Yorkshire and the Humber Co-Design Model Frail Elderly End of Life Care A guide to preparing input data and running the model

Yorkshire and the Humber Co-Design Model Frail Elderly End of Life Care A guide to preparing input data and running the model Introduction This End of Life Care modelling tool was funded by the National Health Service in Yorkshire and the Humber, developed by NHS Hull and Wakefield District NHS in conjunction with Ernst & Young. The goal was to increase efficiency of service improvement and development by doing things once and then adapt the model. Ernst & Young provided the international best practice quality standard against which the design of the model was tested. The Yorkshire and the Humber Co-Design Model is built with a standard core, to enable it to be adapted for individual services. NHS Hull has continued to support development. The goal in this version is to quantify the financial implications of caring for people at the end of their life in a community setting rather than within a secondary care environment, and to provide the basis for informed discussion around alternative pathways. Background Modelling and simulation has a significant impact on the design, planning and implementation of health service processes and pathways of care. Simulating existing and new models of healthcare helps the NHS to anticipate and plan for the impact of changing demands, to design more efficient and effective care pathways and to make more informed strategic decisions about the future of the healthcare system NHS Institute The Frail elderly /End of Life Care modelling tool is designed to allow predictive modelling of local health and social care systems for specific conditions or groups of people. It comes pre-configured with a set of parameters, as the basis of the prediction. These are configured locally, and include population and prevalence data, and with a number of pathways of care. Users modify the defaults to reflect local data, change existing pathways or create new ones. The potential scale and costs of the service change can then be explored using real local data. April 2012 Page 1 of 27

Detailed results for each of four questions are provided. How many of the selected group who are currently treated in hospital might benefit from care closer to home? If that care was provided elsewhere what would the future level of activity be in all settings? If that care was provided in a range of settings what would the cost be? How many hospital bed days could be released for other care or saved? For further details on how the model works visit the Data and Analytical Tools section of the National End of Life Care Intelligence web-site at: http://www.endoflifecareintelligence.org.uk/end_of_life_care_models/commissioner_financial_model.aspx The model is designed to replicate the service configuration and care pathways in a health system by: enabling alternative pathways to be developed that reflect changes in population, health prevalence, service configuration and capacity, and models of care running simulations to examine the impact of changes on flow, capacity, end-to-end transaction times and cost across the whole system indicating where changes to one pathway might have impacts elsewhere in the whole system The tool is intended for commissioners, service developers and clinical teams, to help them to: understand the consequence of, and to plan effectively for, changes to the way that frail elderly /End of Life Care will be delivered explore the changes to demographics and disease prevalence on their existing and future models of service delivery Test possible local approaches to reconfiguration of services, such as the impact of reducing deaths in hospital enabling patients to choose their place of death. design and test the performance of proposed changes in practice (e.g. new referral guidelines, care pathways, models of care, high impact changes) on the performance of their local health system Data The model uses Office of National Statistics (ONS) data related to place of death for a given locality for a twelve month period to generate a cohort, in accordance with local data governance requirements; aligns this with Secondary Use Service (SUS) data on hospital spells within a given time period. This given period of hospital spells is selected locally. Three or six months are recommended initially, as this provides sufficient data for analysis and focuses on end of life care. One stay in hospital (a Spell) may include a number of Episodes of care. http://www.connectingforhealth.nhs.uk/systemsandservices/sus April 2012 Page 2 of 27

SUS provides a range of software services and functionality which enable users to analyse, report and present this data. It is the single, authoritative and comprehensive repository of healthcare data, held within a secure environment that maintains patient confidentiality to national standards. It is collected during the provision of care and treatment and this is primarily used to support and improve individual patient care. To access the SUS suite of applications, you should apply to your local registration authority, which are responsible for allocating the appropriate role-based access controls to their users. Your local registration is usually the Strategic Health Authority (SHA) associated with your organisation. If not, your SHA should still be the contact to advise you who is. They will also advise you what happens after you have applied to register to SUS. To find out how you can access SUS visit 'Guidance for Sponsors & Registration Agents on the Granting of SUS Role Based Access Controls (RBAC)' and under the "Guidance" section what information is in SUS The term SUS is used to cover a core access-controlled data warehouse and a number of reporting and analysis services: Core Data Warehouse SUS Payment by Results (PbR) Online Service SUS Data Quality Dashboards SUS Extract Mart (SEM) Hospital Episodes Statistics (HES) NHS Comparators Glossary Link to suggested glossary http://www.ic.nhs.uk/services/the-casemix-service/new-to-this-service/thecasemix-jargon-buster April 2012 Page 3 of 27

