.png)
What Your SRS Scores May Already Be Telling You About Disclosure: A clinical example of Disclosure Winter in feedback data
- Cindy Hansen
- 9 minutes ago
- 10 min read
Disclosure in psychotherapy shapes what enters the clinical record — and increasingly, what enters the digital mental-health tools built on top of it. A client who does not say the difficult thing can produce a record that looks complete while missing the one piece that mattered.
In this article, I use “Disclosure Winter” to describe conditions under which a population, or a subpopulation, does not disclose important negative, difficult, or corrective information — often for good reason, given prior experience with how that information has been used — while the resulting record continues to appear complete and healthy.
Here I want to make that idea more concrete.
If Disclosure Winter names the system-level problem of what happens when difficult information stops entering the record, SRS data offers one place clinicians may already be able to look for its early traces.
The Session Rating Scale (SRS) was not designed to detect signs that clients may be hesitant to disclose negative or corrective feedback. It was designed to help clinicians monitor the working alliance. But because meaningful alliance feedback depends on whether clients feel able to say what is not working, SRS data may already contain clues about the conditions under which disclosure is possible, limited, or absent.
The SRS therefore offers a familiar place to begin looking for how Disclosure Winter might appear in clinical feedback records.
What SRS measures
The Session Rating Scale operationalizes the working alliance through four areas: Relationship, Goals and Topics, Approach or Method, and Overall. The first three correspond broadly with Bordin’s model of bond, goal consensus, and task agreement. Overall is the client’s own global judgment about the session as a whole, anchored between “There was something missing in the session today” and “Overall, today’s session was right for me.”
The measure is completed at the end of each session and scored out of 40.
The SRS does not ask:
Was there something you wanted to tell your therapist but didn’t?
So it would be wrong to treat an SRS score as a direct measure of disclosure.
But that does not mean disclosure is irrelevant to what the score tells us.
Why disclosure matters to the score
For an SRS score to meaningfully reflect the client’s experience, the client has to be willing to report that experience.
The alliance and the score are not identical. The score is a measurement generated through a social process. If the conditions required for honest reporting change, the measurement may change even if the underlying alliance remains stable.
Research supports this link between alliance and disclosure. In related clinical samples, people report withholding information for relational reasons such as lack of trust, fear of judgment, and concern about losing control. Other studies suggest that when therapist self-disclosure is experienced as helpful, it can strengthen the alliance, while stronger alliance is associated with greater patient self-disclosure and less shame.
The Agnew Relationship Measure makes this point explicit by including a subscale for perceived freedom to self-disclose. That matters because it treats freedom to disclose as a measurable feature of the therapeutic relationship, not as something that can simply be assumed.
None of this means an SRS score tells us what a particular client withheld on a particular day.
It means something narrower — and potentially more useful: the SRS measures aspects of the therapeutic relationship that are often associated with clients’ willingness to disclose.
The pattern is where the SRS becomes useful.
The score is not the signal
The alliance is not the number.
The number is a client’s representation of their experience of the alliance, produced by a person making a decision about what to communicate.
A score of 40 can therefore mean many things. It may mean the client genuinely experienced the session as excellent. It may mean there was little to criticize. It may mean the client tends to score generously. Or, in some circumstances, it may mean that giving a lower score did not feel possible, useful, or safe.
The score alone cannot tell us which. The pattern may give us more to work with.
Two SRS patterns worth looking for
If we are interested in conditions under which difficult information does not enter the record, what might that look like in SRS data?
There are at least two patterns worth examining.
1. The missing dip
A below-cutoff SRS score is not inherently a bad thing. In fact, a low score can be clinically valuable. It tells the therapist that something about the relationship, goals, topics, approach, or overall session experience was not working for the client — information the therapist can respond to.
When a client is willing to give that feedback and the score subsequently improves and holds, the pattern provides something clinically meaningful: evidence that disagreement or dissatisfaction could enter the relationship, be addressed, and then change.
