Module 5a: Data & Digital Infrastructure

MHIA 6160 | Value & Quality in Healthcare | 1.50 | W27

Dr. Cameron B. Chiarot

Module 5a.1

The Data Lifecycle

Tracking Value Across the Data Pipeline


Data Literacy for Strategic Health Leaders

  • Core Premise: Data in modern healthcare is not a passive administrative byproduct; it is a dynamic, high-value strategic asset.
  • Strategic Shift: Move from a linear checklist to a continuous loop where system insights actively optimize future clinical outcomes.
  • Value Framework: Evaluate each stage to see how it alters the value equation.

Operational Reality

If the data lifecycle breaks at any stage, clinical quality decreases and costs increase (potentially compounding).

The Canadian Healthcare Data Landscape

Jurisdictional Silos vs. Commercial Scaling

System Lens (Public Mandates):

  • Governed by provincial ministries under the CHA.
  • Relies on single-payer data repositories focused on population equity (e.g., CIHI).

Industry Lens (Private Innovation):

  • Driven by fast-moving commercial vendors (e.g., Telus Health, WELL Health).
  • Optimized for clinical volume and micro-efficiency.

Core Friction:

  • Patchwork data architectures create systemic barriers, stranding critical clinical information inside disconnected organizational silos.

The Data Lifecycle: The Big Picture


The Five-Stage Value Pipeline

  1. Collection & Ingestion: Capture raw clinical events & ops transactions at POC.
  2. Storage & Security: Secure digital assets & maintain privacy compliance laws.
  3. Integration & Interoperability: Connect separate tech stacks using unified data transmission standards.
  4. Analytics & Knowledge Extraction: Process raw databases into predictive models and macro performance metrics.
  5. Presentation & Action: Feed insights directly into clinical desks and ministry policy tables.

Stage 1: Collection & Ingestion

Public Sector Stream:

  • Standardized hospital EHR/EMR networks, lab registries, & mandatory billing databases.

Private Sector Stream:

  • Consumer virtual care apps, wellness wearables, & private diagnostic lab ops.

The Structural Hurdle:

  • Converting messy, unformatted clinical notes into structured, standardized data tables.

Stage 2: Storage, Security & Sovereignty

Regulatory Guardrails:

  • Strict compliance frameworks across provincial boundaries
  • e.g., Ontario’s PHIPA, FIPPA, and federal Bill C-27

Data Residency Mandates:

  • Legal restrictions requiring Canadian health data to physically reside on servers within domestic borders.

Infrastructure Divergence:

  • Conservative public reliance on local, on-premise hardware vs. rapid private deployment of secure commercial cloud nodes (e.g., AWS/Azure Canada).

Stage 3: Integration & Interoperability

Communication Standards:

  • Implementing HL7 & FHIR (Health Level Seven & Fast Healthcare Interoperability Resources) data models to standardize interface connections.

Public & Private Problems:

  • Public Bottlenecks: Decades of legacy/obsolete hospital tech delay real-time data exchanges.

  • Private Assets: Data hoarding by software vendors seeking to lock clients out of competing platform ecosystems.

Pan-Canadian Strategic Goal:

  • Building a digital health record that travels seamlessly across the care continuum.

Stage 4: Analytics & Knowledge Extraction

Public Sector Objectives:

  • Population-level health indexing, funding model modifications, and strategic wait-time reductions.

Private Sector Objectives:

  • Localized operational turnaround optimization, personalized medicine paths, and health insight commercialization.

The Speed Mismatch:

  • Rigorous, slow academic Research Ethics Board (REB) reviews vs. rapid corporate R&D agile deployment cycles.

Stage 5: Presentation, Dissemination & Action

Policy-Level Translation:

  • Presenting aggregated operational indicators directly to health ministry funding committees.

Point-of-Care Translation:

  • Embedding automated clinical decision support tools straight into front-line medical software.

Stakeholder Visual Standards:

  • Building clean, accessible executive dashboards that let non-technical leaders isolate and act on system friction points.