How to use the model Quick Guide The model is designed for use with partners and in multi-disciplinary discussions, for example with commissioners, clinicians (including GPs, community nurses, palliative care consultants, geriatricians), hospital-based data coding specialists and performance analysts, to decide how best to apply the model for local services. The process is: Download from the web-site http://www.endoflifecareintelligence.org.uk/end_of_life_care_models/commissioner_financial_model.aspx View the EXAMPLE model and GUIDE then press the provided button Clear Input Area at the top left of the Data Input worksheet to create a blank TEMPLATE for your local use. Decide on the data and populate the data input sheets; Confirm the relevant care pathways for the identified population (including no change ); Review and agree the parameters within the model and adapt for local circumstances; Run the model to generate four reports for local discussion of: current year cost and volume (present period); predicted position in five years time with no change in how patients are treated (future present); predicted position in five years time, if patients benefit from being treated through alternative care pathways (future scenario); Potential hospital bed days released by the service change. Assumptions and Prerequisites The model uses Excel to perform the necessary calculations and generate output; therefore it is recommended that the team who use the model includes competent Excel users. The dataset which is required by the model is a combination of two independent sources which require processing into a single combined dataset before being incorporated into the model. Microsoft Access is adequate for this purpose, but experience of manipulating large datasets is strongly recommended as is a good understanding of the datasets themselves. The recommended sources of data for the model are the Public Health Mortality File and the Secondary Uses Services (SUS) admitted patient care (APC) PbR spells extract. April 2012 Page 4 of 27

About the Model The model aims to estimate the difference between costs of care for people during the last days and weeks of life in alternative care settings compared to a traditional hospital. Using the input data it calculates the current cost of hospital care, the cost of hospital care in five years time taking into account population and cost changes, and finally the cost of care in five years time considering potential alternative pathways. The model was developed in and runs best using Excel 2007. However, other than aesthetic elements (i.e. cell colouring), there is nothing which would prevent it from running under Excel 2003 providing Microsoft s Office Compatibility Pack is installed. See the Microsoft Office Website for more detail. Overview of Running the Model An overview of the steps required to prepare data and run the model are: 1. Identify persons who have died in hospital within the relevant time frame (usually a particular financial year). The public health mortality file (PHMF) is used for this purpose and is available from the ONS. 2. For each of the persons identified in step 1, identify all hospital spells, if any, which occurred in the three or six months prior to death. The SUS APC PbR spells extract is used for this purpose, and is a dataset at spell level which includes all required fields such as spell cost and length of stay. Users to date have tended to select six months as the time period of interest. Local discussion determines the time period or periods of interest. 3. Using the data gathered at step 2, insert the required data into the model, and review any data validity issues raised. 4. Change the parameters for the model to reflect local data and then press the provided button Run Calculation Engine at the top left of the Calculations worksheet to calculate the output. 5. Review the output tables and charts. 6. Generate alternative scenarios for comparison by altering the parameters and re-running the model s calculation. Each of these steps will be discussed in more detail in the remainder of this document. April 2012 Page 5 of 27

