The science behind the principles.
Five pillars. A peer-reviewed clinical model - built on the standards every credible clinical psychologist trains on, and extended to cover coaching and self-guided care with the same discipline.
Five interlocking pillars. One clinical model.
Refill Health's clinical model is not a single feature. It is a system of five interlocking pillars - each backed by peer-reviewed evidence, each operationalised in daily practice.
Globally validated instruments. Applied adaptively, never at scale.
Refill Health uses only clinical instruments that have been validated in peer-reviewed research and adopted as standards of care globally - for depression, anxiety, burnout, trauma, wellbeing, and work crisis paths where it matters most.
A short assessment. An adaptive depth.
Every employee completes a small core of always-administered instruments at onboarding. Beyond the core, additional instruments are administered adaptively - selected by the platform's proprietary Issues Screen based on what the employee has actually reported feeling. A typical onboarding takes under 5 minutes. Clinical depth without survey fatigue.
Measurement-based care, as the field defines it.
MBC is a published clinical practice - not a Refill Health innovation. We follow the methodology established in clinical research: validated instruments, administered consistently, with thresholds for clinical action.
What MBC is.
In clinical research, MBC is the routine use of validated instruments to track patient outcomes, with the data systematically used to inform treatment decisions. It is endorsed by NICE, the American Psychiatric Association, and the UK IAPT programme. Meta-analyses across the literature consistently show MBC outperforms non-MBC care on both clinical outcomes and time to clinical significance.
How MBC works at Refill Health.
Two disciplines. Two global standards. One governance model.
Therapy and coaching are different professions with different evidence bases, different competency frameworks, and different regulators. We treat them as such - and we do not let either borrow the other's credibility.
We will not claim that coaching carries the same depth of clinical evidence as psychotherapy - it does not, and the literature is clear about that. What coaching does have is a substantial and growing peer-reviewed evidence base for goal attainment, performance and wellbeing outcomes, and an international credentialing body with an enforceable competency and ethics framework. We hold coaching to that standard in full - and we never allow it to be delivered as a substitute for clinical treatment.
Self-guided care.
The self-care library is built from techniques drawn from established modalities - cognitive and behavioural techniques, mindfulness-based practice, acceptance and defusion exercises, and structured breathing protocols. Each practice is designed and reviewed by our clinical team. We describe this layer as clinically grounded and clinically supervised in its design. We do not present self-guided tools as equivalent to treatment, and we do not attribute clinical outcomes to them in isolation.
Clinical risk stratification, applied to every care decision.
Every employee's clinical profile is risk-stratified before a care plan is generated. When multiple clinical signals are present - and most employees have more than one - a documented prioritisation logic determines which condition becomes primary, which clinician matches, and which crisis path applies if needed.
Without prioritisation, an employee presenting with both moderate depression and acute crisis signals could be routed to a general therapist while the crisis is missed. Without prioritisation, an employee with bipolar indicators could be matched to a non-specialist. Risk stratification is what prevents both. Every decision is recorded in an auditable reasoning trail - visible to the Care Navigator and to clinical supervision.
A clinically-validated recommendation engine.
A risk-stratified clinical profile feeds the recommendation engine, which matches the employee to a clinician, a care modality, and a self-care bundle. The engine is data-driven, clinician-supervised, and built on published methodology.
The general approach - personalised treatment matching using validated clinical instruments and machine-learning techniques - is established in clinical research. Variability is the enemy of clinical outcomes; structured matching is how we reduce it.
- A clinician match - therapist, coach, or both - selected from our network, based on clinical fit and member preference.
- A modality recommendation (CBT, ACT, DBT, ICF-aligned coaching) matched to the clinical presentation.
- A self-care bundle tagged to the clinical profile and stated goals.
- Every recommendation comes with alternates. The member chooses. The Care Navigator can override.
- Does not make clinical decisions. Every recommendation is reviewed by a Care Navigator.
- Does not override clinician judgement. Therapists and coaches can escalate or change track at any point.
- Does not run unsupervised. Outputs are audited monthly for drift, bias and clinical appropriateness.
- Does not see the employer. No individual output is ever surfaced to an organisation.
Continuous monitoring, so care doesn't wait to be asked for.
A reactive programme discovers deterioration at the next appointment, or not at all. Because instruments are administered on a fixed cadence rather than on request, the model detects change between sessions - and acts on it.
