GB2515118A

GB2515118A - Method and System for Healthcare Pathway Modelling

Info

Publication number: GB2515118A
Application number: GB1310686.9A
Authority: GB
Inventors: Dariusz Ceglarek; Prakash Prakash
Original assignee: University of Warwick
Current assignee: University of Warwick
Priority date: 2013-06-14
Filing date: 2013-06-14
Publication date: 2014-12-17
Also published as: GB201310686D0

Abstract

A method and system for modelling a healthcare pathway are provided. The method comprises the steps of receiving at least one set of clinical information 210, generating a patient diversion map 242 containing a plurality of pathways represented by the received clinical information 210, determining a set of factors controlling the pathway to be followed by a patient, generating key performance indicator information associated with each pathway controlled by a respective set of factors, integrating the key performance indicator information and the patient diversion map, and generating at least one suggestion 290 for an optimised set of factors to control a healthcare pathway, each suggestion based on the key performance indicator information integrated with the patient diversion map.

Description

Method and System for Healtheare Pathway Modelling The present invention relates to a computer program product for modelling a health care pathway, and corresponding method and system. More particularly, the present invention relates to a system and method wherein clinical service delivery is managed and optimised.

Even nowadays patients with chronic diseases still frequently experience delays in moving to a dedicated treatment unit from the emergency unit, due to a variety of io factors associated with the hospital environment. While there have been major advances in treatment for these diseases in recent years, a great deal of delay still exists in translating these advances into measurable progress in health care development, due to a lack of integration of services between the key entities that are providing the services. Currently there is a great need to improve the efficiency and performance of i the service system in hospitals, which includes medical staff, ward capacities, and complex medical devices. For example, acute stroke care is often dependent on the time-critical steps that are carried out in the emergency department which involves rapid and accurate diagnosis of the patient condition, and moving the patient into a stroke unit as early in the process as possible, since stroke patients experience better outcomes when they receive majority of their multidisciplinary care in a stroke unit.

The quality ofhealthcare services is generally assessed against government guidelines, for example the Department of Health (UK) defined standards, and US Public Health Service guidehnes. These standards are largàly defined specifically as quantitative targets also known as Key Performance Indicators (KPIs). KPIs are used in various countries such as UK to measure the quality and outcomes of healthcare services. For instance, one of the stroke care KPIs in UK is that 8o% of stroke patients must spend more than 90% of their hospital stay in stroke units -the 80/90 KPI.

The present invention integrates system engineering methods with healthcare modelling analysis to provide hospitals with a method to model their current existing system so as to rapidly and effectively address the shortcomings and constraints of the system and their root causes. Specifically, the present invention provides a formal methodology to develop a detailed quantitative system-level representation of care processes carried out in the clinical environment, by performing efficient data mining and process mining to a large amount of received clinical information, in order to help the hospital meet its performance targets.

According to an aspect to the present invention, there is provided a method for modelling a health care pathway, the method including: receiving at least one set of clinical information, generating a patient diversion map containing a plurality of pathways represented by the received clinical information, determining a set of factors controlling the pathway to be followed by a patient, generating key performance indicator information associated with each pathway controlled by a respective set of io factors, integrating the key performance indicator information and the patient diversion map, and generating at east one suggestion for an optimised set of factors to control a healthcare pathway, each suggestion based on the key performance indicator information integrated with the patient diversion map.

According to another aspect of the present invention, there is provided a computer program product for modefling a health care pathway, comprising: a first instruction to receive cUnical information, a second instruction to generate a patient diversion map containing a plurality of pathways represented by the received clinical information, a third instruction to determine a set of factors controlling the pathway to be followed by a patient, a fourth instruction to generate key performance indicator information associated with each pathway controlled by a respective set of factors, a fifth instruction to integrate the key performance indicator information and the patient diversion map, and a sixth instruction to generate at least one suggestion for an optimised set of factors to control a healthcare pathway, each suggestion based on key performance indicator information integrated with the patient diversion map.

According to another aspect of the present invention, there is provided a system for modelling a health care pathway, comprising: a data receiving unit to receive at least one set of clinical information, a patient diversion pathway generating unit to generate a patient diversion map containing a plurality of pathways represented by the received cUnical information, a data evaluation unit to determine a set of factors controlling the pathway to be followed by a patient and to generate key performance indicator information associated with each pathway controlled by a respective set of factors, an information integration unit to integrate the key performance indicator information and the patient diversion map, and a suggestion unit to generate at least one suggestion for an optimised set of factors to control a healthcare pathway, each suggestion based on the key performance indicator information integrated with the patient diversion map.

