The effects of collaborative care versus consultation liaison for anxiety disorders and depression in Denmark: two randomised controlled trials

Background Collaborative care (CC) and consultation liaison (CL) are two conceptual models aiming to improve mental healthcare in primary care. The effects of these models have not been compared in a Danish setting. Aims To examine the effects of CC versus CL for persons with anxiety and depression in Danish general practices (trial registration: NCT03113175 and NCT03113201). Method Two randomised parallel superiority trials for anxiety disorders and depression were carried out in 2018–2019. In the CC-group, care managers collaborated with general practitioners (GPs) to provide evidence-based treatment according to structured treatment plans. They followed up and provided psychoeducation and/or cognitive–behavioural therapy. The GPs initiated pharmacological treatment if indicated, and a psychiatrist provided supervision. In the CL-group, the intervention consisted of the GP's usual treatment. However, the psychiatrist and care manager could be consulted. Primary outcomes were depression symptoms (Beck Depression Inventory-II, BDI-II) in the depression trial and anxiety symptoms (Beck Anxiety Inventory, BAI) in the anxiety trial at 6-month follow-up. Results In total, 302 participants with anxiety disorders and 389 participants with depression were included. A significant difference in BDI-II score was found in the depression trial, with larger symptom reductions in the CC-group (CC: 12.7, 95% CI 11.4–14.0; CL: 17.5, 95% CI 16.2–18.9; Cohen's d = −0.50, P ≤ 0.001). There was a significant difference in BAI in the anxiety trial (CC: 14.9, 95% CI 13.5–16.3; CL: 17.9, 95% CI 16.5–19.3; Cohen's d = −0.34, P ≤ 0.001), with larger symptom reductions in the CC-group. Conclusions Collaborative care was an effective model to improve outcomes for persons with depression and anxiety disorders.


Methods for post hoc analyses
Analyses not specified in the online protocol registration but specified in the statistical analysis plan (SAP) According to the SAP, count measures/outcomes and safety measures should be analysed using negative binomial regression with robust standard error corrections in cases of severely skewed distributions with adjustment for stratification variables and baseline values.However, the models fitted poorly and Poisson regression with non-parametric bootstrapped confidence intervals was used instead.This model was also used analysing count data on GP, psychologist and psychiatrist contacts.Subgroup analyses were carried out for stratification variables and a dichotomized SAPAS variable (screening for personality disorder).The same ANCOVA model specification as in the primary analyses was used, but with treatment effects estimated separately for each subgroup.Sensitivity analyses were carried out with additional adjustments for unbalanced baseline variables (not registerbased outcomes at baseline), observed data only, and extreme case analyses.In the two extreme case analyses, missing data in both groups were replaced with high mean values (90th percentile of the observed mean value), respectively, low mean values (10th percentile of the observed mean value) to assess the worst-case vs. bestcase impact of missingness.Effect sizes were estimated for the self-reported outcomes and not only the primary outcomes as specified.Effect sizes were estimated by dividing the mean difference by the pooled standard deviation at follow-up.

Analyses not specified in the online protocol registration or the statistical analysis plan
Analyses presenting proportions having contacts with a health care provider were analysed using chi 2 tests.Sensitivity analyses testing three different models that take the nesting of patients into account were carried out: 1) geographical area (three different) and GP as a fixed effect, 2) geographical area and GP as a random effect (mixed model with random intercept), 3) with robust standard errors corrected for clustering within geographical area and GP.In the anxiety trial, a sensitivity analysis taking the unequal distribution of psychiatric outpatient contacts at baseline into account using a bootstrapped linear model was also carried out (data not shown).

Table S1. Treatment information for participants in the collaborative care group
Abbreviations.CBT: Cognitive Behavioral Therapy.a Includes participants who had minimum the first treatment planning consultation.b Treatment length reflects the days from randomization date to registered date of treatment completed.The time from randomization to the first meeting with the care manager was on average 8.5 days.c Includes contacts with a treatment purpose: the first treatment planning consultation (11% of total contacts, assumed average length: 60 min); other consultations/sessions incl.psychoeducation/CBT/CBT booster/other with/without monitoring/reevaluation (83% of total contacts, average length registered: 55 min); patient consultations with next of kin (4% of total contacts, average length registered: 55 min); other contact incl.monitoring/reevaluation face-to-face/phone/electronically and other (2% of total contacts, average length registered: 27 min).Percentages are provided across the two trials and do not include contact with a logistical purpose, e.g., calendar scheduling.d Includes supportive sessions and no treatment.e Includes drop out before or during treatment because of the patient's wish, missed contact, difficulties finding time or other reasons.

Contact between care manager and participant
Participants who had ≥ one treatment contacts a , n (%) 193 ( 98 For GP contacts about anxiety/depression, full information was not available for all participants.Therefore, the total number of participants varied: a n=175, b n=168, c n=129, d n=120.Contacts between the GP and care manager could include face-toface contacts, phone calls, e-mail or SMS contacts.It does not include written status notifications that care managers sent to GPs after each reevaluation and at the end of treatment for collaborative care participants.Contacts between care manager and GPs could include passing on treatment suggestions from the psychiatrist.Contacts between the care manager and psychiatrist concerning consultation liaison participants have not been registered separately as for participants in the collaborative care group (Table S1) and this data is thus not shown.Sensitivity analyses are performed for the primary outcomes; BAI in the anxiety trial and BDI-II in the depression trial.Estimates are based on observed data.The proportion of variance explained (R^2) by GP clustering was 0.15 for BAI and 0.09 for BDI-II.The proportion of variance explained (R^2) by area clustering was 0.01 for both BAI and BDI-II.

Table S2a . Contacts between participants and health care providers in primary care setting during 6- month follow-up (information from registers) Depression trial
Incidence rate.IRR: Incidence rate ratio.Collaborative care is the reference group when reporting IRR.Some cells do not hold a number due to Danish protection rules concerning discretion.

trial Collaborative care Consultation liaison Number of contacts between participants and the general practitioner N IR (95% CI) N IR (95% CI) IRR (95% CI)
Incidence rate.IRR: Incidence rate ratio.Collaborative care is the reference group when reporting IRR.

Table S3 . Proportion of participants having contacts with a private psychologist, private psychiatrist or having talking therapy with a GP during 6-month follow-up Depression trial
General practitioner.Some cells do not hold a specific number due to Danish protection rules concerning discretion.

Table S4a . Sensitivity analyses Depression trial
Sensitivity analyses are performed for the primary outcomes; BAI in the anxiety trial and BDI-II in the depression trial.Estimates for best case, worst case and adjustment for unequal baseline means are based on imputed data.