Method and system of scheduling appointments

Abstract

The present invention seeks to provide a method of scheduling appointments, the method including, in one or more electronic processing devices: accessing historical data relating to past appointments; accessing supply data indicative of a clinician availability; calculating a predicted clinician availability using the supply data and the historical data, wherein the predicted clinician availability is calculated to correct for overestimation of clinician availability; generating a patient flow using the predicted clinician availability, wherein the patient flow is indicative of a queue of patients for future appointments; and, at least one of: scheduling a future appointment using the patient flow; and updating the patient flow to add patients requiring future appointments.

Description

METHOD AND SYSTEM OF SCHEDULING APPOINTMENTSBackground of the Invention

[0001] The present invention relates to a system and method of scheduling appointments. In one particular example, the present invention relates to a system and method of scheduling appointments for outpatients in a clinical setting.Description of the Prior Art

[0002] The reference in this specification to any prior publication (or information derived from it), or to any matter which is known, is not, and should not be taken as an acknowledgement or admission or any form of suggestion that the prior publication (or information derived from it) or known matter forms part of the common general knowledge in the field of endeavour to which this specification relates.

[0003] Every year, there are approximately 3 million new outpatient specialist clinic appointments at local public hospital networks in Victoria, Australia. Growing daily demand for these services leads to high-volume waitlists and long appointment delays for patients.

[0004] Increasing demand for outpatient specialist services in Australia creates many problems for healthcare providers to efficiently manage waitlists. The waitlist includes appointments that have been delayed which may cause patients’ dissatisfaction. It also raises important concerns for patient delay risk by delaying further care in a manner not informed by a clinician. It is, therefore, important to provide tools to assist with scheduling new appointments and / or managing waitlist in outpatient specialist clinics.Summary of the Present Invention

[0005] In one broad form, an aspect of the invention seeks to provide a method of scheduling appointments, the method including, in one or more electronic processing devices: accessing historical data relating to past appointments; accessing supply data indicative of a clinician availability; calculating a predicted clinician availability using the supply data and the historical data, wherein the predicted clinician availability is calculated to correct for overestimation of clinician availability; generating a patient flow using the predicted clinicianavailability, wherein the patient flow is indicative of a queue of patients for future appointments; and, at least one of: scheduling a future appointment using the patient flow; and, updating the patient flow to add patients requiring future appointments.

[0006] In one example, the method includes, in one or more electronic processing devices: receiving an appointment request indicative of a clinician and a recall time; and, at least one of: scheduling the future appointment using the appointment request; and, updating the patient flow using the appointment request.

[0007] In one example, when updating the patient flow, the method includes, in one or more electronic processing devices, providing a queue length of the updated patient flow.

[0008] In one example, the step of generating the patient flow includes: generating an initial patient flow using the clinician availability; and, modifying the initial patient flow to generate the patient flow using the predicted clinician availability.

[0009] In one example, the historical data is indicative of at least one of: appointment data in a previous time period; and, appointment data indicative of scheduled appointments.

[0010] In one example, the previous time period is at least one of: 3 months; 6 months; 12 months; 18 months; and 24 months.

[0011] In one example, the supply data is indicative of at least one of: a number of available sessions in a week for each clinician, unit and service type combination; information of each session within a week; and, a number of patients booked for each session.

[0012] In one example, the appointment request is indicative of at least one of: a unit; a care provider; and, a service type.

[0013] In one example, the step of generating the patient flow includes generating the patient flow using a model equationQk = max Qk-1+ xk- Ck, 0); where Qkis the queue length at time k = 1,2, ...;xkis the number of patient appointments at week / cCkis the number of patients removed from the queue at week k: and is the backlog which represents the patient appointments with a recall time before week k.

[0014] In one example, the step of calculating the predicted clinician availability, in one or more electronic processing devices, includes: extrapolating the supply data for a future time period based on the historical data; calculating an open factor fraction of the clinician availability; and, applying the open factor fraction to thereby calculate the predicted clinician availability.

[0015] In one example, the future time period is at least one of: 3 months; 6 months; 12 months; 18 months; and 24 months.

[0016] In one example, the step of calculating the open factor fraction, in one or more electronic processing devices, includes at least one of: calculating an actual available session divided by a planned available session based on the historical data; and, calculating using an exponentially weighted moving average withO Pnew_predict OPnewjcctual + (1 - a)0Poldpredict where OPnewj]redictis a new prediction of an open factor fraction;0Pnew _actuai is the actual open factor fraction for the most recent week;OP0id_predict is the previous week’s value of the open factor fraction; and a = 0.01 .

[0017] In one example, the step of scheduling the future appointment using the patient flow and the appointment request, in one or more electronic processing devices, includes: determining an appointment availability and patients to be scheduled using the patient flow;determining a recall time of the patients to be scheduled; and, scheduling the appointments using the recall time.

[0018] In one example, the method includes scheduling patients with a most overdue recall time.

[0019] In one example, the step of scheduling the future appointment using the patient flow and the appointment request, in one or more electronic processing devices, includes: determining an appointment availability of the recall time using the patient flow; if the appointment availability indicates available, scheduling the future appointment of the recall time; and, if the appointment availability indicates not available, repeating the above steps for a time interval of the recall time.

[0020] In one example, the time interval is at least one of: ± 1 days; ± 3 days; ± 5 days; ± 1 week; and, selected depending on the appointment request.

[0021] In one example, the method further includes, in one or more processing devices, updating the historical data to reflect the scheduled future appointment.

[0022] In one example, the step of updating the historical data is performed at least one of: as soon as the future appoint is scheduled; daily; weekly; fortnightly; and, monthly.

[0023] In one broad form, an aspect of the invention seeks to provide a system of scheduling appointments, the system including: a database containing historical data and supply data; and, one or more electronic processing devices being configured to: access the historical data relating to past appointments from the database; access the supply data indicative of a clinician availability from the database; calculate a predicted clinician availability using the supply data and the historical data, wherein the predicted clinician availability is calculated to correct for overestimation of clinician availability; generate a patient flow using the predicted clinician availability, wherein the patient flow is indicative of a queue of patients for future appointments; and, at least one of: schedule a future appointment using the patient flow; and, update the patient flow to add patients requiring future appointments.

[0024] In one example, the one or more electronic processing devices is configured to: receive an appointment request indicative of a clinician and a recall time; and, at least one of: schedule the future appointment using the appointment request; and, update the patient flow using the appointment request.

[0025] In one example, when updating the patient flow, the one or more electronic processing devices is configured to provide a queue length of the updated patient flow.

[0026] In one example, the one or more electronic processing devices, at the step of generating the patient flow, is configured to: generate an initial patient flow using the clinician availability; and, modify the initial patient flow to generate the patient flow using the predicted clinician availability.

[0027] In one example, the historical data is indicative of at least one of: appointment data in a previous time period; and, appointment data indicative of scheduled appointments.

[0028] In one example, the previous time period is at least one of: 3 months; 6 months; 12 months; 18 months; and 24 months.

[0029] In one example, the supply data is indicative of at least one of: a number of available sessions in a week for each clinician, unit and service type combination; information of each session within a week; and, a number of patients booked for each session.

[0030] In one example, the appointment request is indicative of at least one of: a unit; a care provider; and, a service type.

[0031] In one example, the one or more electronic processing devices, at the step of generating the patient flow, is configured to generate the patient flow using a model equationQk = max Qk^ + xk- Ck, 0); where Qkis the queue length at time k = 1,2, ...; xkis the number of patient appointments at week k;Ckis the number of patients removed from the queue at week ; andQk-i is the backlog which represents the patient appointments with a recall time before week k.

[0032] In one example, the one or more electronic processing devices, at the step of calculating the predicted clinician availability, is configured to: extrapolate the supply data for a future time period based on the historical data; calculate an open factor fraction of the clinician availability; and, apply the open factor fraction to thereby calculate the predicted clinician availability.

[0033] In one example, the future time period is at least one of: 3 months; 6 months; 12 months; 18 months; and 24 months.

[0034] In one example, the one or more electronic processing devices, at the step of calculating the open factor fraction, is configure to at least one of: calculate an actual available session divided by a planned available session based on the historical data; and, calculate using an exponentially weighted moving average withO Pnew_predict ^-OPnew-aciua[ + (1 - a)0Poidpredict where OPnewj]redictis a new prediction of an open factor fraction;OPnew _actuai is the actual open factor fraction for the most recent week;OP0id_predict is the previous week’s value of the open factor fraction; and a = 0.01 .

[0035] In one example, the one or more electronic processing devices, at the step of scheduling the future appointment using the patient flow and the appointment request, in one or more electronic processing devices, includes: determining an appointment availability and patients to be scheduled using the patient flow; determining a recall time of the patients to be scheduled; and, scheduling the appointments using the recall time.

[0036] In one example, the one or more electronic processing devices is configured to schedule patients with a most overdue recall time.

[0037] In one example, the one or more electronic processing devices, at the step of scheduling the future appointment using the patient flow and the appointment request, is configured to: determine an appointment availability of the recall time using the patient flow; if the appointment availability indicates available, schedule the future appointment of the recall time; and, if the appointment availability indicates not available, repeat the above steps for a time interval of the recall time.

