The minimum data set

At last we have a minimum dataset for mental health care. At the last meeting of the committee which determines information requirements in the NHS, a dataset was finally agreed which describes the work of specialist mental health services. The official notices announcing the dataset were posted in December 1999. After a long and agonising gestational phase, this dataset should transform clinical audit and service planning in mental health services throughout England. This article describes the principles underlying the dataset, how the data will be gathered and the types of question it should be able to answer.

Principles

A minimum dataset is a standard set of data items which it may be assumed all NHS trusts will be recording about all patients treated in a particular area (Korner, 1983). National standardisation of the definitions used allows comparisons between trusts and the production of aggregated national statistics. National datasets have included information about mental health care for many years, however their usefulness has been limited as they have been restricted to individual types of service input (in-patients, out-patients, community psychiatric nurse contacts etc). Mental health care is essentially a multidisciplinary affair. To make sense of what a service is doing, all these elements, along with related data about social care provided, need to be joined up around the patients receiving the care and combined with details of their problems and progress.

A dataset needs a basic recording unit. In-patient statistics are currently reported in terms of ‘consultant episodes’ (a period of time an individual spends under the in-patient care of a particular consultant). In mental health care, patients may have short episodes of clinic attendance seeing only one individual, but many will be assessed by a team and subsequently receive several types of intervention, concurrently or consecutively. The dataset conceptualises this as a ‘mental health care spell’. This begins with the patient's first contact and finishes at the point they are signed off with no further intervention envisaged.

For practical purposes, reporting has to relate to a definable time interval. The dataset thus defines a ‘reporting period’. This will normally be a year for national data, and a quarter for local data use. The dataset includes a record for each mental health care spell falling partly or wholly within the reporting period and comprises a summary of the care received by the patient and their progress within the reporting period.

How it's done

Talk of datasets usually conjures up visions of mountains of new paperwork, sending clinicians either to the tranquilliser bottle or to the adverts for alternative job opportunities in the private sector. This is different.

In designing it, I started by identifying all the types of data currently recorded by health trusts and social services departments for statistical returns and local clinical reasons. I undertook a detailed survey of the ways these are currently recorded. Then, working with information department staff in several trusts, I set about devising a way in which the information from all these sources could be pulled together. This entails trust information staff extracting the various data elements from whatever systems they are collected in, and using a very simple (custom-written) dataset assembler programme to build them into patient-based records.

This type of strategy (called data warehousing) is well known in many commercial fields such as telecommunications, banking and retailing (Kelly, 1996). It is relevant in situations where making sense of the enterprise requires integration of information from several sources, and where data collection is likely to raise complex practical issues better addressed independently.

The only area where the dataset envisages data collection not currently required for routine returns is in the field of the Care Programme Approach (CPA). Here I worked with the team developing the new guidance Effective Care Co-Ordination (NHS Executive, 1999a ). The main justifications this sets out for data collection relate to the need for clinicians to be able to manage case-loads and pick up the details of patients needing urgent intervention. The CPA information systems it proposes are essentially simple electronic patient records.

What's in it for you

The dataset produced is intended first and foremost to provide rich, clinically-focused information for local clinical governance and strategic planning by health trusts and their corresponding social services departments. From epidemiological studies we know a lot about the numbers and distribution of people with mental health problems likely to be found in any district. From therapeutic research we also know about the treatment packages likely to benefit them and the likely time course and outcome profiles of their problems.

The dataset should allow us to compare the extent to which those people are reaching specialist care and the clinical outcomes are being achieved in actual practice. It also gives a detailed picture of the deployment of staff and other resources in their care. It includes a detailed set of geographic markers allowing the data to be easily displayed on maps of trust, primary care group or social services catchment areas. The custom written dataset ‘reader’ software provides highly flexible crosstabulation facilities, linking numbers of patients seen to relevant population data. It is also able to draw a number of graphs of the course of service use, such as the time from hospital discharge to readmission or to first post-discharge contact. All these tools will be free NHS software, using the standard Microsoft Access format for data storage in order to make it as readily available for further local analysis (including addition of special locally collected data) as possible.

When can we have it?

On the basis of the survey work undertaken, rolling out the dataset to all mental health trusts in the NHS in England is expected to take 3-4 years. The National Service Framework (NHS Executive, 1999b ) sets out the agenda for an information strategy for mental health services. This will encompass information for patients, the knowledge base for health and social care practitioners and electronic patient records as well as information to support clinical management, of which the mental health minimum dataset will be a key element. The detailed strategy is in preparation and a draft should be published in spring 2000. The National Service Framework itself sets a target implementation date for the dataset of March 2003. The roll-out plan envisages one or two trusts in each region implementing it in the first year, the bulk in the second, with those busy either with the process of implementing major new systems or with major structural re-organisations, such as trust mergers, following in the third.

The main cost to trusts will be in terms of work by local information staff in setting it up. Because of the way it is designed, it is not expected to require investment in new information systems (although the more modern the systems locally available, the easier will be the local task). Thus, the principal resources the NHS Executive is making available are the technical help of a data specialist to assist trusts in the task of setting up the data extraction routines required, and consultancy from clinicians experienced in data-based audit work to assist in its use.

For the first time, the dataset should give us a detailed picture of what English mental health services are doing in terms of the patients receiving care. By presenting a much more detailed view of the pattern of strengths and weaknesses in the care system this should help us to plan, and to argue for resources much more affectively in the 21st century.