Please select the service you like to visit.i5 Commissioning Suite i5 Watson Browser
Posted on January 26th, 2016 To view the full article from Linkedin click here
Tragically, many people have died in the US shovelling snow in last week's snowstorm. This is often caused by undiagnosed conditions which cause poor quality of life or sudden death as people might not be aware of them and do not receive medical advice. Often care is provided when their condition worsens which stops after discharge if not diagnosed. This results in frequent hospital visits, increased health care cost and negative outcomes.
Fortunately, patients can be classified by algorithms for a probable condition that is not yet diagnosed using Population Health Management and AI methods. If diagnosed promptly, they could receive life-saving education and be treated accordingly to achieve better outcomes at a lower cost. This would be truly a �win-win� situation or Better Value Healthcare.
Recently, algorithms were applied to patients� medical records to identify undiagnosed Diabetes, COPD, Infections, Hypertension, etc. - demonstrating a clear trend in application of algorithms for diagnosis. Such algorithms can use anonymised primary (GP) or secondary (hospital) care data or both. Primary care data contains medication, physiological parameters and some diagnosis and procedures. Secondary care data contains many more diagnosis and procedures than primary care but contains no physiological parameters and medication.
By way of example, the best results in diagnosing Hypertension were achieved with primary care data when medication and physiological parameters were included in the algorithms. Nevertheless, secondary care data has also been successfully used to predict Diabetes, Heart Failure, Atrial Fibrillation (AF), CHD, COPD, Dementia and Depression. Since secondary care data contains many more diagnosis and procedures than primary care. Therefore disease patterns, attendance frequency and co-morbidities in combination with AI Probabilistic Reasoning and Bayesian Networks can be used to develop models for probable disease diagnosis.
Studies show that the majority of patients at risk were successfully classified in the US. A study in Illinois concluded 72% of patients were correctly classified with hypertension; in Million Hearts, a US government initiative, 83% were correctly classified using secondary care data. Identification of probable undiagnosed diabetes based on US primary care records resulted in the discovery of a substantial number of patients with unrecorded diabetes.
In the context of improving health outcomes and identification of commissioning opportunities, i5 Health has been using various AI algorithms for identification of undiagnosed conditions based on secondary care information to find service gaps in local health economies. The table below shows the number of patients identified with undiagnosed conditions for three confidence intervals in England. Patients within the >90% confidence interval would benefit from accurate diagnosis in primary care, >80% includes patients invited to clinics and >70% includes patients invited for screening programmes.
For example, there are 61,924 patients in England that have an over 90% probability of Atrial Fibrillation that would benefit from seeing their GP, 159,929 patients that would benefit being seen by GPs and clinics and 243,333 patients that would benefit from a screening programme.
The English NHS can and must improve identification of Long Term Conditions e.g. AF to reduce preventable myocardial infarctions and strokes. However, improvement can only occur if all patients with undiagnosed conditions are promptly identified, accurately diagnosed by a clinician and provided with evidence-based treatment and support.
For more information about identification of patients with undiagnosed conditions for your CCG or hospital, please contact email@example.com
Posted on January 18th, 2016 To view the full article from Linkedin click here
For the first time in the NHS, the local planning process will have significant central money attached and the plans become the single criteria for receiving transformational funding for 2017/18 onwards. The most compelling and credible plans will secure the earliest additional funding.
The latest NHS planning guidance for 2016/17 - 2020/21 focuses on closing the health and wellbeing gap, the finance and efficiency gap and the care and quality gap for the NHS as a whole. It requires that NHS commissioners and providers focus on Place-Based Planning (PBP) to deliver a five year Sustainability and Transformation Plan (STP) and a One Year Operational Plan (OYOP) for 2016/17 as part of the Five Year Forward View (FYFV).
To deliver the FYFV, major transformational changes are required at local level that need to be delivered faster than originally intended and must be sustainable. Therefore, the NHS is asking each local health and local care system to come together to create their own accelerated area blueprint to deliver the FYFV, using Place-Based Planning (PBP), as a full draft by 8 Feb 2016.
PBP involves planning for geographically local populations instead of planning by organisation with the intention of reducing organisational separation and autonomy as these do not fit in with patient pathways and new models of care. PBP supports the development of a shared vision within the local community which should agree on a set of overarching initiatives and services. Those interventions and services should be founded on experiences elsewhere in the NHS and should drive adapting and shared planning.
