Carolinas HealthCare System: Consumer Analytics Custom Case Solution & Analysis

1. Evidence Brief: Carolinas HealthCare System (CHS) Consumer Analytics

Financial Metrics and Market Context

• System Scale: CHS operates as one of the largest non-profit healthcare systems in the United States, with over 900 care locations, 7,460 licensed beds, and 60,000 employees (Case Paragraph 4).
• Market Shift: Transition from fee-for-service to value-based care models, specifically driven by the Medicare Access and CHIP Reauthorization Act (MACRA) and the rise of Accountable Care Organizations (ACOs) (Exhibit 1).
• Data Investment: CHS purchased consumer data on 2 million individuals across North and South Carolina. This includes 300 to 500 data points per person, including purchasing history and credit scores (Case Paragraph 12).
• Revenue Model: Approximately 25% of CHS revenue is tied to value-based contracts where the system assumes financial risk for patient outcomes (Case Paragraph 8).

Operational Facts

• Data Integration: The system combines clinical Electronic Health Record (EHR) data with external consumer data to generate predictive risk scores for readmission and chronic disease management (Case Paragraph 15).
• The Dictionary: CHS created a Master Patient Index (MPI) to link disparate data sources to a single individual identity across the 900+ locations (Case Paragraph 14).
• Predictive Modeling: Models target high-risk patients for interventions, such as those with a high probability of developing Type 2 diabetes or heart failure (Case Paragraph 18).
• Pilot Scope: Initial implementation focused on the Dickson Advanced Analytics (DA2) group, a centralized team of 50 data scientists and analysts (Case Paragraph 10).

Stakeholder Positions

• Dr. Michael Dulin (Chief Clinical Officer for Analytics): Primary advocate for using non-clinical data to address social determinants of health. Maintains that 80% of health outcomes are determined outside the clinic (Case Paragraph 6).
• Front-line Physicians: Expressed significant concern regarding data accuracy and information overload. Many view consumer data as creepy or an intrusion into the doctor-patient relationship (Case Paragraph 22).
• Patients: Generally unaware that their purchasing habits (e.g., buying tobacco or high-sodium foods) are being tracked by their healthcare provider (Case Paragraph 25).
• IT/Data Teams: Focused on the technical challenge of cleaning and normalizing third-party data to make it actionable within the EHR (Case Paragraph 13).

Information Gaps

• Specific ROI: The case does not provide the exact dollar amount saved per patient through consumer-data-driven interventions.
• Vendor Costs: The annual licensing fee for the third-party consumer data is not disclosed.
• Patient Opt-out Rates: There is no data on how many patients would opt out of this tracking if given an explicit choice.

2. Strategic Analysis

Core Strategic Question

• How can CHS transform consumer data into clinical interventions without alienating physicians or triggering a patient privacy backlash?
• How can the system justify continued investment in non-clinical data as value-based care margins remain thin?

Structural Analysis

The transition to value-based care shifts the hospital’s role from a repair shop to a health manager. The Value Chain analysis reveals that the bottleneck is not data acquisition, but data translation. While CHS has successfully integrated 2 million records, the final mile—physician action—is broken. Doctors lack the time to interpret why a patient’s low credit score or frequent fast-food purchases matters during a 15-minute clinical encounter.

Strategic Options

• Option 1: The Centralized Care Management Model. Remove consumer data from the physician’s view entirely. Funnel insights to a centralized team of social workers and nurse care managers who handle outreach and social determinant interventions.
  • Rationale: Protects the physician-patient relationship and prevents information overload.
  • Trade-offs: Increases overhead costs for non-clinical staff; creates a disconnect between the primary doctor and the care plan.
• Option 2: The Transparent EHR Integration Model. Embed consumer risk scores directly into the EHR with specific, automated clinical decision support (CDS) prompts for the physician.
  • Rationale: Ensures the physician remains the central authority in patient health.
  • Trade-offs: High risk of physician burnout and active resistance to data they deem non-medical.
• Option 3: The Patient-Facing Empowerment Model. Use the data to push insights directly to patients via a portal or mobile app, bypassing the clinician for behavioral nudges.
  • Rationale: Scalable and addresses the 99% of time patients spend outside the clinic.
  • Trade-offs: Significant privacy risks; limited effectiveness for high-risk populations with low digital literacy.

