TAC Claims Management Transformation

Transcription

1 TAC Claims Management Transformation Prepared by Natalie Pocock, Sophie Holdenson and David Gifford Presented to the Institute of Actuaries of Australia Accident Compensation Seminar November 2011 Brisbane This paper has been prepared for the Institute of Actuaries of Australia s (Institute) 2011 Accident Compensation Seminar. The Institute Council wishes it to be understood that opinions put forward herein are not necessarily those of the Institute and the Council is not responsible for those opinions. Transport Accident Commission The Institute will ensure that all reproductions of the paper acknowledge the Author/s as the author/s, and include the above copyright statement: The Institute of Actuaries of Australia Level 7 Challis House 4 Martin Place Sydney NSW Australia 2000 Telephone: +61 (0) Facsimile: +61 (0) actuaries@actuaries.asn.au Website:

2 Table of Contents Introduction... 3 The Recovery Model... 4 Vision and Objective... 4 Visual Representation of Model... 4 What do we mean by a claims management model?... 4 Previous Model... 5 Why did TAC change its claims management model?... 6 Designing, building and implementing the Recovery Model... 8 People... 8 Pilots... 8 Change Management plan... 9 Key Changes in Detail New Teams New Tools and Processes Segmentation Segmentation Algorithm Manual Business Rules Results to date Impact of Recovery on Scheme Objectives Recovery Model Recovery Segmentation Model What s next? Conclusion Acknowledgements References... 30

3 Introduction The Transport Accident Commission supports Victorians injured in transport accidents by paying for appropriate medical treatment and income support. Approximately 16,000 claims are lodged each year at the TAC, with approximately 30,000 active at any time. Minor to moderately injured clients make up 97% of the TAC s overall annual claim numbers, resulting in $550M in annual payments, about 65% of the total for the TAC and $2.5B in claims liabilities, about 30% of the total for the TAC (including common law damages and costs). The remaining 3% of claim numbers, 35% of payments and 70% of liabilities comprises the TAC s severely and catastrophically injured clients. In mid 2009 the TAC Board approved the TAC 2015 strategy and in doing so introduced client outcomes as a third strategic objective in addition to scheme viability and client satisfaction. As part of this six year strategy, in October 2010 the TAC implemented Recovery, a new claims management model for minor to moderately injured clients. This new model aims to improve performance measured using all three objectives client outcomes, client satisfaction and scheme viability, with ambitious targets relating to each of these objectives. The previous Benefit Delivery approach to managing the minor to moderately injured clients did not produce the desired results after five years of operation and in particular never achieved an actuarial release (i.e. liability reduction) (Crossett, 2010). The new model was created following extensive national and international research conducted to identify best practices within the personal injury industry (TAC, 2009). Leading personal injury schemes around the world, such as SUVA in Sweden and the Accident Compensation Commission (ACC) in New Zealand, are moving towards managing claims according to complexity, using early intervention and triage tools (Crossett, 2010). Similarly the Recovery model focuses on managing claims based on risk of high cost, which is made up of a number of factors rather than just focusing on one factor, such as injury type, alone. This paper examines creation of the Recovery model at TAC, including the purpose of the model, how it was implemented, and key differences from the previous model. One particular aspect of the new model, auto segmentation via an algorithm, is examined in particular detail.

4 The Recovery Model Vision and Objective The vision of the Recovery model is to support clients to return to work and health, as quickly as possible. The objective of the Recovery model is to get the right information up front so that clients can be directed to the team best equipped to assist them with their individual needs and support them in achieving their health and return to work goals. Visual Representation of Model The model is represented visually below. It is also explained in more detail in following sections. Client ELIGIBILITY EARLY SUPPORT RETURN TO HEALTH INCOME RETURN TO WORK CLIENT ASSIST CLIENT REVIEW What do we mean by a claims management model? In this context the claims management model refers to the combination of A) The organisational structure (i.e. teams and roles) within which claims are managed; B) The rules and processes (both automated and manual) via which claims move within the structure and the resulting portfolio sizes and claim flows; and C) The strategies, work practices and interventions employed by staff in managing claims given the structure and claim movements under A and B.