To implement shared planning from an operational perspective, the following should be kept in mind:
Expert Systems can support planning by applying hundreds of rules based on Population Health Management (PHM) to match patients to services and interventions. The outputs of such rules help decision makers to measure affected populations, evaluate what and how much can be saved in each setting, facilitate grouping of services into pathways and new models of care and enable what-if scenario modelling to perform impact analysis to assess sustainability.
NHS Halton CCG, already ahead of many organisations in combining the forces of Health and Social Care, has been using the Expert System developed by i5 Health (COP) to identify and evaluate interventions for which sufficient patients are present in their local population. They identified the best options for their Urgent Care, OP and ACS pathways to drive an informed transformation programme. A case study on how an Expert System consisting of PHM rules has been used to identify the most suitable services and interventions can be found here.
Posted on January 7th, 2016
Tags: i5 AI for BI
i5 Health and Brighton and Sussex University Hospitals NHS Trust (BSUH) have developed an Unscheduled Care Predictor that enables accurate short-term demand forecasting at BSUH�s four A&E sites in Brighton (Royal Sussex A&E, Royal Alex A&E, Princess Royal A&E and Sussex Eye A&E).
Short-term forecasting of A&E demand aids resource optimisation, workforce planning and rota management and is a valuable tool to avoid 4 hour breaches and ambulance queues. Forecasting is based on past data which includes daily patient counts, past weather patterns, calendar events, holidays, temperatures etc. The ability to forecast seven days accurately is mainly dependent on the accuracy of the temperature and the weather forecasts which has become more and more reliable.
The Unscheduled Care Predictor uses Neural Networks that require training before they can be properly operated. Training of the Neural Networks involves the presentation of historic data thousands of times so that the �weights� that measure relevancy of the input can be adjusted. Once the network has been trained, seven day forecasts for each A&E site can be made based on the last six weeks of data.
Accuracy of the forecasting at each site differed mainly due to the number of patients attending each of the sites.
The principal A&E site at Royal Sussex forecast accuracy was over 97% - with a max error of +10% and a min error of -8% at worst.
With the number of patients fluctuating between 210 and 310 daily, the forecast was within �5 patients on average.
The diagram below shows the prediction accuracy on a daily basis.
For more information or how to obtain the tool for your A&E departments, please contact
Dr Harald Braun - firstname.lastname@example.org
Posted on October 26th, 2015
Continuing from my last blog Where are the Missing Dementia Patients, this is a summarised update from the i5 Health - Dementia Digest - reports. This update will focus on key numbers for England and is available at CCG or LAT levels on request.
The DoH and NHSE set a national target of increasing dementia diagnosis at GP practices from an assumed 50% to 66% between September 2014 and March 2015 (6 months). For this, �5m was set aside to pay GPs �55 for each patient put on the Dementia Register. This has caused controversy with GPs calling for boycott of the scheme because it may put patients at risk of misdiagnosis by directly linking GP payments and government targets.
The initial floor for the calculation was that, in March 2015, 323k patients were on the register, assumed to be 50% of the likely total. To bring
this to 66%, an additional 103k patients would need to be found. At �55 each, this would cost the DoH �5.7m, not the �5m as allocated.
From September 2014 to March 2015, GPs were encouraged to find dementia patients and add them to their registers. It can be noted that this
incentive scheme has caused a significant increase of dementia registers - as shown in the diagram below. During this period, 85k dementia
patients were added to the register and 7k were taken off due to death since dementia is mainly present in the elderly. The net increase was
78k patients but the DoH paid for 85k patients - paying �4.68m to 5,701 participating GP practices.
At the first glance this seems like a success. But there are two ways of increasing the dementia register: by making a new diagnosis or by changing an existing diagnosis. To support both, LATs introduced two initiatives: Care Home case-finding and Data Harmonisation.
Making a new diagnosis in a practice or at a care home is what we expect from GPs and this is what they generally do very well.
Data harmonisation does not involve a new diagnosis. It is about changing codes on the practice IT system to include more patients in the dementia register. It changes the patient�s clinical record and, where possible, back-dating the date of diagnosis. It involves GP IT Administrators or CCG Medicines Management identifying acetylcholinesterase inhibitor prescription or change codes that suggest the presence of dementia. Basic pilot projects at three GP surgeries showed an almost 20% increase of patients on the dementia register. LATs were financially supporting CCGs to change clinical coding in GP systems in order to increase prevalence rates to reach the 66% target.
