Inclusion and exclusion criteria

Any study reporting the incidence or prevalence of RHD that were conducted in South Africa and published in English or Afrikaans between the periods 27 April 1994 and 26 April 2014 will be considered.

No age restrictions will be used during the article retrieval process. We will include any study that estimated one or more of the following epidemiological measures of RHD burden: incidence, prevalence, remission rate and relative risk of mortality (ie, excess mortality) or cause-specific mortality. Prevalence of RHD has in the recent past been defined by screening programmes of subclinical disease in asymptomatic populations.10––12 Hospital-based studies, however, focus on clinical disease in symptomatic populations. As far as possible, we will elucidate the diagnostic methods in either echocardiography screening studies or hospital-based studies. In the context of RHD, ‘remission’ is due exclusively to surgical intervention; thus, we will explore the surgical literature for rates of disease regression versus long-term progression (ie, mortality). In our clinical experience, the following morbid outcomes are considered the most significant and we will thus include: heart failure, ischaemic/thromboembolic or haemorrhagic stroke, atrial fibrillation, infective endocarditis and valve repair or replacement. In addition, we will consider any study of cardiovascular morbidity that quantified the attributable proportion of RHD cases.

The burden of structural heart disease (including RHD) in South Africans seeking antenatal care was recently reviewed,13 so for the purpose of this review we will exclude studies of RHD in pregnancy. Studies will be excluded if they focused on degenerative heart valve disease, rheumatological conditions other than RHD or solely on acute rheumatic fever. As a rule, we will also exclude autopsy and necropsy studies because the consent rates for these procedures in South Africa are low, and their conclusions about mortality patterns are highly likely to be biased. We will immediately exclude editorials, commentaries and case reports. Although excluded from the final analysis, reviews pertaining to RHD in South Africa will be retained temporarily in order to manually search reference lists. While the WHO recommends considering the inclusion of disease register data in the burden of disease studies, previous experience with RHD registers in South Africa has demonstrated that such data collection is unreliable and not representative of the general population.14 In addition, rheumatic fever registers excluded information regarding RHD, one of our outcomes, during the period of interest. They are thus excluded from our analysis. Similarly, the latest South Africa Demographic and Health Survey (2003) does not contain primary data on RHD and thus could not be reviewed.15

Search strategy

Two clinician researchers (LZ paediatric cardiology; DW internal medicine) will compile lists of articles obtained from three large databases relevant to the South African population: PubMed, ISI Web of Science and EMBASE. Additionally, to identify South African conference proceedings, theses and abstracts, we will manually search the following archives at the University of Cape Town Health Sciences Library: Current and Completed Research (South Africa), and two journals not indexed on MEDLINE: SA Heart and the Cardiovascular Journal of South Africa. References will be managed using EndNote X7 software. We will collect vital registration data from Statistics South Africa. Although the flaws in vital statistics are well known, the Global Burden of Disease study considers them an important source of mortality data and incorporates specific methods for handling misclassifications and inconsistencies.16 Owing to the fact that no RHD registers are used within the country, these cannot be included. Finally, we will hand search the reference lists of all studies included in the final review. Prior consultation with other RHD experts led us to suspect a substantial ‘grey’ literature around RHD. Thus, we will intentionally keep the database search strategy broad and redundant. We will also communicate with other South African cardiovascular disease researchers and practitioners, as well as international experts on RHD when possible, to identify unpublished works. Published and unpublished data will be subject to the same quality assessment. The prespecified search strategy for each database mentioned previously is listed in table 1.

The two researchers (LZ and DW) will independently, in the first instance, select articles on the basis of relevant title and abstract, after which full-text manuscripts will be obtained from potentially eligible reports. When discrepancies arise over the inclusion of titles/abstracts or full-text articles, we plan to resolve them by consensus discussion between the two primary reviewers (LZ and DW), with arbitration by a third reviewer (MEE) as necessary.

Data extraction

The primary reviewers will use a standardised data extraction form to independently extract information from included articles. This extraction will be duplicated (ie, not split between the two authors) in order to improve reliability. The data extraction form will capture basic study characteristics including objectives, study population, sample size, years and location of study, and study design. Disease-related parameters including hospitalisation, secondary events, surgical interventions and mortality will also be recorded. Where study data are unclear, the original author of the manuscript will be contacted to clarify his or her findings.

