WO2009083886A1

WO2009083886A1 - Presenting patient relevant studies for clinical decision making

Info

Publication number: WO2009083886A1
Application number: PCT/IB2008/055456
Authority: WO
Inventors: Colleen M. Ennett; Pradyumna Dutta; Stephen S. Ober; William P. Lord
Original assignee: Koninklijke Philips Electronics N.V.
Priority date: 2007-12-28
Filing date: 2008-12-19
Publication date: 2009-07-09

Abstract

A method of presenting options for relevant studies to be performed on a current patient case includes retrieving a similar patient case or multiple similar patient cases to the current patient case from a medical database. Clinical practice guidelines relevant to the current patient case are determined. The clinical practice guidelines are combined with the similar patient case(s) in order to present options for studies to be performed for the current patient case.

Description

PRESENTING PATIENT RELEVANT STUDIES FOR CLINICAL

DECISION MAKING

FIELD OF THE INVENTION

The invention relates to a method of and system for presenting relevant studies to be performed on a current patient case. The invention relates further to a computer program product, comprising instructions which, when carried out by a computer, causes the computer to carry out such a method.

BACKGROUND OF THE INVENTION

Evidence based medicine has often been a goal of decision support systems, but it can be difficult to get this information to the clinicians in an easily accessible and understandable format. In addition, interpreting the evidence from clinical studies and translating the results into how this information relates to the current patient is not a simple task. Clinical studies and practice guidelines apply to the "average" patient, and do not account for individual variances. Another source of evidence is drawing from a medical database of past patient cases. Clinicians easily understand this type of evidence because this is how they currently make decisions by themselves. Without computer assistance, clinicians draw only from their own experience, which is likely a smaller dataset because only a limited number of cases can be retrieved by memory. Presenting individualized information to the clinician can be achieved by finding similar cases from a medical database of previous cases to show known outcomes from various treatment options for patients with characteristics similar to the current patient.

Case based reasoning paradigms have been used to retrieve past cases that are similar to a present problem, with the recalled information being reused, possibly after an adaptation step. Clinical guidelines briefly identify, summarize and evaluate the best evidence and most current data about prevention, diagnosis, prognosis, therapy including dosage of medications, risk/benefit and cost-effectiveness. Then they define the most important questions related to clinical practice and identify all possible decision options and their outcomes. Some guidelines contain decision or computation algorithms to be followed. Thus, they integrate the identified decision points and respective courses of action to the clinical judgment and experience of practitioners. Many guidelines place the treatment alternatives into classes to help providers in deciding which treatment to use. The USA and other countries maintain medical guideline clearinghouses. In the USA, the National Guideline Clearinghouse maintains a catalog of high quality guidelines published by various organizations (mostly professional physician organizations). In the United Kingdom, clinical practice guidelines are published primarily by the National Institute for Health and Clinical Excellence (NICE). In The Netherlands, two bodies (CBO and NHG) publish specialist and primary care guidelines, respectively. In Germany, the Agency for Quality in Medicine (AEZQ) coordinates a national program for disease management guidelines. All these organizations are now members of the Guidelines International Network, an international not-for-profit association of organizations and individuals involved in clinical practice guidelines. G-I-N is owner of the International Guideline Library - the largest web based data base of medical guidelines worldwide. Additional objectives of clinical guidelines are to standardize medical care, to raise quality of care, to reduce several kinds of risk (to the patient, to the healthcare provider, to medical insurers and health plans) and to achieve the best balance between cost and medical parameters such as effectiveness, specificity, sensitivity, resolutiveness, etc. It has been demonstrated repeatedly that the use of guidelines by healthcare providers such as hospitals is an effective way of achieving the objectives listed above, although they are not the only ones.

US 6,029,138 describes an embodiment of a method that collects results from previous exams of recent patients to provide suggestions to the clinician on whether to follow through with the ordered study. The doctor enters information into the system about the patient's symptoms and a working diagnosis to justify ordering the study. Using this information, the system presents statistics from recent exams on similar patients indicating the number of studies with significant results that aided confirming the diagnosis out of the total number of studies ordered for those specific indications.

However, this method relies upon the physician to provide feedback as to the usefulness of a study.

BRIEF DESCRIPTION OF THE INVENTION In accordance with one aspect of the exemplary embodiment, a method includes retrieving at least one similar patient case to the current patient case from a medical database, determining clinical practice guidelines relevant to the current patient case, combining the clinical practice guidelines with the at least one similar patient case in order to present options for studies to be performed for the current patient case. The insight is that the combination of clinical practice guidelines with similar patient cases allows for certain studies to be recommended which may be incorporated into the guidelines but not be performed for the similar patient cases. In this way the recording and storage of patient cases is decoupled from personal preferences as to the usefulness of a study. Advantageously, the recording and storage of patient cases is also decoupled from local requirements as laid down in clinical guidelines.