5 While making some reference to the changes made under C, this paper primarily focuses on A and B. Previous Model The TAC s previous claims model for minor to moderately injured clients was referred to as Benefit Delivery. Under this model teams were structured according to clients return to work needs; injury type and length of hospital stay. Importantly, no psycho social, demographic or common law components were considered. A visual representation of the Benefit Delivery model is as follows: Minor (Benefit delivery client support) Moderate (Benefit delivery risk teams) Vocational (income support required) Long hospital Early RTW Capacity Standard intervention (10 mth+) Over 50% Quick recovery team Orthopaedic Soft tissue Early RTW Capacity Standard intervention (10 mth+ Capacity RTW Standard Income team Non-vocational (no income support required) Maintenance only Long hospital Early intervention Standard Long tail Other (East) Early intervention Standard Long tail Other (West) Early intervention Standard Long tail Pharmacy team The claims management teams within Benefit Delivery were as follows: Active Management Teams Vocational teams (Soft Tissue and Fractures) rehabilitation coordinators who managed clients with entitlement to income and vocational needs. These teams were delineated by injury type; Vocational, Long Hospital rehabilitation coordinators who managed clients with entitlement to income and vocational needs that had a hospital stay longer than 16 nights and a demonstrable injury;

6 Non Vocational teams (East and West) rehabilitation coordinators who managed clients with no income entitlement and less severe injuries. The teams were delineated by the clients location within Victoria; and Non Vocational, Long Hospital rehabilitation coordinators who managed clients with no entitlement to income, that had a hospital stay longer than 16 nights and a demonstrable injury Support Teams Client Support entry level claims officers who had a dual role in determining initial eligibility of the claim as well as managing low risk and emergency expenses only claims Income entry level claims officers and income assessors who determined the income rate and processed income payments for clients Maintenance entry level claims officers managing low risk, long tail clients Within each team, claims typically moved between portfolios based on life of claim (i.e. duration since accident). The typical life of a claim within Benefit Delivery was as follows: Commence in Client Support where eligibility would be determined and the claim would be initially managed. Transitioned to an active management team based on manual business rules, depending on the type of service requested and/or a request for income. Once within an active management team, the claim would move three more times; firstly to an early intervention portfolio, then to a 10 month plus portfolio and finally to a standard portfolio awaiting the three years of income entitlement to cease. The client would then eventually move to the Maintenance team which would manage any ongoing paramedical needs. This resulted in claims moving through approximately five teams in their first three years with the TAC. When combined with the impact of claims manager attrition and extended leave, a typical active management client within Benefit Delivery would often have 6 8 claims managers or more. This significant number of transitions is well established as a source of dissatisfaction among TAC clients. Why did TAC change its claims management model? There were a number of drivers for the change from Benefit Delivery to Recovery : The introduction of client outcomes as TAC s third corporate strategic objective in 2009 placed a greater emphasis on Recovery as an objective, and necessitated a claims management model better designed to facilitate client recovery

7 The jointly run TAC and WorkSafe Ambassador program had recently derived learnings from international schemes about proactively managing clients by their needs and non injury risk factors, rather than just by injury type and this aligned well with a client outcomes model. Under Benefit Delivery client needs and risk factors other than injury, were not always well identified or managed. Clients with high risk factors such as mental health issues and common law were spread across multiple teams and portfolios, and could even be in low risk teams. Portfolio sizes and staff role types were also not well aligned to client needs. This spread of risk claims also meant that there were missed opportunities for early intervention. There was little if any focus on managing common law risk. As discussed the model involved multiple file handovers through the life of a claim with these handovers being triggered by duration since accident rather than any significant event in the life of a claim. The high number of claim movements resulted in a high administrative burden for senior staff. Other work practices within the model were also inefficient, leaving staff unable to be proactive.

8 Designing, building and implementing the Recovery Model The process of designing the Recovery model began in 2008 (more than two years prior to implementation), comprising research and analysis of best practice claims management models around the world such as ACC (New Zealand) and suva (Switzerland). This research continued in 2009, also incorporating in house knowledge and experience from TAC's own staff. Key features of the design of the new model included: Use of various risk factors (rather than primarily injury type) to assign claims to teams, in particular a greater focus on the management of common law risk The concept of an automated segmentation algorithm to assign claims to team according to the relevant risk factors Process changes to improve efficiency and to aid in early identification of clients at risk of poor outcomes An initial view of the number of claims and portfolio size for each team type The business case to proceed with the Recovery model was approved by the TAC Board in June 2009, with implementation scheduled for October There were a number of steps involved with building the new model. People Development of the Recovery model saw many different areas of the TAC working together, including: The Benefit Delivery management team (the business owners of the change) Benefit Delivery team managers and staff, acting as subject matter experts IT to build and test the necessary system changes Business Intelligence to build the segmentation algorithm, new file review processes, and to provide other analytical support Human Resources to help manage the required cultural change The overall project was managed by dedicated project resources. Pilots In order to test a number of the new processes and components of the Recovery model, a number of pilots were conducted. Return to Work pilot a team of six portfolios was created with portfolios similar to those anticipated for the Return to Work Complex teams in the Recovery model Client Review Team a group was established with a sole focus on reviewing tail claims (i.e. those more than two three years post injury) Client Conversational Tool (a screening tool used to identify clients with high needs in relation to return to work, mental health and/or persistent pain) the tool was used on new claims in suitable teams under the existing Benefit Delivery model.

