WO2017055949A1

WO2017055949A1 - Clinical decision support for differential diagnosis of pulmonary edema in critically ill patients

Info

Publication number: WO2017055949A1
Application number: PCT/IB2016/055322
Authority: WO
Inventors: Caitlyn Marie CHIOFOLO; Srinivasan VAIRAVAN; Nicolas Wadih Chbat; Rodrigo Cartin-Ceba
Original assignee: Koninklijke Philips N.V.
Priority date: 2015-09-28
Filing date: 2016-09-07
Publication date: 2017-04-06

Abstract

A system and method for diagnosing pulmonary edema. The system and method perform the steps of comparing patient information to a list of predictors in a predictor database to identify patient predictors, assigning points based on the identified patient predictors, adding the points to determine an aggregated number of points, and determining a probability of one of acute respiratory distress syndrome or cardiogenic pulmonary edema based on the aggregated number of points.

Description

CLINICAL DECISION SUPPORT FOR DIFFERENTIAL DIAGNOSIS OF PULMONARY EDEMA IN CRITICALLY ILL PATIENTS

Inventors: Caitlyn CHIOFOLO, Srinivasan VAIRAVAN, Nicolas CHBAT and Rodrigo CARTIN-CEBA

Background

[0001] Differential diagnosis of cardiogenic pulmonary edema (CPE) and non-cardiogenic pulmonary edema (NCPE) is a crucial part of the decision-making process for care providers due to the different mechanisms of injury and interventions used to manage or treat the conditions. Both CPE and NCPE are

characterized by the presence of fluid in the alveolar spaces of the lung and impaired oxygenation. While CPE can be quickly reversed if the cause is treated, NCPE has a slow or difficult recovery due to impairment of the alveolar-capillary epithelial barrier properties and failure of repair due to initial insult.

[0002] Several clinical markers have been proposed to aid in the differential diagnosis of pulmonary edema. These clinical markers may include markers derived from monitoring, laboratory, radiography or charting data. For example, pulmonary capillary wedge pressure (PCWP) or pulmonary artery occlusion pressure

(PAOP) has been recognized as a potential marker. However, use of pulmonary catheters to measure PCWP is controversial as the technique is invasive, offers unclear or limited benefits, and carries the burden of increased risk of infection, thrombosis and patient discomfort. Laboratory data including markers such as brain natriuretic peptide (BNP) , NT-proBNP, and the troponins are often elevated in patients with acute CPE. Thus, an absence of elevation of these markers can be used to rule out CPE.

Though these markers carry good negative predictive value, additional tests are often required to make a diagnosis if they are elevated.

[0003] Radiologic data such as the vascular pedicle width

(VPW) has been shown to accurately differentiate CPE from NCPE, with a higher VPW (e.g., greater than 70mm) corresponding to CPE. However, the VPW is not routinely measured. Routine

measurements of VPW may require time, training and/or experience in reading chest x-rays. Echocardiographic markers such as ejection fraction (EF%) or the ratio of mitral peak velocity of early filling to early diastolic mitral annular velocity (e/E' ratio) can be used to differentiate CPE from NCPE. These

markers, however, may have limited benefit late in the disease course, as patients with NCPE can develop concomitant CPE.

Patient history from charting data can also provide key

information to the decision process, as history of heart failure or coronary artery disease, or recent acute cardiac event (e.g., myocardial ischemia or stroke) can increase the likelihood of current CPE. However, charting of such acute events may be delayed or difficult to extract from textual information.

[0004] In addition to differentiating between CPE and NCPE, differentiating between the causes of NCPE may also have great prognostic value and may help care providers better assess the chance of edema progression or recovery. Identifying acute respiratory distress syndrome (ARDS), one type of NCPE, can aid in the patient prognosis and provide necessary context for care providers to better manage the patient's condition. Despite the availability of clinical markers for differentiating CPE and

NCPE, however, they have limited applicability. Obtaining the markers may be invasive, not routinely available, not sensitive or not integrated with other markers. Further, risk of CPE or NCPE is not automatically calculated and none of the clinical markers allows for the further differentiation among causes of NCPE or identification of ARDS .