Overview and Structure of the Input Dataset At its simplest, the data the model uses is a list of hospital spells, with costs, that occurred in the three months prior to a person s death in hospital. Here specific columns of data required, these are mandatory: ID Death - This is a unique identifier for each person who has died in hospital. It is analogous to NHS Number, though the use of patient identifiable data isn t recommended. ID Spell This is a unique identifier for the hospital spell and is analogous to the Provider Spell Number. If a patient has more than one hospital spell in the period being analysed, each one will have its own unique spell identifier even though the ID death remains the same for that person. Age The age of the person at their hospital admission. Establishment Type This is the care setting in which the person died. Currently this will always be Hospital though the model is flexible enough to be adapted for other uses and establishment types. Prim Diag The hospital spell s primary diagnosis code. PbR Cost The cost of the hospital spell. Spell LOS The length of stay for the hospital spell. The model is able to accommodate a number of columns which are useful for establishing the answer to quick questions regarding the data while using the model. April 2012 Page 6 of 27

Preparing the Input Data The input data is a combination of two main datasets: ONS Mortality data, available yearly or monthly (as a provisional publication). Contact the ONS at vsob@ons.gov.uk for more details. The ONS mortality data includes patient details, location of death, date of death, and diagnosis coding. SUS Admitted Patient Care PbR Spell dataset. This dataset is available from the Secondary Uses Service and each row within it represents a person s stay in hospital and includes the cost and length of stay fields. Preparing the Dataset (an example) Mortality Data The ONS data does not include NHS numbers by default, although they are available as a separate dataset from the ONS. Alternatively the Demographic Batch Service 1 can be used to trace the NHS numbers as required. The steps to prepare the ONS mortality data are (this example uses the frail and elderly aged 85 years or older upon death): Restrict the dataset to persons who have died in hospital within a time period and who re aged 85 or older upon death. To identify if the person died in hospital, a lookup on the field called Communal establishment Code to a small lookup dataset is required 2. If the records from the previous step do not have an NHS Number, it is also possible to use the Demographic Batch Service to retrieve them. SUS APC PBR Spells Data Using the list of persons identified in the mortality data: 1 More details regarding the Demographics Batch Service can be found at: http://nww.connectingforhealth.nhs.uk/demographics/dbs (N3 connection required) 2 The communal establishment lookup data can be found at: http://nww.connectingforhealth.nhs.uk/ods/downloads/officenatstats (N3 connection required). April 2012 Page 7 of 27

Restrict the list of hospital spells to those where the NHS number in the spells data exists on the list of persons identified in the mortality data. Further restrict the resulting list of spells to those spells which occurred within 3 months (or other locally selected period, say six or twelve months) of the person s data of death (92 days). The resulting dataset should be a list of spells comprising of the columns discussed in the prior section regarding the structure of the dataset. Appendix A shows an example of the SQL statements that can be used to perform the steps above. Note that the appendix demonstrates the principles and assume both required datasets are held locally. A note about the datasets and alternative sources It may be possible to acquire the data using other methods depending on the requirements; the approached discussed here outlines the approach the development team suggests. Using the PHMF ensures all deaths are identified, wherever they occur, and the use of SUS data ensures that the cost associated with the hospital admission is calculated correctly. Local data sources may be used to replace the suggested approach, or alternative methods of identifying the require data. One example of this would be to omit the use of the public health mortality file and identify deaths in hospital using the SUS data instead. HES Online may be another suitable alternative source for some of this data. HES s website suggests that it may be able to provide mortality data, but not the spell level inpatient data, with the most notable omission from this being costs. Running the Model A Step by Step Walkthrough An Overview of the Model s Contents The model is implemented using Microsoft Excel. An overview of the worksheets contained within it and their purpose follows: Version Control: Shows a version history with comments. Data Input: The worksheet where the input data is placed and validated. Parameters: User changeable parameters which can be changed to produced different scenarios. ICD Groups: The mapping between ICD codes and the type assigned to each hospital spell in the input data. This is discussed further elsewhere in the document. April 2012 Page 8 of 27

