Effect of joint crisis plans on use of compulsory treatment in psychiatry: single blind randomised controlled trial

Setting

We recruited patients in 2000 and 2001 from seven community mental health teams in south London and one in Kent. Catchment areas ranged from near the national average to relatively socially deprived. The proportion of black and ethnic minorities was between 10.5% and 37.6%. Each team provided mental health care to adults with severe mental illness for their catchment areas during office hours, with differing local arrangements for out-of-hours crisis response.

Participants

A member of the community mental health team (either the consultant, the team secretary, or a community psychiatric nurse) provided a list of patients. All possible patients were on these lists. To be eligible, patients had to be in contact with their local community mental health team; have been admitted to a psychiatric inpatient service at least once in the previous two years; and have a diagnosis of psychotic illness or bipolar affective disorder without psychotic symptoms (according to the operational criteria checklist7). We excluded those unable to give informed consent because of mental incapacity or insufficient command of English. Current inpatients were not recruited to avoid any coercion to participate. No other exclusions were made.

Recruitment of participants and data collection

Care staff arranged meetings for recruitment of participants and baseline interviews and were present at the meeting if they and the patient wished. All patients were informed about both arms of the trial and about the randomisation procedure. All participants gave written informed consent. One investigator (CH) collected follow up data and was blinded to treatment group. The nature of the interventions meant that neither participants nor staff could be blinded to allocation.

Randomisation—Randomisation took place immediately after recruitment. The allocation sequence was generated by using minimisation, stratified by team and by severity of the patients' condition (necessitating standard versus enhanced level of care as defined by the CPA (care programme approach), to ensure even distributions of these features, which we expected would influence the production of the plan and its use by staff. Predictability of allocation by the minimisation process was occasionally a problem (when batches of similar patients were forwarded for allocation). To avoid this, ML reassigned the allocation of one patient, chosen at random, within each batch, before we reverted to minimisation. When a patient was recruited, the project worker requested allocation by email, which was returned by a statistician as intervention or control group. Allocation was not revealed to the investigator.

Intervention group—At the first meeting the project worker (CF) explained the procedure to the patient and, if possible, the care coordinator. To finalise each plan, the patient was encouraged to bring a carer, friend, or advocate to a second meeting. This meeting was to discuss the views of patients and professionals on what to do in a crisis and to negotiate agreed solutions. The selection of information to include and the exact wording were the patient's choice alone. Full details of how plans were produced are given in reports of our pilot study.1 2

Control group—Patients in the control group received information leaflets about local services, mental illness and treatments, the Mental Health Act, local provider organisations, and relevant policies. In accordance with standard practice in England, all patients should have received written copies of their care plan, within the care programme approach.8

Baseline and outcome measures—We collected data on sociodemographic variables, clinical details, history of adverse events—for instance, self harm and harm to others, and compliance with mental health treatment, rated by the care coordinator on a 7 point rating scale adapted for use with outpatients.9 Our primary outcomes were admission to hospital and length of time spent in hospital. Our secondary outcome was objective coercion—that is, compulsory treatment under the Mental Health Act 1983. We collected data on outcomes from case notes, the computerised patient administration system, Mental Health Act Office data, and interviews with patients and care coordinators. Follow up was conducted 15 months after randomisation.

Sample size

We calculated the required sample size from the pilot study using the proportion of patients admitted to psychiatric hospital at least once as the primary outcome, 1. The proportion of the pilot sample admitted during one year fell from 70%, averaged over the previous two years, to 45% during the year after the intervention. From these figures we calculated that at a significance level of P = 0.05 (double sided) and power of 80% required a sample size of 68 per group. We increased this to 76 (total 152) to allow for loss to follow up, estimated at 10% based on the rate of 7% at 6-12 months for the pilot study.

Statistical analyses

We compared age, sex, and length of contact with the service between those recruited and not recruited and baseline variables between the groups to establish if there were major differences. Patients discharged or transferred from the service were retained in their allocated group for intention to treat analyses. We used χ² tests, t tests, and non-parametric tests (Mann-Whitney) to compare the groups, the latter for bed days and number of days on a section (that is, compulsory detention for assessment or treatment under the Mental Health Act), both of which were skewed. We used Fisher's exact test to compare proportions of adverse events. Risk ratios (binary variables) and mean differences (continuous variables) and their 95% confidence intervals were estimated, in the case of continuous variables, by bootstrapping 3000 replicates (bias corrected method). Stata version 8 was used for the analysis (StataCorp, College Station, TX). There were no interim analyses. All significance tests were double sided.

