Early warning, late realisation

Early warning scores are in use throughout NHS hospitals to spot patients who are deteriorating. So where did they come from, and what are their limitations?

At Barts in 1997, it was noted that patients admitted to ITU from the wards had a high mortality.   During the summer and autumn of 1997, a new ‘patient at risk’ team (PART) was set up at Barts in Whitechapel. Their role was to assess ward based patients with a view to fixing things before they got worse, and thinking early about potential ITU/ceiling of care decisions.

Patients were selected for review by the PART based on a scoring system, where more deranged physiology led to a higher score. A certain score would trigger the nurse to call the ward doctor or the PART directly.

Noble stuff. Did it help? Of the 97 patients admitted to ITU over this six month period, 28 were seen by the PART within the preceding 48h. The incidence of CPR among this group pre-ITU admission was 3.6% compared to 30.4% among those patients not assessed by the PART (p<0.005). Mortality once at ITU was 25% for the PART group vs 44.9% for the non PART group, although this did not quite reach statistical signficiance (p=0.07).

Over time, various warning scores were developed, ranging from single parameters to aggregated scores. There is now a drive by the RCP to encourage NEWS, which is a nationalized early warning score.

What are the limitations? The National Patient Safety Agency published a report in 2007 which in part assessed how the early warning scores were working. 16 clinicians were interviewed, including 5 junior doctors. It’s worth reading this – key problems were communication, observations being seen as low priority and trigger fatigue (too many bleeps for “Just thought I’d let you know Mr Peters is scoring a 5…”)

From my own experience, there are three things I wish I understood better at the start of FY1 about using early warning scores.

1.    Communication

Personally, I sometimes spent more time trying to write a comprehensive and totally unambiguous management plan than actually with the sick patient. I was often confused when I then got bleeped asking what the plan was.

This was because I assumed that nurses will read the medical notes. It only dawned on me that nursing notes and medical notes are kept in separate folders about two months into my first placement. Lesson learnt: spending 30 seconds explaining the plan to the nurses saves you twenty minutes from having to revisit the patient when your plan wasn’t followed.

2.    Not all unwell patients score

There are some sick patients who don’t always score very highly on early warning scores. This includes patients with meningitis, patients with raised ICP, stroke, sepsis (especially neutropenic sepsis) and overdoses. It is worth mentioning that the elderly have a lower baseline temperature than the young, and that not only is the febrile response is blunted in 20-30% of elderly patients with acute infection, but a blunted response is associated with a poorer prognosis.

A patient’s change in their observations from their baseline is a better measure than the early warning score of how they are doing. A patient who is normally hypertensive with a BP drop from 175/100 to 100/65 in two hours doesn’t score on most systems, but this could lead to end organ hypoperfusion in this patient.

3.    Some not so unwell patients score

Lots of hospital inpatients will have a raised pulse from simply pain and anxiety. Many patients with well controlled COPD on home oxygen may score highly for oxygen and low sats on admission to hospital despite being at their baseline for these criteria.

Again, a patient’s change in their observations from their baseline is a better measure than the early warning score of how they are doing. A stable COPD patient whose sats have been between 88% and 92% for the past 2 days may move from 90% to 89% and suddenly score enough to trigger a colour change to amber, but this doesn’t mean there is a new acute problem.

What does this mean for the junior doctor on call?

Early warning scores are most effective at selecting a select group of patients that the nurses should generally seek a doctor’s opinion on. It may be end up being nothing serious; this is not a reason to get angry with the nurse. Your clinical assessment supersedes the score as the measure of the severity of the problem. For all subsequent dealings with this patient, you should use your knowledge of the situation together with the trend in observations to manage accordingly; you should not be using the early warning score alone to decide if the patient is improving or deteriorating.

Talking directly to the nurses after you have seen the patient will save you time in the long term. Unwell patients and the early warning score are far from perfectly correlated. Whilst it is true that for the whole group of hospital patients there is a correlation between early warning scores and general unwellness, there are specific disease processes that are not well matched to the early warning scores, and an individual’s variation from baseline supercedes the early warning score as a measure of how that individual patient is.

The early warning score is undoubtedly useful, but we need to remember it is a screening tool, nothing more.

The final week

I now have some idea of what sorts of things come up regularly, and also what I need constant reminders about. For example, medicine.

I’m going to post stories/summaries I found helpful, and turn them into proper diagrams after MRCP. These are NOT comprehensive summaries. They are overviews to trigger my memory.


Toxoplasma gondii = an intracellular protozoan parasite. It lives in the cat.

In most people, it doesn’t do much. Occassionally it causes a couple of lymph nodes to enlarge, esp cervical/occipital. If you are really unlucky and the cat gave you an evil eye, you get chorioretinitis.

In pregnant women , the baby may get congenital toxoplasmosis, especially if mum is infected in the first 10 weeks. Watch out for CHC: Intracranial calcification, Hydrocephalus and Chorioretinitis.

In immunofail people, cerebral toxo is the big issue. Multiple, ring enhancing lesions, best seen on MRI.

IgG stays positive for life. Maybe you can use the affinity of IgG to work out if there’s an acute infection. Or maybe you could ask the Vengeful Cat God if they are infected, bringing a suitable supply of mice as an offering. IgM rises acutely, but returns to normal when it feels like it.

Cerebrospinal fluid (CSF) amplification DNA for toxoplasmosis gives the confirmatory diagnosis

Tx is co-trim for AIDS, and pyrido/sulfid for others.

Acute HIV

About 70% of people with HIV infection experience symptoms during seroconversion. It starts 2-6 weeks after exposure.

Think of Infectious Mononucelosis, but with more emphasis on a maculopapular rash that affects the upper body plus mouth/genital ulcers. Almost like a marriage between IM and Behcets.

FBC may show low platelets.

Ix: Bite the bullet and order HIV antibody and p24 antigen tests once the patient has been counselled and consented. The window period is now just one month with this cool 4th gen approach. Don’t use viral load for the initial test as you risk false positives. Get some confirmation assays if these are positive, and establish if its HIV-1 or HIV-2.