Calculations: The model is run from this worksheet; it is also where all calculation and the rowlevel output it stored which backs the main output. Output Tables: The aggregated output of the model is presented on this worksheet and forms the main output for review. Check: The model includes some validity checks. If any of them fail they will be highlighted at the top of all worksheets, with additional detail here. As a convention all cells within the model whose contents are intended to be changed by the user are coloured yellow. Adding the Input Data into the model (The Data Input Worksheet) Once the input data has been prepared, the block to be pasted should be seven columns wide and fit perfectly into the yellow area shown in figure 1. Mandatory input is coloured yellow on this worksheet. The green columns to the right (see figure 1) can be populated with optional input data which has proven useful during team discussions to answer on-the-spot question about the data during use and to provide a richer view of the conditions and treatments of the anonymised persons included in the dataset. Figure 1 - The 'Input Data' worksheet prior to pasting data. Note the yellow area to denote where the mandatory data is to be pasted, with the selected cell being the insert point. The green area is for optional data. April 2012 Page 9 of 27

Even though Excel will accept many rows it is recommended that the input be limited to as few as possible since the volume of data has a direct impact on the time it takes the model to process the data. All models run by NHS Hull have been less than one thousand rows and have run within a reasonable period of time. If the model contains data from a previous run, this can be cleared by clicking the Clear Input Data button. Once data has been pasted into the model, the result should look similar to figure 2. Figure 2 - The 'Input Data' with the relevant data pasted in. Cell D11 remains the active cell as this is the point at which the data was pasted in. Finally, press the Validate Input Data button. This button does two things: It populates the Index column which is required for the model to function correctly, and second, it performs some rudimentary validation. If any validation errors are encountered, the number of errors is shown on row 8, and the error type is shown in column C for each row which failed. April 2012 Page 10 of 27

The validation checks every row in the input area until the first full blank row is encountered. The checks performed are: Ensure all fields do not contain a blank value. Ensure the Age field is a number between 0 and 130 inclusive. Ensure PbR Cost field contains a number equal to or greater than zero. Ensure the Spell LOS field contains a number equal to or greater than zero. Figure 3 shows an example with rows with failed validation. Figure 3 - Failed validation is shown in column K with a brief explanation. The error count on row eight provides a useful indication if there are unseen errors elsewhere on the input data worksheet. Figure 4 shows the input data with successful validation and the index column correctly populated. April 2012 Page 11 of 27

Figure 4 - The input data has passed validation and is ready for use by the model. Adjusting the Model Parameters (The Parameters Worksheet) The Parameters worksheet is split into five sections: Hospital Spell Data Input Period; Growth Parameters; Price in new location; Decision Criteria; and Bed Utilisation Hospital Spell Data Input Period During the preparation of the input data, a period of time prior to death for which spells were included was used. That period of time should be specified here. It is used on the output worksheet for the model to provide a documented record of the parameter used. It should be noted that this parameter does not affect how the model functions in any way and the model can be rerun and saved using different Input Periods where this is useful. The Input Period is a local selection of the time period considered most relevant to the decisions the Model is being used to address. For example where the decision relates to people going into and out of hospital as they enter the End of Life, then Hospital Spells in the last 6 months prior to death is a reasonable period to use. This April 2012 Page 12 of 27

is informed by the definition of End of Life care set out in the National Institute for Health & Clinical Excellence (NICE) End of Life Care Quality Standard published 30 November 2011. The care of adults with advanced, progressive, or incurable conditions who are approaching the end of their life and are expected to die within the next 12 months. The end of life definition also includes adults with existing conditions who are at risk of dying from a sudden, acute crisis in their condition, or those with lifethreatening acute conditions caused by sudden catastrophic events. The standard also covers support for the families and carers of such people. Growth Parameters In the output, the model also considers how the population and tariffs may change in five years time. Section 2.1 provides the facility to input how population and tariff is expected to change during this time. Age is split by age band, which can be changed as required, though care must be taken to make sure each age band runs consecutively as the model doesn t validate this input. Where the focus is on a specific condition such as respiratory diseases, and the predictive data is available, then consider inputting Population Growth by Age Band as a composite of the overall population growth and the predicted growth for that specific condition. This will provide a more realistic future activity levels and costs. Figure 5 - Population and tariff growth parameters The tariff growth by care setting in section 2.2 allows the input of expected growth of tariff by each care setting for the five year projection expressed as a percentage. April 2012 Page 13 of 27

