Building a Parkinson’s service: Demystifying data

by Dr Andrea Lindahl and Dr Lucy Strens Consultant Neurologists and Joint Leads for the West Midlands Parkinson’s Network

Introduction

Following our graduation from the Parkinson’s Academy in the mid-noughties, we embarked on our consultant careers full of vim and gusto, determined to revolutionise services for our local people with Parkinson’s. We were armed with facts, figures, references and the business plan templates from the Academy but, despite this considerable head start, found it difficult to know where to begin, whom to seek out, what data to obtain and then how to analyse the information we had gleaned. The aim of this article is to share our experiences so far and hopefully give new graduates the means to avoid some of the pitfalls we have encountered.

‘It’s not just what you know, it’s who you know….’

If you are new to the area and service (and possibly even if not!) it is useful to start by generating a map of services and people in your area. Where are clinics held? What capacity is there already for new and follow up patients? Are there Parkinson’s nurses and if so where are they? What therapy colleagues have an interest or expertise in Parkinson’s? Who are the local Parkinson’s UK advisors and Service Improvement Advisors? Who are the local Parkinson’s UK Branch chair people? Engaging the people that use the service is vital. A focus group at a local branch meeting of Parkinson’s-UK or a service users’ questionnaire about the ease of accessing the services available and the patient and carer experience can be incredibly revealing.

Generating a database

What quantitative information can you easily obtain about your local population? If you are extremely lucky, there may already be a database of patients or the community Parkinson’s nurses may be able to access data on their caseload. If these data are not easily available, an estimate of new and follow-up patient numbers can be made from prevalence and incidence data. This requires a knowledge of your local catchment population. Population data at CCG level is available from the Health and Social Care Information Centre (HSCIC)¹. There is no charge for the information for NHS employees.As an example: our Acute Trust provides Parkinson’s services to people in Coventry and Warwickshire. There are three local CCGs (Coventry and Rugby, Warwickshire North and South Warwickshire). The CCG populations (2013-14 figures) are 479,768, 184,562 and 274,878 respectively, totalling 939,208. PD incidence is estimated at 15-20 per 100 000 population per year. Prevalence is estimated at between 1 in 500 and 1in 625. In Coventry and Warwickshire, we might therefore expect between 141 and 188 new patients per year and a total prevalence region-wide of between 1,503 and 1,879 (i.e. number requiring follow up appointments). This prevalence figure tallies nicely with our actual database figure of just over 1,500 patients currently known to the service.When assessing likely new patient capacity needs, it is important to consider how many patients you may be referred that are out of area and also the number of people referred with possible Parkinson’s that turn out to have something else. The demand for new clinic slots will also depend whether your clinic is purely for Parkinson’s or attracts other movement disorder cases. When you are in a position to know the strengths and weaknesses of your existing service, any improvements or developments you want to make will be dependent upon working in partnership with your local acute trust, with local partnership trusts (e.g. mental health services, palliative care services) and with local Clinical Commissioning Groups [CCGs] (who currently commission, and therefore pay for, services). It is important (but not always straightforward) to establish who the key managers are in these areas. Many CCGs have a long-term conditions lead and this may well be the key person with whom to build a relationship. However, before meeting any person in power it is essential to arm yourself with data that will (hopefully!) persuade them towards your way of thinking. For those of us based in the acute sector, we are largely dependent on Hospital Episodes Statistics (HES) data.We have found the generation of a database very helpful. As it takes time to input and update data, this has had to be kept simple but even a list of current patients, which consultant and nurse they see and where they live (analysed by their postcodes) can give very useful information on the demand on the service, the caseloads of community Parkinson’s nurses and the geographical spread of patients. Locally, we have identified a particular gap in capacity in one geographical area, resulting in patients having to travel long distances to attend clinics. We now have plans to institute a new consultant-led community clinic in that area.