Summary of the Invention

[0005] A method for diagnosing pulmonary edema. The method including comparing patient information to a list of predictors in a predictor database to identify patient predictors,

assigning points based on the identified patient predictors, adding the points to determine an aggregated number of points, and determining a probability of one of acute respiratory distress syndrome or cardiogenic pulmonary edema based on the aggregated number of points.

[0006] A system for diagnosing pulmonary edema. The system including a predictor database including a list of predictors and a processor comparing patient information to the list of predictors to identify patient predictors, assigning points based on the identified patient predictors and adding the points to determine an aggregated number of points to determine a probability of one of acute respiratory distress syndrome or cardiogenic pulmonary edema based on the aggregated number of points .

Brief Description of the Drawings

[0007] Fig. 1 shows a schematic drawing of a system according to an exemplary embodiment.

[0008] Fig. 2 shows a table of predictors for NCPE/CPE and assigned points for each according to an exemplary embodiment. [0009] Fig. 3 shows a table assigning probability of NCPE/CPE based on aggregated points according to an exemplary embodiment.

[0010] Fig. 4 shows a flow diagram of a method for pulmonary edema detection according to an exemplary embodiment.

[0011] Fig. 5 shows a flow diagram of a method for pulmonary edema differentiation according to another exemplary embodiment.

[0012] Fig. 6 shows a flow diagram of a method for

integrating pulmonary edema detection and differentiation according to an exemplary embodiment.

[0013] Fig. 7 shows a flow diagram of a method for

integrating ARDS detection and pulmonary edema differentiation according to an exemplary embodiment.

[0014] Fig. 8 shows a flow diagram of a method for

integrating ARDS detection and pulmonary edema differentiation according to another exemplary embodiment.

[0015] Fig. 9 shows a flow diagram of a method for

integrating ARDS detection and pulmonary edema differentiation according to yet another exemplary embodiment.

[0016] Fig. 10 shows a flow diagram of a method for

integrating ARDS detection along with pulmonary edema detection and differentiation according to an exemplary embodiment. [0017] Fig. 11 shows a flow diagram of a method for integrating ARDS detection along with pulmonary edema detection and differentiation according to another exemplary embodiment.

[0018] Fig. 10 shows a flow diagram of a method for

integrating ARDS detection along with pulmonary edema detection and differentiation according to yet another exemplary

embodiment .

Detailed Description

[0019] The exemplary embodiments may be further understood with reference to the following description and the appended drawings, wherein like elements are referred to with the same reference numerals. The exemplary embodiments relate to a system and method for the differential diagnosis of CPE and NCPE and assessing the risk of acute respiratory distress syndrome

(ARDS), one particular type of NCPE.

[0020] As shown in Fig. 1, a system 100 according to an exemplary embodiment of the present disclosure differentiates between CPE and NCPE and may further identify a patient's risk of ARDS. The system 100 comprises a processor 102, a user interface 104, a display 106 and a memory 108. The processor 102 may include a pulmonary edema detection module 110 for detecting a presence of pulmonary edema in a patient and a pulmonary edema differentiation module 112 for differentiating between CPE and NCPE and particularly between CPE and ARDS. The processor 102 may further include an ARDS detection module 114. A user (e.g., physician) may input user selection and/or preferences for the system 100 via the user interface 104, which may include input devices such as, for example, a keyboard, mouse and/or touch display on the display 106. [0021] The pulmonary edema detection module 110 automatically detects patients at risk of pulmonary edema by identifying the presence of bilateral infiltrates and calculating the ratio of partial pressure arterial oxygen (PaC>₂) to fraction of inspired oxygen (F1O2) and/or the ratio of peripheral oxygen saturation (Sp0₂) to fraction of inspired oxygen (F1O₂) . Pulmonary edema is detected when bilateral infiltrates are present and low P/F or S/F is present. The presence of pulmonary edema may be

displayed on the display 106.

