A theory-based multi-component intervention to increase reactive balance measurement by physiotherapists in three rehabilitation hospitals: an uncontrolled single group study

A quasi-experimental, multi-site, uncontrolled study that spanned a 27-month time frame was completed. The application of the Knowledge-to-Action Framework [25] as a guiding framework for this research program has been previously reported [26]. With this framework, we previously completed the identify problem and assess barriers to knowledge use phases [21, 23, 24, 27]. The present study addressed the phases of adapting to local context; selecting, tailoring and implementing interventions; monitoring use; evaluating outcomes and sustaining knowledge use. Institutional approval including research ethics approval was obtained at all sites.

The study was conducted at three urban adult rehabilitation hospitals in Ontario, Canada. All sites provide inpatient and outpatient rehabilitation care for multiple clinical populations with balance impairment, including older adults (aged 65 years and older) and people living with neurological disorders, musculoskeletal impairments, trauma, amputation, and general deconditioning.

Physiotherapists working in relevant clinical programs at the study sites who were not already involved in ongoing clinical balance research were eligible to participate in the study. There were 67 eligible physiotherapist full-time equivalent (FTE) positions at the time of the study, representing 75 individual physiotherapists (64 full-time and 11 part-time). Eligible physiotherapists were identified by physiotherapist research team members (onsite local champions- described below), and were actively recruited by a research coordinator who introduced the study at each site during a clinical meeting, then individually invited all eligible physiotherapists via email.

A 12-month intervention was administered at each site. Details of intervention development have been reported elsewhere [28] and are described briefly here. Previously-identified factors influencing reactive balance assessment informed the selection of the reactive balance measure [29] and were mapped to the Theoretical Domains Framework (TDF) [30, 31]. The intervention (Table 1) targeted eight TDF domains through eight established behavior change techniques [32] that were incorporated into three intervention components that were supported by existing behaviour change and implementation evidence: (i) onsite local champions who were physiotherapist research team members with an advanced practice or management role at each site and were identified at study conception through networking and were involved in all aspects of study planning, intervention delivery, data collection and analysis, (ii) seven group meetings that included didactic education, hands-on practice, and ongoing discussion, and (iii) health record modifications that included development of a reactive balance measure administration form customized for the study that included instructions and scoring guidance, and met institutional specifications (Additional file 1).

With regards to the intervention components, the local champions worked actively to engage participants and foster adoption of the reactive balance measure through a variety of activities as needed throughout the study [28]. These included scheduling and coordinating group meetings, communicating study goals and objectives, co-facilitating group meeting discussions to tailor activities to each site, modeling use of the reactive balance measure in their caseload and encouraging use among participants, providing ongoing mentorship and problem solving with participants between meetings, answering questions and brokering interactions between participants and the larger offsite research team, as well as co-developing and customizing the reactive measure administration forms, and assisting with study administrative coordination as needed. The group meetings included one 60-min didactic education session, one 60-min hands-on practice session, and five 60-min bi-monthly check-in discussions. The didactic education session was delivered by the principal investigator and included foundational content on reactive balance, its measurement, and the reactive balance measure used in the study. The hands-on practice session was led by two physiotherapists who were regular users of the measure and included viewing of training administration and scoring videos and test practice among participants. The bi-monthly check-in sessions included informal discussion of questions and concerns about administration and scoring issues, experiences using the test, trouble shooting, etc. Bi-monthly check-in discussions were co-facilitated by the principal investigator, local champion, physiotherapist hands-on practice trainers and the research coordinator. Additional resources, such as test scoring interpretation guidance and visual handouts to use with patients when administering the test, were developed and distributed in an iterative fashion in response to participant requests to the study team. Sample agendas for all three meeting types are included in Additional file 2.

The reactive balance measure selected was the Postural Responses section of the Balance Evaluation Systems Test [29] (Fig. 1). The Balance Evaluation Systems Test is a comprehensive balance measure with established interrater reliability [33], normative data [34], and documented use in multiple clinical populations [35‐37]. Each section of the comprehensive measure has been individually tested, and developers have indicated that sections can be used independently [33]. The Postural Responses section is a 6-item measure of reactive balance requiring no equipment. Interrater reliability for this section was found to be excellent (intraclass correlation coefficient = 0.92, 95% confidence interval = 0.85–0.97). This measure was selected because it addressed multiple identified factors influencing reactive balance measurement including availability of tools, appropriate for multiple populations, and (lack of) time [29].

Multiple data collection methods were used to address each objective and were tailored to align with site-specific conditions and data availability.

The primary outcomes were the proportion of included health records with a completed reactive balance measure pre- (objective i), during- (objective ii) and post-intervention (objective iii). Secondary outcomes included differences in reactive balance measure use during- and post-intervention, differences in reactive balance measure use across sites, differences in reactive balance knowledge and measurement confidence pre- and post-intervention, participant satisfaction with intervention content, delivery, and intervention components, and intent to continue reactive balance measurement.

Sample size decisions were based on the during-intervention outcome of proportion of health records with the reactive balance measure. Decisions were guided by a review of previous implementation research effect sizes [38]. Anticipating a 20% effect size during-intervention, reference calculations indicated that a sample size of 91 health records was required to demonstrate a 20% increase in the proportion of health records with the reactive balance measure completed during-intervention (alpha level = 0.05; 80% power) [39]. This estimate was increased to 100 for each of the three sites (i.e. total during-intervention sample size n = 300). Sample size decisions for pre- and post-intervention periods were based on feasibility of data collection considerations within the funded granting period.