Optiona' features are set out in the dependent claims.

Embodiments of the present invention will now be described, by way of example, with reference to the accompanying drawings, in which: Figure 1 is a flow chart representing the steps of obtaining optimised models for a io healthcare pathway according to an embodiment ofthe invention; Figure 2 is a b'ock diagram of a heallhcare pathway modefling system according to an embodiment of the present invention; Figure 3 illustrates the flow chart representing the operation steps of the method according to an embodiment of the present invention; Figure 4 is an example of a patient diversion map containing a phiraBty of heakhcare pathways; Figure 5 is a flow chart representing the operation steps of sim&ating patient flow in a healthcare pathway; Figure 6 is a screenshot of a computer program product according to an embodiment of the present invention; and Figure 7 is a plurality of examples of confusion matrices as graphical representations for different clinical tests.

Figure 8 shows a plurality of graphs illustrating the effects of various types of factors on the performance of a clinical environment in the aspect of patient diversion.

Detailed Description of Embodiments

Certain embodiments will now be described in greater detail with reference to the accompanying drawings. In the following description, like drawing reference numerals are used for like elements, even iii differcift drawings. The matters defined in the description, such as detailed construction and dements, are provided to assist in a comprehensive understanding of the exemplary embodiments. However, the embodiments can be practiced without those specifically defined matters. Also, weB known functions or constructions are not described in detail since they would obscure the embodiments with unnecessary detail. Moreover, expressions such as "at least one of', when preceding a list of elements, modify the entire list of elements and do not modify the individual elements of the list. Unless features of "one embodiment" are inconsistent with features of other embodiments, the term "one embodiment" shall be construed to be a disclosure of the associated features in conjunction with all other features that are consistent therewith.

Figure 1 illustrates a flow chart representing the steps of identifying processes in a healthcare model and updating the model, according to an embodiment of the invention. The procedure as illustrated in Figure 1 starts with a dynamic model and/or a static model of the operation of a healthcare facility, such as a hospital.

Jo Constraints, interactions, sources of error, conforming and non-conforming causes (collectively referred to as "parameters") are identified (5102) from the model(s). The identification may comprise carrying out data mining and process mining processes on the information contained in the dynamic and/or static model(s). The step of identification results in an estimated healthcare model (5103), which is based on the is initially provided model, but effectively reconfigured in a way which is indicative of the effect of the identified constraints, interactions, sources of error and conforming / non-conforming causes.

In the next step of the procedure (5104), a selection is made by a user as to the basis of the optimisation of the model. The selection is based on one of three bases -optimisation for conformance with a set of guidelines, optimisation to reduce diversion from an ideal healthcare pathway, and optimisation to reduce process bottlenecks.

In the process conformance branch (5105-5108), a non-conforming process in the estimated healthcare model is identified (Sio) from the information identified in step S1o2. The parameters associated with the non-conforming process are then identified in the subsequent step (Sio6) from the information obtained in step 5102. Variations are then introduced into the healthcare pathway model (5107) to reflect a set of changes to some or all of the identified parameters. Simulations (Sio8) are then carried out in order to simulate the effect of each of the variations in the model to determine whether the variations ead to an improved or conforming model. The nature of the simulations will be described in more detail below.

In the process diversion branch (S1o9-S112), process diverting nodes in the estimated healthcare model are first identified (8109). A node in the model can be defined as an event, such as point of making a clinical decision, or the existence of a patient in a particular location. A process diverting node is a node from which the model shows the possibility of deviating from an optimum pathway. For example, two process pathways may be derived from a single node. If on'y one of those pathways is optimal, the fact that the node allows for the possibility of movement along the non-optimal path means that the node represents a candidate event leading to process diversion. Process diverting nodes can be identified by firstly identifying all those nodes associated with more than one output pathway, and secondly, identifying whether any of those pathways is sub-optimal, using logic similar to that used in step Sins, based on the information identified in step S1o2. I0

The parameters associated with a diverting node are then identified in the subsequent step (Siio), through analogy with step Sio6. Diverting nodes in the model are then modelled (Sni) in relation to the identified types of interactions, and simulations (5112) are then carried out, as with steps 5107 and SioS.

In the process bottleneck branch (S113-S116), process bottleneck nodes in the estimated healthcare mod& are first identified (S113). Abotrieneck node is one which is associated with a delay in a healthcare process due to a processing speed being limited by a parameter, such as a number of hospital beds, at the node. The types of interactions causing the process bottlenecks are then identified in the subsequent step (S113). Bottleneck nodes in the model are then modelled in relation to the identified types of interactions, and simulations (Sn6) are then carried out.