The table below shows that 7,119 practices participated in the DES and earned �4.68m between them. Although most practices earned less than �4k, the
top 50 practices earned almost �300k between them.
Although 85k patients were added to the dementia risk register, there are still large differences between hospital and GP practice diagnosis of dementia. As outlined in my previous blog, this discrepancy consists of patients being diagnosed in hospital and not being followed up in primary care.
Has this investment improved dementia patient care?
Our latest research shows that there are still 62,881 more patients diagnosed with dementia in hospital than in primary care. An increase of over 14k patients compared to 2013/14 costing a total of �150m over three years.
The table below shows the top 20 CCGs (out of 211 CCGs) where patients were clinically diagnosed in hospital but not added to the GP dementia
register. The Additional Patients in Hospital column shows the number of patients that were diagnosed in hospital which are not on the GP
dementia register and the Hospital Spend column shows how much those patients were costing the commissioner in hospital care over a three year
By increasing the dementia registers by those additional 62,881 patients clinically diagnosed with dementia, the government would improve dementia care and reach a diagnosis rate of 71.8% - well above their 66% target.
Posted on October 22nd, 2015
Based on the following key facts, many dementia patients are sub-optimally treated:
The Alzheimer�s Society states that an estimated 850,000 people in the UK have dementia which they suggest will rise dramatically by 2021 to over 1 million . A deep dive analysis by i5 Health into commissioning opportunities for NHS commissioners has shown that, as at 2015, there are 584,657 patients in England that were diagnosed in hospital with dementia. This number can be pro-rated to the UK population which equates to 694,409 patients (22% less compared to the Alzheimer�s Society�s estimation).
i5 Health has projected that the number of patients diagnosed with dementia for the UK should reach 725,000 by 2020, which is also lower than the projections of the Alzheimer�s Society but still very significant to the NHS.
GPs working in the NHS maintain QOF registers on which patients with long term conditions are recorded . The registrations provide extra income for challenged practices. One of these registers is the dementia register. The highest number of patients on this register is 561 - for the Summerhill Surgery in Dudley. When aggregating all dementia risk registers for England we get 348,973 patients. This means, only 348,973 patients are known to GP practices out of 584,657 patients diagnosed in hospital. Overall, a staggering 234,684 patients are missing from the GP registers; they are therefore not being looked after properly and additionally present a huge long-term clinical and financial risk to the NHS.
Using the i5 Health case finding tool for dementia, 48,025 patients in England were identified that are not on the GP register . Large CCGs like N, E & W Devon have 716 dementia patients missing on their GP registers with one practice alone missing 50 dementia patients (58 on the register and 128 in hospital). The CCG with the most missing patients of 1,295 is Birmingham Crosscity CCG with 115 GP practices: one of the practices is missing almost 80 patients.
This is low hanging fruit. CCGs or GPs can obtain lists of de-identified patient information from i5 Health that they can send to their hospitals for identifying patients on their PAS system. With this information, GPs can contact patients for dementia review and diagnosis. In England, around 4,400 practices have on average 11 patients to contact, with the worst affected practice in York having to contact 135 patients. The cost of a second class letter is �0.54 and the overhead to write and print it is �8.50. Informing 48,025 patients would cost around �400,000 nationally. This can be offset with avoided dementia related hospital admissions and the associated long term impact for late diagnosis.
Those missing 48,025 patients have caused over 171,000 episodes of hospital care directly related to dementia in England. This equates to approximately 95,000 hospital spells with an average cost of �1,078 (Cerebral Degenerations). The potential saving could be over �102 million in acute care only, not counting the additional care costs required for advanced dementia patients.
(Data sources used: QoF, ONS, Dementia Assessment and Referral Data Collection, CQUIN, National Tariff 2014/15, HRG Grouper)
Posted on July 28th, 2015
This is part 3 of a 7 part blog on reducing Non-Elective Readmissions (NELR).
Part 3: How can savings be achieved?
In a saturated local health economy where demand outstrips supply, reducing readmissions will achieve financial savings only in the long term. This is because any reduction in readmissions reduces inefficiencies which will lower the cost of treatment per patient but will not release capacity that would release cash. Any capacity freed will be utilised immediately e.g. for elective care to reduce waiting lists. This in turn causes less patients to deteriorate driving less emergency admissions and better outcomes.