9 The learnings from these pilots were instrumental in preparing for implementation and used as lead indicators to the model s success. Change Management plan In order for the Recovery model to meet its objectives, a change management plan was required. With 150 staff having their roles directly affected by the change of model it was important that they were taken along on the journey. The change management plan began in January 2010 (ten months prior to implementation) and focussed on the behaviour and mind set changes required. There was a key focus on being more proactive and working in a more client focussed way. There were several components to the plan: In February 2010 a Recovery Expo was run by the Benefit Delivery leadership team to introduce all TAC staff to the new teams and roles as well as the principles of the new model. This was a key factor in increasing staff understanding and buy in to the new model, before they were to be realigned to twelve new teams and in some cases new roles. There was also a significant focus on supporting the existing Benefit Delivery leadership group to lead change within their teams. Workshops were conducted with team leaders prior to and following implementation to develop a consistent understanding of the model and the required mindsets and behaviours. These enabled the leadership group to role model and provide support to their teams. Two months prior to implementation cultural change workshops were run for staff members aligned to each new team, with a focus on the mindset and behaviours required to achieve their team s vision. Change Ambassadors from each team were nominated with these staff members playing an integral role in influencing and supporting their peers through the change. Post implementation various activities have continued with numerous feedback sessions conducted giving staff the opportunity to share their views on the strengths and weaknesses of the new model. A number of the suggestions made by staff have resulted in refinements to the model, and in other cases further education sessions have been held to help provide staff with a deeper understanding of the new model. The TAC were one of three finalists in the Excellence in Personal Injury awards, run by the Personal Injury Education Foundation (PIEF) for the Change Management plan for Recovery.

10 Key Changes in Detail New Teams The Recovery model teams are broadly divided into two streams: Active Management managing risk claims with a return to work focus Pathways managing risk claims with a return to health focus, all low risk claims and the eligibility and income processes Within these teams portfolio sizes are differentiated according to client complexity and needs. Further detail about the new teams is provided below: Active Management teams Early Support Manages clients with return to work needs or an entitlement to income within six months of accident The key objective of this team is to return clients to work as early as possible Team comprises highly skilled rehabilitation coordinators with a health or strong claims management background Portfolio size of 40 clients Return to Work Complex & Less Complex Manages clients with ongoing return to work needs Two complex teams manage claims with greater risk, and two less complex teams manage claims with lower risk Teams comprise highly skilled rehabilitation coordinators with a health or strong claims management background Portfolio sizes of 50 clients in RTW Complex and 100 clients in RTW Less Complex Pathways teams Return to Health Manage clients with ongoing return to health needs who exhibit significant risk factors Team comprises highly skilled rehabilitation coordinators with a health or strong claims management background Portfolio sizes of 120 clients Client Assist Team Recovery s largest team, managing clients with low needs and low risk with an intention to fast track entitlements and provide excellent customer service Claims advisors with customer service backgrounds manage these clients Portfolio size of 1000 clients

11 Client Review Team Reviews long tail claims and outlier providers from the low risk Client Assist team. Team does not manage claims as such but reviews the reasonableness of ongoing benefits (primarily medical and paramedical) Team comprises rehabilitation coordinators with a health or strong claims management background New Tools and Processes Client Conversational Tool A series of questions to identify high needs in relation to return to work, mental health and/or persistent pain Derived from two existing tools: the Rehab Progress Checklist to predict persistent pain (from ACC in NZ) and the Trauma Screening Questionnaire to predict psychological injury (from the UK) Was initially run as a pilot prior to implementation New tools and processes to manage Common Law (CL) risk Automated identification of claims with high risk of Common Law based on data available at intake CL flag included on claims management software to increase visibility of CL risk Shared KPIs between the Recovery and Lump Sum (who manage Common Law benefits) Stronger reliance on proactive data analytics to identify trends Claims analytically selected for review by Client Review, Return to Health and Return to Work teams using a variety of triggers based on cost, age of claim and recent activity. These triggers are able to be adjusted to address current areas of performance concern. Segmentation Algorithm used to automatically allocate claims to the most appropriate team at acceptance Revised manual business rules for moving claims as client circumstances change More detail is provided in the following section

