WO2016029272A1

WO2016029272A1 - Analysing medical data

Info

Publication number: WO2016029272A1
Application number: PCT/AU2015/050504
Authority: WO
Inventors: Paul NICOLARAKIS
Original assignee: Capital Markets Crc Ltd
Priority date: 2014-08-29
Filing date: 2015-08-28
Publication date: 2016-03-03
Also published as: AU2015309696A1

Abstract

This disclosure relates to analysing medical services data related to services provided by medical service providers. A processor first determines a network of the medical service providers based on the medical services data, the network comprising a measure of relationship between any two of the medical service providers. The processor then determines a network measure based on the measure of relationship between any two of the medical service providers and generates a report indicating the network measure. Since the network measure in the report is based on the measure of relationship, the report is more meaningful than reports using other measures that are not based on networks or relationships. The report provides an insight into the underlying structural characteristics of the medical data.

Description

"Analysing medical data" Cross-Reference to Related Applications

[0001] The present application claims priority from the Australian provisional application 2014903444 filed on 29 August 2014 with Capital Markets CRC Ltd being the applicant and the contents of which are incorporated herein by reference.

Technical Field

[0002] This disclosure relates to analysing medical services data related to services provided by medical service providers, for example, insurance claims data from doctors and hospitals.

Background

[0003] Doctors perform treatments on patients and claim their charges from a health insurance, such as a private health insurance. Similarly, hospitals also claim their charges from the health insurance. The claims typically include details about the symptoms, treatment, complications, the cost for the doctor's services, length of stay in hospital and material costs. As a result, this medical services data related to the services provided by the doctors and hospitals contains a large amount of data.

[0004] However, it is difficult with current methods to analyse this data in order to obtain meaningful reports.

[0005] Any discussion of documents, acts, materials, devices, articles or the like which has been included in the present specification is not to be taken as an admission that any or all of these matters form part of the prior art base or were common general knowledge in the field relevant to the present disclosure as it existed before the priority date of each claim of this application. [0006] Throughout this specification the word "comprise", or variations such as "comprises" or "comprising", will be understood to imply the inclusion of a stated element, integer or step, or group of elements, integers or steps, but not the exclusion of any other element, integer or step, or group of elements, integers or steps.

Summary

[0007] A computer implemented method for analysing medical services data related to services provided by medical service providers comprises:

determining a network of the medical service providers based on the medical services data, the network comprising a measure of relationship between any two of the medical service providers;

determining a network measure based on the measure of relationship between any two of the medical service providers; and

generating a report indicating the network measure.

[0008] Since the network measure in the report is based on the measure of

relationship, the report is more meaningful than reports using other measures that are not based on networks or relationships. For example, an average cost per medical service provider has limited value for addressing the underlying factors that influence medical performance. Instead, the reports generated by the above method have increased usability because they identify the network measure based on relationships. As a result, the reports provide an insight into the underlying structural characteristics of the medical data.

[0009] Determining the network may comprise determining the measure of relationship such that the measure of relationship between two medical service providers is indicative of how many medical services the two medical service providers provided together.

[0010] The network measure may be related to a quality of care parameter. [0011] The quality of care parameter may be one or more of:

length of stay;

complication rate;

medical cost;

infection rate;

planned readmission rate;

unplanned readmission rate;

revision procedures;

removal of a prosthesis; and

reinsertion of a prosthesis.

[0012] Determining the network may comprise determining a collaboration network comprising the medical service providers.

[0013] Determining the network measure may comprise determining a count of subnetworks by determining how many sub-networks have three medical service providers and a measure of relationship indicating that any two of the three medical service providers have provided at least one medical service together.

[0014] Determining the network may comprise determining a provider centred collaboration network associated with one of the medical service providers.

[0015] The medical service providers may comprise at least one surgeon and the provider centred collaboration network is a surgeon centred collaboration network associated with the at least one surgeon.

[0016] Determining the network measure may comprise:

determining a betweenness centrality measure for each of the medical service providers in the network; and

determining the network measure based on the betweenness centrality measure of each of the medical service providers in the network. [0017] Determining the network measure may comprise:

determining multiple distance values, each distance value being indicative of a distance of the betweenness centrality measure for a service provider in the network to a maximum betweenness centrality measure of all medical service providers; and

determining an aggregated distance value based on the multiple distance values; and

normalising the aggregated distance value based on a total number of the medical service providers.

[0018] The report may comprise a graphical visualisation of the network.

[0019] The graphical visualisation of the network may comprise for each of the medical service providers a node having a colour that indicates a type of that medical service provider.

[0020] The method may further comprise displaying the report on a computer display.

[0021] The method may further comprise generating an electronic document of the report and storing the electronic document on a data store.

[0022] The method may further comprise determining a team of medical service providers to maximise a quality of care parameter.

[0023] Determining the team may comprise determining the team based on diagnostic data associated with a patient.

[0024] Software, when installed on a computer causes the computer to perform the above method.

[0025] A computer system for analysing medical services data related to services provided by medical service providers comprises:

an input port to receive the medical services data;

a processor to determine a network of the medical service providers based on the medical services data, the network comprising a measure of relationship between any two of the medical service providers,

to determine a network measure based on the measure of relationship between any two of the medical service providers, and

to generating a report indicating the network measure; and a data store to store the report.