Hospital Episodes Statistics (HES) Data

The first thing anyone will tell you when you mention HES data is that the data are fraught with inaccuracy. This is, no doubt, true. However, it is the basis of how CCGs are charged (the ‘tariff’). It is also the only data set that most trusts have which will give any information on numbers and costs of admissions. In order to demystify HES data it is important to understand the jargon. This is perhaps best understood using an example.Scenario A: Mrs Johnson, who has Parkinson’s, attends the emergency department of Trust A having had a fall. She is admitted to the medical assessment unit (MAU) at 10 pm. By the following morning she is found to have sustained no serious injury, she already has a package of care in place and she is discharged at 2 pm.In this case, the admission (known as a ‘spell’) started at the point she was transferred to the MAU. She did not move wards at any stage, therefore the ‘spell’ consisted of one ‘episode’. As she stayed overnight (defined as past midnight), her length of stay would be recorded as one day. Had she been discharged before midnight on the same day, this would have been recorded as a zero bed day stay.Let us consider a more complicated example:Scenario B: The same Mrs Johnson, one month later, is readmitted with a more serious fall. She is brought in by ambulance to the emergency department and requires admission to an orthopaedic ward as she has sustained a fractured neck of femur. She spends 10 days on the orthopaedic ward then suffers an aspiration pneumonia which requires a two week stay in intensive care. She is transferred to a medical ward for her ongoing management and further discharge planning and is discharged to a nursing home five weeks later.This spell now consists of three episodes: her stay on the orthopaedic ward, her stay in ITU and her stay on the medical ward. She will also be classed as a readmission as she presented with the same complaint within a 28 day period.The system by which a CCG is charged for an admission is incredibly complex. Acute trusts get paid significantly more for their planned activity (elective and day case) than their unplanned work (non-elective and emergency). As we all know, however, non-elective work forms the bulk of what trusts do and often at the expense of planned activity due to bed occupancy rates. It is in everyone’s interest (not least the person with Parkinson’s) if non-elective admissions are kept to a minimum. This requires an anticipatory as well as responsive model of care, detailed discussion of which is outwith the scope of this article.By complex means, hospital admissions are allocated something called a ‘trim point’. This is defined as the length of stay (LOS) at which 25% of the admissions have a higher LOS (third quartile) plus 1.5 times the difference between the third and first quartile². If this is confusing, don’t worry, the essence is that the trim point is a means by which excessive lengths of stay can be calculated. As far as the money goes, any LOS that is below the trim point is paid at the basic tariff. Any LOS exceeding the trim point is paid at the excess bed day tariff (bad news for the acute trust as this is generally less than the basic rate). Safely minimising lengths of stay is therefore not only good medicine but attractive to the acute trust.By contrast, thinking back to scenario A: had Mrs Johnson been discharged before midnight on the day of admission to MAU, she would have been classed a ‘zero bed-day stay’. This admission would have attracted the full tariff for an admission despite the fact she was only in hospital a few hours. This is clearly bad news for the CCG who will be keen to work with you to develop community-based or anticipatory models of care that prevent these brief admissions.In scenario B, Mrs Johnson was readmitted within a 28-day period. In 2011, the Department of Health introduced a system whereby local commissioners would not pay for any emergency readmissions to hospital within 30 days of discharge from a previous hospital stay. In the NHS confederation briefing paper in 2011³, it was estimated that this would reduce hospital annual income by around £790 million. The driver behind this initiative was that patients have a right to expect proper care in the first instance with the necessary level of support in place after discharge. However, the trail of readmissions can be complex and may not involve the same trust. What is certain is that any hospital trust will welcome any initiative you can generate that will keep 28 day readmissions to a minimum.

