I have an uncertain relationship with statistics.
The relationship fell two standard deviations below the mean once I discovered qualitative research. I realised that important things like happiness, success, feelings, opinions and attitudes need words not numbers to measure them properly. I realised that more than one version of truth can exist, that truth depends on viewpoints and that someone decides what version of the truth is shown.
Truth was no longer an absolute.
We tend to think of statistics as the truth. We have to hold a deep understanding of statistics in our jobs in healthcare, but we also need to know about its flaws, biases and errors.
We need to know why ‘practice changing’ papers should not always change practice and why other papers which absolutely must change practice don't.
You might think you always follow the evidence but if I asked you a little about your personal life, I might get a very different story. I could show you undeniable evidence for an action which could save you money, save the environment, reduce your carbon footprint, reduce stress, lower your blood pressure, lose weight, reduce your risk of cancer, halve your risk of cardiovascular disease and even allow you to enjoy guilt-free doughnuts. But fewer than 6% of you reading this blog will be doing the simple action that could do all this. Why? I wanted to understand.
I started cycling on the 23 March 2020 – the first day of lockdown. There was a bike at the back of my garage that hadn’t made it out much. I began by cycling on all the nice sunny days that followed the announcement. I was on open roads and there wasn't a car to be seen. I became more and more confident. Lockdown eased and I moved from the roads to the wagonways that weave their way through the coal mining history of Newcastle Upon Tyne. Cycling became the only way I travelled. It has transformed my working day. Commuting is now a complete pleasure.
I really don't know why more people don't cycle, so I thought I would science it for a bit of fun.
A quick check of a few social media posts I made about cycling to work identified the weather as a major issue for people. I wondered if I could prove that it didn't rain as much as everyone thinks. My background research identified statistics such as average rainfall, cumulative rainfall and number of wet days per month. It did not identify the chances of rain during a single defined period in a day in the North East of England. This was going to be new science.
I prospectively monitored every commute I made over a 1-year period from 22 April 2022 (new bike day) to 21 April 2023. I travel at about 08:00 and 17:30 every day, with occasional trips between hospital sites. My distance and time taken was measured with a smart watch and recorded. The watch also estimated my calorie consumption.
Using Google Sheets, I recorded the daily details of every journey along with my rating of the weather.
The first problem was how to assess the weather when trying to show that it is better than you think. This single decision could completely alter my results.
Sounds simple to define a dry trip. But does one drop of rain on your glasses mean it wasn't dry? Are puddles from the night before which splash as you speed through them on a sunny day really defining damp? I needed to standardise how I was going to record the weather.
So, I invented my own classification system – introducing the Tiplady Index of Precipitation score (TIPs).
No significant rain falls out of the sky during the journey.
Such little rain falls that it dries off in moments and you remain completely dry.
Also includes nice snow, the kind that doesn’t stick because it is cold, crispy and lovely.
Puddles are OK, just go round them if you can.
Clothes do not require any drying at all.
Significant rain falls from sky during the journey, may only be for part of the journey. Visible droplets.
Sufficient standing water from recent rain that it might as well be raining because of the degree of spray from your wheels.
Sleet. I hate sleet.Clothes will reliably dry in less than one hour near a window or radiator.
Rains for all or most of the journey. Can include brief torrential downpours.
Strong wind can turn damp rating to soaked.
‘Tipping it down’. Wet through.
Clothes require wringing out, drying in a drying room or on a radiator for more than 1 hour. Shoes often unpleasantly squelchy, socks drip when squeezed.
Too dangerous to travel
Very subjective and depends on confidence levels.
Named storms, dangerous cross winds, weather warnings or snow that is so deep you can't stay upright.
Data was analysed using pen and paper and detailed subgroup analysis used BETH, a Natural Intelligence (NI) currently studying quantum mathematics at Durham University. BETH was asked to find whatever she could in the data as I had a hunch some nonsense would show up by chance.
I made a total of 433 commuting journeys. Total distance travelled was 5,979.5 km. Total time taken was 15,143 minutes, giving an average journey time of 35 minutes and an average journey length of 13.8 km. Average speed was 23.7 km/hr. Travel by bike was impossible (TIPs 3) on only 5 days (recorded as 10 journeys) due to solid ice.
Each journey used approximately 330 calories, about the same as in a large jam doughnut.
|Number of journeys
Subgroup analysis showed that if it rained on Thursdays, it was more likely (p < 0.05) than average to be heavier (TIPs 2) than light (TIPs 1) rain. Further subgroup analysis was paused as the NI “had more important things to do Dad”.
I therefore concluded that 84% of my journeys were dry, proving my hypothesis that it rains a lot less than people think.
Definitions influence outcomes
I hope that you can see how my definition of dry significantly influenced the results. We see this so much in medicine. How we define an illness, a symptom or an end point influences study design and results. The broader the diagnostic criteria the more early stage patients get included and better results could appear. New definitions of disease response such as progression-free survival depend completely on how you define progression. I am unashamedly biased about cycling and set out to prove a point. Be wary of conflict of interests and be wary of changing disease or response criteria.
The study took place in the North East of England and will therefore not be relevant to commuters in the entire rest of the world. Consider medical practice and how different healthcare systems and social determinants of health are. Be wary of transferability.
The wetter Thursdays, I hope, proves another point. It was not something the study set out to do, it took a skilled mathematician to spot and it is, of course, just chance nonsense. Easy for us to see when it is something like weather but far less easy when convoluted analyses are made in medical papers. Be wary of subgroup analysis and clever statisticians.
Understanding the ‘why’
I had set out to understand why people don't cycle and what could change this. Perhaps a few of you have raised an eyebrow at the weather statistic but I suspect little will change for most of you. What might really change how you travel to work? This required some qualitative work.
I did a bit more science and interviewed 25 medical students. Only 1 was a regular cyclist. The main reasons for the others not cycling included the weather, distance, baggage, clothes, drying and changing facilities, security and other journeys. There were more complex issues like confidence, knowledge of routes, roads, infrastructure, drivers, safety, cost, fear of travelling alone, the dark, health, fitness and public perception. Then there was the complete surprise for me that some people have never learnt how to ride a bike.
This stuff was way beyond me. This is politics, public health, education, financial, psychological and sociological. This is real medicine. These are the factors that influence the health of nations.
This small and simple study is my small and simple commentary on modern medicine. We could significantly lower the national risks of cardiovascular disease, cancer and all-cause mortality by some small and simple personal actions but we do not. Our duty should be to understand determinants of health and the barriers to implementation. Pair up qualitative and quantitative researchers and put more pathologists on pedal power.