12 Segmentation One of the key differences between the new Recovery model and the previous Benefit Delivery model is a new approach to allocating claims to teams, referred to as Segmentation. Segmentation has two main components: The Segmentation algorithm this is a logistic regression model which delivers automated allocation of claims to the most appropriate team at acceptance based on data available at that time Manual business rules enabling movement of claims to a more appropriate team as client circumstances change over time Segmentation Algorithm TAC s initial research identified five primary complexity factors that contribute to risk of high cost: 1. Injury type (TAC, 2009); 2. Common law risk (TAC, 2009); 3. Psychological wellbeing (Victorian WorkCover Authority, 2006; Australian Safety and Compensation Council, 2009; LaMontagne, Louie and Ostry, 2006); 4. Presence of persistent pain (Waddell, 2002); and 5. Employment status (TAC, 2009). To develop an automated data driven process at claim acceptance it was necessary that the factors used in the model be available at claim acceptance. Of the primary complexity factors listed, versions of the first, second and fifth are available at acceptance, while psychological wellbeing and persistent pain are hard if not impossible to identify at acceptance. A model was therefore developed based on information available at acceptance (i.e. information collected in initial interview and on the claim form). Overall Model Structure The overall model used to allocate claims to teams comprises two sub models One which predicts the probability of a claim being high cost (defined as receiving a no fault payment in the seventh month post accident); and One which predicts the probability of a claim ultimately lodging a serious injury (common law) application. The models were developed based on five years of historical data (approximately 80,000 claims).

13 High Cost Model The target variable within this model was the probability of a claim being high cost as measured by the receipt of a no fault payment between 180 and 210 days post acceptance. Various other definitions of high cost, using different time periods post acceptance, or cost thresholds were tested before arriving at the definition adopted. The factors found to be significant within the high cost model exhibited some similarities with those indicated by the initial research, in particular injury category and employment status. In total there are seventeen significant variables used in the model. The model also revealed some surprising findings. In particular it was expected that the presence of pre existing conditions and/or treatments, such as heart conditions, neck pain, or previous physiotherapy treatment, would tend to increase the risk of a claim becoming high cost. In fact, the opposite was the case. Anecdotally it is suspected that this does not actually arise from clients with (say) a pre existing heart condition necessarily recovering from an accident more quickly than someone without a heart condition. Rather it suggests that clients who declare such conditions and treatments when filling in their claim form are more likely to be motivated to return to work thus enabling them to achieve earlier outcomes. Common Law model The target variable within this model was the probability of a claim ultimately lodging a serious injury (common law) application at any point in its life. The significant factors for the common law model had similarities with those in the high cost model and with the initial research. Interestingly the existence of psychological impairment, which was present in the initial research was not significant in the high cost model but was in the common law model. In total there are nine significant variables used in the model. Perhaps not surprisingly, the most significant factor by far within the common law model was whether or not the client was at fault in the accident. Claims will only be successful at common law when it is demonstrated that there is an at fault driver other than the injured client. An initial model built excluding the at fault variable was considerably weaker than that including the at fault variable. Combined model The high cost and common law models were combined to create the overall segmentation algorithm. This was done via a series of thresholds on the high cost and common law scores, and also via some overriding rules.

14 For example for claims assigned to Return to Work Complex, return to work claims are flagged based on whether the client has requested a loss of earnings review on the claim form, or whether they have indicated that while currently unemployed they do intend to return to work in the future. Following the return to work flag being assigned the probabilities determined via the two models are used to determine whether the claim is sufficiently complex to assign to the Return to Work Complex team. Not all claims requesting a loss of earnings review on the claim form will go to a return to work team, with claims with lower probabilities (based on other complexities) going to the higher volume Client Assist team. Many of these claims will either not receive income at all, or will only receive it for a short time, and so therefore do not need active management. The probability thresholds used to assign new claims to each team were based on the capacity of each team as well as expected flows out of the team each month. The use of these probability thresholds has the advantage of being relatively easy to adjust if it is determined that the number of claims assigned to particular teams needs to be changed. Prior to implementation, manual reviews were also undertaken by claims staff on claims being assigned to each team via the algorithm, as a check that the algorithm was appropriate, while recognising that the nature of a model such as this means that not all claims will end up in the correct team.

