Valenstein 2011.

Methods	Randomized controlled trial
Participants	The study location was 4 Departments of Veterans Affairs: San Diego California; Hines, Illinois; Ann Arbor, Michigan; and Detroit, Michigan, USA 58 participants were randomized to the intervention group and 60 participants were randomized to the control group The inclusion criteria were clinical diagnoses of schizophrenia, schizoaffective, or bipolar disorder; a treatment plan that included long‐term antipsychotic treatment, antipsychotic medication possession ratios (MPRs) of < 0.8 in the prior 12 months; and no clinical contraindications to study participation The exclusion criteria were no longer active in VA care, receiving antipsychotics outside of VA system, life expectancy of less than 1 year, receiving "stronger interventions" to promote adherence (e.g. the VA adaptation of Assertive Community Treatment, staff administration of medications, depot antipsychotics, or clozapine with its attendant close supervision), medications being administered by either paid staff or a care giver
Interventions	Intervention: MEDS‐HELP INTERVENTION A composite intervention consisting of: 1) unit‐of‐use packaging that included all patients' medications for psychiatric and general medical conditions; 2) a medication and packaging education session; 3) refill reminders mailed 2 weeks before scheduled refill dates; and 4) notification of clinicians when patients failed to fill antipsychotic prescriptions within 7 to 10 days of a fill date. The medication education session was conducted by a pharmacist, usually in person, but occasionally by telephone. During this session, the pharmacist reviewed patients' prescribed medications, including treatment indications. The pharmacist also explained unit‐of‐use medication packaging and plans for interim use of pill boxes when medication changes were made by clinicians before the next shipment of medication packages Control: USUAL CARE Usual care was treatment in Veterans Affairs outpatient mental health clinics and included psychiatrist visits, non‐MD mental health visits, and group visits. During the study period, patients completed an average of 8 visits with psychiatrists; 49% had visits with non‐MD mental health clinicians and 23% had group visits
Outcomes	The measures of adherence were Medication Possession Ratio (MPR) and a Composite Adherence Measure (CAM). MPR was the primary adherence measure. It was based on pharmacy data and was calculated for 3 periods: baseline, 0 to 6 months, and 6 to 12 months. Patients were considered adherent on the CAM only if: (1) their MPR during the study time periods was < 0.8, (2) they reported they "always" took their antipsychotics or only missed antipsychotics "a couple of times" in response to questions from Schizophrenia Outcomes Module, and (3) their blood test indicated the presence of some antipsychotic medication. The patient outcomes were psychiatric symptoms as measured by the Positive and Negative Symptom Scale (PANSS), quality of life as measured by the Quality of Well‐Being Scale (QWB), and patient satisfaction as measured by Client Satisfaction Questionnaire (CSQ‐8) scale. The PANSS and QWB were completed by the research associates at each of the study sites. All these scales are widely used and have demonstrated reliability and validity. Exploratory outcomes included VA psychiatric hospitalizations, ascertained through chart reviews
Notes	―
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Central randomization. (pg.729) "Randomization was completed centrally at the coordinating site, using a blocked randomization scheme by site based on the patient's level of adherence in the prior 12 months (MPR of < 0.4, 0.4–0.59, or 0.6–0.79) and the presence of concurrent substance use."
