Extracts from CHR protocol describing a method of collecting, managing, and analyzing video data

Nicolas Sheon, PhD, UCSF Center for AIDS Prevention Studies

From Description of Qualitative Data Analysis

Qualitative Data Analysis of the focus groups will begin immediately following the focus group in the form of a debriefing session among the researchers present at the focus group to review the salient points, dynamics and unexpected findings. These debriefing sessions will be audio-recorded as well to help guide later analysis of the focus group data. While these debriefing sessions are useful for capturing the immediate impressions, we will use Transana software for more systematic coding of the video recordings of the focus groups.

Why we need to video-record the focus groups: We recognize the sensitivity of video data and the added risk of loss of privacy over audio data alone. While digital audio may be sufficient for individual interviews, analysis of focus groups requires the ability to distinguish among several speakers talking simultaneously. In addition, video enables us to pay attention to non-verbal behavior, direction of gaze, and the physical performance of social identities and status, such as clothing, gaze, gesture, posture, physical proximity and facial expressions. A key research question in the proposed study is how race and ethnicity relate to attitudes and motivations towards participation in clinical research and prevention screening of anal cancer. Video data are also essential in order to identify the demographic characteristics of focus group participants and correlate them with their contributions to the discussion. We have worked with audio recordings of focus groups, but without the video, it is often difficult to determine whether the current speakers is a new speaker or the same person, especially when there are several people speaking simultaneously. It is therefore imperative that we be able to see who talks when, to whom, and how others react in order to make sense of the focus group data in all their richness and complexity. Although inexpensive video cameras typically have poor audio quality, a new class of recording device designed for musicians, the Kodak Zi8, incorporates professional quality stereo recording with a modest VGA (640X480 pixel) camera on the same miniature device. The video and audio data are recorded onto a SDHC card in QuickTime h.264, a compressed video format that can be readily analyzed using Transana software. Because the Zoom Q3 does not use tapes, the video data do not need to be converted or compressed, and can be immediately transferred to a computer and stored securely in an encrypted volume as described in the section Part 5 Section D.8. The lack of video tapes greatly decreases the chances of data loss and enables us to employ a strict security protocol for data storage and transfer as described in Part 5 Section D.8.

Transcription, segmentation, and coding of the video: Transana allows data analysis to begin prior to transcription by coding the audio and video itself. Our first analytic pass of the data will select and code the most analytically useful segments of the focus groups for systematic transcription and thereby reduce transcription time and costs. These segments will be transcribed by a transcriptionist who has been trained to insert Transana time codes at each speaker change while transcribing. These time codes serve several important functions. First, time codes synchronize the transcript to the video much like subtitles are synchronized to the dialogue in films. Second, they segment the video into individual speaker turns to facilitate coding of who is speaking on a particular theme. The ability to see who is speaking during a clip will enable us to correlate visible aspects of identity (race, class, education, etc.) with themes, such as distrust of clinical research, health disparities, etc. Finally, time coded transcripts allow us to see where overlapping speech occurred so that we can distinguish simultaneous speech with the help of the video image.

Using Sequence Maps to Discover Patterns within and across Focus Group Discussions. Dr. Sheon has worked with the Transana software developer on several new features for Transana to facilitate sequential analysis of digital video/audio files. One feature, called "Sequence Maps" uses colored bars along the timeline of the video to facilitate the visual analysis of patterns of discussion such as who spoke when, for how long, and on what theme. This method requires video so that we are able to visually distinguish participants in terms of who is speaking, the direction of gaze while speaking, and how others in the group are responding and listening. Once speaker turns are coded by theme and participant, Transana sequence maps will help us visualize patterns of agreement and disagreement among participants. This will enable us to determine whether certain themes or beliefs are widely shared among the group, or idiosyncratic among certain participants. For example, a common pattern in focus groups is that a view expressed by one participant is rejected by others as non-normative, and this is displayed by a silence or shift in topic in the next turn. Alternately, a view may be embraced and elaborated on by the group in subsequent turns. In this way, the focus groups provide insight into the process of social construction of attitudes and beliefs. We therefore expect to encounter considerable ambivalence about trial participation and to see patient views change during the focus groups as they encounter new arguments for and against participation in the trial. Being able to track these shifts by specific participants over the course of the focus group will help us to identify key arguments for or against participation and how they are deployed in group dynamics. New or unexpected themes identified in the focus groups will be explored in subsequent focus groups and, where appropriate, included in the patient survey. The focus groups will enable us to identify particularly rich descriptions that will be useful for a more in-depth understanding the determinants of patient participation in the future RCT and other trials as well.