According to an aspect of the exemplary embodiment, options may be presented for one or more studies to be performed for the current patient case. These options for studies may include one or more of: one or more tests which may be performed on the patient; and one or more treatments for the patient. The test or treatment may be one that is recommended by the guidelines and/ or have been shown to be associated with good outcomes in prior patient cases.

According to an aspect of the present application, the method comprises presenting information about the presented studies to be performed for the current patient case. By presenting information about the studies, a clinician gets more insight into all the studies that can be performed for the current patient case. It may help the clinician to forecast potential responses to studies and treatment for the current patient based on past cases, while still adhering to the appropriate clinical practice guidelines.

According to another aspect of the present application, the information includes at least one of raw data or statistics, and the statistics may include at least one of: the occurrence of a study ordered for a similar patient; the response to treatment of a similar patient. The information that the method outputs may be presented to the clinician as raw data, or in the form of statistics to provide evidence to make the most appropriate decision based on patient-specific information. These statistics may be presented in the form of tables or graphs that aggregate information and compare the statistics of the matched cases and the unmatched cases from a medical database to the current patient.

In accordance with another aspect of the exemplary embodiment, a system includes a retriever designed for retrieving at least one similar patient case to the current patient case from a medical database, a determiner designed for determining clinical practice guidelines relevant to the current patient case, a combiner designed for combining the clinical practice guidelines with the at least one similar patient case in order to present options for studies to be performed for the current patient case. In accordance with another aspect of the exemplary embodiment, a computer program product includes instructions which, when carried out by a computer, cause the computer to carry out a method according to the invention.

The advantages and effects achieved by the system and computer program product are as described with reference to the method according to the invention.

In accordance with another aspect of the exemplary embodiment, a computer implemented method of presenting options for relevant studies to be performed on a current patient case includes inputting data about a current patient extracted from patient records to a reasoning engine, automatically identifying a set of matching patient cases in a medical database of cases using a similarity metric to measure how similar the cases are to one another, automatically identifying an outcome of an intervention for each of the matching patient cases in the set, automatically sorting the matched patient cases based on the desirability of their outcome, automatically identifying at least one clinical practice guideline relevant to the current patient case from a set of stored clinical guidelines, and automatically combining the at least one clinical practice guideline with at least one of the sorted patient cases in order to present options for studies to be performed for the current patient case.

BRIEF DESCRIPTION OF THE DRAWINGS

These and other aspects of the invention will be apparent from and elucidated with reference to the embodiments described hereinafter as illustrated by the following Figures:

Figure 1 illustrates a method according to the invention in a schematic way.

Figure 2 illustrates a raw representation of information of patients in a schematic way; Figure 3 schematically illustrates a statistical representation of information of patients summarized in graphical format;

Figure 4 schematically illustrates a statistical representation of information of patients;

Figure 5 illustrates a system according to the invention in a schematic way.

DESCRIPTION OF EMBODIMENTS

Figure 1 illustrates a method according to the invention in a schematic way. Within step 102 data about the current patient is extracted from patient records. Such records may include electronic medical records, nursing notes, doctors' notes, discharge notes and similar sources of information e.g., admission diagnosis, working diagnosis, and final diagnosis. This includes, but is not limited to: information about the patient's demographics, e.g., age, gender, weight, height, medical history, e.g., clinical, therapeutic and diagnostic events, comorbidities, previously prescribed medications, current physiological data, e.g., heart rate, respiratory rate, mean arterial blood pressure, blood glucose, administration of medications, e.g., date, time, dose, frequency, test results e.g., ECG testing, imaging, blood test, and treatment results e.g., surgical procedure, radiation treatment, medical treatment.

These items may be nominal, discrete-valued or continuous-valued data. The extent of information about the current patient will depend on the intended use of the decision support system i.e., for determining diagnosis or treatment options. The information about the current patient includes all information collected up until the decision point under consideration.

Similarly, a medical database of previous patient cases includes data on patient demographics, medical history, physiological data, test results, treatment results, and other relevant data.

These cases contain all diagnosis and treatment information necessary for matching with the current patient. These patient cases may be derived from a local, regional or national medical database that is representative of the clinical presentation of the current patient. The data set of previous patients is representative of the diagnosis and treatment options available for the current patient.

