Protocol of a Scoping Review on the Use of Transactional Data for Early Diagnosis (TRADED-ScR)

Kurdo Araz; Benjamin M Jacob; James Flanagan; Patrick Redmond

doi:10.12688/hrbopenres.14015.2

Home Browse Protocol of a Scoping Review on the Use of Transactional Data for...

ALL Metrics

Views

Downloads

Get PDF

Get XML

Export

▬

✚

Study Protocol

Revised

Protocol of a Scoping Review on the Use of Transactional Data for Early Diagnosis (TRADED-ScR)

[version 2; peer review: 2 approved with reservations]

Kurdo Araz ^1,2, Benjamin M Jacob², James Flanagan³, Patrick Redmond²

PUBLISHED 05 May 2026

Author details Author details

¹ School of Medicine, RCSI University of Medicine and Health Sciences, County Dublin, County Dublin, Ireland
² Department of General Practise, RCSI University of Medicine and Health Sciences, Dublin, County Dublin, Ireland
³ Department of Surgery and Cancer - Faculty of Medicine, Imperial College London, London, England, UK

Kurdo Araz
Roles: Conceptualization, Methodology, Visualization, Writing – Original Draft Preparation, Writing – Review & Editing

Benjamin M Jacob
Roles: Conceptualization, Writing – Original Draft Preparation, Writing – Review & Editing

James Flanagan
Roles: Conceptualization, Writing – Review & Editing

Patrick Redmond
Roles: Conceptualization, Writing – Original Draft Preparation, Writing – Review & Editing

OPEN PEER REVIEW

REVIEWER STATUS

Abstract

Background

The early detection of many diseases is crucial for effective treatment, as it increases the likelihood of successful management and helps prevent or delay complications. However, early detection is often hampered by the asymptomatic nature of initial disease stages and delays in patients seeking care. In cancers such as ovarian, gastrointestinal, and haematological types, delays may result from self-management of non-specific symptoms with over-the-counter medications. Recent studies, such as the Cancer Loyalty Card Study (CLOCS), suggest that transactional data can reveal early self-medicating behaviours indicative of cancer. Despite this potential, a comprehensive understanding of the use of transactional data for early diagnosis across diseases is lacking.

Aim

This scoping review aims to systematically collate and analyse the literature on the use of transactional data for early detection of any disease, assessing its viability as a predictive tool.

Methods

The review will follow the Arksey and O’Malley methodological framework, with enhancements by Levac et al., and will be reported according to PRISMA-P and PRISMA-ScR guidelines. A comprehensive search will be conducted across databases including MEDLINE, Embase, Scopus, Web of Science, and the Cochrane Database of Systematic Reviews. The review will include studies that utilise transactional data to identify early signs of disease, focusing on both peer-reviewed articles and conference abstracts. The data will be thematically charted and synthesised to compare methodologies, disease types, types of transactional data used, key findings, and limitations.

Implications

By mapping the use of transactional data as a non-invasive tool for early detection, this review aims to inform and potentially transform approaches to screening and diagnosis. The findings will provide insights for healthcare professionals, researchers, and policymakers, supporting the development of targeted interventions that leverage transactional data for disease surveillance and early detection.

Keywords

Early diagnosis, Screening, Transactional data, Health informatics, Consumer behaviour, Scoping review

Corresponding author: Kurdo Araz

Competing interests: No competing interests were disclosed.

Grant information: This research was supported by a grant from the HSE National Cancer Control Programme. SDAP-2023-033 Tailoring Lung Cancer Screening for Ireland: Using Existing Datasets to Generate Evidence-Based Recommendations for Health Policy.
The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Copyright: © 2026 Araz K et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

How to cite: Araz K, Jacob BM, Flanagan J and Redmond P. Protocol of a Scoping Review on the Use of Transactional Data for Early Diagnosis (TRADED-ScR) [version 2; peer review: 2 approved with reservations]. HRB Open Res 2026, 8:3 (https://doi.org/10.12688/hrbopenres.14015.2) First published: 10 Jan 2025, 8:3 (https://doi.org/10.12688/hrbopenres.14015.1) Latest published: 05 May 2026, 8:3 (https://doi.org/10.12688/hrbopenres.14015.2)

