Hospital Admissions Statistics – A fair representation of the true position?


John Roberts


COVID-19 Actuaries Response Group – Learn. Share. Educate. Influence.


The UK Government provides, as part of its coronavirus dashboard, daily information on hospital admissions where the patient has tested positive for COVID-19. We believe this measure represents a reasonable indication of the level of infections in the community, assuming a constant proportion of those infected are admitted. This assumption does have drawbacks, as the age distribution of infections can distort the picture. Nevertheless, the measure has the obvious advantage that it represents those ill enough to require hospitalisation with the virus, and is a leading indicator both of deaths, and hospital capacity utilised.

Challenges have been raised that analysis of hospital admissions overstates the current position, either due to inpatients catching the virus, or because admissions for other reasons happened to have COVID too (the all-too-familiar “with COVID” or “from COVID” debate). So, does this challenge have merit, either in terms of materiality, or in terms of the relevance of the measure?

Hospital Acquired Infections

The NHS now provides a daily update of hospital admissions which includes a breakdown of those acquired “in the community”, which for clarity includes care homes, even though these are also identified separately. Community figures include those diagnosed in hospital where the test was carried out within 7 days of admission. It is thus possible to work out how many infections have been acquired beyond 7 days, at which point it is likely to have occurred within the hospital. It should be noted that there is clearly some approximation, and an exact figure is unlikely to be known.

Taking the North West region as an example, as this is currently the largest in terms of absolute numbers, with rapid growth in early October, we can see that the proportion of admissions suspected to be hospital acquired is volatile when numbers are very low, as might be expected. However, with increasing numbers the proportion settled above 10%, and has subsequently been slowly rising to a current level of around 18%.

Whilst this is undoubtedly of concern in terms of management of the virus within hospitals, it should not detract from the fact that most admissions are still from the community, and these have increased rapidly. It is also the case that if levels of COVID were much lower, both in hospitals and more generally in the community, the management of the virus would be much easier. It is thus the growth in community infections that drives hospital acquired infections.

“With” or “From” COVID

A frequent challenge to those reporting admissions data is that there is no evidence that these relate to people sick with COVID. They could equally be patients admitted for other reasons, who then happen to have tested positive as part of the standard infection control process of testing all admissions (we refer to these as “with COVID” cases).

In the first instance, we can discount patients presenting for elective surgery as contributing to “with COVID” cases. All such patients, even for day surgery, are asked to self-isolate beforehand and have a COVID test shortly before admission. A positive test results in postponement of the surgery.

Thus the only cases where an admission may turn out to be “with COVID” are emergency admissions, where clearly admission is the priority and only then is a test carried out. We know from NHS statistics the attendances at A&E that require admission (September was 16,000 per day), and we can estimate, using population levels of infection, how many will subsequently test positive. This can be compared with the overall admission levels at a point in time to estimate the proportion of “with COVID” cases.

Using the latest ONS data for age related infections, together with a typical age distribution of A&E admissions, we see that for the most recent 7 days matching the ONS data (10th to 16th October), out of 112,000 emergency admissions, around 600 might test positive for COVID. That compares with 4,700 COVID admissions reported for England, so around 12.5% to 13%. It is a moot point as to whether some of those emergency admissions seemingly for other conditions are indeed related to the presence of COVID. In the absence of any available data we discount it.


The fact that hospital acquired infections are occurring is clearly of concern, not simply because it inflates the figures, but because people, already ill and vulnerable, are being put at increased risk. It is perhaps inevitable that as infection levels rise in the community, such cases will increase, despite the best efforts of those involved to try and keep the virus under control within healthcare settings. As such, it is valid that such numbers are included in the totals, when understanding the impact of the pandemic on the health system.

The issue of “with COVID” cases is different, if it is clear that COVID is unrelated to the cause of admission. Here it is valid to suggest that the inclusion of these cases overstates the scale of the problem. However, in understanding the dynamics of the situation, it is the rate of growth of numbers that is of particular interest. Since we might expect both “from COVID” and “with COVID” admissions to grow in line with infection levels in the community, we do not believe that the rate of growth is distorted by the inclusion of the latter. Neither do we believe the level of overstatement is of a level to sufficient to materially distort the picture.

27 October 2020

About henry tapper

Founder of the Pension PlayPen,, partner of Stella, father of Olly . I am the Pension Plowman
This entry was posted in actuaries, coronavirus and tagged , , , . Bookmark the permalink.

1 Response to Hospital Admissions Statistics – A fair representation of the true position?

  1. Derek Benstead says:

    The graphic is not very helpful as it is. It would be better if it extended back at least to March, so we can see where we are now relative to where we have already been in March/April/May.

Leave a Reply