My latest DigitalHealth.net column: The Paul Score!

My latest column is out. Click on the link to see the original and comments. Copy of text below.

I was recently reminded of the story behind the “Bacon number” and the fact that all actors have one.

I’ve spent a lot of time talking to business intelligence people usually about how their lovely graphs and data mean almost nothing to me as a clinician as I’m not really bothered about describing what is, or has been I want to know where I can inject effective change that makes a difference to perhaps longevity. Don’t get me know I believe in quality of life but knowing Ethel is 95, has 14 comorbidities and that her blood pressure isn’t quite at target doesn’t get me thinking about recommissioning blood pressure services, it makes me want to ask her what she wants.

However knowing that some of my atrial fibrillation patients, who should be aren’t on an anticoagulant gives me a nice project to work on and trying to work out how much each person has to work on has led me to the proposition that every patient has what I’m going to call the “Paul number of modifiable risks”.

Spotting trends

In a previous column I looked at population health, big data analysis and the concept of segmentation. I still think it is fair to say, a lot of people (myself included) don’t really get population health, maybe this is why there were so many conferences on it advertised before the lockdown.

Population health is in its purest form is about understanding the non-medical determinants of health especially when planning things. For example, people need clean air, clean water, green spaces, places to exercise to be healthy not just doctors.

Big data is to some extent about being able to spot trends in linked data sets that would not have been found otherwise be found.

Population segmentation for me is about trying to find small enough groups that a small intervention team with limited resources can make a difference to. You could review everyone on medication, or you could have a team / project to review everyone on 10 or more medications.

GPs as gatekeepers

The job of a GP as gatekeeper is a difficult one. A lot of what I see is normal or self-limiting illness or mild disease, that either needs self-management or simple treatments and not referring on into the wider system where they get sucked into a merry go round of tests and treatments.

My job is to see lots of people and filter out the abnormal. However, it does feel sometimes like I’m not adding much value or extending lives. Of course, a lot of what we do is screening, vaccination and chronic disease management. I’d like to explore this measurability as I think it’s an area where we are lacking some tools.

Risk assessment 

Let’s take primary prevention of cardiovascular disease. This involves assessing a patient’s risk for developing cardiovascular disease and advising them on how to reduce it.

To help us we have something called the QRisk which calculates the risk of a patient having a cardiovascular event in the next 10 years.

The QRisk is based on a whole host of independent risk factors, that have been mathematically derived from real data.

Now some of these factors you can do little about, for example age, sex, presence of certain diseases, even post code can be a difficult one. However, some of these risk factors are modifiable, ie Smoking status, blood pressure control, control of lipids and control of sugar.

Now it strikes me that analysing the number of modifiable risks and working to reduce them is very measurable and co-incidentally very worthwhile.

The Paul Score

If we were to score 1 for each modifiable risk a person has, we could give each patient a score of how many risks they have – this is what I am calling the Paul score.

Now for simplicity, I’m assuming all risks are the same either in terms of added risk or in terms of ease of reducing (the version that takes these differences into account is called the weighted Paul score!).

A person with a Paul Score of 4 has four modifiable risks and is at more risk than someone with a score of 1. It should be possible to create an average score per patient per practice and look at whether this differs significantly. Does a high score indicate where resource should be targeted?

Practice variation 

I’m interested to see how much this varies from practice to practice and from area to area. It’s almost certainly linked to deprivation, but I wonder if there are practices with higher or lower Paul numbers than you would expect?

It also raises some interesting questions. Is it better reducing everyone’s Score by 1 by some project or are you better looking at the highest risk people and reducing their scores? Should a team target the patients with scores of 4 or above? Perhaps if it takes equal time to reduce a risk by 1 it doesn’t matter however if you can engage a patient and reduce their risk by 2 or 3 in the same time you are working very productively.

Looking at how many risks you have reduced perhaps gives you a measure of how much good you have done not just how many patients you have seen.

Links to Covid-19

Now I’m sure there are health informaticians laughing at this and able to point out all sorts of problems but when I see a patient, medically I’m often thinking are they are the blood pressure target,  are they at the cholesterol target, are they on beta blocker etc. Would this help prioritise or direct work?

Is this relevant to Covid? Well at some point there will be a vaccination programme and we know it will be an odd one.

For flu campaigns we get people in often in their hundreds to a clinic and jab them as quick as we can. This will be a disaster for Covid. It’s out there already, some will have it. We can score people based on the number of risks they have and the higher their score the more a priority there are to be vaccinated.