Following any of the three branches as described above, it is then determined if the estimated model is to be updated in step Sn7, according to the results of the simulations. For example, in the bottleneck branch, the identified parameters which control the bottleneck node in a patient diversion map would be simulated, and a solutions portfolio containing a number of sets of optimised parameters, each set containing a different combination of suggested modifications to the identified parameters would be generated. The estimated model can then be updated automatically according to a prioritised characteristic associated with one of the suggested modifications, or altemativ&ymanuafly by the hospital management.

Examples for performing such simulation and how the results of simulation are used to determine whether or not to update the estimated model will be described in more detail below.

Conventional methods for improving healthcare pathway models usually only involve the elimination of bottlenecks in healthcare pathways, and therefore there is a need to develop and integrate additional techniques to (i) straightforwardly identi1r non-conforming syntax of an existing healthcare model when compared with that suggested by government clinical guidelines; and (2) identifr diverting nodes in a healthcare pathway and their root causes. The present invention therefore provides an improved methodology of modelling healthcare pathways, by incorporating all three branches shown in Figure 1 to provide an accurate analysis technique for optimising healthcare io services provision.

Figure 2 is a block diagram of a healthcare pathway modelling system according to an embodiment of the present invention, which can be used to carry out the method shown in Figure 1. In Figure 2, a healthcare pathway modelling module 200 comprises a data receiving unit 220, a data evaluation unit 230, a patient diversion pathway generating unit 240, an information integration unit 250, a suggestion unit 260, a contrcil unit 270, and a storage unit 280.

The data receiving unit 220 receives clinical information 210. Specifically, the data receiving unit 220 may receive clinical information 210 from a plurality of external sources including but not limited to healthcare-related literature 212, government guidelines 214, hospital workflow and medical device data 216, and patient data 218.

Clinical information 210 received may be heterogeneous data related to hospital staff such as interview information, training records, medical qualifications, dynamic workfiow models, multiple static Role Activity Diagrams (RADs), and historical data such as statistics and trends related to patient admission date and time and decision processes made by hospital staff.

Heakhcare rethted literature 212 may include medical journals, books, and pubhcations related to hospital management, treatment of diseases, and medical training.

Government guidelines 214 may inchide those issued by a local government for treatments of diseases; and/or literature on clinical decision-making.

Hospital data and medical device data 216 may include sample onset-to-admission time, sample time of arrival, A&E intervention phases time windows, patient category distribution, ward capacities, electronic tracking and planning data from healthcare software (e.g. iPM, Dendrite), equipment productivity assessment of medica' devices (e.g. CT (computerised tomography) scanner, MRI (magnetic resonance imaging) scanner) including but not limited to downtime data, event logs, service cost, and technical information such as types of scan data, frequency, speed, radiation, dosage.

Patient data 218 may include information such as medical history, personal Jo information, prescription, admission date and time, ength of stay, ward transfers, etc. The clinical information 210 received by the data receiving unit 220 may be stored in the storage unit 280.

The data receiving unit 220 provides received data to the data evaluation unit 230 for evaluation. Specifically, the data evaluation unit 230 acquires information related to healthcare modefling from clinical information 210 that may be used to generate patient diversion pathways.

The data evaluation unit 230 performs step 8102 of Figure 1 and identifies and determines factors affecting a healthcare pathway in a hospital, such as pharmacological interventions, flow of clinicians, job duties of hospital staff, and performance of medical devices, from the clinical information 210. The data evaluation unit 230 may also calculate Key Performance Indicator (KPI) related information using data included in clinical information 210.

The data evaluation unit 230 may develop a healthcare pathway model, such as a Role Activity Diagram (RAD), using at least one set of clinical information 210. The data evaluation unit 230 may also generate a Discrete Event Simulation (DES) model based on the one or more Role Activity Diagrams. The Discrete Event Simulation (DES) models may contain a chron&ogical sequence of events representing the operation of the Wile Activity Diagrams.

The information generated by the data evaluation unit 230 may be stored in the storage unit 280.

The patient diversion pathway generating unit 240 generates at least one patient pathway diversion map 242, based on inputs from the data evaluation unit, and optionafly, information from the data receiving unit which has not been evahiated. The patient diversion pathway generating unit 240 may generate a plurahty of pathways in the patient diversion map 242 by simulating system variations in the Role Activity Diagrams (RADs) using different clinical scenarios.