Of the 300 records reviewed in the pre-intervention period, 211 included evidence of balance assessment and were eligible. Among these, a standardized reactive balance measure was not documented in any health record. Within one record (0.5%), there was a completed non-standardized reactive balance assessment, and 15 records (7%) mentioned reactive balance in the physiotherapy notes section of the health record.

Thirty-three physiotherapists enrolled in the study. Five participants (15%) withdrew during the intervention, either voluntarily (n = 2) or due to change in job status (n = 3). Twenty-eight participants completed the study (Table 2). Participants had on average 12 ± 2 years of clinical experience, and most participants (n = 19, 68%) estimated that in a typical week, at least 80% of their caseload included adults with balance impairment at risk of falls. Participants worked with a diverse range of clinical populations, including neurological (n = 11, 39%), orthopedic (n = 8, 29%), geriatric (n = 3, 11%), and those with multiple or complex conditions in need of general reconditioning (n = 5, 18%). Most participants (n = 20, 71%) worked in inpatient settings.

All three local champions at each site were actively involved throughout the intervention. All champions were in attendance and co-facilitated all seven meetings at each site and responded to all participant inquiries and brokered questions to the research team. Local champion contributions were fluid and responsive to participant needs. All seven group meetings were held as planned at each site (total n = 21 group meeting sessions). Average attendance at the group meetings was six out of seven meetings. Seven participants (25%) attended all meetings. Five participants (18%) attended four or fewer meetings. Attendance details are reported in Table 2. The reactive balance measure administration form was co-developed and revised over multiple drafts and included content and layout feedback from participants. Health record modification processes were tailored to each site and practice setting as needed. Reactive balance measure administration forms were stored in the additional notes sections of the health records.

Initial published intervention components [28] included using regular health record audits as an additional feedback strategy to participants. However, this component was not delivered due to feasibility challenges extracting participant-linked data in real-time during the intervention. There were also unanticipated challenges and delays in linking the reactive balance measure administration form to the health record and participant caseload at the start of the intervention for all sites. Participants were encouraged to use the measure during this period, however the period of health record eligibility was reduced to the final 8–9 months of the 12-month intervention.

During-intervention, the reactive balance measure was documented in 31% of eligible health records (n = 94). Post-intervention, the reactive balance measure was documented in 19% of eligible health records (n = 18). The proportion of eligible health records with a completed reactive balance measure was significantly lower post-intervention compared to during-intervention (X² = 4.34, p = 0.037). In both during- and post-intervention periods, there was a significant difference in the proportion of eligible health records with a completed reactive balance measure across sites that ranged from 7 to 52% of records during-intervention and 3–35% of records post-intervention (all p < = 0.012, Fig. 2). Health record characteristics are reported in Additional file 4.

Reactive balance knowledge significantly improved across all three sites (pre: 65.3 ± 13.5%; post: 74.8 ± 12.4%; time main effect (F (1, 25) = 9.1, p < 0.01). Self-reported confidence for reactive balance measurement also significantly improved across all three sites (pre: 54.3 ± 20.9%; post: 76.2 ± 14.9%; time main effect (F (1, 25) = 20.1, p < 0.001). There were no significant interactions or differences in reactive balance knowledge or confidence for reactive balance measurement across sites (all p > 0.05).

Most participants reported positive satisfaction with the intervention content (n = 20, 71%) and delivery (n = 19, 68%). With respect to utility of intervention components, the hands-on practice meeting received the most positive ratings (n = 27, 96%). Eighty-two percent (n = 23) of participants positively rated the utility of the initial didactic education meeting, 71% (n = 20) positively rated utility of the health record modifications, 61% (n = 17) positively rated the utility of the bi-monthly check-in discussion meetings, and 46% (n = 13) positively rated the utility of local champions. Forty three percent of participants (n = 12) indicated that they intended to use the reactive balance measure on completion of the study. There were no significant differences in any ratings across sites (all p > 0.05).

The study’s quasi-experimental design limits the extent of causational conclusions that can be drawn due to persisting threats to internal validity. Future studies should explore alternative study designs (e.g. stepped wedge [44]) to address this limitation. The inability to directly compare during- and post-intervention reactive measure use with pre-intervention use due to differences in data sampling is also a limitation. The decision to conduct bi-monthly check-in discussion meetings was not based on published evidence and warrants continued research. The knowledge questionnaire did not undergo a full psychometric evaluation and needs to be validated and further developed to fully understand the clinical relevance of the changes in scores. Recent advancements in standardized approaches for measuring clinical practice and professional development (e.g. [45]) could strengthen this component of data collection. The post-intervention data collection was smaller than the during-intervention and the reduced precision of estimating documented use at that stage is a limitation. Finally, it is acknowledged that the intervention protocol was modified by not delivering health record audit and feedback as planned. While such audit and feedback techniques have demonstrated effectiveness [46], the TDF domain targeted through the monitoring technique (skills) was addressed by other techniques used in the intervention (e.g., rehearsal of relevant skills, problem solving, decision making, goal setting).