15 The Daily Process A visual representation of the process via which claims are allocated to their new team following acceptance is as follows: Day of acceptance Claim accepted by Eligibility team Claim automatically moves to To Be Segmented Portfolio Data extracted from the claims system into SAS Scheduled process runs in SAS Overnight following acceptance List produced of claims and their appropriate team List fed back into Claims Management System Claims moved to new team Day following acceptance Senior officers in teams move claims to a portfolio in their team Manual Business Rules In addition to the segmentation algorithm, which deals with the allocation of claims to their initial team following acceptance, manual business rules to move claims between teams are also required as clients circumstances change over time. In the Recovery model each team has a different set of business rules based on the claims within that team. The Benefit Delivery model (which did not have any automatic data driven segmentation) consisted of many manual business rules based on various events in the life of a claim, and these formed the starting point for the manual business rules under the Recovery Model.

16 As discussed above the primary rules under the Benefit Delivery model were time based (i.e. driven by duration since accident). The Recovery model specifically did not make use of such rules and these were therefore removed. While movements via these rules occur manually, the rules generally consist of claim events which were able to be tested using historical data. Each rule was tested by looking at claims where the relevant event occurred, and the cost of these claims post the event. Manual claim by claim reviews were also used to determine whether claims were likely to benefit from receiving active management post the event occurring. Under the Benefit Delivery model only numbers of claim movements were monitored, but with Recovery came the ability to also record and therefore monitor the reason for claim movements. These reasons are based on the business rules, and therefore allow TAC to see over time which business rules are relevant and working well. Number of claim movements The combination of the clear purpose of each new team, the automated segmentation algorithm, and the revised business rules are intended to significantly reduce the number of claim movements between portfolios. As discussed above this has been identified as a risk to client satisfaction. Ensuring claims are in the most appropriate team should also result in the most appropriate level of services being provided to the clients. For a client in the new Recovery model, the typical life of an active management claim would be Commence in Eligibility where the claim would be accepted and then automatically moved to the appropriate team via the algorithm. Move to Early Support where the claim would be managed for up to six months. If within this time return to work was achieved and no significant treatment remained, the claim would move to Client Assist. If at the end of six months the client still required ongoing return to work support the claim would move to one of the Return to Work teams. It would stay in this team until the client had achieved return to work, and then the claim could be moved to Client Assist for ongoing maintenance. This results in only two or three moves for this particular claim, compared to the five moves (or more) that the client would have experienced under the Benefit Delivery model. Supporting staff to work with the new model Due to the significant differences between the Recovery segmentation approach and the Benefit Delivery model it was important that staff were supported in understanding and effectively implementing the new model.

17 Training was delivered to the new teams pre implementation covering aspects such as: The significant factors in the segmentation algorithm Examples of the types of claims in each team The business rules applying to particular teams There were a number of challenging concepts for claims staff to come to terms with within this training: The combination of factors that impacts the algorithm score for a claim and therefore the team to which it is assigned. For example in the Benefit Delivery model if a claim had a relatively minor orthopaedic injury, it would automatically be assigned to an active management team. Under Recovery if this claim did not exhibit any other risk factors it would most likely be assigned to a lower risk team The distinction between similar concepts differently within the model: o Complexity, being claims at risk of becoming high cost and/or a common law lodgement, and o Activity, based on workload required to manage a claim. While there is a correlation between activity and complexity, some complex claims result in relatively low levels of activity and some claims with high activity exhibit low complexity. Some components of the training have been reinforced with further sessions post implementation. Feedback from claims staff regarding claims felt to be inappropriate for particular teams has also been considered, with some changes to the initial business rules being made. Transition of existing claims at implementation In order to implement the Recovery model existing claims as at implementation were re assigned to the new teams. This was done based on the factors identified in the initial research as well as a consideration of the new business rules. Each claim was assigned a rank, with the ranks being used to assign claims to teams. For example, a claim receiving income benefits, with recent psychological treatment and active common law would be given a rank of 2, which would see it moved to a Return to Work complex team. The allocation was tested by running the same allocation rules over active claims as at June 2008 (i.e. two years prior to implementation), then looking at the spread of high cost claims and common law lodgements between the active management teams and Client Assist (based on actual experience in the following two years). A further constraint was to minimise claim movements at implementation to the extent possible.