International Classification of Diseases Codes

Having located the appropriate data manager in your employing organisation, what questions do you ask? How do you identify how many people with Parkinson’s have been admitted to the local acute trust and what other data are useful? HES data is dependent on hospital coding. As you will all know, when a patient is discharged, the cause of their admission, co-morbidities and procedures are coded. The combination of these codes and the lengths of stay for each episode will ultimately generate the Healthcare Resource Group (HRG) code that will determine what the local commissioners will pay to the trust for the admission. These codes are determined by the International Classification of Diseases codes (ICD) and if you crunch data for long enough you will become very familiar with these! The code for Parkinson’s is G20X. However, if a patient with Parkinson’s is admitted for a reason other than their PD, G20X may not fall in the primary code (codes are determined hierarchically). For example, aspiration pneumonia, urinary tract infection or fracture may be the primary reason for admission. In this case, G20X may appear further down the coding. If you are trying to identify all patients with Parkinson’s admitted in a defined period for whatever reason, it is important to ask your data manager to search G20X in all coding domains, not just the primary code. You can then work out by looking at the primary code why the patient was admitted. Thus we are dependent upon the accuracy of coding, which we all appreciate is suboptimal.Another potentially huge pitfall is the coding system itself. For example, there are many ways of coding a pneumonia (e.g. J180 bronchopneumonia; J181 lobar pneumonia; J189 pneumonia, unspecified; J22X unspecified acute lower respiratory tract infection; J690 pneumonitis due to food and vomit). If you were interested in looking at the number of admissions with pneumonia, for example to form a business case to support earlier swallowing and nutritional assessment in complex Parkinson’s patients, it would be essential to take all of these codes into account, not just the ‘pneumonitis due to food and vomit’ J690 code.In order not to get completely confused by an overwhelming array of data, it is important to think carefully about what you are trying to find out. After many iterations, we have identified a core set of data that we analyse annually, shown in the table below.Length of stay can be difficult to interpret. If a significant number of your patients are admitted for very brief periods (the 0, 1 and 2-day stays) then your average length of stay will be skewed downwards. It could be argued that those patients coming in for two days or less could potentially have been dealt with in the community and an admission avoided. If you look at the reasons why these patients are coming in, you may be able to put anticipatory pathways in place that may reduce these short stay admissions. Your local CCG will be very co-operative as these admissions are expensive for the CCG.If lengths of stay are particularly prolonged, then the inpatient pathways need to be examined. Why are patients staying in hospital? Is it that they have complex co-morbidities, did they develop in-hospital complications or are they difficult to discharge due to issues with social care? Length of stay by discharge location in our trust showed (unsurprisingly) that the longest lengths of stay were those patients discharged to non-NHS institutional care. The shortest lengths of stay were in those patients discharged to their normal place of residence. Does this reflect issues with social care or are the first group of patients inherently more complex to manage? Often data analysis will create as many questions as it answers.HES data will also give you information on the number of people with Parkinson’s who have died in hospital. In many cases these patients will have had a number of admissions in the year leading up to death. It is impossible to say how many of these patients might have preferred to die at home but consideration must be given to advanced care planning, allowing patients to outline their preferred priorities for their own care.

Other sources of data

We must remember that HES data only records details of patients admitted to a hospital ward. It does not, therefore, take into account any visits to hospital that do not result in an admission. Emergency Department data is coded differently, usually by presenting symptom (e.g. back pain, syncope, headache) and not by underlying diagnosis. It is therefore challenging to obtain data on the number of ED visits people with Parkinson’s might make in any given time period.You might be in the lucky position of working at a trust where outpatient clinic diagnoses are coded. In this case, information on the numbers of people with Parkinson’s being followed up in clinics within that trust is easy to obtain. Otherwise, you may have to rely on counting numbers of patients seen in specific Parkinson’s clinics, thus potentially missing any people with Parkinson’s seen in general clinics.Both of the above problems could be overcome to some extent by performing audits. An audit could at least capture, for example, numbers of Parkinson’s patients seen in the Emergency Department or the number of Parkinson’s patients seen in clinic for any given time period. This might help when analysing the demand on your Parkinson’s service.In addition to trust and CCG data, pharmaceutical representatives can be a very useful source of information, both qualitative and quantitative. They will usually know other Parkinson’s specialists in the area and can facilitate your networking. They are also often in possession of HES and other data that is difficult to access at regional level.

How can data analysis drive change?