Revised Amendments from Version 1

Methods – Search strategy: Specified OSF reference to Appendix A, indicating where to find detailed search strategy.
Methods – AI screening: Specified the AI tool (Research Screener/Rayyan), described validation using Phase 1 results (sensitivity ≥95% threshold), and outlined safeguards including human verification of excluded articles.
Methods – Screening criteria (Table 1): Expanded with operational definitions of "transactional data" and "early detection," plus explicit rules for borderline cases.
Methods – Data extraction (Table 2): Added a code legend explaining all abbreviations (e.g., P-Journal, M-Cohort) and a new field for "definition of early detection."
Methods – Predictive value analysis: Provided detailed methodology for calculating implied PPV from sensitivity/specificity or OR/HR, including stratification by study design and handling of incomplete data.
Methods – Quality assessment: Added domain-specific adaptations of ROBINS-E/ROBINS-I for transactional data studies (confounding, selection bias, information bias), presented as a traffic-light table.
Methods – Ethics and commercial interests: New subsection on handling ethical approval, privacy/GDPR compliance, and commercial conflicts of interest.
Methods – Timeline: Added a 13-month project timeline.
Terminology: Replaced "event horizon" with "lookback period" throughout; added glossary."

See the authors' detailed response to the review by Paul A. Townsend

Introduction

Many diseases begin with subtle symptoms, which can lead to delayed diagnoses that occur only after significant, sometimes irreversible, damage has been done. These late diagnoses are linked to lower survival rates and higher morbidity, highlighting the importance of early detection to improve patient outcomes.¹^–³ Early diagnosis refers to identifying a condition at a stage when interventions can prevent or minimise irreversible damage, thereby improving survival rates, reducing morbidity, and enhancing quality of life.¹^–³

However, the early detection of diseases presents substantial challenges. Many diseases in their early stages are asymptomatic or present with non-specific symptoms, which complicates timely diagnosis. Conditions such as cardiovascular diseases, renal diseases, and various forms of cancer exemplify this issue, where delays in seeking care are common due to the subtlety of early symptoms.⁴^–⁶ These delays are often exacerbated by self-management behaviours, where individuals attempt to alleviate vague symptoms with over-the-counter (OTC) medications. While self-medication can temporarily mask symptoms, it may also obscure the early signs of more serious underlying conditions, such as malignancies.⁷^,⁸

Recent studies, including the Cancer Loyalty Card Study (CLOCS), have explored the use of transactional data—specifically, the purchase of OTC medications—as a potential tool for early disease detection. These studies suggest that patterns in transactional data can reveal self-medicating behaviours that may be indicative of conditions like cancer, offering a novel and non-invasive method for early diagnosis.⁹ Beyond cancer, transactional data has also been used to enhance the detection of infectious disease outbreaks, such as Influenza, by tracking sales of specific OTC medication.¹⁰^,¹¹

The relevance of this approach is further supported by research showing that patients with certain cancers, such as ovarian, haematological, and gastrointestinal cancers, often delay seeking professional advice due to non-specific symptoms, which they manage with OTC medications.¹²^–¹⁴ For example, a survey of over 1,500 women with ovarian cancer revealed that many attributed early symptoms to non-threatening causes such as irregular menstrual cycles, menopause, or aging, leading to delayed diagnosis.¹⁴ Additionally, focus group studies have demonstrated that these women had higher purchases of specific medications prior to diagnosis compared to healthy controls, indicating a potential pattern that could be leveraged for earlier detection.¹⁵

The potential for using transactional data in medical research is further supported by regulatory frameworks like the General Data Protection Regulation (GDPR), which grants patients the right to access their personal data, including transaction records.¹⁶ This regulation could facilitate the ethical use of consumer transactional data for medical research, provided that safeguards such as the ability to withdraw consent, data security measures, and transparency in data management are maintained.¹⁵

Given these developments, transactional data, such as purchase histories of OTC medications, has emerged as a promising avenue for the early detection of cancer and other diseases.⁹^–¹¹ The ability to detect early warning signs through consumer purchasing patterns represents a novel, non-invasive diagnostic approach that could significantly improve early diagnosis strategies. However, the existing literature on this topic is fragmented, with studies varying widely in design, disease focus, and outcomes. To date, there has been no comprehensive synthesis that could clarify the overall efficacy of using transactional data for disease detection and its potential applications in public health and clinical practice.

Aims and objectives

The aim of this scoping review is to provide a detailed overview of how transactional data can be used in the early detection of disease. This review intends to map the range, scope, and nature of the existing research, providing critical insights into the feasibility of utilising such data to enhance early diagnosis strategies.