[0026] A method for visualising medical services data related to services provided by medical service providers comprises:

generating a graphical user interface comprising node symbols representing the medical service providers and graphical connections between pairs of two node symbols representing the measure of relationship between these two node symbols.

[0027] The user interface may comprise user controls to control filtering of the medical services data.

[0028] The user interface may comprise a user input for entering an identifier of one of the medical service providers and the method may further comprise, in response to a user entering an identifier of one of the medical service providers generating a provider centred collaboration network.

[0029] Software, that when installed on a computer causes the computer to perform the above method for visualising medical services data related to services provided by medical service providers.

[0030] A computer system for visualising medical services data related to services provided by medical service providers comprises:

an input port to receive the medical services data; a processor

to determine a network of the medical service providers based on the medical services data, the network comprising a measure of relationship between any two of the medical service providers, and

to generate a graphical user interface comprising node symbols representing the medical service providers and graphical connections between pairs of two node symbols representing the measure of relationship between these two node symbols; and

a display to display the graphical user interface.

[0031] Optional features described of any aspect of method, computer readable medium or computer system, where appropriate, similarly apply to the other aspects also described here.

Brief Description of Drawings

[0032] An example will be described with reference to

Fig. 1 illustrates a computer system for analysing medical data related to services provided by medical service providers.

Fig. 2 illustrates a method as performed by processor for analysing medical services data.

Fig. 3 illustrates an example of a collaboration graph.

Figs. 4a and 4b illustrate examples and of a Surgeon-Centric Collaboration Network.

Figs. 5a, 5b, 5c and 5d illustrate a star, a line, a circle and a complete graph, respectively.

Fig. 6 illustrates a surgeon node with its surrounding assistants and anaesthetists.

Fig. 7 illustrates a table of the impact of the network structure around a specialist (based on SCCN) on quality of cares.

Fig. 8 illustrates a table of the impact of non-network attributes on quality of cares.

Fig. 9 illustrates a table of the impact of network position of individual specialist in the complete network (CN) on quality of cares.

Fig. 10a illustrates table that shows the average LoS and percentage of admissions of the four knee categories performed in the dataset used for analysis in one example.

Fig. 10b illustrates a table which shows that in terms of all the four treatment types, group B consistently has a lower Average LoS compared to group A and also the whole dataset as shown in Table of Fig. 10a.

Fig. 11 illustrates a screenshot of a graphical user interface (GUI).

Fig. 12 illustrates a further example of a GUI.

Description of Embodiments

[0033] Fig. 1 illustrates a computer system 100 for analysing medical data related to services provided by medical service providers, such as claims data and hospital discharge data provided by doctors and hospitals to a health insurance.

[0034] The computer system 100 comprises a processor 102 connected to a program memory 104, a data memory 106, a communication port 108 and a user port 110. The program memory 104 is a non-transitory computer readable medium, such as a hard drive, a solid state disk or CD-ROM. Software, that is, an executable program stored on program memory 104 causes the processor 102 to perform the method in Fig. 2, that is, the processor determines a network comprising measures of relationships, determines a network measure and generates a report.

[0035] The processor 102 may then store the report on data store 106, such as on RAM or a processor register. Processor 102 may also send the report via

communication port 108 to another computer.

[0036] The processor 102 may receive data, such as medical services data, from data memory 106 as well as from the communications port 108 and the user port 110, which is connected to a display 112 that shows a visual representation 114 of the network or the report to a user 116 . In one example, the processor 102 receives medical services data from a doctor's computer via communications port 108, such as by using a Wi-Fi network according to IEEE 802.11. The Wi-Fi network may be a decentralised ad-hoc network, such that no dedicated management infrastructure, such as a router, is required or a centralised network with a router or access point managing the network.

[0037] In one example, the processor 102 receives and processes the medical services data in real time. This means that the processor 102 generates the report every time medical services data is received from a doctor or hospital and completes this calculation before the doctor or hospital sends the next record.

[0038] Although communications port 108 and user port 110 are shown as distinct entities, it is to be understood that any kind of data port may be used to receive data, such as a network connection, a memory interface, a pin of the chip package of processor 102, or logical ports, such as IP sockets or parameters of functions stored on program memory 104 and executed by processor 102. These parameters may be stored on data memory 106 and may be handled by-value or by-reference, that is, as a pointer, in the source code.

[0039] The processor 102 may receive data through all these interfaces, which includes memory access of volatile memory, such as cache or RAM, or non-volatile memory, such as an optical disk drive, hard disk drive, storage server or cloud storage. The computer system 100 may further be implemented within a cloud computing environment, such as a managed group of interconnected servers hosting a dynamic number of virtual machines.

[0040] It is to be understood that any receiving step may be preceded by the processor 102 determining or computing the data that is later received. For example, the processor 102 pre-processes the medical services data and stores the pre-processed data in data memory 106, such as RAM or a processor register. The processor 102 then requests the data from the data memory 106, such as by providing a read signal together with a memory address. The data memory 106 provides the data as a voltage signal on a physical bit line and the processor 102 receives the medical services data via a memory interface.

