters were discarded. Variable selection algorithms or models may be used to perform the feature or variable selection to select different candidate model parameters for the candidate wound closure prediction models. Examples of the variable selection algorithms and models include prediction algorithms with embedded variable selection methods (e.g., a Bayesian belief network (BBN), LASSO, and/or the like), a fdter variable selection method (e.g., Relief-based feature selection), and/or a wrapper selection method (e.g., Stable Iterative Variable Selection).

[00167] Early model exploration showed a preference for an elastic-net algorithm that regularizes a generalized linear model. The predictive ability of each candidate model was evaluated using indices of model discrimination with an AUROC, calibration (at large and a slope), and overall accuracy (scaled Brier score). The wound closure prediction model selected from the candidate models was chosen based on overall fit indices as well as demonstrating consistency across important subgroups as described in association with Example 10.

[00168] As FIG. 13 illustrates, the best-performing candidate wound contains six cytokine plate biomarkers (GM-CSF, IL-5, IL-7, IL-12/IL-23p40, IL-15, and IL-17A) collected both overnight pre -debridement (e.g., for 12 hours before debridement) and for two hours post debridement and wound characteristics, which may include, wound size (e.g., length, width, and depth), wound location if the wound is on an upper extremity, and wound type (e.g., soft tissue injury, open fracture, amputation, or fasciotomy). Candidate models with large numbers of predictors (e.g., over 20) exhibited poor calibration indices with extremely wide confidence intervals across bootstrapped replication, thereby indicating that such candidate models were unstable. In addition, these candidate models show high within-subject variability across debridements, which may indicate that these candidate models would be unsuitable for clinical use.

[00169] FIG. 13 illustrates an example table 1300 of candidate wound prediction models. The model content section indicates tire model parameters for each candidate model. The table 1300 also includes, for each candidate model, pcrfonnancc metrics including a discrimination AUROC value for each candidate model, calibration values (intercept and slope), a Scaled Brier Score, and an Average Number of Covariates. The values within the parenthesis for each performance metric indicate a confidence interval forthat performance metric. The table 1300 is also ordered from the candidate model that is most suitable to be selected as the wound closure prediction model at the top to the candidate model least suitable for selection at the bottom of the table. In embodiments, the candidate model may be selected as the wound closure prediction model based on the candidate model having an AUROC value closest to 1.

[00170] Based on the variable selection and the performance metrics of each candidate model, the generated wound closure prediction model is a regularized logistic regression model containing 21 model parameters. The model variables include six cytokine plate biomarkers (GM-CSF, IL-5, IL-7, IL-12/IL-23p40, IL-15, and IL-17A) from biological samples collected overnight pre- debridement and two-hour post-debridement and wound characteristics including wound size (wound length, widths, and depth), wound location (if wound is on an upper extremity), and wound type (soft tissue injury, open fracture, amputation, or fasciotomy). In equation form, the wound closure prediction model is: where p(success) is the probability of successful wound closure, i is an index for K = 21 predictor variables including an intercept, e indicates taking the exponent, and p_t indicates the indexed coefficient of the x_t predictor variable. Table 4 below includes the p coefficient for each of the 21 model parameters including an intercept. [00171] Table 9: coefficients, means, and standard deviations of biomarkers needed to calculate a probability of successful wound closure.

Standard

Model Parameters coefficient Mean Deviation

Intercept -15.00746

GM-CSF overnight -0.89881361 0.95167 0.495629

IL- 12/IL-23p40 overnight -0.31377901 2.042366 0.264377

IL- 15 overnight -0.37379907 1.262377 0.127867

IL- 17A overnight 0.71727238 0.978654 0.323231

IL-5 overnight -0.51077253 1.637538 0.438534

IL-7 overnight 0.72038653 0.261215 0.304764

GM-CSF two-hour 1.50806731 0.240582 0.407213

IL- 12/IL-23p40 two-hour 0.72373065 1.949873 0.201348

IL- 15 two-hour -0.03990937 1.124534 0.159242

IL-17A two-hour 0.52606664 0.800342 0.361174

IL-5 two-hour -0.92380753 0.896539 0.390423

IL-7two-hour 0.38169924 0.729817 0.187269

Amputation 3.38546221

Open fracture 10.79290581

Fasciotomy 12.53863519

Soft tissue injury 9. 14580752

Wound location upper body 0.21557153

Wound length -0.06046686

Wound width 0.32314661

Wound depth 0.19999832

[00172] Example 10: Model Evaluation and Validation.

[00173] The predictive ability of each candidate model was evaluated using indices

[00174] The wound closure prediction model described in association with Example 9 and also illustrated as the top candidate model in FIG. 13 showed good discrimination with an AUROC value of 0.82 and a calibration values of -0.46 for the intercept and 0.45 for the slope, which indicates that the wound closure prediction model could slightly under-predict successfill wound closure.

