Adapting Open Clinica for International Collaborations Andy Lin

Adapting Open. Clinica for International Collaborations Andy Lin University of Michigan June 21, 2013 Open. Clinica Global Conference 2013

Overview • • • Background Optical mark recognition Open. Clinica on bootable USB drive Embedding audio files into Open. Clinica CRFs Open. Clinica on i. Pad mini

Background • UMHS-PUHSC Joint Institute – Collaboration between University of Michigan Health System and Peking University Health Science Center – Started in October 2010 to support collaborative clinical and translational research – Studies conducted jointly by teams at each university

Joint Institute Structure

BRBI Core • Biorepository and Biomedical Informatics • Set up technological infrastructure for supporting research data collection • Establish standardized process for securing and storing biological specimens and clinical data • Facilitates information sharing and data management

Research Data Collection • Selected Open. Clinica after comparison with other solutions – Using Enterprise Edition in at UM, run by MICHR – Using Community Edition in China • Enterprise Edition not available in China at the time • Majority of clinical data collection occurs in China – Data collection in China presents a variety of challenges

Challenges of Data Collection in China • Paper forms requested by study teams – Required to retain the original data on paper forms by IRB or departments – Not comfortable with computer interface for direct data entry • Limited internet access in rural areas • Language barrier / low literacy in rural areas – Potential for subjects to not be able to read and understand survey – Not all interviewers are native speakers – Variety of dialects in China – not everyone is fluent in Mandarin, especially in rural areas with older population • Data collection via tablets requested – Superior battery life to laptops – Generate additional interest in study participation

Overview • • • Background Optical mark recognition Open. Clinica on bootable USB drive Embedding audio files into Open. Clinica CRFs Open. Clinica on i. Pad mini

Optical Mark Recognition • What is OMR? – Recognizes check box and multiple choice bubbles – Bubble forms / Scantrons • Standardized tests, election ballots – OMR software allows desktop computers to process data • History of usage in clinical research – Proposed for use with health care data in the 1970 s – Popular in 1990 s and early 2000 s for transcribing to database – Recent shift towards EDC and tablets for data entry

Optical Mark Recognition • Current implementation – Paper forms still used by many studies – Transcribe to EDC or other study management software – Scanners are faster and cheaper • OMR software – Remark Office OMR – Allows users to design their own forms – Works with all common image types – Exports data in many formats – Excel, SPSS

Optical Mark Recognition • For studies using paper data collection forms – Requirements of some studies – Paper forms can be used without internet access or power – Multiple surveys can be administered at a time with self-reporting forms • Increases efficiency and accuracy of data transfer from paper forms to Open. Clinica • Cost-effective solution if reused for multiple studies

Process Overview Design forms using Remark software Create template for reading scanned images Collect data and scan completed forms Extract data using Remark image reader Use Python script to convert data into Open. Clinica XML import format • Import data into Open. Clinica • • •

1. Capture data using paper data collection forms 3. Use OMR software to read data from the scanned image 2. 4. Scan completed form as an image Import data into Open. Clinica

Form Design • Can use any word processor or desktop publishing software • Use bubbles for multiple choice questions – Special font provided by Remark • Can also set up areas for text recognition – Remark does not have handwriting recognition • Requires plenty of whitespace between bubbles and text recognition areas

Form Design

Template Design • Print and scan blank form for template base – Insures layout will match scanned completed forms • Open image in Remark template editor • Highlight areas with bubbles and text entry – Assign labels and values to bubbles – Allows grouping of questions with same responses (i. e. Likert scale) • Create Open. Clinica CRF first, and match the region names to the Open. Clinica item names

Template Design Blue region: Text Green region: OMR Specify number of bubbles and respective values in properties

Data Collection and Scanning • Forms should be completed by filling in bubbles – Check and ‘X’ marks can be too light or extend into other bubbles – Dark pen gives optimal results • B&W setting works well if forms are filled out correctly – Grayscale can pickup lighter marks or text • Minimum of 150 dpi recommended for OMR

Image Reading Process • Displays form region during reading process • Detects missing or extraneous marks – Pauses reading for correction • Detects written text in text regions – Pauses reading so user can enter text • Setting to identify and ignore large ‘X’ marks – Allows users indicate incorrect marks

Conversion to XML Format • Export data from Remark in tab-delimited text file • Add subjects to Open. Clinica – Allows script to query subject IDs • Get study and form IDs from Open. Clinica • Python script reads data from text file and converts into Open. Clinica import format

Conversion to XML Format Tab-delimited format with header XML Format

Importing Data • XML file configured for Open. Clinica’s Import Data feature • Performed by interviewer if tracking interviewer is important • Subjects must be created in advance with our method – Could use web services to automate the subject creation process