[0041] Fig. 2 illustrates a method 200 as performed by processor 102 for analysing medical services data. Processor 102 receives medical services data via

communication port in the form of an XML file, for example, and may contain medical claims data from a hospital or a doctor. Other formats may also be used, such as comma separated values. The medical services data may contain data from all doctors or hospitals of a particular area or country, such as Australia.

[0042] Based on the medical services data processor 102 determines 202 a network of the medical service providers. The network may be represented as a graph where each node represents a provider, such as a doctor or hospital. In one example, the originators of the medical services data may not be identical to the medical service providers. For example, the originator of the services data may be hospital and processor 102 extracts the names of the surgeons, assistants, anaesthetists and other providers from the data. Processor 102 uses these extracted names as nodes of medical service providers.

[0043] The graph may be stored on data memory 106 implicitly. For example, the processor 102 stores each service provider as one record or row in a provider table of a database, such as an SQL database. The database also comprises a relationship table, where each row represents a relationship between two of the medical service providers. For example, each row in the provider table stores for each provider a unique provider identifier and provider name, such as the name of a doctor or surgeon. Further, this table may store an indication of the provider's role, such as Ί ' to indicate a surgeon, '2' to indicate an assistant, '3' to indicate an anaesthetist and '4' to indicate a general practitioner.

[0044] The relationship table may comprise in each row a first provider identifier of a first provider and a second provider identifier for a second provider. Further, each row comprises a measure of relationship between the first provider and the second provider. In one example, the measure of relationship is the number of services that the first and second provider have provided together, such as procedures that a particular surgeon has performed together with a particular anaesthetist.

[0045] Processor 102 reads the first record of the medical services data and identifies the name of a first medical service provider, such as a surgeon. Processor 102 then queries the provider table of the database to check whether that first provider is already registered in the database. If the first provider is not found in the database, processor 102 adds a new record. Processor 102 then retrieves the identifier of that first provider.

[0046] Similarly, processor 102 extracts from the medical services data the name of a second service provider, such as an anaesthetist, queries the database, potentially adds a record and retrieves the identifier of the second provider.

[0047] Based on these two identifiers processor 102 can query the relationship table for a matching entry. If no entry exists this is an indication of a measure of relationship of zero and processor 102 adds a record with the two identifiers with the measure of relationship set to T. If an entry exists, processor 102 increments the measure of relationship for that entry by Ί ' .

[0048] This way the processor 102 determines a network comprising a measure of relationship between any two of the medical service providers. It is noted here that the number of entries in the relationship table may be less than the number of possible combinations of providers because many combinations have a measure of relationship of zero which is indicated by a missing record for that combination.

[0049] As a next step, processor 102 determines 204 a network measure based on the measure of relationship in the relationship table explained above. The network measure may be a betweenness centrality measure or other measure as described in more detail further below. [0050] Finally, processor 102 generates 206 a report indicating the network measure. For example, processor 102 may generate a pdf document with a graphical

representation of the network measure. Where the network measure is a numerical value, the report may contain a bar chart and the length of the chart indicates the value of the network measure. This way, an operator or analyst of the medical system can visually compare multiple network measures and draw conclusion or develop action items to enhance the service level or reduce the overall cost of the medical system.

[0051] The report may also me machine readable data, such as an XML file or data provided as a web-service and transmitted as part of an HTML POST command or as part of an AJAX communication.

[0052] Further, the report may contain a textual explanation of the network measure to indicate whether the network measure is high or low relative to an expected measure. Processor 102 may also evaluate the network measure and provide an indication of whether the network measure is satisfactory or whether parameters of the medical system should be adjusted to enhance service levels.

[0053] In this context, social network analysis (SNA) has the unique ability to play a new role in exploring the context and situations that lead to efficient and effective healthcare. The following description describes an example of method 200 in Fig. 2 in a specific context of private healthcare in Australia and describe an SNA based approach (applied to health insurance claims) to understand the nature of collaboration among doctors treating hospital inpatients and explore the impact of collaboration on cost and quality of care.

[0054] In particular, processor 102 performs network analysis to (a) design collaboration models among medical service providers, such as surgeons, anaesthetists and assistants, who work together while treating patients admitted for specific types of treatments (b) identify and extract specific types of network topologies that indicate the way doctors collaborate while treating patients and (c) analyse the impact of these topologies on cost and quality of care provided to those patients. [0055] In one example, processor 102 uses data from health insurance claims to design network-based models of collaboration among medical providers and analyses the impact of social networks and their underlying network structures, to discover provider communities and analyse the topology of the emerging community structure (of surgeons, anaesthetists and assistant surgeons) on treatment outcomes for patients who undergo specific category of surgeries, for example knee surgeries.

[0056] The methods described herein may be based on semantics and the data available in a private health insurance (PHI) claim, and the claiming patterns of hospitals and medical providers. Within the context of Australian PHI there are two types of claims. A medical claim is sent by a doctor - also referred to as a provider - who performs a service to treat a patient who is a member of a particular private health insurer. The medical claim has information about the provider, the member, the hospital where the patient was treated, the details of the treatment and the cost of the services provided.

[0057] A hospital claim is sent by a hospital's billing department and includes details of treatment, theatre charges, accommodation charges, prosthetics charges and charges for other services provided.

