Care management outreach data trust: what population health targeting requires
Care management outreach fails when the targeting data is wrong. Roughly 30% of outreach attempts never reach the intended patient due to stale contact information, misattributed risk scores, and incomplete social determinant records. Population health targeting requires data trust infrastructure that most health plans and systems do not yet have.
A care manager calls a patient flagged as high-risk for readmission. The phone number is disconnected. The address on file belongs to a prior residence. The risk score was built on claims data that lagged 60 days behind the patient's actual clinical state. The outreach fails before it starts.
This is not an edge case. It is the default operating condition of most population health targeting programs. Care management outreach data trust is not a theoretical concept. It is the operational prerequisite that determines whether a billion-dollar investment in population health actually reaches the people it is designed to help.
The outreach failure rate nobody reports
Most health plans and health systems do not track outreach failure as a formal metric. They track enrollment, engagement, and outcomes. But the denominator problem sits upstream: how many of the people your algorithm identified were actually reachable, correctly identified, and appropriately prioritized?
Studies on Medicaid managed care populations show that between 25% and 40% of outreach attempts fail on the first contact. The reasons split roughly into three categories: wrong or outdated contact information (40% of failures), patient no longer attributed to the targeting entity (25%), and clinical data used for risk stratification that was incomplete or stale (35%).
These are not technology failures. They are care management data quality failures.
What population health targeting actually requires
Population health targeting is the process of identifying individuals within a defined population who would benefit from specific interventions, then directing resources toward those individuals. The concept sounds simple. The data infrastructure required is not.
Targeting requires at minimum five data layers working in concert: clinical data (diagnoses, labs, medications), claims data (utilization patterns, cost), demographic data (age, geography, contact information), social determinant data (housing stability, food access, transportation), and attribution data (which entity is responsible for this patient's care).
When any one of these layers contains errors, the targeting breaks. When multiple layers contain errors, which is the norm, the targeting produces lists that care managers learn to distrust. That distrust is rational. It is also the single biggest barrier to population health program effectiveness.
Key statistics
Claims data used for population health targeting lags 30 to 90 days behind real-time clinical status, meaning risk scores reflect a patient's past, not their present. Address data in health plan files carries error rates between 15% and 30%, depending on the population segment and churn rate. The CDC estimates that chronic diseases account for 90% of the $4.1 trillion in annual U.S. healthcare spending, making accurate targeting of chronic disease populations a multi-trillion-dollar data quality problem. SuperTruth's work with imaware demonstrated that standardizing 105,000 diagnostic records reduced processing time from 3 weeks to 2 hours, a 95% reduction, while identifying the patient segment driving 20% of revenue. Organizations that implement trust-scored data for outreach targeting report engagement rate improvements of 2x to 3x over unstandardized approaches.
What are the four pillars of population health?
The four pillars of population health are chronic care management, quality and safety, public health, and health policy. Each pillar generates and depends on distinct data types. Chronic care management requires longitudinal clinical records with consistent coding. Quality and safety depend on outcome measurement data that is timely and complete. Public health requires aggregated surveillance data with geographic precision. Health policy depends on cost and utilization data that can withstand actuarial scrutiny.
The common thread across all four pillars is that none of them function without data you can trust. A chronic care management program built on misattributed diagnoses will target the wrong patients. A quality improvement initiative built on incomplete outcome data will optimize for the wrong metric. Data trust is not a fifth pillar. It is the foundation all four stand on.
What are the four areas of intervention identified by the CDC to promote population health?
The CDC identifies four key areas: epidemiology and surveillance, environmental approaches, healthcare system interventions, and community programs linked to clinical services. Each area has a specific data dependency that most organizations underinvest in.
Epidemiology and surveillance require real-time or near-real-time data feeds with geographic granularity. Environmental approaches require address-level social determinant data that is current, not census-vintage. Healthcare system interventions require clinical and claims data that is reconciled across sources. Community programs require referral tracking data that closes the loop between clinical identification and community-based service delivery.
The CDC's framework is sound. The data infrastructure that most organizations bring to it is not. When address-level SDOH data relies on census tract approximations rather than verified geocoding, environmental targeting becomes a guessing exercise.
Which population health management approach identifies excessive healthcare utilization?
High-utilizer identification, sometimes called super-utilizer or rising-risk stratification, is the population health management approach that seeks to identify populations accessing the healthcare system excessively. This approach typically uses claims-based algorithms to flag patients with frequent emergency department visits, multiple inpatient admissions, or disproportionate cost relative to their diagnosed conditions.
