Does the weather affect the demand for healthcare?

Here in Accurx’s Data team, we’ve heard lots of anecdotes over time about how the busyness of GP practices is not only influenced by the practice’s coverage area and demographic but by something much more basic: the weather.

This was recently confirmed when one of my colleagues, Jonny, visited a practice in east London, where a GP offhandedly mentioned that “everyone forgets they’re sick when it’s nice outside”. It’s an interesting idea. Are people really less likely to seek medical assistance simply because of the weather?

In February, our team decided to kick off some research (or perhaps more accurately “a quick analysis”, since this analysis lasted a few days between larger projects) to examine this qualitative trend and put concrete numbers to what had till then been only a hunch — the hunch that maybe, just maybe, people use health services more when it’s cold, grey and rainy outside.

Here’s what we did…

Given 98% of GP practices use Accurx, we’re lucky to be in the unique position to use our software as a means of testing this theory. We decided to do this by analysing patient-to-GP communications i.e. Patient Triage requests. These are queries which patients send to their practice asking for medical advice/assistance or administrative information related to their care.

We analysed this data at 1,181 practices over a period of two years, using London as our sample area (for simplicity above anything else since this was a very quick analysis compared to typical projects!). We examined the daily volume of Patient Triage and GP patient SMS messages. We then collected day-by-day data on weather patterns over this period such as (1) the temperature, (2) the type of weather (i.e. rainy, cloudy, clear skies), (3) the amount of rain and even (4) the moon cycle.

As you can imagine, there were a few variables we needed to control for:

• The Christmas slowdown. We removed data related to this as this dramatically skewed the results due to messaging slowing down significantly during the holidays.
• Our growing user base. There’s been an increase in our total users over this period so we needed to control for this also.
• The day of the week. Given Accurx usage decreases substantially over the weekend, we needed to control for whether it was a weekday or the weekend.
• The season. In the colder months of the year, the NHS often faces increased pressure due to cold and flu season and the prevalence of viral diseases. So we controlled for this by adding a variable for the season of the year (Autumn, Winter, Spring and Summer).

Here’s what we found…

When it came to Patient Triage requests there were some clear differences in usage — and therefore in patient demand — depending on the weather. Here’s a quick summary of our findings:

• Clear days showed the least number of Patient Triage requests, so we used this as our baseline for the other data.
• However, cloudy days resulted in around 12% more Patient Triage requests than on clear days. The difference wasn’t that significant statistically (p<0.1), but it’s clear there’s a quantitative trend here and we’d detect it with more certainty over a longer sample period.
• Rainy days showed a 17% increase in Patient Triage requests compared to clear days, after controlling for the variables listed above.
• More Patient Triage requests were sent in Autumn compared to other seasons — 26.4% more compared to Spring.

Figure 1: Effect of season on patient triage requests relative to spring time (where we receive the least amount of patient triage requests). On average practices receive 26.4% more requests in the autumn compared to the Spring. Other seasons were not statistically significant (note that the asterisk’ next to the numbers in the plot indicate *p<0.1 and **p<0.05)

Figure 2: Effect of Weather type on patient triage requests relative to spring time (where we receive the least amount of patient triage requests). On average practices receive 12.7% more requests during cloudy days and 17% more requests on rainy days compared to days with clear skies. Note that the asterisk’ next to the numbers in the plot indicate *p<0.1 and **p<0.05.

Interestingly, the temperature and the amount of rain on rainy days had no significant effect on the data. And unfortunately — despite hoping for a creepy, werewolf-based discovery — neither did the moon phase.

So it turns out, there’s something in our hunch! All else being equal, on days with bad weather, GPs receive more patient triage requests (about 12% more on cloudy days ☁️ and 17% more on rainy days 🌧️ relative to days with clear skies ☀️). They also receive around 26% more requests in the autumn 🍂 all else being equal, showing clear seasonality effects on usage. In short, the ‘better’ the weather the less Patient Triage requests practices receive.

The obvious question…

Why? Why do practices receive more Patient Triage requests and patient demand go up when the weather is worse? Is it because, like Jonny’s GP said, people ‘forget’ they’re unwell when it’s nice outside. Do they care less when they’re bathed in sunshine? Or is there some other contextual, psychological or socio-cultural factor we haven’t even thought of? The honest answer is, we don’t know! But if you’ve got any theories or insights of your own, we’d LOVE to hear them.

A few reflections…

As a Product Data Scientist, I find it really interesting to dig into theories and hunches and testing how much truth is in them. When Accurx was a younger company, the team often had to use a healthy dose of gut feeling and intuition to make big decisions, so being able to back up more of our assumptions with data today is a really important shift. Having a more established data infrastructure enables us to do this much quicker and much easier.

Although this project was about testing a hunch for the sake of our own curiosity, most of the projects we work on in the team helps the NHS in so many different ways. In fact, I’ve been at Accurx for a year now and have been lucky to see how big the scale of Accurx’s impact is, whether that’s transforming the way a service communicates with patients or improving efficiency and cutting costs. Given Accurx is used in 98% of all GP practices in England, it’s such a privilege to do work that has such a powerful, widespread impact for people working in healthcare.

If you’re thinking of applying to a role at Accurx, I’d definitely say go for it. What I enjoy most is the people, the culture and the values. Having moved to London from Canada a year ago, it’s been a great environment for making lots of close friends. It’s also great spending time getting to know people over lunch, since people often sit with different people and have quite interesting stories about how they got to working here. The company culture is pretty fast-paced where everything changes all the time, but I really enjoy that everything stays fresh and you get to learn a ton all the time.

Our data team — The left image includes Fran (Head of Data), Ali (Data Delivery Manager), Imo (Staff Data Engineer), Beth (Senior Data Analyst), Elena (Data Analyst), Rebecca (Data Analyst) and myself (Data Scientist). The image to the right includes Emily (Senior Data Analyst, to the right of the image) being silly during one of our meetings.

Interested in joining the Accurx team? Check out our latest job opportunities on our careers page.