Price in New Location The section named Price in new location provides a way to estimate the per day cost for each of the alternative care pathways. There are three sub-sections: Sub section 3.1 provides a selection of different resource types and the ability to configure the cost of each. Sub section 3.2 allows the definition of alternative care pathways, including the resulting care setting and the amount of resources required for that pathway. Early Adopters have found it useful to characterise each Care Pathways as providing Supportive, Intermediate or Intensive levels of care in order to determine workforce levels and skills input. The appropriate period of care as short, medium, or long, should to be selected for each pathway. Sub section 3.3 defines the number of days associated with short, medium or long periods for each pathway. These are pre-set at 28, 42 and 56 days to reflect current experience for end of life patients. Figure 6 - Defining the cost of resources used in alternative care pathways At the time of writing the Independent Palliative Care Funding Review reported in July 2011 and pilots are underway to develop tariffs. Until a tariff is available the model uses estimated cost derived from workforce skill mix as the basis of Price in New Location. April 2012 Page 14 of 27

Figure 7 Defining the alternative care pathways and the amount of resources they use Figure 8 - Defining the duration for each short, medium or long care pathway. Decision Criteria The cause of death decision criteria, section 4.1, is a table where the relationship between the cause of death types and the potential alternative pathways are defined. First each record is mapped to a Type. Then an algorithm is applied to allocate the record to the Care Pathways as shown in the Price in New Location section above and calculate the future predicted volume and cost. This mapping and algorithm has been developed by clinicians and tested during the Early Adopter phase led by the Department of Health National End of Life Care Programme www.endoflifecareforadults.nhs.uk and will continue to be validated as the model is used. April 2012 Page 15 of 27

The left-most column shows the record Type, numbered 1 to 10. During the running of the model each spell is associated with a type derived from its primary diagnosis (discussed further in the ICD Groups section), and it is here that these spells can be allocated to a potential alternative pathway. New types can be added as desired and the proportions for allocation to Alternative Pathways can be changed to align with up-to-date published evidence and local understanding as discussed below. The potentially avoidable admission column provides a way to specify if the spells associated with the type have the potential for alternative pathways of care or not. The columns labelled care pathway type if potentially avoidable allow the split of any given type across multiple pathways. For example, in the figure below type 3 is allocated 50%, 14%, and 36% to care pathways A, B and C respectively. This will mean that the total cost of alternative care for those spells which are associated with type3 will be apportioned across the alternative care pathways. Again, these percentages can be altered as required to decide what percent of a particular type of admission are directed to an alternative pathway. Figure 9 - The decision criteria determines the conditions of when a spell is avoidable, and if so how the alternative cost is calculated from the potential alternative pathways. A note about deriving what proportion of activity should be directed to each alternative care pathway In the example above, Type 2 represents avoidable admissions, with 28% being directed to care pathway B (Own home) and 72% to care pathway C (Care Home). For the purposes of testing the model, these figures were arrived at by looking at the location of deaths in this cohort and calculating the split between those that died in a care home/residential home and those which occurred at home. April 2012 Page 16 of 27

Location Deaths Percent Own Home 61 28% Nursing Home/Residential Home 158 72% Total 219 100% Table 1 The % split between deaths at home and in a residential/care home. Bed Inputs Figure 10 - The Bed Inputs part of the parameters worksheet. The bed inputs section allows the utilisation level of a hospital bed. It is used in the output section to estimate how many potential beds-per-year can be made available. The default value is 85%, but can be changed to suit an organisation s requirements. Specifying a Spell s Type (The ICD Groups Worksheet) The ICD Groups worksheet contains the mapping between each record and the type it will be assigned. This mapping was created by clinicians during the development of the model based on detailed review of all the Input Data for people over 85 years of age. Following that analysis, spell primary diagnosis was tested and applied as a reasonable indicator for mapping the selected population (over 85 years old and died in hospital) to one of three Types. This does not indicate a clinical decision. However it is an indicator of the likely numbers and types of people who have the potential to be cared for in an Alternative Care Setting. Types are mapped to the ICD code at block level, which encompasses the first three characters of the ICD code. The mappings can be changed to suit requirements, provided changes are based on validated clinical review. This is not recommended without reference to the national programme, so it can be tested and validated as Early Adopters report back. Only those primary ICD Groups which appear in your input data need be maintained. Clinical input is required to set these values correctly. The model has a pre-defined set for Frail Elderly/ End of Life Care. April 2012 Page 17 of 27

