Mental Health Care and Outcomes

Question
How do adolescents and new mothers with depression move through care, and how can routine data measure that path and where it leads?
Status
Current research: In development
Approach
Longitudinal cohorts in EHR and claims data, care-cascade measures with explicit time windows, NLP and large language models for clinical notes
Updated

With collaborators in mental health research, I am planning studies of adolescent and maternal and postpartum depression that follow recognition, treatment start, engagement and outcomes in real-world data, and testing whether language models can recover the symptoms and outcomes that only clinical notes record.

Depression care as a set of possible pathsSix stages from symptoms to outcomes, with paths that skip diagnosis, return to recognition after relapse, or leave observed care, and the stages that structured EHR data, claims and clinical notes typically capture.relapse or return to caretreated without a recorded diagnosisLeaves observed care: no follow-up, a move, or care outside the health systemSymptomsemerge or worsenRecognitionscreening or aclinical visitDiagnosisrecorded, or notTreatment starttherapy ormedicationEngagementsustained orinterruptedOutcomessymptoms andfunctionWhat routine data usually captureStructured EHRClaimsClinical notesusually recordedpartlyrarelypaths that skip a stage, loop back, or leave careDepression care as a set of possible pathsSix stages from symptoms to outcomes, with dashed paths that skip diagnosis or return to recognition after relapse, and the stages that structured EHR data, claims and clinical notes typically capture.EHRClaimsNotesSymptomsemerge or worsenRecognitionscreening or a clinical visitDiagnosisrecorded, or notTreatment starttherapy or medicationEngagementsustained or interruptedOutcomessymptoms and functionusually recordedpartlyrarelyskipped stages and returnsAny stage can end in leaving observed care.

Conceptual illustration

Depression care is a set of paths, and each data source sees part of it

Stages and paths are simplified. Coverage marks show where each source typically holds information; they are not measured completeness.

Figure description

Six boxes run left to right: symptoms, recognition, diagnosis, treatment start, engagement and outcomes. Dashed curves show treatment that begins without a recorded diagnosis and a return from engagement to recognition after relapse. Dotted lines drop from four stages into a band labeled as leaving observed care. Below, three rows mark coverage for structured EHR data, claims and clinical notes: notes usually record symptoms and recognition, which structured data and claims rarely or only partly capture.

Work in this direction

  1. In development

    Care pathways in adolescent and maternal depression

    Current research

    Study planning with collaborators. We are defining measures of recognition, treatment initiation, engagement and longitudinal outcomes for adolescents and for the perinatal and postpartum period.

  2. Exploratory

    Symptoms and outcomes from clinical notes

    Proposed research

    Testing whether NLP methods and large language models can extract depression symptoms, severity and outcome information that structured fields do not hold, and how to validate that extraction against manual chart review.

  3. Proposal in development

    A larger study of depression care pathways

    Proposed research

    Developing a proposal for a multi-year study of how care pathways relate to outcomes. Its scope will be described here once it has been submitted.

Why this question

Depression often begins in adolescence, and pregnancy and the first year after birth are another period of high risk. For both groups the distance between needing care and receiving effective care can open at several points: symptoms that are never recognized, a diagnosis that does not lead to treatment, treatment that stops after a visit or two. A quality measure that counts one event, such as a completed screening or a first prescription, cannot show where that distance opens or whether it closes.

Electronic health records and claims follow people across visits and years, which makes it possible to describe care as a sequence rather than a snapshot. They also have blind spots. Symptoms and their severity are usually written in notes rather than coded. Care received outside a health system is invisible to it. A person who stops coming back looks, in the data, the same as a person who recovered.

Approach

  • Measures that respect order and time. Recognition, treatment start and engagement are defined with explicit windows anchored to a stated index event, so the same cascade can be computed again in another dataset or next year.
  • Paths, not a single funnel. The analysis allows for skipped steps, re-entry after relapse and loss to follow-up, and reports them rather than forcing every patient onto one line.
  • Information from notes. NLP methods and large language models are candidates for recovering symptoms, severity and outcomes from clinical text. No extracted variable enters an analysis before it has been checked against manual chart review.
  • Observability first. Every measure states which data source can observe it and what a missing value means.

Earlier work this builds on

Earlier projects measured two aspects of pediatric mental health care quality: whether children in Florida Medicaid received first-line psychosocial care before starting an antipsychotic, and how body mass index changed among children taking second-generation antipsychotics, where clinical information added what claims alone could not show. Work on clinical-note de-identification and on real-world cancer endpoints showed how much any measure depends on what the record actually contains.

Next steps

  1. Settle cohort and measure definitions with collaborators, written so another team could reproduce them.
  2. Build and validate note-based extraction on a manually reviewed sample before scaling it.
  3. Report what the cascades show, including where the data cannot answer the question.

Published work from my record in the areas this direction draws on. None of it reports results of the work above.