Live Webinar with Dr. Pamela Semanik

Welcome everyone. Thank you all so much for attending our oi.edu live webinar series hosted by the OAI core knowledgebase. I want to thank today's speaker Dr. Pamela Semanik who is associate professor at Rush University College of Nursing. Title of today's talk is accelerometer data from the OAI movement measured. Following the presentation, we should have a few minutes for Q&A. If you have any questions, feel free to type them into the chat and we'll try to get to them at the end. and I'll let Dr. Semanik take over. Okay. Thank you so much for that great introduction. Um I'm going to try to advance stuff here. All right, here we go. Um, delighted everybody could join us today and wanted to thank our Northwestern accelerometry team. Um, which was headed by Dorothy Dunlop and uh, Jen Song, Julia Lee who's actually on the call, Rowland Chang who's on the call and me. These were supported by some grants uh, from NIAMS for which we are very grateful. So today's objectives are to let you know how the data was acquired, the form of the data, how to access it, now what has been done with the data and what can be done with it and what you should know before using the data. The data acquisition occurred in a subcohort of the osteoarthritis initiative. Uh these were people with and without baseline radiographic OA at the scheduled 48-month follow-up visit. In order to be eligible, people required a scheduled OA 48month visit follow-up between August 2008 and July 2010. And these were staggered across many months in the OAI sites. Um there were roughly 4,796 uh eligible people um but many were excluded. Um probably half of them were excluded because their visits occurred prior to the study start date. Then you have a few um deceased and no-show. So, of the 2712 eligible participants, um 2,127 consented to participate, which was about 78% of the available folks. Um, of those people, 1,223 had radiographic knee OA.

Uh, how did they get this? While the OAI site researchers performed accelerometer initialization, application and participant instruction there at the site in person as people came in. Um the accelerometry units that we used at the time um was the acttograph model GT1M which was one of their earlier models. Um I think there was a 7131 before that. Um but so this was you know uh for us it was their latest model. It was worn on the waist. Um it's a small unaxial accelerometer that measures vertical acceleration and deceleration. Now even the least expensive Fitbit these days has a triacial accelerometer. So just by way of comparison this was um you know a very um different kind of model. uh these things were collected um in 60-second epics and the sample rate was 30 hertz in this group. So the staff at the sites issued the instructions to wear for seven consecutive days except for sleep and water activities and downloaded the data. This would be minute-by-minute accelerometer counts checked for valid wear time and stored on the server. Physical activity variables were synthesized from the minute-by-minute accelerometer output. These were aggregated on a dayby-day basis and aggregated to the person level for general users. Accelerometry data were merged with OAI public data from baseline to the 48month examination. Um at that time the condition of the data is no known data errors and the process this whole process of issuing the accelerometers and uh getting them mailed back and put on the server uh was repeated for the 72-month visit and that data from the 72-month visit was released in 2016. So what do we have here? When you get into the data set, you will see that there's um the minute-by-minute data, which we've had a lot of interest in lately. Um that's the designation of each minute to a physical activity intensity level. Um so that's by the minute. You can see here the you know number of records is quite large. Then we have the day-by-day total minutes spent and physical activity intensity levels meaning light and moderate vigorous. Um these are restricted to the valid monitoring days. Um days in which the device was worn at least 10 hours and we had um at least four days of data. accelerometry um minutes per day spent in physical activity intensity level for each participant having four or more valid days is at the person level and these are also available. So those are the things that you will see when you get into the data site. Um what can you see in the minute to minute data? Well, you can see that um there's activity counts available. You can see the day of the week that they wore it. You can see the PA monitoring month. We've had some people, you know, that were interested in what the weather was like and how that might have affected um physical activity. And of course, the you know, the National Weather Service has all of that information from way back. So, you know, if people were interested in activity according to the weather, um that's certainly doable with this minute-to-minute data. The data was processed using three different sets of cut points or thresholds um of activity intensity for comparison to different populations. Um so the activity intensity cut points that we have distilled the data into uh for public release are the Schwarz data um thresholds you can see there for light moderate and vigorous or greater uh the traditional freeden and then the national cancer institute uh cut points that were formulated by Troyano um in that study. So um most of our publications have dealt with Triano but a lot of people have been interested in the other cut points as well. So how do you access this? Now um the NDA the uh National Institute of Mental Health uh National Data Archive is the home of the publicly available um images and data. Now you'll need to log in and sign the data user certification. uh complete the data use agreement. Uh you can see the link there. If you don't have a login uh identity, that's highly advised. Most things are tied to that. Now, um go to my account, you click on the upper right data permissions, and you scroll down to request access. Um we had um our people test this out a little bit and um oops, I don't want to go there. I'm gonna go back. Um, sometimes, uh, you have to call the help desk so they can create, um, a role for you. Uh, they assign a role so you can access the data. And that, um, worked out very well. They were super responsive and our people got in, you know, within minutes once they figured out what the issue was. So, what has been done with this data? Um I'm I can only tell you uh with high confidence what we've been doing with it. Uh the first thing we were very interested in was our physical activity guidelines being met. Uh the aim of that study was to describe objectively measured levels of physical activity in adults with knee OA and report the prevalence of meeting public health physical activity guidelines in the OAI group. uh we used a cross-sectional accelerometry data from those adults uh age 49 to 84 with radiographic knee OA and assessed those folks for meeting the aerobic component of the 2008 physical activity guidelines for Americans which of course is the equal to or greater than 150 minutes per week of moderate to vigorous intensity activity lasting an at least 10-minute bouts quantile regression tested median sex differences in physical activity levels. Um so what we found was that the aerobic physical activity guidelines were met by almost 13% of men and about 8% of women with knee OA. uh 41 sorry 40% of men and 56% of women were inactive meaning they had no moderate to vigorous activity that lasted in bouts of 10 minutes or more during the seven monitored days. uh the proportions of men and women meeting public health physical activity guidelines were substantially less than what was previously reported based on self-report activity in arthritis populations. So that was a wake up uh physical activity minimum threshold uh predicting improved function was another one of our interests. For a long time we were interested uh not knowing um so much how many people could do 10 minutes at a time of aerobic activity uh which was you know tough when you have knee pain. Uh what we really wanted to know was how can you maintain function and what level of physical activity was required in order to do that. So to identify an evidence-based minimum PA threshold to predict improved or sustained high function for adults with lower extremity joint symptoms. Uh we examine this data again. This was a prospective multi-sight uh data from 1,629 adults over 49 years of age with symptomatic lower extremity joint pain. Now these people didn't necessarily have radiographic knee away. they just had symptoms. Uh they were participating in the accelerometer study and clinically they were assessed two years apart and we saw improved high function and two-year gate speed and patient reported outcomes were based on improving or remaining in the uh best meaning the maintaining high function quintile compared to baseline status. Uh the optimal threshold that we saw was 45 minutes uh based on total MODVIG minutes uh per week which predicted improved or high function outcomes more strongly than the 150 uh legacy amount or other investigated me metrics like total physical activity. So meeting the 45 minute total modvig per week threshold had increased relative risk for improved or high function. Uh the for gate speed the relative risk was 1.8 with a 95% confidence interval. Um and for patient reported outcome physical function the relative risk was 1.4 with a 95% confidence interval. uh that was compared to less active adults. So we were excited that we had some um objective evidence that people to maintain function didn't necessarily need the 150 minutes per week which is a little different than the cardiovascular uh health aims. So what can be done with this data? Well, you can merge the physical act data uh physical activity data with any of the other data sets for analysis. It can be merged with imaging data for analysis. Um you can describe the PA levels of any demographic subset in terms of total and mean minutes per day spent various physical activity intensity levels. uh you can perform comparisons of those using the data from two collection time points because the 72 month data is out there as well. So there are a few things you should know before you go. Uh when using the other OAI data sets, you should merge or join the data using the participant ID. Analysts should always output and view the SAS variable labels in their entirety to ensure important information about uh the variables is not lost and the maximum SAS label length is 160 characters. You should know that valid days in this group are defined as those with wear time of at least 10 hours and only participants with four or more valid days of wear are included in the summary data. You should recognize that daily wear was determined by a rolling window algorithm that allowed for zero counts for up to 90 minutes. Uh this is different than the typical 60 minutes that was used in the uh Troano study. For example, uh this was validated u by our team um in a different paper. The coding was modified to fit the unique monitoring characteristics of adults with OA um in the OAI who are at high risk or have knee osteoarthritis.

