Eurographics Workshop on Visual for Biology and Medicine (2019) Short Paper K. Lawonn and R. G. Raidou (Editors)

MedUse: A Visual Analysis Tool for Medication Use Data in the ABCD Study

H. Bartsch1,2 and L. Garrison5 and S. Bruckner5 and A. (Szu-Yung) Wang4 and S. F. Tapert3 and R. Grüner1

1Mohn Medical and Centre, Haukeland University Hospital, Bergen, Norway 2Center for Multimodal Imaging and Genetics, University of California San Diego, La Jolla, United States 3Department of Psychiatry, University of California San Diego, La Jolla, California, United States 4Laboratory of , National Institute on Alcohol Abuse and Alcoholism, Bethesda, Maryland, United States 5Department of Informatics, University of Bergen, Norway

58930 Zyrtec #209 1186679 Zyrtec Pill #29 865258 Aller Tec #18 203150 cetirizine hydrochloride #15 1186677 Zyrtec Oral Liquid Product #6 1152447 Cetirizine Pill #5 1086791 Wal Zyr #5 1366498 Ahist Antihistamine Oral Product #4 1428980 Cetirizine Oral Solution [Pediacare Children's 24 Hr Allergy] #3 1020026 cetirizine hydrochloride 10 MG Oral Tablet [Zyrtec] #3 1020023 cetirizine hydrochloride 10 MG Oral Capsule [Zyrtec] #3 371364 Cetirizine Oral Tablet #3 1366499 Ahist Antihistamine Pill #2 1296197 Zyrtec Disintegrating Oral Product #2

R06AE Piperazine derivatives #209 1186680 Zyrtec D Oral Product #2 1020021 cetirizine hydrochloride 1 MG/ML Oral Solution [Zyrtec] #2 759919 levocetirizine Oral Solution #2 398335 Xyzal #2 1595661 1595661 #1 1366500 Chlorcyclizine hydrochloride 25 MG Oral Tablet [Ahist Antihistam 1296338 Zyrtec Chewable Product #1 1187910 Wal Zyr Pill #1 1186678 Zyrtec Oral Product #1 1186638 Xyzal Pill #1 1020025 cetirizine hydrochloride 10 MG Oral Tablet [Alleroff] #1 1020016 cetirizine hydrochloride 1 MG/ML [Aller Tec] #1 1011483 cetirizine hydrochloride 10 MG [Zyrtec] #1 855168 levocetirizine dihydrochloride 0.5 MG/ML Oral Solution #1 371362 Cetirizine Oral Solution #1 353102 Zyrtec D #1 1694 Bonine #1

203576 Claritin #133 324026 Allegra #37 316167 Loratadine 10 MG #15 1171569 Claritin Pill #9 1369748 Allerclear #8 1297046 Claritin Chewable Product #6 707591 Alaway #6 1166659 Allegra Oral Liquid Product #4 1296601 Claritin Disintegrating Oral Product #3 R06 ANTIHISTAMINES FOR SYSTEMIC USE #209 R06A ANTIHISTAMINES FOR SYSTEMIC USE #209 1296571 Alavert Disintegrating Oral Product #3 1171571 Claritin D Pill #3 352401 Alavert #3 316168 Loratadine 5 MG #3 261705 Zaditor #3 1186869 Zaditor Ophthalmic Product #2

R06AX Other antihistamines for systemic use #133 1173504 Alavert D Pill #2 1171570 Claritin D Oral Product #2 1166661 Allegra Pill #2 1151461 Cyproheptadine Pill #2 836338 Loratadine 10 MG Oral Capsule [Claritin] #2 744830 Loratadine 10 MG Disintegrating Oral Tablet [Claritin] #2 1488053 Fexofenadine hydrochloride 180 MG Oral Tablet [Mucinex Non Drowsy Allergy] #1 1369752 Allerclear Pill #1 1171568 Claritin Oral Product #1 1171567 Claritin Oral Liquid Product #1 1171562 Clarinex D Pill #1 1165337 Loratadine Pill #1 1090510 Loratadine 10 MG [Wal vert] #1 1020126 Loratadine 10 MG Oral Tablet [Loradamed] #1 1020125 Loratadine Oral Tablet [Loradamed] #1 1020124 Loratadine 10 MG [Loradamed] #1 311372 Loratadine 10 MG Oral Tablet #1 216076 Claritin D #1

