WO2008056128A1

WO2008056128A1 - Method and apparatus for providing medical information

Info

Publication number: WO2008056128A1
Application number: PCT/GB2007/004228
Authority: WO
Inventors: Oliver John Gibson; Lionel Tarassenko
Original assignee: T+ Medical Limited
Priority date: 2006-11-06
Filing date: 2007-11-06
Publication date: 2008-05-15
Also published as: GB0622091D0

Abstract

A method and system for providing information relating to a chronic medical condition to a patient comprises receiving a measurement of the patient's current medical condition, accessing a stored database of past cases for that patient using the current measurement as an index, and retrieving similar cases. Each of the cases consist of a record of the initial measurement of the patient's condition, the treatment following in that instance and the result of the treatment. Rather than one measurement a series of measurements may be used to access cases with a similar series of measurements. Cases in the database can be classified by the measurement or series, and weighting can be used to direct attention towards significant values or recent measurements. Pattern matching or a distance metric can be used to find similar cases in the database. For example in the case of insulin-dependent diabetes the initial measurement is a measurement of blood glucose level, the record of treatment followed is a record of the insulin dose taken, and the outcome is a record of the subsequent blood glucose level. The method and system display to the patient the treatments followed in each of the similar cases retrieved together with the outcome of those treatments, thus allowing the patient to select a treatment based on past experience. In the case of diabetes the display is of the subsequent blood glucose level against insulin dose taken.

Description

METHOD AND APPARATUS FOR PROVIDING MEDICAL INFORMATION

The present invention relates to a method and apparatus for providing information relating to a medical condition, in particular a chronic medical condition, and more particularly to the provision of the information from a database accessed by reference to a measurement of the patient's current condition, which information is then of assistance to the patient in deciding on treatment.

In the case of many chronic medical conditions improved clinical outcomes can be obtained by allowing patients to monitor and, to an extent, manage the treatment of their condition themselves (self-management). Where such management and treatment are performed well, the closer day-to-day adjustment of treatment improves control of the patient's condition compared to the more intermittent adjustments based on patients' visits to a clinic. Further, the less quantifiable health benefits deriving from the psychological improvement in a patient who feels that they are in control of their condition should not be ignored.

Thus in recent years systems allowing patients to monitor their condition and store and review measurements of their condition have been made available conveniently on mobile data processing terminals such as Personal Digital Assistants (PDAs), mobile telephones etc. as disclosed, for example, in WO/2004/027676 and WO/2005/087091. Such systems (or at least those put into practical use) have stopped short of giving the user advice as to future treatment. This is because of the serious liability issues involved if the treatment proves inappropriate. For example, in the case of insulin-dependent diabetes an injection of too much insulin could lead to coma and possibly death of the patient. Thus automated systems for providing advice are not trusted. Furthermore, health professionals do not want chronic sufferers of medical conditions to become over-reliant on physical devices because it leaves the patient in a difficult situation if the device is forgotten, lost or ceases to function. Finally, dependence on automatically-provided advice tends to go against the concept of patient empowerment (that is to say where the user makes the decision of how to treat themselves based on information provided to them). Although insulin-dependent (type 1) diabetes has been mentioned above, it will be appreciated that similar situations arise with other chronic medical conditions such as type 2 diabetes treated through diet and exercise or oral medication, insulin- treated type 2 diabetes, hypertension, asthma, Chronic Obstructive Pulmonary Disease (COPD)₅ etc.