The patient diversion pathway generating unit 240 may acquire time related information for the generation of the patient diversion map 242 using radio frequency io identification (RFID) tracking data.

The patient diversion map 242 may be stored in the storage unit 280.

The information integration unit 250 integrates information from both the patient diversion map 242, which has been generated by patient diversion pathway generating unit 240, and the data evaluation unit 230.

The information integration unit 250 may perform sensitivity analysis to identify throughput issues in the patient diversion map 242 by simulating changing different factors identified by the data evaluation unit 230 and evaluating the effects caused by such changes, in other words, the sensitivity to those changes. Specifically, the effect on KPI information of such changes in factors may be measured quantitatively, by integrating the information from both the patient diversion map 242 and the KPI information generated by data evaluation unit 230. These simulation results may be recorded and stored in the storage unit 280.

The suggestion unit 260 provides at least one suggestion 290 for an optimised set of factors to control a healthcare pathway based on the integrated data. One example of a set-based solution suggestion to be implemented in a hospit& may bc a combination of factors which include changes in decision making, specifically redesigning clinical decision making to improve diagnostic accuracy of referrals in the emergency department, training ambulance crews to alert the hospital to all incoming patients with certain symptoms, changes in capacity such as increasing the number of beds in certain wards in order to manage increase patient flow and eliminate bottlenecks, balancing bed capacity to manage patient flow variations.

Specifically, different sets of optimised factors may comprise different suggestions of modifying the same identified factors. For example, a first suggestion maybe rapidly identifying patients with FND symptoms in the emergency department and adding two more beds to the stroke ward, and a second suggestion may be pre-alerting the stroke team about patients with FND symptoms in the emergency department and adding four more beds to the stroke ward. These different combinations of changes to identified factors may produce different effects on the diversion of patients as well as other key performance indicators (KPIs).

io Other examples of suggested improvements may include modifications to technical hardware and/or system of medica' devices, in order to improve efficiency, accuracy, hardware lifetime of the medical devices. These may further include reconfiguration of hardware components and/or modifying the functionalities of the embedded software.

Yet other examp'es of suggested system kvel improvements may include changing the sequence of cBnical decision making processes; promoting improved recognition of certain disease in the emergency department thereby reducing the number of patients diverted from the specialised care pathway for that disease; redesigning the hospital's specialised care unit; relocating the entire specialised unit or merging the unit with another ward to allow clinicians the flexibility to effectively handle variations in patient flow.

The suggestion unit 260 may prioritise suggestions based on at least one of the potential impact, the cost for changes, and the ease of implementations of the one or more optimised set of factors.

Figure 3 is a flow chart illustrating a healthcare pathway modelling method according to an embodiment of the present invention. In Figure 3, the healthcare pathway modelling systcm 200 receives clinical information (S301).

Subsequently, the clinical information 210 received is either used to generate a patient diversion map 242 (S3o2) or used to determine factors controlling the pathway to be foflowed by a patient.

In the step of generating patient diversion map (S3o2), the patient diversion map is generated using forward analysis methods, based on clinical information received. A -10-plurality of pathways may be generated for a patient diversion map 242 by performing data-flow analysis on system variations, using Role Activity Diagrams (RAI)s) obtained from cUnical information 210. Pathways taken by patients maybe dependent on medical decisions made by certain medical staff members, and this information may be obtained from Role Activity Diagrams (RAPs).

In step S3o2, patient diversion in the pathway may be identified based on the decision making process represented in dynamic workflow models provided by the hospital, and/or decision making related static data from electronic medical records. Time io related information acquired using radio frequency identification (RFID) tracking data from cUnical information 210 may be used for the generation of patient diversion map.

The healthcare pathway modelling system 200 determines factors controlling pathways (S3o3) in the patient diversion map 242. In this operation step, the data evaluation unit 230 may identii' a set of factors controlling the pathway to be followed by a patient. A number of different factors and parameters may be assessed based on the received clinical information 210. For example, the cUnical information 210 may contain medical guidance from a document which specifies certain clinical tests which affect a treatment pathway for a particular illness. In addition, historical treatment data may contain various data patterns or trends, which indicate dependence on a certain factor on a treatment pathway, such as the time of year.

Decision making processes causing diversion in the healthcare pathway may be dependent on decision makers such as medical staff, decision making steps based on clinical tests, interdepartmental KPIs, and the like. Decision-making regarding patient conditions within the healthcare pathway may depend on the results of multiple medical tests performed by various medical specialists along the pathway. These tests may be critical to the final decision made by the decision-maker on patient conditions.