796199 Dimetapp Cold #8

R06AB Substituted alkylamines #8 353035 Triaminic Cold and Allergy #2 1294014 Brompheniramine Maleate 2 MG #1 1090458 1090458 #1 202384 Chlor Trimeton #1

153889 Singulair #46 1296363 Singulair Chewable Product #16

R03DC Leukotriene receptor antagonists #46 330047 montelukast 5 MG #13 115713 montelukast sodium #10 Figure 1: MedUse: Medication use analysis for the largest long-term study of brain development and child health1183694 Singulair Pill #7 in the United States. R03D OTHER SYSTEMIC DRUGS FOR OBSTRUCTIVE AIRWAY DISEASES #46 1157312 montelukast Pill #2 153892 montelukast 10 MG Oral Tablet [Singulair] #1

R03DX Other systemic drugs for obstructive airway diseases #1 356767 356767 #1

647295 ProAir #32 1649960 ProAir Inhalant Product #14 8887 Ventolin #13 R RESPIRATORY SYSTEM #209 142153 Albuterol Sulfate #10 1187868 Ventolin Inhalant Product #9 745752 200 ACTUAT Albuterol 0.09 MG/ACTUAT Metered Dose Inhaler [ProAir] #7

R03AC Selective beta 2 adrenoreceptor agonists #32 202908 Proventil #4 R03 DRUGS FOR OBSTRUCTIVE AIRWAY DISEASES #46 1008406 Albuterol / Ambroxol #2 346188 Albuterol 0.83 MG/ML #2 1182471 Proventil Inhalant Product #1 1154606 Albuterol Pill #1 Abstract 801095 60 ACTUAT Albuterol 0.09 MG/ACTUAT Metered Dose Inhaler [Ventolin R03A ADRENERGICS, INHALANTS #32 801094 Albuterol Metered Dose Inhaler [Ventolin] #1 330935 Albuterol 0.5 MG/ML #1 10369 10369 #1

50121 Fluticasone propionate #24 The RxNorm vocabulary is a yearly-published biomedical resource providing normalized names for medications.301543 Advair #17 It is used R03AK Adrenergics in combination with corticosteroids or other drugs, excl. anticholinergics #24 1372704 Dulera #5 30145 mometasone furoate #5 R03C ADRENERGICS FOR SYSTEMIC USE #0 1171309 Advair Inhalant Product #3 R03CC Selective beta 2 adrenoreceptor agonists #0 896212 60 ACTUAT Fluticasone propionate 0.25 MG/ACTUAT / salmeterol 0.05

R05C EXPECTORANTS, EXCL. COMBINATIONS WITH COUGH SUPPRESSANTS #23 R05CA Expectorants #23 to capture medication use in the Adolescent Brain Cognitive Development (ABCD) study, an active and352777 Mucinex #23 publicly available 1171649 Cough Out Oral Liquid Product #1

702274 Nyquil Cough #16

R05 COUGH AND COLD PREPARATIONS #23 1114038 DayQuil Cough #8 longitudinal research study following 11,800 children over 10 years. In this work, we present medUse, a3146 Delsym visual#8 tool allowing 353037 Triaminic Cough #4 672676 Mucinex Cough #3

