Is diabetes going undiagnosed?

Monday February 25 2008

“Some 60,000 people may be living with diabetes without realising they have it,” said the Daily Mail and several other news sources. They report on a large survey of GPs electronic health records, that found that thousands of people have blood sugar levels that suggest diabetes, but remain undiagnosed. The Guardian reports that “more than half a million people could have diabetes or a high risk of developing the condition without knowing it”.

This study examined an electronic database containing the health records of more than nine million people registered with GP surgeries in the UK. Although there are some limitations to the study, the findings are based on a large amount of reliable information. The results may prompt a debate about electronic health records in the UK and systems that could ensure that all blood results are being noted and acted upon where necessary.

Where did the story come from?

Dr Tim Holt from the University of Warwick and colleagues from the University of Nottingham, Imperial College London and EMIS, a commercial supplier of IT and electronic patient records systems for primary healthcare, carried out the research. Some of the researchers worked for QRESEARCH, the UK healthcare database that was used to conduct this study. The authors acknowledge that publication of the paper could lead to increased awareness of the scope of the database for research purposes.

The study was published in the peer-reviewed: British Journal of General Practice.

What kind of scientific study was this?

Previous research has suggested that about one per cent of people in the UK have undiagnosed diabetes and are therefore not receiving the required management and follow up care.

In this cross-sectional study, the authors carried out a survey of an electronic healthcare record database to come up with an estimate of how many people in the UK have evidence of diabetes, but remain undiagnosed. They hope that this will help to develop simple database search techniques within practices that will help to recognise early cases of diabetes.

The database contained the health records of more than nine million patients from 499 general practices in the UK. The records included patient’s personal details, GP consultations, clinical diagnoses, investigation results and prescription medications.

The researchers only included practices and their patients if their data were available around the date of the search;  June 1 2006. The researchers identified and categorised patients who had been diagnosed with diabetes or impaired glucose tolerance, had had “normal” results from a glucose tolerance test or if they had ever had a random or fasting glucose test. The patients were categorised with “Read” codes, codes that are used to transmit results to practices from hospital laboratories.

The researchers then used two separate “search strategies” on this group. Search strategy A included patients whose last random or fasting blood glucose test was above the recognised World Health Organisation (WHO) cut-off level for diagnosing diabetes. The second, strategy B, had a lower cut-off for random blood glucose tests and was therefore more inclusive.

The researchers then excluded patients who had been diagnosed with diabetes or who had been further investigated by a glucose tolerance test (positive or negative).

What were the results of the study?

There were 480 GP practices and 3.63 million people included in the search. There were 128,421 people with diagnosed diabetes, equivalent to a prevalence of 3.54%. Following the exclusion of those with diagnosed diabetes, and those whose diagnosis had been excluded or resolved, the researchers were left with 3.49 million people. Of these, about 30% had had their blood sugar checked on at least one occasion, with older people more likely to be checked. One-fifth of all of these people had had measurements taken within the past two years, and in most cases, it was unclear whether the glucose test had been random or fasting. The researchers assumed that all instances where the type of test was unclear were random.

When the researchers used Search Strategy A, 3,800 people (10% of the study population) were found to meet or exceed the requirements for the WHO definition of being diagnosed with diabetes using random or fasting glucose. Only 1.3% of these people had been tested and had diabetes excluded as a possible condition.

When Search Strategy B was used, (the lower fasting limit, regardless of matching to the actual blood test) the researchers found 33,057 people with blood sugar above this level (90% of the study population). Only 1% of the people had been tested and had diabetes excluded as a possible condition.

This gave an average per GP practice of eight non-diagnosed Search Strategy A patients and 68 non-diagnosed Search Strategy B patients.

What interpretations did the researchers draw from these results?

The authors conclude that a significant proportion of the UK population have had their blood sugar measured at some point, and that those who have undiagnosed or borderline diabetes can easily be identified using an electronic database system. They say that “all but one of the 480 practises in the study’s sample included people whose most recent blood glucose level probably requires further follow-up according to current guidelines.”

What does the NHS Knowledge Service make of this study?

This study is based on a large amount of reliable information contained in electronic databases in the UK. However, conducting a large cross sectional survey that relies upon codes to identify diagnoses and investigation results may introduce some recording or measurement errors.

  • As the authors state, only a relatively small proportion of practices coded whether the sample was fasting or random. Similarly, the more inclusive parameters of Search Strategy B would incorrectly consider many patients as having undiagnosed diabetes when they did not.
  • The time period during which the patient’s last blood sugar measurement was taken varied from within the past two weeks to over two years previously. For blood tests taken recently, it cannot be assumed that the patient would remain undiagnosed and untreated. As the authors acknowledge, “It was not possible to determine the proportion of patients identified who were subsequently diagnosed with diabetes or impaired glucose regulation.”
  • None of the background information to these tests or preceding consultations is available. It is possible that some of the GPs acted upon their patient’s slight increase in blood sugar (for instance, by giving the patient dietary advice) and just not coded it in a way that could be recognised by the search. Likewise, the slight raise in blood sugar may have been expected due to another medical condition.

However, despite its limitations, this study highlights the potential for using electronic databases for the simple recognition of people who would benefit from further care and follow up, or may have been missed.

Sir Muir Gray adds...

Things are getting better; five years ago, the headlines were “the missing million”.

Analysis by Bazian
Edited by NHS Choices