Worked example: MHSDS Expert Agent conversation

Demonstration only. Synthetic aggregate data. Not an official Dorset HealthCare report or live Trust agent.

What this example shows

This is an illustrative worked example — not a verbatim chat transcript. It demonstrates how the MHSDS Expert Agent would trace a local figure to source, separate data-quality from operational hypotheses, and refuse to approve publication without human sign-off.

This example uses synthetic demonstration data only. It is a different strand from the public RDY briefs — there is no June 2026 improvement pass or automated validation script for this SME demo.

Evidence chain

Agent rule

• MHSDS Expert Agent rule

Evidence data

• Synthetic aggregate data — pre-aggregated demo figures cited in the dialogue
• MHSDS source pack register — approved national source references

Definitions

• Synthetic local dictionary — local metric lineage and calculation logic

Related public brief (separate strand)

• Public MHSDS access profile — uses real public MHSDS aggregates for RDY; not connected to the synthetic SME demo above

Checks and scripts

No automated validation script applies to this synthetic SME strand — figures are illustrative only.

Turn 1 — Why has this gone up?

User:

I'm drafting the March directorate pack. Synthetic All Age MH access contacts have jumped from 1,240 to 1,610. Why has this gone up?

MHSDS Expert Agent:

The first thing I would check is whether this is a real activity increase or a recording / extraction change.

What this figure is based on locally

In the synthetic local dictionary (docs/synthetic-mhsds-local-dictionary.md), Synthetic All Age Mental Health Access Contacts is a local demo metric based on attended mental health contacts recorded in RiO for selected all-age mental health teams. It is not an official Trust definition or a direct MHSDS return output.

Synthetic lineage (demonstration only)

RiO contact / appointment record (DemoMHContactForm)
  → StgMHSDSContactExtractDemo
  → FactMHSDSContactDemo
  → join DimMHSDSServiceTeamDemo (AllAgeMHFlag = 1)
  → count where ContactStatusCode = 'ATT' and ContactMonth = reporting month
  → aggregate in synthetic_mhsds_sme_demo.csv

Key fields on the synthetic RiO form: ContactStatusCode, ContactDate, AppointmentType, TeamCode.

What the data shows

Movement: +370 contacts, approximately +29.8% month-on-month.

March also includes 180 late-recorded contacts in the synthetic data (LateRecordedFlag = 1), so I would not assume this is purely a demand increase.

Data-quality issues that could cause the figure to rise

1. RiO contact form or appointment recording process changed
2. Teams using DemoMHContactForm differently this month (e.g. recording contacts that were previously missed)
3. New team added to DimMHSDSServiceTeamDemo with AllAgeMHFlag = 1
4. Duplicate contacts appearing in FactMHSDSContactDemo (dedup rule not applied consistently)
5. Attended / DNA / cancelled status mapping changed in ContactStatusCode
6. Late-recorded contacts counted in March rather than the month they occurred
7. Extract logic change in StgMHSDSContactExtractDemo (nightly load, filter or join change)

Operational changes that could cause the figure to rise

• Backlog clearance clinic or additional capacity in March (noted in demo Crisis Resolution scenario)
• Higher referral demand (referrals rose ~7.1% in synthetic data — smaller than the contact increase)
• Staffing or session template changes increasing available appointments

Checks I would run before using this in a pack

1. Has the RiO contact form or appointment recording process changed?
2. Are teams using the form differently this month?
3. Has a new team been added to the All Age MH mapping in DimMHSDSServiceTeamDemo?
4. Are duplicate contacts appearing in FactMHSDSContactDemo?
5. Has attended / DNA / cancelled status mapping changed?
6. Are late-recorded contacts being counted in March rather than the month they occurred?
7. Was there a backlog clearance clinic or additional capacity in March?

National MHSDS context (background only)

MHSDS is a person-based secondary uses data set for mental health services (mhsds_standards_directory). This local demo metric is inspired by contact-count reporting concepts but is not validated against a specific national published measure. I am not quoting ETOS derivations for this count.

Draft directorate pack wording (pending human review)

All Age MH access contacts rose from 1,240 to 1,610 (+29.8%) in March 2026. This may reflect increased activity and/or recording effects, including 180 late-recorded contacts in the synthetic data. The metric counts attended RiO contacts for selected all-age MH teams per local warehouse logic. Definition and data quality checks are in progress — confirm with Information Lead before finalising.