The clinically meaningful signal is the trajectory: a lower score appears, the therapist responds, the score improves, and the improvement holds. That sequence requires something a perfectly flat series of high scores does not: the client had to be willing to say that something was not right.
So one question to ask of your own SRS records is: across a substantial body of cases, how often do I see clients giving lower scores early in treatment and then improving as the alliance develops?
If that pattern is unusually rare, it merits examination. Not because the absence proves nondisclosure, but because it removes one of the clearest opportunities in the data to observe disagreement, repair, and change.
2. The persistent ceiling
The second pattern is almost the mirror image.
Look at cases where SRS scores remain at or near 40 across many sessions.
Again, this is not proof of a problem. Some clinicians genuinely achieve consistently high SRS scores. The question is not whether high scores are possible, but whether unusually little variation invites further inquiry.
Across dozens of clients and hundreds of sessions, how much variation is there? How often do scores dip, and by how much? Do clients ever disagree, or say the approach was not helpful, or report that a session did not feel quite right?
If very little of that appears anywhere in a large body of data, the absence itself deserves exploration.
A near-ceiling score is not necessarily evidence of unusually strong alliance. An unusually low rate of any negative feedback may instead be a signal about the conditions under which feedback is being produced.
The SRS cannot tell us which explanation is true. But it can tell us that the question is worth asking.
A clinical example
Imagine a new clinical manager who begins examining raw SRS data rather than relying only on summary dashboards.
One therapist stands out. Almost every client, almost every session, scores in the high 30s or 40. Across dozens of cases and hundreds of sessions, there is barely a dip anywhere.
The manager then looks at the therapist’s outcome data. The annual pre-post effect size has consistently ranked near the bottom of the team.
The two datasets do not appear to make sense together. The alliance data suggests unusually strong therapeutic relationships. The outcome data suggests comparatively weak results.
The manager brings in someone experienced in Feedback-Informed Treatment to examine the pattern more closely.
The consultant notices something the dashboard did not show. There are no meaningful early-session dips followed by repair and improvement. There is almost no evidence of disagreement entering the measurement process at all. And the near-perfect scores continue across an unusually large number of clients.
The consultant does not conclude: “These clients were not disclosing.” The data cannot establish that.
Instead, the pattern raises a clinically useful question: what were the conditions under which these clients were giving their feedback?
That is exactly the kind of anomaly this concept is meant to make visible: not proof that clients withheld information, but a reason to examine whether the feedback process was capable of carrying negative or corrective information into the record.
From one therapist to a data system
Inside a clinic, a pattern like this can potentially be investigated. A manager can look at individual cases. A clinician can discuss the feedback process. A consultant can examine trajectories. Context is still available.
The significance of this distinction increases when feedback data moves beyond clinical supervision and becomes part of larger analytic systems. Most of what a client discloses in therapy does not remain contained within the therapeutic relationship. It becomes part of a record that may be stored, linked, analyzed, reused, and increasingly incorporated into digital mental-health systems and AI development.
A model does not have the room. It does not know that the therapist’s client smiled while giving a 40. It does not know that the client was afraid to disagree. It does not know that the client had previously tried to raise a concern and decided not to do it again.
It sees the data. And if a dataset contains years of near-ceiling scores produced under conditions in which clients were reluctant to provide negative feedback, the dataset does not necessarily look broken. It may look remarkably healthy.
The absence of disclosure can become indistinguishable from the absence of the underlying problem.
When silence scales
This is the mechanism Disclosure Winter is meant to name. The term is a working label, not an established clinical construct.
At the individual level, the SRS gives us one way of thinking about how this might show up. At scale, the consequences become much larger.
A population whose members are genuinely doing well may produce high scores. A population whose members are unable or unwilling to report that things are not going well may also produce high scores. The resulting datasets can look remarkably similar.
The difference exists in the human relationship that produced the data — a context that may disappear as the data moves downstream.
