PolyClinic Visits for Acute Upper Respiratory Tract Infections (AURTI) over weeks
The graph above inspired it all. While browsing through the data.gov.sg datasets for a potential project to work on, I stumbled upon this graph which showcased a steep decline in polyclinic visits for AURTI during the COVID-19 pandemic.
The Dataset we’re working with:
Columns: Year-Week, Year, Type of Disease and No. of cases
Since we’re still in the midst of the COVID-19 pandemic, I thought it’ll be interesting to focus on Acute Upper-Respiratory Tract Infection.
From the graph, we can see that there is a general increasing trend before the COVID 19 Pandemic hit Singapore in 2020.
This could be due to 2 main factors: population growth and the increasing number of polyclinics developed over the years. For example, in 2018, 2 new polyclinics were established in Yishun and Ang Mo Kio. With more polyclinics available, this would increase accessibility to these very polyclinics, increasing the number of patients that these clinics see.
I noticed that there were high peaks every 2 years! Upon investigating further, I found out that 1 possible reason could be that a novel influenza strain would appear around every 2 years that the general population wasn’t immune to. This is also why in the following year, there wouldn’t be that high of a spike as the community would have developed immunity towards it too! Interesting insight that I thought was worth sharing!
Now, let’s move on to the interesting bit — Looking at the plunge in polyclinic visits in 2020. We all know that it’s due to COVID-19 — but how does it exactly affect the visits? Is there a delay in the clinical visits to the increasing number of COVID-19 cases? Let’s find out!
I went into the ‘our world in data’ website and found a dataset on the Daily number of COVID-19 Cases in Singapore
As the Polyclinic visits dataset is weekly, I had to then aggregate the COVID dataset with the WEEKNUM Function and Pivot Table where I did a group by weekly and found the summation of the new COVID-19 Cases discovered on that day.
I then went on to plot the number of polyclinic visits for AURTI and the Number of COVID-19 cases on Tableau! The results were interesting! Just as the COVID-19 unfolded, before the cases even rose, Singaporeans stopped going to polyclinics once DORSCON orange was announced! Moreover, the next catalyst in the downtrend was the 2 death cases when people really got scared to visit the clinic. This shows that with COVID-19, people feared going out in general, including to the polyclinics as they were afraid of being infected in a place where the likelihood of being infected is high. This fear of being infected drove the decrease in polyclinic visits by 90% from its high in Late January.
With a correlation co-efficient of -0.642, it seems like there is a moderately strong negative correlation between AURTI polyclinic visits and COVID-19 Cases. We can now quantitatively conclude that the fear of COVID-19 caused polyclinic visits to dovetail.
This is also true in other countries such as the US where doctor visits fell across the region!
Now that we’re done looking at the 2020 portion of the graph, let’s take a look at the earlier drops in AURTI polyclinic visits!
But after messing around with the data, I realised that the graph shown on the data.gov.sg website was wrong! Apparently, they had mis-plotted the points such that it looks like there was a sharp drop in visits! <Explain in video>
The real graph is at the bottom and it looks like there wasn’t any sharp drop in 2012.
However, it seems like the plot had some form of seasonality, so I decided to explore it further.
As I thought about the factors that affected AURTI, I remembered an old saying — ‘Don’t play in the rain, else you’ll catch a cold!’ Thus, I decided to explore that saying and proceeded to search for rainfall data on weather.gov.sg! I used 2018 as the sample year to test out my hypothesis! After stringing together the results for the year 2018 with daily weather data, I managed to plot rainfall with 2018 AURTI polyclinic visits! However, it does seem like there’s little to no correlation between these 2 variables as my correlation coefficient yielded only -0.04. I’m sure there could be other reasons for the seasonality feature in the AURTI visits, but because of time constraints, I wasn’t able to explore that in this post! However, off hand, I guess I could look at temperature data for a start too!
And with that, we’ve come to the end of my data story. It was really fun and I hope I managed to tell the story well with data! Special thanks to Dr Charles for this really fun mod! :D