From the behavioural analysis, the following behaviour change intervention functions were identified: i) training: imparting practical skills conducted by the National TB Programme (NTP); ii) modelling: providing an example for people to aspire/imitate by champions/clinical leaders; iii) persuasion: using communication to induce positive or negative feelings or stimulate action via audit & feedback by the ward clinical leaders and/or TB programme staff; iv) environmental restructuring: changing the physical or social context e.g. availability of record forms for better documentation; and v) education: increasing knowledge or understanding by the champions. From these, the following policy categories were identified: i) guidelines: ensuring availability and access to child TB diagnostic protocols by the NTP; ii) regulation: establishing principles of best practice by the NTP; and iii) communication/marketing: conducting mass media campaigns to educate the public on TB by the NTP and mass marketing to target health workers on need to scale up TB testing.
From discussions with the various child TB stakeholders, a multi-faceted intervention package composed of redesigning of training to focus on practical skills, selection of champions, better use of audit and feedback and work flow restructuring was proposed. Table 2 summarises the process that was followed in linking the proposed intervention package with theory while the logic model (figure 4) conceptualises the theory of change of how the intervention package might work.
The subsequent section looks at each component in turn, elucidating the definitions as per the Expert Recommendations for Implementing Change (ERIC) taxonomy [20], briefly reviewing available evidence for how each may impact health worker practice and how they would be delivered in our context.
Redesigning of Training
Training is defined as giving instruction and/or actual demonstration of the desired action and works to improve physical and psychological capabilities of health workers, and with time, their reflective and automatic motivation [20]. The theoretical constructs through which training work are physical skills; memory, attention & decision processes [27]. The child TB training has traditionally been didactic/classroom based, usually away from the providers’ facility (to remove interruptions from work) and NTP has trained hundreds of health workers in this way. Training is a key component of the NTP national strategic plan and receives a considerable budget every year [10]. Feedback from Kenyan health workers was that they felt they still lacked competence in specimen collection in children and how to interpret test results. There was also concern about the selection of participants for training-key frontline actors were often left out [31].
ERIC recommend that training should be made dynamic i.e. vary the information delivery methods to cater to different learning styles and work contexts, and shape the training to be interactive [20]. The evidence however shows that training on its own has modest effects on health worker performance and propose that it should be combined with other strategies like supervision and group problem solving [57]. We recommend that child TB training be made more hands-on, with skills being demonstrated and participants given opportunities to practice under supervision until competence is attained. The mode of delivery should be both to individuals and groups, preferably at their work places, initially using video demonstrations and then with actual patients. Ongoing training in the form of continuous medical education/refresher sessions can be arranged ideally led by the champions. Training can be supplemented with educational outreach visits- having a trained person meet with providers in their practice settings to educate providers about TB in children with the intent of changing their practice. Redesign and distribution of printed material like guideline booklets and posters to remind health workers of the correct steps and procedures are an additional suggested mode of delivery of training as an intervention.
Champions/Local Opinion Leaders
Champions are usually local opinion leaders, individuals perceived as credible and trustworthy and disseminate and implement best evidence, for instance through informal one-to-one teaching [59]. These are individuals who dedicate themselves to supporting, marketing, and driving through an implementation, overcoming indifference or resistance that the intervention may provoke in an organisation [20]. They provide clinical leadership, mentorship and supervision through modelling/demonstrating the correct procedures and this should impact health workers’ reflective and automatic motivation positively- important in places where leadership is largely lacking [31]. The main theoretical construct through which champions work to improve health worker practices is through social influence and reinforcement. A recent Cochrane review found that local opinion leaders alone or in combination with other interventions probably improve health workers’ compliance with evidence-based practice but effect on patient outcomes is uncertain [59]. Another review found that combining training and supervision had somewhat larger effects than use of either strategy alone [57]. We recommend selection of willing mid-level managers like paediatricians, senior medical or clinical officers and nurse managers in county hospitals to perform this champion role, together with the TB clinic teams. Our work found that paediatricians in particular are often left out of child TB trainings and policy decisions, and yet a final decision to start TB treatment in difficult to diagnose patients is often left to them [31]. The NTP now recognises them as opinion leaders and has had several sensitisation meetings to update them on the latest guidelines and engage them as partners in improving care. The champions should be supported with leadership training to enable them to perform their roles.
