After many years of experience submitting to NDAR, Prometheus has learned a thing or two. With that knowledge, we’ve put together a list of tips for approaching your NDAR data submission. With these tools in your back pocket, there’s no need to worry; NDAR submissions will be clean, quick, and leave you feeling confident about your data.
#1 Check your data. And then check it again
#2 Communicate regularly with NDAR
One easy way to complicate your submission is by miscommunicating exactly what data should be expected in the submission. This is easy to do since you’re in a busy clinic, and NDAR manages the submissions for many clients. Eliminate this potential risk by keeping your own documentation, that you share with NDAR, that explicitly outlines exactly which forms will be submitted during each submission. We usually send this documentation to NDAR after each submission as well, just to confirm we’re on the same page.
Another difficulty that can arise is when your Principal Investigator decides to make a change to the protocol midway through the grant. Let’s be honest, this happens all the time. Make your life a little easier and immediately contact NDAR regarding the change. Discuss the changes with them and come to an agreement regarding what changes will be made and how those changes impact the submission agreement. Open communication sounds like an age-old mantra, but it can really make a difference.
#3 Design your data to match the NDAR data dictionaries
Are you just getting started with a new study and are planning to fund it with an NDAR grant? Do yourself a huge favor and look at the NDAR-supported forms. NDAR has a large collection of data dictionaries including most published measures. If you’ll be using an ADOS or a Mullen, take a look at the NDAR version and see if that will work for your study. If it will, we recommend using the exact questions, answer options, and unique identifiers for that form because when it comes time to submit to NDAR, it will require little to no effort to submit. This isn’t always possible since each study is unique, but if you have a form that is similar to the NDAR version, contact NDAR and ask if they can update the standard to include your change.
Ultimately the goal of the National Database for Autism Research (NDAR) is to collect and consolidate data that can be used for future research. If data is being collected similarly across the country, you want it to be stored similarly and accessed altogether.
#4 Regularly validate your data
The more work you do at the beginning of the NDAR submission process, the easier each submission will be for you. That being said, inevitably something will change during the lifetime of your grant. In 2016, the NIMH Data Archive (NDA) announced changes to the NDAR data dictionaries. One major change announced was a large list of newly required fields, which came from the phenotypical concept definitions defined for ASD by Alexa Crawford of Harvard. Since these fields weren’t previously required, most submissions did not include them. This resulted in some work to update the process for pulling datasets to make sure these new fields were included. While this was no small task, we mitigated potential problems by catching it early and resolving issues prior to the submission deadline.
This is one example where validating your data using the NDAR Validation Tool, will help you catch issues that could make submission time more stressful. Another example is if the data is being collected in a manner that is not consistent with the NDAR data dictionaries. Typically these data collection issues can be resolved by making a few in-house process changes and/or recontacting a few families from whom you need a little more information. Since things like this can take weeks rather than days, it’s nice to catch this a month or two before the deadline.
We recommend using the NDAR Validation Tool monthly to validate your datasets. That way your submission will only include at most a month’s worth of data that you haven’t already seen and validated.
#5 Be aware of the data you’re submitting
We’ve saved the best and in our opinion the most important, for last. What’s the number one thing that we all worry about when it comes to submitting our data? Yep, that’s right, PHI. We work hard to ensure that absolutely no PHI finds its way into one of our de-identified datasets. One thing that gives us a real headache around NDAR submission time is checking, double checking, and triple checking to make sure there’s no PHI in our datasets.
One way to make this help ease this anxiety is to be aware of the data you’re submitting. This is another task that requires attention upfront but, trust us, it’s worth it. Take the time to make sure that the data fields you’re planning to submit aren’t text or memo fields. Those fields will require extensive review to make sure nothing slips by. Don’t do that to yourself. Create more restrictive data types that will allow you to collect the same data without the concern of potential PHI being in your dataset. There are a lot of text fields in the NDAR data dictionaries, but you don’t have to submit a free text box for that field, and we strongly recommend you don’t. This will make you confident that the data is truly de-identified.
Following our top five rules will make the NDAR submission process much more efficient and simpler. NDAR submission doesn’t have to be an error-prone and anxiety-filled process, and in fact, it shouldn’t be. If it currently is, we recommend assessing the ways in which you could improve your data collection and data auditing, or reach out to us. We’ve handled a lot of NDAR data submissions and would be happy to help.