Do psychotherapy sessions cluster at 53 minutes?

Across 652,956 recorded psychotherapy sessions, 29,169 lasted 53 minutes and 23,343 lasted 52 — 25% more sessions at 53 than at 52. The four preceding minutes increased by about 11% each. CPT 90837, which reimburses 15 to 25 percent more than 90834, begins at 53 minutes.

Why does 53 minutes matter in therapy billing?

Outpatient psychotherapy is billed in time bands. CPT 90834 covers 38 to 52 minutes. CPT 90837 begins at 53 minutes and reimburses 15 to 25 percent more, roughly $30 to $35 per session on Medicare's 2026 national rates. The 53-minute figure is the midpoint between the two codes' targets of 45 and 60 minutes, rounded up.

Does this show that therapists lengthen sessions for money?

No. It shows that the distribution of session lengths changes at a reimbursement boundary across a large population. The data cannot distinguish sessions being extended from appointments being booked in slots that clear the threshold, or from recording behavior. It identifies no clinician, and we have no billing claims data.

How do you know it isn't a measurement artifact?

Physical, occupational and speech therapists have a 53-minute boundary in the CMS 8-minute rule and use the same recording software. Their sessions increase 0.7% from minute 52 to 53, compared with 25% for psychotherapy. The result also holds under both floor and nearest-minute rounding, and 53 is not a round number, so scheduling habits at 45, 50 or 60 minutes don't produce it.

What happens when clinicians enter the duration manually?

Among 8,374 manually entered durations — 1.3% of sessions, excluded from the main analysis — 53 minutes appears 1,392 times, against 58 entries at 51 minutes and 86 at 52. The only comparable value is 60 minutes, at 1,301.

Did the data match the prediction?

Partly. We predicted both an excess above 53 minutes and a shortfall at 51 and 52, which is the pattern produced when sessions are moved across a threshold. The excess is present; the shortfall is not. Minutes 51 and 52 hold 21,288 and 23,343 sessions, close to what the preceding minutes predict. That means appointments booked in slots that clear the threshold are also consistent with the pattern.

What data is this based on?

661,330 recorded psychotherapy sessions from 4,105 clinicians between 1 February and 1 August 2026, captured by an ambient AI scribe. The primary analysis uses the 652,956 system-measured durations. Figures are aggregate counts per minute with no access to session content, and the minute-level dataset is published as a CSV under CC BY 4.0.

25% More Therapy Sessions at 53 Minutes Than 52

Key takeaways

There are 25% more psychotherapy sessions lasting 53 minutes than 52 minutes, exactly where the higher-paying CPT 90837 code begins.
All four of the largest psychotherapy professions have more sessions lasting 53 minutes than 52 minutes, with increases ranging from 18% to 44%.
Physical, occupational and speech therapy sessions increase just 0.7% from 52 to 53 minutes, compared with 25% for psychotherapy, despite using the same recording system and having their own billing boundary at 53 minutes.
When clinicians enter session length manually, they record 53 minutes 1,392 times, compared with 86 times for 52 minutes and 58 times for 51 minutes.
Psychotherapy session lengths change exactly at the 53-minute threshold for the higher-paying CPT 90837 code, but the data cannot determine whether sessions are extended, scheduled around the threshold, or recorded differently.

Why we looked at this

In an earlier analysis of session lengths across 48 clinical specialties, we measured twelve psychotherapy professions separately, expecting them to differ. Ten of the twelve had median session lengths within two minutes of 53: mental health counseling at 54.6, psychotherapy at 53.9, clinical social work at 53.8, cognitive behavioral therapy at 51.2. Different licensures, different client populations, nearly the same number.

53 minutes is also where CPT 90837 takes over from CPT 90834 and reimbursement rises 15 to 25 percent. That could be a coincidence — the therapy hour has existed far longer than the code set, and a distribution centered near an hour will have its median land somewhere in the low fifties regardless.

Medians can't separate the two. A minute‑level histogram can. We wrote down what each outcome would mean before running the query, and said we would publish whichever we got.