[00175] FIG. 11 illustrates a calibration plot 1100 associated with the wound closure prediction model and an AUROC plot 1102 associated with the wound closure prediction model. Tire nonparametric calibration curve illustrated in FIG. 11 indicates that the under-prediction of successful wound closure may primarily occur at high predicted probabilities. FIG. 11 also plots sensitivity and specificity values at four points (0.65, 0.75, 0.85, and 0.95). [00176] The wound closure prediction model may be further evaluated in clinically relevant subgroups illustrated in FIG. 14. The goal is to validate that the model demonstrates consistent quality in predictive performance across the subgroups.

[00177] FIG. 14 illustrates an example table 1400 of fit indices for the clinically relevant subgroups for the wound closure prediction model. Overall, measures of the discrimination AUROC and calibration performed consistently in terms of overall injury severity, injury mechanism, wound location, wound size, and single versus multiple injuries.

[00178] Hie model displayed exaggerated performance in amputation wounds (as indicated by a perfect 1.0 AUROC value, and highly inflated calibration values). Hie model also had limited performance in patients with open fractures and patients with wounds closed after 4 or more debridements. As such, the model cannot be considered reliable for predicting successful wound closures in amputation wounds, open fractures, and patients with wounds that have had 4 or more debridements.

[00179] Example 11: Determination of Prediction Threshold for Wound Closure.

[00180] After the wound closure prediction model was selected and generated, the model was used to predict wound closure in subjects at various debridements (e g., first debridement, second debridement, third debridement, and so forth) to identify and select a prediction threshold that makes the most sense from a clinical perspective in terms of sensitivity and specificity. To facilitate this, the sensitivity and the specificity were calculated over a range of high-risk thresholds plotted on an AUROC curve, a plot of the component of the AUROC curve (which indicates a true positive rate and a false positive rate) over the different candidate thresholds, and a clinical impact plot that plots a simulated number of wounds closures over the range of candidate thresholds.

[00181] A simulation of 380 wounds following multiple debridements, using real-world data from current clinical practice wound closure rate and estimated wound closure rate developed from the wound closure prediction model, was generated for the selected prediction threshold to evaluate and verify that the selected prediction threshold is suitable for predicting successfill wound closure using the model.

[00182] FIG. 12 illustrates a clinical impact plot 1200 between threshold values ranging from 0.6 to 1.00 in the penultimate wound debridement before attempting wound closure. The clinical impact plot 1200 is created by simulating a number of successful wound closures (tire dotted line) and a total number of wound closures (solid line) in 100 patients over each threshold value within the range. Experienced clinicians favor high threshold values between 0.80 and 0.90 as this indicates that a high proportion of attempted wound closures would be successful. These clinicians reasoned that higher thresholds would give wounds time to enter a pro-healing state across successive debridements. In embodiments, a prediction threshold of 0.85 may be selected [00183] To test this reasoning, a simulation of 380 wounds followed over multiple debridements was instantiated using data from current clinical practice wound closure rates (for comparison to current practice) and estimated wound closure rates using the wound closure prediction model for the detennined threshold.

[00184] FIG. 15 illustrates an example table 1500 of simulated wound closure rates for current practice and estimated wound closure rates from the wound closure prediction model at the selected prediction threshold of 0.85. The table 1500 shows a substantially higher cumulative successful wound closure rate of 88% compared to 72% that is associated with current practice. This higher success rate is primarily a function of identifying more wounds ready for closure as seen in the second debridement interval where the successfid wound closure rate is 90% compared to 72% under current practice. The simulation assumes that most subjects would have had their fourth or fifth debridements (if needed) within 14 days of injury'. However, a small number, 39 (10%) in the current practice and 23 (6%) in the wound closure prediction model simulation would not have had their first closure attempt within this 14-day window that the wound closure prediction model is considered valid in.

Claims

CLAIMS WHAT IS CLAIMED IS:

1. A method of generating a wound closure prediction model for a wound of a subject comprising: generating a data structure storing a plurality of reference clinical parameters associated with a plurality of subjects and a plurality of reference biomarker parameters associated with one or more biomarkers: executing a plurality of variable selection models on the plurality of reference clinical parameters and the plurality of reference biomarker parameters to select a plurality of model parameters, the plurality of model parameters comprise a subset of at least one reference clinical parameter of the plurality of reference clinical parameters and at least one reference biomarker parameter of the plurality of reference biomarker parameters; determining a plurality of candidate wound closure prediction models, each candidate model comprising at least one model parameter of the plurality of model parameters; determining a performance metric for each candidate model of the plurality of candidate models; and selecting, based on the performance metric, a candidate model from the plurality of candidate models as the wound closure prediction model, the wound closure prediction model configured to determine a prediction metric associated with a successful closure of the wound.