These may include blood tests, computerised tomography (CT) scan, electro-cardiogram (ECG) scan, a Face Arm Speech Time (FAST) test, and any other medical tests.

Next, the healthcare pathway modeffing system 200 generates Key Performance Indicator (KPI) information associated with each pathway (s3o4) in the patient diversion map 242.

-11 -The generating may further comprise calculating key performance indicator information by measuring the impact of one or more sub-groups of patients on the diversion of other patients in a healthcare pathway. One or more sub-groups of patients may have an effect on the diversions in the pathway, due to co-morbidities or co-occurring illnesses, severity of patient conditions, and mimics. Different subgroups of patients may require different times in receiving particular tests, or present unclear symptoms which may delay the decision making process, thus affecting pathway KPI.

The generating may further comprise taking into account a plurality of factors that have io been determined in the determination step for caictilating key performance indicator information associated with each pathway in the patient diversion map.

The generating may further comprise recording and computing the impact of clinical decisions, and using the information to calculate key performance indicator information.

The generating may further comprise measuring at east one of the sensitivity and specificity of the clinical tests identified in the determination step s303, using data related to true positives, true negatives, false positives and false negatives.

Other examples of Key Performance Indicator (KPI) information may include confusion matrices, as described in more detail below, as representations of the accuracies and reliabilities of clinical tests. The clinical tests may include FAST, FND (Functional Neurological Disorder), NIHSS (National Institutes of Health Stroke Scale).

Other examples of Key Performance Indicator (KPI) information may also include patient length of stay in specialised care unit, and/or patient throughput from the pathway, interdepartmental KPI and other hospital operations parameters. For example, stroke care pathway in general can involve emergency department, hypcr acute stroke ward, and rehabilitation ward.

The generated KPI information is then integrated with the patient diversion map (S3o5).

This integrating step may comprise comparing at least two or more Role Activity Diagrams (RADs) to identify methods of identifying process issues. The process issues -12 -may include at least one of throughput and patient diversion, sen'ice monitoring, simulation modelling, and future process change analysis. For example, the RAI of cUnical guidelines and RAD of the actual hospital floor maybe compared to investigate the non-conforming factors in the healthcare pathways, with the help of r&ated tracking data and IT systems data.

The integrating step may also comprise using sensitivity analysis on the KPI information and the patient diversion map, to determine the decision-making processes that may be performed in a different sequence in order to maximise the total io time spent by a patient in a dedicated treatment unit.

Thereafter, the healthcare pathway modelling system 200 generates at least one suggestion for an optimised set of factors to control a healthcare pathway (8306).

The generating of suggestions may further comprise carrying out system impact assessment, wherein impact analysis is performed based on Discrete Event Simulation (DES) models. In this case, the effect of changing system parameters, such as the number of staff and equipments, work shifts, patient arrival rates, may be studied in detail, and optimum values may be suggested.

Figure 4 is an example of a patient diversion map constructed based on received clinical information. The patient diversion map in Figure 4 is comprised of a plurality of nodes representing ward levels, a plurality of arrows indicating the direction of patient flow, and a plurality of numbers corresponding to the arrows indicating the number of patients of the respective patient flows. Each healthcare pathway starts at either at the node labelled "AE", which stands for "Accident and Emergency unit", or "0TH" which stands for "Other units". In this particular example, 298 patients are admitted into the hospital, wherein 273 originate from AE and 25 originate from other units.

At the starting point of the patient diversion map, patients in either ward "AE" or "0TH" are further diverted into a plurafity of different wards/unit in a subsequent leveL In this examp'e, 131 patients from AE and 10 patients from 0TH are directed to W43, which stands for ward 43. The other wards in the patient diversion map are Clinical Decision Unit (CDU), Wards 4°, 41, 42, Clinical Research Department (CRD), various Medical Wards (MED), Surgery (SRG), satellite site (RGB). Each pathway terminates in the final treatment ward of a patient.

-13 -Figure 5 illustrates a flow chart representing the operation steps of simulating patient flow in a dedicated hospital unit. The simulation is designed to measure and record the effects of changing a pathway-controthng factor on the overafi performance of a healthcare pathway model.

The starting point of the simulation procedure is the generation of a decision table (DT) (5501) from statistics obtained from data mining of treatment of known conditions. The decision table (DT) may contain a number of such known conditions and a number of actions to be performed for each condition. In this particular example, information related to patient data, onset-to-admission time, time of arrival, and time window of admission time are contained in the decision table.