[0038] In one example, the time interval is at least one of: ± 1 days; ± 3 days; ± 5 days; ± 1 week; and, selected depending on the appointment request.

[0039] In one example, the one or more electronic processing devices is configure to update the historical data to reflect the scheduled future appointment.

[0040] In one example, the one or more electronic processing devices is configure to update the historical data at least one of: as soon as the future appoint is scheduled; daily; weekly; fortnightly; and, monthly.

[0041] It will be appreciated that the broad forms of the invention and their respective features can be used in conjunction and / or independently, and reference to separate broad forms is not intended to be limiting. Furthermore, it will be appreciated that features of the method can be performed using the system or apparatus and that features of the system or apparatus can be implemented using the method.Brief Description of the Drawings

[0042] Various examples and embodiments of the present invention will now be described with reference to the accompanying drawings, in which: -

[0043] Figure l is a flow chart of an example of a method for use in an appointment scheduling process;

[0044] Figure 2 is a flow chart of an example of a method for use in an appointment scheduling process;

[0045] Figure 3 is a schematic diagram of a specific example of an appointment scheduling system;

[0046] Figure 4 is as schematic diagram of an example of a processing system;

[0047] Figure 5 is a schematic diagram of an example of a client device;

[0048] Figure 6 is a screenshot of an exemplary of a supply table;

[0049] Figure 7 is a schematic diagram of a patient flow before / after a review / new appointment;

[0050] Figure 8A is a graph illustrating a count of yearly registered records for review appointments in the waiting list during the study period;

[0051] Figure 8B is a graph illustrating a count of yearly overdue review appointments during the study period;

[0052] Figure 9A is a graph illustrating a distribution of session closure (in percentage) by category;

[0053] Figure 9B is a graph illustrating a distribution of overdue appointments (in weeks) using different scenarios;

[0054] Figures 10(a) to 10(c) are screenshots of an exemplary of a prototype of the electronic booking form;

[0055] Figure 11 is a schematic diagram of a system design of a next available appointment tool;

[0056] Figure 12 is a schematic diagram of system structure of web application of implemented tools;

[0057] Figure 13 is a screenshot of an exemplary display of the next available appointment tool;

[0058] Figure 14 is a screenshot of an exemplary display of the template planning tool; and,

[0059] Figure 15 is a screenshot of an exemplary display of the template planning tool output.Detailed Description of the Preferred Embodiments

[0060] An example of a process for scheduling an appointment will now be described with reference to Figure 1.

[0061] For the purpose of illustration, it is also assumed that the process is performed at least in part using one or more electronic processing devices forming part of one or more scheduling appointment systems. Whilst the system can use multiple processing devices, with processing performed by one or more of the devices, for the purpose of ease of illustration, the following examples will refer to a single device, but it will be appreciated that reference to a singular processing device should be understood to encompass multiple processing devices and vice versa, with processing being distributed between the devices as appropriate.

[0062] In this example, the scheduling appointment process involves the processing device receiving an appointment request, the appointment request including appointment request data indicative of one or more scheduling parameters, at step 100. The appointment request may be received from a patient, a care provider, an administration personnel of a clinic or the like using web forms, apps, email messages or text messages. In one example, the appointment request may be generated and received from a process for scheduling appointments on waitlists. In this example, the appointment request data may be extracted from a database including the waitlist. Alternatively, the appointment request might be received from a medical practitioner, such as a referring clinician or similar.

[0063] The appointment request data may be indicative of one or more of a unit, a recall time, a service type, and / or a care provider. The unit is indicative of the specific outpatient department or clinic. The recall time may be the date or week of the appointment, such as 31 August 2023 or week 38 2023, or a period from the time of requesting the appointment, for example, 5 days later or in 6 weeks. The service type is indicative of services provided by the units and clinicians, for example, new liver, Rev Pri Derm Dr Bruce, a regular check-up, a chemotherapy session, or the like. The care provider indicates the specific clinician orphysician assigned to the appointment. If the care provider is not indicated in the appointment request, it is defaulted to all clinicians of the unit.

[0064] At step 110, upon receiving the appointment request that includes the above-described appointment request data, the processing device proceeds to schedule an appointment using the scheduling parameters as indicated in the appointment request data.

[0065] The scheduling appointment process, at step 111, involves the processing device accessing a database containing appointment data indicative of a patient flow to identify one or more available time slots. The patient flow indicates available appointments over a subsequent time period, and may be based on supply data provided by a clinician, historical appointment data for a previous time period of 3 months, 6 months, 12 months, 18 months, or 24 months, which can be used in predicting appointment availability as will be described in more detail below. In one example, the database including data in a form of a table, each row indicates a week and a number of available time slots for a care provider on the week. The database may include further tables containing detailed information of each time slots, such as the day of the week and the number of sessions or patients in the time slot.

[0066] At step 112, the process selects an available time slot to thereby schedule an appointment.

[0067] At step 120, the processing device generates an indication of a scheduled appointment based on the selected available time slot. In one example, the indication is provided to a user to notify the user that the appointment is successfully scheduled.

[0068] At step 130, the processing device updates the appointment data with the scheduled appointment. In one example, the selected available time slot in the appointment data is marked as scheduled and / or unavailable, so that the available time slot is not double-booked by subsequent scheduling. In one example, the appointment data may be updated real-time as the appointments are being scheduled, or periodically, such as daily, weekly, fortnightly and / or monthly.

[0069] Accordingly, the above described process schedules an appointment based on a request, using a patient flow, which is in turn indicative available appointment times based on aclinician availability. This approach thereby schedules an appointment that is less likely to be cancelled. This allows clinician to manage appointment scheduling and minimise the risk of delayed appointment, and in turn, provide improved patient experience and outcome by increasing successful appointment rate. By using the appointment scheduling process as described, overdue appointments may be minimised and thereby historical backlog of appointments is reduced or eliminated over a shorter period of time.

[0070] A number of further features will now be described.

[0071] In one example, the scheduling parameters include at least one of an appointment window and clinician availability parameters. In one example, when the appointment request indicates a certain recall time being a date but there is no available time slots on the date, the process then selects potential available time slots in a day after the date for scheduling an appointment.

[0072] The process may further include having the processing device access a waitlist backlog in the database, wherein the waitlist backlog including a plurality of past appointment requests, each appointment request in the waitlist backlog includes a past recall time; determining an appointment priority for each of past appointment request; and scheduling an appointment based on the appointment priority. In one example, when a past appointment request includes a past recall time that is earliest, the appointment priority is determined to be the highest, and thus, the past appointment request will be scheduled first using the process described above in order to clear the waitlist backlog. In one example, one patient had an appointment and the clinician asked for a follow-up in three months, for a patient who had an appointment one week after that, the clinician may ask for a follow-up in two months. In this situation, the second patient came later but will have a higher priority as its recall date is sooner.

[0073] The processing device may further predict the clinician availability by extrapolating clinician availability for a future time period at least in part using one of clinician availability parameters, a number of clinicians allocated to the unit, historical appointment data and an open factor fraction. In one example, the clinician availability is extrapolated based on historical appointment data, such as 3 days every week. The clinician availability may beextrapolated for the future 3 months, 6 months, 12 months, 18 months or 24 months from the date of scheduling appointment.

[0074] The clinician availability parameters may include one of hours worked, public holidays, school holidays, influenza season, and pandemic. The historical appointment data includes scheduled and successful appointment data; and, scheduled and cancelled appointment data. In one example, when predicting clinician availability for the recall time, the process takes into considerations that there is no available time slot on public holidays and weekends. Furthermore, it is also taken into considerations that the clinician availability will be low on days that fall near public holidays, such as between Christmas Day and New Y ear's Day, during school holidays, flu seasons or when there is an outbreak of a pandemic.

[0075] In one example, the process can further take into consideration of a general overestimation of the clinician availability. This can be used for example to account for the fact that appointments overrun and there may be other unexpected reasons why clinicians are unavailable, such as illness or the like. This can therefore help provide a better prediction of the clinician availability. In one example, the processing device calculates an open factor fraction which is then used in an extrapolating step. The open factor fraction may be calculated based on an actual available session divided by a planned available session in a past period of time. Additionally or alternatively, the open factor fraction can be calculated using an exponentially weighted moving average withwhere 0 Pnew predictis a new prediction of an open factor fraction;OPnew_actuai is the actual open factor fraction for the most recent week;0 Poid_predict is the previous week’s value of the open factor fraction; and a = 0.01 .The open factor fraction may also be calculated based on the unit, the service type, and / or the care provider.

[0076] In one example, the processing device can schedule a plurality of appointments using multiple sets of different appointment request data including different scheduling parameters; determining a waitlist backlog for each of the multiple sets of different scheduling parameters; and, using the waitlist backlogs to optimise the scheduling parameters.

[0077] Accordingly, the above described process can schedule an appointment based on a request, determine a next available time in accordance with the clinician availability and thereby schedule an appointment. The process further predicts the clinician availability to provide better or more realistic predictions of the clinician availability. This provides improved patient experience and outcome by increasing successful appointment rate. By using the appointment scheduling process as described, overdue appointments may be minimised and thereby historical backlog of appointments is reduced or eliminated after over a shorter period of time.Exemplary Embodiment - Optimised Scheduling

[0078] An example of a process for scheduling appointment will now be described with reference to Figure 2.