2. The method of claim 1, wherein determining the plurality of candidate models comprises: determining that a first value of at least one reference biomarker parameter associated with a candidate model of the plurality of candidate models is missing; and estimating, based on the missing first value, the first value using a reference value.

3. The method of claim 1, wherein the performance metric comprises a magnitude of an area under a receiver operator curve associated with each candidate model.

4. The method of claim 1, wherein selecting the plurality of model parameters comprises: executing the plurality of variable selection models on the plurality of reference clinical parameters and the plurality of reference biomarker parameters to select first model parameters and second model parameters, tire first model parameters comprising at least one parameter different from the second model parameters; and determining tire plurality of candidate wound closure prediction models comprises determining a first candidate wound closure prediction model and a second candidate wound closure prediction model, the first candidate wound closure prediction model comprising the first model parameters and the second candidate wound closure prediction model comprising the second model parameters.

5. Tire method of claim 1, wherein the reference biomarker parameters are determined by analyzing effluents or serums collected from the wound, and wherein analyzing the effluents or serums comprises performing an immunoassay, the immunoassay comprising one or more of electrochemiluminescence assay, affinity capture assay, immunometric assay, heterogeneous chemiluminescence immunometric assay, homogeneous chemiluminescence immunometric assay, ELISA, western blotting, radioimmunoassay, magnetic immunoassay, realtime immunoquantitative polymerase chain reaction (iqPCR), and surface-enhanced Raman spectroscopy (SERS) label-free assay.

6. The method of claim 1, wherein the plurality of reference clinical parameters comprise a plurality of wound characteristics, wherein a wound characteristic of the plurality of wound characteristics comprises: a wound type; a number of wounds; a wound location; a wound length; a wound width; or a wound depth.

7. The method of claim 1, wherein a reference biomarker parameter of the plurality of reference biomarker parameters comprises: a biomarker type of a biomarker; a concentration of the biomarker; a first period of time associated with collecting a first biological sample associated with the biomarker from the wound before debridement of the wound; a second period of time associated with collecting a second biological sample associated with the biomarker from the wound after the debridement of the wound; a coefficient of variation associated with a standard deviation of a first concentration of the biomarker associated with the first period of time and a second concentration of the biomarker associated with the second period of the time and a mean of the first concentration and the second concentration; or a difference between the first concentration and the second concentration.

8. Tire method of claim 7, wherein the first period of time comprises a period of time of at least 12 hours prior to the debridement of the wound and the second period of time comprises a period of time of at least 2 hours after the debridement of the wound.

9. Tire method of claim 1, wherein the plurality of variable selection models comprises one or more machine learning algorithms comprising decision trees, random forests, neural networks, logistic regression, support vector machine, Bayesian belief network, LASSO, Relief-based feature selection, Stable Iterative Variable Selection, or Elastic -net.

10. The method of claim 1, wherein the plurality of reference biomarker parameters is associated with at least one serum biomarker comprising: basic fibroblast growth (FGFBASIC), epidermal growth factor (EGF), eotaxin-1 (CCL-11), eotaxin-3 (CCL-26), fins related receptor tyrosine kinase-1 (FLT-1), granulocyte-colony stimulating factor (GCSF), granulocyte-monocyte colony stimulating factor (GMCSF), hepatocyte growth factor (HGF), interferon alpha-2 (IFNa2a), interferon gamma (IFNy), interleukin- 10 (IL- 10), interleukin- 12/intcrlcukin-23p40 (IL-12/IL- 23p40), interleukin- 15 (IL-15), interleukin- 16 (IL-16), interleukin- 17A (IL-17A), interleukin-1 receptor antagonist (IL IRA), interleukin-22 (IL22), interleukin-2 receptor subunit alpha (IL-2RA), interleukin-3 (IL-3), interleukin-5 (IL-5), interleukin-6 (IL-6), interleukin-7 (IL-7), interleukin-8 (IL-8), interleukin-9 (IL-9), interferon gamma-inducing protein- 10 (IP- 10), monocyte chemoattractant protein-1 (MCP-1), monocyte chemoattractant protein-4 (MCP-4), macrophage- derived chemokine (MDC), monokine induced by gamma interferon (MIG), macrophage inflammatory protein-lb (MIP-1B), placental growth factor (PLGF), thymus and activation regulated chemokine (TARC), angiopoientin-1 receptor (TIE-2), tumor necrosis factor alpha (TNFA), vascular endothelial growth factor (VEGFA), vascular endothelial growth factor C (VEGFC), or vascular endothelial growth factor D (VEGFD).

