The evidence is encouraging and limited. A Cochrane review that sought randomised trials comparing styles of nursing handover found none that met its criteria (Smeulers et al., 2014). A review of studies of bedside shift-to-shift handoff reported improvements in patient and nurse satisfaction and some safety indicators, in designs that were mostly before-and-after comparisons on single units (Mardis et al., 2016). Stronger evidence exists for structured handover in general: a multicentre study of a handoff programme for resident physicians found a reduction in medical errors and preventable adverse events after its introduction (Starmer et al., 2014). The structured format most widely used in nursing is SBAR (situation, background, assessment, recommendation), brought into health care as a shared mental model for communication between clinicians (Haig et al., 2006).

Patient involvement is the feature that sets bedside handover apart, and the research on it is nuanced. A mixed-methods review found that many patients value being included while preferences vary, and that both patients and nurses worry about sensitive information being overheard (Tobiano et al., 2018). Bedside handover was introduced in several of the early Australian sites as a quality improvement initiative and not as a research intervention, which is how most wards will meet it (Chaboyer et al., 2009).

Several features of a busy surgical ward in a Ghanaian public hospital shape the design. Wards are commonly open, with beds close together and curtains or screens the only partition, so that anything said at one bedside may be heard at the next. Relatives are often at the bedside and take part in care. English is the official language and the language of records, while patients may be most at ease in Akan, Ewe, Ga, Dagbani or another Ghanaian language. Nurses are responsible for many patients each; the global shortage of nurses is concentrated in low- and middle-income countries, those of the African region prominent among them (World Health Organization, 2020). Handover is traditionally led by the senior nurse of the outgoing shift, and hierarchy within the team is marked.

The legal and policy setting matters too. The Patient's Charter of the Ghana Health Service sets out the patient's rights to information about their condition and to privacy and confidentiality (Ghana Health Service, 2002), and the Data Protection Act treats information about a person's health as special personal data (Republic of Ghana, 2012). Bedside handover serves the first of those rights and can endanger the second. A framework for Ghana has to treat confidentiality as a design problem from the start.

The first phase establishes why the ward wants to change and who will lead. Kotter (1996) argues that change efforts fail when urgency is not established and no guiding coalition is formed, and the Australian experience bears this out (McMurray et al., 2010). The ward manager, two or three respected staff nurses from different shifts, a night nurse and a representative of the surgical team form a small group, with the backing of nursing administration. They write an aim that is specific and modest, for example that within a stated period every morning handover will include a bedside exchange for every patient.

Before designing anything, the team describes what happens now. Members observe several handovers on each shift and note where they take place, how long they last, what is said for each patient, what is written, how often they are interrupted and what the incoming nurses go on to discover for themselves. Staff are asked what works. The purpose is to identify the information a surgical nurse needs and the constraints, such as the time at which the night staff's transport leaves, that any new routine must respect.

The protocol should be written by the nurses who will use it. A two-part structure is proposed. The first part is a brief safety briefing at the station for the whole incoming shift, covering the ward state, patients causing concern, staffing and anything too sensitive for an open ward. The second is the bedside exchange, patient by patient, following a fixed order such as SBAR. At each bed the outgoing nurse introduces the incoming nurse, gives the reason for admission, the operation and the day since it, the current condition and the plan, and both look at the patient, the wound site, drains, infusions and charts together. The patient is asked whether anything has been missed.

Three design decisions are specific to the setting. The first concerns sensitive information. The team agrees a list of matters that are never spoken aloud at the bedside, such as a new diagnosis the patient has not yet been told, HIV status and social circumstances, and these are passed on at the station or in writing. The second concerns relatives: the patient is asked, before handover begins, whether those present may stay. The third is language. The clinical exchange between nurses may be in English, and the summary and the question to the patient are given in a language the patient understands. On a ward with many beds, time is the binding constraint, and the script should be short enough that the round fits within the overlap between shifts.

Nurses in published studies have feared that bedside handover takes longer, that they will be exposed in front of patients and that confidentiality will suffer (Tobiano et al., 2018). Preparation should address these concerns directly. Short practice sessions on the ward, in which pairs rehearse the script at an empty bed, are more useful than a lecture. Junior nurses practise presenting in front of seniors, and seniors practise inviting questions. Patients and relatives are told on admission, in plain terms and in their own language, that nurses will come to the bedside at the change of shift, why, and that they may decline to have their care discussed there.

The protocol is first tried on one shift change and a handful of beds. After each test the pair involved and the team spend a few minutes on what happened: whether the script was followed, how long each patient took, what the patients said and what was awkward. Changes are made and the next test is slightly larger. This is the plan-do-study-act cycle of the Model for Improvement (Langley et al., 2009), and its merit here is that problems of time, noise and privacy are met on a small scale, where they can be solved before the whole ward is committed.

Once the protocol runs smoothly on some beds and one shift it is extended to all. Sustaining it needs a few simple supports: a prompt card at each bed or on the handover sheet, the ward manager or a deputy joining handover regularly in the first weeks, a brief monthly audit, and inclusion of bedside handover in the orientation of new and rotating staff. Night-to-morning handover, when the outgoing staff are most tired and patients may be asleep, deserves particular attention and may justify a shorter script.