Healthcare Data Lifecycle Acronyms


Acronym Definition Primary Context & Purpose
CACDI Canadian Core Data for Interoperability Standardizes specific data elements to allow secure, seamless sharing across provincial borders.
CHIMA Canadian Health Information Management Association The national organization that certifies and represents health information professionals.
CIHI Canadian Institute for Health Information Provides essential, pan-Canadian data and analysis on hospital care, financing, and workforce trends.
DGSI Digital Governance Standards Institute Establishes the national consensus standards for data trust and governance.
EMR Electronic Medical Record The primary clinical software application used by providers to record daily patient encounters.
HIA Health Information Act Alberta's strict provincial legislation governing the access and privacy of health records.
OCAP® Ownership, Control, Access, and Possession The foundational principles ensuring First Nations communities retain ownership over their health data.
PHI Personal Health Information Any core medical data—such as lab results or diagnoses—that can be linked to a specific individual.
PHIPA Personal Health Information Protection Act Ontario's primary health-specific privacy law regulating custodians of patient records.
PIPEDA Personal Information Protection and Electronic Documents Act The overarching federal privacy law governing commercial and private-sector data management.

Section 5a.1 Think-Pair-Share: The Lifeline Disconnect


Interactive Think-Pair-Share

A patient records persistent heart rate anomalies and cardiac spikes using a commercial, private smartphone application. Hours later, the patient presents at a public hospital Emergency Department experiencing acute chest pain. The attending ER physician has no technical method to access or review the application’s recorded data stream.


  • Think (1 min): At exactly what stage of the data lifecycle did this healthcare system break down?
  • Pair (2 mins): Is this disconnect a standard failure (technical), a corporate choice (commercial), or a regulatory wall (legal)?
  • Share (3 mins): As a future executive, which specific policy or operational lever would you pull to bridge this gap?


Section 5a.1 Discussion Prompt


Core Governance Question

If an innovative, private digital platform is entirely funded via public health insurance dollars to deliver virtual clinical care to Ontario residents, who should maintain ultimate ownership and secondary use rights over the generated patient data: the private enterprise or the public sector?


Module 5a.2

Dual Perspectives on Artificial Intelligence

From Manual Curation to Deep Learning Scalability


The Core Breakthrough (Rajkomar et al., 2018)

  • The Interoperability Bridge: Raw EMR data is notoriously messy and siloed. FHIR standardizes this data into a uniform temporal timeline, eliminating the traditional, labour-intensive manual data harmonization that discards up to 80% of clinical records.
  • Direct Feature Learning: Because FHIR standardizes the format, deep learning models can ingest raw, unharmonized timelines directly and outperform hand-crafted risk scores.
  • Strategic Scalability: Health systems can now deploy a single, unified deep learning architecture across multiple disparate medical centres.

The System Lens: Macro Efficiency & Capacity Forecasting


Operationalizing the Aggregate

  • Predictive Targets: Scale models to forecast (e.g., mortality, readmissions, & prolonged length of stay).
  • System-Wide Value: Ingest FHIR feeds for optimization (e.g., hospital bed availability, allocation of nursing teams, & manage resources).

The System Imperative

  • Goal: Maximize clinical throughput and protect fiscal sustainability.
  • Blindspot: Models optimized for net system throughput can accidentally reinforce historical biases, resulting in unequal care patterns for marginalized groups.

The Industry & Patient Lens: Personalization vs. Privacy


The Micro-Level Value Equation

  • Hyper-Personalized Medicine: Shifting clinical interventions from generic protocols to highly tailored pathways.
  • Patient-Centred Benefits: Faster diagnostic turnarounds, automated outpatient safety alerts, and reduced duplicate testing.
  • Commercial Reality: Private technology vendors use advanced predictive layers to build competitive, high-margin software features that attract premier hospitals.

The Patient Risk Profile

As clinical predictive power scales, how do we protect absolute individual data privacy when large corporate entities require access to comprehensive public EMR data blocks to train their algorithms?

The Interface Friction: When Lenses Collide

The Explainability Wall:

  • Deep learning models function as complex black boxes. System administrators value the macro accuracy, but clinicians often reject warnings they cannot verify or interpret.

