| Methods |
Random allocation by means of a table of random numbers |
| Participants |
Patients between the ages of 18 and 65 who were admitted to hospital with acute psychosis over 14 months. DSM III‐R diagnoses of subjects included schizophrenia, severe affective disorders, schizophreniform, schizoaffective disorder, delusional disorders, and psychotic disorder not otherwise classified. Non‐English speakers and subjects with low IQ scores, deafness, or organic brain disease were excluded |
| Interventions |
Control group treatment consisted of 4 to 6 supportive counseling sessions with the same therapist. Therapists listened to patients' concerns but when medication issues were broached, patients were directed to discuss such issues with their treatment teams. Experimental intervention treatment consisted of 4 to 6 sessions of "compliance therapy" ‐ a strategy that borrows from motivational interviewing. During session 1 and session 2, patients reviewed their illness and conceptualized the problem. In the next 2 sessions, patients focused on symptoms and the side effects of treatment. In the last 2 sessions, the stigma of drug treatment was addressed |
| Outcomes |
Adherence scores were measured using a 7‐point scale (1 = complete refusal to 7 = active participation and ready acceptance of regimen)
Clinical outcome measures included ratings on a brief psychiatric rating scale, global functioning assessment, schedule for assessment of insight, drug attitudes inventory, attitude to medication questionnaire, Simpson‐Angus Scale for extrapyramidal side effects. Measures were obtained in‐hospital preintervention and postintervention. Following discharge, measurements were made at 3, 6, 12, and 18 months |
| Notes |
Initial compliance was rated by the patient's primary nurse. Follow‐up compliance ratings were obtained using the 7‐point scale, based on corroboration from as many sources as possible (mean number of sources was approximately 2) |
| Risk of bias |
| Bias |
Authors' judgement |
Support for judgement |
| Random sequence generation (selection bias) |
Unclear risk |
Random sequence generation is not explained. (pg 414) "Patients were randomly assigned to the intervention consisting of 4‐6 sessions (mean = 4.7) of compliance therapy, lasting 20‐60 minutes" |
| Allocation concealment (selection bias) |
Unclear risk |
No mention of method of allocation concealment. (pg 414) "Patients were randomly assigned to the intervention consisting of 4‐6 sessions (mean = 4.7) of compliance therapy, lasting 20‐60 minutes, approximately twice weekly, or the control treatment consisting of an equal number" |
| Selective reporting (reporting bias) |
Unclear risk |
No protocol available |
| Other bias |
High risk |
The authors note the following: (pg 417) "A weakness of the study is the drop‐out rate over follow‐up; the statistical method used to deal with the subsequent missing values has already been discussed. A high rate of patient refusal and drop‐out would not be unexpected with this relatively unselected and highly disturbed patient sample, of which 59.5% were admitted involuntarily. In fact, the actual initial refusal rate was I5%, and subsequent attrition by one‐third included only 11 overt refusers. The therapy sessions were very well accepted by patients, with all those recruited completing the course of therapy. Substantial follow‐up drop‐out rates have been re‐ ported in: (a) drug studies; (b) prospective studies of inner‐city samples where at least 10% relocate beyond catchment area boundaries each year; and (c) studies of psychosocial interventions in similar populations (Tarrier et al, 1993). Longer follow‐ up periods inevitably lead to greater attrition. A practical issue is that of tracking subjects for follow‐up in a highly mobile inner‐city sample. Another methodological issue concerns the measure of compliance. This was an indirect composite measure based on information from a number of sources, which correlated strongly at each time point with measures of treatment attitudes, lending it concurrent validity. A mean advantage for the compliance therapy group of between one and two points on the observer‐rated compliance scale is a clinically meaningful difference (for example it could amount to the difference between a patient with a bipolar disorder accepting lithium with some supervision rather than being reluctantly maintained on a less appropriate depot