A range of expert systems have been proposed in the medical field for allowing clinicians, and more occasionally patients, to access expert knowledge without needing to refer to a specialist. The paper "Advancements and Trends in Medical Case-Based Reasoning: An Overview of Systems and System Development" by Nilsson and Sollenborn, American Association for Artificial Intelligence, 2004, reviews a variety of such systems which utilize case-based reasoning. In this technique a plurality of cases are stored in the database and each case is provided with some form of numerical parameterisation which allows the assessment of which of the cases in the database are similar to any given other case, for example a new case presented to the system, to be made using a suitable similarity metric. One such project reviewed in that paper (the Auguste project) uses case-based reasoning to plan treatment of Alzheimer's disease by retrieving from the database a nearest neighbour match to a presented case and suggesting whether or not to treat with a drug based on whether or not a drug was used to treat the case corresponding to the nearest neighbour match. The system then uses 100 features of the presented case in a rule-based reasoning module to determine the specific treatment. The system has, however, not been put into practical use or commercialised and reasons for this may, again, be lack of trust in the answers provided. The present invention aims to overcome these problems while utilizing more effectively the enhanced processing power now provided in mobile data processing terminals such as PDAs and mobile telephones. It aims to do this without the disadvantages of specifying a particular treatment which could result in difficulties if the treatment is wrong or ineffective and disempowerment of the patient. Accordingly, a first aspect of the present invention provides a method of providing information relating to a chronic medical condition to a patient suffering from that condition, the method comprising: receiving a measurement of the patient's current medical condition; accessing a stored database of past cases on the basis of the measurement to retrieve similar cases, each of said stored cases comprising the value of a previous measurement of the patient's condition and the treatment followed; characterised in that: each of said stored cases further comprises the outcome of the treatment; and by the step of: displaying to the patient the treatments followed in each of the similar cases and the outcome of the treatments.

Thus with the present invention each of the stored cases includes the outcome of the treatment, and the display to the patient is a display of the treatments followed in each of the similar cases and of the outcome of those treatments. This allows the patient themselves to consider and select an appropriate treatment based on their view of the range of past treatments and outcomes for conditions similar to their current condition. It thus avoids the problems of dictating a particular course of action to the patient.

Preferably the past cases in the stored database are each classified to one of a plurality of classes based on the measurement of the patient's condition. For example each class can relate to a range of values of the measurement. Further, the displayed cases can be the cases in that class to which the received measurement of the patient's current conditions corresponds.

The stored cases may each comprise a plurality of previous measurements e.g. a time series of measurements, and the past cases in the database may each classified to one of a plurality of classes based on the plurality of previous measurements of the patient's condition. The classification may be based on patterns in the series of measurements, e.g. increasing/decreasing values over time or oscillatory patterns indicating high variability, and may be based on a non-uniform weighting of the series. The weighting may be non-linear or time-based.

The selection of cases for display may be on the basis of a time series of measurements (e.g. the latest measurement and some preceding ones) of the patient's condition again with weighting or pattern matching. A distance metric such as Euclidian distance may be used to decide which stored cases to display.

It may be preferred that only past cases with a positive outcome are displayed, and optionally cases with positive outcomes may be stored in a different database from those with less good outcomes. Thus in addition to storing all the cases (for record purposes) - which can be on the patient's device or at a remote server, there may be a database on the patient's device storing only a selected subset of cases suitable for display.

In the case of application of the invention to insulin-dependent diabetes, for example, the "measurement of the patient's condition" is the blood glucose level measurement and the "treatment followed" is the insulin dose. The "outcome of treatment" is the resultant blood glucose level measured at a later time. In this application the past cases can be classed by initial blood glucose measurement. For example, four ranges could be defined as: "amber" corresponding to 10-13 mmol/1; "amber-red" corresponding to, say, 13-16 mmol/1; "red" corresponding to 16-20 mmol/1; "very red" corresponding to greater than 20 mmol/1. Of course the specific ranges can be different for different patients.

The display can be a two-dimensional plot of the resultant blood glucose level against the insulin dose taken for the cases in the class corresponding to the received measurement of blood glucose level. More specifically, and using the example above, if the patient measures their blood glucose level as being 18 mmol/1, this places them in the red zone and so the display would be a display of the resultant blood glucose level for each different insulin dose for the different cases in that class. This allows the patient to see what results were obtained by different treatments when they were previously in a similar condition. This allows them to choose the appropriate insulin dose. Correct behaviour may be reinforced by only displaying those cases (and the corresponding insulin doses) which led to a positive outcome (i.e. a resultant blood glucose level in the normoglycaemia range, say 4 to 10 mmol/1). The invention is, of course, applicable to other chronic medical conditions such as asthma, type 2 diabetes treated through diet and exercise or oral medication, insulin-treated type 2 diabetes, hypertension, Chronic Obstructive Pulmonary Disease (COPD), etc..