Turn 2 — Where does the figure come from?

User:

Where does this figure actually come from — is it straight from an MHSDS return?

MHSDS Expert Agent:

No — this is not a figure pulled directly from an MHSDS submission. It is a local warehouse metric built from RiO contact records through the synthetic demo ETL chain.

Source to aggregate — step by step

Included teams (synthetic mapping)

• Demo Community MH Access Team
• Demo All Age Crisis Resolution Team
• Demo Primary Care Liaison MH Service

Relationship to MHSDS

National MHSDS (mhsds_data_model_v60) includes person-level care contact tables that Trusts ultimately submit from PAS/EHR sources. This demo metric uses similar concepts (attended contacts by team and month) but has not been mapped to a specific MHSDS derivation or published measure code. Mapping validation would be an Information Lead task on real data.

Turn 3 — How is it calculated?

User:

Walk me through how Synthetic All Age Mental Health Access Contacts is calculated.

MHSDS Expert Agent:

Calculation logic (synthetic demo)

From the local dictionary pseudologic:

SELECT COUNT(*)
FROM FactMHSDSContactDemo f
JOIN DimMHSDSServiceTeamDemo d ON f.TeamCode = d.TeamCode
WHERE f.ContactStatusCode = 'ATT'
  AND f.ContactMonth = @ReportingMonth
  AND d.AllAgeMHFlag = 1

Filters applied

March 2026 breakdown (synthetic)

February combined total: 1,240. Movement: +370 (~29.8%).

Late-recorded contacts

180 contacts in March have LateRecordedFlag = 1 — the contact occurred in an earlier month but was recorded or extracted into March. These inflate the March count unless you understand and caveat them.

Dedup note

If the ROW_NUMBER() dedup in FactMHSDSContactDemo changed between February and March (e.g. ETL release), the count could move without any change in clinical activity. Worth checking with the BI team.

Turn 4 — Does this mean performance has deteriorated?

User:

Does this increase mean performance has deteriorated?

MHSDS Expert Agent:

Not necessarily — and I cannot give a performance verdict. An increase in attended contacts can mean different things depending on whether the driver is data quality or operations.

If the rise is data-quality driven

• Form usage or recording practice changed → more contacts captured, not more care delivered
• Status coding changed → contacts reclassified from DNA/cancelled to attended
• Team mapping expanded → more teams now counted in All Age MH scope
• Late recording or extract timing → contacts attributed to March that belong elsewhere
• Duplicate rows in FactMHSDSContactDemo → inflated count

If the rise is operationally driven

• More people seen (which could indicate improved access)
• Backlog clearance or additional clinic capacity
• Higher demand following referral increase (~7.1% in synthetic data)

What the data shows (observation only)

Contacts rose ~29.8% while referrals rose ~7.1%. The contact increase is disproportionate to referral growth, which points toward recording/extraction effects or a step-change in conversion from referral to attended contact — but I cannot determine which without the checks from Turn 1.

Questions for service leads

• Did March include a backlog clearance initiative or extra clinic sessions?
• Have any teams changed how they record contacts in RiO?
• Is the increase concentrated in one team (Community MH contributed 720 of 1,610)?

I can help you frame these as investigation questions for the directorate pack — I cannot conclude whether performance has improved or deteriorated.

Turn 5 — Can I use this in the pack?

User:

Can I use this figure in the March directorate report?

MHSDS Expert Agent:

Not yet as a final figure. It can appear as draft wording with caveats only until the checks below are complete and a human owner signs off.

Minimum checks still outstanding

Data quality (Information Lead):

• Confirm DimMHSDSServiceTeamDemo mapping has not changed between February and March
• Validate dedup logic in FactMHSDSContactDemo is consistent across both months
• Review the 180 late-recorded contacts — decide whether to caveat or re-attribute
• Confirm ContactStatusCode mapping has not changed

Operational context (Performance Manager):

• Confirm with service leads whether backlog clearance or capacity changes explain part of the movement
• Agree proportionate caveats for the directorate narrative

Sign-off required

Information Lead (MHSDS accountability) and Performance Manager must review and sign off before this figure is treated as final in the directorate pack. I cannot sign off, submit an MHSDS return, or treat this demo as operationally validated against live Trust systems.

Suggested status label for the pack

Draft — pending Information Lead and Performance Manager review.

Closing summary