R05DA Opium alkaloids and derivatives #16 702442 Zicam Cough #2 R05D COUGH SUPPRESSANTS, EXCL. COMBINATIONS WITH EXPECTORANTS #16 1373743 Dextromethorphan Hydrobromide 1 MG/ML / Guaifenesin 20 MG/ML Ora R05DB Other cough suppressants #0 1297101 Delsym Oral Liquid Product #1 researchers to explore and analyze the relationship of drug category to cognitive or imaging derived measures1183325 Mucinex DM Pill #1 using ABCD 1179871 Nyquil Cough Oral Liquid Product #1 727556 Dextromethorphan 5 MG / Guaifenesin 100 MG Oral Granules [Mucinex Children's Cough] #1 692906 Triaminic Day Time Cold & Cough #1 236146 dextromethorphan polistirex #1

study data. Our tool provides position-based contextR01B NASAL DECONGESTANTS FOR SYSTEMIC for USE #8 tree traversal and selectionR01BA Sympathomimetics #8 granularity of both study203302 Sudafed #8 participants and 1180816 Sudafed Oral Liquid Product #1 643092 Wal Phed #1 drug category. Developed as part of the Data Exploration and Analysis Portal (DEAP), medUse is available215051 Afrin #4 to more than 600 ABCD researchers world-wide. By integrating medUse into an actively used research product we are able to reach a wide audience and increase the practical relevance of visualization for the biomedical field. CCS Concepts • Human-centered computing → Information visualization; Activity centered design;

1. Introduction store, merge, and filter the available project's data. All of these data management steps incur costs for the data sharing platform and the The evolution of data sharing platforms into data analysis sys- individual researcher, which can hinder biomedical research. Data tems is driven by the complexity of the analysis required to gen- sharing platforms such as the NIMH Data Archive [Nata] there- erate publishable results, the complexity of the data shared, and fore add features for online data analysis such as the Data Analysis the costs involved in downloading and post-processing data. Us- and Exploration Portal (DEAP) [DEA]. This provides a unique op- ing conventional methods to access the data generated with obser- portunity for visualization to impact ongoing research in the fields vational studies such as the Adolescent Brain Cognitive Develop- of developmental science and neuropsychology. The DEAP sys- ment (ABCD) study [CCC∗18], a researcher needs to download, tem provides access to ABCD study data to individual researchers

© 2019 The Author(s) Eurographics Proceedings © 2019 The Eurographics Association. H. Bartsch, L. Garrison, S. Bruckner, A. Wang, S.F. Tapert & R. Grüner / medUse

ATC level L1 L2 L3 L4 RxNorm spiral arrangements [Sch11, Cui] with the goal to produce visually pleasing graphical representations. Space filling visualizations such as treemaps [Shn92] are more space conserving compared to many node-linked methods. They are well suited for displaying information located at the leaf nodes of the [GBP02], but they are more difficult to use for the task of selecting non-leaf nodes in large trees. Selection and nav- igation between neighboring nodes in the tree can result in focus movements in both horizontal and vertical planes with distance as the conserved organizational principle. Treemaps can be adjusted to allow for parent labels to be displayed [LMS∗18,KH14] but exam- ples presented in those implementations suggest that they are more tailored for short labels. Support for a focused display of labels at different scales has been suggested [KvK∗19]. Using camera dis- tance and occlusion to selectively show and position hierarchical labels of multi-instance objects for volumetric displays. Focus and context approaches have been suggested to help users to show particular tree nodes while providing context for the place- ment of the node in the hierarchy [LRP95, SCF11]. The 2D dis- ATC classified medications classified ATC tortion introduced to highlight a focus point in the graph can be severe and require global changes in the visual representation dur-

Medications ing navigation. Focus and context approaches can also be applied to more traditional approaches for displaying large hierarchical struc- Figure 2: ABCD medication use data organized by medUse. The tures such as indented lists [SQX∗06]. The large available horizon- data is separated into medications recognized by ATC. Individual tal space is ideal for longer node names but is more difficult for the ATC levels are marked as L1 through L4. The label RxNorm in- user to navigate and highlight the current location without extensive dicates the position of the captured medication names in RxNorm mouse or keyboard interactions. nomenclature. The RxNav browser is a web application for exploring the