[0058] Disclosed herein are two types of networks to explore collaboration among medical providers: (i) collaboration networks (CN) designed to capture the

collaboration among surgeons, anaesthetists and assistant surgeons (ii) surgeon centric collaboration networks (SCCN) which explore an individual surgeon's connections.

[0059] In terms of the network representations used herein, a node in the network represents a (medical) provider such as surgeon, anaesthetist, assistant surgeon; the node size indicates the total amount charged by that provider; the thickness of the edge (or tie strength) connecting two nodes represents the number of common hospital admissions between the two providers. An admission refers to a single episode of admitted patient care. The time interval between the date of admission and the date of discharge represents the length of stay for that admission. [0060] In addition to size of nodes and tie strength, other network measures - closeness centrality and betweenness centrality that are related to the position of the node in the network, and centralization measures that indicate how central its most central node is in relation to how central other nodes are, can provide interesting insights about the influence of the node in the overall communication control capacity and the network. For example, the larger nodes with a more influential position in the network have the capacity to provide additional meaning within the context of the graph.

[0061] Questions that may be answered are:

• Is there a team structure that emerges as providers work together on a number of shared admissions?

• What is the impact of an individual surgeon's network on cost and quality of care of the surgeries performed?

• What types network structures have positive or negative impact on cost and quality of care?

[0062] In some examples, betweenness centralisation in the SCCN network is the variable that has significant positive influence on Length of Stay (LoS), Complication rate and Medical cost. This gives an indication that nodes with high betweenness centrality are likely to be in more demand. In another example, surgeons who collaborate with more number of teams appear to have a lower average LoS.

[0063] The proposed determination of surgeon collaboration offers a unique perspective as it combines theoretical analysis with empirical investigations of a PHI large dataset. The design of the collaboration model is influenced by domain experts who wish to understand the nature of team structures that have an impact on cost as well as quality of care provided to patients for specific types of treatments. In one example, the claims data is related to knee procedures. However, the methods is applicable to similar models for other orthopaedic procedures as well other treatment groups such as cardiology and cardio-thoracic procedures. [0064] The hospital and medical claims processed by an insurer contain data that specify the type of service provided during an admission, the length of stay for that admission, and the cost of that service. The service is specified as a Medicare Benefit Schedule (MBS) code [22], as stipulated by the Australian Government. The hospitals also send additional data related to an admission, once the patient is discharged. For any given hospital admission, processor 102 may receive three sets of data:

1. Medical claims - these show the provider- ID i.e. who performed a service, and the service is indicated by the MBS code for the specific type of treatment performed while the patient was in hospital;

2. Hospital claims - these are sent by the billing department of the hospital, and include MBS codes, accommodation cost, prosthetic costs, laboratory and radiology costs; and

3. Hospital discharge data that consolidates the patient's clinical care during that particular admission, and includes details such as length of stay, whether this was an unplanned readmission, and additional diagnosis codes that indicate complications or infections that occurred during that admission. Therefore the discharge data provides valuable information pertaining to quality of care.

[0065] Processor 102 combines these three sets of data using the patient's health insurance number as a unique patient identifier or primary key in a database.

[0066] Processor 102 may use data from all three sources to design the network models. A network graph represents the collaboration among three specific types of medical providers; the surgeons, the anaesthetists and assistant surgeons, as they perform knee-related surgical procedures.

[0067] In one example, the aim may be to investigate the quality of care provided by a specific provider or a group of providers who collaborate while performing knee surgeries. The PHI domain experts are interested in understanding the impact of collaboration among the three types of providers: surgeons, anaesthetists, assistant surgeons (also refer to assistants herein). This may be represented by a tripartite graph in which the nodes correspond to the three types of providers. [0068] The three sets of data offers content-rich health information about each admission episode. The admissions are categorized by the treatment codes as specified in the MBS coding taxonomy. For example, knee surgeries are coded in the following hierarchy: "Therapeutic— » Surgical Operation— » Orthopaedic— » Knee'. The nodes represent the providers, and the edges as the number of common admissions shared by the two providers. Processor 102 associates the node size with the total medical charges of the corresponding provider.

[0069] Fig. 3 illustrates an example of a collaboration graph 300. The three different types of providers are shown in three distinct gray levels: white for surgeons, such as indicated by reference numeral 302, black for anaesthetists, such as indicated by reference numeral 304 and grey for assistants, such as indicated by reference numeral 306. The size of the node is modeled as the medical charge of the provider. The edge thickness is modeled as the number of collaborating claims by two types of providers. A thicker edge indicates a higher number of shared admissions.

[0070] Just by glancing at the graph, one can immediately identify the 'big' providers, i.e. providers with high medical charges, as well as highly connected providers. We can also see isolated cluster of providers. Often such isolated clusters indicate providers working in a specific geographic region. This network graph offers a powerful visualization to study collaboration among providers.

[0071] One example use of the proposed methods is to study the impact of the collaboration network structure on quality of care.

[0072] Figs. 4a and 4b illustrate examples 400 and 420 of a Surgeon-Centric

Collaboration Network (SCCN) graphs comprising nodes for anaesthetists in black and nodes for assistants in white. In one example, processor 102 investigates a specific surgeon node in the CN and build the lower level SCCN 400 or 402. The SCCN 400 is a network of a specific surgeon while the SCCN 402 is a network of a different surgeon. The grahps show how a specific surgeon collaborates with the assistants and anaesthetists, and the hospital(s) in which they work together while performing knee surgeries. The individual surgeon node is not shown in the SCCN as all admissions (which are modeled as the edges) relate to a particular surgeon.

