This study will adhere to the Preferred Reporting Items for Systematic Reviews and Meta-analyses extension for Diagnostic Test Accuracy studies extension (PRISMA-DTA) standards in reporting the findings of this review. [19] This content of this protocol has been written in accordance with the Preferred Reporting Items for Systematic Review and Meta-analyses Protocols (PRISMA-P) guidelines.[20] This systematic review was registered with the International Prospective Register of Systematic Reviews (PROSPERO) on 27 September 2019 with registration number CRD42019141498.
Criteria for considering studies for this review
A study will only be deemed eligible for review if it fulfils the inclusion criteria and if it does not fulfil any of the exclusion criteria as demonstrated in Table 1.
Types of studies
We will include prospective, retrospective, cross-sectional and randomized-controlled trial studies which report the accuracy of sputum colour in identifying the presence of bacteria in AECOPD. We will not restrict studies by language or publication date. We will only include studies in which the data required to populate 2 X 2 tables are reported, can be reconstructed from reported summary estimates or can be provided by the authors of primary diagnostic studies.
Case reports, case series, diagnostic case-control studies with healthy controls and studies presenting insufficient data for the construction of a 2 X 2 table will be excluded from this review.
Participants
We will include studies of participants diagnosed with COPD that have been complicated by acute exacerbation of any severity. We will include studies involving participants of any age greater than or including 18 and any gender. In studies where participants are not limited to adults with COPD, only data pertaining to this patient groups will be reviewed.
We will exclude studies involving participants with stable COPD or other respiratory diseases including pneumonia, bronchiectasis, asthma, acute bronchitis and interstitial lung disease.
Index tests
Our index test is sputum colour as a marker for the presence of bacteria in the sputum of adults with AECOPD. We will include studies in which sputum colour has been macroscopically assessed by a professional health care worker or reported by patients. Professional health care workers will for this purpose include doctors, nurses, physiotherapists, respiratory therapists and laboratory based medical scientists. An index test will be considered positive if sputum colour is assessed as “purulent”, “green”, “yellow” and “brown”. An index test will be considered negative if sputum colour is assessed as “mucoid”, “colourless”, “grey” or “white”.
Target conditions
AECOPD secondary to bacterial infection is the target condition. This is defined as the acute worsening of respiratory symptoms, in a patient with COPD, which requires further therapy and is likely due to bacterial aetiology.[5]
Reference standards
The reference standard is the detection of potentially pathogenic bacteria on bacterial culture of an adequate sputum sample. A sputum sample will be regarded as adequate if it satisfies either the Murray-Washington or Bartlett microscopic assessment criteria. The Murray-Washington criteria define an adequate sputum specimen by the presence of less than 10 squamous epithelial cells per low-power field.[21] The Bartlett criteria derives a score based on the number of neutrophils per low-power field, the presence of mucous strands and the number of squamous epithelial cells per low-power field.[22] A score of 1, 2 or 3 defines and adequate sample.[22]
Search methods for identification of studies
Electronic searches
We will search the following databases for studies to be included in this systematic review: PubMed, Scopus, Web of Science, Africa-Wide, CINAHL, EMBASE and Health Source Nursing Academy. We will use a combination of Medical Subject Heading (MeSH) terms and free text terms on these platforms. All databases will be searched from inception until present. Table 2 demonstrates our main search strategy for our PubMed search which will be modified for the other databases.
Searching other resources
We will review citation indexes and the references lists of the studies identified through the electronic search for additional articles not found during the initial search. We will also conduct a grey literature search to include conference papers, theses and other unpublished papers (Global Index Medicus, OpenGrey, OpenUCT and OpenDoor).
Data collection and analysis
The screening process and study selection will be completed in accordance with the guidelines published in the Cochrane Handbook for Systematic Reviews of Diagnostic Test Accuracy.[23]
Selection of studies
We will use Rayyan QCRI to assist with the screening of titles and abstracts.[24] Two primary reviewers will independently screen all titles identified by the search strategy. The reviewers will complete a standardised coding sheet (Google Forms) indicating whether a study has met the inclusion criteria or the reason a study has been excluded. Duplicated studies will be removed. The more recent publication with the most complete dataset will be included in the event that duplicate publications for the same data are reported.
The reviewers will select studies from the search strategy in two phases:
Phase 1: Screening of titles and abstracts
The primary reviewers will evaluate all titles and abstracts from the search strategy against the predetermined inclusion criteria. The full text of a study will be reviewed if it is not apparent from the title and abstract whether a study has met the inclusion criteria, or if both primary reviewers do not exclude the study.