Shape	Interpretation	Result
Step up at 53, deficit at 51–52	Billing thresholds affect session length	Step observed, deficit not observed
Spikes at 45, 50 and 60	Scheduling convention rather than billing	Not observed in measured durations
Smooth curve through 38 and 53	The earlier finding was an artifact of using medians	Not observed

The distribution

Across 652,956 recorded psychotherapy sessions, 29,169 lasted 53 minutes and 23,343 lasted 52. Minute 54 is 1.9% above minute 53.

Histogram of psychotherapy session lengths by minute from 30 to 76 minutes across 652,956 system-measured sessions. Session counts climb steadily to minute 52, rise sharply at minute 53 where CPT 90837 begins, then flatten through minute 57 before declining.

Figure 1. Recorded psychotherapy sessions by duration, minutes 30 to 76. System‑measured durations only.

Session length	Sessions	Change from previous minute
48 minutes	15,428	+11.0%
49 minutes	17,129	+11.0%
50 minutes	19,683	+14.9%
51 minutes	21,288	+8.2%
52 minutes	23,343	+9.7%
53 minutes	29,169	+25.0%
54 minutes	29,722	+1.9%
55 minutes	30,635	+3.1%
56 minutes	30,030	−2.0%

At the average rate of increase across minutes 49 to 52, minute 53 would hold about 25,900 sessions.

Why the boundary falls at 53

53 is not a scheduling interval, which is part of what makes it useful as a test. It comes from the fee schedule: the AMA places the boundary between the two psychotherapy codes at the midpoint of their targets, 45 minutes for 90834 and 60 for 90837, giving 52.5, rounded up.

CPT code	Face-to-face time	Reimbursement
90832	16–37 minutes	Base
90834	38–52 minutes	Middle
90837	53 minutes and up	15–25% above 90834; roughly $30–35 more per session on Medicare's 2026 national rates

A 52‑minute session bills as 90834 and a 53‑minute session bills as 90837. There is no rounding at the boundary, and payers audit it.

The control cohorts

Physical, occupational and speech therapists bill under the CMS 8‑minute rule, whose unit steps fall at 23, 38 and 53 minutes. Psychiatry has a 53‑minute threshold for the 90838 psychotherapy add‑on. Both cohorts use the same recording software and have a billing boundary at the same minute.

Three line charts of sessions per minute from 46 to 60, indexed to minute 48 equals 100. Psychotherapy rises steeply and increases 25 percent at minute 53. Physical, occupational and speech therapy increases 0.7 percent. Psychiatry increases 4.3 percent.

Figure 2. Sessions per minute by cohort, indexed to minute 48 = 100.

Cohort	Sessions 48–52 min	Sessions 53–57 min	Ratio	Change 52→53
Psychotherapy	96,871	148,767	1.54×	+25.0%
Physical / occupational / speech therapy	3,745	4,112	1.10×	+0.7%
Psychiatry	5,402	5,769	1.07×	+4.3%

The comparison makes several simpler explanations less likely. The recording software is identical across cohorts. The result holds under both floor and nearest‑minute rounding. A general tendency to drift past billing thresholds would be expected to appear in physical therapy, which has the same threshold. And 53 is not a round number, so a 45‑, 50‑ or 60‑minute scheduling habit does not produce it.

Profession	Sessions 48–52 min	Sessions 53–57 min	Ratio	Change 52→53
Clinical Social Work	15,021	25,880	1.72×	+43.9%
Mental Health Counseling	35,290	57,840	1.64×	+39.3%
Marriage and Family Therapy	6,851	10,264	1.50×	+21.4%
Psychotherapy	29,186	41,962	1.44×	+18.1%

3,046 clinicians contributed sessions between 50 and 57 minutes, and the largest single contributor accounts for 0.4% of them.

The part of the prediction that failed

We predicted the excess above 53 would come with a shortfall at 51 and 52 — the pattern produced when sessions that would otherwise end below a threshold are pushed across it. The excess is there. The shortfall is not.