Audit and Feedback with Group Problem Solving
Audit and feedback involves collecting and summarising clinical performance data over a specified time period and giving it to clinicians and administrators to monitor, evaluate and modify provider behaviour [20]. We found that the NTP regularly collect data from patients started on treatment but the hospital teams were not consumers of these data. The audit is done at county level, but feedback is mainly given to the county TB co-ordinators and clinicians at the TB clinics, excluding those on the wards [31].
Audit and feedback has been widely used based on the belief that healthcare workers will be prompted to modify their practice when given feedback showing their behaviour is inconsistent with a desirable target [12]. Ivers et al showed that audit and feedback generally leads to small but potentially important improvements in professional practice [12]. The effectiveness depends on baseline performance and how the feedback is provided. We propose that feedback from national level Could be given by the TB county co-ordinators or by the champions to all the clinical teams on quality of care given to children with possible TB. This then sets the stage for local audits and group problem solving led by the champions/clinical leads or TB co-ordinators. Audit and feedback will target health workers’ psychological capability and eventually their reflective and automatic motivation. The theoretical constructs through which audit and feedback work include reinforcement and behavioural regulation.
Group problem solving has been shown to have moderate to large effects on improving health worker practices [57]. Group problem solving could work through clinician implementation team meetings. Initiating these may require some coaching and they would require protected time to reflect on the implementation effort, share lessons learned and support one another’s learning [20]. These teams should ideally bring together representation from clinicians from the TB clinic, laboratory personnel, biomedical teams and clinicians in the wards and out-patient departments, as our work showed gaps in team work leading to bottlenecks in patient flows [31]. For feedback to work well, there needs to be credible data, and this requires good documentation as the initial step. Good documentation requires environmental restructuring to ensure consistent availability of structured record forms, laboratory forms and other records.
Work flow restructuring
We observed several bottlenecks in patient flow and processes that were a hindrance to identifying potential TB patients in hospitals, as illustrated in the following vignette:
Workflow vignette
An example is given of a child with possible TB in a busy outpatient department. The patient was sent to the laboratory with a request form for investigations, as the clinician was alone with long queues and had no designated space or time to collect specimens. The laboratory technician said it was not his job to collect samples and he was also alone, so the patient was sent to the ward to request the junior doctor to assist. Unfortunately, she was new in the ward and had never done specimen collection for TB in children and was busy with other procedures for the ward admissions and could not help. After spending the whole day in and out of various departments, the child and the caregiver were sent back where they started, only to find their clinician left for the day, and a new clinician had started a shift.
ERIC describe an intervention strategy of changing physical structure and equipment. This requires one to evaluate current configurations and adapt as needed the physical structure and/or equipment e.g. changing the layout of a room, adding equipment to best accommodate the targeted innovation [20]. Reorganising patient flow and processes targets physical & social opportunity as well as reflective & automatic motivation and works through TDF constructs of reinforcement, knowledge and behavioural regulation. Work flow also encompasses social restructuring with clear definition of roles and expectations e.g. who should collect samples, where and when. We propose that work flow restructuring be done with the local clinical implementation teams, as part of earlier described group problem solving activities, where they restructure and keep adapting until they reach the best local solutions. The use of process maps such as Figure 1 can help with this. It is important to ensure holistic care of all patients, so that improved TB care for children is done in tandem with improving quality of care for all. Work flow restructuring has been shown to improve health worker practices as they are based on local problem analysis and generation of solutions. The health workers get empowered because they gain control over their own work [60].