[0079] In this example, the scheduling appointment process involves the processing device accessing historical data relating to past appointments, at step 200. The historical data is accessed from a database. The database may contain historical data relating to past appointment. In one example, the historical data includes appointment data in a previous time period of 3 months, 6 months, 12 months, 18 months, or 24 months. Additionally, the historical data may include appointment data indicative of appointments that have already been scheduled in the future.

[0080] At step 210, the processing device accesses supply data indicative of a clinician availability. In one example, the supply data is accessed from a database. The supply data may be indicative of a number of available sessions in a week for each clinician, unit and service type combination, information of each session within a week, and / or a number of patients booked for each session. In one example, the supply data is in a form of a table, each row indicates a number of available time slots for a clinician / care provider, for the particular unit,of a week. The supply data may include further tables containing detailed information of each time slots, such as the day of the week and a number of sessions or patients in the time slot. In one example, the supply data is provided by the clinician(s) or on behalf of the clinician(s). The supply data may be provided periodically, such as weekly, fortnightly, monthly, annually or on an ad-hoc basis depending on the operation of the scheduling system and / or the database. In one example, this is provided annually, allowing scheduling to be performed up to a year in advance.

[0081] At step 220, a predicted clinician availability is calculated using the supply data and the historical data. The predicted clinician availability is calculated to correct for overestimation of clinician availability. Clinician availabilities are often overestimated as failing to take into account of their regular off days, such as public holidays, annual leave or sick leave days, and missed or overdue appointments that they need to reschedule. The predicted clinician availability is therefore more reliable and better reflects the actual availability of the clinician, in turn leading to more accurate patient flow, and hence a reduced likelihood of appointment cancelation.

[0082] The processing device generates a patient flow using the predicted clinician availability, at step 230. The patient flow is indicative of a queue of patients for future appointments. In one example, the patient flow is generated as a queue, or a queue length using a model or a model equation. The queue may be indicative of the number of patients that need to be seen by a clinician each week, and the queue length may represent the number of available slot of the week.

[0083] At step 241, a future appointment can be scheduled using the patient flow. For example, this can be performed in response to an appointment request received in respect of a patient, as will be described in more detail below. Additionally, and / or alternatively, this might be performed in order to allocate appointment to patients currently on a waiting list. In either case, if there are available slots for the week of the recall time, the future appointment canthen be scheduled. Alternatively or additionally, the patient flow can be updated to add patients requiring future appointments, at step 242, for example to add one or more patients to the patient queue, for example to clear a patient backlog, waitlist, or the like.

[0084] In the event that an appointment is scheduled, optionally, an indication of the future appointment is provided, at step 250. In one example, the indication indicates that the future appointment has been scheduled, and provides relevant information, such as the time and date of the scheduled appointment.

[0085] Accordingly, the above described process can model a patient flow, for example to predict numbers of patients that require appointments in given time periods and / or provide a real-time indication of whether one or more appointments can be successfully scheduled within a given time period, which in one example is performed based on an appointment request by modelling a patient flow for the week of the recall time of the appointment request. The process improves the success rate of appointment scheduling, and allows for improved patient handling and reductions in wasted time for clinicians by avoiding cancelled appointments. This in turn, provides improved patient experience and outcome. By using the appointment scheduling process or patient flow modelling as described, overdue appointments may be minimised and thereby historical backlog of appointments is reduced or eliminated after over a shorter period of time.

[0086] It should be appreciated that the above-described steps may be performed in different orders depending on the preferred implementation. Alternatively, a patient flow could be generated periodically, for example, by performing steps 200 to 241 and / or 242 once a week, with the resulting patient flow being used to schedule multiple appointment requests received during the following week.

[0087] A number of further features will now be described.

[0088] In one example, the processing device receives an appointment request indicative of a clinician and a recall time. Similar to the above described process, the recall time may be the date or week of the appointment, or a period from the time of requesting the appointment. In one example, the appointment request is further indicative of a unit, a care provider, and a service type. The unit is indicative of the specific outpatient department or clinic. The service type is indicative of service provided in the appointment, for example, a regular check-up, a chemotherapy session, or the like. The care provider indicates the specific clinician or physician assigned to the appointment. If the care provider is not indicated in the appointmentrequest, it is defaulted to all clinicians of the unit. In this example, a future appointment is scheduled using the appointment request. In one example, if there are available slots for the week of the recall time, the future appointment is then scheduled. Alternatively or additionally, the patient flow is updated using the appointment request.

[0089] In one example, when updating the patient flow, the process provides a queue length of the updated patient flow. In this example, a user may simulate a patient flow with one or more future appointments, and the process provides an indication of the update patient flow to thereby allowing the user to reconfigure the future appointments and optimise the patient flow accordingly.

[0090] It should be appreciated that the above-described steps may be performed in different orders depending on the preferred implementation. For example, the patient flow could be calculated in real time when an appointment request is received. In this case receiving the appointment request occurs before steps 200 to 241 and / or 242. This allows the latest available data to be used in calculating the patient flow, potentially making this more accurate. However, this could be expensive from a computational resource perspective and so alternatively, a patient flow could be generated periodically as describe above.

[0091] In one example, the processing device generates the patient flow by generating an initial patient flow using the clinician availability and modifying the initial patient flow to generate the patient flow using the predicted clinician availability. This allows the patient flow to better represent an actual patient flow by using the predicted clinician availability, and thereby, minimising the effects of overestimating clinician availability and eliminating overdue and / or backlog of appointments.

[0092] The patient flow may be generated or simulated using a model equation. The model equation may be a queue or a queue length. In one example, the model equation isQk = max Qk^ + xk- Ck, 0); where Qkis the queue length at time k = 1,2, xkis the number of patient appointments at week k:Ckis the number of patients removed from the queue at week k: andQk-i is the backlog which represents the patient appointments with a recall time before week k.

[0093] In one example, the process may further include the processing device extrapolating the supply data for a future time period based on the historical data, calculating an open factor fraction of the clinician availability, and applying the open factor fraction to thereby calculate the predicted clinician availability. The future time period may be 3 months, 6 months, 12 months, 18 months, and / or 24 months. As the extrapolated supply data is likely to overestimate the clinician availability, the open factor fraction, which is indicative of a corrective fraction, is the applied to avoid overestimating clinician availability. This allows the clinician availability for the week of the recall time to better reflect the actual clinician availability, and hence, provides a more realistic indication of availability.

[0094] The open factor fraction may be calculated based on an actual available session divided by a planned available session in a past period of time. Additionally or alternatively, the open factor fraction can be calculated using an exponentially weighted moving average withwhere OPnew predictis a new prediction of an open factor fraction;OPnew_actuai isthe actual open factor fraction for the most recent week;0 Poid_predict is the previous week’s value of the open factor fraction; and a = 0.01 .

[0095] The processing device, in one example, determines an appointment availability and patients to be scheduled using the patient flow, determines a recall time of the patients to be scheduled, and scheduling the appointments using the recall time. In one example, the processing device schedules patients with a most overdue recall time. This allows patients with earliest recall time or most overdue recall time to be scheduled first, and thereby, reducing or minimising backlog.

[0096] The processing device, in one example, determines an appointment availability of the recall time using the patient flow, and if the appointment availability indicates available, the future appointment of the recall time is then scheduled. If the appointment availability indicatesnot available, the above steps are repeated for a time interval of the recall time. The time interval may be ± 1 days, ± 3 days, ± 5 days, ± 1 week, and / or selected depending on the appointment request. In one example, the patient flow of the recall time indicates that there is no available slots of the week, in this event, the processing device then looks into a week after the recall time and schedules the future appointment when the patient flow of the week after indicates that there are available slots. This allows the user to schedule an appointment in a time near the requested recall time when the requested recall time is not available.

[0097] In one example, the processing device updates the historical data to reflect the scheduled future appointment. The update may be performed as soon as the future appointment is scheduled, daily, weekly, fortnightly, and / or monthly. This allows other / subsequent future appointments to be scheduled based on the up-to-date data.

[0098] Accordingly, the above described process provides a real-time indication of whether an appointment can be successfully scheduled based on the appointment request by modelling a patient flow for the week of the recall time in the appointment request. This process assists in improving the success rate of appointment scheduling, and in turn, provides improved patient experience and outcome. By using the appointment scheduling process as described, overdue appointments may be minimised and thereby historical backlog of appointments may be reduced or eliminated over a shorter period of time.

[0099] A specific example of a scheduling appointment system will now be described in more detail with reference to Figures 3 to 6.

[0100] In this example, the scheduling appointment system includes a database 310. A number of client devices 320 are provided, with some of these being in communication with the database 310. A processing system 330, such as one or more servers, is provided in communication with the client devices 320 via one or more communications networks 340.

[0101] It will be appreciated that the configuration of the networks 340 are for the purpose of example only, and in practice the client devices 320 and the processing system 330 can communicate via any appropriate mechanism, such as via wired or wireless connections, including, but not limited to mobile networks, private networks, such as an 802.11 networks,the Internet, LANs, WANs, or the like, as well as via direct or point-to-point connections, such as Bluetooth, or the like.