In the case of asthma, the "measurement of the patient's condition" is the peak flow measurement (Peak Expired Flow - PEF - or Forced Expired Volume - FEVl) and the "treatment followed" is the number of puffs of inhaler (reliever and/or preventer). In the case of hypertension, the "measurement of the patient's condition" is the blood pressure and the "treatment followed" is the dosage of drug taken.

Preferably each time the patient takes a measurement of their condition, treats their condition and makes a further measurement, these are added as a new case to the database. This allows the database to build up a good level of knowledge of the patient's own response to treatment. It will be appreciated, therefore, that preferably the database comprises cases only for the particular patient concerned. However it may that for a new user the database is populated with typical cases, these preferably being replaced gradually by the patient's own cases. It is possible also that for self- management purposes the database comprises mostly cases with positive outcomes, whereas the cases with negative outcomes (resultant blood glucose levels outside the normoglyceamia range) could be stored in an another database, primarily for education purposes when the patient discusses the management of their condition with their clinician either at the Diabetes Clinic or during a telephone consultation. The display may comprise the average or modal outcome of each of the different past treatments in the similar cases retrieved from the database.

Preferably the measurement is received by a mobile data processing terminal, such as a mobile telephone or PDA, access to the stored database is via that terminal and the display is on the terminal. This allows the invention to fit-in easily with a modern lifestyle when many people own a PDA or mobile telephone or both. Preferably the database is stored on the mobile data processing terminal itself.

The invention may be embodied as a computer program comprising program code means which are effective to execute the method when run on a data processing device. The invention also extends to a mobile data processing terminal programmed with such a computer program.

The invention will be further described by way of example with reference to the accompanying drawings in which: Figure IA and B are flow diagrams illustrating the process flow of an embodiment of the invention;

Figure 2 shows the information displayed to the user in one embodiment of the invention; and Figure 3 shows the information displayed to the patient in another embodiment of the invention.

An example of the invention applied to the treatment of insulin-dependent diabetes will now be described. Typically a person suffering from Type 1 diabetes is required to measure their blood glucose level using a commercially-available meter periodically during the day, for example four times per day. The patient records those measurements and also is expected to inject themselves with insulin to control their condition, with the dosage depending on the measured blood glucose level. Patients who are skilled and experienced manage to control their blood glucose level within reasonably close tolerances, and this significantly improves their long-term health prognosis. Inexperienced patients may find control difficult and their blood glucose level tends to fluctuate greatly. The particular difficulty is determining which insulin dose is appropriate given the current blood glucose measurement. The present invention provides the patient with a method executable on a normal mobile data processing device such as a mobile telephone or PDA which indicates to them what outcomes have followed what insulin dosages in the past when their condition resembled their current condition. The cases may either be stored in the mobile telephone or PDA memory from a remote server to which the phone or PDA is periodically connected. The patient can then select an appropriate dosage with knowledge of the outcome that has been achieved in the past. Figure IA illustrates the process flow. The patient is expected to measure their blood glucose level as normal and input it to the device resulting in the device receiving the blood glucose measurement in Step 100. The device compares the blood glucose measurement to a set of predefined ranges to classify it to one of the plurality of classes. In the particular Example which will be illustrated below the classes are: amber 10-13 mmol/1 amber-red 13-16 mmol/1 red 16-20 mmol/1 very red more than 20 mmol/1

As mentioned above these ranges may be different for different patients. In Step 104 the device then retrieves from a database of stored cases those cases which have an initial blood glucose measurement in the same class as the received blood glucose measurement. Each of the stored cases consists of a record of the initial blood glucose measurement (regarded as taken at time t-1), the insulin dose taken by the patient (again at time t-1), and the subsequent blood glucose level obtained at the next normal measurement at time (regarded as time t). In Step 106 the device then displays the resultant blood glucose level (i.e. at time t) for the retrieved previous cases against the insulin dose taken by the patient at time t-1.