RxNorm and ATC classification schemes [ZBKN06]. The soft- other ware supports a class view visualization that supports zoom and and performs direct hypothesis tests. Its modular design and large pan after searching for a specific medication name. Only a sin- user base provides a unique opportunity for new visual analysis gle parent chain from a single medication is displayed. Naviga- tools to be tested by the scientific community. In this work we have tion to neighboring nodes is cumbersome as it involves repeated extended DEAP, creating the first visualization-based application folding operation using drop-down menus. The tool does not al- focusing on the exploration of medication use data collected by low the user to visualize medication data for a study cohort. An the ABCD study, a long-term study of brain development and child interesting service that allows computations on study cohort data is health in the United States, collecting longitudinal data from 11,878 RxMix [PMNB13]. RxMix provides a graphical user interface to children. This data resource includes current medication use and compose medication data specific processing pipelines. Applied to medication use by the mother during pregnancy captured in parent user-provided lists of medications such a pipeline can, for example, questionnaires as structured medication names from the RxNorm translate between different medical nomenclatures. vocabulary [U.S11]. This data is shared annually as a spreadsheet with collected medication identifiers and names as columns per par- ticipant [ABC]. 3. Tasks The objective of medUse is to a) enable the user of the ABCD A typical analysis using the ABCD data resource would look at study to interactively browse the RxNorm coded medication use medication use behavior or psychological and cognitive side effects data using a meaningful level of abstraction and to b) allow the of classes of medications. Assigning the class labels to each par- user to create novel scores that sum over the individual medication ticipant in the cohort the researcher compares multi-dimensional entries and integrate into the DEAP system as first-class scores. class assignments statistically. For example, if there is a correla- tion between the reported use of ADHD medications and a cogni- tive outcome measure that correlates with attention or short-term 2. Related Work memory. The large number of similar prescriptions and brand and The literature on tree visualization is extensive and mainly focused generic versions of the same drug requires significant effort and ex- on the effective use of space [KHL∗07]. Separated into space filling pertise for the researcher to derive medication classes for such a and node-linked visualization methods these algorithms use a wide system level analysis. For ADHD related medications, for example, range of visual organizational principles such as 2D or 3D embed- our tool reports that there are 53 individual medications selected ding [DE01], distortions [LRP95], rectangular [SCF11], radial, or by the parents. As the study includes healthy kids, a large number

© 2019 The Author(s) Eurographics Proceedings © 2019 The Eurographics Association. H. Bartsch, L. Garrison, S. Bruckner, A. Wang, S.F. Tapert & R. Grüner / medUse

Table 1: ACT classification for the medication Adderall by organ system, therapeutic use, pharmacological subgroup, and chemical sub- group. An additional ATC level 5 for chemical substances is not used to simplify the overall tree layout.

level principle use class id node name root Medications drug in ATC Anatomical Therapeutic Chemical (ATC1-4) 1 organ system N Nervous System 2 therapeutic subgroup N06 Psychoanaleptics 3 pharmacological subgroup N06B Psychostimulants, Agents used for ADHD and Nootropics 4 chemical subgroup N06BA Centrally acting sympathomimetics RxNorm ABCD captured medication 84815 Adderall

