PAPrKA: Physical Activity Patterns after Knee Arthroplasty 

Knee Osteoarthritis (OA) is a degenerative knee joint disease that affects around 250 million people globally. That means one in five people aged over 50 in the UK are affected.  

OA can be treated with an operation called a Knee Arthroplasty, also known as joint replacement. During this operation the damaged knee surfaces are replaced to reduce pain in patients and improve mobility. Before their operation, patients want to know how long their recovery will take and how much mobility they will regain. 

Our study, led by The University of Manchester, will answer those questions. To do this we’ll use the vital data our patients will share with us from their smartphones and wearable fitness devices. 

Project aims 

The Physical Activity Patterns after Knee Arthroplasty (PAPrKA) study is aiming to do three things: 

• To understand physical activity in people with knee osteoarthritis before and after joint replacement. 

• To understand factors that might influence different recovery pathways. 
• To share our findings about how to work successfully with people and access and link their data for research. 

Latest updates

Currently, progress on PAPrKA remains at stage 5 – design, development and testing of the study. Once we have completed final checks on the data collection pipeline, we will be ready to go live this month.

We will collect up to 18 months of data per participant between July 1, 2016 to the end of 2024 from Apple devices (iPhone and Apple Watch), Fitbit, and Oura. This data consists of physical activity data such as steps, distance, activity, minutes, energy used, and heart rate.

The data collection will be done through the help of colleagues at RADAR-Base developed by King’s College London.

The exact types of physical data differs based on the device types, with all irrelevant data being filtered out prior to the analysis stage.

Data collection for Fitbit and Oura can be collected through web APIs through a simple button on the website taking the participant directly to the consent and data collection stage. However, for Apple, we will be using an iOS app which will allow us to connect to Health Kit rather than pulling the data from the cloud server. As the Apple data is stored in the participants own devices, they will need to download a new app on their iPhone and go through the steps within the app to connect their data.

To make a setup easy, we have provided participants with instructions on the web site – downloadable PDFs for each device type – along with walkthrough videos showing the entire process.

Data processing challenges

Once the data is flowing, the data processing work can begin.

Firstly, the priority is to ensure the data taken from all the different devices are consistent. For example, step data from Apple comes within the day, but at uneven gaps – sometimes every two minutes, sometime times by hours – whereas Fitbit’s step data comes with even bigger gaps.

Another challenge are those posed by people who may own multiple devices. We need to line up their timeline and decide how to merge the data. For example, even within Apple, iPhone and Apple Watch can report different step counts for the same day. Apple and Fitbit use different algorithms, so when both are worn we need to be more careful to explore the more reliable stream, avoiding duplicates and being aware of potential missing data.

For studies like PAPrKA, there are particular challenges because we work retrospectively with data from participants own devices in free living conditions and so don’t have as much control.
Some days participants may not wear their device, or the device is on charge for multiple hours and so there are concerns surrounding missing data that we need to prepare for and handle.

We are refining our data processing method, testing ways to handle instances of multiple devices, missing data, and other collection issues and we will share the approach we adopt once it’s validated.