[0073] Therefore, processor 102 only models two types of edges, one is the edge between assistants and hospitals, the other is between anesthetists and hospitals. Since it's a surgeon centric network, the links between assistants and anesthetists are not shown.

[0074] However such links are shown in the CN graph 300 in Fig. 3. The SCCN networks 400 and 420 also shows the hospitals 402, 422 and 424 where the surgeon performs knee surgeries. The hospital node is represented symbolically in the form of a building. The size of the building indicates the total medical cost. Edge thickness is modeled as the number of admissions of the specific surgeon with an anaesthetic or assistant in that hospital.. The graph 400 in Fig. 4a shows a surgeon who only works in one hospital 402 and collaborates with nine anaesthetists or assistants. The graph 420 in Fig. 4b shows a surgeon who works in two hospitals 422 and 424 and collaborates with anaesthetists or assistants who also work in those hospitals.

[0075] Figs. 5a, 5b, 5c and 5d illustrate a star, a line, a circle and a complete graph, respectively.

[0076] Degree centrality is one of the basic measures of network centrality. It is the proportion of nodes that are adjacent to a particular actor in a network. Degree centrality highlights the node with the most links to other actors in a network, and can be defined by the following equation for the actor (or node) in a network having N actors:

[0077] C_D =

N - l

[0078] The subscript D for degree' and ά(η_{) indicates the number of actors with whom actor is connected. The maximum value for C_D is 1 when actor is linked with all other actors in the network. The set of degree centralities, which represents the collection of degree indices of N actors in a network, can be summarised by the following equation to measure network degree centralisation:

[0080] Where, {C_D («_;) } are the degree indices of N actors and C_D (n ) is the largest observed value in the degree indices. For a network, degree centralisation (i.e. the index C_D ) reaches its maximum value of 1 when one actor chooses all other (N - 1) actors and the other actors interact only with this one (i.e. the situation in a star graph as illustrated in Figure 3). This index (i.e. C_D ) attains its minimum value of 0 when all degrees are equal (i.e. the situation in a circle graph as illustrated in Figure 3). Thus, C_D indicates varying amounts of centralisation of degree compared to both star and circle graphs.

[0081] Closeness centrality, another view of actor centrality based on closeness or distance, focuses on how close' an actor is to all the other actors in a network. The idea is that an actor is central if it can quickly interact with all other actors in a network. In the context of a communication relation, such actors need not rely on other actors for the relaying of information. For an individual actor, it can be represented as a function of shortest distances between that actor and all other remaining actors in the network. The following equation represents the closeness centrality for a node in a network having N actors:

[0082] C ,. ) =

[0083] Where, the subscript C for "closeness', din^ n^ is the number of lines in the shortest path between actor and actor j , and the sum is taken over all i≠ j . A higher value of C_c(n.) indicates that actor i is closer to other actors of the network, and will be 1 when actor has direct links with all other actors of the network. The set of closeness centralities, which represents the collection of closeness indices of N actors in a network, can be summarised by the following equation to measure network closeness centralisation:

[0084] C_c =

- l][N - 2] / [2N

[0085] Where, (C_c(n.)} are the closeness indices of N actors and C_c(n_i) is the largest observed value in closeness indices. For a network, closeness centralisation (i.e. the index C_c ) reaches its maximum value of unity when one actor chooses all other (7V -1) actors and the other actors have shortest distances (i.e. geodesies) of length 2 to the remaining (N— 2) actors (i.e. the situation in a star graph as illustrated in Figure 5a). This index (i.e. C_c ) can attain its minimum value of 0 when the lengths of shortest distances (i.e. geodesies) are all equal (i.e. the situation in a complete graph and circle graph as illustrated in Figs. 5c and 5d, respectively). Thus, closeness centralisation indicates varying amounts of centralisation of closeness compared to star, circle and complete graph.

[0086] Betweenness centrality is obtained by determining how often a particular node is found on the shortest path between any pair of actors (or nodes) in the network. It views an actor as being in a favoured position to the extent that the actor falls on the shortest paths between other pairs of actors in the network. That is, nodes that occur on many shortest paths between other pairs of nodes have higher betweenness centrality than those that do not. Therefore, it can be regarded as a measure of strategic advantage and information control. In a network of size n , the betweenness centrality for an actor (or node) can be represented by the following equation:

[0088] Where, i≠ j≠k ; represents the number of shortest paths linking the two actors that contain actor ; and g_jk is the number of shortest paths linking actor j and k . From the set of betweenness centralities of N actors in a network betweenness centralisation can be defined by the following equation:

N

[0090] Where, (C_B (n.) } are the betweenness indices of N actors and C_B (n ) is the largest observed value in betweenness indices. For a network, betweenness

centralisation (i.e. the index C_B ) reaches its maximum value of unity when one actor chooses all other (N - l) actors and the other actors have shortest distances (i.e.

geodesies) of length 2 to the remaining (N— 2) actors (i.e. the situation in a star graph as illustrated in Fig. 5a). This index (i.e. C_B ) can attain its minimum value of 0 when all actors have exactly the same actor betweenness index (i.e. the situation in a line graph as illustrated in Fig. 5b). Thus, C_B indicates varying amounts of centralisation of betweenness compared to both star and line graph.