of research topics would benefit from the inclusion of medication method to ignore hierarchical information and instead search for data. The effort required to generate, document and update these keywords found in node descriptions. Our application supports all manual classifications prompted the development of our medUse combinations of the available navigation modes i) mouse/touch, ii) application. keyboard/cursor, and iii) text searches and updates all components of the user interface accordingly for maximum view flexibility. 4. Approach Based on its intended use, one medication item is always dis- played in the center of the screen. From there it can be exported To support research in the field of developmental science and neu- as a binary classification vector coding 1 for study participants at a ropsychology we selected the Anatomical Therapeutic Chemical given visit that had at least one medication use reported for the class Classification System (ATC) [Wor], an ontology that provides in- and 0 otherwise. With the integration of medUse into the DEAP ap- formation to group individual medications hierarchically by organ plication this new data vector becomes part of the ABCD study data system, the mechanism of action or therapeutic use, and by phar- and is available in the statistical modules of DEAP. macological and chemical subgroup (see 1). An active class level removes the need for frequent mouse clicks The ATC classification system provides a fixed depth of five hier- to select a new level for inspection and provides automatic updates archy levels. We introduce one additional level to include the more of context information during all navigation steps. The informa- than 800 observed RxNorm medication types in Figure2. Using a tion presented includes access to online documentation as well as data-driven generation of the medication class hierarchy we limit textual and visual summary information for the number of partici- the ATC classifications to entries that have at least one reported pants selected by the current class, as depicted in Figure1 and using medication use in the selected group of study participants. For med- color to indicate the current class level. This allows the user to eas- ication items that belong to multiple ATC classes the specific ther- ily judge the specificity of the selection during the navigation step. apeutic use is unclear, so we use multiple assignments to allow for Sorting the branches of the tree by the number of the participants a correct statistical coverage of the multiple choices. This approach reporting drug use, the most frequent used medication class appears leads to a combined tree with more than 1,000 nodes and an aver- near the top of the tree, whereas less frequently used medications age branching factor of 24 (see Figure 2). are grouped closer to the bottom. Navigation from parent nodes to For visualization-naive users, the main target group of medUse, child nodes favors the top branches with the most frequently re- we developed a four-point navigation paradigm to guide naviga- ported medication classes. The height of the selected location in tion. This allows for more control by separating movements in the the medication tree therefore has meaning in the context of the user horizontal direction to indicate class specificity from movement task of finding an appropriate class for analysis. vertically in the tree to select alternative drug classes at the same specificity (Figure3). To support class comparison across differ- 4.1. Implementation ent branches we changed the default navigation strategy, allowing MedUse is implemented as a single-page web-application utilizing users to move between branches using the Up and Down arrow nav- the DEAP user authentication and authorization framework. This igation (see example in Figure 3). Each branching node also tracks ensures that users with valid access to the project data also have the last navigation choice as the user traverses the tree (right to access to the application. The integration as a DEAP component left). If the navigation direction is reversed, the previously recorded allows the application to focus on a single workflow leading to a choice is selected during tree traversal. Both of these changes in- cleaner interface design. Other DEAP components are contacted crease saliency of user choice and can prevent feelings of being by medUse to add derived data to the data dictionary and saving lost after navigation steps. Overall, this four-point navigation sys- co-variate data accessible to the statistical analysis tool. tem supports the user in recovering already visited nodes in the tree. Limiting the mental load required to perform the task medUse Whereas access to the study data is provided by the DEAP attempts to improved the overall user satisfaction. For experienced framework, the hierarchical ATC data is downloaded from the Na- users we also included a full text search as a drop-down option. This tional Library of Medicines ATC web-API [Natb] as JSON en- allows them to navigate quickly to a specific location, providing a coded text and cached by the user's client browser to improve the

© 2019 The Author(s) Eurographics Proceedings © 2019 The Eurographics Association. H. Bartsch, L. Garrison, S. Bruckner, A. Wang, S.F. Tapert & R. Grüner / medUse