Qualitative analysis of the individual interviews will be same as focus groups, except that only one interviewer will be present during the interview and we will not need to use video. The interviewer will write up or dictate impressions immediately following the interview and these notes will be included in the Transana database. Like the focus groups, we will use Transana software for thematic analysis of the audio recordings of individual interviews.

Description of Data Security

We recognize the sensitivity of video data and the added risk of loss of privacy over audio data alone. While digital audio may be sufficient for individual interviews, analysis of focus groups and clinical encounters requires attention to non-verbal behavior, direction of gaze, and the physical performance of social identities and status, such as clothing, gaze, gesture, posture, physical proximity and facial expressions. A key research question in the proposed study is how race and ethnicity relate to attitudes and motivations towards participation in clinical research and prevention screening of anal cancer. It is therefore imperative that we be able to see who talks when, to whom, and how others react in order to make sense of the focus group data. However, because video data are necessary for our research questions, and indeed have become the norm in many types of qualitative research, we have devised a comprehensive data security protocol that addresses the primary risks of data loss and secure storage.

Our data collection, storage and security procedures are an improvement over traditional methods involving tapes. Until the advent of affordable digital audio recorders (ca. 2001) and affordable tapeless digital video recorders (2007), video and audio data were primarily stored as cassette tapes. These tapes were difficult to work with during analysis because they required specialized playback equipment and were difficult to index and playback in a non-linear manner. The advent of affordable digital video recorders with good audio along with Transana video analysis software, co-developed by Dr. Sheon with David Woods at the University of Wisconsin have revolutionized the way qualitative data are collected and analyzed by UCSF researchers. The growing ease with which researchers can collect, store and analyze video and audio data without transcription, has resulted in a move away from text-based analysis of transcripts. Over the last year, two other popular qualitative software packages, ATLAS.ti 6 and Nvivo 8 have added video analysis as a core features, reflecting the growing trend towards the use of video analysis across disciplines, particularly in fields of health communication, counseling, and health education, where attention to non-verbal aspects of communication has grown along with the use of video data.

As Co-Director of the CAPS Methods Core, Dr. Sheon is an expert in qualitative methods and has conducted dozens of trainings and served as a consultant on dozens of UCSF research projects to improve the quality and security of qualitative data since 1999. The protocol presented here represents the current state of the art keeping in mind available technology, cost, and feasibility. These methods are particularly useful for studies involving research in other countries where the disposition of the audio and video data during translation and transcription present risks for data loss. Video data will be recorded onto digital media (SDHC card) using a portable video recorder (Kodak Zi8) that records directly into QuickTime h.264 format and can accommodate an external stereo microphone to capture better sound than typical video cameras. Using this format avoids the need for a digital video tape (mini DV) which must be compressed and converted to a format that is compatible with Transana prior to analysis. The disposition of physical tapes and optical media such as miniDVD represents a significant risk of loss. However, because flash-based video recorders, such as the Kodak Zi8 no longer use tapes or optical media, the video data can be easily transferred from the recorder and stored in an encrypted volume that is then stored on a secure computer or burned on a DVD in a locked cabinet. At the conclusion of each focus group, the video data will be transferred by the Project Director from the camera's USB port to an encrypted volume created on a laptop computer using TrueCrypt software. TrueCrypt software is free and can be used in conjunction with other encryption software that are designed to encrypt access to the computer itself. Dr. Sheon has produced an online guide for using TrueCrypt at http://www.palmpal.org/truecrypt.pdf. TrueCrypt creates an encrypted volume on your computer hard drive using AES-256-bit encryption. The encrypted volume can only be opened using a password. Only study personnel will have access to this password. In case of hard drive failure, backup copies of the focus group video data will be stored for safekeeping as encrypted volumes on burned onto DVD-R discs that will be kept in a locked cabinet at the UCSF Center for AIDS Prevention Studies. The encrypted volume will be opened (mounted) only when it is being analyzed using Transana. At all other times, the encrypted volumes containing the data will remain closed (dismounted) and thereby encrypted. Because the video data are stored in an encrypted format, the risk of unauthorized access is eliminated. Even if a laptop computer or DVD-R containing the video is stolen, the encrypted volume would prevent access to the data contained therein.

