'In the absence of an adequate system to record and monitor the numbers and circumstances of deaths, the detection of Shipman's high numbers of deaths was dependent on the chance of observations of individual practitioners or medical referees. . . they [medical referees] only have their own memories to help them detect patterns and numbers of deaths'.

1The data required to produce the startling statistics in Professor Richard Baker's report was really very limited, and it is tragic that it was so difficult to retrieve and collate the information. The fields required were the patient's name and age, the GP, the date and cause of death from the death certificate, and the time and place of death and persons present from the cremation form.

A relatively simple analysis of this information would have highlighted the enormous differences between Dr Shipman and other doctors in the area.

The single fact that he issued more than twice as many death certificates as any of the other GPs in the control group would have triggered alarms, and there can be little doubt that if computerised records were available it would be far easier to spot such anomalies.

It is perfectly possible to use computers to collect data reliably and to use it effectively. There is a mass of information stored on every one of us: information is collected by individual agencies and organisations and then distributed and collated with data from others.

The NHS has shown that it is able to collect data of reasonable quality. To date, this is primarily driven by the largely clerical requirements of the Korner central returns. However, we are on the verge of a host of central initiatives to collect and use a much broader range of clinical information details towards creating electronic patient and health records. This is also driven by the move towards formal clinical governance following Bristol and by specific initiatives such as the collection of cancer data following the Calman-Hine report. The key now is to ensure that as the demands to capture and disseminate information soar, the pitfalls so evident in the Shipman report are avoided.

Take POSSUM, for example. It is a methodology for measuring the outcomes of surgical procedures on a risk-adjusted basis. Basically, it uses measures such as age, cardiac and respiratory signs, systolic blood pressure, ECG and blood results in order to calculate the likely morbidity and mortality rates for operations. When this information is combined with operative data such as the severity of the operation, malignancy and peritoneal soiling and with complications it provides an audit tool which allows comparisons between surgeons and against POSSUM calculated norms. It does this by compensating for patient fitness and operative severity when analysing outcomes. In short, it does precisely what is required for clinical governance.

The implementation of POSSUM, however, highlights the classic difficulties in data collection and quality. Despite the small amount of data being collected, users have reported difficulty in collecting complete and accurate data, and the POSSUM scoring system is very sensitive to data errors.

Data-quality problems of this type are often due to the use of paper data collection systems - a piece of paper has no intelligence and is not able to validate information. Professor Baker found many instances of missing and incorrect data.

The key to this problem is to use computers to collect information as close as possible to the point of patient contact.

This means that when data is entered, the details are fresh in the mind of the operator and the information is validated on entry.

Also, the data entered should always be a by-product of an operational requirement. In the case of POSSUM, entering the 12 pre-operative measures could automate the production of the consent form. In the case of a death certificate, entry of the data once could print the form, print the burial or cremation certificate and transmit the data electronically to where it is needed.While this method will not prevent the deliberate falsification of data, it will at least provide a single consistent set which will be used to generate a number of outputs which are likely to be reviewed by others as they do their jobs.

Once accurate information is collected in this way, the service will reap all the benefits. Many of Dr Shipman's patient records were found in a cardboard box in his garage and in a plastic bag on his dresser.

REFERENCE

1 Harold Shipman's Clinical Practice 1974-1998. Clinical audit commissioned by the chief medical officer. 2001.