OMR in Practice • Used for 2012 summer project for UM medical students – Survey on knowledge of Hepatitis C and motivation for clinical research participation – OMR forms used at UM and at rural area near Beijing • Subjects at UM completed forms on their own – Multiple subjects taking survey at a time – Clear instructions are essential • Forms filled out by interviewers in China – Accuracy was much higher than UM forms

Efficiency and Accuracy • Used OMR process for JI-supported medical student project – – – 64 forms entered with entry from paper form, 111 forms entered with OMR process 7 minutes per form for data entry 2 minutes per form for entire OMR process • • Scanning, reading, and importing Scanned 90 sheets at a time Efficiency of OMR improves as more data is proccesed at a time OMR error rate of 2. 2% compares favorably with data entry error rates from published studies of 2 -10%**

Overview • • • Background Optical mark recognition Open. Clinica on bootable USB drive Embedding audio files into Open. Clinica CRFs Open. Clinica on i. Pad mini

Bootable USB Drive • Developed by Manhong Dai • Bootable USB drives running Open. Clinica on Linux OS – Can run Open. Clinica on laptops without internet access – Cost-effective if interviewers already have their own laptops – No need to set up software on interviewer’s laptops – Single configuration for multiple computers – USB drive can be transferred between computers • Interviewers could share a single USB drive if not conducting surveys at the same time

Features • Complete setup of Linux, Apache, Tomcat, Postgres, and Open. Clinica on 8 GB USB drive • Persistent – Allows changes to be saved between sessions – Entire setup can be backed up and/or copied to new USB drives • Automated backup of Open. Clinica database • Direct connection to server at U-M to provide IT support

Operating System • Unetbootin – Software to create bootable USB drives for Linux – Supports many distributions out of the box • Can supply ISO file to use other distributions • Lubuntu – Lightweight version Linux distribution based on Ubuntu – Targeted at users who don’t use command line – Very low hardware requirements – Up-to-date drivers supporting most laptop hardware

Operating System • Casper – Creates “Live” USB • • Run preinstalled operating system from read-only media Configured automatically by Unetbootin – 4 GB allocated to persistent storage space • Need enough disk space for applications and Open. Clinica database – Overlays the file system and copies all changes – Image file can be saved as regular file on Windows or Mac

Drivers and Software • Some laptop hardware may require additional drivers • Older version of Postgres required by Open. Clinica – Open. Clinica requires 8. 4, default Lubuntu version is 9. 2 • Default browser Chromium causes Javascript errors in Open. Clinica – Firefox available for Linux

Backup and IT Support • USB drives have a direct connection with U-M server – Drives identified with a configuration file • Files backed up automatically every 30 minutes with shell script – Requires internet access – Script can also be run manually – Only creates new backup if changes are detected • Allows access to the Linux instance on the USB drive – Access to Open. Clinica on USB if needed

USB Drives in Practice • Set up for 2013 summer project for UM medical students – Survey on motivations for participating in clinical research – Students will stay in China for 2 months • Two study sites: PKU clinic and rural clinic • No internet access at rural clinic – Small budget for project; needs to cover flight and living expenses – Using student’s personal laptops

Wireless Local Network Laptop booting from USB drive operates as a local Open. Clinica server Laptop server connects to router wirelessly or via Ethernet Mobile devices connect to the laptop server through a wireless network

Overview • • • Background Optical mark recognition Open. Clinica on bootable USB drive Embedding audio files into Open. Clinica CRFs Open. Clinica on i. Pad mini

Embedding Audio Files • Requested for surveys collecting data in rural China – Concerns over literacy of subjects and language barrier • Ensures consistent administration of survey • HTML entered in “RIGHT_ITEM_TEXT” column of Open. Clinica CRF Excel spreadsheet – HTML 5 <audio> tag – <embed> tag included for older browsers

Embedding Audio Files <audio controls preload> <source src="/Open. Clinica/audio/section 1/1. 1. mp 3" type="audio/mpeg"> <source src="/Open. Clinica/audio/section 1/1. 1. ogg" type="audio/ogg"> <embed autostart="false” src="/Open. Clinica/audio/section 1/1. 1. wav"> </audio> Audio controls in Firefox

HTML 5 <audio> tag • Standard for playing audio files • Supported in all popular modern browsers • Different browsers require different formats – IE only supports MP 3 files • Audio player built into browsers – User does not need to install plugins

Implementation • IE only supports MP 3 format, but Firefox does not support MP 3 in all versions – Need both MP 3 and Wav/Ogg formats • Audio files stored in directory in top Open. Clinica directory • Using <audio> tag avoids need for additional browser plugins