Phase 2: Screening of full-text studies
The full text of all potentially eligible studies will be reviewed. A third reviewer will adjudicate any discrepancies between the primary reviewers. The reasons for the exclusion of studies will be documented and presented in a table of excluded studies.
Data extraction and management
Two primary reviewers will develop a data extraction form using Google Forms and will independently extract data from all studies fulfilling the eligibility criteria. The data extraction form will be piloted on at least three potentially eligible studies. A third reviewer will adjudicate any discrepancies between the primary reviewers. The following data will be extracted from the included studies:
- Study characteristics: authors, year of publication, geographical region of study conduct, study design, sample size, clinical setting, index test, reference standard and funding source.
- Participant characteristics: Age, gender, disease and exacerbation severity.
- Outcomes measures: values of diagnostic 2 x 2 table(s) (number of true positive (a), false positive (b), false negative (c), and true negative (d)) that cross‐classified the disease status (as determined by the reference test) and the index test’s outcome for each index test evaluated in the included studies) and number of inconclusive results.
We will contact the relevant authors of primary diagnostic studies in the event of missing data.
Assessment of methodological quality
Two investigators will independently evaluate the risk of bias for each article reviewed. Any disagreement with be resolved by a third reviewer. Findings will be reported in accordance with the Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2) tool.[25]
QUADAS‐2 is the redesigned and improved version of the Quality Assessment of Diagnostic Accuracy Studies list (QUADAS). It comprises four domains: patient selection, index test, reference standard, and flow and timing. Risk of bias is assessed for each domain, and for the first three domains, a statement on concerns regarding applicability is given. Each key domain has a set of signalling questions to help judge the risk of bias and concerns regarding applicability. Signalling questions are answered as 'yes', 'no', or 'unclear'. Risk of bias is rated as 'low risk of bias', 'high risk of bias', or 'unclear risk of bias'. Concerns regarding applicability are rated as 'low', 'high' or 'unclear'. In Appendix 3 we provide the precise criteria, tailored for this review, with which we expect to assess risk of bias and applicability.
Statistical analysis and data synthesis
Statistical analysis and data synthesis will be completed in accordance with the guidelines published in the Cochrane Handbook for Systematic Reviews of Diagnostic Test Accuracy.[26]
Our analysis will be conducted at the level of the sample and not at the level of the participant. For example, if a participant in a study produces more than one sputum sample, each sample will be regarded as an independent index test. We will plot the identified sensitivities and specificities for each index test on Forest plots using RevMan.[27] This will allow for visual examination of variation in test accuracy across studies. We will aim to use values reported in the diagnostic 2 X 2 for each included study. If these data are unavailable, we will attempt to reconstruct these values from reported summary measures. Studies with incomplete or inconclusive index test results will be excluded from statistical analysis but will still be summarized in the descriptive analysis. The results of the index test will be dichotomous; “Positive” or “Negative” based on colour of the sputum. Meta‐analysis using a bivariate logistic regression model with random effects will be conducted around this common threshold (Positive/Negative) using the xtmelogit function in Stata V.14.2 (Stata Corp, College Station, Texas, USA). This will estimate pooled sensitivity and specificity (with 95% confidence intervals).
Investigations of heterogeneity
We will explore heterogeneity through the visual examination of the forest plots of sensitivities and specificities and through the inclusion of possible sources of heterogeneity as covariates in a meta-regression model if sufficient studies are included in the analysis. We will investigate the following sources of heterogeneity:
- Study setting, that is, outpatient versus inpatient.
- Antibiotic uses, that is, if patients were exposed to antibiotics within 4 weeks of participation in the study; “Yes” if patients were antibiotic exposed and “No” if patients were not antibiotic exposed,
- Source of index test, that is, if sputum colour was assessed by a healthcare professional or patient reported; “Yes” if sputum colour was assessed by a healthcare profession and “No” if sputum colour was patient reported.
Sensitivity analyses
We will conduct sensitivity analyses to explore the influence of study quality on our findings, drawing primarily on our assessment of study bias using the QUADAS-2 tool. We will first explore the effect of excluding studies in which the index test or reference standard domains are judged as having a high risk of bias or unclear risk of bias as these are considered the most relevant sources of bias. We will then explore the effect of excluding studies in which two or more domains of the QUADAD-2 are judged as having a high risk of bias or unclear risk of bias. We will also use sensitivity analyses to explore the effect of potentially influential studies, such as studies with accuracy estimates markedly different from the rest of the included studies.
Assessment of reporting bias
We will not undertake any formal assessment of reporting bias in our review due to current uncertainty about how to assess reporting bias in reviews of diagnostic test accuracy, particularly in the presence of significant heterogeneity.[26]