Alternation as a form of allocation for quality improvement studies in primary healthcare settings: the on-off study design

Abstract

BackgroundRandomized controlled trials are considered the “gold standard” for scientific rigor in the assessment of benefits and harms of interventions in healthcare. They may not always be feasible, however, when evaluating quality improvement interventions in real-world healthcare settings. Non-randomized controlled trials (NCTs) are designed to answer questions of effectiveness of interventions in routine clinical practice to inform a decision or process. The on-off NCT design is a relatively new design where participant allocation is by alternation. In alternation, eligible patients are allocated to the intervention “on” or control “off ” groups in time series dependent sequential clusters.MethodsWe used two quality improvement studies undertaken in a Canadian primary care setting to illustrate the features of the on-off design. We also explored the perceptions and experiences of healthcare providers tasked with implementing the on-off study design.Results and discussionThe on-off design successfully allocated patients to intervention and control groups. Imbalances between baseline variables were attributed to chance, with no detectable biases. However, healthcare providers’ perspectives and experiences with the design in practice reveal some conflict. Specifically, providers described the process of allocating patients to the off group as unethical and immoral, feeling it was in direct conflict with their professional principle of providing care for all. The degree of dissatisfaction seemed exacerbated by: 1) the patient population involved (e.g., patient population viewed as high-risk (e.g., depressed or suicidal)), 2) conducting assessments without taking action (e.g., administering the PHQ-9 and not acting on the results), and 3) the (non-blinded) allocation process.ConclusionsAlternation, as in the on-off design, is a credible form of allocation. The conflict reported by healthcare providers in implementing the design, while not unique to the on-off design, may be alleviated by greater emphasis on the purpose of the research and having research assistants allocate patients and collect data instead of the healthcare providers implementing the trial. In addition, consultation with front-line staff implementing the trials with an on-off design on appropriateness to the setting (e.g., alignment with professional values and the patient population served) may be beneficial.Trial registrationHealth Eating and Active Living with Diabetes: ClinicalTrials.gov identifier: NCT00991380 Date registered: 7 October 2009.Controlled trial of a collaborative primary care team model for patients with diabetes and depression: Clintrials.gov Identifier: NCT01328639 Date registered: 30 March 2011.

Examination of the Effects of an Intervention Aiming to Link Patients Receiving Addiction Treatment With Health Care

“…The study was a nonrandomized clinical trial 31 comparing the LINKAGE intervention and usual care (UC). Intervention allocation was determined by a nonrandomized, alternating off-and-on design over 30 months (ie, 5 alternating 3-month periods for each condition).…”

Cost-Effectiveness Evaluation of Collaborative Care for Diabetes and Depression in Primary Care

“…First, the comparison groups were not randomly allocated, but rather the study used a monthly time series (onÀoff design). Such a design is a viable alternative to randomization, 36 which was not employed, given the abundance of available evidence of the efficacy of collaborative care from RCTs. 15 Second, the main effectiveness study was powered based on a change in PHQ scores; the available sample size had more-limited power to address secondary objectives where greater variability in measures were observed, such as comparison of overall health-related quality of life and healthcare costs.…”

A qualitative study examining healthcare managers and providers’ perspectives on participating in primary care implementation research

“…Indeed, we observed an imbalance in TeamCare enrollment with 95 patients assigned to the intervention group and 62 patients assigned to the control group [ 27 ]. While the groups had similar characteristics, the imbalanced numbers may reflect attempts of healthcare providers to enrol patients into the intervention group.…”

“…While the groups had similar characteristics, the imbalanced numbers may reflect attempts of healthcare providers to enrol patients into the intervention group. However, this was not the case for HEALD, where the groups were more balanced numerically [ 27 ]. As a strategy to alleviate healthcare managers and providers’ concerns with the design, TeamCare study design allowed control group patients to cross over into the intervention group once they completed the control group cycle and if they were still experiencing depressive symptoms [ 27 ].…”

“…To recruit for both interventions, healthcare providers telephoned potential participants to establish willingness to participate and confirm eligibility. Once confirmed, healthcare providers allocated eligible patients using a controlled “on-off” time-series method [ 26 , 27 ]. This ‘alternation’ form of allocation is credible for health care quality improvement studies, and resulted in no imbalances or detectable biases between the intervention and control groups [ 27 ].…”