fore sufficient for other users to replicate ABCD findings related to

X medication use. Our tool allowed the ABCD test users to visually explore hi- erarchical relationships on medication use for parents (during pregnancy) and youth by navigating from the most general class U (drug/supplement or no-drug) down to individual medications. This R1 + is done in the context of a study participant cohort dataset that is in- L terpreted and visualized with the help of the ATC ontology. This is R2 D in contrast to many previous approaches on graph visualization for ontological data [KHL∗07] that focus on the ontology itself, with- Figure 3: (Top) Example focus information for Melatonin linking out regard to a specific task in which that ontology is utilized. to search field and list of alternative uses and touch/mouse enabled Based on results from the user studies medUse is well suited four-point navigation fields (right). (Bottom) Navigation example to support visualization naive users in the selection of medication omitting the display of actual node labels for clarity. Current loca- classification groups from the ATC ontology. Users tend to use a tion is marked with (+). The four node-linked neighboring positions single navigation style and readily adopt the four-point navigation in the tree are Left (level L2), Right (e.g. R1, L4), and Down (L3). using the keyboard without additional training (see user study). Test The Up position (L3) is local only relative to the selected ATC level users not familiar with the nomenclature used to identify medica- 3, but not local given the ATC hierarchy. tion use data or the classification system for organ specific or thera- peutic uses of such medications have nevertheless been able to use medUse successfully. Compared to a manual re-coding of the med- ication information, the medUse approach is less error prone and startup time of the application. All data manipulations are imple- can be repeated easily to include other medication classes. MedUse mented client-side in JavaScript. This limits the requirements on fully automates the mapping of data to the ATC ontology and can the DEAP server infra-structure and allows the application to scale be used by users not trained in the classification of individual med- to a large number of simultaneous users. The tree visualization is ications. implemented using the D3.js library [BOH11]. Focus-level infor- mation was added as additional HTML/CSS elements placed on 6. Discussion and Conflusions top of the tree at a fixed screen location. During navigation the fo- cus level information is updated and the tree view moves so that the MedUse makes assumptions that might limit its use for analyzing new selection is in the center of the browser window. Animations other medication related information. First of all, the restriction of are used to move the tree improving the user's location awareness. only supporting a single classification system limits the choice for the user. One might argue that this reduces researcher degrees of freedom, but this is unhelpful if a required medication grouping 5. User Study and Feedback is not supported. The application cannot be used to create arbitrary user defined classifications. A possible extension of medUse to sup- To test the usability of the application we asked 3 researchers to port alternative ontologies are possible and requires further discus- participate in a user test. They were familiar with the DEAP anal- sion. MedUse also limits itself to the display of ontology elements ysis system but unfamiliar with medUse. They had varying levels relevant for a specific data set. It cannot be used to explore branches of expertise in analyzing medication use data. We instructed them in the ontology that are not supported by the current data in the to create a regression model analyzing the effects of brain struc- study. There are also problems associated with the use of the ATC tures on cognitive measures while correcting for reported medica- classification. According to the WHO [Wor], ATC codes are as- tion use related to attention deficit hyperactivity disorder (ADHD). signed frequently according to the mechanism of action rather than The task required them to create the summary measure for ADHD the therapy. Medication groups can therefore include medicines for medication in medUse and to use the generated score in another different indications, and medicines with similar therapeutic use DEAP application. All users did understand the hierarchical visu- may be classified in different groups. MedUse tries to support these alization using a node-linked tree and appreciated the level of detail ambiguities with highlighting multi-use medications. it provides for individual medications in relation to the therapeuti- The need to zoom in order to localize visually the current class cal subgroups. in relation to the overall ATC hierarchy can be problematic. At low Hypothesis generation in the DEAP statistical analysis involves zoom factors the hierarchy is visible, but font-sizes are scaled down the combination of multiple measures. We manually reviewed the and ATC text entries become unreadable. This favors desktop users variability of the user-generated hypotheses and observed a pref- with larger screen sizes and is not an issue for users that explore the erence of experts for adding medUse generated medication groups hierarchy locally at higher zoom factors, or for users that use full- at varying levels of specificity. Overall they tended to generate less text search to navigate larger ATC distances in the tree. It would than 10 alternative analysis models. MedUse supports the docu- have been possible to add additional visualization components to mentation of this user choice by generating fixed names for the ex- provide two simultaneous level of detail or geometric distortions, ported medication use measures based on the name for the node in but the benefits of such an addition may not outweigh the risk in- the tree. Sharing of the hypothesis specification on DEAP is there- volved with making the user interface more complex, and therefore