Next, a set of patients is simulated by sampling the decision table (5502), based on distribution information of patients derived from clinical information 210. For each simulated patient, the onset-to-admission time, time of arriva', and time window of admission time is sampled from the decision table (S5o3, S5o4, S5o5).

Subsequently, in step So6, it is determined for each patient whether or not the time window is less than, a threshold for example, 4 hours, according to the model such as that derived in step S1o7 of Figure 1, which is the basis for the subsequent steps in this flow chart. In the present example, the factors associated with a process to be analysed are the number of patients in a stroke ward and a clinical decision unit (CDU), as determined in steps Sio and S1o6 of Figure 1. The operation proceeds to S5o7 if the decision is "no", where the patient count in the CDU is updated, and So8 if "yes", where it is determined whether the time window is within "regular" hours of a hospital, such as 8am to 8pm.

If it is dctcrmincd that thc timc window is within regular hours, the operation proceeds to another decision process where it is determined whether the patient is to be sent to a stroke ward, or to a CDU. A predetermined proportion of patients are sent to the stroke ward (S91o). If it is determined that the time window is not within regular hours, the operation proceeds to yet another decision process where it is determined whether the patient is sent to stroke ward, again based on a predetermined proportion.

If it is determined that the patient is sent to stroke ward in either of the decision process 5509 and 5910, the patient count in stroke ward (SW) is then updated (8911, S913). Otherwise, the patient count in cbnical decision unit (CDU) is updated accordingly (8912).

In the subsequent step S514, it is then decided whether or not all the patients in the generated set of patients have undergone the simulation procedure. If the decision is "yes", the total number of patients sent to clinical decision unit (CDU) and stroke ward (SW) is computed (5515), otherwise, the procedure returns to steps 5503, 5504 and io 5505 for the next patient to be simuated.

The output of the simulation procedure can then be provided to a decision making unit, for example, the data evaluation unit 230 or the information integration unit 250, for determining whether to update the model, based on whether the division of patients i5 into the clinical decision unit (CDU) and stroke ward (SW) is optimaL Figure 6 is a screenshot of a computer program for cakulating and monitoring key performance indicator information and a patient diversion map. In this exemplary computer program, there is provided a user interface showing an analysis summary specifically for a stroke care services pathway for a selected duration.

The computer program product calculates a number of parameters using data set(s) provided by the received clinical information 210. As shown in Figure 6, these parameters may include the average number of cases where tPA (Tissue Plasminogen Activator) is given to the patient, the average number of cases where a CT scan is performed within 4 hours of admission, the average percentage of patients seen by a physiotherapist within 24 hours, the average length of stay of patients, and any other information that may be useful for subsequent analysis with the patient diversion map 242 or related to the calculation of pathway KPI information.

The computer program also provides the total number of stroke patients for the duration of time being analysed, and the percentage of patients who spent a predetermined percentage time during their length of stay in a dedicated treatment ward, in this case, the percentage of patients who spent more than 90% of their length of stay in the stroke unit is calculated (80/90 KPI).

-15 -The computer program also provides a graphical representation of the patient diversion map, in a similar layout as presented in Figure 4. The patient diversion map in this example is comprised of a pthrality of nodes representing ward units, a plurality of Bnes between nodes indicating patient flow (from eft to right), and a plurality of numbers corresponding to the lines indicating the number of patients of the respective patient flows.

The user can select a particular node in the model and obtain data for that node, which is represented on the user interface. I0

Figure 7 shows a plurafity of confusion matrices constructed as a representation of the outcomes of a number of clinical tests, including FAST (Face Arm Speech Time), FND (Functional Neurological Disorder), and NIHSS (National Institutes of Health Stroke Scale).

Each confusion matrix contains a specific table thyout that allows representation of the performance of each cUnical test it represents. Each column of the matrix represents the instances of a diagnosis (e.g. stroke or non-stroke) resulting from the clinical test, while each row represents the instances of positives and negatives resulting from the clinical test.

For example, the True Positive (TP), False Positive (FP), True Negative (TN), and False Negative (FN) values for stroke diagnosis respectively represent the number of stroke patients correctly diagnosed with stroke, non-stroke patients incorrectly identified with stroke, non-stroke patients correctly identified as non-stroke, and stroke patients incorrectly identified as non-stroke.