The data trust problem with high-utilizer identification is acute. Claims data is the primary input, and claims data lags 30 to 90 days behind clinical reality. A patient who visited the ED three times last quarter may have already been connected with a primary care provider. A patient whose claims show low utilization may be avoiding care due to transportation barriers that claims data cannot capture.
High-utilizer models also suffer from attribution errors. When a patient switches plans or providers, their utilization history fragments. The new plan sees a clean slate. The old plan sees a patient who stopped utilizing. Neither view is accurate.
What are the five major strategies to promote health?
The five major strategies recognized in population health practice are building healthy public policy, creating supportive environments, strengthening community action, developing personal skills, and reorienting health services. These strategies originated from the Ottawa Charter for Health Promotion and remain foundational in global public health frameworks.
From a data trust perspective, each strategy requires different data types with different trust thresholds. Building healthy public policy requires population-level aggregate data with provenance you can defend in legislative testimony. Creating supportive environments requires environmental and SDOH data at the community and address level. Strengthening community action requires data sharing between clinical and community-based organizations, which raises consent and governance challenges that most data sharing agreements do not address. Developing personal skills requires patient-generated health data with trust thresholds appropriate for clinical use. Reorienting health services requires performance data with the recency and completeness to drive operational change.
None of these strategies work if the underlying data scores poorly on provenance, recency, or consent. The strategy is only as good as the data feeding it.
The six dimensions where outreach data breaks
Care management outreach data trust fails along predictable dimensions. Understanding these dimensions is the first step toward fixing them.
Contact accuracy. Phone numbers and addresses decay at a rate of roughly 15% to 20% per year in Medicaid populations and 8% to 12% in commercial populations. If your outreach file was last validated six months ago, a meaningful percentage of your contacts are already stale.
Risk score recency. Most risk stratification models run on quarterly or monthly claims refreshes. A patient's risk profile can change dramatically within that window. A new diagnosis, a hospitalization, a medication change. The model does not know until the next refresh.
Attribution integrity. Patient-to-provider and patient-to-plan attribution files contain systematic errors. Provider directory data carries error rates near 50%, which cascades into attribution problems. If a patient is attributed to a provider who no longer participates in the network, the care management outreach is directed to the wrong care team.
Social determinant completeness. Most health plans have SDOH data on fewer than 20% of their members. The members most likely to benefit from care management outreach are precisely the ones most likely to have incomplete SDOH records. Social risk factor screening data requires specific instrument quality to be usable.
Consent status. Outreach requires consent. Consent status changes. A patient who opted into care management six months ago may have revoked consent, changed plans, or moved to a different program. If your consent tracking is not real-time, you risk regulatory violations and patient trust erosion.
Coding consistency. Population health targeting relies on diagnosis codes to define cohorts. But ICD-10 coding accuracy varies widely across providers and settings. A patient coded with unspecified diabetes in one setting and Type 2 diabetes with complications in another may appear in two different risk strata depending on which record the algorithm sees first.
Why population health models fail without trust scoring
Population health management models, whether they follow the Triple Aim framework, the Accountable Care model, or payer-driven value-based approaches, all share a common assumption: the data used for targeting is good enough to act on.
That assumption is wrong.
The gap between the data health systems have and the data population health targeting requires is wide. Risk stratification algorithms are only as accurate as their inputs. When those inputs include stale claims, fragmented clinical records, and unverified SDOH data, the algorithm produces ranked lists that do not reflect clinical reality.
This is not a model problem. It is a data trust problem. A model can be well-designed, well-validated, and statistically sound. If it runs on data that scores 40 out of 100 on a trust index, its output will reflect that 40, not the model's theoretical capability.
SuperTruth's DTI Engine scores every record across eight dimensions: Provenance (25%), Consent (20%), Recency (15%), Quality (10%), Concordance (10%), Validation (10%), Breadth (5%), and Stability (5%). These weights were calibrated in health data because health is where a wrong record costs the most. For care management outreach, the three dimensions that matter most are Recency, Concordance, and Consent.
What trust-scored outreach targeting looks like
A trust-scored approach to population health targeting works differently from the standard approach. Instead of running a risk model on whatever data is available and generating a ranked outreach list, a trust-scored approach evaluates the data quality of each record before it enters the model.