[0102] Whilst the processing system 330 is shown as a single entity, it will be appreciated that in practice the processing system 330 can be distributed over a number of geographically separate locations, for example as part of a cloud-based environment. However, the above described arrangement is not essential and other suitable configurations could be used.

[0103] An example of a suitable processing system 330 is shown in Figure 4. In this example, the processing system 330 includes at least one microprocessor 400, a memory 401, an optional input / output device 402, such as a keyboard and / or display, and an external interface 403, interconnected via a bus 404 as shown. In this example the external interface 403 can be utilised for connecting the processing system 330 to peripheral devices, such as the communications networks 340, databases 310, other storage devices, or the like. Although a single external interface 403 is shown, this is for the purpose of example only, and in practice multiple interfaces using various methods (e.g. Ethernet, serial, USB, wireless or the like) may be provided.

[0104] In use, the microprocessor 400 executes instructions in the form of applications software stored in the memory 401 to allow the required processes to be performed. The applications software may include one or more software modules, and may be executed in a suitable execution environment, such as an operating system environment, or the like.

[0105] Accordingly, it will be appreciated that the processing system 330 may be formed from any suitable processing system, such as a suitably programmed client device, PC, web server, network server, or the like. In one particular example, the processing system 330 is a standard processing system such as an Intel Architecture based processing system, which executes software applications stored on non-volatile (e.g., hard disk) storage, although this is not essential. However, it will also be understood that the processing system could be any electronic processing device such as a microprocessor, microchip processor, logic gate configuration, firmware optionally associated with implementing logic such as an FPGA (Field Programmable Gate Array), or any other electronic device, system or arrangement.

[0106] As shown in Figure 5, in one example, the client device 320 includes at least one microprocessor 500, a memory 501, an input / output device 502, such as a keyboard and / or display, and an external interface 503, interconnected via a bus 504 as shown. In this example the external interface 503 can be utilised for connecting the client device 320 to other devices, such as the database 310, the communications networks 340, other storage devices, or the like. Although a single external interface 503 is shown, this is for the purpose of example only, and in practice multiple interfaces using various methods (e.g. Ethernet, serial, USB, wireless or the like) may be provided.

[0107] In use, the microprocessor 500 executes instructions in the form of applications software stored in the memory 501 to allow for communication with the processing systems 330, the database 310 as well as to allow user interaction for example through a suitable user interface.

[0108] Accordingly, it will be appreciated that the client devices 320 may be formed from any suitable processing system, such as a suitably programmed PC, Internet terminal, lap-top, or hand-held PC, and in one preferred example is either a tablet, or smart phone, or the like. Thus, in one example, the client device 320 is a standard processing system such as an Intel Architecture based processing system, which executes software applications stored on nonvolatile (e.g., hard disk) storage, although this is not essential. However, it will also be understood that the client devices 320 can be any electronic processing device such as a microprocessor, microchip processor, logic gate configuration, firmware optionally associated with implementing logic such as an FPGA (Field Programmable Gate Array), or any other electronic device, system or arrangement.

[0109] In one example, the one or more processing systems 330 are servers, which communicate with the client devices 320 via a communications network, or the like, depending on the particular network infrastructure available. The servers 330 typically execute applications software for performing required tasks including storing, searching and processing of data, with actions performed by the servers 330 being performed by the processor 400 in accordance with instructions stored as applications software in the memory 401 and / or input commands received from a user via the I / O device 402, or commands received from the client device 320.

[0110] For the purpose of the following examples, it is assumed that the user interacts with the client device 320 via a GUI (Graphical User Interface), or the like presented on a display of the client device 320, and in one particular example via a browser application that displays webpages, or an App that displays relevant information. Actions performed by the client devices 320 are performed by the processor 500 in accordance with instructions stored as applications software in the memory 501 and / or input commands received from a user via the I / O device 502.[oni] However, it will be appreciated that the above described configuration assumed for the purpose of the following examples is not essential, and numerous other configurations may be used. It will also be appreciated that the partitioning of functionality between the client devices 320, and the servers 330 may vary, depending on the particular implementation.

[0112] In one example, the client devices 320 may be used by the users / patients requesting an appointment, or by the clinicians / care providers examining schedules or operating the appointment scheduling process.

[0113] Examples of the processes for scheduling appointment will now be described in further detail with reference to a specific example performed on real appointment data.

[0114] Every year there are approximately 3 million new outpatient specialist clinic appointments at local hospital networks in Victoria, Australia. Growing daily demand for these services leads to high-volume waiting lists and therefore long appointment delays for patients. This phenomenon emphasizes the importance of providing analytics and tools to assist with waiting list management in outpatient specialist clinics. In the following descriptions, a novel Next Available Appointment (NAA) tool is developed, to assist clinicians to manage delayed- appointment risk and improve the patient experience by aligning the expected and actual day of the appointment. The NAA uses simulation to determine the earliest available week for a patient appointment on or after the timeframe requested by the clinician, considering the current waiting list and future planned clinician availability. It was validated using three years of historical waiting list information across several scenarios chosen to capture operational diversity. As a practical example, a scenario chosen for implementation within the clinic’s operational setting achieved a simulated reduction in overdue appointments from 41% to 25%(i.e., a reduction of 47,000 overdue appointments over three years). Early details are also provided on the implementation of the tool currently underway.

[0115] Increasing demand for outpatient specialist services in Australia creates many problems for healthcare providers to efficiently manage waiting lists. The lack of transparency in treating clinicians on their own clinic waiting lists and appointment availability can have important direct and indirect consequences. For example, creating an expectation within a patient to return in a specific time frame (e.g., 6 months), when the waiting list means the appointment will be in 12 months can cause patients’ dissatisfaction. It also raises important concerns for patient delay risk by delaying further care in a manner not informed by a clinician. Lack of waiting delay visibility also breaks the feedback loop between a specialist’s discharge practices and an ever-growing list of patients needing Review appointments. Where it is medically appropriate to return a specialist care patient to primary care, there can be an improvement to the waiting list. Furthermore, visibility in waiting list delay and ever-growing patient lists make it harder for hospitals to achieve mandated “access targets” for timely care of “New-Routine” and “New-Urgent” patients.

[0116] Innovative and strategic solutions are required to help clinics manage the ever-growing waiting lists. Improving appointment processes is one of the approaches that can help manage waiting list backlogs. Previous research has focused on different aspects of appointment systems: optimising appointment intervals (i.e., the interval between two successive appointment times), determining patient priorities (i.e., categorising patients into different groups to determine their priorities), determining the exact day / time of appointment, determining the exact appointment, and determining the order in which a list of patients should be scheduled. However, these tools are largely designed around the mechanics of scheduling patients into day / time slots. Tools that can provide clinicians, at the end of each consultation, an accurate timeframe (possibly more than 12 months in the future) for the patient’s next appointment are still lacking. The situation where most patients are placed on first in-first out waiting lists with the precise day / time slot of the appointment to be determined later (i.e., appointments are not immediately scheduled into day / time slots) is considered. More precisely, for appointments within 7 weeks, day / time slots are scheduled immediately, but appointments for more than 7 weeks in the future (by far the majority) are not scheduled into a day / time slotuntil 7 weeks in advance. This “convenience -based” system is used by the hospital, in part, to reduce the non-presentation rate at appointments. The focus is on predicting the next available appointment weeks beyond the 7-week window. Thus, the administrative mechanics of scheduling into exact day / time slots is not be concerned.

[0117] In this example, there is provided a mathematical solution that helps clinicians, considering the current appointment load, to advise patients on realistic future appointments. By having an appointment prediction tool, specialist clinicians will be able to ensure timeframes given to patients are realistic. Clinicians will also obtain feedback on the state of their waiting lists and can take the first steps to reduce delays through actions like an increased discharge of patients to primary care.

[0118] The following descriptions aimed to outline the development of a novel tool, called the “Next Available Appointment”, to provide informed decision-making capability using near- real-time analytics to predict the next available Review appointments, while taking into consideration the current state of waiting lists and the clinics’ scheduled capacity. This tool immediately notifies the clinician if the requested timeframe for a future patient appointment is not available. It then prompts the clinician on alternative options such as the next predicted available timeframe, thereby creating realistic expectations on appointment timeframes for patients. The clinician can determine the risk to patient safety if the appointment is delayed relative to the initial unavailable timeframe.MaterialsData Sources

[0119] Data used in the following were provided by one of Australia’s major metropolitan hospitals, for the period of 1 July 2014 to 30 June 2019. Patient data were de-identified using a unique identifier created by the hospital.

[0120] Data was provided for the tool development in three tables.

[0121] Table 1 lists the provided data tables along with the number of records for the study period (July 1, 2014 - June 30, 2019). The data can be viewed as contributing to one of twobasic quantities: patient demand and clinician supply. Patient demand was obtained from the waiting list table which stores patients registered on the waiting list. Clinician supply was recorded in two tables: “OP Services Scheduled Resources” (i.e., OP Table) and “Vacant Slots” (i.e., VS Table).

