QualDash: Adaptable Generation of Visualisation Dashboards for Healthcare Quality Improvement

1 INTRODUCTION

• Healthcare QI dashboards must accommodate diverse performance metrics, audit data models, and user classes across specialties and hospitals.This diversity makes the appropriate balance between human involvement and automated dashboard adaptation an open research question.Source: §1 · p. 1
• QualDash uses Metric Specification Structures (MSSs), JSON templates that generate rearrangeable QualCards for metric-specific task sequences.The engine accepts an array of MSSs and generates the corresponding visualisation containers while presetting shared patterns.Source: §1 · p. 1
• The project followed a design-study process involving 54 interviews, co-design workshops, nine analyst meetings, prototype evaluation, and deployment in five NHS hospitals.Deployment included ten meetings with clinical and IT staff to adapt MSSs to newly arising tasks and inform a second version of QualDash.Source: §1 · p. 2
• The paper contributes task characterisation, the QualCard and MSS design, a dashboard generation engine, and reflection on 62 hours of deployment observation.These contributions connect healthcare QI task sequences to concise dashboard configuration and adaptation.Source: §1 · p. 2

2 BACKGROUND AND RELATED WORK

• Task spaces consolidate task taxonomies to help researchers reason about similarities and differences among domain-specific operations.Prior work uses task classifications as systematic design guides, view-specification constructs, and design checklists.Source: §2 · p. 2
• Quality dashboards inform standardized performance metrics at unit or organizational level, serving broader users and tasks than clinical dashboards.The paper focuses on healthcare quality dashboards because their users, contexts, and tasks vary widely.Source: §2 · p. 2
• Healthcare visualisation tools commonly allocate fixed screen space to data facets, supporting specialized events or patient cohorts rather than heterogeneous QI tasks.Examples include tools for ICU admissions or stroke patients.Source: §2 · p. 2
• Unlike Tableau, Vega, Vega-Lite, and Draco, QualDash encodes healthcare QI task sequences in concise templates for dynamic dashboard generation.The MSS covers a smaller design space but supports co-design and deployment, which the paper presents as the first attempt to match healthcare QI task sequences to dynamic dashboard generation.Source: §2 · p. 2

3 TASK ANALYSIS

• The analysis asked how healthcare QI task sequences, metric definitions, visual features, and dashboard update needs vary across contexts.It used National Clinical Audit data from PICANet and MINAP as the empirical setting.Source: §3 · p. 2, §3.1 · pp. 2–3

3.1.1 Interviews with Stakeholders

• Interviews with 54 healthcare professionals produced a thematic account of audit metrics and an initial pool of 124 user tasks.Participants included clinical, quality, safety, board, support, and information staff; a representative subset was selected by task granularity, type cardinality, and target.Source: §3.1.1 · p. 3

3.1.2 Co-design Workshop

• A half-day co-design workshop with seven clinical and non-clinical staff used task cards to elicit task context, relevance, inputs, outputs, beneficiaries, and current visualisations.Participants then selected entry-point questions, sketched minimally sufficient dashboards, and identified follow-up tasks.Source: §3.1.2 · p. 3

3.1.3 Structure of Task Sequences

• NCA data were used mainly at clinical-team level, where dashboards needed to monitor five to six key metrics for each audit.The analysis identified independent task sequences for individual metrics, with entry-point monitoring followed by subsidiary investigation.Source: §3.1.3 · p. 3
• Each metric’s entry-point tasks monitor a small number of measures over time before expanding into breakdowns, linked measures, or different temporal granularities.These recurring subsidiary tasks were labelled ST1, ST2, and ST3.Source: §3.1.3 · pp. 3–4

4 DESIGN REQUIREMENTS FOR QUALDASH

• Nine meetings with audit coordinators and clinicians iterated QualCard prototypes and identified minimally sufficient configuration properties.The participants included consultant cardiologists, a consultant pediatrician, and audit coordinators.Source: §4 · p. 4
• Analysts used their knowledge of audit data to specify field, type, and aggregate properties and to exclude unsupported or inaccessible tasks.Cross-site comparisons were excluded when they required data-sharing agreements.Source: §4 · p. 4
• The design process explored layered and multi-view plots within each QualCard to address the identified task sequences.The meetings produced a set of design requirements for configuring QualCards.Source: §4 · p. 4

