Contribution
DBT says skills are learned in group and carried into daily life by homework and by phone coaching, but, by the authors’ account, no published study had related DBT homework to outcomes (p. 506), and the phone-coaching literature was mostly clinical anecdote (p. 507). This study relates both, measured over the first three months, to change in urges and behaviours by the fourth month, in a programme run in a day-hospital.
Key insight
The authors read homework and coaching as the components of DBT built for generalization, so their association with outcome is evidence for generalization as a mechanism of change. In their abstract’s words, “These generalization methods appear to be important ingredients of DBT effectiveness” (p. 504). What the study can show is narrower: patients who did more homework, or called more, also reported larger falls in some urges.
Assumptions
- The programme (pp. 508–509). DBT modified for intensive outpatient care: six one-hour groups on two days a week for 16 weeks, about 96 hours; one skills module a month; a primary clinician, a trainee, for each patient; individual therapy outside the programme and not necessarily DBT. Handouts and worksheets followed Linehan’s manual (LIT-1492). Phone coaching was available to all, with DBT’s rule of no coaching within 24 hours after self-harm or a suicidal act.
- The sample (p. 509). 88 charts, 2017–2019; 4 excluded for not finishing the cycle and 28 for fewer than 7 diary-card days in month 1 or month 4, leaving 56. 79% female, 89% White; 32 with a BPD diagnosis, 35 with major depression. No diagnostic criteria for entry.
- Measures (pp. 510–512). Homework: the share of assignments complete in attended review groups, weeks 1–12, from progress notes, with partial completion counted incomplete. Coaching: the number of skills-focused calls in weeks 1–12. Outcomes from diary cards: change in mean daily highest urge (month 4 minus month 1) and in the share of reported days with the behaviour, and whether any target behaviour occurred in month 4. Attendance as a covariate. No reliability check of the chart coding (p. 518).
- Temporal order (p. 510). Predictors from weeks 1–12 and the outcome window weeks 13–16, which the authors call cross-lagged. Month-1 urges enter the change score, and they overlap the predictor window.
Key results
- Descriptives (p. 512). Attendance 88%; homework completion 86% (median 91%; 24 of 56 completed all); coaching calls a mean 5.39 (median 3, range 0–31), with 5 patients making none. Diary data on 68% of possible days. Urges were “relatively stable” from month 1 to month 4, and behaviour frequency fell slightly.
- Urges (p. 513; Table 4). Homework correlated with change in suicide urges (r = .35) and substance-use urges (r = .36 in the text, .37 in the table), and at r = .25 (p = .06) with alcohol urges. Coaching correlated with change in self-harm urges (r = .31). The other five pairings were not significant. Attendance correlated with nothing.
- Behaviour frequency (p. 513). “Regarding frequency of engagement in target behaviours, results reflected no significant associations.”
- Any behaviour in month 4 (p. 513). 27 patients reported a target behaviour in month 4 and 27 did not. Controlling for attendance, the omnibus test differed (Wilks’ λ = 0.86, p = .03). Those with none had completed more homework (91.3% against 79.8%, p = .02) and made more calls (6.00 against 3.85), the latter not significant (p = .096).
- Coaching’s contingency (p. 512). Not measured, but reported: “patients also noted that knowing they would lose access to phone coaching was helpful in motivating them away from acting on urges for crisis behaviours”.
- Coaching outside DBT (pp. 507, 517). The authors set the result against trials in which unstructured telephone access did not help, or went with 85% more self-harm among patients with repeated prior self-harm (Evans et al. 1999), and suggest that DBT’s orientation, brevity and rules make the difference. Suggested, not tested.
Claims
| id | claim | strength | support |
|---|---|---|---|
| C1 | More homework completion goes with larger falls in suicide and substance-use urges over treatment | weak to moderate: r ≈ .35, n = 56, no correction for multiple tests | Table 4, p. 513 |
| C2 | More phone coaching goes with a larger fall in self-harm urges | weak: one of four urge correlations, r = .31 | p. 513 |
| C3 | Neither goes with change in how often target behaviours occur | as reported | p. 513 |
| C4 | Patients free of target behaviours in month 4 had completed more homework | weak to moderate: p = .02, controlling for attendance only | p. 513 |
| C5 | Homework and phone coaching are active ingredients of DBT, via generalization | weak: correlational, selected sample, no measure of skill use | the authors’ reading, pp. 504, 517 |
Concepts
- Generalization: carrying behaviours learned in treatment into the patient’s natural environment, proposed as a mechanism of change in behaviour therapies and in DBT (p. 505).
