Why We Need A Better Way Of Measuring Demand Within Primary-Care

SQUARING THE CIRCLE:

Why we need a better way of measuring demand within primary care

By Peter Milmer, August 2023

In our latest Perspectives blog, Optum consultant and practising GP Dr Peter Milmer describes his experience of measuring and applying data-on-demand flows across his practice to draw out new ways of supporting patients and releasing time to care.

I recently published an article on the importance of the personal connections which take place within healthcare – in our pursuit of efficiency, we must never lose sight of the deep human interactions that go to the heart of what general practice is about.

On the flip side, we won’t be able to adequately meet these basic human responsibilities unless we get on top of the demand and capacity challenges all primary care services face. In my experience, population health management is a good way of squaring the circle.

Demand vs activity

First, though, we need to be a little more precise about how we define what demand really is. In simple terms, demand is all about a patient or a professional colleague first perceiving a need for something and then requesting it from the organisation.

In other words, demand represents every single request for the practice to provide a formal service. It can arrive digitally, by phone, by paper, or by someone walking through front door. At the point that the practice does work on these requests, it then becomes activity.

This is an important distinction because too often NHS organisations use activity measures as a proxy for demand. What a patient actually asked for what – rather than what they received by way of service delivery – isn’t something we routinely code for.

The problem, as we’ll see, is that measuring the activity generated by a request doesn’t necessarily tell you whether that work was necessary or desirable or led to the right outcome. And this can sometimes take us in the wrong direction.

The drivers of demand

My starting point, therefore, was to try to capture the volume and nature of all incoming demands – these raw requests for assistance – as accurately as I could across my practice to give us a more accurate picture.

When you do this analysis, you find a single demand contact often ripples out to numerous additional pieces of work across the organisation and sometimes beyond. And by understanding the full contours of demand, you can do three things:

You can start to make predictions on what types of demand is going to come into your practice, how it is likely to affect different teams, and how you can manage it more effectively.
You can understand the intricacies of patient behaviour in a more sophisticated way, by developing patient clusters that are segmented not by pathology or age but on the type of demand-stimulating actions they take.
And you can start to track and measure the way demand flows across the system, teasing out the way that activity being carried out (or not being carried out) elsewhere connects to the demand experienced at a practice level.

Let me give you some examples what this means in the real world, as they help to illustrate why measuring demand rather than activity matters so much and how it can lead you to some simple but often game-changing actions.

Reducing downstream demand linked to prescribing

The first relates to prescribing. Imagine you get a patient telephone query about a prescription. You might review the dosage and check how long they've been taking it but most likely you wouldn’t book the patient in for an appointment – in many cases, therefore, this wouldn’t register in any activity data as most practices don't have the ability to link phone calls to the patients that made them.

However, we did do this analysis in a practice and immediately found about 30% of their inbound telephone calls related to prescribing. This felt counter-intuitive as they had a digital system for requesting prescriptions. The problem was quite easy to spot: the digital system didn't give a response back to the patient to say that the request had been received and people were therefore requesting prescriptions digitally, then phoning the practice to check that the request has been received. When this was changed, so they got a message back, those calls stopped entering the practice system and for the same demand, the corresponding activity was reduced.

A more sophisticated analysis of the same data then helped us to understand why certain patients chose that moment to call about their prescription and what decisions were taken in their care pathway that led to it. When we did this, we found there was a clear correlation between the practitioner that have previously seen the patient and the probability of a subsequent demand query.

The data also pinpointed the types of patients most likely to make these demands. We found the patients most likely to need to query their prescriptions were patients not already on repeat prescriptions that were phoning 24 to 28 days after the consultation. They were usually between the ages of 20 and 40, and most commonly had been prescribed analgesics or antidepressants.

The solution here wasn’t signposting patients to a community pharmacy or employing more in-house support – which might have been our reflex response if we’d just looked at activity levels showing a spike in prescribing-related calls. Instead, we worked with the specific prescribers concerned and created a standard message for any patient prescribed these drugs for the first time.

Online consultations and predictive modelling

A second example relates to online consultations. We wanted to explore whether this technology actually improved efficiency and reduced demand in practice – and if so, how do we maximise the benefits?

We wanted to reduce to a minimum the decision-making time for both reception and GPs. So, we spent time segmenting our patients and analysing their eConsult data and how this manifested in terms of demand across the practice workforce. Out of this, we were able to develop a pretty accurate predictive modelling system that helped us understand what the patient’s most likely need was based on their known behaviours. A big part of this was understanding how you can use continuity in digital care with a part time workforce.

For some patients, this meant the receptionist could book appointments with a specific GP directly because they knew the conversion rates from eConsult to face-to-face appointments was so high. Indeed, in some cases, our analysis showed we were going to be seeing them up to 70% of the time. Interestingly this had more to do with the patient than their presenting complaint. This saved us triage work overall.

Wider system perspective on demand

A few other examples show how you can also make connections with what’s going on across the wider system. The first relates to A&E attendance, where we found that a large proportion – 20% – of patients attending our local A&E contacted the practice within 24 hours of a visit.

Importantly, much of this demand was related to administrative queries, fit notes, prescription changes, and so on – intelligence which would again have been lost if we focused solely on measuring activity. This deeper insight enabled us to have better conversations with our system partners about how we manage patient demand across different parts of the health service.

A second example related to mental health. Like many PCNs, our network uses ARRS funding to provide mental health appointments into which its member practices can book patients. We wanted to test the theory that referring a patient onto the PCN service would result in less activity for practice GPs.

When we crunched the numbers, we found that this wasn’t always the case. For complex mental health patients, referral to PCN team made zero difference to the referring GP’s resulting activity. There was no demand reduction at all. They still had to do as much work as before.

Where it did make a difference, however, was for patients who hadn’t yet been diagnosed, partly because there tends to be a lot of follow-up activity after diagnosis that a dedicated PCN service can absorb.

Conclusions

I’ve drawn a few conclusions from all of this.

First: we need to rethink the way we measure demand at all levels, up to and including national policy. Relying on measuring activity as a proxy simply doesn’t help you get to the root of what is driving demand. As we’ve seen, it may occasionally take you in a direction that adds demand unnecessarily.
Second: a more forensic study of demand can help you create a more personalised and proactive way of reshaping your care pathway and showing where you need to prioritise and build capacity. Through use of intelligent software, it should be possible to create “smart systems” that can capture demand at the point of entry and advise practice staff and clinicians on the most effective way to manage this – one of the exciting areas of development I’m working on with Optum.
And third: this analysis shows we’re not powerless in the face of rising demand. We can adjust a significant proportion – a double-digit percentage – of the demand our practices receive. By recording and interrogating patient demand data, we can create better solutions for patients, helping them get where they need to be quicker. And, with fewer demand contact points, we can release essential time to help us care for our patients with the level of excellence, dignity and humanity to which we all aspire.

This article was prepared by Dr Peter Milmer in a personal capacity. The views, thoughts and opinions expressed by the author of this piece belong to the author and do not purport to represent the views, thoughts and opinions of Optum.