There are four columns which can be changed: Has Alt Type Used to identify cases where two types can be assigned to a spell based on some other condition. For example, in the version of the model NHS used for the frail and elderly it was decided that patients who were admitted for some cancer related ICD codes should be allocated to type 3, if the length of stay was two days or less, or type 1, if it was greater than 2 days. Type This is the primary type allocated to an ICD block. Alt Type This is the alternative type which can be allocated to an ICD block. Condition This is a comment regarding the condition upon which the alternative type is used instead of the primary type. Note that this is a comment only. To build decision logic into the model requires changes to the calculation worksheet which will require a higher technical knowledge of the model s workings. Figure 11 Each ICD block code is associated with a type. These are then used to identify a type for each spell based on the primary diagnosis. Type 1 unavoidable admissions and require hospital care. Type 2 appropriate to receive the care elsewhere (i.e. in the commmunity). Type 3 may require hospital care, so allocated proportionately between hospital and community April 2012 Page 18 of 27

It is important to note that by mapping a record as a specific type enables a discussion at a population level of what kinds of people could benefit from an alternative pathway of care, other than hospital admission. It does not replace clinical decision- making. Types are allocated to appropriate Care Pathways which again are adapted for local use. The default Pathways are: Care Pathway A Care delivered within a hospital setting only Care Pathway B Supportive level of care (defined by skills mix and time allocated to care) delivered in a Community setting. Care Pathway C Intermediate Level of Care (defined by skills mix and time allocated to care) delivered in a Community setting Additional Pathways can be added to the model for example Hospice at Home and other services where Intensive levels of care can be delivered outside of a hospital setting. Running the Calculations (Calculations Worksheet) Once the input data has been validated and the parameters changed as required, the model s calculations are performed by pressing the Run Calculation Engine at the top of the calculation worksheet. A pop-up message box will ask to confirm the running of the model (it may take some time for large datasets), after which the model performs the necessary calculations. The progress can be viewed in the status bar at the bottom left of the window. When the progress has ceased updating, the running is complete and the output can be reviewed. Structure of the Calculation Worksheet There are a number of sections on the calculations worksheet; they re introduced below for the purposes of familiarity, though detailed discussion will be omitted: Data Call Up As the model works through the list of the spells on the Data Input worksheet, the data for the relevant spell is temporarily held in this section while the model performs the necessary calculations on it. Calculation Controls This section contains general numbers required for the model to run correctly. Where relevant these numbers are automatically updated, for example, No of Rows should match the total number of records to be processed. April 2012 Page 19 of 27

Calculation Engine The calculation engine section is where the logic of the calculation exists. The Result Row at the top is the output of all of the calculations which are built up from the subsections below. Changing anything in this section inadvertently will either prevent the model from running correctly, or produce false results. It is recommended that only confident Excel users make changes to this section. All of the calculations can be viewed in Excel and should be fairly understandable by experienced Excel users. Output Copy Area This area is a combination of the data call up and the result row in the calculation engine area and serves as a temporary storage of the data before being added to the output paste area. Output Paste Area As each row of the input data has been calculated, it is pasted here and forms the dataset which drives the output tables. Reviewing the Output (Output Tables Worksheet) There are six output sections: General Information; Alternative Pathways; Number of Spells by Care Setting or Pathway; Local Tariff for Each Care Pathway; Commissioner spend in each Care Setting ; and Bed Days Saved in Hospital. General Information This section contains some general information about the model and its data. The two current pieces of information are: the number of people in the model, and the number of months prior to death for which spells have been included. Alternative Pathways This section reports the total number individuals included in the model and the number of hospital spells which were identified having a potential alternative pathway. Using the example below, 628 spells were provided on the original data input, of which just under half, 49.5%, met the criteria for an alternative pathway. April 2012 Page 20 of 27