In the long term, savings can be achieved after waiting lists are shorter and unnecessary procedures avoided - releasing capacity and cash. To achieve savings in the short term, waiting lists would need to remain unchanged and the provision of services throttled - also releasing capacity and cash.
Efficiency interventions can be applied either inside or outside the provider boundaries. To evaluate which interventions deliver most benefit for the local population case-mix, sophisticated Population Health Management (PHM) rules and impact modelling should be used. Each PHM rule represents a successfully implemented intervention within the NHS for which a specific patient group has been defined. An intervention should be considered if sufficient patients can benefit and activity and cost impact modelling shows no adverse reaction in the local health economy. By executing hundreds of PHM rules and testing hundreds of interventions, decision makers will obtain an accurate picture of their change opportunities to make informed decisions.
There are basically two settings in which interventions can break the readmissions cycle. Which are either inside or outside provider boundaries. Interventions that reduce readmissions inside the provider boundaries must prevent medically unfit patients from being discharged and may include (sample list below):
Interventions that reduce readmissions outside the provider boundaries must provide sufficient follow-up support for discharged patients and may include (sample list below):
There is a substantive body of evidence for many of those interventions listed that concludes e.g. that integrated care can deliver better outcomes for populations with long term conditions and reduce preventable hospital readmissions. Because of this large body, an automated PHM system is required that utilises clinical rules to identify population groups to guide and inform decision makers.
Posted on July 26th, 2015
This is part 2 of a 7 part blog on reducing Non-Elective Readmissions (NELR).
Part 2: What is the financial impact?
Spending on health in England in 2015/16 is expected to be �115.4bn of which �94.94bn is allocated to CCGs of which �13.55bn for Specialist Services, �12.3bn for Primary Care, �66.8bn for Secondary Care and �2.3bn for BCF and CCG Admin. Secondary Care can be broken down to �45.8bn for Hospitals, �6.9bn for Community, �11.7bn for Mental Health and �2.3bn for others.
Nationally there are approximately 280k Preventable 30 Day Non-Elective Readmissions (P30NELR) per year costing over �500m. Of those, 130k patients were unfit for discharge (�230m) and 90k patients had insufficient follow-up support (�160m) totalling to 220k patients costing �390m. The remaining 60k patients are more complex patients with clinical complications or social care requirements costing over �110m.
Two work streams are required to address those 220k patients being readmitted. The first one dealing with 130k patients that were unfit for discharge which requires providers to improve clinical practice and discharge processes � saving �230m. The second one dealing with 90k patients that were unsupported which requires the commissioner to implement better care coordination and integration � saving �160m.
The delivery of improving clinical practice and discharge processes for 130k patients would not cause an additional cost to the commissioner sine it would be delivered at cost to the provider. The delivery of better care coordination and integration for 90k patients would cause an additional cost to the commissioner since intermediate care teams and additional nursing staff would have to be commissioned.
The additional cost would be the following: 90k patients requiring better care coordination and integration could receive 4.2 days (33.6h) post discharge support totalling to 1,890 clinicians costing �56m per year (90k patients * 4.2days each / 200 days per FTE @ �30k band 5/6). By investing �56m a total net saving of approximately �104m could be achieved each year (�160m-�56m).
This equates for the average CCG with over 1,000 preventable readmissions to a potential saving of around �1.6m net by investing �250k to pay for 9 clinicians for discharge support per year.
Posted on July 24th, 2015
This is part 1 of a 7 part blog on reducing Non-Elective Readmissions (NELR).
Reducing Non-Elective Readmissions (NELR) improves patient experiences, avoids unnecessary procedures and reduces waiting times for other patients whilst saving costs. In particular - Preventable 30 Day Non-Elective Readmissions (P30NELR) - are often related to low clinical outcomes or insufficient care integration. Low clinical outcomes can be improved by e.g. continuous development, uptake of new technologies and use of high volume units. Insufficient care integration can be improved by identification of care gaps in pathways and patient education by implementing new models-of-care including Multispecialty Community Providers (MCP) or Primary and Acute Care Services (PACS). Reducing preventable readmissions offers commissioners and providers the opportunity achieve better outcomes and save money without capping provision of services.
Part 1: What is the readmissions problem?