[0091] Density measures how connected a graph is. For example, if a graph G has N nodes, V edges. Then the density D_G of the graph G is calculated as:

2 *V

[0092]

N(N - l)

D_G reaches maximum value as 1 when the graph is fully connected, and minimum value as 0 when there is no edge. [0094] In terms of non-network variables, in one example the following four admission related features are selected:

%. of distinct anaesthetists

%. of distinct assistants

%. of distinct pathologists

%. of distinct imaging providers

[0095] The admission data shows all the medical providers who are involved in treating a patient during that admission. We consider four types of providers that includes: anaesthetists, assistants, pathologists and imaging providers. In some examples, a surgery has one principal surgeon and assistant and anaesthetist who work with the surgeon during the surgery. Specifically, processor 102 considers the number of distinct providers a surgeon collaborates with while performing a knee surgery. For the data analysis, processor 102 considers the percentage of distinct providers who collaborate with the surgeon rather than the absolute number of providers. The percentage is calculated with the denominator as the sum of distinct number of the four types of providers collaborating with the surgeon in knee procedure.

[0096] For network features, we first consider CN graph as depicted in Fig. 3. There are three types of nodes in the graph, out of which, around 500 nodes are surgeon' nodes. Amongst several possible network variables, the following five network features may be selected:

[0097] Network features of CN

1 Clustering coefficient

2 Number of triangles

3 Degree centrality

4 Closeness centrality

5 Betweenness centrality

[0098] These variables were chosen as they have the potential to offer insights into the collaboration patterns among providers. [0099] Clustering-coefficient: The local clustering coefficient of a vertex (node) in a graph quantifies how close its neighbors are to being a clique (complete graph). This represents the strength of the surgeon's network.

[0100] Number of triangles: The global clustering coefficient is based on triplets of nodes. A triplet consists of three nodes that are connected by either two (open triplet) or three (closed triplet) undirected ties. A triangle consists of three closed triplets, one centred on each of the nodes. In our context, a triangle shows the three types of providers i.e. surgeons, anaesthetists and assistants working together while performing knee surgeries.

[0101] Fig. 6 illustrates a surgeon node 602 with its surrounding assistants 604 and 606 and anaesthetists 608, 610 and 612. This is a small subset extracted from the CN graph shown in Fig. 3. The surgeon node 602, which is represented hatched, has three triangles connected to it. This indicates that the surgeon performs knee surgeries frequently with three pairs of assistants and anaesthetists. The three centrality measures are explained above. The features from the CN graph may be all node-level features.

[0102] Next, processor 102 considers network-level features. For each surgeon, processor 102 stores one SCCN graph on data memory 106. In one example, processor 102 considers the following four network measures for the SCCN graph:

1 Degree centralisation

2 Closeness centralisation

3 Betweenness centralisation

4 Density

[0103] In one example, processor 102 considers the following quality of care parameters:

LoS - the average length of stay for all admissions of the surgeon. This information is available in hospital discharge data as explained above.

Medical cost - the average medical charge for all the knee related admissions teated by the surgeon. Complication rate - calculated as the percentage of admissions with complications out of all the admissions of a surgeon.

[0104] In one example, the dataset for knee surgeries includes a total of 59, 256 admissions performed by 870 surgeons. However, in order to make more robust conclusions, processor 102 may only consider surgeons who had more than ten claims. Processor 102 may further remove surgeons who appeared as outliers. In one example, processor 102 carries out the analysis on a set of 559 surgeons. For the variables above, processor 102 applies a z-score standardization. Thus each variable has a mean of zero and a standard deviation of one. In the simple regression analysis, this allow to interpret the constant and the slope" term appropriately.

[0105] Processor 102 may perform the standardization before performing the analysis for each feature in the form of a column x according to:

_* x - x

x =

std(x)

[0106] Fig. 7 illustrates a table 700 of the impact of the network structure around a specialist (based on SCCN) on quality of cares (i.e. LoS, Complication rate and Medical cost).

[0107] Fig. 9 illustrates a table 900 of the impact of network position of individual specialist in the complete network (CN) on quality of cares (i.e. LoS, Complication rate).

[0108] The quality of care parameters introduced above are the dependent variables in a regression analysis. Since the independent variables are semantically distinct in the healthcare domain, they have been dealt with independently. Hence an individual linear model has been constructed for each variable to be executed by processor 102.

Although most of the linear models have a low R² value, our focus are the β values, which are significant. [0109] Table 700 in Fig. 7 explores the impact of admission-related features on the dependent variables (i.e. LoS, and Medical cost). We can see that a higher percentage of anaesthetist and assistant indicates a lower LoS and Medical cost, while a higher percentage of pathologists and imaging providers indicates a higher LoS and Medical cost. This shows that admissions with more imaging may be more severe situations and thus incur longer LoS and higher Medical cost.