Figure 12 The number of persons in the model and how many of their spells which may be attributed to an alternative care pathway. Number of Spells by Care Setting or Pathway This report shows the number of spells by pathway. The Current column displays the count of spells as they re supplied on the input data, the Future Present columns shows the number of spells including any population growth (as defined on the parameters worksheet), and the Future Scenario column shows the number of spells split by the alternative care pathway. April 2012 Page 21 of 27

Figure 13 Comparing the number of spells for the current situation, future situation if no alternative pathways are explored and future scenario situation if potential alternative pathways are explored. Local Tariff for each Pathway This table is a summary of the pathways and their proposed cost as defined on the parameters worksheet. It is not driven by the model s calculation but is included as a reference. Figure 14 - The costs of care for each alternative pathway at low, medium and high levels. At the time of writing the Independent Palliative Care Funding Review reported in July 2011 and pilots are underway to develop tariffs. Currently therefore the model uses the estimated cost derived from workforce skill mix as the basis of Local tariff. April 2012 Page 22 of 27

Commissioner Spend in each Care Setting ( ) This output shows similar data to that of the Number of Spells by Care Setting or Pathway output but from a financial perspective instead of activity. As with the activity report, the current column shows the current cost, the future present column shows the future cost including population and tariff growth, and the future scenario shows the future cost considering any potential alternative pathway. Figure 15 Shows the financial comparison of exploring alternative care pathways. Bed Days Saved in Hospital This report shows the number of potential bed days released. The Beds Released; represents the bedcapacity released at the utilisation level as specified in the parameters section. Figure 16 - The potential number of bed days which may be released. April 2012 Page 23 of 27

Appendix A Below is an example of the SQL required to extract the data required for the model assuming the PHMF and hospital spell data are held on one table each. The purpose of the example is to show how the two datasets are connected and which column comes from which dataset; it s probable that local circumstance will mean that this example will be unusable without modification. This example was written to run using Microsoft SQL Server. -- Step 1: Identify the persons to be included in the model. -- First, conditionally drop the temporary -- table if it exists from a previous execution. if object_id('tempdb..#tdeaths') is not null drop table #tdeaths -- Second, identify persons in the PHMF data with -- and auto-generated pseudonym to be used instead -- of the NHS number. Select [NHS Number], identity(int, 1, 1) as [ID Death], [Date Of Death], [Age], [Establishment Type] into from where #PHMFTemp dbo.phmf [NHS Number] is not null and [PCT Of Residence] = '5NX' and [Date Of Death] between '01-Apr-2010' and '31-Mar-2011' and Age >= 85 and [Age Unit] = 1 -- Age unit of measure is years. -- Step 2: Identify spell data for the persons identified in previous -- step. Output is suitable for pasting into the DataInput -- worksheet within the model select #PHMFTemp.[ID Death], -- NHS no pseudonym from step 1 SpellData.[Provider] + '_' + SpellData.[SpellNo] as [ID Speall], #PHMFTemp.[Age] #PHMFTemp.[Establishment Type] SpellData.[Prim Diag], SpellData.[Pbr Cost], SpellData.[Spell LOS] -- Total spell LOS (Dischage - admit date) from dbo.spelldata inner join #PHMFTemp on SpellData.[NHS Number] = #PHMFTemp.[NHS Number] where datediff(d, SpellData.[Discharge Date], #PHMFTemp.[Date Of Death]) < 92 April 2012 Page 24 of 27

