Care Plan Goal Drift Across Encounters
Issue: Agent Regenerates the Care Plan at Each Visit From the Current Problem List Alone, Losing Continuity With Previously Agreed Patient Goals
Frequency: Common
Symptoms
- Long-term, patient-specific goals set in an earlier visit (e.g., “avoid insulin if achievable through lifestyle changes for 6 months”) are absent from the next visit’s regenerated plan
- Each visit’s care plan reads as if generated fresh from the diagnosis list, with generic goals rather than the negotiated, patient-specific ones
- Progress toward a previously set goal (e.g., weight loss target, A1c target) is not tracked or referenced in the new plan
- Care team members receive inconsistent goal framing across visits, leading to conflicting guidance given to the patient
Root Cause Care plan generation is often implemented as a stateless per-visit task: given the current problem list and latest note, produce a plan. Without an explicit mechanism to retrieve and carry forward the previously negotiated goals and their progress status, each generation event is independent and has no memory of prior shared decision-making, so the plan silently reverts to generic, diagnosis-driven defaults rather than continuing the patient’s actual care trajectory.
Example
Scenario: Visit 1 — patient and clinician agree to a 6-month trial of lifestyle modification before considering insulin
Visit 2 (3 months later): Agent regenerates care plan from current A1c and problem list
Generated plan: "Recommend initiating insulin therapy" (generic guideline-driven default)
Missed: Prior 6-month trial agreement, 3 months of progress already made
Impact: Erodes patient trust in care continuity; may override a deliberate, agreed-upon care strategy
Key Statistics
- Loss of care plan continuity across encounters is a recurring theme in studies of fragmented, episodic clinical documentation systems
- Shared decision-making goals that are not explicitly carried forward in structured form are disproportionately likely to be lost or contradicted in subsequent encounters
- Structured, longitudinal goal-tracking fields in care plans have been associated with improved goal-concordant care continuity compared to per-visit plan regeneration
Mitigation Strategies
- Persistent Goal Object: Store patient-specific care goals as a structured, persistent object (goal, agreed timeframe, progress checkpoints) separate from and carried into every visit’s plan generation, not re-derived from the problem list each time
- Progress-Aware Regeneration: Require the agent to explicitly state progress against each active goal before proposing any plan changes
- Guideline-Override Justification: When a generated recommendation would override a previously agreed patient-specific goal, require explicit justification and patient re-discussion rather than silent replacement
- Goal Continuity Audit: Periodically audit care plans for goals that disappeared between visits without documented resolution
Metrics
- % of active patient-specific goals correctly carried forward into the next visit’s plan
- Rate of generated plans that silently override a prior agreed goal without justification
- Goal-concordance rate measured via patient/clinician review
Alerts
- Active patient-specific goal absent from new plan with no documented resolution → P2
- Generated plan contradicts a previously agreed goal without explicit override justification → P1