Study design and sample size

This study was designed as a before/after study in line with the recommendations by the Medical Research Council on trial design for complex interventions on a single site

(Campbell et al., 2000, Council, 2008). The study was divided into two consecutive phases, the ‘before’ group named phase 1 (observational controls) and the ‘after’ group, phase 2 (implementation of bundle).

3.4.5.1 Admission and recruitment process

All adults admitted to AETC in both phases with suspected ABM were screened and included using identical inclusion criteria to ensure the participants in both phases were

117 matched. To minimise confounding, active recruitment, prospective data collection and follow up of all study subjects were identical in both phases.

3.4.5.2 Out of hours recruitment

One main difference in recruitment between the two phases was present. Funding for an additional study nurse became available at the start of Phase 2b, facilitating 24 hour recruitment between Sunday 5pm and Saturday 8am. Phase 1 recruitment was between Monday to Friday 8am -5pm. Patients admitted out of hours in Phase 1 were recruited from the wards following notification from the admitting medical team and the laboratory, acute deaths out of hours on the wards and in the AETC in patients meeting the CSF inclusion criteria were noted.

3.4.5.3 Recruitment start and end for each phase

The study timing was planned to ensure data collection was running at peak capacity during the cold dry season when it has been noted that more meningitis cases may present to the hospital (Gordon et al., 2000). Patient recruitment to phase 2b started at the same time point as phase 1, but one year later to minimise the effect of seasonal variation in the presentation of ABM. Phase 1 ran between the 2nd _{of January 2012 and data collection ended on the 31}st of October 2012. Phase 2a ran between November and December 2012, and Phase 2b ran between the 2nd _{of January 2013 and the 31}st _{of October 2013.}

3.4.5.4 Division of phase 2 into pilot and active phases

Phase 2 was divided into 2a and 2b. 2a was a short pilot phase that involved training the study team in the care bundle delivery, where the team were introduced to each bundle element sequentially. They underwent bedside assessments of their skill in delivering each care bundle element. Phase 2a was followed immediately by phase 2b, the active bundle. This element of the study design was chosen due to the complexity of the care bundle to

118 ensure that all study team members were familiar with each target and intervention before full active data collection started in Phase 2b.

3.4.5.5 Sample size considerations

A formal sample size calculation was not performed for this study. No formal studies on care bundles for meningitis in adults or children have been done prior to this study from which to estimate a potential effect size. Care bundle studies on adults with sepsis in high resource settings, which have been used as a model for the design of this study design do not provide directly translatable estimates of effect size that could be used for this study. A pragmatic plan to recruit equal numbers to each phase with a target of 100 patients per phase was therefore made instead. It was decided from the outset that this study would act as a

feasibility and safety pilot study to assess the care bundle, and it would be under-powered to detect any significant difference in mortality between the two phases.

3.4.5.6 Drug prescription and medical care in the AETC

Medical care in phase 1 was entirely prescribed by the AETC and medical clinical teams, and delivered by the BAM study nurses. In Phase 2, the BAM study clinician and nursing team instituted the care bundle protocol, and gave any additional treatments prescribed by the AETC or admitting medical clinical teams, such as oral fluconazole for suspected cryptococcal meningitis.

3.4.5.7 Follow up

The study team performed daily follow up of all screened patients on the medical wards from Monday-Friday until a decision for inclusion or exclusion on CSF grounds was made. Once patients were excluded, follow up ceased. Patients and guardians whose CSF results met the inclusion criteria were invited to remain in the study, and those who agreed gave written informed consent. Included patient follow up on the ward was daily for 10 days or until discharge if that was earlier. The patients were then followed weekly by phone call or text

119 message for 6 weeks and invited to a follow up appointment at 6 weeks. If they were unable to attend or unreachable by phone, community based follow up was done. This was only achieved in the second phase of the study after it became apparent that a significant number of patients were lost using telephone based follow up only in phase 1.

It was made clear to the inpatient team that the study team was not medically responsible for the study patient once they were admitted to the ward. At the day 40 outpatient appointment, further medical referrals were made by the study team as necessary, such as to antiretroviral clinic, if these had not been done at discharge by the medical team.

3.4.5.8 Trial Registration

This study was registered online with two clinical trials registries. The Pan-African Clinical Trials Registry (PACTR) number 201111000338157, and the International Standard Randomised Controlled Trials Network (ISRCTN) 96218197. The details of the study were published on both registries websites, and kept updated as the study progressed.