The specific objectives include:

1. to identify all peer-reviewed articles which describe the use transactional data for early detection, cataloguing them according to methodology, disease type and data type;
2. to report, in the case of observational and interventional studies, the population characteristics, the event horizon studied, and how the transactional data and disease diagnosis data were operationalised;
3. to quantify, where the published data allow, the predictive power of the disease signal identified in the transactional data, by calculating the implied positive predictive value;
4. to discuss the logistical, legal and ethical challenges related to the use of transactional data in disease screening and early detection reported by the authors of the primary studies.
5. to conduct this review in two phases, the first phase will focus specifically on the early detection of cancer, while the second phase will broaden the scope to include the early detection of any disease.

Methods

Reporting and registration

We will conduct a scoping review in accordance the Arksey and O’Malley framework enhanced by Levac et al.¹⁷ The protocol will adhere to the PRISMA-P reporting guidelines and will be pre-registered on the Open Science Framework. The results will be reported following the PRISMA-ScR guideline.¹⁸ Supplementary files and relevant datasets will be made available via the Open Science Framework.¹⁹

Eligibility criteria

The review will include studies that use transactional data for early detection of diseases. Eligible studies must specifically address the use of transactional data, such as loyalty card purchases or OTC medication sales, with incident disease or disease recurrence as an outcome, regardless of the disease type. Studies from all geographical locations will be considered.

Since our aim is to describe the scope of the literature, all article types will be eligible for inclusion: interventional studies, observational studies, qualitative studies, review articles, letters and editorials will all be included. Both peer-reviewed articles (including doctoral theses) and conference abstracts will be included. We will exclude studies that do not focus on transactional data or are not concerned with diagnosis as an outcome.

Information sources

In Phase 1, the databases to be searched include MEDLINE, Embase, Scopus, Web of Science, and the Cochrane Database of Systematic Reviews. In Phase 2, additional sources such as CINAHL, ProQuest Dissertations and Theses Global (PQDT Global), three clinical trial registries (ClinicalTrials.gov, EU Clinical Trials Register, and International Standard Randomised Controlled Trial Number), and two systematic review registries (PROSPERO and Joanna Briggs Institute) will also be searched.

Search strategy

The search strategy, developed with an information specialist, will combine keywords and subject headings related to the themes of “transactional data” and “early detection”. To ensure thorough coverage, the strategy will also incorporate synonyms and related terms such as “purchase data,” “loyalty card data,” “retail data,” and “consumer behaviour data.” The detailed search strategies for both phases are available via the Open Science Framework in Appendix 1 – Search Strategy: https://doi.org/10.17605/OSF.IO/ZRJT5.²⁰ Based on pilot searches, it is anticipated the number of papers to be screened in Phase 1 will be less than 20,000, while the number of papers to be screened in Phase 2 is expected to be over 100,000. References of included studies will be reviewed to identify additional relevant works.

Selection of sources of evidence

Two reviewers will independently screen titles and abstracts for eligibility using Rayyan, a software that designed to facilitate efficient screening of abstracts and titles.²¹ Reviewers will follow the screening algorithm ( Table 1) to determine included studies and excluded studies. Full-text screening will follow for potentially relevant studies. Discrepancies will be resolved through discussion or consultation with a third reviewer. A PRISMA flow diagram will document the selection process. The screening algorithm, which will be piloted using a limited search of the MEDLINE database, is outlined on Open Science Framework: https://doi.org/10.17605/OSF.IO/ZRJT5.²⁰

Table 1. Screening algorithm.

Step	Concept	Question	Operational definition	Exclusion label
1	Transaction	Does this paper deal with transaction data?	Transactional data = loyalty card records, OTC medication sales, retail purchase histories, supermarket scanner data, pharmacy transaction logs. Exclude: prescription-only medication data, self-reported purchases, or healthcare claims data without individual purchase records.	“NotTransactional”
2	Diagnosis	Does this paper discuss the association between patterns in transaction data and a subsequent disease diagnosis/incidence/risk?	Early detection = diagnosis at a stage preceding typical clinical presentation (pre-symptomatic or early symptomatic phase). Include if diagnosis is confirmed by medical records, registry, or clinical assessment.	“NotDiagnosis”
3	(INCLUDE)	If negative to all the above, then include		n/a