[0022] The pulmonary edema differentiation module 112

differentiates between ARDS (NCPE) and CPE by using NCPE/CPE predictors, which may be stored in the memory 108 via a

predictor database 116. Some of the predictors shown in Fig. 2, have been previously identified in Schmickl et al . (Chest. Jan 2012; 141 (1) : 43-50) . New predictors such as NT-pro BNP, EF%, E/E%, E/E', massive transfusion and high risk surgery have been added. As shown in Fig. 2, each predictor is assigned a weight with magnitude and sign. ARDS predictors have positive weights while CPE predictors have negative weights. The magnitude of the weights indicates the strength of the predictor. Patients are assigned points for each known predictor and the aggregated points may be used to assign a probability of ARDS or CPE according to, for example, the table shown in Fig. 3. For example, the probability of CPE is equal to 1 minus the

probability of CPE. The determined probability may be displayed on the display 106.

[0023] The ARDS detection module 114 calculates an ARDS risk score, as described in PCT Published Application WO2013121374 entitled "Acute Lung Injury (ALI) /Acute Respiratory Distress Syndrome (ARDS) Assessment and Monitoring" and filed on February 14, 2013, the entire disclosure of which is incorporated herein by reference. The ARDS detection module 114 may calculate the ARDS risk score using patient vitals such as, for example, heart rate, arterial systolic and diastolic blood pressure,

respiratory rate along with addition patient data such as, for example, labs, vent settings and/or the administration of one or more drugs to the patient. The ARDS risk score may be

calculated using algorithms such as, for example, the Lempel-Ziv complexity metric, a logistic regression model, a log-likelihood ratio, an inference system, a Bayesian network and/or any of a variety of techniques for aggregation of inference algorithms. A system and method for predicting a medical condition is also described in PCT Published Application WO 2012085750 entitled "Patient Condition Detection and Mortality" and filed on

December 12, 2011, the entire disclosure of which is

incorporated herein by reference. The calculated ARDS risk score may be displayed on the display 106.

[0024] Fig. 4 shows an exemplary method 200 for automatically detecting pulmonary edema in a patient using the system 100. In a step 210, the pulmonary edema detection module 110 may detect the presence of bilateral infiltrates in a patient. The

presence of bilateral infiltrates may be extracted from a chest x-ray image via imaging informatics, from a chest x-ray report via a Natural Language Processing (NLP) program, or from manual entry, either directly through the specific application or indirectly through charting in an electronic medical record. If bilateral infiltrates are detected, in a step 220, the ratio of partial pressure arterial oxygen (PaC^) to fraction of inspired oxygen (FiC^) and/or the ratio of peripheral oxygen saturation

(F1O2) to fraction of inspired oxygen (Fi02) is calculated. In a step 230, it is determined whether ( PaC>₂ ) / ( F1O₂ ) and/or

( SpC>2 ) / ( F1O2 ) are considered to be low - i.e., below a threshold value. For example, when (Pa0₂)/(Fi0₂ ) is less than 300 or

( SpC>₂ ) / ( F1O₂ ) is less than 357, the values are considered to be low. In a step 240, the pulmonary edema detection module

determines whether the patient has pulmonary edema. If the bilateral infiltrates are detected and either ( PaC>₂ ) / ( F1O₂ ) or ( SpC>₂ ) / ( F1O₂ ) are considered to be low, the patient is identified as having pulmonary edema. If either of these conditions are not met, it is determined that the patient does not have

pulmonary edema. The result may be displayed on the display 106. Alternatively, the presence or absence of edema may be indicated via a binary value 1 or 0, respectively.

[0025] Fig. 5 shows a method 300 differentiating between ARDS

(NCPE) and CPE using the system 100. In a step 310, pulmonary edema differentiation module 112 compares available patient history and/or information stored in, for example, a patient's electronic medical record, to the predictors stored in the predictor database 116. In a step 320, points are assigned for each of the known predictors. As described above in regard to the predictor database 116, each predictor is assigned a weight with magnitude and sign, ARDS predictors having positive weights and CPE predictors having negative weights. In a step 330, the points for all of the known patient predictors are added to calculate the aggregated number of points. In a step 340, the aggregated number of points are used to determine the

probability of ARDS and/or CPE using, for example, the table shown in Fig. 3. The probability of either ARDS and/or CPE may be displayed on the display 106, in a step 350. Additional information such as, for example, identified patient predictors and/or aggregated number of points may also be displayed on the display 106.