• Each QualCard has an entry-point state for main measures and an expanded state with subsidiary views for ST1, ST2 and ST3 tasks.Source: R1 Support pre-configured reusable… · p. 4
• GUI controls allow users to adapt subsidiary measures, while data updates, quality indicators, exports, and on-site privacy support operational requirements.Source: R1 Support pre-configured reusable… · p. 4
• The architecture reads arrays of MSS configurations and audit data to generate the corresponding QualCards dynamically.Source: Figure 4 · p. 4, R1 Support pre-configured reusable… · p. 4

5 QUALDASH DESIGN

• QualDash runs client and server components locally within each hospital, using preprocessed audit data and server-side JSON configurations to generate dashboards.Source: §5 · pp. 4–5

5.1 The Metric Specification Structure (MSS)

• The MSS is a constrained, self-contained JSON specification that configures concatenated views and task sequences while leaving shared defaults to the engine.Source: §5.1 · p. 5
• MSS fields define metric descriptions, compact visual encodings, inclusion filters, legends, and subsidiary categories, quantities, and temporal context.Source: §5.1 · p. 5
• The yfilters extension records valid values and start/end events, enabling data-quality handling and temporal aggregation such as bed and ventilation days.Source: §5.1 · p. 5

5.2 The QualCard Interface

• QualCards present a grid-based entry point and expandable task-specific sub-views, with tabs mapping MSS categories, quantities, and times to user navigation.Source: §5.2 · p. 5, Figure 5 · p. 6, Figure 6 · p. 6
• Co-design favored bar charts for main measures, while MSS logic can produce grouped or stacked bars and dual-axis views combining counts with running averages.Source: §5.2 · p. 6
• Linked brushing connects categorical and temporal views, allowing selected periods to reveal category distributions such as 51% recovering from surgery and 4% with bronchiolitis.Source: §5.2 · pp. 6–7
• Evaluation led participants to prefer navigable multi-series line charts over the early small-multiples design for temporal sub-views.Source: §5.2 · p. 7

6.1 Off-site Evaluation

• Off-site feedback supported consistent data views while identifying aggregation, validation, timeliness, readability, and new-functionality requirements.Participants requested cumulative SMR aggregation, clearer visual presentation, and an adverse-event feature that was implemented through enhanced MSS and QualCard fields.Source: §6.1 · p. 8
• QualDash accommodated requested task adaptations through MSS changes, including a “No PCI data available” measure for call-to-balloon analysis.The measure identified STEMI patients with missing reperfusion-treatment dates, distinguishing absent intervention from missing audit data.Source: §6.1 · p. 8
• SUS scores of 74 and 89.5 indicated very good usability across the two think-aloud groups.The evaluation combined qualitative feedback with System Usability Scale questionnaires administered after participants completed dashboard tasks.Source: §6.1 · p. 8

6.2.1 Support for Data Timeliness

• QualDash supported site-specific upload schedules, reducing one quarterly extraction task from up to two hours to 10 minutes.Sites used monthly, weekly, or daily uploads according to local needs, although schedules were generally ad hoc and driven by need.Source: §6.2.1 · p. 8

6.2.2 Safe Interpretation: A Case Study

• The Gold Standard Drugs case showed that removing and restoring a metric card allowed eligibility definitions to be verified before visualisations were used for interpretation.The card was temporarily removed when its population definition was uncertain, then updated after supplier consultation and reintroduced.Source: §6.2.2 · p. 9

6.2.3 Perceived Usefulness: A Case Study

• The mortality case supported investigation and reporting by combining SMR trends, deaths, case-mix measures, PIM scores, and exportable contributing records.A consultant reported extracting deaths contributing to a high SMR in 15 seconds, while PIM-score views supported staffing discussions.Source: §6.2.3 · p. 9

Conclusions

• QualDash’s concise MSS specification made it straightforward to correct misunderstood tasks, add new tasks, and support site-specific dashboard changes across five hospitals.Examples include correcting the call-to-balloon metric, accommodating adverse-event tasks, and making local changes.Source: §7 · p. 9
• Moderated MSS authoring, modular QualCards, and sequenced view expansion supported safer interpretation, focused communication, and task-oriented exploration.The authors note that further evidence is needed to establish MSS usability as an authoring tool.Source: §7 · p. 9

Future Work

• Current QualDash supports only time and population referrer types, while future development could add geographic location and other referrers.These extensions were motivated by discussions about risk factors, vulnerable communities, and geospatial visualisation needs.Source: §7 · p. 9
• Cross-QualCard transitions remain a future design requirement because they were unnecessary for the authors’ healthcare QI dashboards but may matter in other applications.The modular, self-contained QualCard is expected to localize future transition design decisions.Source: §7 · p. 9