Quality assessment

For all studies deemed eligible for inclusion, we will assess overall study quality according to four basic criteria: (1) representativeness of cases to the general population with RHD, (2) completeness of dataset (including follow-up), (3) validity of case definitions and methods of ascertainment and (4) appropriateness of the study design to the research question. These criteria were adapted from general criteria used in the Global Burden of Disease study17 and will each be assigned one point on a four-point scale. For specific estimates reported in a study, we assigned each estimate a grade A, B or C based on the quality of the data and study methods. For instance, a population-based study of incidence would receive an ‘A’, a hospital-based study of incidence accounting for the catchment area would receive a ‘B’, and a study reporting rates of hospitalisation as incidence would receive a ‘C’. The quality assessment will evaluate the reliability of the estimates of the outcome measures. Again, we will resolve discrepancies in data extraction or quality assessment of study quality by consensus discussion between the two primary reviewers (LZ and DW), with arbitration by a third reviewer (MEE) as necessary.

Risk of bias assessment

In addition to the quality assessment, we will also include an assessment of risk of bias. The Cochrane collaboration suggests that the phrase ‘risk of bias’ is the preferred terminology in reflecting the risk of underlying bias in study design or execution, in addition to the effect of the exposure of intervention under study. The risk of bias will be assessed using the design-specific criteria outlined in the publication by the Agency of Health-related Research and Quality and listed in table 2. These permit the assessment of selection, performance, attrition, detection and reporting biases.18

Data synthesis

Prevalence data from individual studies will be combined by random-effects meta-analysis according to the Mantel-Haenszel method. Heterogeneity will be evaluated using the χ²-based Q statistic (significant for p<0.1) and the I2 statistic (>50% to be indicative of ‘notable’ heterogeneity.19 STATA software V.11.2 (STATA Corporation, College Station, Texas, USA) will be used to perform calculations and the meta-analysis and to produce the forest plots using the metan routine. Should standard error (SE) not be provided, CIs will be incorporated into the formula, SE=(upper limit—lower limit)/3.92.

Presenting and reporting of results

We plan to make use of flow diagrams to summarise the study selection process and detail reasons for exclusion. This will follow the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines for reporting of systematic reviews.20 We will publish our search strategy and quality-scoring tool as supplementary documents.

Primary outcomes

Secondary outcomes

Fatal and non-fatal outcomes data will be expressed in the pooled analysis where possible. We will tabulate estimates of crude age-specific mortality rates from RHD per 100 000 persons per year along with their 95% CIs. Attributable proportions and relative risks of fatal and non-fatal outcomes will be calculated if data are available and 95% CIs generated. A measure of the consistency of results will be included and in cases where a meta-analysis cannot be performed, a narrative summary will be presented. Any trends will be reported on, either using meta-analytic methods such as meta-regression if possible or a detailed qualitative assessment.

Dissemination

This protocol will result in a systematic review of the contemporary incidence, prevalence and fatal and non-fatal outcomes of RHD in South Africa. The burden of RHD in developed nations has changed dramatically over the past century while far fewer reductions have occurred in developing countries.10 Emerging economies such as South Africa, although demonstrating some important medical and healthcare advances, struggle with the continued health burden of diseases of poverty and marginalised communities, such as those living with RHD. The conclusions of this review will be critical to provide high-level healthcare providers, public health officials and policymakers with pooled contemporary data regarding RHD, in particular regarding the effect the new political dispensation has had on the burden of a preventable disease within the country.

A recent publication outlining the most important research priorities for RHD has stressed that high-quality country-specific data are crucially important to determine the research priorities for a particular region.22 We propose that this review will provide these data to affect policy decisions regarding prevention, management and control of RHD. Contemporary measures of fatal and non-fatal outcomes of RHD are also crucially needed to change or amend current practice and plan outcome-based research in RHD.

Finally, we anticipate that this review will identify trends in the burden of disease over the past two decades within South Africa. These trends may, in turn, identify key areas of intervention for future research and practice, such as specific evidence-based interventions or innovative secondary prophylaxis measures.