Examples of such displays are shown in Figures 2 and 3. Figure 2 illustrates cases corresponding to the four different classes mentioned above for a particular patient. Each case gives one data point, marked as a circle, plotting the resultant blood glucose level against insulin dose taken. Although Figure 2 shows all four classes, in practice only the class corresponding to the received measurement would be displayed to the patient. For example if the patient's current glucose level is 18 mmol/1, this places them in the "red zone" and so only the second plot of Figure 2B would be shown. The plots of Figure 2 also indicate clearly the boundaries of the normal or desirable blood glucose level (normoglycaemia) for the patient of 4-10 mmol/1. Referring to Figure 2B the patient can see that most often insulin dosages of 14 or 16 units were taken, and by comparing the density of data points in the normal range can see that a dose of 16 units most often led to the blood glucose level subsequently being in the normal range. This can guide the patient in choosing an appropriate dose for the current condition. The plot of individual data points in Figure 2 allows the patient to see the modal result of each treatment. However, a simplified display may be used as illustrated in Figure 3 in which the mean and standard deviation for each dosage in each class is displayed, with treatments only being included if more than a minimum number of data points are present. This display allows a patient to see, for example, that if their current glucose level is in the red zone (Figure 3B) then insulin dosages of 12-16 units on average resulted in the blood glucose level coming down at the next measurement to the normal region. On the other Tiand for the "very red" zone of Figure 3 C, dosages of 16-20 units resulted, on average, in Wood glucose levels coming down to the top of the normal zone.

It should be appreciated that the particular form of display is not limited to the specific form of Figure 2 or Figure 3. Specifically, if the aim is to reinforce correct behaviour by the patient (selection of appropriate insulin dose), only a sub-set of cases may be displayed.

In addition, the patterns of measurement of the patient's condition (typically blood glucose levels for diabetes or PEF/FEV1 for asthma) may be recorded for longer periods of time than simply the initial measurement at time t-1. Thus measurements at times t-2, t-3, ..., t-N may be included to define a pattern of measurements and hence a case. A transform may also be applied to each of the measurements, for example a non-linear transform to emphasize the importance of low blood glucose levels (hypoglycaemia) or a time-dependent transform to place greater emphasis on more recent measurements. The metric used to identify nearest neighbours in these patterns, or cases, may typically be Euclidean distance but other distance metrics are also suitable.

Although the embodiment described here has been concerned with the management of high blood glucose levels (hyperglycaemia) and the appropriate treatment (choice of appropriate insulin dose) to return to normoglycaemia, the invention can also be used to highlight patterns in successive blood glucose measurements which are likely to lead to hypoglycaemia at time t (cases of "high risk of hypoglycaemia") and alert the patient to this possibility. Recovery from this situation would again be aided by case-based reasoning using the databases of high risk of hypoglycaemia cases and their resultant outcomes (the blood glucose level at time t). In the case of hypoglycaemia, the treatment, based on the recognition of the possible risk of hypoglycaemia occurring at time t, may be the rapid intake of food or glucose tablets at time t-1.

Advantageously in this embodiment the database of stored cases is updated with use of the system so that the system gradually improves its "knowledge" of the patient's response to treatment. This is achieved, as illustrated in Figure IB, by receiving the blood glucose measurement taken later by the patient, (following their insulin injection), and creating a new case for storage in the database in which the later ^"blood glucose measurement is set as the ^"blood glucose at time t as in Step 200, the initial blood glucose measurement from Step 100 of Figure IA is taken as the ^"blood glucose at time t-1, and the insulin dose is input by the patient. These together as a new case are stored in the database at Step 202. As mentioned above in the case of the new user of the system it may be necessary to populate the database with cases of other users or of a typical population of users. Preferably as the patient uses the system the patient's own new cases are added, gradually replacing the initial set. This may be performed by an algorithm implemented on the phone or wireless PDA, or remotely on a server with the updated set of cases (possibly after validation by the patient's clinician) downloaded periodically to the phone or wireless PDA over the air.