Allocation concealment (selection bias)	Unclear risk	Insufficient information to permit judgment of 'Low risk' or 'High risk'. "In all, 118 patients were randomized to UC alone or UC plus Meds‐Help. Randomization was completed centrally at the coordinating site, using a blocked randomization scheme by site based on the patient's level of adherence in the prior 12 months (MPR of < 0.4, 0.4–0.59, or 0.6–0.79) and the presence of concurrent substance use. Site research associates (RAs) called into the central research coordinator to receive newly enrolling patients' treatment assignment." (pg 729)
Selective reporting (reporting bias)	Low risk	None detected
Other bias	Unclear risk	"Potentially, patients may have refilled their medications more regularly with Meds‐Help but failed to increase their medication ingestion. However, our a priori CAM included components related to ingestion and still showed robust increases with the intervention. In addition to patients with schizophrenia, we included patients with bipolar disorder on long‐term antipsychotics. These patients had lower levels of psychotic symptoms at baseline, which could potentially limit our ability to detect symptomatic changes. However, decreases in PANSS scores from baseline to 12 months were similar for patients with schizophrenia and bipolar disorder. We note that the increase in overall dose from improved adherence may simply have been too small to impact symptoms (25% more outpatient days with medications ''on hand''), the follow‐up period may have been too short to fully capture the benefits of increased adherence, or the relatively stable outpatients enrolling in this study may have been receiving the maximum benefit from their antipsychotic medications despite incomplete adherence. Patients with SMI may need broader‐ based interventions than those focused on medication adherence if their outcomes are to be improved. Other Limitations Measures of adherence in this study are approximations of actual adherence behaviors. A direct measure of patients' medication taking behavior would be the ultimate gold standard but is seldom achievable in adherence research.33 Although our participation rate is similar to that of most RCTs, patients recruited into a randomized trial may not be typical of all patients with SMI. More paranoid patients or those who were completely unwilling to take medications may have been underrepresented in the study sample—and this should be considered when generalizing study results to clinical populations. However, study participants' reported attitudes toward antipsychotic medications were similar to those reported for another population of patients with schizophrenia.25 Our study was not double blind, and a blinded assessment of outcomes would have been desirable. However, primary study outcomes were based on longer term patterns of medication filling which may be less susceptible to subtle biases on the part of interviewers or patients." (pg 734)
Blinding of outcome assessment (detection bias) Adherence measure	Unclear risk	(PRIMARY) MEAN POSSESSION RATE ‐ No information on whether pill count was computerized. There is insufficient information to permit judgment of 'Low risk' or 'High risk'
Blinding of outcome assessment (detection bias) Patient outcome	High risk	(PRIMARY) PSYCHIATRIC SYMPTOMS ‐ Research assistants were not blinded. (pg 730) "Research associates were also not blinded due to the costs and logistics of hiring blinded assessors for each site and the likelihood that assessors would be "unblinded" by patient comments."
Blinding of participants (performance bias) Adherence measure	Low risk	(PRIMARY) MEAN POSSESSION RATE ‐ No information on whether pill count was computerized. There is insufficient information to permit judgment of 'Low risk' or 'High risk'
Blinding of participants (performance bias) Patient outcome	High risk	(PRIMARY) PSYCHIATRIC SYMPTOMS ‐ "As with many health services interventions, patients could not be blinded to study assignment."(pg 730)
Blinding of personnel (performance bias) Adherence measure	Unclear risk	(PRIMARY) MEAN POSSESSION RATE ‐ No information on whether pill count was computerized. There is insufficient information to permit judgment of 'Low risk' or 'High risk'
Blinding of personnel (performance bias) Patient outcome	Unclear risk	(PRIMARY) PSYCHIATRIC SYMPTOMS ‐ Insufficient information to permit judgment of 'Low risk' or 'High risk'
Incomplete outcome data (attrition bias) Adherence measure	Low risk	(PRIMARY) MEAN POSSESSION RATE ‐ There was a similar rate of completion for both groups
Incomplete outcome data (attrition bias) Patient outcome	Unclear risk	(PRIMARY) PSYCHIATRIC SYMPTOMS ‐ "Unlike adherence measures, the secondary outcomes of the PANSS,QWB, and CSQ‐8 required that patients complete in‐person interviews within the appropriate time frame. These outcomes were calculated for 56 (97%) of intervention patients and 50 (83%) of UC patients at 6 months and for 49 (84%) of intervention patients and 49 (82%) of UC patients at 12 months. Patients missed in‐person interviews for a variety of reasons including the deaths noted above, no shows or interview refusals, loss to follow‐up, geographic moves, or jail time." (pg 731). At 6 months there is more missing data in the UC (83%) group than MedHelps (97%) group. It is not clear whether this difference is significant or not