Taking the confusion matrix for FAST as an example, the number of true positives (TP) is 40-44, the number for false positives (FP) is 5-8, the number for f&sc negatives (FN) o is 5-9, and the number of true negatives (IN) is 34-37. In this particifiar example, these are data obtained from the chnical information 210 direcfly supplied by the hospitaL The Positive Predictive Value (PPV), Negative Predictive Value (NPV), Sensitivity, and Specificity of a clinical test may then be calculated using the values of TP, FP, FN and TN. PPV is the proportion of positive test results that are True Positives (TP), and NPV is the proportion of negative results that are True Negatives (TN). Sensitivity relates to the ability of the test to identify positive results, and is the probability of a positive test given that the patient is Hi. Specificity relates to the ability of the test to identify negative resuks, and is the probability of a negative test given that the patient is weTh The formulae below show how values for PPV, NPV, sensitivity, and specificity maybe calculated. These values may be used for generating KPI information in subsequent operation steps in the method of the present invention.

= numberofirue Positives number of True Positives + number of False Positives NPV = numherof'I'rue Negatives number of True Negatives + numberof False Negatives * number of True Positives Sensitivity = nuinberof True Positives + nwnberof False Negatives * * nwnberof True Negatives Specijicitv= -numberof 1 rue Negatives + number of /a/se Positives Figure 8 includes three graphs respectively demonstrating the effects of changes in three different factors (i.e. inter-departmenta' KPI, decision maker and/or decision making procedures, and capacity) respectively on the performance of a clinica' environment in the aspect of patient diversion.

Figure 8A is a graph of number of patients on an optima] heakhcare pathway and number of patients dive,rted away from the optim& healiheare pathway, against the change in inter-departmental KPI.

Mong the x-axis, the value of inter-depaitment& KPI increases from "current", representing a current value of the inter-departmental KPI, to "ai", "a2" ... and so forth towards "aN", which respectively represent a number (i.e. N) of increasing inter-departmental KPI values that can be incorporated as a factor of an optimised set of factors of suggested improvement. The y-axis represents the number of patients. -17-

As illustrated in Figure 8A, the number of patients on an optimal healthcare pathway increases as a particular inter-departmental KPI increases from "current" to "al", "af, and so forth. Convers&y, the number of patients diverted away from the optimal healthcare pathway decreases as the particular inter-departmental KPI increases from "current" to "al", "a2", and so forth.

Figure 8B is a step chart of number of patients on an optimal healthcare pathway and number of patients diverted away from the optimal healthcare pathway, against the io change in decision makers and/or decision making procedures.

Along the x-axis, a plurality of combinations of decision-makers and their respectively assigned decision-making procedures are respectively represented as "current" (i.e. current setting in the clinical environment), "bi", "b2", and so forth to "bN". The y-axis represents the number of patients.

As illustrated in Figure 8B, the number of patients on an optimal healthcare pathway increases in discrete steps as the decision-maker and decision-making changes from "current" to "hi", "b2", and so forth. Conversely, the number of patients diverted away from the optimal healthcare pathway decreases in discrete steps as the decision-maker and decision-making changes from "current" to "bi", "b2", and so forth.

Changes in decision-making may include redesigning clinical decision-making to improve diagnostic accuracy of referrals in different ward levels (e.g. emergency department), training ambulance crews to alert the hospital to all incoming patients with certain symptoms, balancing bed capacity to mange patient flow variations.

Specifically, for example, "bi" may be a change in decision-making to rapidly identify patients with FND symptoms in the emergency department, and "b2" may be a change in decision-making to pre-alert the stroke team about patients with FND symptoms in the emergency department.

Changes "bi" to "bN" may also include the reassignment of a particular cbnical decision-making procedure to a different type of clinical staff in the hospital, and changing the sequence of steps in a particular decision-making process. -18-

Figure 8C is a graph of the number of patients that can be accommodated in a ward level against the capacity of that ward level.

A]ong the x-axis, the value of capacity in a ward eve1 increases from "current", representing a current value of the capacity in a ward level, to "ci", "cs" ... and so forth towards "c9", which respectively represents increasing capacity values that can be incorporated as a factor of an optimised set of factors of suggested improvement. The y-axis represents the number of patients that can be accommodated in the ward level.

io Capacity may include number of beds in certain wards and number of certain types of staffs in certain wards, or any other factor which affects the quantity of patient flow in the ward level or the clinical environment.

The graph may also be representative of the effect of the number of patients that can be accommodated in a cUnical environment in genera], against the capacity of the c]inica] environment.

Accordingly, a set-based solution suggestion, which is an optimised set of factors to control a healthcare pathway, may be provided by integrating changes of different factors (e.g. "al, bi, ci, c2, c3", "a2, b, c4, c5, c6"), or any other combinations of factors, based on a desired criterion, which may include potential impacts on the clinical environment, the cost for changes, and the ease of implementation.

