Detecting Treatment Non-Response Early: A Data-Driven Approach

A patient has been coming to therapy for two months. They're engaged, insightful, do their homework. Sessions feel productive. You're both doing good work, or so it seems.

Then you look at their PHQ-9 scores. Session 1: 18. Session 4: 17. Session 8: 16.

Eight sessions in, and this patient's depression has barely moved. Without the numbers, you might not have noticed for another month or two.

This scenario is far more common than most therapists realize. Research estimates that 5–10% of therapy patients actually deteriorate during treatment, and another 30–40% show no reliable improvement. The majority of these cases go undetected without systematic outcome monitoring.

What Non-Response Looks Like in the Data

Non-response isn't the same as deterioration. It's the absence of expected improvement: a flat line where you'd expect a downward trajectory. It looks like this:

• PHQ-9 scores that fluctuate within 3–4 points of baseline without a clear trend
• GAD-7 scores that temporarily dip after specific sessions but return to baseline
• CORE-OM scores that remain in the same severity band for 6+ sessions

The challenge is distinguishing genuine non-response from normal variation. Scores bounce around session to session due to life events, mood on the day of assessment, and measurement error. A single elevated score isn't cause for alarm. But a pattern of no sustained improvement over 4–6 sessions is a clinically significant signal.

Why Non-Response Goes Undetected

The "good session" bias. A productive therapy session creates the impression of progress. Insight feels like change. Emotional processing feels like improvement. And sometimes it is. But sometimes it isn't. The patient feels better for an hour and then returns to baseline. Without measurement, the in-session experience creates an overly optimistic picture.

Gradual deterioration is invisible. A patient who drops 1 point per session on the PHQ-9 is measurably worsening, but the change from session to session is imperceptible. It takes a trajectory to see the trend.

Patients want to please. Many patients report improvement to their therapist because they sense what the therapist wants to hear. This isn't manipulative. It's relational. They value the relationship and don't want to seem ungrateful or difficult. Self-report questionnaires completed privately (ideally digitally) are less susceptible to this bias.

Clinician optimism bias. Therapists are, by nature and training, hopeful. This is mostly a strength. But it can lead to overestimating progress and underestimating stagnation. Multiple studies show that therapists' predictions of their patients' outcomes are only weakly correlated with actual outcomes.

What to Do When the Data Says Treatment Isn't Working

Detecting non-response is only valuable if it changes what you do. Here's a systematic approach:

Step 1: Verify the Signal

Check that the pattern is genuine rather than a measurement artifact:

• Has the patient been completing assessments consistently? Irregular administration can create misleading patterns.
• Are there life events explaining the lack of improvement? A patient dealing with a major stressor may be holding steady rather than declining, which could actually be a good outcome.
• Is the patient completing the assessment honestly? If you suspect social desirability bias, discuss it directly.

Step 2: Revisit the Formulation

The most common reason for non-response is an incomplete or inaccurate case formulation. Ask yourself:

• Is the primary problem correctly identified? A patient treated for depression who actually has an undetected eating disorder, trauma history, or substance use problem won't improve on depression measures until the underlying issue is addressed.
• Are maintaining factors identified? Sometimes the formulation correctly identifies the problem but misses what's keeping it going: avoidance behaviors, relationship dynamics, environmental stressors, or unaddressed comorbidity.
• Is the treatment approach matched to the problem? CBT for a patient whose primary difficulty is emotional avoidance may need to shift toward acceptance-based or experiential approaches. Insight-oriented work for a patient who needs behavioral activation may not produce symptom change.

Step 3: Discuss Openly with the Patient

This is the most important step and the one most often skipped. Sharing the data with the patient opens a collaborative conversation:

"I've been looking at your scores over the past two months, and I want to be honest with you: they haven't changed much. I don't think that's because you're not trying. I think it means we need to look at whether our approach is the right fit or whether there's something we're missing."

This conversation does several things: it demonstrates your accountability, it gives the patient permission to share frustrations they may have been withholding, and it establishes a collaborative problem-solving frame.

Step 4: Adjust the Treatment

Based on the formulation review and the conversation with the patient, consider:

• Changing the therapeutic approach. If you've been doing primarily cognitive work, try behavioral activation. If you've been doing insight-oriented therapy, try more structured, skills-based work. If individual therapy is stalling, consider adding group therapy.
• Addressing barriers to engagement. Is the patient completing homework? Attending consistently? Applying skills between sessions? Non-adherence is informative, not frustrating. It tells you something about the treatment fit.
• Adding or modifying medication. If you're not a prescriber, consider recommending a medication review. If you are, consider adjustment.
• Consulting a colleague. Peer consultation is one of the most effective responses to non-response. A fresh perspective on the case often identifies blind spots.
• Referring. Sometimes the best intervention is a referral to someone with different expertise. This isn't failure. It's clinical competence.

Step 5: Set a Timeline for Reassessment

After making changes, set explicit expectations: "Let's try this new approach for four sessions and check the scores again. If we're still not seeing movement, we'll have another conversation about next steps."

This creates accountability for both therapist and patient and prevents the drift back into unmeasured treatment-as-usual.

The Expected Trajectory

Research provides useful benchmarks for expected improvement:

• Most patients who will respond to treatment show measurable improvement within the first 6–8 sessions
• The dose-response curve suggests diminishing returns after roughly 13–18 sessions for many conditions
• Early change (in the first 3–4 sessions) is a strong predictor of eventual outcome

These are averages, not rules. Complex cases, personality difficulties, and chronic conditions have longer expected trajectories. But the principle holds: if you're not seeing any signal of improvement within the first 2 months, something needs to change.