Implementation and Evaluation
This intervention is considered complex due to the number of interacting components, number of behaviours being targeted, range of possible outcomes and the need to adapt implementation to the local settings- which has implications for evaluation, especially in assessing fidelity. Guided by the Medical Research Council Framework for designing and evaluating complex interventions [61] , we present a plan for evaluation and implementation of the intervention (see Additional File 3). We propose to select four hospitals as learning sites/case studies to test feasibility and acceptability of the intervention. The hospitals will be selected from counties that have different TB case notification rates (high vs low), in which we are able to collect reliable estimates of the outcomes of interest (see figure 4). We propose to choose hospitals from the Clinical Information Network where we started the preliminary work, as they have already shown readiness and willingness to improve care for children with TB and have reliable medical records. All the hospitals will undergo a sensitisation to the project and a process of getting champions to emerge with a strategy to further support them including leadership training. All will also receive the redesigned child TB training, followed by regular audits of performance in the care given to children with possible TB. Two hospitals will receive feedback with supervision by the hospital TB champion and the other two will receive feedback with supervision by outreach from TB programme officers. This will test feasibility of these two strategies with qualitative determination of differences in preference for supervisors.
Mechanisms for delivery of feedback i.e. how frequently, to groups or individuals, written or verbal feedback, will be allowed to adapt to each site, guided by the champions/supervisors, with each team deciding how they will go about problem solving, frequency of meetings, what goals to set for improvement etc. The data for feedback will however be standard, reporting on similar variables for the quality of care given. Workflow restructuring will be site dependent, and will evolve from the group problem-solving efforts. External support and mentorship will be available from the TB programme and the research team, who will be responsible for documenting the implementation process. The intervention will initially be delivered over six months in all the participating hospitals, to learn what aspects of the intervention work as intended, what are the resource costs, are the processes acceptable, practical while causing minimal disruption. Aspects that need refinement will go back to the development stage, and those that are effective will be adopted for implementation, learning and refining iteratively over an 18-month period.
After feasibility has been established, evaluation will be done to establish effectiveness of the intervention, understand the change process and assess cost-effectiveness. Simultaneous quasi-experimental interrupted time series studies will be conducted with data prospectively collected from medical records of all paediatric admissions in the selected hospitals. Quantitative data outcomes as outlined in the logic model (Figure 3) will include proportion of paediatric admissions including pneumonia cases with suggestive signs of TB who get correctly evaluated for TB; number of TB tests done and results; proportion of patients who get a documented differential diagnosis of TB; proportion who get started on treatment; and time spent from diagnosis to treatment. While a cluster randomised control trial would have been a more robust approach, this interrupted time series design is chosen for feasibility and to enable learning and refining of the intervention with local adaptations. Conduct of parallel studies in two sets of case study hospitals powered for effect will explore replication and provide effect estimates for interventions that share major components but differ in supervision, feedback and activities prioritised for problem solving. Consistent results will increase plausibility that effects are attributable to the intervention.
The quasi-experimental design will be strengthened by qualitative work which will explore the intervention process, the pathway to effect, validity of the pre-specified theory while describing the modifying effect of differences in context. We will collect data on the health workers’ experiences, their confidence levels, their beliefs about capabilities, decision processes etc as guided by the logic model, to assess how well the BCW intervention functions explain what works about the intervention. For process evaluation, we will document the quality of delivery of the intervention at each site and any variabilities, assess how well the champions take up their roles, frequency of feedback and group problem solving, goals set and how all these contribute to the desired outcomes of interest, and whether there are any unintended disruptions to other clinical services. We will be looking to identify how well the starting theory explains the causal mechanisms of the outcomes, and whether other contextual factors can explain variation at the case study sites.
We also propose to also carry out an economic evaluation that will be of great use to policy makers when planning for scale up. We will document the time and effort as well as material resources used to deliver the intervention, compared to status quo. We propose to use participant observation by the champions and TB programme supervisors, and non-participant observation by the research team, all of whom will be documenting their reflections in diaries. For analysis, we will use the theoretical domains framework to assess theoretical fidelity (to what extent the intervention was delivered in tandem with the intervention theory). We will also borrow from realist philosophies [62], to learn and document: “what works for whom, in what respects, in what contexts and how?” This will be important for predicting the outcomes and translating and adapting interventions for other contexts.