18 The table below shows the numbers of claims moved for Day 1 of Recovery. Slightly more than 50% of active claims at implementation moved portfolios. Claims moved at implementation Active Inactive Overall Changed Owner 13,013 55% 54,453 42% 67,466 44% Same Owner 10,824 45% 73,829 58% 84,653 56% Total claims 23, % 128, % 152, %

19 Results to date It is now just over twelve months since implementation of the Recovery model. While the ultimate benefits (particularly those relating to common law) will take some time to be realised, there are a number of early indicators in relation to the success of the model. Impact of Recovery on Scheme Objectives Scheme Viability Early results are positive with Recovery/Benefit Delivery receiving its first actuarial release in four years at June 2011 Client Satisfaction The most recent results for client satisfaction have shown significant improvement from February to August 2011, with a score of 7.1 in February 2011 and 7.4 in August 2011, with the most significant improvement in Client Assist. However, immediately post implementation there was a reduction in client satisfaction. This is likely to have related to the significant number of claims who transferred from one claims manager to another (despite attempts to minimise these transitions) as well as claims staff adjusting to the new model. Recovery Model Return To Work (RTW) Pilot Results from the RTW Pilot team which ran for six months prior to implementation give an early indication of the impact the Recovery Return to Work Complex teams can have. Results were encouraging in the 12 months after the pilot began, with the proportion of active income claims from this pilot group being 9% less than a control group of like claims. RTW Pilot - Comparison of Number of Income Active claims Start Month Post 12 Mths # Difference % Difference Control Group % RTW Pilot % Client Review Team Results from the first year of Client Review conducting reviews on analytically selected files show 24% of claims reviewed had a Cease, Reduction, Reduction with Cessation or Denial decision made. This is a higher rate of decision than was achieved within Benefit Delivery. Client Review - Outcome of completed analytically driven reviews Decision Type # % Denial, Cease, Reduction, Reduction with cessation 70 24% Other % Total %

20 Client Review Outcome of completed analytically driven reviews 60% 50% 40% 30% 20% 10% 0% Dec-10 Jan-11 Feb-11 Mar-11 Apr-11 May-11 Jun-11 Jul-11 Aug-11 Sep % Denial, Cease, Reduction, Reduction with cessation Files reviewed Client Conversational Tool Client conversational tool is used to identify clients at risk of pain, psychological or return to work issues. Results to June 2011 show pain to be the risk most often flagged as high need Client Conversational Tool Identified as Pain Psych RTW High Need % 98 6% 62 4% Low Need % 1,097 73% % Not Assessed % % % (blank) 76 5% 76 5% 76 5% Grand Total 1, % 1, % 1, % Client Conversational Tool High Needs Identified 25% % % % 100 5% 50 0% Oct-10 Nov-10 Dec-10 Jan-11 Feb-11 Mar-11 Apr-11 May-11 Jun-11 0 % High Need - Pain % High Need- Psych % High Need - RTW Number of CCT conducted

21 Accounts Processing Exceptions Accounts processing exceptions arise where an account triggers one or more rules which lead to the account requiring manual intervention by claims staff. In order for Client Assist to operate with large portfolio sizes the number of exceptions generated by claims in that team needed to be reduced. January to September 2011 has seen a 64% reduction in average exceptions per month relative to the previous year. Account Processing Exceptions Average exceptions per month % Reduction Oct09 - Sep10 11,577 Oct10 - Dec10 8,592 26% Jan11 -Sep11 4,138 64% Recovery Client Assist Account Processing Exceptions Jul-09 Aug-09 Sep-09 Oct-09 Nov-09 Dec-09 Jan-10 Feb-10 Mar-10 Apr-10 May-10 Jun-10 Jul-10 Aug-10 Sep-10 Oct-10 Nov-10 Dec-10 Jan-11 Feb-11 Mar-11 Apr-11 May-11 Jun-11 Jul-11 Aug-11 Sep-11 Service Limits Other Common Law Potential The new common law process as part of Recovery aims to see common law potential flags being applied earlier resulting in a greater focus earlier in claim life. Earlier flagging of Common Law potential allows for earlier intervention and a greater focus on return to work. There has been some increase in potential flags being applied within 6 months compared to the same time last year, however it is expected that further increases are possible. Specifically, there has been a 12% increase in the number of claims flagged as having Common Law potential (CLP) within six months of accident.