Advantageously, the data is pre-processed as necessary to meet the needs of a selected reasoning engine, as they are well known in the art. Pre-processing includes analysing the data and identifying potential outliers and errors, which are treated appropriately using standard statistical approaches. To attempt to correct for systematic bias in the data, pre-processing of the data may include normalizing the data to account for skewedness, kurtosis, differences in orders of magnitude of the data, etc.

The choice of parameters that can be used for matching is preferably based on one or more of: an expert opinion, e.g., as identified in clinical practice guidelines specific to the clinical presentation of the current patient, clinical studies, etc., system optimized, e.g., using principle component analysis, weight-extraction from artificial neural networks, logistic regression or other analysis tools, physician selected, e.g., the clinician has the option to include or exclude parameters as he or she feels is appropriate to find the closest matches to the current patient.

Each input parameter is assigned match weights to be used for identifying similar cases. The weights are selected based on one or more of: an expert opinion, e.g., indicated emphasis on particular parameters as identified in clinical practice guidelines specific to the clinical presentation of the current patient, results of randomized controlled trials, system optimized, e.g., using principle component analysis PCA, weight- extraction from artificial neural networks, logistic regression or other analysis tools, uniform weighting, e.g., no emphasis on any particular parameter.

The reasoning engine draws together the information contained in the current patient's data and the medical database of previous patient cases to identify similar patient profiles using a similarity metric to measure how similar the cases are to one another. The number of similar cases may be user-specified or system-specified based on the desired number of closest-matching cases, or based on the value of the similarity metric. The underlying technology of the reasoning engine could involve case-based reasoning, k-nearest neighbors, adaptive neuro-fuzzy inference system, clustering techniques, artificial neural network, etc. The information extracted from the reasoning engine may be combined with data from one or more additional decision support modules to forecast the patient's response to diagnostic testing and/or treatment procedures.

Within step 104 the clinical guideline that is applicable for the current patient is retrieved from a database comprising clinical guidelines based upon the patient's admission diagnosis or working diagnosis automatically. Alternatively, a physician selects the clinical guideline himself from the database. The specific information about the current patient to match on can be determined either by the physician/user selecting data entries or, if used in conjunction with established clinical guidelines, protocols, etc., can be automatically selected and potentially fine-tuned by the user.

Within step 106 the selected clinical guideline is combined with the similar patient cases in order to present (e.g., suggest or recommend) optional studies to be performed for the current patient case. The advice is in the form of raw data, or in the form of statistics to provide evidence to make the most appropriate decision based on patient-specific information. These statistics may be presented in the form of tables or graphs that aggregate information and compare the statistics of the matched cases and the unmatched cases from a medical database to the current patient's profile. The information can include not only the test or treatment results of similar patient cases, but also the physiological data either static or continuous, occurring after the intervention to show that it was the "right" thing to do for these similar patients, i.e., that their condition improved, etc. For example, it's not enough to simply know that certain drugs were given to 90% of similar patients that are similar to the current patient. The care provider wants to know that the drugs achieved the result they intended to achieve. The physiological data shown would be relevant to the test and/or treatment under consideration. For example, if one intervention was the administration of an anti-hypertensive drug, the physiological data shown to the care provider might be mean arterial blood pressure, heart rate, etc.

An appropriate time window is associated with the intervention and its expected outcome to determine the effectiveness of the intervention (e.g., taking into consideration how long it takes the intervention to impact the patient's health status). If the similar patient case includes a good expected outcome satisfying the time window, then it is more meaningful since a good outcome should have occurred within the time window if the intervention was going to be effective.

There may be multiple outcomes that are desirable, as well as multiple interventions that can achieve the same outcome. For example, if the event is "Acute blood pressure drop," the desired outcome is "Blood pressure increased," and the interventions may be "Intravenous fluids given" or "Drug A given." Also, another coding system will indicate what the desired response from the intervention is so that the system knows whether the desired response was achieved. The system may also analyze the past patient information ahead of time to automatically catalogue whether the test or treatment achieved the desired results.

Within optional step 108, the outcome is presented in the form of trend graphs, a discrete subsequent measurement, or summary statistics that demonstrate whether the associated intervention had the desired effect. The outcome may be one or more physiological parameters that would be monitored for the specified clinical event. For example, Table 1 describes the monitored variables for sepsis such as temperature, respiratory rate, heart rate, and blood pressure, and also test results such as white blood cell count, blood gases, platelet count, and blood/urine culture. By clicking on an individual patient profile that was identified as a close match to the current patient via a similarity matching algorithm, the user could open a new window showing trend graphs for these variables before and after the intervention as evidence that the intervention had the desired effect.