Preliminary Recommendation

CHS should adopt Option 1: The Centralized Care Management Model. The current physician workflow is too rigid to accommodate non-clinical data. By using the DA2 group to trigger interventions by social workers, CHS addresses the social determinants of health (the 80%) without burdening the clinical staff (the 20%).

3. Operations and Implementation Planner

Critical Path

• Month 1-2: Segmenting the Risk Pool. Identify the top 5% of patients where consumer data (e.g., housing instability, food desert residence) correlates most strongly with 30-day readmission.
• Month 3-4: Care Manager Deployment. Hire or reassign 20 social workers to act as Data Interventionalists. They will receive daily alerts from the DA2 dashboard, not the EHR.
• Month 5-6: Protocol Standardization. Develop scripts for patient outreach that frame the intervention around support services (transportation, food pharmacy) rather than disclosing the specific consumer data used to find them.
• Month 9: Physician Feedback Loop. Provide doctors with a monthly summary of how many of their patients were helped by the centralized team, proving value without requiring their active participation in data analysis.

Key Constraints

• Data Privacy Regulation: While HIPAA governs clinical data, consumer data sits in a gray area. Any change in FTC or state-level privacy laws regarding purchasing data could invalidate the entire model.
• Labor Market: Finding care managers who can bridge the gap between data analytics and empathetic patient outreach is a significant hiring hurdle in the Carolinas market.

Risk-Adjusted Implementation Strategy

The primary risk is the perception of surveillance. To mitigate this, the implementation will avoid the term consumer data in all patient-facing and physician-facing communications, replacing it with Community Support Indicators. If physician resistance exceeds 30% in initial surveys, the system will pivot to an opt-in model for clinicians, starting only with those in high-risk primary care pilots.

4. Executive Review and BLUF

BLUF (Bottom Line Up Front)

CHS must immediately decouple consumer analytics from physician workflows. The current strategy of pushing non-clinical data to doctors creates friction and risks a professional revolt. Instead, CHS should utilize its DA2 analytics unit to power a centralized Care Coordination Hub. This hub will deploy social workers to intervene in the social determinants of health that clinical staff are neither trained nor timed to handle. This shift protects the doctor-patient relationship while capturing the 25% of revenue currently at risk in value-based contracts. Success depends on treating consumer data as an operational trigger for support services, not a clinical diagnostic tool.

Dangerous Assumption

The analysis assumes that purchasing behavior is a persistent and accurate proxy for health risk. Consumer data is often noisy, outdated, or reflects household-level rather than individual-level choices. Basing clinical interventions on a patient buying high-sodium food that might actually be for a healthy family member risks misallocating resources and damaging patient trust.

Unaddressed Risks

• Regulatory Shift: Increased scrutiny of data brokers could lead to a sudden loss of access to the 500 data points CHS relies upon, rendering the predictive models obsolete. (Probability: High; Consequence: Critical).
• Algorithmic Bias: Consumer data, particularly credit scores and zip code data, often mirrors historical socioeconomic disparities. Using this data to prioritize care may unintentionally bake systemic bias into the CHS delivery model. (Probability: Medium; Consequence: High).

Unconsidered Alternative

Data Monetization Partnership: Instead of internal clinical use, CHS could partner with major insurers (Payers) to co-develop these predictive models. This would shift the financial burden of data acquisition and the liability of privacy concerns to the insurers, who already have a mandate for population health management and established actuarial frameworks for non-clinical data.

MECE Assessment

• Mutually Exclusive: The proposed options (Centralized vs. Integrated vs. Patient-Facing) represent distinct operational paths.
• Collectively Exhaustive: The analysis covers the primary avenues for data utilization: the provider, the coordinator, and the patient.

VERDICT: APPROVED FOR LEADERSHIP REVIEW