Referral Criteria in Mental Health Services: Designing Safe, Defensible Access Thresholds

Referral criteria determine who gets in, who is redirected, and who waits — and those decisions carry clinical, safeguarding and reputational risk. In practice, the strongest access models treat criteria as an operational tool, not a static list: they align thresholds to capacity, define escalation triggers, and create safe alternatives for people who do not meet the service offer. This article sits within our Access, Referral & Clinical Triage content and links to wider thinking on Mental Health Service Models & Care Pathways, because criteria only work when the pathway around them is real.

Why referral criteria are a safety issue, not a paperwork issue

Referral criteria are often written as if they are simply an “eligibility filter”. In reality, they are a risk-management system. If they are too loose, teams become overwhelmed, triage becomes inconsistent, and urgent need can be missed in noise. If they are too tight, people are bounced between services, crises escalate, and the system absorbs avoidable harm.

A credible criteria set does three things at once:

• Defines the service offer (what you do, for whom, at what intensity, and for how long).
• Creates defensibility (clear reasons for accept/redirect/decline decisions, evidenced and auditable).
• Builds safe alternatives (what happens next if the person does not meet criteria today).

What good criteria contain (and what they avoid)

Core components that make criteria operational

Strong criteria are specific enough to be used consistently, but not so rigid that they ignore complexity. They typically include:

• Inclusions: presentation types the service is designed for (e.g., moderate-to-severe anxiety/depression with functional impairment; trauma presentations requiring structured intervention; relapse prevention needs).
• Exclusions: needs the service cannot safely meet (e.g., acute psychosis requiring secondary care crisis input; intoxication-driven presentations requiring substance pathways; immediate high-risk safeguarding requiring urgent statutory response).
• Priority rules: what is seen first (risk, deterioration, safeguarding, recent discharge, caring responsibilities, perinatal considerations where relevant).
• Red flags: triggers that override “routine” categorisation (suicidal intent with plan, escalating self-harm frequency, domestic abuse disclosure with immediate danger, medication destabilisation, psychotic symptoms with risk).
• Route rules: who can refer and what minimum information is required (GP, A&E liaison, social worker, self-referral; what happens if information is missing).
• “Redirect with safety” pathways: named alternative routes and how you ensure the person is not abandoned.

What weak criteria look like

Weak criteria often rely on vague descriptors (“mild/moderate/severe” with no operational meaning), or they hide a capacity problem behind “not suitable” language. They can also unintentionally discriminate if they rely on communication-heavy referral forms or assume advocacy is available. If you cannot apply the criteria consistently across staff members on a busy Monday morning, it is not operational.

Operational example 1: Minimum referral information that prevents unsafe triage

Context: A community mental health access team receives referrals with variable detail. Decisions are inconsistent, and urgent safeguarding risk is sometimes discovered only after allocation.

Support approach: The team introduces a minimum dataset for triage, applied to all referral sources. Referrals that do not meet the dataset are not “rejected”; they are placed into a rapid clarification loop.

Day-to-day delivery detail:

• Referral form requires: current risk narrative (not tick-box), recent self-harm/attempt history, safeguarding concerns, current medication, current supports, presenting problem duration, functional impact, and reason for referral now.
• A duty clinician (or senior practitioner) runs two daily “clarification slots” (e.g., 11:30 and 15:30) to call referrers for missing information.
• If a referrer cannot be reached, the case is temporarily risk-coded based on available information, and the person is offered a short safety check call if contact details are present.

How effectiveness is evidenced: Weekly audit checks the proportion of referrals meeting the dataset at first submission, time-to-triage decision, and the number of “risk upgrades” after first contact. A thematic log captures what information is most often missing and feeds back to GP practices or internal referrers.

Designing thresholds: matching need, offer, and capacity

The real challenge is not writing a long list of inclusions/exclusions; it is aligning thresholds to what your service can actually deliver. Criteria should reflect:

• Intensity and duration of your intervention (brief work vs longer-term coordination).
• Clinical leadership available (who holds risk, prescribing interfaces, supervision structure).
• Partnership routes (crisis, safeguarding, housing, substance misuse, LD/autism, domestic abuse services).
• Local demand patterns (high DNA risk, high deprivation, rural access, language needs).

When criteria are not aligned to capacity, staff compensate informally — and that creates inequality (who gets accepted becomes dependent on who triages, not the criteria).

