One-way SMS and healthcare outcomes in Africa: Systematic review of randomised trials with meta-analysis

Background The impact of one-way SMS on health outcomes in Africa is unclear. We aimed to conduct a systematic review of one-way SMS randomised trials in Africa and a meta-analysis of their effect on healthcare appointments attendance and medicine adherence. Methods PubMed, Embase, CENTRAL, The Global Health Library, ClinicalTrials.gov, ICTRP, and PACTR were searched for published and unpublished trials in Africa without language restriction (up to April 2018). Trials reporting effect estimates on healthcare appointment attendance and medicine adherence were assessed for risk of bias and included in meta-analyses using random-effects models. Other outcomes were reported descriptively. The protocol is registered in PROSPERO, ID:CRD42018081062. Results We included 38 one-way SMS trials conducted in Africa within a broad range of clinical conditions. Eighteen trials were included in the meta-analyses, and four were assessed as overall low risk of bias. One-way SMS improved appointment attendance, OR:2·03; 95% CI:1·40–2·95 (12 trials, 6448 participants), but not medicine adherence, RR:1·10; 95% CI:0·98–1·23 (nine trials, 4213 participants). Subgroup analyses showed that one-way SMS had the highest impact on childhood immunization attendance, OR:3·69; 95% CI:1·67–8·13 (three trials, 1943 participants). There was no clear evidence of one-way SMS improving facility delivery, knowledge level (reproductive/antenatal health, hypertension), diabetes- and hypertension management. Conclusion In an African setting, the clinical effect of one-way SMS is uncertain except for appointment attendance where the effect seems to vary depending on which clinical condition it is used in.


Review question
The aim of this review is to landscape the number of randomised controlled trials in Africa involving SMS interventions where the intervention is compared to standard care. Further, an analysis is conducted of how effective SMS interventions are at improving attendance to health appointments and medicine adherence are compared to standard care.

Problem/Population: Africa
Intervention: SMS Comparison: Standard Care Outcome: Attendance to health care appointments / Adherence to medicine.

Searches
The primary author developed the specific search strategies in co-operation with an experienced clinical research librarian. Firstly, an electronic search for headings and sub-headings was conducted so that the search string would include non-indexed studies. Literature searches will be updated before final analysis to retrieve recently published studies. Types of study to be included We will include randomized controlled trials (RCTs), in any language, including cluster RCTs and pilot Page: 1 / 5 S5 File: Protocol PROSPERO International prospective register of systematic reviews studies. All other study-design will be excluded.

Condition or domain being studied
The effect of one-way SMS interventions on health appointment attendance and medicine adherence in Africa.

Participants/population
Inclusion criteria: -Male and female healthcare clients/patients.
-Guardians for healthcare clients, e.g. parents for child patients.
Exclusion criteria: -Health personnel and healthcare providers.
Intervention(s), exposure(s) Inclusion criteria : • Designed to assess the effect of SMS interventions on participants healthcare behaviour, such as appointment attendance or medicine adherence.
• At least one SMS intervention arm that is exclusive SMS, such as SMS reminders, health educative SMS or SMS quiz • A control group which receives standard care.
If co-interventions (e.g. written materials supplied to participants) are received by participants in both intervention and control arms we consider this to be part of standard care, and such studies will be included.If we find any studies that are described as pilot studies, these we will be included if they live up o the other inclusion criteria.
Exclusion criteria: • SMS intervention arm is interactive. For example, if a study has two arms (1) SMS and (2) standard care, but the SMS service is interactive, meaning the participant can respond to the SMS and receive e.g. a phone call from a health professional, this study will be excluded.
• SMS intervention arm is integrated with other services, such as phone calls or face-to-face education. For example, a study that has a two-arm intervention with (1) SMS &#43; phone call versus (2) standard care, this study will be excluded. However, if a study has a three-arm intervention with (1) SMS, (2) phone call and (3) standard care, this study will be included as it is possible to evaluate the isolated effect of the SMS.

Comparator(s)/control
A control group which receives standard care.

Context
The included randomised controlled trials must be conducted in Africa. Multicentre studies will be included if separate data is available.

Main outcome(s)
We will include any randomised trial meeting the inclusion criteria above irrespective of which outcomes are reported in the identified publication. -Change in health behaviour as defined by trial authors, for example healthcare appointment attendance, treatment adherence, clinical outcomes Data extraction (selection and coding) The reference management software Endnote X8.0.2 (Thomson Reuters, New York, NY, USA) will be used to organise articles identified in the search. Search results from difference database will be combined in an Endnote library and duplicate records will be removed. The remaining literature will be uploaded to Covidence which is a systematic review tool designed to facilitate the process of screening and enable two reviewers to work efficiently through the steps of a systematic review. Covidence will be used to manage the second duplicate search, title-abstract screening and full text screening.
Using Covidence, two reviewers (DSL, JK) will independently screen titles and abstracts of all retrieved articles for inclusion. Any inconsistency in the identification of potentially relevant papers will be discussed until consensus can be reached. For studies selected for full text screening, two reviewers (DLS, MK) will independently assess the full text articles for inclusion. Disagreements will be resolved through discussion and if consensus cannot be reached and arbiter (AL) will make the final decision.
One author (DSL) extract data and a second author (MK) verifies data extraction for trial outcomes. Data will be extracted into Excel. The extraction will include:

Risk of bias (quality) assessment
Two reviewers (DSL, MK) will independently will independently assess trials for risk of bias using the Cochrane Risk of Bias Tool. We will assess contamination bias and assess the following domains: random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective reporting and other biases. For cluster randomised trials we Page: 3 / 5 PROSPERO International prospective register of systematic reviews will assess risk of bias related to recruitment bias and baseline imbalance. Disagreements will be resolved through discussion and if consensus cannot be reached and arbiter (AL) will make the final decision.

Strategy for data synthesis
We will analyse all data by performing intention-to-treat analysis using available case analysis. Using Review Manager 5 (RevMan 2012), we will calculate pooled RRs and estimate 95% confidence intervals (CIs) using the random-effects model with the Mantel-Haenszel method for dichotomous data. We will use randomeffects model instead of a fixed effect model due to the anticipated clinical heterogeneity of trials.
For cluster randomized trials we will include data in meta-analysis, if possible, using methods described in Cochrane Handbook on calculating effective sample sizes.
We assessed statistical heterogeneity by using I².

Analysis of subgroups or subsets
We will conduct a subgroup analysis to compare low risk of bias trials with high risk of bias trials. We defined low risk of bias trials as trials with low risk of selection bias, detection bias and selective reporting, and all other trials as having high risk of bias.
Further, we will conduct a subgroup analysis on level of disease. We will conduct a subgroup analysis to compare trials randomized on individual level with trials using cluster randomization.
To test the robustness of our findings, we will perform a sensitivity analysis in which we re-analyse our primary outcome using a fixed-effect instead of a random-effects model. Further, we will conduct sensitivity analyses (1) where we exclude cluster trials, and (2)  PROSPERO This information has been provided by the named contact for this review. CRD has accepted this information in good faith and registered the review in PROSPERO. The registrant confirms that the information supplied for this submission is accurate and complete. CRD bears no responsibility or liability for the content of this registration record, any associated files or external websites.