Records that score below a defined DTI floor do not enter the targeting model at all. They are flagged for remediation: contact verification, clinical data reconciliation, consent re-confirmation, or SDOH data enrichment. Records that meet the DTI floor enter the model with their trust scores attached, so care managers can see not just who to call, but how confident the system is in the data behind that recommendation.
This changes the care manager's workflow fundamentally. Instead of working a list of 200 patients and reaching 120, they work a list of 150 patients with verified data and reach 140. The outreach volume drops. The contact rate rises. The engagement rate rises further because the patients they reach are the right patients, identified with current data, prioritized by accurate risk scores.
The care gap identification process becomes more precise. The discharge planning handoff becomes more reliable. The entire care management workflow shifts from volume-based outreach to trust-scored outreach.
The consent dimension in outreach targeting
Consent is the most overlooked dimension of care management outreach data trust. Most population health programs assume that plan enrollment constitutes consent for care management outreach. This assumption is increasingly fragile.
State Medicaid programs vary widely in their consent requirements for care management contact. Some require explicit opt-in. Some allow opt-out. Some have specific restrictions on outreach modality: you can call but not text, or you can mail but not email. TCPA regulations add another layer of complexity for phone and text outreach.
The DTI Engine weights Consent at 20% of the total trust score for good reason. A record with perfect clinical data but ambiguous consent status is not actionable for outreach purposes. It is a liability. Organizations that treat consent as a binary field (yes/no) rather than a multi-dimensional, time-variant attribute will find their outreach programs increasingly exposed to regulatory and reputational risk.
From data quality to data trust in population health
Data quality and data trust are related but not identical concepts. Data quality asks: is this record complete, accurate, and timely? Data trust asks: can I act on this record with confidence, and can I prove why?
For population health targeting, the distinction matters. A record can be high quality by traditional metrics, containing complete fields, valid codes, and recent dates, but still lack trust if its provenance is unclear, its consent status is ambiguous, or its concordance with other records about the same patient has not been verified.
Care management outreach data trust requires all eight dimensions of the DTI framework working together. Provenance tells you where the record came from and whether that source is reliable. Consent tells you whether you are permitted to act on it. Recency tells you whether it reflects the patient's current state. Quality tells you whether the record is internally consistent. Concordance tells you whether it agrees with other records about the same patient. Validation tells you whether it has been checked against external standards. Breadth tells you whether you have enough data dimensions to make a targeting decision. Stability tells you whether the record has been consistent over time or subject to frequent changes that signal data integrity issues.
No single dimension is sufficient. All eight together produce a trust score that care management teams can use to make targeting decisions they can defend.
Building the outreach trust infrastructure
The operational path from untrusted outreach data to trust-scored outreach data follows a clear sequence.
First, score the existing data. Run every record in the outreach file through a DTI assessment. This creates a baseline understanding of where trust gaps exist and which dimensions are weakest.
Second, set a DTI floor. Define the minimum trust score required for a record to enter the targeting model. This floor will vary by program type, population, and regulatory context. A Medicaid care management program with strict consent requirements will need a higher floor than a commercial wellness outreach program.
Third, remediate records below the floor. This is where the real work happens. Contact verification, clinical data reconciliation, consent re-confirmation, SDOH data enrichment. Each remediation action improves the record's DTI score.
Fourth, run the targeting model on trust-scored data. The model's output is now constrained to records that meet the trust threshold. The resulting outreach list is smaller but more accurate.
Fifth, measure outreach outcomes against trust scores. This creates a feedback loop: which trust score ranges produce the highest contact rates, engagement rates, and clinical outcomes? Over time, this data refines both the DTI floor and the targeting model.
The DTI Engine scores every record 0 to 100 across eight dimensions before your AI model sees it. If your team is building population health targeting for care management outreach and needs data you can trust before you act on it, talk to the SuperTruth commercial team. Schedule a conversation or call (215) 918-4140.
Further reading:
Jason Alan Snyder
Co-founder of SuperTruth and Artists & Robots, and an inventor on the Data Trust Index patents. Twenty-plus years building technology inside Interpublic Group. He writes here nearly every day on data trust, provenance, and what AI should be allowed to act on, and publishes essays on his Substack.
About SuperTruth · LinkedIn · Substack · jasonalansnyder.com
See it in practice
DTI scores the record, not the patient.
8 dimensions. 0 to 100. Travels with every record permanently.