22 Timing of Common Law Potential flag % of CLP flagged in month that was flagged within 6 months of accident Number of CLP flagged that were flagged within 6 months of accident Common Law Potential Raised Number of CLP flagged - total Oct09-Aug10 33% 895 2,701 Oct10-Aug11 37% 1,006 2,756 Diff 3% Timing of Common Law Potential being raised 50% 45% 40% 35% 30% 25% 20% 15% 10% 5% 0% % of CLP flagged in month that was flagged within 6 months of accident Total number of CLP raised Recovery Segmentation Model Segmentation Algorithm Numbers of claims to teams Approximately 11,000 claims, about 900 per month, have progressed through the algorithm since Recovery went live. There were some initial minor problems with the new at fault variable, partly caused by data entry problems, and partly due to the variable being collected differently to how it was collected in the data used to build the model. This resulted in more claims than expected moving to Client Assist. This was addressed by adjusting the score thresholds used to assign claims to teams. Claim numbers to Client Assist and Active Management teams have broadly been in line with expectations since May Algorithm moves to teams - New claims Average claims per month Percentage to Active Management Percentage to Client Assist Oct-Dec % 80% Jan-Mar % 78% Apr-Jun % 70% Jul-Sep % 70% Overall % 74%

23 Algorithm moves to teams - New claims Active Management - Actual Client Assist - Actual Active Management - Expected Client Assist - Expected Claims moving to the right team The segmentation model is aiming to predict whether a claim will have services days post acceptance, and whether the client will lodge at common law. Given the time lag involved (only 3 months of claims have received payments for services days post acceptance, and common law can take years to lodge), results below are relatively early. Examining claims accepted in October to March 2011 the percentage of claims going to o Client Assist that were not high cost and have not lodged common law (i.e. were in the appropriate team) is lower than expected (with the expected percentages being based on the results achieved using historical data in development of the algorithm). This is linked to the threshold changes noted above, and would be expected to improve for more recent months o Active management that were high cost or did lodge common law (i.e. were in the appropriate team) is higher than expected. Algorithm months Oct10 - Dec10 Algorithm Team Algorithm Claims Claims In 'Right' team Actual % Expected % Active Management % 37% Client Assist 1,973 1,578 80% 88% The algorithm is expected to evolve over time and its accuracy is expected to improve. Nonetheless the initial results are cautiously encouraging.

24 Algorithm Accuracy High cost, common law lodged 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Active Management % 'Right' Team - Actual Client Assist % 'Right' Team - Actual Active Management % 'Right' Team - Expected Client Assist % 'Right' Team - Expected Claims moving out of team assigned by the algorithm The number of claims moving out of their team assigned by the algorithm can be used as an early indicator of its accuracy, as claims staff will move claims they believe are in the wrong team via the manual business rules. From October 2010 to August 2011 the number of claims remaining in Client Assist is in line with expectations. The number of claims remaining in active management is higher than expectations. This could be due to active management claims having common law potential and claims staff consequently retaining them (although these may never go on to lodge common law). Algorithm months Oct10 - Aug11 Actual % Still in Algorithm team Expected % Still in Algorithm team Algorithm Team Claims to team from Algorithm Claims leaving algorithm team Client Assist 8,124 1,098 86% 88% Active Management % 53%

25 Percentage of claims still in algorithm team New claims and inactive segmentations 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Active Management (excl Early Support) % still in team - Actual Active Management (excl Early Support) % still in team - Expected Client Assist % still in team - Actual Client Assist % still in team - Expected Movements via manual business rules The Recovery model aims to have as few claim movements between teams as possible and recent months have seen claim movements in line with expectations. In particular the number of claims moving from Client Assist into Active Management is low, which suggests that Client Assist is receiving predominantly low risk claims from the algorithm as intended. October and November 2010 saw a large number of movements. o This was mainly driven by re allocation of claims between the Return to Work Complex and Less Complex teams which was necessary to even out the distribution of active income claims. The allocation for Day 1 had placed a number of claims with active common law in the Complex teams, even though these claims were not necessarily receiving income payments. o Some claims were not able to be moved automatically at Day 1, and were moved manually. These contributed significantly to the high number for October 2010, which was a one off event and should not be regarded as a failure of the model. May 2011 had a large number of movements, with several active management teams having a clean up of claims that were now ready for Client Assist. Claims staff have since been encouraged to move claims as they become ready for Client Assist, rather than moving them in bulk.

