X Close

Research Department of Primary Care and Population Health Blog

Home

Menu

Archive for the 'Infections' Category

Google Flu Trends is dead – long live Google Trends?

23 January 2018

In this post Fabian Sailer discusses using Google Trends and the current flu season 

Google Flu Trends (GFT) was a project which aimed to predict flu outbreaks based on web searches of google users. The project was launched in 2008 but abandoned in 2015. Why did this happen? Mainly because the predictions were not accurate enough; disease models based on surveillance data, e.g. of the Centre for Disease Control (CDC), have shown higher accuracy. GFT also did not predict the non-seasonal 2009 H1N1 pandemic, and from 2011 until 2013 the flu prevalence predicted by GFT was wrong in 100 out of 108 weeks. Approaches to estimate flu epidemics of another independent research team based on google trends showed a better accuracy then GFT.

There was enough reason to cancel this project, many also say it was cancelled because it was able to predict the past better than the future. The promising results of Yang et al. motivated me to have a look at trends.google.com, a web page which allows users to analyse the popularity of search terms over time.

I was interested in the current flu season which is why I started to look at the term “flu”.

Graph 1: Popularity of the term “flu”

This figure (Graph 1) shows that “flu” searches on Google started to increase at the beginning of September 2017, which is also the beginning of the vaccination period. The terms “vaccine” and “flu vaccine” have been searched more frequently in this period as well. In December the number of searches rose faster and peaked on the 09.01.2018. I compared the shape of this graph with the surveillance data provided by Public Health England (PHE) in their weekly national influenza report. PHE documents that the flu season picked up speed in December and peaked in the first week of January.

Due to this similarity between the PHE data and the graph obtained from google trends I decided to have a little play around with trends.google.com.

As a next step, I wanted to know whether I could also recognize the “flu” pattern from the first figure for searches for flu symptoms. Therefore, I took the first 5 symptoms listed on nhs.uk/conditions/flu.

Graph 2: Popularity of terms relating to flu symptoms

Generally this chart (Graph 2) does not indicate a clear trend as observed in the flu chart. But the searches for some symptoms (“fever”, “cough”, and “sore throat”) seem to rise in line with the general “flu” trend. On the other hand, other more unspecific symptoms like “aching” and “tired” did not show this trend at all.

Graph 3: Popularity of generic terms relating to health

 

Looking at other generic terms (Graph 3 – “sick”, “ill”, and “healthy”) I could not observe the “flu” trend. On a side note, I could  also see that during Christmas people are searching less frequently for “healthy”. But presumably the New Year’s resolutions overcompensate this drop-off with searches for “healthy” peaking in the first week of January.

I wanted to know whether these observed “flu” trend connects to current news.

It was reported that the National Health Service (NHS) is currently working at a 100% capacity. This maximum workload is partially due to the flu season. Therefore, I wanted to see whether the “flu” pattern can be observed in search terms looking for the NHS or health care in general as well.

Graph 4: Popularity of search terms relating to health care in general

Interestingly, there is a completely different pattern in this data. First of all, the three curves displayed here do not start rising between October and December. But all of them show a weekly repeated pattern. Most searches for “nhs”, “gp”, or “hospital” happen on Mondays/Tuesdays. Over the week there is a slight slope until Friday before the number of searches drop to their lowest value of the week on Saturday/Sunday (Graph 4).

There might be lots of reasons for this pattern. My first theory was that employees might need to get a doctor’s certificate on these days to call in sick for work.

Graph 5: Popularity of terms relating to medical certificates

But this pattern cannot be observed within the search terms for sick notes (Graph 5).

A similar weekly pattern can be observed if we examine the searches for other working places like a “bank”. Even the searches for “flu” and the symptoms for the flu seem to follow this weekly repetition to a certain extent. But none of these patterns is as distinct as the one for “nhs” and “hospital”. I was not able to find the one reason for this pattern, and I guess it is a combination of factors which result in this unique shape.

As all of this data is freely available I invite you to have a play with google trends and see which interesting bits and pieces you might discover.

HPRU in Blood Borne and Sexually Transmitted Infections

5 April 2017

In this post Tom Hartney, a PhD student at PCPH, talks about the Health Protection Research Unit and its recent Academy day.

The Health Protection Research Unit (HPRU) is a collaboration between universities, including University College London and Imperial College London, and Public Health England (PHE). This is part-funded by the National Institute of Health Research (NIHR) in order to promote high quality multi-disciplinary health research. It was set up in 2012, and supports the funding of PhD students to work on projects related to the HPRU’s research objectives.

These students make up the HPRU Academy – this isn’t a university in itself, but instead represents PhD students based across several universities, working in 10 research units each of which focuses on a different health issue (such as emerging infections or environmental hazards) or methodology (such as modelling or evaluation of interventions). This January saw the first HPRU Academy Retreat, which brought together students from across the country to present results from their work and hear about future plans and priorities for the Academy.

T Hartney April 17

Me, second left, and Josina, third left, at the Academy Retreat with our fellow PhD students from the HPRU in Blood Borne and Sexually Transmitted Infections

Myself and Josina Calliste from PCPH attended, both representing the HPRU in Blood Borne and Sexually Transmitted Infections. With 19 students giving oral presentations, as well as talks from academics and PHE staff, over two days, we knew this would be a packed schedule, but it gave us a chance to meet other students working on public health topics, to talk about our projects and find opportunities to work with them in future.

The format was divided into two types of talks, with those from students talking about their work making up the majority of the programme, and each day introduced and ending with talks from HPRU academics on the public health research landscape and aspects of methodology. The student presentations covered a huge breadth of healthcare topics, from the effects of traffic pollution to the risks from ticks in urban areas, with the research opportunities provided by genomic sequencing being a common theme. Given there was so much to cover, students somehow still managed to convey their enthusiasm for their chosen topic and explain their key results while sticking to time.

Professor Bernie Hannigan gave the PHE perspective on the Academy, emphasising its focus on applied research – identifying gaps in the evidence and addressing them with research outputs – as well as the need for researchers to think about how evidence influences public health policy & practice. Dr Satnam Sagoo, PHE’s Head of Education and Training, discussed routes to develop a career in public health, setting out the options available and ways to enhance transferable skills during the PhD process.

Karen Wilding from the University of Liverpool presented on research ethics and governance, a fundamental aspect of every research project. She described the variety of clinical research governance systems and frameworks, as well as the key considerations for ethics, and resources available for help and advice. Finally, Professor Jackie Cassell from Brighton and Sussex Medical School talked about getting patients and the public involved in research. She described ways to implement this at every stage of the lifespan of the doctorate, some of the challenges involved in healthcare research in general, and on transient and stigmatised conditions in particular. Imaginative approaches to public engagement were illustrated  these using the example of an exhibition on scabies at the Eastbridge Hospital in Canterbury – originally a medieval hospital for pilgrims – set up in collaboration with the Textile Department of the University of Brighton.

To break things up the Retreat included interactive elements – firstly, in the form of an outbreak response exercise led by Dr Ruth Ruggles, Head of Public Health Training at PHE. Dr Ruggles used data from a real outbreak to put students in the role of public health response teams with crucial decisions to make as the outbreak developed. This culminated in a role played press conference, with volunteers fielding tricky questions from the audience on the media and public reaction to the outbreak. Also, students were able to vote online on their favourite talks and poster presentations, with prizes awarded by Professor Tom Solomon from the University of Liverpool on the final day.

Despite the challenge of packing so much into a two day schedule, the inaugural HPRU Academy retreat provided plenty of food for thought on how to use the expertise available within the Academy – here’s looking forward to next year’s event.