Multi-stage model for predicting probabilities of mortality in adult critically ill patients

• US 8,417,541 B1
• Filed: 08/27/2009
• Issued: 04/09/2013
• Est. Priority Date: 08/27/2009
• Status: Active Grant

• Current Assignee: Cerner Innovation, Inc. (Oracle Corporation)
• Original Assignee: Cerner Innovation, Inc. (Oracle Corporation)
• Inventors: Kramer, Andrew A.
• Primary Examiner(s): Morgan, Robert
• Assistant Examiner(s): LAM, ELIZA ANNE
• Application Number:
  US12/549,153
• Time in Patent Office: 1,321 Days
• Field of Search: None
• US Class Current: 705/2
• CPC Class Codes: G06Q 10/10 Office automation; Time management
• CPC Class Codes: G16H 50/30 for calculating health indices; for individual health risk assessment
• CPC Class Codes: G16Z 99/00 Subject matter not provided for in other main groups of this subclass

Abstract

Computerized methods in a clinical computing environment for predicting mortality in critically ill patients, that is, patients admitted to Intensive Care Units, are provided. In accordance with embodiments hereof, at least two distinctly different mortality prediction models (e.g., the Acute Physiology and Chronic Health Evaluation (APACHE®) model and the Mortality Probability Model at Admission (MPM₀) are utilized in a multi-stage fashion to determine, with better accuracy than may be provided by either mortality prediction model alone, the probability of mortality for critically ill adult patients.

Claims

1. One or more non-transitory computer-storage media having computer-executable instructions embodied thereon for performing a method in a clinical computing environment for predicting a probability of mortality for a patient admitted to an Intensive Care Unit, the method comprising:

determining a length of a first Intensive-Care-Unit-day for the patient;

calculating a first mortality probability utilizing a first mortality prediction model, the first mortality prediction model being based on physiological data items collected within the patient'"'"'s first Intensive-Care-Unit-day;

calculating a second mortality probability utilizing a second mortality prediction model, the second mortality prediction model being based on clinical data items collected within one hour of the patient being admitted to the Intensive Care Unit;

determining a difference between the second mortality probability and the first mortality probability;

calculating the probability of mortality for the patient utilizing Equation 1;

e^Yi/(1+e^Yi),

wherein Y_i is calculated utilizing Equation 2;

Y_i=β

₀+(β

₁)(A)+(β

₂)(λ

)+(β

₃)(δ

_(MA)),

wherein Y_i equals zero if the patient survives and one if the patient dies, wherein A equals the logit of the first mortality probability, wherein λ

equals the length of the first Intensive-Care-Unit-day, wherein δ

_(MA) equals one if the difference between the second mortality probability and the first mortality probability exceeds 0.15 and zero if the difference between the second mortality probability and the first mortality probability is less than or equal to 0.15, and wherein values for β

₀, β

₁, β

₂, and β

₃ are obtained through a logistic regression procedure using Equation 2; and

assessing ICU performance by comparing observed and predicted mortality.

2. The one or more computer-storage media of claim 1, wherein determining the length of the first Intensive-Care-Unit-day for the patient comprises:

providing a time that the patient is admitted to the Intensive Care Unit;

determining if a minimum of eight hours exists between the time that the patient is admitted to the Intensive Care Unit and the next subsequent day'"'"'s start time,

wherein if a minimum of eight hours exists between the time that the patient is admitted to the Intensive Care Unit and the next subsequent day'"'"'s start time, the length of the first Intensive-Care-Unit day for the patient is equal to the number of hours between the time that the patient is admitted to the Intensive Care Unit and the next subsequent day'"'"'s start time,

and wherein if a minimum of eight hours does not exist between the time that the patient is admitted to the Intensive Care Unit and the next subsequent day'"'"'s start time, the length of the first Intensive-Care-Unit day for the patient is equal to the number of hours between the time that the patient is admitted to the Intensive Care Unit and the next subsequent day'"'"'s start time plus twenty-four.

3. The one or more computer-storage media of claim 1, wherein calculating the first mortality probability utilizing the first mortality prediction model comprises calculating the first mortality probability utilizing the Acute Physiology and Chronic Health Evaluation mortality prediction model.

4. The one or more computer-storage media of claim 3, wherein calculating the first mortality probability utilizing the Acute Physiology and Chronic Health Evaluation mortality prediction model comprises calculating the first mortality probability utilizing the Acute Physiology and Chronic Health Evaluation IV mortality prediction model.