[0110] In Table 800 in Fig. 8, we observe that betweenness centralisation is the only variable that has significant positive influence on LoS, Complication rate and Medical cost. This can be interpreted as follows: From the perspective of a SCCN structure, a high betweenness centralisation indicates that the structure of the corresponding SCCN follows a star-like or centralized structure since betweenness centralisation reaches its highest value of 1 for a star network. A star-like or centralized network has few actors with higher betweenness centrality values and the rest actors have very low

betweenness centrality values. In this type of network, only a small number of actors play major collaboration and communication role. That indicates there is a presence of network hubs" in this type of network.

[0111] On the other side, if network actors have almost equal level of network connectivity (as like a line graph) then betweenness centralisation will be small and in such networks there does not present any network hub". Therefore, SCCN, where participating actors have almost equal level of network connectivity, will produce lower LoS, Complication rate and Medical cost. In the context of health care domain, this offers an interesting insight. In their corresponding hospitals, healthcare managers or administrators could encourage a practice culture where each member will have equal level of network connectivity.

[0112] Table 900 in Fig. 9 explores the impact of the network position of the individual specialist in the complete network (CN) on independent variables (i.e. LoS, Complication rate). We can observe that in the case of both LoS and Complication rate, the variable "Number of triangles' has a negative correlation. That is, when a surgeon works with a large number of distinct groups, LoS and Complication rate are lower. [0113] Intuitively, we have two assumptions with respect to the variable Number of triangles': (i) Surgeons who work with large number of distinct assistants or anaesthetists could be involved in more complicated surgeries and thus resulting longer Los and higher complication rate, (ii) Surgeons who consistently work with only a few distinct assistants or anaesthetists have a lower number of triangles. For these cases, the analysis shows a higher LoS. Our conjecture is that this limits external influence of other providers on the surgeon. The converse case where the number of triangles is higher clearly shows lower LoS.

[0114] Fig. 10a illustrates table 1000 that shows the average LoS and percentage of admissions of the four knee categories performed in the dataset used for analysis in one example. The dataset used includes about 59, 256 knee surgeries performed by 559 surgeons over a period of 2 years. As per the MBS descriptions, there are four broad categories of knee surgeries with varying degrees of complexity. Accordingly, the average length of stay for each category of knee surgery varies. Column 3 of table 1000 also shows the distribution of the four categories of knee surgery.

[0115] We conducted an empirical investigation to analyze the performance of teams of surgeons indicated by the No, of triangles as shown in Table 900.

[0116] Fig. 10b illustrates table 1020 which shows that in terms of all the four treatment types, group B consistently has a lower Average LoS compared to group A and also the whole dataset as shown in Table 1000. Table 1020 shows two groups of providers: Group A and group B. Group A represents the 200 surgeons having the least Number of triangles, and group B represent 200 surgeons with largest Number of triangles. The purpose of this analysis is to compare the Average LoS for each category of knee surgery for the two groups of providers.

[0117] Next we compare the Average LoS for each category of knee surgery for the two groups in Table 1020 with the Average LoS of the complete dataset summarised in Table 1000. We can observe that in all four categories of knee surgeries, group B consistently has a lower Average LoS compared to group A, as well as the whole dataset shown in Table 4. The empirical investigation implies that surgeons who work with a higher number of teams appear to have a lower length of stay.

[0118] The above results provide a strong indication that network features like degree, betweenness and closeness centralization and number of triangles have a statistically significant impact on efficiency metrics. In particular, for surgeon centric networks, betweenness centralization is significant for all three metrics: Length of Stay, Complication rate and Medical cost. This observation can potentially be used by health care providers to reorganize surgical teams and improve the overall efficiency of health care delivery.

[0119] Fig. 11 illustrates a screenshot of a graphical user interface (GUI) 1100 as displayed on display device 112. In this example, processor 102 generates the GUI 1100 to display the collaboration network 300 determined in step 202 of method 200 in Fig. 2. GUI 1100 comprises selection elements to allow input from operator 116. In particular, GUI 1100 comprises hierarchical drop-down menus 1102, 1104, 1106 and 1108.

[0120] GUI 1100 further comprises drop-down menus to select a treatment code 1110, a diagnosis code 1112, an hospital ID or name 1114 and a provider ID 1116. GUI 1100 also comprises sliders to control filtering of the graph elements, such as edge weight 118, node degree 1120, node benefit 1122 and time series 1124. Operator 116 can make selections using the drop-down menus 1110, 1112, 1114 and 1116 and move the sliders 1118, 1120, 1122 and 1124 up and down. GUI 1100 sends these filtering commands as parameter values, such as in an XMLHttpRequest or an event handler, to processor 102. Processor 102 receives the filtering parameters and changes the CN 300 according to the selected drop-down menu items and slider values and generates a display of the updated CN in GUI 1100. Operator 116 can observe the change in the CN 300 to visually inspect the characteristics of the medical services data.

[0121] In this example, when operator 116 presses Ctrl and click on a surgeon node, GUI 1100 sends a request to processor 102 to view additional details about the cluster of providers. The request comprises the provider ID associated with that node.

Processor 102 receives the request, retrieves the details from the data base and sends the data back to the GUI 1100 such that they are displayed in GUI 1100. In one example, simply hovering the mouse over a node triggers the above process to display more details of that done. Operator 116 can also zoom in using the mouse scroll wheel and pan using click and drag operations.