11. The method of claim 1, wherein tire plurality of reference biomarker parameters is associated with at least one effluent biomarker comprising: FGFBASIC. FLT-1. GMCSF. IFNy, IL-10, IL-12/IL-2pP40, IL-15, IL-16, IL-17A, interleukin-1 alpha (IL-1A), interleukin- 1 beta (IL- 1B), IL-2RA, IL-5, IL-7, MDC, MIG, PLGF, TARC, TIE-2, VEGFA, VEGFC, or VEGFD.

12. The method of claim 1, further comprising: determining fit indices associated with using tire w ound closure prediction model, wherein the fit indices are determined based on at least one of: an injury mechanism; a number of injuries; a w ound location; a wound type; a wound size; or a current number of debridement.

13. The method of claim 12, further comprising: generating, based on clinical wound closure data and estimated wound closure data from the wound closure prediction model, a simulation to determine a probability threshold for wound closure.

14. A method of predicting successful closure of a wound of a subject comprising: receiving a first value associated with a clinical parameter associated with the subject and a second value associated with a biomarker parameter; executing a wound closure prediction model using the clinical parameter and the biomarker parameter to generate a prediction associated with closing the wound, wherein the wound closure prediction model is generated by performing operations comprising: generating a data structure storing a plurality of reference clinical parameters associated with a plurality of subjects and a plurality of reference biomarker parameters; executing a plurality of variable selection models on the plurality of reference clinical parameters and the plurality of reference biomarker parameters to select a plurality of model parameters, the plurality of model parameters comprising a subset of at least one reference clinical parameter of the plurality of reference clinical parameters and at least one reference biomarker parameter of the plurality of reference biomarker parameters; determining a plurality candidate wound closure prediction models, each candidate model comprising at least one model parameter of the plurality of model parameters; determining a performance metric for each candidate model of the plurality of candidate models; and selecting, based on the performance metric, a candidate model from the plurality of candidate models as the wound closure prediction model; and outputting, by the wound closure prediction model, a prediction metric associated with the successful closure of the wound.

15. The method of claim 14, wherein the prediction metric comprises a probability associated with successful wound closure and a recommendation associated with the probability.

16. Tire method of claim 15, wherein the recommendation is determined based on the probability and a probability threshold.

17. The method of claim 14, wherein the clinical parameter corresponds to a reference clinical parameter of the model parameters and the biomarker parameter corresponds to a reference biomarker parameter of the model parameters.

18. Hie method of claim 14, wherein, the prediction metric indicates closing the wound, the method further comprising: executing the closure of the wound; determining a status associated with the closed wound; and updating, based on the first value, the second value, and the status associated with the closed wound, the wound closure prediction model.

19. A system for generating a wound closure prediction model for a wound of a subj ect comprising: one or more processors; and one or more non-transitory computer-readable media storing computer-executable instructions that, when executed, cause the one or more processors to perform operations comprising: generating a data structure storing a plurality of reference clinical parameters associated with a plurality of subjects and a plurality of reference biomarker parameters; executing a plurality of variable selection models on the plurality of reference clinical parameters and the plurality of reference biomarker parameters to select a plurality of model parameters, the plurality of model parameters comprising a subset of at least one reference clinical parameter of the plurality of reference clinical parameters and at least one reference biomarker parameter of the plurality of reference biomarker parameters; determining a plurality of candidate wound closure prediction models, each candidate model comprises at least one model parameter of the plurality of model parameters; determining a performance metric for each candidate model of the plurality of candidate models; and selecting, based on the performance metric, a candidate model from the plurality of candidate models as the wound closure prediction model, the wound closure prediction model configured to determine a prediction metric associated with a successful closure of the wound.

20. A system for predicting a successful closure for a wound of a subject comprising: one or more processors; an input component; an output component; and one or more computer-readable media storing computer-executable instructions that, when executed, cause the one or more processors to perform operations comprising: receiving, from the input component, a first value associated with a clinical parameter associated with the subject and a second value associated with a biomarker parameter; executing a wound closure prediction model for predicting wound closure for the wound using tire clinical parameter and the biomarker parameter to generate a prediction associated with closing the wound, wherein the wound closure prediction model is generated by performing operations comprising: generating a data structure storing a plurality of reference clinical parameters associated with a plurality of subjects and a plurality of reference biomarker parameters; executing a plurality of variable selection models on the plurality of reference clinical parameters and the plurality of reference biomarker parameters to select a plurality of model parameters, the plurality of model parameters comprising a subset of at least one reference clinical parameter of the plurality of reference clinical parameters and at least one reference biomarker parameter of the plurality of reference biomarker parameters; determining a plurality of candidate wound closure prediction models, each candidate model comprising at least one model parameter of the plurality of model parameters; determining a performance metric for each candidate model of the plurality of candidate models; selecting, based on the performance metric, a candidate model from the plurality of candidate models as the wound closure prediction model; and outputting, by the wound closure prediction model and at the output component, a prediction metric associated with a successful closure of tire wound.