medication schedule). However, it may be argued that lack of a direct compliance measure is a weakness in this study, though all these measures have their drawbacks. Pill counts have been widely used in the past even though despite the well‐recognised potential for inaccuracy (Rudd, 1976; Kane, 1983; Pullar et a[, 1989). Furthermore, they are time‐consuming, demanding of manpower, and ultimately not particularly suitable for: (a) in‐patient studies where medication is largely administered by staff; (b) studies involving patients on depot medication; (c) patients with individually tailored regimes. Serum assays are expensive, invasive and are not available for the full range of antipsychotics; we believe their use in this study would have further adversely affected the drop‐out rate. Urine tests for a drug or its metabolite may overestimate compliance with drugs like antipsychotics with a long half‐life (Churchill, 1985); also currently available assays may not detect some antipsychotics at the lower dose range. Both serum and urine assays are of limited value in assessing partial compliance (Babiker, 1986). While simple self report compliance measures are poorly predictive of behavior, Hogan et a1 (1983) found that DAI ratings could accurately assign 89% of a large out‐patient sample to compliant and non‐compliant groupings. Observer bias Observer bias was another potential problem. The initial ratings and most of the three‐month follow‐up ratings of functioning and compliance were made by the research psychiatrist (R.K.) who was not blind to treatment status, though the compliance ratings were based on information from impartial sources including community psychiatric nurses and out‐patient doctors on the clinical teams. However, all six‐month, 12‐month and 18‐month ratings on the other hand were carried out by two researchers trained in their use, who were blind to treatment status. Inspection of the results shows consistency between ratings over the serial points, and satisfactory inter‐rater reliability coefficients were obtained. Despite these shortcomings, the findings suggest that compliance therapy does have a measurable and positive effect on treatment adherence." |
| Blinding of outcome assessment (detection bias)
Adherence measure |
Unclear risk |
(PRIMARY) NURSE ASSESSMENT ‐ 7 POINT SCALE ‐ (pg 414) "Compliance. Initial compliance was rated blind to intervention status by the patients' primary nurse according to a seven‐point rating scale." However, the following is also indicated in the article: "The initial ratings and most of the three‐month follow‐up ratings of functioning and compliance were made by the research psychiatrist (R.K.) who was not blind to treatment status" (pg 418) |
| Blinding of outcome assessment (detection bias)
Patient outcome |
Low risk |
(PRIMARY) GLOBAL ASSESSMENT OF FUNCTIONING SCALE ‐ All ratings except for the initial and 3‐month rating were carried out by blinded researchers. (pg 415): "All six‐month, 12‐month and 18‐month ratings on the other hand were carried out by two researchers trained in their use, who were blind to treatment status." |
| Blinding of participants (performance bias)
Adherence measure |
High risk |
(PRIMARY) NURSE ASSESSMENT ‐ 7‐POINT SCALE ‐ Likely to be unblinded because of the nature of the intervention |
| Blinding of participants (performance bias)
Patient outcome |
High risk |
(PRIMARY) GLOBAL ASSESSMENT OF FUNCTIONING SCALE ‐ Patients likely to be unblinded due to the nature of the intervention |
| Blinding of personnel (performance bias)
Adherence measure |
Unclear risk |
(PRIMARY) NURSE ASSESSMENT ‐ 7‐POINT SCALE ‐ Insufficient information and a bit unclear. "The initial ratings and most of the three‐month follow‐up ratings of functioning and compliance were made by the research psychiatrist (R.K.) who was not blind to treatment status" (pg 418) |
| Blinding of personnel (performance bias)
Patient outcome |
Unclear risk |
(PRIMARY) GLOBAL ASSESSMENT OF FUNCTIONING SCALE ‐ There is insufficient information available regarding other personnel, if any |
| Incomplete outcome data (attrition bias)
Adherence measure |
High risk |
(PRIMARY) NURSE ASSESSMENT ‐ 7‐POINT SCALE ‐ As noted in limitations, dropouts lead to high risk of bias |
| Incomplete outcome data (attrition bias)
Patient outcome |
High risk |
(PRIMARY) GLOBAL ASSESSMENT OF FUNCTIONING SCALE ‐ High loss to follow‐up during the study |