[0122] Fig. 12 illustrates a further example of a GUI 1200 comprising similar dropdown selection menus as GUI 1100 in Fig. 11. In Fig. 12 operator 116 has entered a particular provider ID in field 1116. Processor 102 receives this provider ID and consequently, switches the GUI 1200 to display the SCCN 400 instead of the CN 300 in Fig. 11. The SCCN 400 also has nodes and edges as described above, where the size of the nodes represents the benefit paid to the provider and the thickness of the line indicates the number of common admissions. Star shape nodes represent pathology providers and diamond shape nodes represent radiology providers. These node shapes may also be used in GUI 1100 in Fig. 11.

[0123] GUI 1200 further comprises a provider profile field 1202 showing the name 1204, number of admissions 1206, total amount charged 1208, total benefit paid 1210 and top 10 codes 1212 by benefits paid.

[0124] It will be appreciated by persons skilled in the art that numerous variations and/or modifications may be made to the specific embodiments without departing from the scope as defined in the claims.

[0125] It should be understood that the techniques of the present disclosure might be implemented using a variety of technologies. For example, the methods described herein may be implemented by a series of computer executable instructions residing on a suitable computer readable medium. Suitable computer readable media may include volatile (e.g. RAM) and/or non-volatile (e.g. ROM, disk) memory, carrier waves and transmission media. Exemplary carrier waves may take the form of electrical, electromagnetic or optical signals conveying digital data steams along a local network or a publically accessible network such as the internet.

[0126] It should also be understood that, unless specifically stated otherwise as apparent from the following discussion, it is appreciated that throughout the description, discussions utilizing terms such as "estimating" or "processing" or "computing" or "calculating", "optimizing" or "determining" or "displaying" or "maximising" or the like, refer to the action and processes of a computer system, or similar electronic computing device, that processes and transforms data represented as physical (electronic) quantities within the computer system's registers and memories into other data similarly represented as physical quantities within the computer system memories or registers or other such information storage, transmission or display devices.

[0127] The present embodiments are, therefore, to be considered in all respects as illustrative and not restrictive.

Claims

CLAIMS:

1. A computer implemented method for analysing medical services data related to services provided by medical service providers, the method comprising:

generating a report indicating the network measure.

2. The method of claim 1, wherein determining the network comprises determining the measure of relationship such that the measure of relationship between two medical service providers is indicative of how many medical services the two medical service providers provided together.

3. The method of claim 1 or 2, wherein the network measure is related to a quality of care parameter.

4. The method of claim 3, wherein the quality of care parameter is one or more of:

length of stay;

complication rate;

medical cost;

infection rate;

planned readmission rate;

unplanned readmission rate;

revision procedures;

removal of a prosthesis; and

reinsertion of a prosthesis.

5. The method of any one of the preceding claims, wherein determining the network comprises determining a collaboration network comprising the medical service providers.

6. The method of claim 5, wherein determining the network measure comprises determining a count of sub-networks by determining how many sub-networks have three medical service providers and a measure of relationship indicating that any two of the three medical service providers have provided at least one medical service together.

7. The method of any one of the preceding claims, wherein determining the network comprises determining a provider centred collaboration network associated with one of the medical service providers.

8. The method of claim 7, wherein the medical service providers comprise at least one surgeon and the provider centred collaboration network is a surgeon centred collaboration network associated with the at least one surgeon.

9. The method of claim 7 or 8, wherein determining the network measure comprises:

determining the network measure based on the betweenness centrality measure of each of the medical service providers in the network.

10. The method of claim 9, wherein determining the network measure comprises: determining multiple distance values, each distance value being indicative of a distance of the betweenness centrality measure for a service provider in the network to a maximum betweenness centrality measure of all medical service providers; and

determining an aggregated distance value based on the multiple distance values; and normalising the aggregated distance value based on a total number of the medical service providers.

11. The method of any one of the preceding claims, wherein the report comprises a graphical visualisation of the network.

12. The method of claim 11, wherein the graphical visualisation of the network comprises for each of the medical service providers a node having a colour that indicates a type of that medical service provider.

13. The method of any one of the preceding claims, further comprising displaying the report on a computer display.

14. The method of any one of the preceding claims, further comprising generating an electronic document of the report and storing the electronic document on a data store.

15. The method of any one of the preceding claims, further comprising determining a team of medical service providers to maximise a quality of care parameter.

16. The method of claim 15, wherein determining the team comprises determining the team based on diagnostic data associated with a patient.

17. Software, that when installed on a computer causes the computer to perform the method of any one of the preceding claims.

18. A computer system for analysing medical services data related to services provided by medical service providers, the computer system comprising:

an input port to receive the medical services data;

19. A method for visualising medical services data related to services provided by medical service providers, the method comprising:

20. The method of claim 19, wherein the user interface comprises user controls to control filtering of the medical services data.

21. The method of claim 19 or 20, wherein the user interface comprises a user input for entering an identifier of one of the medical service providers and the method further comprises, in response to a user entering an identifier of one of the medical service providers generating a provider centred collaboration network.

22. Software, that when installed on a computer causes the computer to perform the method of any one of the claims 19 to 21.

23. A computer system for visualising medical services data related to services provided by medical service providers, the computer system comprising:

an input port to receive the medical services data; a processor

a display to display the graphical user interface.