[0026] The methods 200, 300 described above along with the calculation of ARDS risk score, as described above in regard to the ARDS detection module 114, may be integrated with one

another in any of a variety of ways. For example, as shown in Fig. 6, a method 400 integrates pulmonary edema detection

according to the method 200 with NCPE/CPE differentiation

according to the method 300 so that differentiation between

ARDS/CPE only occurs if pulmonary edema is detected. According to the method 400, in a step 410, the pulmonary edema detection module 110 detects the presence of pulmonary edema by detecting bilateral infiltrates and calculating the Pa0₂)/(Fi0₂) and/or ( SpC>₂ ) / ( F1O₂ ) , as described above in regard to the method 200.

In a step 420, the processor determines whether the pulmonary edema is present. If present, the method 400 proceeds to a step 430. If pulmonary edema is determined not to be present, no further action is taken.

[0027] Upon positive detection of the presence of pulmonary edema, the method 400 proceeds with the steps of the method 300. In particular, in the step 430, the pulmonary edema

differentiation module 112 compares patient history/information with the predictors stored in the predictor database 116. In a step 440, the patient's known predictors are assigned points and the points summed, in a step 450, to determine an aggregated number of points. In a step 460, the aggregated number of points is used to determine the probability of either ARDS or CPE according to, for example, the table shown in Fig. 3. The determined probability may be output or displayed on the display 106, in a step 470. Additional information such as, for example, the presence of pulmonary edema may also be displayed on the display 106.

[0028] According to another exemplary embodiment, the ARDS risk score calculated via the ARDS detection module 114, as described above with respect to the system 100, may be

integrated with the method 300, which differentiates between ARDS and CPE via the pulmonary edema differentiation module 112. This integration, however, may be achieved in a number of different ways. In a first example, as shown in Fig. 7, a method 500a integrates the calculated ARDS risk score and the differentiation between ARDS and CPE using a suppression method. In a step 510a, the ARDS risk score is calculated using the ARDS detection module 114. Steps 520a - 550a correspond to the steps 310 - 340, described above with respect to method 300, to determine the probability of CPE using the calculated aggregated number of points. In a step 560a, the processor 102 determines whether the CPE probability is greater than a threshold value (e.g., 50%) . If the CPE is greater than the threshold value the method 500 proceeds to step 570a, in which any ARDS output, alarms and/or advisories indicating the patient's risk for ARDS are suppressed. If the CPE probability is not greater than the threshold value, no further action is taken. In a further embodiment, the CPE probability may be displayed on the display 106.

[0029] In another example, as shown in Fig. 8, a method 500b may integrate the ARDS detection module 114 with the method 300 of the pulmonary edema differentiation module 112 using a revision method. The method 500b may be substantially similar to the method 500a described above, with steps 510b - 550b substantially corresponding to the steps 510a - 550a described above. In a step 560b, however, an ARDS risk score calculated in step 510b is multiplied by an ARDS probability determined in the step 550b to determine a revised ARDS risk score. This revised ARDS risk score may be output (e.g., displayed) in a step a 570b.

[0030] In yet another example, as shown in Fig. 9, a method

500c may integrate the ARDS detection module 114 with the method 300 of the pulmonary edema differentiation module 112 using a revision method only if the probability of ARDS is high. The method 500c may be substantially similar to the method 500b described above. In a step 510c, an ARDS risk score is

determined using the ARDS detection module 114. In a step 520c, the processor 102 determines whether the calculated ARDS risk score exceeds a threshold value (e.g., 0.8) . If the ARDS risk score exceeds the threshold value and is thus representative of a high probability of ARDS, the method 500c proceeds to steps 530c - 580c, which substantially correspond to the steps 520b - 570b, described above with respect to the method 500b. If the ARDS risk does not exceed a threshold value in step 520c, no further action may be taken. Alternatively, where the ARDS risk score does not exceed the threshold value in the step 520c, the method 500c may display the ARDS risk score determined in the step 510c.