Commissioner expectation (explicit)

Commissioner expectation: Access criteria must support equitable access and system flow. Commissioners typically expect you to show that criteria are applied consistently, that urgent needs are prioritised, and that people who do not meet criteria are redirected safely with clear handoffs. In tender or contract monitoring terms, that means you should be able to evidence decision timeliness, outcomes of redirects, and how you prevent “bounces” that drive crisis demand.

Regulator / inspector expectation (explicit)

Regulator / inspector expectation (CQC): Decisions about access and exclusion must be safe, person-centred, and well-led. Inspectors will look for evidence that risk is identified early, safeguarding concerns are escalated appropriately, and people are not left without support because they “do not fit” a pathway. They will also look for learning: how incidents, complaints or near-misses lead to criteria updates, staff briefing, and improved consistency.

Operational example 2: “Redirect with safety” so exclusions don’t become abandonment

Context: The team excludes people whose primary need is substance misuse, but many have co-occurring anxiety, trauma and safeguarding concerns. People are repeatedly re-referred, often in crisis.

Support approach: The service defines a co-occurrence rule: substance use does not automatically exclude if there is clear mental health need and the pathway can be delivered safely. Where exclusion applies, the redirect must include a documented safety action.

Day-to-day delivery detail:

• Triage template includes: “If redirected, what immediate safety action is taken today?” (e.g., call to referrer, signpost to crisis line, safeguarding referral, appointment booked with alternative service).
• A shared protocol is agreed with the local substance service: rapid appointment slots for redirected cases plus joint review option when mental health risk is present.
• Weekly “bounce list” meeting reviews repeat referrals (e.g., third referral in 12 weeks) to decide whether criteria are being misapplied or the system offer is insufficient.

How effectiveness is evidenced: Monthly dashboard tracks redirect outcomes: proportion successfully engaged with the alternative pathway, repeat referrals, and crisis presentations within 30 days of redirect. Findings are reported through governance and used to refine thresholds and partnership protocols.

Handling high-risk and urgent presentations within criteria

Criteria should never be the mechanism that delays urgent risk response. A robust model makes urgent response explicit:

• Immediate risk pathway: clear triggers for same-day contact or crisis handover.
• Safeguarding pathway: domestic abuse, exploitation, neglect, risks to children, and immediate vulnerability have named escalation routes.
• Post-discharge vulnerability: where someone has recent inpatient discharge, the pathway includes rapid stabilisation checks and liaison with discharge coordinators.

In operational terms, this often means a duty function that can act same-day, rather than placing urgent cases into routine allocation queues.

Operational example 3: Criteria review as a live governance process

Context: Staff report that criteria are “ignored” in practice because demand is high and decisions feel pressured. Complaints arise about unfairness and inconsistent decisions.

Support approach: The service introduces a criteria governance cycle so thresholds evolve with evidence rather than informal practice.

Day-to-day delivery detail:

• Quarterly criteria review chaired by the clinical lead, with triage staff, safeguarding lead, and a commissioner or system partner invite (as appropriate).
• Inputs include: serious incidents/near misses, complaints themes, equality monitoring, DNA rates, time-to-triage, and redirect outcomes.
• All changes are converted into a one-page “criteria update” briefing, built into supervision and team huddles for two weeks after release.

How effectiveness is evidenced: Pre/post review comparison: variation in triage decisions between staff (inter-rater consistency checks), reduction in repeat referrals due to bouncing, and improved timeliness of urgent contact. Equality audit checks whether changes have improved access for people needing interpreters, advocacy, or reasonable adjustments.

Making criteria fair: inclusion, accessibility and reasonable adjustments

Criteria can unintentionally exclude people who struggle with forms, who need advocacy, or who present in non-standard ways. Practical mitigations include:

• Accessible referral routes (supported self-referral, phone triage option, interpreter booking pathways).
• Reasonable adjustment prompts in the triage template (communication needs, sensory needs, trauma triggers).
• Functional impact measures that don’t rely on polished narratives (structured questions, collateral from carers/support workers).

Fair access is not a slogan; it is a set of operational choices that can be audited.

What to measure to prove criteria are working

To evidence a safe and defensible access model, track:

• Decision timeliness: referral received to triage decision; triage decision to first contact (by priority band).
• Risk outcomes: number of urgent escalations; risk upgrades after first contact; crisis presentations within 30 days.
• Redirect outcomes: engagement with alternative pathway; repeat referrals; complaints about access decisions.
• Equity indicators: access outcomes by protected characteristics and by deprivation; interpreter/advocacy needs met.

When criteria are evidence-led, they become a stabiliser for the system: staff trust decisions, partners understand thresholds, and people experience clearer, safer routes to help.

---