- Skills homework: practising the week’s skill with a worksheet; in DBT “delivered and reviewed in a group context following a curriculum and is not individualized” (p. 506), unlike CBT homework.
- Phone coaching: “Phone coaching is intended to provide patients in-the-moment instruction and reinforcement surrounding application of skills learned in treatment” (p. 507), with crisis reduction and repair of the therapeutic relationship as its other functions.
- The 24-hour rule: no coaching for 24 hours after self-harm or a suicidal act, so that contact does not reinforce the crisis behaviour (pp. 507, 510).
Connections
It draws on Linehan (1993; the skills manual, 2015, LIT-1492) for the model; on the CBT homework meta-analyses of Kazantzis and colleagues for the expected effect; on Neacsiu, Rizvi & Linehan (2010) for skill use as a mediator; and on Chalker et al. (2015) and Oliveira & Rizvi (2018) for phone contact in standard DBT.
Bearing on the record
- The exchange’s description. The exchange cited this study at U10,
unnamed, as one in which homework and phone coaching went with better
behavioural outcomes, practice carried into daily contexts being where
the skill is generalized; the LIT’s summary repeats the abstract. Two
corrections.
- Urges, not behaviour frequency. The associations are with change in urges; with change in how often the behaviours occurred there were none. The one behavioural result is whether any target behaviour occurred in month 4, and only homework reached significance there.
- Not all completers. The 56 are those of 88 who finished the cycle and also gave at least a week of diary data in months 1 and 4; 28 were excluded for missing diary data. The authors say this may bias the result, in either direction (pp. 518–519).
- CLAIM-443 (a learned prior over which processes govern). The claim uses DBT as its worked example of practice changing which process is recruited. This study is about practice and outcome, and it fits the claim’s premise that practice matters, weakly. It cannot bear on the claim’s two limits. Nothing measures what changed: no skill-use measure (the authors’ fourth limitation, p. 518), no measure of the cost of regulating, and no situation in which the practised response stopped being right. Homework completion may also index motivation, which the design cannot separate (“analyses could not exclude the possibility of reverse-order effects”, p. 518).
- CLAIM-349 (access, authority, action; its note on DBT skills as acquired governance operations). The note distinguishes knowing how to respond from being able to deploy the response under pressure and having practised it until deployment costs little. Phone coaching is the case where deployment is supported from outside at the moment of pressure: another person recruited “in-the-moment”. The association with falling self-harm urges is consistent with that reading and too thin to support it. The 24-hour rule is a contingency on access to the external regulator, which patients reported using to hold off acting on urges.
- CLAIM-358 and QUESTION-095 (amortized cost; caching). Nothing here measures cost. The study does not show that practised skills become cheaper to deploy.
- Filed: THEORY-922, the associations as measured.
- No instruction for machine-learning practice; nothing for the anthology.
Limitations
The authors list most of these (pp. 517–519).
- Correlational, from charts, with no reliability check; reverse causation and confounds such as motivation, personality and pre-treatment functioning cannot be excluded.
- A selected sample: completers with adequate diary data, 64% of the charts.
- “statistical adjustments for multiple comparisons were not feasible, increasing the likelihood of type I error” (p. 519). Of eight urge-change correlations with the two predictors, three were significant at p < .05.
- Low base rates. Two patients reported any suicidal behaviour in month 1 and none in month 4 (Table 2); behaviour frequencies were low and likely under-reported.
- One modified programme, mostly White and female; individual therapy outside the programme was not controlled.
- The abstract overstates. It says homework and coaching were significantly associated with reduced urges for alcohol use and with a lower likelihood of any target behaviour in month 4. The alcohol correlation was p = .06, and the coaching difference in month 4 was not significant (p. 513).
- Reporting slips. r = .36 in the text against .37 in Table 4 for substance-use urges; univariate tests reported as F(1, 54) on 54 patients with attendance as a covariate, where 51 error degrees of freedom would be expected; the homework score’s example divides by “skills-coaching groups” where homework review groups are meant (p. 510). The sign convention of the change scores (month 4 minus month 1) would make reductions negative, yet positive correlations are read as reductions; the text’s reading is taken as intended.
Open questions
- Does assigning more homework or coaching, rather than observing who does more, change urges or behaviour? A dismantling or randomized component trial would tell.
- Does skill use carry the association, as the generalization account says? Measuring use of skills between sessions as a mediator, as the authors propose, would.
- Does practice make a skill cheaper to deploy under pressure, the question CLAIM-349’s note and CLAIM-358 ask? That needs a measure of effort or latency, not only of outcome.