Not all NELR can be prevented but approximately 18% of them are P30NELR where patients were unfit for discharge or where follow up services were insufficient. P30NELR doubles or triples the cost of the initial admission and blocks other patients requiring care - causing longer waiting times. They are also causing negative healthcare experiences which may lead to non-adherence to clinical advice and drugs which, if the condition does not improve, can lead to mental health issues in the long term.
Calculating NELR and P30NELR is not trivial because the clinical reference point is the initial or index admission. Often, NELR baselines are overstated when patients have a chain of readmissions and admissions are referenced against each other instead being referenced against the index admission. Calculation of the P30NELR baseline is also complex since this involves clinical algorithms that use 1,000�s of admission and discharge diagnosis and procedures code pairs. Without the NELR and P30NELR baselines, setting achievable targets and interventions for readmission reduction becomes an uninformed exercise.
Posted on May 22nd, 2015
The 30 day readmission rule introduced in 2011/12 is an incentive for hospitals to reduce avoidable unplanned emergency readmissions within 30 days of discharge. Section 6.3.2 in the 2014/15 National Tariff Payment System states that �Providers should not be reimbursed for the proportion of readmissions judged to have been avoidable�. Readmissions relating to maternity and childbirth, cancer, chemotherapy and radiotherapy, renal dialysis, organ transplant, young children, emergency transfers, cross border activity and where patients self-discharged against clinical advice are all payable to the provider.
The scheme was designed to encourage providers and commissioners to manage emergency admissions through planned discharges, preventative initiatives, and greater involvement of experienced clinicians. Commissioners must reinvest money they retain from not paying in post discharge services that support rehabilitation and re-ablement. Commissioners are also required to identify patient groups that would most benefit from those services; they must discuss with providers where this money will be reinvested and must insure coordination with other commissioning decisions.
Commissioners are required to set an agreed readmissions threshold and determine the amount that will not be paid for readmissions above this threshold. Setting a threshold requires measuring how many readmissions could have been avoidable, which is a challenge in itself. Separate thresholds can be set for readmissions following elective admissions and readmissions following non-elective admissions.
To perform this process efficiently, a shortlist of patients that experienced an �avoidable readmission� should be made available to the review team. In this list, each patient should be categorised by the provider where an action could have prevented the readmission. This will inform the commissioner where a service gap exists e.g. hospital, primary care, community, social services etc.
The disadvantage of setting a threshold is that if might put pressure on the provider to reduce readmissions but it does not accurately reflect clinical need of the patient or better outcomes. Instead a more advanced systematic solution should be used utilising algorithms that identify avoidable readmissions consistently, month-by-month, case-by-case that are not payable to providers. Also, it might not have been in the provider�s control where follow-up care failed to deliver or the patient did not adhere to the rehabilitation and an emergency readmission was required.
Setting thresholds are a budgetary solution to a clinical problem where a lot of time and money is spent in discussions about what is over the threshold and not payable. Instead a shortlist of patients should be compiled by clinical algorithms that are subsequently reviewed by a clinically led team to decide if the provider gets paid or not.
Readmissions are generally indicative of ineffective patient management and call the quality of care provided across the continuum into question. However, while many readmissions are preventable, some are clinically necessary or unavoidable. Our research at i5 Health shows that over 10% of non-elective readmissions within 30 days are on the same day, over 20% on the next day and over 50% after 7 days of being discharged. Considering the short timeframes after discharge, those readmissions are unlikely to be caused by support services outside the control of the provider and are more likely to be due to low quality care.
Readmissions within 30 days generally account for 12%-16% of all admissions whereby avoidable readmissions account for only 2%-3%. If avoidable readmissions can be reduced, capacity can be released at the provider so that more patients can be treated for, the provider will be paid and the healthcare event will be a much more positive experience for the patient.
Posted on November 9th, 2014
While the original QIPP programme was initially designed to cover 2010/11, current planning assumptions are based on a continued squeeze up to 2021/22 to reduce spending from a current level of 8% of GDP to just over 6%. To consolidate the �3.8 billion Better Care Fund budget for 2015/16, providers are required to cut emergency and elective activity to free an additional �1.8 billion of NHS allocations. The current funding constraints are more severe and sustained than ever. They come during a major reorganisation and cannot be addressed by reduction of quality (4h A&E, RTT, 12h admission, 2w cancer, etc.) due to CQINN targets. In consequence, large deficits are now forecast for trusts and CCGs where unsustainable measures are in place (e.g. freezing pay, PbR tariff reductions, prescribing). In the face of this, trusts and CCGs need to think and act more collectively about their local health economy and seek guidance on how services in all local healthcare settings are provided.