What you can look for in your own records
The SRS was not designed to detect whether clients are withholding difficult feedback. But that does not mean it has nothing to contribute to detecting the conditions that might produce that kind of silence.
If you have access to session-level SRS data, consider looking beyond the average score. Ask:
How much variation is there across clients?
How often do scores fall below the usual clinical cutoff?
Do lower scores occur early in treatment and then improve?
Do those improvements hold?
Are there therapists whose scores remain near the ceiling across unusually large numbers of cases?
How often do clients disagree about the approach, goals, or relationship?
Are there cases where outcome data and alliance data appear strangely disconnected?
None of these patterns proves nondisclosure. None diagnoses a therapist. None establishes that difficult information is being withheld from the record. They are questions the data can invite you to ask.
What the SRS can show us
The point is not to turn the SRS into a disclosure instrument. It is to recognize that clinical measures carry information about the conditions under which they are produced.
A score is never just a number. Someone had to decide to give it. Someone had to feel able to answer. Someone had to believe that their answer could be given without unacceptable consequences.
That human contribution is easy to overlook once the response becomes a row in a dataset. But it is precisely what makes the data clinically meaningful in the first place.
The SRS gives us a small, familiar example of a much larger problem. Before there is data, there is a person deciding what to disclose. And if that decision changes, the data changes with it — even when the data continues to look perfectly healthy.
That is why your SRS records may already be telling you something about Disclosure Winter: not by revealing exactly what was withheld, but by showing whether disagreement, repair, and corrective feedback were able to enter the record at all.
What We’ve Been Working On — August 2026
August was a busy month at Holistic Research Canada, with new research, publications and ongoing work exploring a common question: what happens to people and their information as healthcare becomes increasingly digital and AI-enabled?
Here’s a look at what we’ve been working on this month.
August 13
When health records become representations of people: Composition, inference, and why de-identification no longer settles the question
In this new paper, Cindy Hansen examines what happens when health information is linked across time and systems, combined with unstructured information, and used to generate new inferences. The paper argues that health records can become representations of the people they describe—and that this creates governance questions that cannot be resolved simply by removing identifiers.
August 16
Illustrating emerging personal health data governance challenges in AI-enabled health data ecosystems: An international scenario-based study using the Clinical Data Governance Checklist
This international scenario-based study applies the Clinical Data Governance Checklist (CDGC) to emerging AI-enabled health-data environments. The scenarios illustrate how governance challenges can arise across the data lifecycle, including collection, linkage, transformation, inference and secondary use.
The work is intended to make emerging governance challenges more tangible and provide a practical way to examine where existing approaches may no longer be sufficient.
August 21
What De-identification Misses: Trust and Secondary Use in Digital Mental Health
Published by eMHIC, this article looks beyond the question of whether information has been technically de-identified.
In digital mental health, information originates in relationships where people disclose highly personal experiences in a particular context and for a particular purpose. When that information is subsequently linked, transformed, analysed or reused, removing identifiers does not necessarily remove the relationship or the obligations associated with the original disclosure.
The article explores what this means for trust, secondary use and the governance of digital mental-health information.
August 27
The Disclosure Winter
Our latest article explores a different point in the data lifecycle: before the data exists.
What happens when people become less willing to disclose difficult or sensitive information because they no longer trust the environment in which that information will be collected, stored or reused?
The Disclosure Winter proposes this as a potential—and largely invisible—data-quality problem in mental health. A record can appear complete while important information has never entered the record at all.
The article connects the quality of AI and other data-driven systems back to something much more fundamental: the conditions under which people are willing to tell us what is actually happening.
Looking ahead
Taken together, this month’s work has continued to develop a connected line of inquiry at Holistic Research Canada: from clinical information, to data, to representations of people, and ultimately back to the human relationship from which the information began.
As AI becomes increasingly embedded in healthcare, we think these questions need to be considered before systems are designed—not after the data has already been collected and transformed.