[0122] The OP Table describes high-level information as it shows the effective periods for each clinician supply record (called a “roster” here). Figure 6 shows a snapshot of the supply (OP) Table, split into two parts to accommodate its width. The roster in the first row would run from 27 March 2015 to 02 August 2015, using “EffectiveFrom” and “ScheduleGenerateUntil” columns, with 5 spots available for appointments (i.e., MaxAllowed represents the capacity for each roster) on every Friday for the given information (e.g., given unit, care provider, and service type).

[0123] The VS Table contains low-level clinician supply information about the clinic roster similar to information in the OP Table. The VS Table breaks down each OP Table record into weekly sessions. For example, for the first roster in OP Table, there is a record for each Friday with the same information for the given time window. A row of the VS Table is referred to as a “session”. The VS Table also includes additional information such as the number of patients booked for each session. In the VS Table, the number of spots planned for a given session (i.e., the session capacity) is represented by a variable called Template, which is analogous to the MaxAllowed value in the OP Table.

[0124] The three tables include over 100 variables in total. However, for this example, seven key variables were used, as shown in Table 2. Unit, Service Type and Care provider were provided in all three tables. Registration date and Recall date were provided only in the waitinglist data, while Template and Bookings data fields were only provided in the clinician supply tables.Table 2 Summary of variables used for simulated queuingExemplary Embodiment - Optimised Planning

[0125] Table 3 shows the exploratory data for all three tables. It includes the missing values of each variable, the number of unique values for each categorical variable, the range for the datevariables, and the median for the numeric variables. The waiting list dataset included 903,910 records. 38,273 records had missing values for the service type and 3,206 records for which the care provider was “unknown”. Those 41,479 records, representing 4.5% of the waiting list records, were excluded since they did not relate to a (unit, service type, care provider) queue. Of the remaining waiting list records, 557,609 (65%) were Review records. For 30,306 of these Review records, the recall date was after the end of the study period (30 June 2019). As they represented a demand for an appointment after the study period, they were safely excluded. There were 21,824 Review records where the recall date was missing, and these were also excluded because the recall date represents the priority of patients in the queue and is a key variable in the simulation. In the VS Table, 5,645 records with missing values for the unit were excluded as they were mapping variables to the waiting lists in the example. Similarly, the OP Table’s records with a missing Template value were also filtered out as this is a necessary variable for working out the clinician supply of the given roster.Table 3 Data summary1 - number of unique values for “categorical” variables, Median for “numeric” variables, and Range for “date” variables.2 - means that the variable is not in that dataset.The final data used in this example represented 60 unique units, 820 unique service types, and 370 unique care providers, which contained almost 1400 different (unit, service type, care provider) unique combinations.MethodsPatient flow modelling and the Lindley equation

[0126] The process of patients going through a waiting list to see a specialist clinician was modelled as a first-in / first-out Lindley queue. In a Lindley queue, the queue length Qkat time k = 1,2, ... can be defined using the Lindley equationQk = max Qk-1+ xk- Ck, 0); where xkis the number of patient appointments at week k: Ckis the number of patients removed from the queue at week k: and Qt-i is the backlog which represents the patient appointments with a recall time before week k.Patient demand modelling

[0127] Following an appointment, specialist clinicians decide to either discharge the patient back to their GP for further monitoring or continue their care in follow-up specialist appointments. For the latter, clinicians provide a recommended timeframe (i.e., recall date),and patients are added to a particular waiting list specified by unit, service type, and care provider, with a timeframe fortheir next appointments. The patient journey is shown in Figure 7.Clinician Supply

[0128] NAA runs simulated queuing on a weekly basis, therefore, if a Lindley queue of interest included multiple sessions (e.g., different days of the week, moming / aftemoon sessions) in a week, the Template values are summed to find their total supply for each week.

[0129] There are two problems in using session data to form the clinician supply. First, clinician supply data are not available for some Lindley queues for the entire next two years. To address this problem, the clinician supply is extrapolated to the end of the 2-year prediction window. Second, the historical data used for modelling overestimates the clinician supply. This is a known limitation of the data, as confirmed by the hospital administrators. To deal with this issue, models for an “open factor fraction” are created to capture the fraction of clinician supply provided for patient appointments. These models leam over time to capture any effect from the use of the tool on these fractions.Extrapolation Model

[0130] The basic approach of the extrapolation models is to simulate “virtual” VS Table records, including a Template column, into the future using the OP Table and then extending those records to complete a two-year future period. For example, the first roster shown in Figure 1 covers less than two years (i.e., from 27 March 2015 to 02 August 2015). So, to simulate the next two years, roster to cover the entire two years (e.g., from 27 March 2015 to 02 August 2017) is extrapolated. The biggest issue with this approach is overestimating the clinician supply because it considers that all sessions occur without any interruption. However, as described by hospital administrators, and later confirmed by the data, sessions are often closed / cancelled for various reasons.Overestimation

[0131] Analysis of historical VS data indicates that around 15% of sessions were closed. There is a range of reasons for session closures, but, in short, the Template field in the clinician supply data does not capture a range of situations where a session is not offered. Thus, relying on the Template variable for predicting future clinician supply could greatly overestimate the actual supply.

[0132] To avoid overestimating the clinician supply, the extrapolation model adopts two approaches. First, it assigns no clinician supply for sessions on public holidays. These are knowable well in advance, providing an easy way to improve predicted Template supply. Analysing historical data confirms that in over 99% of sessions listed on a public holiday, the Bookings field is 0.

[0133] A second key approach to avoid overestimating clinician supply recognises that even after allowing for public holidays, a session may still not be offered. To address this, the fraction of non-public holiday opportunities that sessions are open for each (unit, service type, care provider) combination, which is called the “open factor fraction”, is calculated. Using the historical data, this fraction provides a more accurate clinician supply (on average) for the next two years. Using the historical data, the open factor fraction for each (unit, service type, care provider) triple is calculated as the actual total supply (i.e., summation of the Template field of sessions that are not closed) divided by the planned total supply (i.e., summation of the Template field).

[0134] To capture large-scale changes in open factor fractions over time, an exponentially weighted moving average approach was employed. This approach uses past information (memory) to predict values of the open factor fraction according to the formulawhere OPnew predictis a new prediction of an open factor fraction; OPnewactual is the actual open factor fraction for the most recent week; OP0id predictis the previous week’s value of the open factor fraction; and a is a weight that controls the mixing of old and new information. If the most recent week includes a public holiday (i.e., a “knowable” zero Template value), theprevious prediction is carried over. Based on the extensive investigation, a value of a = 0.01 was determined as a suitable compromise between responsiveness and smoothing. This provides time for the smoothed value to become more realistic in the event the first few weeks produce an exceptional initial value. To initialise the sequence, the average over a bum-in period of 26 weeks was used to create the first open factor fraction forecast. In the production environment, new types of (unit, service type, care provider) sessions necessarily have no historical data, so 26 weeks is a compromise between bum-in time for a usable open factor fraction and waiting time where there is insufficient data to generate a forecast.Next Available Appointment method

[0135] The approach of the Next Available Appointment (NAA) tool is to predict a suitable week for a Review appointment on or after a date requested by the clinician (i.e., recall date), considering the current waiting list and an accurate estimate of the future clinician supply. The NAA tool creates weekly simulated queues and does not focus on dynamics within the day of the appointment or administrative details around the day / time of appointments in a week.

[0136] The NAA tool performs simulated queueing for each (unit, service type, care provider) combination. Therefore, the NAA tool for a given week k first finds the overdue (also called ‘backlog’) patients (i.e., Qk-i), which are the patients still on the waiting list with a recall date before week k. Then it creates a table, called Availability table, that includes xkand Ck, where xkrepresents the number of recall dates of each week after k (i.e., k + 1, k + 2,..., k + 104) for a two-year future period, and Ckrepresents the future clinician supply for each week after k using the extrapolation model. Since the NAA runs simulated queuing on a weekly basis, the Template values of all sessions that happen in a given week k are added to capture the weekly Ck. Using Qk-, and the created Availability table featuring xkand Ck, the NAA tool operates the simulated queueing (the Lindley equation) to create the additional column Qkfor weeks (k + 1, k + 2,... , k + 104).

[0137] For implementation purposes, the Qkcolumn in the Availability table for each (unit, service type, care provider) combination is created once during the weekend and used on any day of the following week until new tables are created on the next weekend. Understandably, the small effect that appointments placed on the waiting list earlier in the week can have onpredictions made later in the week is ignored. This approach greatly reduces the computational burden compared to nightly calculation, while still facilitating a near-real-time query for the approximate waiting list. This is implemented using a quick look-up table with two columns: week and Spot- A vailable. The week represents each week after k (i.e., k + 1, k + 2,... , k + 104) weeks for a two-year future period, and Spot- A vailable is a yes / no binary column which indicates whether there is sufficient clinician supply to accommodate additional patient appointments in that future week.Experimental design

[0138] To evaluate the performance of the predictions, the historical data is used to compare the week of actual appointments with four predictions, each based on a choice of data to inform the clinician supply. All but the recall date-based estimations involve simulated queueing and a choice for the clinician supply volume. Since historical data are used, it is possible to know exact numbers by the benefit of hindsight. Some choices for clinician supply volume serve more as hypothetical, best-case comparisons.Current (recall date)

[0139] The clinician-requested recall date represents the timeframe for the next appointment. In the case of a large waiting list, the actual appointment may be delayed until after the recall date. In this sense, the recall date is more properly a “request” than a prediction, but its role as a measure for overdue appointments makes it appropriate for comparison here. Since this comparison between the recall date and the actual appointment date is the approach currently used in the hospital, the baseline is adopted for comparisons, and it serves as the gold standard for comparison purposes.