Minute 51 holds 21,288 sessions and minute 52 holds 23,343, both close to what the preceding minutes predict. Instead of a hole followed by a spike, the distribution steps up at 53 and stays at the higher level through 57.

The absence of a deficit below 53 means scheduling around the threshold is also consistent with the pattern: a practice that books 55‑minute appointments rather than 50‑minute ones shifts its whole distribution without extending any individual session. The data cannot distinguish between these explanations.

Manually entered durations

Twofold records session length automatically, and a clinician can overwrite it. While checking whether edited durations could produce the pattern, our engineer noticed that the edit form only accepts whole minutes and multiplies by 60 — so every edited duration is an exact multiple of 60 seconds and lands on an integer minute by construction. Edited rows can generate a spike at any round value on their own. We excluded all 8,374 of them, 1.3% of sessions, from everything above.

Bar chart of 8,374 manually entered session durations from 44 to 66 minutes. A tall bar at 53 minutes holds 1,392 entries, against 58 at minute 51 and 86 at minute 52. A second tall bar at 60 minutes holds 1,301 entries.

Figure 3. Manually entered session durations, minutes 44 to 66. Excluded from all other figures.

Among those 8,374 sessions, 53 minutes is entered 1,392 times, against 58 entries at 51 minutes and 86 at 52. The only comparable value is 60 minutes, at 1,301 entries. A typed duration doesn't measure how long a session ran, but it does record what the clinician wrote down.

What this doesn't establish

The distribution changes at a reimbursement boundary. Several mechanisms would produce that, and this dataset can't separate them.

Consistent with the data

Appointments booked in slots that clear 53 minutes

Sessions occasionally extended near the boundary

Recording started or stopped with the boundary in mind

Some combination of these across different practices

Not shown by the data

That any individual clinician did any of these

That care changed rather than documentation

That these sessions were billed as 90837 — we have no claims data

Intent

Booking a 55‑minute appointment so it comfortably clears the threshold is ordinary practice management.

The 38-minute boundary

We also predicted a smaller step at 38 minutes, the floor of 90834. Sessions increase 14.6% from 37 to 38 minutes, against roughly 8% in the surrounding minutes. It is present but much smaller than the change at 53, and it separates a smaller financial gap — the lowest code from the middle one. We would not rest a claim on it alone.

Limitations

Recording duration is not encounter duration

We measure how long the recording ran, not when the session began and ended. If clinicians start or stop recording with the threshold in mind, the same pattern appears without any change to the session. This is the largest limitation.

No claims data

We don't observe billing. Some of these sessions are cash-pay, some bill under other codes, some aren't billed at all. That should dilute rather than create the effect, since the sample includes clinicians for whom the boundary is irrelevant.

No confidence intervals, and sessions cluster within clinicians

Every figure is a point estimate. Sessions are not independent observations — 4,105 clinicians contributed them, and a clinician with a fixed schedule contributes many similar durations — so the effective sample size for inference is smaller than the session count. We have not run a clustered test, and we would want one before making a formal statistical claim about the size of the discontinuity.

One platform's users

US outpatient practices, weighted toward behavioral health and smaller practices. The control cohorts come from the same population, which is what makes the comparison useful even where absolute levels don't generalize.

Method

Parameter	Value
Window	1 February 2026 to 1 August 2026, fixed
Psychotherapy sessions	661,330 total: 652,956 system-measured, 8,374 manually entered
Clinicians	4,105
Primary analysis	System-measured durations only
Included	Live conversation recordings with successfully generated, non-deleted notes
Excluded	Dictation and typed notes; 7,231 sessions under 1 minute; 1,005 over 4 hours
Professions	The twelve psychotherapy specialties from the earlier study; all twelve resolved to specialties in the database
Control cohorts	Physical, occupational and speech therapy (31,955 sessions); psychiatry (229,038)
Privacy	Aggregate counts per minute. No visit, clinician or patient identifiers and no session content. Buckets with fewer than 5 contributing clinicians are suppressed.

The earlier study used a rolling six‑month window from its run date, which isn't reproducible. This one uses fixed dates.