Liability Fragmentation:

  • When a model generates an incorrect risk prediction, who is legally accountable for the adverse outcome: public system, doctor, or private software developer?

Resource Conflict:

  • System algorithms may tag an administrative readmission risk as low to save public beds, directly conflicting with a family’s personal desire for protective inpatient observation.

Module 5a.3

Digital & Algorithmic Governance

Pillars 1 & 2: Human-Centric Systems

Beneficence (Do Good)

  • Focus: Ensure systems actively promote population well-being, maximize resource utilization, and improve clinical quality.
  • System Lens: e.g., Deploy predictive clinical models to reduce provincial ER wait times or catch diagnostic anomalies early.
  • Industry Lens: e.g., Create user-friendly remote patient tracking portals to improve adherence.

Non-Maleficence (Do No Harm)

  • Core Focus: Protecting individual privacy, enforcing cybersecurity, and maintaining capability caution.
  • The Risk Vector: Data leaks from third-party vendor or rogue clinical alerts causing physician desensitization.
  • System Safeguard: Mandatory algorithmic risk assessments before clinical deployment.

Pillars 3 & 4: Equity and Agency Boundaries

Autonomy (The Power to Decide)

  • Core Focus: Striking a clear balance between human choice and automated decision delegation.
  • The Clinical Border: Guarding against automation bias (e.g., doctors blindly follow algorithmic directives).
  • The Mandate: Ensuring AI acts exclusively as an supportive advisor, leaving final diagnostic accountability to the clinician.

Justice (Preserving Fairness)

  • Core Focus: Eradicating algorithmic bias and preventing digital inequalities.
  • The Healthcare Mismatch: Algorithms trained on data from well-funded urban academic hospitals often fail when deployed in underfunded rural or remote communities.
  • The Metric: Measuring performance across disaggregated cohorts to verify equitable resource distribution.

The Additional Pillar 5: Explicability

The Data Pipeline Connector:

  • Explicability is what bridges technical data engineering (5a.1) and clinical prediction (5a.2).
  • If data collectors cannot audit how variables are mapped, and clinicians cannot interpret how a model predicts risk, the entire data-to-action collapses.

The Double Definition:

  • Explicability demands answering two distinct questions for any active clinical algorithm:

    1. Epistemological: How does it work? (Intelligibility / Interpretability)
    2. Ethical: Who is responsible for the way it works? (Accountability / Liability)

Pillar 5: Explicability (cont.)


System Lens Requirements

  • Public administrators require clear audit trails to prove regulatory compliance, safeguard funding alignment, and maintain patient safety.

Industry Lens Realities

  • Commercial tech companies protect proprietary neural weights as trade secrets, directly conflicting with public demands for open-source clarity.

Algorithmic Governance in the Canadian Regulatory Context


The Legislative Frontier

  • Federal Oversight: Bill C-27 and the Artificial Intelligence and Data Act (AIDA) set strict regulatory compliance thresholds for high-impact AI systems.
  • Provincial Enforcement: Information and Privacy Commissioners enforcing strict penalties for non-compliant secondary health data utilization.
  • The Executive Reality: Ignorance of algorithmic mechanics is no longer an acceptable legal defence for hospital boards or corporate wellness CEOs.

Module 5a Summary: Data & Digital Infrastructure

  • Data Lifecycle (Value Pipeline): Health data is a strategic asset. Navigating collection, secure storage, interoperability, analytics, and action requires bridging public sector mandates and private industry innovations.
  • Dual Perspectives on AI: Moving from manual curation to deep learning scalability offers massive gains in macro system capacity and micro-personalization, but introduces black-box friction and liability challenges.
  • Algorithmic Governance: Ethical deployment demands adapting the four pillars of bioethics alongside Explicability to ensure transparency, equity, and accountability under evolving Canadian regulations.

Core Executive Takeaway

Effective health leaders do not just manage technology—they actively govern the intersection of public equity, private innovation, and patient trust across the entire digital lifecycle.

Intentionally Blank

{end}