To this end, they should rely on information that is clinically relevant and can drive clinical change and is not based purely on statistics and variances. Information for QIPP/CIP initiatives that is based on variances and statistics does not take into account real people with clinical needs. It often informs clinicians that there may be a certain number of patients that have similar conditions or that there are too many referrals, attendances or admissions based on national or local benchmarks. However, such information alone seldom leads to successful initiative implementations due to the cross over from statistical to clinical information domains. For example, weather data should be used to predict the weather, not the number of birds in the sky � although there might be a loose correlation.
On the other hand, QIPP/CIP information based on clinical information including patient history, presenting acuity and severity, and future risk scores can be mapped against service definitions of initiatives successfully implemented elsewhere. Identification of patient cohorts based on clinical metrics that are in need of a local service leads to successful services and better patient outcomes. Since no information domain cross-overs occurred (clinical information for clinical services), the �language� and service definitions are consistent supporting implementation and change management.
The challenge here is that there are hundreds of successfully implemented QIPP initiatives within the NHS with different service specifications making it hard to find the ones relevant for your CCG or provider trust. Also, without the ability to monitor ongoing implementations, incentives to encourage improvements are difficult to manage in the short term. A system is required that helps CCGs and trusts identify, impact-measure and monitor initiatives before they are included in medium to long term strategic planning or piloted for testing. Such a system should also have the ability to monitor ongoing initiatives on a monthly basis. Furthermore, this system should use clinical information to find clinical services, calculate the total opportunity and also allow demand simulations for many different scenarios (not just best and worst case).
The King�s Fund has highlighted the fact that health economy or region-wide service changes can be planned and progressed via collective action by national organisations and academic health sciences identifying and modelling service reconfiguration. With the provision of clear evidence based on clinical metrics, greater clinical engagement and leadership can be achieved leading to successful change management. In other words, by addressing these clinical drivers, unwarranted variations can be tackled and sustainable levels of change based on suitable QIPP/CIP initiatives can be achieved.
i5 Health as developed the Commissioning Opportunities (COP) report and tool-set that enables identification of QIPP/CIP, Co-Commissioning and BCF initiatives based on clinical metrics, not statistics, using de-identified local or regional activity data. By applying hundreds of service specification rules to a CCG�s or provider trust�s activity data, i5 Health has that ability to match real patients to real initiatives providing actionable results that lead to real results.
Posted on September 9th, 2014
The key priority amongst the commissioning intentions for Prescribed Specialised Services for 2015/16 is collaboration in order to achieve the most efficient service models through delivering change. In other words, the challenge to identify the most efficient service models and deliver change may be best tackled in partnership between CCGs and Specialised Services.
But how should services and initiatives be prioritised and what is the impact of existing services and initiatives on specialist care? National initiatives such as Early Diagnosis, Care at Home, CfV, QIPP, etc. are delivering the NHS agenda to reduce hospital attendances and admissions and to constrain levels of spend to match available resources. The difficulty has always been the identification of the most promising initiatives for a specific locality, their successful delivery and the reduction of costs whilst improving outcomes.
Identifying correctly the most suitable initiatives based on real activity, case-mix and clinical parameters takes away the guesswork most commissioners are facing which often results in the non-delivery and uncertainty of QIPP. By utilising advanced algorithms that identify initiatives with the highest impact based on actual activity, commissioners now have a tool that provides certainty. The i5 Commissioning Opportunities (COP) report enable commissioners to focus on initiatives that deliver the highest impact based on clinical conditions of the local population. It creates cohorts of patients, which would benefit from a previously successfully implemented initiative within the NHS, to enable commissioners to find suitable initiatives from a catalogue of hundreds.
The upcoming Co-Commissioning arrangements between CCGs and NHS England will, for the first time since the restructuring, enable CCGs to implement Out-of-Hospital strategies that affect specialist and non-specialist patients. By focusing, though COP, on initiatives that can deliver healthcare services for patient groups, commissioners of specialist services and CCGs can achieve a reduction in healthcare costs that leads to a delivery of QIPP savings.