[0031] According to another exemplary embodiment, ARDS detection module 114, the pulmonary edema detection module 110 and the pulmonary edema differentiation module 112 may be integrated with one another. This integration, however, may be achieved in a number of different ways. In a first example, as shown in Fig. 10, a method 600a may be substantially similar to the method 500a, described above, with the addition of the pulmonary edema detection module 110. In a step 610a, an ARDS risk score is determined via the ARDS detection module 114. In a step, 620a, the presence of pulmonary edema is detected via the pulmonary edema detection module 110 using the method 200, as described above with respect to Fig. 4. If it is determined, in a step 630a, that pulmonary edema is present, the method 600a may proceed to steps 640a - 690a, which substantially correspond to the steps 520a - 570a, described above with respect to the method 500a in Fig. 7. In steps 640a - 690a, predictors are used to determine the probability of CPE. If it is determined that the probability of CPE is greater than a threshold value, outputs indicating a risk of ARDS are suppressed. If, in the step 630a, it is determined that pulmonary edema is not present, no further action is required.

[0032] In a second example, as shown in Fig. 11, a method

600b integrates the pulmonary edema differentiation module 112 with the method 500b, which integrates the ARDS risk module 114 with the pulmonary edema detection module 110. In a step 610b, an ARDS risk score is determined using the ARDS risk module 114. In a step 620b, the presence of pulmonary edema is detected via the pulmonary edema detection module 110 using the method 200, as described above with respect to Fig. 4. If it is determined, in a step 630b, that pulmonary edema is present, the method 600b may proceed to steps 640b - 690b, which substantially correspond to the steps 520a - 570b, described above with respect to the method 500a in Fig. 8. In the steps 640b - 690b, predictors are used to determine the probability of ARDS. This ARDS

probability is multiplied with the ARDS risk score determined in step 610b to determine a revised ARDS risk score which may be output in the step 690b. If it is determined, in the step 630b, that pulmonary edema is not present, no further action is required .

[0033] In a third example, as shown in Fig. 12, a method 600c integrates the ARDS risk module 114 along with the pulmonary edema detection and differentiation modules 110, 112, the

pulmonary edema differentiation module 112 is integrated with the method 500c. In a step 610c, an ARDS risk score is

calculated using the ARDS detection module 114. In a step 615c, it is determined whether the calculated ARDS risk score is greater than a threshold value. If the ARDS risk score does not exceed the threshold value, no further action is required.

Alternatively, the ARDS risk score calculated in the step 601c may be displayed. If the ARDS risk score is greater than the threshold value, the method 600c proceeds to a step 620c in which the presence of pulmonary edema is detected via the

pulmonary edema detection module 110 using, for example, the method 200, as described above in regard to Fig. 4. In a step 630c, it is determined whether pulmonary edema is present. If it is determined that the patient does not have pulmonary edema, no further action is required. Alternativelt , the ARDS risk score calculated in the step 601c may be displayed. If, however, pulmonary edema is present, the method 600c proceeds to steps 640c - 690c, which substantially correspond to steps 530c - 580c of the method 500c. In the steps 640c - 690c, predictors are used to determine the probability of ARDS. This ARDS

probability is multiplied with the ARDS risk score determined in step 610b to determine a revised ARDS risk score, which may be output in the step 690c.

[0034] It is noted that the claims may include reference signs/numerals in accordance with PCT Rule 6.2(b) . However, the present claims should not be considered to be limited to the exemplary embodiments corresponding to the reference

signs/numerals .

[0035] Those skilled in the art will understand that the above-described exemplary embodiments may be implemented in any number of manner, including, as a separate software module, as a combination of hardware and software, etc. For example, the pulmonary edema detection module 110, the pulmonary edema differentiation module 112 and the ARDS risk detection module 114 may be programs containing lines of code that, when compile may be executed on a processor.

[0036] It will be apparent to those skilled in the art that various modifications may be made to the disclosed exemplary embodiments and methods and alternatives without departing from the spirit or scope of the disclosure. Thus, it is intended that the present disclosure cover the modifications and

variations provided that they come within the scope of the appended claims and their equivalents.