Disposition of the data over time. Because video and audio files stored in an encrypted volume are not accessible without Truecrypt sofware and a password, there is less need to destroy the data for security purposes. In fact, as video and audio data have become the primary focus of analysis, reliance on the text transcript as the primary unit of analysis has diminished. As qualitative methods and analysis software evolve in response to the increased use of video data, there are important reasons for retaining the source data for future secondary analyses of the same data using different methods of analysis. For example, researchers who initially focus their analysis on content using a Grounded Theory approach, may develop a new methodological approach or wish to collaborate with another researcher who is interested primarily in a different aspect of the data, such as interactional process using discourse analytic approach. Grounded Theory and Discourse Analysis use very different units of analysis, transcription formats and levels of detail regarding what aspects of the audio or video are transcribed, e.g. pauses, breaths, and other non-verbal gestures. If the source audio and video have been destroyed after the initial transcription, there is no way to go back and reanalyze the data using a different methodological lens. This is akin to asking statisticians to destroy their quantitative data sets, or pathologists to destroy their tissue samples as soon as they have completed their preliminary analyses, thus precluding them from reanalyzing the data as diagnostic techniques, analytic methods and theoretical approaches evolve. For these reasons, it is unreasonable to set an arbitrary time limit for analyzing video and audio data before destroying it. Protocols for the destruction of tapes and physical media should not be applied in the same manner to digital media that is properly stored and encrypted. We will store all video data securely in a locked cabinet on encrypted DVD for use in future studies about participation in clinical research. We will destroy all video data by December 31, 2020.

Our transcriptionist, Paul Garton, is experienced in focus group and video transcription. He has been trained to use Transana software and follows a strict data management and security protocol developed by Dr. Sheon. Because Mr. Garton works off campus, we use the following steps in order to ensure data security during the transcription process. The encrypted volume containing the focus group video will be transmitted to Mr. Garton using his YouSendIt service account. Dr. Sheon uploads an encrypted volume containing video using Mr. Garton's YouSendIt portal on his web site: transcriptcoop.com. Once the encrypted volume has been uploaded to the YouSendIt data center, Mr Garton receives an email that a new file has been sent, Mr. Garton clicks on a link in the email which downloads the file to his computer from one of YouSendIt's data centers. In order to protect data integrity during file transfer, YouSendIt employs the Secure Socket Layer (SSL) that implements industry-standard, 128-bit SSL encryption deployed using Class 3 certificates and Server-Gated Cryptography (SGC). A description of YouSendIt's security protocols are described here: http://www.yousendit.com/cms/security. Once Mr. Garton has downloaded the file to his computer, he will enter the password provided by Dr. Sheon to mount the encrypted volume so that he can transcribe the video using Transana software. In this way, we use multiple layers of redundant encryption, email authentification, and passwords for both Truecrypt and Transana to secure the video and audio data sent to the transcriptionist. Once Mr. Garton has finished transcribing the data, he uploads the transcript onto the Transana database and then adds time codes at each new speaker turn in order to synchronize the video with the transcript. Once the transcript has been timecoded, he then dismounts the encrypted volume and deletes it from his computer and the YouSendIt server.