5. The one or more computer-storage media of claim 1, wherein calculating the second mortality probability utilizing the second mortality prediction model comprises calculating the second mortality probability utilizing the Mortality Probability Model at Admission.

6. The one or more computer-storage media of claim 5, wherein calculating the second mortality probability utilizing the Mortality Probability Model at Admission comprises calculating the second mortality probability utilizing the Mortality Probability Model at Admission, version III.

7. The one or more computer-storage media of claim 1,

wherein calculating the first mortality probability utilizing the first mortality prediction model comprises calculating the first mortality probability utilizing the Acute Physiology and Chronic Health Evaluation mortality prediction model,

and wherein calculating the second mortality probability utilizing the second mortality prediction model comprises calculating the second mortality probability utilizing the Mortality Probability Model at Admission.

8. The one or more computer-storage media of claim 7,

wherein calculating the first mortality probability utilizing the Acute Physiology and Chronic Health Evaluation mortality prediction model comprises calculating the first mortality probability utilizing the Acute Physiology and Chronic Health Evaluation IV mortality prediction model,

and wherein calculating the second mortality probability utilizing the Mortality Probability Model at Admission comprises calculating the second mortality probability utilizing the Mortality Probability Model at Admission, version III.

9. A method in a clinical computing environment for predicting a probability of mortality for a patient admitted to an Intensive Care Unit, the method comprising:

determining a length of a first Intensive-Care-Unit-day for the patient;

calculating, via a first computing process, a first mortality probability utilizing the Acute Physiology and Chronic Health Evaluation mortality prediction model;

calculating, via a second computing process, a second mortality probability utilizing the Mortality Probability Model at Admission mortality prediction model;

determining a difference between the second mortality probability and the first mortality probability;

calculating, via a third computing process, the probability of mortality for the patient utilizing Equation 1;

e^Yi/(1+e^Yi),

wherein Y_i is calculated via a fourth computing process utilizing Equation 2;

Y_i=β

₀+(β

₁)(A)+(β

₂)(λ

)+(β

₃)(δ

_(MA)),

wherein Y_i equals zero if the patient survives and one if the patient dies, wherein A equals the logit of the first mortality probability, wherein λ

equals the length of the first Intensive-Care-Unit-day, wherein δ

_(MA) equals one if the difference between the second mortality probability and the first mortality probability exceeds 0.15 and zero if the difference between the second mortality probability and the first mortality probability is less than or equal to 0.15, and wherein values for β

₀, β

₁, β

₂, and β

₃ are obtained through a logistic regression procedure using Equation 2,

and wherein the first, second, third, and fourth computing process are performed utilizing one or more computing devices; and

assessing ICU performance by comparing observed and predicted mortality.

10. The method of claim 9, wherein determining the length of the first Intensive-Care-Unit-day for the patient comprises:

providing a time that the patient is admitted to the Intensive Care Unit;

determining if a minimum of eight hours exists between the time that the patient is admitted to the Intensive Care Unit and the next subsequent day'"'"'s start time,

wherein if a minimum of eight hours exists between the time that the patient is admitted to the Intensive Care Unit and the next subsequent day'"'"'s start time, the length of the first Intensive-Care-Unit day for the patient is equal to the number of hours between the time that the patient is admitted to the Intensive Care Unit and the next subsequent day'"'"'s start time,

and wherein if a minimum of eight hours does not exist between the time that the patient is admitted to the Intensive Care Unit and the next subsequent day'"'"'s start time, the length of the first Intensive-Care-Unit day for the patient is equal to the number of hours between the time that the patient is admitted to the Intensive Care Unit and the next subsequent day'"'"'s start time plus twenty-four.

11. The method of claim 9, wherein calculating, via the first computing process, the first mortality probability utilizing the Acute Physiology and Chronic Health Evaluation mortality prediction model comprises calculating the first mortality probability utilizing the Acute Physiology and Chronic Health Evaluation IV mortality prediction model.

12. The method of claim 9, wherein calculating, via the second computing process, the second mortality probability utilizing the Mortality Probability Model at Admission mortality prediction model comprises calculating the second mortality probability utilizing the Mortality Probability Model at Admission, version III.

13. The method of claim 9,

wherein calculating, via the first computing process, the first mortality probability utilizing the Acute Physiology and Chronic Health Evaluation mortality prediction model comprises calculating the first mortality probability utilizing the Acute Physiology and Chronic Health Evaluation IV mortality prediction model,

and wherein calculating, via the second computing process, the second mortality probability utilizing the Mortality Probability Model at Admission mortality prediction model comprises calculating the second mortality probability utilizing the Mortality Probability Model at Admission, version III.