Building Alliance With the Hardest Patients: What Actually Works in Severe Mental Illness
- Systematic review of 48 studies (2000–2025) on therapeutic alliance in severe mental illness — the most comprehensive synthesis on alliance in SMI to date
- Six factor domains identified: client factors, clinician factors, clinical factors, social factors, care factors, and other factors — alliance is not just about technique
- Key asymmetry: clinical symptom severity affects clinician-rated alliance but NOT client-rated alliance — patients see the relationship differently than their clinicians do
- Three strongest predictors of good alliance: client illness insight, client secure attachment style, and early positive interactions — two are modifiable
Therapeutic alliance predicts outcome across virtually every treatment modality. But the SMI literature has always been treated as a special case — these patients are "hard to engage," the alliance is "fragile." This systematic review from the University of Groningen and Lentis Psychiatric Institute does something overdue: it synthesizes 48 studies to identify what actually predicts alliance quality in this population, separating myths from mechanisms.
The rating asymmetry
The most clinically striking finding: symptom severity affects how clinicians rate the alliance, but not how patients rate it. Clinicians see a floridly psychotic patient and rate the alliance as poor. The patient, from their side, may experience the relationship as supportive regardless. This asymmetry has direct implications: if we make treatment decisions based on clinician-rated alliance alone, we may terminate or reduce engagement with patients who actually feel connected.
What builds alliance in SMI
Three factors emerged most consistently. Illness insight — patients who understand their condition form stronger alliances. Secure attachment — patients with a secure attachment style navigate the therapeutic relationship more flexibly. And early positive interactions — the first sessions predict the trajectory.
Two of these three are modifiable. Illness insight can be developed through psychoeducation and motivational interviewing — not by lecturing, but by building shared understanding. Early interactions can be shaped: warm, non-judgmental first sessions, shared agenda-setting, explicit discussion of what therapy will involve. Attachment style is harder to change, but understanding a patient's attachment pattern helps the clinician adapt their stance.
The supportive technique signal
The review found that supportive techniques — providing feedback, validating experience, shared decision-making — predicted better client-rated alliance. Directive or symptom-focused techniques did not harm the alliance, but they did not build it either. The practical implication: in SMI, how you are is at least as important as what you do.
For your practice
Three actionable takeaways. First, assess alliance from the patient's perspective, not just your own — your rating may be contaminated by symptom severity. Use brief alliance measures (WAI-SR, Session Rating Scale) regularly. Second, invest disproportionately in the first 2–3 sessions. Early positive connection is the strongest modifiable predictor. Third, lead with supportive techniques before symptom-focused work. The alliance built through warmth and validation creates the container for the challenging interventions that follow.
Clinicians rate the alliance as poor when symptoms are severe. Patients do not. We may be withdrawing from people who feel connected to us.
Heterogeneous populations under the SMI umbrella (schizophrenia, bipolar, severe depression). Correlational designs predominate — causal direction uncertain. Many studies use clinician-rated or observer-rated alliance only.