Audio Files in Practice • Embedded in CRFs for 2013 summer project for UM medical students – Rural site has older population with low education and literacy levels • Higher chance for older people to not speak Mandarin fluently • Difficulty of reading survey increased by medical terms – Students are not native Chinese speakers – Subjects that need to have questions read to them will hear the same file

Overview • • • Background Optical mark recognition Open. Clinica on bootable USB drive Embedding audio files into Open. Clinica CRFs Open. Clinica on i. Pad mini

Open. Clinica on i. Pad mini • Study team request to support data collection on i. Pad mini • Want to keep Open. Clinica as standard software for data collection • Chrome instead of Safari – Also need to update Javascript libraries for full functionality • Limit number of questions per section in CRFs to reduce scrolling • Modifications to interface required – Input elements and buttons are too small to select – Constantly zooming in and out undesirable – Images needed to be resized for some interface buttons

Modifying Open. Clinica with CSS • CRFs can be modified using inline styles – Increasing size of text and input elements for CRF not enough • CSS (Cascading Style Sheets) allow us to: – – increase spacing between elements increase size of elements not in CRF modify other forms (Add New Subject, Schedule Study Event) hide input elements not used by studies • Insert separate style sheet instead of editing current file – Avoid changes to code; makes upgrading difficult

Modifying Open. Clinica with CSS • Include new style sheet – Required to add one line to code for pages we want to modify to link to our style sheet – Use media query to only change appearance when resolution is below a specific value • Resolution of i. Pad mini is 1024 x 768 • <link rel="stylesheet" href="includes/ipad. css" type="text/css" media="screen and (max-width: 1024 px) "> • Desktop version is unaffected unless they decrease width of browser

Modifying Open. Clinica with CSS • Target elements with CSS selectors to avoid making global changes – CSS attribute selector to specify form names form[action='Add. New. Subject'] td, form[action='Create. New. Study. Event'] td { padding: 3 px; } – ID selector #login form, #login form input { font-size: 16 px; }

Modifying Open. Clinica with CSS • Hide elements on Add New Subject form[action='Add. New. Subject'] input[name='submit. Enroll'], form[action='Add. New. Subject'] input[name='submit. Done'] { display: none; } • Resizing buttons – Buttons use a fixed image that we can’t resize – Remove image and recreate button using CSS #login form. loginbutton { background-image: none; background-color: #789 ec 5; padding: 0. 2 em 0. 5 em; border-radius: 0. 3 em; }

Login Form

Add Subject Form

Schedule Study Event Form

Case Report Form

Open. Clinica on i. Pad mini • Planning to use for study on Fatty Liver Disease in Pinggu – Most data collection will take place in very highvolume clinics – Interviewers will have to move around and cannot keep a laptop charged • Tablets have superior battery life – Study team is also expecting that using tablets for data collection will generate extra interest – Uncertain about reliability of internet access – may try to use laptop as portable server

Combining Methods • 2013 student project is using elements of all four methods – Data will be collected at both U-M and China – OMR forms are being used by interviewers at U-M clinics • Waiting times are short; patients can finish form in examination room – USB drives will be used by interviewers in China • Internet access unreliable • Small budget didn’t leave room for new hardware – Audio files embedded in CRFs on USB drives • Interviewers are not native speakers – CSS modifications applied to Open. Clinica on USB drives • Default Open. Clinica font size is much too small for Chinese

Future Improvements • OMR process – Create subjects with web services – Hoping to implement this for U-M subjects of current 2013 summer project • USB drives – Initial testing with USB 2. 0 drives, later testing showed significantly increased performance with USB 3. 0 drives – Test portable laptop server and wireless network • May use setup for Fatty Liver Disease project • Open. Clinica on i. Pad mini – Working on converting 30 page survey into Open. Clinca CRF • Anticipating additional modifications for later parts of the form – Searching for better solution than resizing button images

Contact Information Andy Lin Bioinformatics Programmer, Molecular and Behavioral Neuroscience Institute, University of Michigan andylin@umich. edu

Acknowledgements Molecular and Behavioral Neuroscience Institute Microarray Lab Manhong Dai, Fan Meng UMHS-PUHSC Joint Institute University of Michigan Peking University BRBI Core Kai Zheng, Danny Wu, Ching-Ping Lin Hepatitis C Project Anna Lok, Sherry Fu, Elizabeth Wu Fatty Liver Disease Project Liz Speliotes BRBI Core Yanfang Wang, Yao Chen, Yuliang Huo Hepatitis C Project Lai Wei, Huiying Rao Fatty Liver Diease Project Linong Ji Funded by the UMHS-PUHSC Joint Institute http: //www. puuma. org