The present invention, and particularly the healthcare pathway modelling module 200, can also be implemented as computer-readable program on a computer-readable recording medium. Codes and code segments constituting the computer-readable program can be easily construed by programmers skilled in the art to which the present invention pertains. Also, the computer-readable program is stored in a computer-rcadablc rccording mcdium, and rcad and cxccutcd by a computer systcm to accomp]ish the present invention. Examp]es of the computer-readab]e recording medium inc]ude magnetic tapes, optica] data storage device, and carrier waves.

A]ternativey, the invention can be imp]emented in hardware, or a combination of hardware, such as an integrated circuit, and software.

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it will be understood by those of ordinary skill in the art that various changes in fbrm and details may be made therein without departing from the present invention as defined by the appended claims.

Claims

Claims 1. A method of modeffing a heakhcare pathway, comprising the steps of: receiving at least one set of clinical information; generating a patient diversion map containing a plurality of pathways represented by the received clinical information; determining a set of factors controlling the pathway to be followed by a patient; generating key performance indicator information associated with each pathway controlled by a respective set of factors; io integrating the key perfoniiance indicator information and the patient diversion map; and generating at least one suggestion for an optimised set of factors to control a healthcare pathway, each suggestion based on the key performance indicator information integrated with the patient diversion map.
2. A method according to claim 1, wherein determining a set of factors controlUng the pathway to be followed by a patients comprises identifying at east one of constraints, interactions, conforming and non-conforming causes of a healthcare pathway fmm the received clinical information.
3. A method according to claim 2, wherein the generating further comprises generating a plurality of nodes in the patient diversion map, each node being a point of clinical decision-making and is representative of at least one of a clinical ward and a clinical processing stage.
4. A method according to claim 2, wherein the identifying comprises performing at least one of data mining and process mining processes on the received clinical information.
5. A method according to claim 2, wherein the identifying comprises identifying at least one process diverting node among the phirality of nodes in the healthcare pathway, the process diverting node being a point of clinical decision making which leads to the possibility of deviating from an optimum pathway.
6. A method according to claim 2, wherein the generating further comprises determining a direction and number of patients between a first node and a second node in the patient diversion map.
7. A method according to claim 2, wherein the identi'ing comprises identiing at least one bottleneck node in the healthcare pathway, the bottleneck node being a point in the healthcare pathway associated with a delay.
8. A method according to claim 1, wherein integrating comprises introducing io variations into the healthcare pathway to reflect a set of changes to at east one of the identified factors.g. A method according to claim 8, wherein integrating further comprises performing simulation to simulate the effect of each of the variations in the healthcare is pathway.10. A method according to claim 1, wherein the at least one suggestion for an optimised set of factors contains at least one of the parameters related to changes in decision making, changes in sequence of clinical decision making processes, changes in training of hospital staff, changes in technical hardware of medical devices, and changes in ward capacities.ii. A method of modelling a healthcare pathway according to claim 1, wherein the at least one suggestion for an optimised set of factors is such that the healthcare pathway conforms to a set of predetermined guidelines.12. A computer program which, when executed by a processor, is arranged to perform the method of any one of claims 1 to 11..13. A system for modeffing a heakhcare pathway, comprising: a data receiving unit to receive at least one set of clinica' information; a patient diversion pathway generating unit to generate a patient diversion map containing a phirality of pathways represented by the received clinical information; a data evaluation unit to determine a set of factors controlling the pathway to be followed by a patient and to generate key performance indicator information associated with each pathway controlled by a respective set of factors; -22-an information integration unit to integrate the key performance indicator information and the patient diversion map; and a suggestion unit to generate at least one suggestion for an optimised set of factors to control a healthcare pathway, each suggestion based on the key performance indicator information integrated with the patient diversion map.14. A system according to claim 13, further comprising a storage unit for storing data from at least one of the other units in the system.Jo 15. A system according to claim 13, fuither comprising a control unit for controlling the operations of at east one of the other units of in the system.i6. A system according to claim 13, wherein the data receiving unit acquires time related information using radio frequency identification tracking data.17. A system according to claim 13, wherein the information integration unit performs sensitivity analysis on the patient diversion map by simiflating the effects of changing different factors determined by the data evaluation unit.iS. A system according to claim 13, wherein the patient pathway generating unit performs forward analysis on received clinical information to generate a patient diversion map.19. A system according to claim 13, wherein the suggestion unit prioritises the suggestions based on one of the potential impact, costs of changes, and the ease of implementations of the one or more optimised set of factors.