Interventions, such as medications, may be grouped into a hierarchy, starting with the most general, and proceeding to as specific a description as possible. This information can be displayed in an expandable/collapsible format for the user to view more specific information as desired. For example, a hierarchy for a medication to increase blood pressure could be (from general to specific):

Hypotensive medication → Catecholamines → Central acting → Dopamine

Table 1

The information may also be displayed to the user by sorting the matched patients based on the desirability of their outcome (good, bad, no change) and providing statistics indicating the interventions used for this cohort of patients. Using "Acute blood pressure drop" as an example, the outcome categories would be: good ("Blood pressure increases or stabilizes"), bad ("Blood pressure drops at a faster rate"), and no change ("Blood pressure continues to drop"). Then the statistics of the interventions used in each of the categories would be presented to the user.

Another way to display information is to sort the matched cases based on the interventions used and provide statistics on the desirability of their outcomes (good, bad, or no change). Alternatively, the interventions could be categorized based on whether the patient responded to the intervention or not. Figure 2 illustrates a raw representation of information of patients in a schematic way.

Figure 3 illustrates a statistical representation of information of patients summarized in graphical format in a schematic way. A feature of the graphical display shown in Figure 3 could be to allow the clinician to double-click on the desired bar to reconfigure the display to show the statistics for just that specific patient population in comparison with the current patient. Also for Figures 2 and 3, the key parameters chosen for case matching could be highlighted on these displays to remind the clinician which parameters were used to find the matches and also to show the match weights of those parameters. At this point, one possible scenario may be that the physician may integrate his or her clinical experience into the decision-making process with respect to previous cases he or she has treated or consulted on for other clinicians, and current clinical knowledge drawn from reading medical publications, textbooks and clinical practice guidelines, etc. The patient's values with respect to unique preferences, concerns and expectations possibly including quality of life, financial constraints, etc. may also be incorporated by the physician into the decision-making process. The physician may aggregate his or her clinical experience, the patient's values and the results of the case-matching system to select the best next procedure based on evidence from those three sources.

Figure 4 illustrates a statistical representation of information of patients in a schematic way.

Figure 5 illustrates a system according to the invention in a schematic way. The system 500 comprises memories 502, 504, 506, and optionally 508 designed to comprise instructions which, when carried out by a processor 512 causes the system 500 to carry out the method according to the invention. The system further comprises a patient database 518, a clinical guidelines database 520. The memories, processor, and databases are connected to each other through communication software bus 510. The system 500 is connected through an output/input interface 514 to a display 516. The memory 502 ("retriever") comprises instructions for carrying out retrieving a similar patient case or multiple similar patient cases to the current patient case from a medical database. The memory 504 ("determiner") comprises instructions for carrying out determining clinical practice guidelines relevant to the current patient case. The memory 506 ("combiner") comprises instructions for carrying out combining the clinical practice guidelines with the similar patient case(s) in order to present options for studies to be performed for the current patient case. The combiner 506 may include a classifier for classifying (e.g., ranking) the outcomes of the interventions performed in the retrieved similar patient cases, taking into consideration the time window between the intervention and any expected good outcome, or a retriever for retrieving those classifications from associated memory and, based on the classification of the outcomes, identifying studies which may be appropriate for the current patient. The memory 508 ("presenter") comprises instructions for carrying out presenting information about the presented studies to be performed for the current patient case. The patient database 518 comprises patient data as previously described. The clinical guidelines database 520 comprises clinical guidelines as previously described. The display 516 allows for user interaction with the system, by enabling a user to provide input to the system and displaying the output of the system to the user.

The output comprises feedback of information about the current patient and matching patients as previously described. The whole system may have a distributed nature in which for example the databases are located at a different geographical position then the memories, software bus and the processor. In this case the different parts of the system can communicate with each other making use of well known wired or wireless communication protocols.

The method illustrated in Figure 1 may be implemented in a computer program product that may be executed on a computer. The computer program product may be a tangible computer-readable recording medium on which a control program is recorded, such as a disk, hard drive, or may be a transmittable carrier wave in which the control program is embodied as a data signal. Common forms of computer-readable media include, for example, floppy disks, flexible disks, hard disks, magnetic tape, or any other magnetic storage medium, CD-ROM, DVD, or any other optical medium, a RAM, a PROM, an EPROM, a FLASH-EPROM, or other memory chip or cartridge, transmission media, such as acoustic or light waves, such as those generated during radio wave and infrared data communications, and the like, or any other medium from which a computer can read and use. It should be noted that the above-mentioned embodiments illustrate rather than limit the invention, and that those skilled in the art will be able to design many alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. The word "comprising" does not exclude the presence of elements or steps other than those listed in a claim. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The invention can be implemented by means of hardware comprising several distinct elements, and by means of a suitably programmed computer. In the system claims enumerating several means, several of these means can be embodied by one and the same item of computer readable software or hardware. The mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measures cannot be used to advantage.