Transana software, which we will use to analyze the video data from the focus group and individual interviews, provides an additional layer of security. To access the study database, Transana software requires a user name and password to access annotations such as codes and transcripts of the data which are stored on the Transana server at the Center for AIDS Prevention Studies. These user credentials are administered by Dr. Sheon. All annotations will be anonymized and will not contain any personal identifiers. To access the study database on Transana, users must also have a copy of the video data available, and to do that, they would need to have TrueCrypt and know the encryption password. Unless the encrypted volume is mounted, attempts to open a transcript or video in Transana will return an error message saying that the media in question could not be found.

Further language related to our choice of encryption software.

First, in response to the committee's concern that a "free encryption software might not be a sufficient security measure", we have consulted with colleagues in the CAPS Methods Core, CAPS IT staff and SOM IT security experts about available encryption solutions that would fit the needs of our study. To quote Sarah Mays, Security Administrator for Enterprise Information Security (EIS), "Most of the laws do not focus on what type of encryption or how it's encrypted, just generally that it has encryption." Therefore it is useful to distinguish between what is the preferred encryption algorithm, XTS-AES for example, and the software used to encrypt the data. TrueCrypt uses an HHS approved algorithm, XTS-AES, for securing PHI. This algorithm is also considered sufficient for top-secret military data. In addition to using XTS-AES, TrueCrypt is open source software and thus its encryption procedures are open to public scrutiny and testing. Lively public discussion and testing of TrueCrypt's potential weaknesses have resulted in regular updates and improvements by TrueCrypt's developers. In our discussions of CHR's concerns with Stephen Lau, EIS Security Policy and Program Manager, Mr. Lau acknowledged that "TrueCrypt is a decent encryption mechanism" and that "I do not see any issues with how you are using TrueCrypt at this time."

TrueCrypt is currently the best option for this project for the following reasons. First of all, TrueCrypt is easy to use and is very fast. We feel that these are among the most important factors when choosing among encryption methods because if the method is cumbersome to use or if the software slows down the computer, users will tend to bypass it. My colleagues and I have used TrueCrypt and its predecessor, called E4M, since 2001. We therefore feel that our choice of TrueCrypt is justified in terms of stability, ease of use, and security.

Our choice of encryption software has also been influenced by compatibility considerations. The computers we are using for the data analysis of the video data are running Mac OS 10.6, aka "Snow Leopard." SOM has been working on a school-wide encryption solution but it has not yet been rolled out. The encryption software that is being considered by SOM IT is not yet compatible with the Mac 10.6 operating system. Thus, until a SOM-approved encryption product is available for us to evaluate and test on our computers for compatibility with Transana, we will continue to use TrueCrypt. In response to the CHR's request that we use "more rigorous security measures," we have found that Snow Leopard includes new FileVault encryption utility that encrypts the home folder of the Mac. Assuming this FileVault utility is compatible with Transana, we will use FileVault encryption in addition to TrueCrypt until such time that the SOM approved product is available and meets our needs.

Second, in response to the question of how long we will keep the data, we had proposed to keep the data indefinitely in order to be able to use the data in future studies. As we explained in the protocol, Transana's unit of analysis is the media file itself. Therefore, a transcript of the media file cannot be analyzed without the source media from which it was created. In light of the CHR's concerns, we will destroy the media files by December 31st, 2020 and have changed the protocol and the consent forms to reflect this. This date was chosen as our best estimate of the completion date of the larger RCT that relies on the recruitment and retention data we will collect in the current IMPACT study. Data from IMPACT may be useful in interpreting recruitment and retention issues in the larger trial and should therefore be available for future research.

Questions or comments about this document email nicolas.sheon@ucsf.edu. For questions about Transana, visit www.transana.org.