Our data showed (in line with published studies^4,5) that common reasons for people with Parkinson’s to be admitted were urinary tract infections, constipation, falls, fractures and chest infections. In response to this, members of our Parkinson’s Excellence Network (PD nurses, continence advisors, therapists and PD specialist doctors) developed leaflets for patients, carers, care home staff and GPs advising how best to prevent and manage urinary tract and chest infections, dehydration and constipation. Attending to oral fluid intake, toilet hygiene, dietary modifications and, more often than not, daily use of laxatives can help prevent some of these common admissions. Optimal management of the bowels also helps with medication absorption and is therefore critical to managing motor control.We are also looking at earlier referral to therapy services (ideally referral to physiotherapy at diagnosis) to try and prevent falls and fractures and to promote the benefits of exercise. New patient programmes, lifestyle management programmes and self-management programmes are being developed across the region. Earlier recognition of swallowing problems and referrals for swallow and nutrition assessments may help prevent aspiration. Better medications management in hospital and education of non-specialist staff (in our case largely driven by our tireless trust PD nurse specialist) should help reduce lengths of stay and improve the patient experience. Similarly, written management guidelines and regular education of junior staff also contributes to better medications management in hospitals. In our region, a number of trusts have developed an ‘alert’ system to highlight (via email) to the PD team when a patient with PD is admitted. Early review by the PD team can then ensure that the patient is written up for, and receiving, the correct formulation of medications at the correct time, are not being given contraindicated medications, have been referred to the relevant therapists and that attention is being paid to postural blood pressure, bladder, bowels and cognition.

‘No man is an island….’

Small changes can lead to big results. However, any change is difficult to implement without a team of enthusiastic professionals. It is also impossible to implement meaningful change without developing a relationship with your local acute trust management and the CCG. But where to start? The UK Parkinson’s Excellence Network was launched in February 2015 and is a joint collaboration between Parkinson’s UK and professionals with expertise in dealing with Parkinson’s. The aim of the Network is to enable professionals to work together for change (sharing ideas so we are not constantly reinventing the wheel), building up an expert workforce (by disseminating examples of good practice and ensuring that no-one need work in isolation), equip professionals to drive change and strengthen the voice of those affected by Parkinson’s. Your region will have an Excellence Lead and for details email excellence@parkinsons.org.uk. The current leads are shown in Figure 1. The regional lead should have access to a network of medical and allied health professionals who are developing services in your area.Once you have identified medical, nursing and allied health professionals that are treating people with Parkinson’s, a regular team meeting involving trust and community staff is very beneficial. We have found that holding three or four meetings a year breaks down some of the barriers between hospital and community and facilitates multidisciplinary working.It is also critical to ensure that links are made with wider members of the multidisciplinary team involved in complex Parkinson’s, for example palliative care, gastroenterology and the nutrition team are valuable allies.

In conclusion

Effecting change in the NHS is not always easy but it is definitely possible. The changes do not have to be costly or involve drastic reorganisations in service to result in significant changes to patients and carers. When meeting with managers in the acute sector or in the community it is essential to be able to back up your arguments with figures and, if you can demonstrate that your ideas could save money, you will certainly get their attention. In our view, getting things right for patients and carers, anticipating and preventing problems, keeping inpatient admissions to a minimum, and ensuring that care towards the end of life remains patient-centred, will not only improve the quality of the service we are offering but are likely to be much more cost-effective in the long run than our current reactive and often poorly accessible services.Figure 1: Leads for the UK Parkinson's Excellence Network (with permission) (Click to enlarge)

References

1. http://www.hscic.gov.uk/psd
2. Jones, R (2009) Limitations of the HRG tariff: The trim point. British Journal of Healthcare management 14 (11): 510-513
3. www.chks.co.uk
4. Woodford H and Walker R. Emergency Admissions in idiopathic Parkinson’s disease. Mov Disord. 2005 Sep;20(9):1104-8.
5. Low, V et al. Measuring the burden and mortality of hospitalisation in Parkinson's disease: A cross-sectional analysis of the English Hospital Episodes Statistics database 2009-2013. Parkinsonism Relat Disord. 2015 May;21(5):449-54. doi: 10.1016/j.parkreldis.2015.01.017. Epub 2015 Feb 17.