[0140] Three simulation scenarios are adopted that are detailed below:1) “Realistic” scenario

[0141] In this approach, the weekly totals of the Template variable of the VS Table for a (unit, service type, care provider) combination are used as the clinician supply volume. The Templatedata field does not capture closures (e.g., public holidays) and is therefore optimistic about clinician supply.2) “Impossible” scenario

[0142] In this approach, the weekly total of the Bookings column of the VS Table is used as the clinician supply volume. This captures the volume of appointments that were scheduled to occur, including overbooked appointments. ("Overbooked” appointments occur when the number of Bookings appointments exceeds the number of Template appointments.) Since the values of Bookings are not knowable for a whole two-year period in the future, using it as clinician supply volume is hypothetical and impossible for real operation, but revealing to assess data quality and helpfulness.3) “Extrapolation ” scenario

[0143] In this approach, clinician supply volume is the weekly total for a (unit, service type, care provider) combination using the proposed extrapolation model. This approach is practical for use within an operational setting.

[0144] For predictions based on simulated queueing, predictions were based on the active waiting list as of the week the records were registered on the waiting list (“now week”) for a (unit, service type, care provider) combination. Waiting list records with earlier recall dates had higher priority and had their appointment earlier. An appointment occurring later than the week of the prediction was considered an overdue appointment. The number of weeks after the prediction was the amount of delay.Current existing overdue appointments in the hospital

[0145] In this example, 85.4% of appointments lead to follow-up appointments, while 14.6% of appointments do not (e.g., administrative discharge, patient transfer). This information was confirmed by analysing the number of Review appointments registered in the waiting list data in each year during the study period in Figure 8A, and also the number of overdue Review appointments (i.e., appointments which did not occur by their recall date during the study period) in Figure 8B. As can be seen in Figure 8A and Figure 8B, both the number of Reviewappointments on the waiting list and the number of overdue Review appointments are increasing. This is a cause for concern, especially for hospital resource planning.Overestimation of clinician supply

[0146] As described in previous section, sessions in the vacant slot data are often closed / cancelled for various reasons. The session closure rate is not the same throughout the units, service types, care providers, and therefore (unit, service type, care provider) combinations. The session closure for each (unit, service type, care provider) combination with at least 100 sessions is calculated. The calculation is repeated separately for units, service types, and care providers. The distribution of the session closure percentages for each of the scenarios is given in Figure 9A.NAA tool validation

[0147] A bum-in period comprised of the first 2 years of data was used to accommodate the backlog of patient appointments arising from before the study period. The validation period started on the week of 4 July 2016. Waiting list records that were either active on the waiting list on that date or had a registration date in a later week, but before 1 July 2019 were included. Active waiting list records at the end of the study period were assumed to have an appointment in a future week. For these censored records, the appointment week is set to the week after the study period. Note that this approach to censoring does not affect the median (i.e., the 50th percentile) or the 75th percentile if the fraction of censored records is less than 50% or 25%, respectively, of the data. In this data, the rates of censoring were always less than 20%. The results are shown in Table 4. Besides the overall results (considering all data sets), similar results for 5 units with the highest numbers of patients are presented.Table 4 Percentage of the -waiting list overdue records using current (recall) or hypothetical timeframes

[0148] In total, 301,778 waiting list records were considered. Compared to the week of the recall date, about 41% of appointments were overdue. The Impossible scenario suggests that, with impossibly good clinician supply volume numbers, simulated queueing could provide the appointment week so that only about 24% of patient appointments are overdue. The Extrapolation approach, which can be utilised in practice, provides a prediction with fewer overdue appointments, reducing the overdue percentage by 15.5% to 25.6% compared with using the week of the recall date. This translates to a reduction of 47,000 fewer overdue appointments in this cohort. It is also worth mentioning that the Extrapolation scenario provides fewer overdue appointments (e.g., about 6,000 fewer overdue appointments) than the “Realistic” scenario which emphasises the effectiveness of the proposed Extrapolation model.

[0149] Additional rows in Table 4 show results for five specific units, sorted by the number of waiting list records. Units are identified with pseudonyms to preserve confidentiality. From these results, the overdue percentage was the same or lower for these units, sometimes by over 25%. This suggests the simulated next available appointment using the Extrapolation model could be widely helpful for scheduling within specialist clinics. Figure 9B presents the distribution of overdue appointments in different scenarios.

[0150] Table 5 shows the median and 75th percentile of the delay across delayed appointments. Note that delayed appointments are specific to the prediction scenarios, so the number of delayed appointments varies between scenarios. Using the current system, amongst the 41. 1% of appointments that were delayed, half the delays (i.e., the median) were 5 weeks or more, and 25% (i.e., the 75th percentile) were 12 weeks or more. The mean delay is not calculated,in part due to issues from the censoring of records that haven’t had an appointment yet. Additional results for five individual units are also shown in Table 5.

[0151] It is worth elaborating on why using the model-based prediction can have far fewer overdue patients yet simultaneously have delays that are longer than the current system. The key is the 47,000 records that are delayed in the current system and included in those calculations but are excluded from the Extrapolation calculations (i.e., not overdue using the Extrapolation model). When the predicted week is after the week of the recall date (in most cases), the actual appointment week is necessarily larger than the week of the recall date (i.e., overdue), but less than the Extrapolation prediction (i.e., not overdue). They are small delays, and there are many of them. When excluded from the Extrapolation calculation, the calculated statistics shift upwards.Table 5 Median and 75thpercentile (75,h-p) of delay size in -weeks (excludes non-delayed records)The Implemented version of the NAA tool

[0152] This section describes the NAA tool that is being implemented in the outpatient specialist clinic setting. Figure 7 shows a screen capture of the tool. The design of this interface remains in the development stage. The tool includes three main variables described above: unit,service type (i.e., appointment type), and care provider. After clinicians specify the unit, appointment type, and care provider, they select the timeframe in weeks or months. Then, they specify the recommended Review period for the patient and run the tool by clicking on the “Calculate Next Available” button. The NAA tool, using these inputs and the simulated queuing tables already created, shows the availability of spots for potential appointments. As shown in Figure 10, the tool either confirms availability in the requested time frame, as shown in (a), or displays the timeframe that is not available and provides the next available timeframe, as shown in (b). Finally, to improve and validate the efficiency of the tool, a feedback form is provided for the user to select the action taken after running the tool, shown in (c).

[0153] In the context of outpatient clinic scheduling, most of the studies in the literature mainly focus on improving micro-level waiting times (i.e., the time that patients waited in the clinic on the day of their appointments to be called by a clinician for the treatment) which is generally a matter of minutes. However, improving macro-level waiting times (i.e., the time patients waited between their need for treatment (i.e., referral date or recall date) and the date of their actual appointment) which is a matter of days / weeks / months and sometimes years is more important and significant for patients. There have been multiple efforts on improving the waiting lists at the macro-level. An appointment scheduling system based on overbooking policies was developed in a way that benefits of serving additional patients (e.g., generated revenue) could balance the possible extra costs of staff overtime and patient waiting times. Although initial results indicated the efficiency of this method, the variety of overbooking patterns throughout different clinics makes it impossible to provide a general overbooking policy that works for all clinics. Advanced access is another approach focusing on providing faster access for patients with urgent needs by keeping open about 50 percent of appointment slots each day and using them to offer a same-day appointment to all patients who call. This model was implemented in a Chicago primary care network and showed a significant improvement in patient satisfaction. However, it provides multiple challenges such as increasing the waiting times for the patients on the waiting lists and also wasting some of the available daily slots as a result of limited urgent patients. STAT (Specific Timely Appointments for Triage) is another model which focused on timely appointments for new patients (i.e., patients added to the waiting list using a new referral). STAT calculates how many new appointments are required each week to accommodate future demand. Althoughanalysing the trial results showed a reduction in waiting times for community rehabilitation services, it may provide multiple challenges including longer waiting times for follow-up appointments. According to the authors’ knowledge of the literature, the developed tool in this example is one of a kind, which provides informed decision-making capability using near-realtime analytics to simulate the next available Review appointments considering the current waiting list load.

[0154] The proposed tool assists the medical staff in several ways: 1) Assist the clinician during a consultation to provide accurate appointment information to the patient (e.g., setting realistic expectations with a patient on their next appointment), 2) Allow the clinician to manage the risk involved in a patient’s appointment being delayed. In many appointments, there is an acceptable period of delay, and the consulting clinician is best placed to determine if it is a suitable decision for a particular patient, 3) Increase clinician accountability for appointment delays related to waiting list volume, including the option to discharge patients to primary care rather than requesting further Review appointments when medically appropriate.