Claims

CLAIMS:

1. A method of presenting options for relevant studies to be performed on a current patient case, the method comprising: retrieving at least one similar patient case to the current patient case from a medical database; determining at least one clinical practice guideline relevant to the current patient case; combining the at least one clinical practice guideline with the at least one similar patient case in order to present options for studies to be performed for the current patient case.

2. The method according to claim 1, comprising presenting information about the presented options for studies to be performed for the current patient case.

3. The method according to claim 2, wherein the information includes at least one of raw data or statistics, and the statistics may include at least one of: the occurrence of a study ordered for a similar patient; the response to treatment of a similar patient.

4. The method according to claim 1, further comprising: identifying a time window between an intervention used in the treatment of a patient and an expected good outcome of the intervention and determining whether the similar patient case includes an outcome satisfying the time window.

5. The method according to claim 4, further comprising showing summary statistics regarding the rate of occurrence of treatment for patient cases retrieved from a medical database that are similar to the current patient case.

6. The method according to claim 1, further comprising: sorting a plurality of the matched patient cases based on the desirability of their responses to interventions.

7. The method according to claim 6, further comprising: providing statistics indicating the interventions used for the plurality of patient cases and their responses to the interventions.

8. The method according to claim 6, further comprising: establishing a coding system of discrete codes for categorizing responses of patients to interventions and categorizing the matched patient cases with codes based on their responses.

9. The method according to claim 8, wherein the discrete codes include codes for at least a good response, a bad response and no response.

10. The method according to claim 1, wherein the determining of the clinical practice guideline includes retrieving the clinical practice guideline for the current patient from a database comprising clinical guidelines based upon the patient's admission diagnosis or working diagnosis.

11. The method according to claim 1 , wherein the combining includes identifying, from the guidelines, interventions recommended by the guidelines for the current patient, indentifying outcomes of at least some of the recommended interventions from the similar patient cases, and generating a representation of the interventions recommended by the guidelines and their identified outcomes from the similar patient cases for presenting to a user.

12. A system for performing the method according to claim 1 comprising a processor which executes instructions stored in memory for performing the method and memory which stores the instructions and further comprising a medical database of patient cases accessible to the processor and optionally a set of stored clinical guidelines accessible to the processor.

13. The system according to claim 12, wherein the processor further executes instructions for sorting a plurality of matched patient cases based on the desirability of their outcomes and for providing statistics indicating the interventions used for the plurality of patient cases.

14. A system of presenting options for relevant studies to be performed on a current patient case, the system comprising: a retriever designed for retrieving at least one similar patient case to the current patient case from an associated medical database; a determiner designed for determining clinical practice guidelines relevant to the current patient case; and a combiner designed for combining the clinical practice guidelines with the at least one similar patient case in order to present options for studies to be performed for the current patient case.

15. The system according to claim 14, further comprising a presenter designed for presenting information about the presented options for studies to be performed for the current patient case.

16. The system according to claim 15, wherein the information includes at least one of raw data or statistics, and the statistics may include at least one of: the occurrence of a study ordered for a similar patient; the response to treatment of a similar patient.

17. A computer program product comprising instructions which, when carried out by a computer, cause the computer to carry out a method according to claim 1.

18. A computer implemented method of presenting options for relevant studies to be performed on a current patient case, the method comprising: inputting data about a current patient extracted from patient records to a reasoning engine; automatically identifying a set of matching patient cases in a medical database of cases using a similarity metric to measure how similar the cases are to one another; automatically identifying an outcome of an intervention for each of the matching patient cases in the set; automatically sorting the matched patient cases based on the desirability of their outcome; automatically identifying at least one clinical practice guideline relevant to the current patient case from a set of stored clinical guidelines; and automatically combining the at least one clinical practice guideline with at least one of the sorted patient cases in order to present options for studies to be performed for the current patient case.

19. The method according to claim 18, comprising presenting information about the presented options for studies to be performed for the current patient case.

20. The method according to claim 18, further comprising: automatically identifying a time window between an intervention used in the treatment of a patient and an expected good outcome of the intervention and determining whether the similar patient case includes an outcome satisfying the time window.

21. The method according to claim 18, further comprising: providing statistics indicating the interventions used for the plurality of patient cases and their responses to the interventions.

22. The method according to claim 21, further comprising: an established coding system of discrete codes for categorizing responses of patients to interventions, categorizing the matched patient cases with codes based on their responses.

23. A computer program product comprising instructions which, when carried out by a computer, cause the computer to carry out a method according to claim 17.