The data

Minute‑by‑minute counts for every figure above — psychotherapy split by system‑measured and manually entered, both control cohorts, and the four largest professions — are published as session-length-by-minute-2026.csv under CC BY 4.0. The specialty‑level dataset from the earlier study is also available. Questions: info@trytwofold.com

About Twofold

Twofold is an ambient AI scribe for clinicians. It records the visit and drafts the clinical note, and session duration is stored as part of that workflow, which is what made this analysis possible. Audio is discarded once the note is drafted and patient data is not used to train models. More about the product.

Data and sources

Session data: Twofold operational dataset. 661,330 recorded psychotherapy sessions from 4,105 clinicians, 1 February to 1 August 2026. Primary analysis restricted to 652,956 system‑measured durations; 8,374 manually entered durations analyzed separately. Durations bounded between 1 minute and 4 hours. Counts are aggregates per minute; buckets with fewer than 5 contributing clinicians are suppressed. No session content, patient data or clinician identifiers were accessed.

Billing rules: AMA CPT time bands for outpatient psychotherapy (90832, 16–37 minutes; 90834, 38–52; 90837, 53 and up) and CMS Medicare Physician Fee Schedule national rates. The 53‑minute boundary is the midpoint of the 45‑ and 60‑minute code targets, rounded up. Control cohort thresholds follow the CMS 8‑minute rule for time‑based therapy codes, with unit steps at 8, 23, 38 and 53 minutes, and the CPT 90833, 90836 and 90838 psychotherapy add‑on codes.

FAQ

Frequently asked questions

Do psychotherapy sessions cluster at 53 minutes?

Across 652,956 recorded psychotherapy sessions, 29,169 lasted 53 minutes and 23,343 lasted 52 — 25% more sessions at 53 than at 52. The four preceding minutes increased by about 11% each. CPT 90837, which reimburses 15 to 25 percent more than 90834, begins at 53 minutes.
Why does 53 minutes matter in therapy billing?

Outpatient psychotherapy is billed in time bands. CPT 90834 covers 38 to 52 minutes. CPT 90837 begins at 53 minutes and reimburses 15 to 25 percent more, roughly $30 to $35 per session on Medicare's 2026 national rates. The 53‑minute figure is the midpoint between the two codes' targets of 45 and 60 minutes, rounded up.
Does this show that therapists lengthen sessions for money?

No. It shows that the distribution of session lengths changes at a reimbursement boundary across a large population. The data cannot distinguish sessions being extended from appointments being booked in slots that clear the threshold, or from recording behavior. It identifies no clinician, and we have no billing claims data.
How do you know it isn't a measurement artifact?

Physical, occupational and speech therapists have a 53‑minute boundary in the CMS 8‑minute rule and use the same recording software. Their sessions increase 0.7% from minute 52 to 53, compared with 25% for psychotherapy. The result also holds under both floor and nearest‑minute rounding, and 53 is not a round number, so scheduling habits at 45, 50 or 60 minutes don't produce it.
What happens when clinicians enter the duration manually?

Among 8,374 manually entered durations — 1.3% of sessions, excluded from the main analysis — 53 minutes appears 1,392 times, against 58 entries at 51 minutes and 86 at 52. The only comparable value is 60 minutes, at 1,301.
Did the data match the prediction?

Partly. We predicted both an excess above 53 minutes and a shortfall at 51 and 52, which is the pattern produced when sessions are moved across a threshold. The excess is present; the shortfall is not. Minutes 51 and 52 hold 21,288 and 23,343 sessions, close to what the preceding minutes predict. That means appointments booked in slots that clear the threshold are also consistent with the pattern.
What data is this based on?

661,330 recorded psychotherapy sessions from 4,105 clinicians between 1 February and 1 August 2026, captured by an ambient AI scribe. The primary analysis uses the 652,956 system‑measured durations. Figures are aggregate counts per minute with no access to session content, and the minute‑level dataset is published as a CSV under CC BY 4.0.

There are 25% more psychotherapy sessions at 53 minutes than at 52