[0155] Same as other studies of this type, there are limitations to this example. First, the data used in this example was extracted from one hospital in a particular demographic area, thus it may lead to a distinct pattern and different outcomes using data from different hospitals. Second, each data set used for the example had some missing data that were fdtered out from the analysis. However, using other approaches to addressing missing data may lead to different results. Third, the NAA tool is based on weekly simulated queues, and it is not focussed on dynamics within the day of the appointment or administrative details around the day / time of appointments in a week. However, a daily simulation may lead to a small improvement in the tool’s results but was not appropriate for the descriptions.

[0156] Several aspects can be considered for future work. First, in the existing model, it is assumed that all patients waiting on the waiting list with a given recall date led to having an appointment. However, predicting a no-show behaviour for the records already on the waiting list may increase the accuracy of the tool. Second, the extrapolated clinician supply uses an exponentially weighted moving average approach, however, the data can be further analysed to investigate any “trend” or “seasonality”.

[0157] Third, the NAA uses a simulated queuing approach to find the next available appointment timeframe, however other techniques (e.g., machine learning) can also be applied.

[0158] This example develops a novel tool, called Next Available Appointment (NAA), to provide waiting list visibility to outpatient specialist clinicians to assist with scheduling Review appointments. The NAA uses advanced analytics to compare the existing waiting list and clinician capacity to provide an estimate of the next available Review appointment. The tool is evaluated using five years of historical waiting list information across several operational scenarios, ranging from hypothetical best-case scenarios to more practical scenarios. The practical scenario, which was chosen to be implemented within specialist clinics, achieved a simulated reduction in overdue appointments from 41% to 26% (i.e., a reduction of 47,000 overdue appointments over a three-year validation period).

[0159] Further examples of the processes for scheduling appointment will now be described in further detail with reference to a specific example performed on real appointment data.

[0160] Every year there are approximately three million new specialist clinic appointments at local hospital networks in Victoria. Two algorithms are developed to assist with waitlist management in their outpatient specialist clinics. The following describes the implementation of these algorithms in software tools developed to support their use and trial in the clinical setting at a local hospital. The following descriptions discuss the system design and development of both these software tools. The implemented workflow of the tools is reviewed and how these tools seek to improve current systems.

[0161] Health is the second largest area of government expenditure and is now Australia’s largest employer. In Australia, specialist outpatient clinics, run by public hospitals, provide a health service to patients who are waiting to be assessed by a medical / surgical specialist. Across Australia, in the year 2020-21, there were 46.8 million non-admitted patient care service events provided in public hospitals, with demand increasing from the previous year by 23%. The growing demand for healthcare services is creating pressure on the sustainability of the government-funded healthcare system. To be sustainable, it is needed to deliver healthcare services efficiently.

[0162] In this example, the selected hospital is a public teaching hospital in Victoria, Australia. The Specialist Clinics department in the hospital facilitates over 265,000 appointments per annum, across 57 Medical / Surgical specialities, 5 geographical locations (across 2 separate campuses) and approximately 1,500 different clinics to manage the complexities of running a tertiary referral centre with state-wide service obligations. The Specialist Clinics department is made up of 85 EFT of staff (clerical and nursing) who manage the variation in different units daily.

[0163] Currently, many units have high volume waiting lists causing appointment delays in some specialities, with patients sometimes experiencing >18-month delay for an appointment. Also, one of the Department of Health and Human Services (DHHS) Key Performance Indicators for Victorian hospitals is “90% of Urgent appointments seen within 30 days”. However, the magnitude of the intervention involved is variable given the demand of some units from a community referral base and the hospital's current performance for Urgent appointments seen within 30 days is <50%. Whilst there has been considerable training and education in queue management theory / capacity management principles, and various other efforts to increase efficiencies, compliance with DHHS KPIs remains challenging. These delays can lead to clinical risk for patients due to worsening conditions, staff stress and longer waiting times for patients due to re-scheduling required.

[0164] Two solutions are developed, a “Next Available Appointment” (NAA) and a “Template Planning Tool” (TPT) to support improved clinic scheduling and management. The TPT uses a collection of models and the current waitlist to provide a prediction capability for specialist clinic waitlist management. It also provides the ability for an end user to investigate “what if’ scenarios up to two years into the future by varying key model parameters. This tool comprehensively captures the flow of patients to accurately capture the real demand met by clinicians. The NAA tool provides a method to predict a suitable date for a patient’s next appointment on / after a date requested by the clinician. This tool provides informed decisionmaking capability to determine, in real-time, the next available appointment at the time a specialist makes a booking for a patient.

[0165] The following describes the subsequent co-design and development of implementations of both these tools for a trial at the hospital.Methodology & System Design

[0166] The implemented system (as shown in Figure 11 for the NAA tool) was designed in a modular way with three main layers:1. Pre-processing step where data is gathered, cleaned and collated.2. Modelling step using Monte Carlo simulations3. User interface where a user can enter configurable parameters

[0167] The system was designed to be run in standalone mode on a virtual machine (VM) using Docker containers (Figure 12). There are three components of the system which manage the user interaction, the running of the simulation and the display of results:1. A User Interface (UI) has been developed in Vue (i.e., an open-source JavaScript framework used for building front-end user interfaces) and provides a user management system. A user can self-register allowing them access to the tool. The user has the option of running either tool. The user can then select the required services and clinicians and, for the TPT tool, adjust a range of “what-if’ parameters. For the NAA tool, the results are displayed immediately, however, a downloadable report is generated for a TPT one.2. A backend server has been developed in NodeJs Express which is a web application framework that provides a REST API (i.e., an interface that two computer systems use to exchange information securely over the internet) for communication between the UI, the database, and the simulation processes. Data, e.g., registered users, dropdown lists, request parameters and responses, is managed in a local SQLite database.3. For Simulation Processing, simulation code has been developed in R and is managed in a Docker container which runs an R API server. Requests for simulations are passed to the R API server and trigger the R code. TPT Simulation results are presented in a PDF report which is uploaded to the database and madeaccessible to the user via the UI. The NAA results are displayed directly to the user in near-real time.Implemented Workflow

[0168] Both NAA and TPT solutions have been implemented in the outpatient specialist clinic setting of the hospital system. For the NAA tool, clinicians specify the unit, appointment type, and care provider and then select the timeframe in weeks or months. Then, they specify the recommended review period for the patient and run the tool. The NAA tool, using these inputs and the simulated queuing, shows the availability of a spot for a potential appointment. If the next appointment can be accommodated in the given timeframe, NAA confirms the availability, otherwise displays the timeframe is not available and provides the next available timeframe (Figure 13). In either scenario, a feedback form is provided forthe userto report the action they will likely take after running the tool. This form helps to understand how the timeframe predicted by the tool affects the decision of clinicians for the next appointment.

[0169] For the TPT tool, after the user selects the clinic, appointment type, and care provider, they can examine different scenarios by varying key model parameters such as changing the ratio of weekly discharge or the inflow of patients (Figure 14). Also, the tool allows them to investigate the effect of overbooked appointments on the clinic flow.

[0170] The output of the simulation is given in the form of plots and tables which summarise analytics representing the main measures of the selected clinic two years into the future (Figure 15). The output includes 4 plots which are representing the simulated / predicted values for the next two years, on a weekly basis (i.e., 104 values each). The upper-left panel shows the weekly number of registrations to the waitlist. In this plot, the distribution of registration sources, either new or review, is also presented. The upper-right panel shows the weekly number of appointments. The lower-left panel shows analogous results forthe weekly waitlist net change. That is the difference between the number added to the waitlist and the number removed from the waitlist, for each week.

[0171] The lower-right panel shows the number of post-appointment discharges, defined as the number of appointments with no later scheduled appointments or waitlist registrations. Post-appointment discharges have been included as an indication of clinician-initiated discharges, as opposed to “administrative” discharges (e.g., referral expiry, automated waitlist clean-up). The plots are summarized in the upper table. The analytical results for each waitlist priority (review, routine, urgent) are shown in the lower table. This table shows the total records for the next two years for each priority and the percentage of overdue appointments.

[0172] The Next Available Appointment (NAA) and Template Planning Tool (TPT) are webbased software applications that provide visual prompts using validity algorithm / s to provide clinicians with upfront information about the ability to manage the risk associated with delays and provide additional insight for planning and scheduling of outpatient appointments at the hospital, whilst continuing to improve the patient experience. TPT can simulate patient flow 2 years into the future based on user-selected parameters to support clinic management. The implemented solutions are transferable to other health jurisdictions to manage specialist clinic waitlist demand and improve performance against key performance metrics. Outputs from these solutions lay the foundation for evidence-based discussions with hospital decisionmakers for improving access to timely care in outpatient specialist clinics.

[0173] Throughout this specification and claims which follow, unless the context requires otherwise, the word “comprise”, and variations such as “comprises” or “comprising”, will be understood to imply the inclusion of a stated integer or group of integers or steps but not the exclusion of any other integer or group of integers. As used herein and unless otherwise stated, the term "approximately" means ±20%.

[0174] Persons skilled in the art will appreciate that numerous variations and modifications will become apparent. All such variations and modifications which become apparent to persons skilled in the art, should be considered to fall within the spirit and scope that the invention broadly appearing before described.