Appendix B Healthy Ambitions End of Life Care Project Capability assessment questions for financial modelling (EXAMPLE Completed by an Information Manager) The purpose of this self-assessment is to identify the skills that you have, and the ones that you can expect to develop or improve during the market analysis project, by working closely with others, and as a platform for discussing who else in your organisation can contribute any additional skills needed to complete the work. On a scale of 1 to 5 (1 being low/none ; 3 being yes, reasonable 5 being expert) Please rate your own ability by answering the following questions (** must have at least a score of 3 to be capable of providing analytical support to the team who use the Model): 1. I have experience of gathering a wide variety of information from different sources quickly (scored before / after the experience) 5 / 5 2. I am well networked within my organisation 5 / 5 3. I have experience of completing analyses of the implications of changing prices on commissioner spend 3 / 4 4. I have experience of completing analyses of the implications of changing prices on provider viability 3 / 4 5. I have experience of completing detailed activity and financial forecasting** 5 / 5 6. I have experience of completing options identification, validation and refinement** 4 / 4 7. I have experience of considering risk/reward mechanisms that influence provider behaviour 2 / 2 8. I have an understanding of government policy in relation to procurement, tendering and the principles of market-based economics 2 / 2 9. I have a clear understanding of how to complete a comprehensive utilisation assessment of a healthcare provider (determining the extent of the available capacity that's used by a provider. The given units might include theatre time, time of a specified practitioner etc depending on what PCTs looking at). 5 / 5 10. I have a clear understanding of how to complete benchmarking of services, markets and organisations 5 / 5 11. I have experience of building complex financial and activity models** 5 / 5 April 2012 Page 25 of 27

12. I have experience of completing category analysis (this refers to recognising that different care pathways may be comprised of services that may or may not constitute a typical market) 5 / 5 13. I have experience of completing impact and risk assessments and converting these assessment into commercial insights relating to financial risk** 1 / 3 14. I have experience of completing scenario planning and analysis** 5 / 5 15. I have experience of completing data analysis** 5 / 5 16. I have experience of conducting market analysis 1 / 2 If you are confirmed as the PCT representative for financial modelling please complete the following questions in order to monitor your personal development. These questions should be completed at the start and end of the exercise. On a scale of 1 to 5 (1 being low/none ; 3 being yes, reasonable 5 being expert) Please rate your own ability by answering the following questions: 1. I am confident that I understand how to go about a market analysis 1 / 2 2. I am confident that I understand the part that financial modelling has within a market analysis 5 / 5 3. I know how to construct a financial model to inform market analysis 3 / 3 4. I have the knowledge and am ready to share that with other financially/analytically competent folk on how to go about constructing a financial model to support market analysis 3 / 4 5. I have read and understand the processes Guide created by EY on how to create a financial model 3 / 4 6. I am confident that I could explain the Guide to other financially/analytically competent folk within a PCT to help them create a model 3 / 5 General Comments: The learning experience has been good and (in hindsight) wished I would have been brought sooner into the project in order to be able to plan/schedule my time and get more out of it. With regards to the methodology used for building the scalable model, there are aspects of it that I will definitely be able to thread into our own methodology and also share with the rest of my team. That is an additional benefit from my personal perspective. Having to use the model within a workshop environment where clinicians, commissioners and other professionals were present was a new experience for me. It built my skills in being able to work with others to use the data and intelligence that the model provided April 2012 Page 26 of 27

Acknowledgements: This work could not have been undertaken without the continued enthusiasm and willingness to adapt and improve the model by a core of people. Special thanks to: NHS Hull The Maltings, Silvester Street, Hull HU1 3HA Mark Ward Performance Analyst Performance Governance and Information Emma Smith Performance and Information Manager Secondary Care and SLA Contract Sue Craven Commissioning for the original development to: NHS Wakefield District White Rose House, Wakefield Alejandro Arnes Information Manager Julie Thorpe Senior Commissioning Manager Calderdale, Kirklees and Wakefield PCT Cluster Mike Williams Finance Manager CONTACT At the time of writing organisations are changing shape. For ongoing support and information on Yorkshire and the Humber Service Co-Design Model Contact: lois.bentley@nhs.net or lois@bridgesfm.com who is Custodian for the Model April 2012 April 2012 Page 27 of 27