Claims

1. A method of providing information relating to a chronic medical condition to a patient suffering from that condition, the method comprising: receiving a measurement of the patient's current medical condition; accessing a stored database of past cases on the basis of the measurement to retrieve similar cases, each of said stored cases comprising the value of a previous measurement of the patient's condition and the treatment followed; characterised in that: each of said stored cases further comprises the outcome of the treatment; and by the step of: displaying to the patient the treatments followed in each of the similar cases and the outcome of the treatments.

2. A method according to claim 1 wherein the past cases in the stored database are each classified to one of a plurality of classes based on the previous measurement of the patient's condition.

3. A method according to claim 1 or 2 wherein the stored cases each comprise a plurality of previous measurements.

4. A method according to claim 3 wherein the plurality of previous measurements are a time series of measurements.

5. A method according to claim 4 wherein the past cases in the stored database are each classified to one of a plurality of classes based on the plurality of previous measurements of the patient's condition with a non-uniform weighting being applied to the cases.

6. A method according to claim 5 wherein the weighting is non-linear or time- ^"based.

7. A method according to claim 2, 5 or 6 wherein each class relates to a range of values of the previous measurement of the patient' s condition.

8. A method according to claim 7 wherein each class relates to a particular pattern in the time series of previous measurements.

9. A method according to claim 2 or any one of 5 to 8 wherein the similar cases displayed are past cases from that class to which the received measurement or measurements of the patient's current condition corresponds.

10. A method according to any one of the preceding claims further comprising the step following treatment of adding the received measurement, treatment and outcome to the stored database as a case.

11. A method according to any one of the preceding claims wherein the stored database comprises cases only for the patient using the method.

12. A method according to any one of the preceding claims wherein the display to the patient comprises the display of the average outcome of each of different past treatments in the cases retrieved.

13. A method according to any one of the preceding claims wherein the display to the patient comprises the display of the modal outcome of each of different past treatments in the cases retrieved.

14. A method according to any one of claims 1 to 11 wherein the display to the patient comprises the display of only past cases with positive treatment outcomes.

15. A method according to any one of the preceding claims wherein the outcome is the value of a further measurement of patient's condition following the treatment.

16. A method according to any one of the preceding claims wherein the measurement is received by a mobile data processing terminal, access to stored database is via the mobile data processing terminal, and said display is on the mobile data processing terminal.

17. A method according to claim 16 wherein the mobile data processing terminal stores the database.

18. A method according to claim 16 or 17 wherein the mobile data processing terminal is one of: a mobile telephone, a personal digital assistant, a hybrid thereof.

19. A method according to any one of the preceding claims wherein the cases relate to insulin-dependent diabetes and the previous measurement of the patient's condition is an initial blood glucose level measurement, the treatment followed is the insulin dose taken, and the outcome is the resultant blood glucose level measured at a later time.

20. A method according to claim 19 wherein the past cases are classed by said initial blood glucose measurement

21. A method according to claim 19 or 20 wherein the display is a 2-dimensional plot of resultant blood glucose level against insulin dose taken, for a plurality of cases in the class corresponding to the received measurement of blood glucose level.

22. A method according to claim 19 or 20 wherein the display is a 2-dimensional plot of resultant blood glucose level against insulin dose taken, for a plurality of cases in the class corresponding to a pattern of successive measurements of blood glucose level.

23. A method according to claim 22 wherein the class of stored cases is identified ^"by means of a non-uniform weighting of the successive measurements.

24. A method according to claim 22 wherein the class of stored cases is identified lay means of a metric to identify nearest-neighbour patterns.

25. A method according to any one of claims 1 to 18 wherein the cases relate to asthma and the previous measurement of the patient's condition is a Peak Expired Flow or Forced Expired Volume measurement, the treatment followed is the number of puffs of inhaler.

26. A computer program comprising program code means for executing the method of any one of the preceding claims.

27. A mobile data processing terminal programmed with a computer program according to claim 26.