Claims

THE CLAIMS DEFINING THE INVENTION ARE AS FOLLOWS:1) A method of scheduling appointments, the method including, in one or more electronic processing devices: a) accessing historical data relating to past appointments; b) accessing supply data indicative of a clinician availability; c) calculating a predicted clinician availability using the supply data and the historical data, wherein the predicted clinician availability is calculated to correct for overestimation of clinician availability; d) generating a patient flow using the predicted clinician availability, wherein the patient flow is indicative of a queue of patients for future appointments; and, e) at least one of: i) scheduling a future appointment using the patient flow; and, ii) updating the patient flow to add patients requiring future appointments.2) A method according to claim 1, wherein the method includes, in one or more electronic processing devices: a) receiving an appointment request indicative of a clinician and a recall time; and, b) at least one of: i) scheduling the future appointment using the appointment request; and, ii) updating the patient flow using the appointment request.3) A method according to claim 1 or 2, wherein when updating the patient flow, the method includes, in one or more electronic processing devices, providing a queue length of the updated patient flow.4) A method according to any one of claims 1 to 3, wherein the step of generating the patient flow includes: a) generating an initial patient flow using the clinician availability; and, b) modifying the initial patient flow to generate the patient flow using the predicted clinician availability.5) A method according to any one of claims 1 to 4, wherein the historical data is indicative of at least one of: a) appointment data in a previous time period; and, b) appointment data indicative of scheduled appointments.6) A method according to claim 5, wherein the previous time period is at least one of: a) 3 months; b) 6 months; c) 12 months; d) 18 months; and e) 24 months.7) A method according to any one of claims 1 to 6, wherein the supply data is indicative of at least one of: a) a number of available sessions in a week for each clinician, unit and service type combination; b) information of each session within a week; and, c) a number of patients booked for each session.8) A method according to any one of claims 1 to 7, wherein the appointment request is indicative of at least one of: a) a unit; b) a care provider; and, c) a service type.9) A method according to any one of claims 1 to 8, wherein the step of generating the patient flow includes generating the patient flow using a model equationQk = max^Q^ + xk- Ck, 0); where Qkis the queue length at time k = 1,2, ...; xkis the number of patient appointments at week k:Ckis the number of patients removed from the queue at week k: and<2fc-i is the backlog which represents the patient appointments with a recall time before week k.10) A method according to any one of the claims 1 to 9, wherein the step of calculating the predicted clinician availability, in one or more electronic processing devices, includes: a) extrapolating the supply data for a future time period based on the historical data; b) calculating an open factor fraction of the clinician availability; and, c) applying the open factor fraction to thereby calculate the predicted clinician availability.11) A method according to claim 10, wherein the future time period is at least one of: a) 3 months; b) 6 months; c) 12 months; d) 18 months; and e) 24 months.12) A method according to claim 10 or 11, wherein the step of calculating the open factor fraction, in one or more electronic processing devices, includes at least one of: a) calculating an actual available session divided by a planned available session based on the historical data; and, b) calculating using an exponentially weighted moving average with0 Pnew_predict OPnewjcctual + (1 - a)0Poldpredict where OPnewj]redictis a new prediction of an open factor fraction;OPnew _actuai is the actual open factor fraction for the most recent week;OP0id_predict is the previous week’s value of the open factor fraction; and a = 0.01 .13)A method according to any one of the claims 1 to 12, wherein the step of scheduling the future appointment using the patient flow and the appointment request, in one or more electronic processing devices, includes: a) determining an appointment availability and patients to be scheduled using the patient flow; b) determining a recall time of the patients to be scheduled; and, c) scheduling the appointments using the recall time.14) A method according to claim 13, wherein the method includes scheduling patients with a most overdue recall time.15)A method according to any one of the claims 1 to 14, wherein the step of scheduling the future appointment using the patient flow and the appointment request, in one or more electronic processing devices, includes: a) determining an appointment availability of the recall time using the patient flow; b) if the appointment availability indicates available, scheduling the future appointment of the recall time; and,c) if the appointment availability indicates not available, repeating step (a) for a time interval of the recall time.16) A method according to claim 15, wherein the time interval is at least one of: a) ± 1 days; b) ± 3 days; c) ± 5 days; d) ± 1 week; and, e) selected depending on the appointment request.17) A method according to any one of the claims 1 to 16, wherein the method further includes, in one or more processing devices, updating the historical data to reflect the scheduled future appointment.18) A method according to claim 17, wherein the step of updating the historical data is performed at least one of: a) as soon as the future appoint is scheduled; b) daily; c) weekly; d) fortnightly; and, e) monthly.19) A system of scheduling appointments, the system including: a) a database containing historical data and supply data; and, b) one or more electronic processing devices being configured to: i) access the historical data relating to past appointments from the database; ii) access the supply data indicative of a clinician availability from the database; iii) calculate a predicted clinician availability using the supply data and the historical data, wherein the predicted clinician availability is calculated to correct for overestimation of clinician availability; iv) generate a patient flow using the predicted clinician availability, wherein the patient flow is indicative of a queue of patients for future appointments; and, v) at least one of:(1) schedule a future appointment using the patient flow; and,(2) update the patient flow to add patients requiring future appointments.20) A method according to claim 19, wherein the one or more electronic processing devices is configured to: a) receive an appointment request indicative of a clinician and a recall time; and, b) at least one of: i) schedule the future appointment using the appointment request; and, ii) update the patient flow using the appointment request.21) A method according to claim 19 or 20, wherein when updating the patient flow, the one or more electronic processing devices is configured to provide a queue length of the updated patient flow.22) A system according to any one of claims 19 to 21, wherein the one or more electronic processing devices, at the step of generating the patient flow, is configured to: a) generate an initial patient flow using the clinician availability; and, b) modify the initial patient flow to generate the patient flow using the predicted clinician availability.23)A system according to any one of claims 19 to 22, wherein the historical data is indicative of at least one of: a) appointment data in a previous time period; and, b) appointment data indicative of scheduled appointments.24) A system according to claim 23, wherein the previous time period is at least one of: a) 3 months; b) 6 months; c) 12 months; d) 18 months; and e) 24 months.25)A system according to any one of claims 19 to 24, wherein the supply data is indicative of at least one of: a) a number of available sessions in a week for each clinician, unit and service type combination; b) information of each session within a week; and, c) a number of patients booked for each session.26) A system according to any one of claims 19 to 25, wherein the appointment request is indicative of at least one of: a) a unit; b) a care provider; and, c) a service type.27) A system according to any one of claims 19 to 26, wherein the one or more electronic processing devices, at the step of generating the patient flow, is configured to generate the patient flow using a model equationQk = max Qk^ + xk- Ck, 0); where Qkis the queue length at time k = 1,2, ...; xkis the number of patient appointments at week k:Ckis the number of patients removed from the queue at week k: and is the backlog which represents the patient appointments with a recall time before week k.28) A system according to any one of the claims 19 to 27, wherein the one or more electronic processing devices, at the step of calculating the predicted clinician availability, is configured to: a) extrapolate the supply data for a future time period based on the historical data; b) calculate an open factor fraction of the clinician availability; and, c) apply the open factor fraction to thereby calculate the predicted clinician availability.29) A system according to claim 28, wherein the future time period is at least one of: a) 3 months; b) 6 months; c) 12 months; d) 18 months; and e) 24 months.30)A system according to claim 28 or 29, wherein the one or more electronic processing devices, at the step of calculating the open factor fraction, is configure to at least one of: a) calculate an actual available session divided by a planned available session based on the historical data; and, b) calculate using an exponentially weighted moving average withPnew_predict OPnewjictual + (1 — a) OPoidpredict where 0Pnewj]redictis a new prediction of an open factor fraction;OPnew _actuai is the actual open factor fraction for the most recent week;OP0id_predict is the previous week’s value of the open factor fraction; and a = 0.01 .31) A system according to any one of the claims 19 to 30, wherein the one or more electronic processing devices, at the step of scheduling the future appointment using the patient flow and the appointment request, in one or more electronic processing devices, includes: a) determining an appointment availability and patients to be scheduled using the patient flow; b) determining a recall time of the patients to be scheduled; and, c) scheduling the appointments using the recall time.32) A system according to claim 31, wherein the one or more electronic processing devices is configured to schedule patients with a most overdue recall time.33)A system according to any one of the claims 19 to 32, wherein the one or more electronic processing devices, at the step of scheduling the future appointment using the patient flow and the appointment request, is configured to: a) determine an appointment availability of the recall time using the patient flow; b) if the appointment availability indicates available, schedule the future appointment of the recall time; and, c) if the appointment availability indicates not available, repeat step (a) for a time interval of the recall time.34) A system according to claim 33, wherein the time interval is at least one of: a) ± 1 days; b) ± 3 days; c) ± 5 days; d) ± 1 week; and, e) selected depending on the appointment request.35)A system according to any one of the claims 19 to 34, wherein the one or more electronic processing devices is configure to update the historical data to reflect the scheduled future appointment.36) A system according to claim 35, wherein the one or more electronic processing devices is configure to update the historical data at least one of: a) as soon as the future appoint is scheduled; b) daily; c) weekly; d) fortnightly; and, e) monthly

Images