© 2019 The Author(s) Eurographics Proceedings © 2019 The Eurographics Association. H. Bartsch, L. Garrison, S. Bruckner, A. Wang, S.F. Tapert & R. Grüner / medUse less approachable for the target audience. It is worthwhile to ex- [KvK∗19]K OURILˇ D., Cˇ MOLIK L.,KOZLIKOVAˇ B.,WU H.,JOHNSON plore dynamic labeling approaches in more detail as they can re- G.,GOODSELL D.S.,OLSON A.,GRÖLLER M.E.,VIOLA I.: Labels duce visual complexity at different zoom levels. on levels: Labeling of multi-scale multi-instance and crowded 3d biolog- ical environments. IEEE Transactions on Visualization and Computer As the ABCD study releases more data we plan to continue op- Graphics 25, 1 (Jan 2019), 977–986. doi:10.1109/TVCG.2018. 2864491 timizing and extending this visualization tool to support futher re- .2 ∗ search in developmental science and neuropsychology. [LMS 18]L I D.,MEI H.,SHEN Y., SU S.,ZHANG W., WANG J.,ZU M.,CHEN W.: Echarts: A declarative framework for rapid construction of web-based visualization. Visual Informatics 2, 2 Acknowledgments (2018), 136 – 146. URL: http://www.sciencedirect. com/science/article/pii/S2468502X18300068, This work was supported in part by the National Institutes of doi:10.1016/j.visinf.2018.04.011.2 Health, grant number U24DA041123-01, the Visualizing Data [LRP95]L AMPING J.,RAO R.,PIROLLI P.: A focus+context tech- Science for Large Scale Hypothesis Management in Imaging nique based on hyperbolic geometry for visualizing large . Biomarker Discovery (VIDI), and by the University of Bergen and In CHI’95: Proceedings of the SIGCHI Conference on Human Factors in Computing Systems (1995), Katz I. R., Mack R., Marks L., Rosson the Trond Mohn Foundation in Bergen (#813558). M. B., Nielsen J., (Eds.), ACM Press/Addison-Wesley Publishing Co., pp. 401–408. doi:10.1145/223904.223956.2 References [Nata] re3data.org: National Institute for Mental Health Data Archive. re3data.org - Registry of Research Data Repositories. URL: http:// [ABC] ABCD Study data release 1.1. online. URL: http:// data-archive.nimh.nih.gov, doi:10.17616/R3XV5P.1 abcdstudy.org, doi:10.15154/1460410.2 [Natb] National Library of Medicine. online. URL: [BOH11]B OSTOCK M.,OGIEVETSKY V., HEER J.:D3 data-driven doc- https://www.nlm.nih.gov/research/umls/rxnorm/ uments. IEEE Transactions on Visualization and 17, sourcereleasedocs/atc.html.3 12 (Dec 2011), 2301–2309. doi:10.1109/TVCG.2011.185.4 [PMNB13]P ETERS L.,MORTENSEN J.,NGUYEN T., BODENREIDER ∗ [CCC 18]C ASEY B.,CANNONIER T., CONLEY M.I.,COHEN A.O., O.: Enabling complex queries to drug information sources through BARCH D.M.,HEITZEG M.M.,SOULES M.E.,TESLOVICH T., DEL- functional composition. Studies in health technology and informatics LARCO D. V., GARAVAN H.,ORR C.A.,WAGER T. D., BANICH 192 (2013), 692–696. URL: https://www.ncbi.nlm.nih.gov/ M. T., SPEER N.K.,SUTHERLAND M. T., RIEDEL M.C.,DICK pubmed/23920645.2 A.S.,BJORK J.M.,THOMAS K.M.,CHAARANI B.,MEJIA M.H.,HAGLER D.J.,CORNEJO M.D.,SICAT C.S.,HARMS [SCF11]S ONG H.,CAI X.,FU Y.: Rectangular tree browser: M. P., DOSENBACH N.U.,ROSENBERG M.,EARL E.,BARTSCH a navigation and visualization tool for large hierarchies. Journal H.,WATTS R.,POLIMENI