For use of the Mental Health Act, we compared numbers of patients subjected to any section and the number of days spent on a section per patient. We performed sensitivity analyses in which we repeated the analysis adjusting for any variables that were expected to be associated with the outcome variable—for instance, ethnicity, male sex, younger age, and compliance.

Participants

We assessed 466 sets of case notes for eligibility using OPCRIT 5 (figure).7 There were no significant differences for age, sex, or length of service contact between eligible patients who were or were not recruited. Information on hospital admission was available for all participants. Bed days were available for all except one known admission. Information on use of the Mental Health Act was available for 77/80 of each group (total 154/160 = 96%) (figure). Of the 31 participants known to have been compulsorily admitted or treated, we knew the duration for each patient. One participant in the intervention group died from a longstanding cardiac condition. Table 1 shows other adverse events during follow up. Table 2 shows that the sociodemographic and clinical features of the two groups were similar.

Trial profile: participant flow and follow up

• Download figure
• Open in new tab
• Download powerpoint

Hospital admissions

Table 3 shows that a smaller proportion of the intervention group were admitted (30% ν 44%, risk ratio 0.69, 95% confidence interval 0.45 to 1.04, χ² = 3.25, P = 0.07). There was no significant difference in mean bed days (difference 4, −18 to 26, P = 0.15, for the whole sample; difference −24, −72 to 24, P = 0.39, for those admitted). Overall about a quarter of patients were admitted for more than one month (23% in the intervention group and 29% in the control group).

Use of the Mental Health Act

Compulsory admission and treatment were significantly less common in the intervention group (13% ν 27%, risk ratio 0.48, 0.24 to 0.95, χ² = 4.84, P = 0.03, table 3). Sensitivity analyses did not alter this conclusion. The mean number of days of detention for the intervention group was 14 compared with 31 for the control group (difference 17, 0 to 36, P = 0.04). For those admitted on a section, the mean number of days on a section was similar in the two groups (difference 3, −61 to 67, P = 0.98).

The pattern of types of section was similar in both groups (table 4), but those in the intervention group were less likely to be put on a section (table 4) and more often had their sections terminated early (table 5).

Methodological considerations

This study has several important limitations. The rate of hospital admission among the control group was lower than expected from our pilot study, which reduced the power of the study to detect a difference in this outcome and resulted in wide confidence intervals for the mean differences in bed days, consistent with either an increase or a decrease in length of hospital stay. Only 36% of eligible patients agreed to participate, so the results may not be widely generalisable. Those who declined to participate when interviewed reported that the plan would not help them, they were unlikely to become ill again, that a plan was already in place, or that no one would take any notice of it. On the other hand, generalisability was strengthened by the various settings for recruitment (inner city, suburban, small town) and the broad ethnic representation of patients. The follow up rate for the outcomes reported was high.

Implications for services

The reduction in use of the Mental Health Act has important implications for mental health services. Although the provision of a written care plan, signed by the patient, is now required in England, the joint crisis plan is substantially different. Making a joint crisis plan is voluntary, while the standard care plan is a statutory requirement. Thus, joint crisis plans can be used only when staff and patients want to formulate and use them. Furthermore, a third party, with knowledge of severe mental health problems and who is not a team member, mediates between the parties in producing each joint crisis plan. Such facilitation requires extra resources. The joint crisis plan is therefore different from a self completed advance directive10 because it is fully agreed with staff, increasing the likelihood that it will be implemented.

Finally, the process of writing a joint crisis plan is deliberately one of negotiation. We intend to undertake further investigation in future to understand what such negotiation means for staff and patients, to explore the power relationships between staff and patients, and to investigate more fully other contextual factors which may impact on such a complex intervention.10–15 We can find no other evidence in the literature that a structured clinical intervention can significantly reduce compulsory psychiatric admission and treatment. This study suggests that the committee reviewing the Mental Health Act 1983 was correct in its assertion that advance agreements can promote more consensual and less coercive care.6