Claims

What is claimed is:

1. A method for diagnosing pulmonary edema, comprising:

comparing patient information to a list of predictors in a predictor database to identify patient predictors;

assigning points based on the identified patient predictors ;

adding the points to determine an aggregated number of points; and

determining a probability of one of acute respiratory distress syndrome or cardiogenic pulmonary edema based on the aggregated number of points.

2. The method of claim 1, further comprising detecting

pulmonary edema in a patient.

3. The method of claim 2, wherein detecting pulmonary edema

includes :

detecting a presence of bilateral infiltrates; calculating a ratio of one of ( Pa0₂ ) / ( Fi0₂ ) or (Sp0₂) / (Fi0₂) ; and

determining whether the ratio of the one of

(Pa0₂) / (Fi0₂) or ( Sp0₂ ) / ( Fi0₂ ) is less than a threshold value, wherein when bilateral infiltrates are present and the ratio of the one of ( Pa0₂ ) / ( Fi0₂ ) or ( Sp0₂ ) / ( Fi0₂ ) is less than the threshold value, the pulmonary edema is detected.

4. The method of claim 1, further comprising determining an

ARDS risk score using at least one of patient vitals or labs .

5. The method of claim 4, wherein the patient vitals include one of heart rate, arterial systolic and diastolic blood pressure or respiratory rate.

6. The method of claim 4, further comprising multiplying the ARDS risk score with the probability of ARDS to determine a revised ARDS risk score.

7. The method of claim 6, further comprising outputting the revised ARDS risk score.

8. The method of claim 4, further comprising suppressing an ARDS risk output when the CPE probability is greater than or equal to a threshold value.

9. The method of claim 4, wherein the ARDS risk score is

determined via one of a Lempel-Ziv complexity metric, a logistic regression model, a log-likelihood ratio, an inference system and a Bayesian network.

10. The method of claim 1, wherein each predictor from the list of predictors is assigned a weight with a magnitude and a sign .

11. A system for diagnosing pulmonary edema, comprising:

a predictor database including a list of predictors; and

a processor comparing patient information to the list of predictors to identify patient predictors, assigning points based on the identified patient predictors and adding the points to determine an aggregated number of points to determine a probability of one of acute respiratory distress syndrome or cardiogenic pulmonary edema based on the aggregated number of points.

12. The system of claim 11, wherein the processor detects

pulmonary edema in a patient by detecting a presence of bilateral infiltrates, calculating a ratio of one of

( PaC>2 ) / ( F1O2 ) or ( SpC>2 ) / ( F1O2 ) , and determining whether the ratio of the one of ( Pa0₂ ) / ( Fi0₂ ) or ( Sp0₂ ) / ( Fi0₂ ) is less than a threshold value, wherein when bilateral infiltrates are present and the ratio of the one of ( PaC>2 ) / ( F1O2 ) or

( SpC>₂ ) / ( F1O₂ ) is less than the threshold value, the

pulmonary edema is detected.

13. The system of claim 11, wherein the processor determines an ARDS risk score using at least one of patient vitals or labs .

14. The system of claim 13, wherein the patient vitals include one of heart rate, arterial systolic and diastolic blood pressure or respiratory rate.

15. The system of claim 13, wherein the processor multiplies the ARDS risk score with the probability of ARDS to determine a revised ARDS risk score.

16. The system of claim 15, further comprising a display

outputting the revised ARDS risk score.

17. The system of claim 13, wherein the processor suppresses an ARDS risk output when the CPE probability is greater than or equal to a threshold value.

18. The system of claim 13, wherein the ARDS risk score is determined via one of a Lempel-Ziv complexity metric, a logistic regression model, a log-likelihood ratio, an inference system and a Bayesian network.

19. The system of claim 11, wherein each predictor from the list of predictors is assigned a weight with a magnitude and a sign.

20. A non-transitory storage medium including a set of

instructions executable by a processor, the set of

instructions operable to:

compare patient information to a list of predictors in a predictor database to identify patient predictors;

assign points based on the identified patient predictors ;

add the points to determine an aggregated number of points; and

determine a probability of one of acute respiratory distress syndrome or cardiogenic pulmonary edema based on the aggregated number of points .