So you should not start using the data without first looking at the handbook that is on the site and I'll uh leave this up here uh for a few seconds. So this is the um accelerometry data description uh the PDF which is in the database. Um you should not expect to be able to identify the type of activity uh for example jogging or cycling um etc that was engaged in by participants. Um that's only that's not reflected in this particular data set um or any others that we have. Uh don't expect to be able to reprocess raw data. This is uh the GT1M model which did not it distilled its data onto the the hardware of the device before it got to download. So there's no raw data. There is we are aware a later version of the GT1M that did have a sort of um sort of raw data available but that was not our model. So, we don't have any of that um either publicly available or stored in our in our database. Um only summary v variables and minute-to-minute data are available. Um we do not have any step count files. People have asked about that in the past. Um so that's um everything I had to say. We're ready for questions now if anybody has any.

All right. Thank you, Dr. Semanik for that excellent presentation. Yeah. So, we're ready for Q&A. If you haven't done so and have any questions, feel free to type them into the chat. Um, to get us started, I have a question here. Who should we contact if we have questions about the accelerometer accelerometry data? Um,

I would assume you mean the the data that's publicly available. Yeah. Um Jeff, do you feel those questions when they come in? Yeah, we can certainly be the first point of contact and if it's not something that the knowledge base is able to address, we can certainly refer to you and your team. Sure.

Well, anyone else has any questions, feel free to type them out. Um but in that case, uh we can end here for today. Okay, I will send some links in the chat really quickly. Um, want to thank thank you again Dr. Semanik for taking the time to come out and present today and I want to thank everyone for joining us. And as we wrap up, I also want to invite you all to upcoming webinar on November 16th with Drs. David Felson, Michael Nevitt, and John Lynch who will be talking about knee radioraph readings from the OEI. Um, oh actually we got another question. So can this data be considered for Q of pat of OA patients quality of life? I don't know why not. I I don't know. I know the um did they if the OAI database has quality of life measures that they collected regularly? Yeah. So at the annual visits, the osteoarthritis initiative collected the cous quality of life subcore. So that is available and there's also general health questions such as the SF12 scores. Right. I I would think that would be a terrific project. Another question here is accelerometry data available only at 48 months? No. The 72 months data was released in 2016.

Okay, great. Um, same same data, different day. So, the same, you know, packaging and the same types of data, same files are available. Great. Thank you. Um, I think that's all the questions we have today. So, thank you again for uh your presentation and uh again, I sent the links to the next webinar in the chat and a post-weinar survey. So, that concludes today's webinar. Thank you and have a great rest of your day. Thanks very much. It was a pleasure.