Score deterioration, missed sessions, disengagement from a care plan, and crisis-relevant responses each generate a signal. Signals route to the Care Navigator with a recommended next step. The member is contacted; they don't have to re-present.
On predictive capability - precisely stated.
Machine-learning techniques applied to routinely collected outcome data are an established and active research area, and the literature supports their use in identifying patients at risk of non-response. Our current deployment surfaces risk signals from measured data against defined clinical thresholds. Predictive modelling strengthens as longitudinal data accumulates across cohorts. We describe this as continuous risk detection today, and we will publish validation before describing it as anything more.
Every practitioner supervised. Every outcome tracked.
Refill Health is not a marketplace. Every practitioner in our network is credentialed, onboarded, supervised, and continuously developed. An employee matched with a Refill Health therapist or coach receives the same standard of supervision, training and outcome accountability as one matched with any other - because variability is how we eliminate it.
Crisis is treated as a clinical condition with a documented lifecycle.
Most platforms respond to crisis as an event - a phone call, a one-time intervention, a closed ticket. Refill Health treats it as a managed clinical condition with a defined lifecycle. The condition is not closed until clinical leadership signs off.
Refill Health uses the standard instrument adopted by the US FDA, NIMH and the WHO. It is administered by a trained clinician when crisis triggers fire. Crisis is the only clinical situation in which individual context flows beyond the clinical care relationship - and only under a documented consent framework. This is the single biggest predictor of poor crisis outcomes: the non-closeable safeguard, the lifecycle, and the clinical leadership sign-off, all exist to eliminate it.
Every claim on this page traces back to a published source.
Refill Health does not invent its clinical model. It applies what the field has already established - and it is explicit about where the evidence is strong and where it is still developing.
- Lambert (2015) - routine outcome monitoring and feedback
- Unützer et al. (2002) - collaborative care in integrated settings
- Kroenke, Spitzer & Williams (2001) - PHQ-9 validation
- Spitzer et al. (2006) - GAD-7 validation
- Posner et al. (2011) - C-SSRS validation
- Chekroud et al. (2016) - machine learning in treatment outcome prediction
- Scott & Lewis (2015) - measurement-based care in practice
- Fortney et al. (2017) - MBC implementation in mental health
- Theeboom et al. (2014) - meta-analysis of coaching outcomes in organisational settings
- Jones, Woods & Guillaume (2016) - effectiveness of workplace coaching
- Firth et al. (2017) - smartphone-delivered interventions for depression
- Linardon et al. (2019) - meta-analysis of mental health apps
What we project - and exactly what that means.
Refill Health is early in its deployment. We do not yet publish longitudinal outcome data from our own cohorts, and we will not present projections as if they were results. Here is what we are targeting, the basis for it, and how we will report the real numbers.
| Care model | Guided Outcomes Care |
| Clinical instruments | PHQ-9, GAD-7, burnout and risk instruments |
| Coaching measures | ICF-aligned goal and progress measures |
| Basis of projection | Published effect sizes from measurement-based and collaborative care research |
| Measurement cadence | Before every clinical session; mid-engagement for coaching |
| Reporting | Quarterly, cohort-level, anonymised, in Refill Insights™ |
| First measured data | After the first full quarter of deployment |
These are projections, not results. They are targets we hold ourselves to, derived from what the peer-reviewed literature reports for care delivered under this methodology. They are not outcomes we have measured in our own cohorts, and they should not be read as a guarantee of what any individual organisation will experience.
Outcomes vary with engagement, workforce profile, clinical presentation and duration of deployment. A projection is a commitment to measure, not a promise of a number.
What we commit to. Every deployment is measured from a pre-deployment baseline using the instruments named above. Results are reported quarterly, at cohort level, anonymised, whether or not they meet these targets. As cohorts mature, we will publish our measured outcomes and the methodology behind them - and this section will be replaced by real data.
FAQ
Common questions about our clinical model
Everything you need to know about our validated instruments, measurement-based care, and governance.
Our clinical model is built on five interlocking pillars: Validated Instruments, Measurement-Based Care, Dual-Discipline Standards, Clinical Decision Logic, and Clinical Governance. Each pillar is backed by peer-reviewed evidence and actively operationalized in daily practice.
The science is the standard. The measurement is the proof.
Book a clinical review with our team and we'll walk through the instruments, the decision logic, the governance model, and the full bibliography.