26 Movements between teams Average moves per month Average moves from AM to CA Average moves from CA to AM Oct-Dec10 1, Jan-Mar Apr-Jun Jul-Sep Overall 1, Recovery Claims moving between teams 0 Oct-10 Nov-10 Dec-10 Jan-11 Feb-11 Mar-11 Apr-11 May-11 Jun-11 Jul-11 Aug-11 Sep-11 Arriving in team Arriving in team - Expected

27 What s next? Having implemented the initial phase of Recovery in October 2010, work has now commenced on Phase 2. A key focus area for Phase 2 is effective interventions and pathways for Return to Work clients, particularly those with mental health, persistent pain and return to work barriers. It has been identified that claims with all three of these barriers cost the scheme three times as much as a claim with only one of these barriers. This is shown on the following diagram: Recovery No Fault Costs Recovery Combined Liabilities (CL & NF) Total average cost per claim per year % of claims Recovery RTW $20K 32% No Complexities $9K 54% Total ave cost per year Recovery RTW $81M No Complexities $56M $32K 3% Mental Health $79K 3% $151K 2% $77K 1% $81K 2% $36K 2% Persistent Pain $15M Mental Health $42M $54M $12M $29M $11M Persistent Pain Phase 1 of Recovery implemented the systems and structure to identify and allocate these clients to teams. Phase 2 is intended to change the systems and processes used to better support these clients to recovery. Some of these initiatives include: early contact with clients prior to claim acceptance; motivational interviewing skills training for rehabilitation coordinators; and a Return to Work plan within our claims management system to clearly monitor the barriers and interventions for clients. The change management plan for Recovery Phase 2 will be focused on influencing the mindset and behaviours of rehabilitation coordinators to embed the goal of working proactively and holistically in a client focused way.

28 Conclusion Recovery Phase 1 achieved the restructuring of 150 staff members in to twelve new teams with clear purposes and portfolio sizes aligned to client needs. Implementation saw 24,000 active claims and 128,000 inactive claims move into the new structure. New processes such as the segmentation algorithm and client conversational tool were introduced to more effectively identify claims at risk of becoming high cost. While recognising that it is relatively early in the journey, results in the first year have been encouraging. The main components of the new model including the new teams and the segmentation algorithm are largely working as intended. Early indicators of success include algorithm accuracy, income durations, flagging of common law potential and reduced claim movements. Some benefits will take some time to be fully recognised, in particular those relating to common law, but early signs are encouraging. The impact of these changes, combined with the supportive pathways for our clients to achieve their goals to come from Phase 2, will result in the Recovery vision being met to support clients to get back to work and health as quickly as possible.

29 Acknowledgements We would like to thank Rod Peel, Alan Woodroffe and David Attwood for their review of our paper and helpful comments. We would also like to particularly thank Chris Latham for his peer review and suggestions. Any remaining typographical errors, grammatical errors and lack of clarity are wholly the responsibility of the authors. The Recovery model would never have become a reality without the vision and energy of Tracey Slatter, the Head of Claims at the TAC and Bruce Crossett, the Senior Manager of the Recovery branch. Bruce and Tracey did a fantastic job of maintaining and communicating this vision throughout the journey. Finally and most importantly we would like to recognise the 150 staff in the Recovery branch who were the most important ingredient in making the Recovery vision come to life. It is only via the energy and contribution of the staff that the model will be truly successful.

30 References Adam Walsh, Senior Analyst, WorkSafe Victoria correspondence 27 July Australian Safety and Compensation Council, March Compendium of Workers Compensation Statistics Australia 2006/2007, Commonwealth of Australia, Department of Employment and Workplace Relations Crossett, B. (2010). Recovery & Health Services Group Workshop Recovery Overview. Presentation delivered by Mr Bruce Crossett Senior Manager of Recovery at TAC on 20 August LaMontagne, A., Louie, A., and Ostry, A. (2006). Workplace Stress in Victoria: Developing a Systems Approach, Victorian Health Promotion Foundation, Victoria, Australia. Peter Neilson, Branch Manager, Accident Compensation Commission correspondence 26 July 2010 Transport Accident Commission (TAC) (2009). TAC Recovery Business Case. Transport Accident Commission (TAC). (2010). Recovery KPI Driver Trees (Internal document). Victorian WorkCover Authority April, Stressing the point: A study into what causes stress in a budget sector workplace, and how it can be controlled. Waddell, G. (2002). Recent developments in low back pain: UK Paper for World Congress on Pain San Diego August 2002