J.R.,KUPERMAN J.M.,FAIR D.A., of Information and Computational Science 8, 2 (2011), 354–361. http://www.joics.com/publishedpapers/2011_8_ DALE A. M.: The Adolescent Brain Cognitive Development (ABCD) URL: study: Imaging acquisition across 21 sites. Developmental Cognitive 2_354_361.pdf.2 Neuroscience 32 (2018), 43 – 54. The Adolescent Brain Cognitive [Sch11]S CHULZ H.: Treevis.net: A tree visualization reference. IEEE Development (ABCD) Consortium: Rationale, Aims, and Assessment Computer Graphics and Applications 31, 6 (Nov 2011), 11–15. doi: Strategy. URL: http://www.sciencedirect.com/science/ 10.1109/MCG.2011.103.2 article/pii/S1878929317301214, doi:10.1016/j.dcn. [Shn92]S HNEIDERMAN B.: Tree visualization with tree-: 2-d 2018.03.001.1 space-filling approach. ACM Trans. Graph. 11, 1 (Jan. 1992), 92–99. [Cui]C UI W.: A Survey on Graph Visualization. online. URL: http: doi:10.1145/102377.115768.2 //www.cse.ust.hk/~weiwei/PQE/WeiweiPQE.pdf.2 [SQX∗06]S ONG H.,QI Y., XIAO L.,ZHU T., CURRAN E. P.: Lenstree: [DE01]D ACHSELT R.,EBERT J.: Collapsible cylindrical trees: A fast Browsing and navigating large hierarchical information structures. In hierarchical navigation technique. In InfoVis’01: Proceedings of the 16th International Conference on Artificial Reality and Telexistence– IEEE Symposium on Information Visualization (2001), Andrews K., Workshops (ICAT’06) (Nov 2006), pp. 682–687. doi:10.1109/ Roth S., Wong P. C., (Eds.), IEEE Computer Society, pp. 79–86. doi: ICAT.2006.81.2 10.1109/INFVIS.2001.963284.2 [U.S11] U.S. National Library of Medicine, National Institutes of Health, [DEA] DEAP - Data Exploration and Analysis Portal, Bethesda, Maryland, USA, RxNorm API. online, 2011. URL: http: RRID:SCR_016158. online. URL: https://deap.nimhda.org/. //rxnav.nlm.nih.gov/RxNormAPI.html.2 1 [Wor] World Health Organization Collaborating Center for Drug Statis- [GBP02]G ROSJEAN J.,BEDERSON B.B.,PLAISANT C.: Spacetree: tics Methodology. online. URL: https://www.whocc.no/atc/ Supporting exploration in large node link tree, design evolution and structure_and_principles/.3,4 empirical evaluation. In Information Visualization, IEEE Symposium [ZBKN06]Z ENG K.,BODENREIDER O.,KILBOURNE J.,NELSON S.: on (Los Alamitos, CA, USA, oct 2002), IEEE Computer Society, RxNav: a web service for standard drug information. AMIA ... Annual p. 57. URL: https://doi.ieeecomputersociety.org/ Symposium proceedings. AMIA Symposium 2006 (2006), 1156. URL: 10.1109/INFVIS.2002.1173148, doi:10.1109/INFVIS. https://www.ncbi.nlm.nih.gov/pubmed/17238775.2 2002.1173148.2 [KH14]K IM K.,HAN J.: Label embedded treemapping: A label over- lap prevention technique for zoomable treemaps and a user interaction technique. In Human Interface and the Management of Information. Information and Knowledge Design and Evaluation (Cham, 2014), Ya- mamoto S., (Ed.), Springer International Publishing, pp. 44–53.2 [KHL∗07]K ATIFORI A.,HALATSIS C.,LEPOURAS G.,VASSILAKIS C.,GIANNOPOULOU E.: Ontology visualization methods - a sur- vey. ACM Computing Surveys 39 (11 2007), 10. doi:10.1145/ 1287620.1287621.2,4

© 2019 The Author(s) Eurographics Proceedings © 2019 The Eurographics Association.