Given the high volume of articles anticipated in Phase 2, title and abstract screening may be assisted by an artificial intelligence tool. We will use Research Screener²² or Rayyan’s AI functionality. The validity of this approach will be assessed using the Phase 1 results (where human reviewers have already made decisions) as a gold standard. We will calculate the AI’s sensitivity (proportion of included articles correctly identified) and specificity. Only if sensitivity is ≥95% will we proceed with AI-assisted screening for Phase 2. In that case, two reviewers will independently verify all articles flagged for inclusion by the AI and a random 10% of excluded articles. Any disagreements will be resolved by a third reviewer. Regular calibration checks will be performed every 500 abstracts. The detailed approach will be outlined in a subsequent protocol akin to a previous study implementing a similar approach.²²

Data collection and data management

A standardised form will be used to extract data from selected studies based on Table 2. Extracted data will include study characteristics (e.g. author, year, disease type), transactional data specifics (data sources, items tracked), methodology (e.g. data collection, analysis techniques), and key findings (e.g. disease symptoms identified, effectiveness of detection, and limitations).

Table 2. Data charting pro-forma.

Step	Concept	Question	Inclusion label example
1	Publication details	(Automatically extracted by Rayyan software)	Title, Authors, Publication Year, Data Source (e.g., Journal, Registry)
2	Publication type	What type of record/publication is this?	P-Journal, P-Conference, P-Registration,
3	Methods	What type of study is this?	M-Review, M-Trial, M-CCC, M-Cohort, M-Other
4	Location	Where did the study population live?	[Coded as “Note”] L-Korea, L-UK, L-NA
5	Disease type	What disease was being studied?	[Coded as “Note”] C-Pancreas, C-CRC, C-Ovarian, C-Lung
6	Disease data	What data was used to determine disease diagnosis?	[Coded as “Note”] D-CPRD, D-THIN
7	Transactional data	What type of transactional data was used?	[Coded as “Note”] T-Statement, T-Loyalty

Data items

This review aims to identify and catalogue peer-reviewed articles that utilise transactional data for early disease detection, categorising them based on methodology, disease type, and data type. For observational and interventional studies, it will report population characteristics, the time window between data collection and diagnosis (the ‘lookback period’ studied, and operationalisation of transactional and disease diagnosis data. Additionally, it will quantify the predictive power of the studied transactional data signal, when possible, by calculating the implied positive predictive value.

Critical appraisal

While scoping reviews do not typically require quality assessment, we will critically appraise included studies to inform the interpretation of findings. For observational studies of exposures (e.g., purchase patterns), we will use ROBINS-E; for non-randomised intervention studies, we will use ROBINS-I.²³^,²⁴ However, we acknowledge that these tools require adaptation for transactional data research. We will therefore assess the following domain-specific biases: (1) Confounding: whether studies adjusted for socioeconomic status, healthcare access, comorbidities, and prior healthcare utilisation; (2) Selection bias: whether the transactional data source (e.g., loyalty card holders) is representative of the target population; (3) Information bias: validity of purchase data as a proxy for medication consumption (purchased ≠ ingested); (4) Temporal bias: whether the lookback period is appropriately aligned with disease natural history. We will not calculate a summary quality score but will present domain-level assessments in a narrative summary and a traffic-light table (green/yellow/red) for each included study. Meta-biases and confidence in cumulative evidence will not be reported since we are conducting a scoping review that does not focus on a single primary outcome.

Data synthesis

The data synthesis will proceed through four steps. First, a bibliometric analysis will map the research landscape by categorising identified articles by methodology, disease type, and data type, providing a high-level overview of the field. Second, study details will be analysed, including population characteristics, study durations, and the operational definitions and uses of transactional and disease diagnosis data, using simple descriptive statistics and narrative approaches as appropriate. Third, where data are sufficient, we will quantify the implied positive predictive value (PPV) of the transactional data signal. For studies reporting sensitivity and specificity, we will calculate PPV = (sensitivity × prevalence) / [(sensitivity × prevalence) + ((1 − specificity) × (1 − prevalence))] using prevalence estimates from population-based cancer registries or epidemiological studies matched for age, sex, and region. For studies reporting odds ratios or hazard ratios, we will convert these to approximate PPVs. Where published data do not allow PPV calculation, we will narratively report the available measures of association (e.g., OR, HR, AUC) and explicitly state that PPV could not be derived. Results will be stratified by study design (case-control vs. cohort vs. cross-sectional), as PPV interpretation differs fundamentally across these designs. Finally, the review will discuss the logistical, legal, and ethical challenges highlighted in the studies, addressing the complexities of using transactional data in disease detection.

Handling of ethical and commercial considerations: We will systematically extract and report the following for each included study: (1) Ethical approval: whether the study reports obtaining ethics committee approval; (2) Data governance: any reported GDPR compliance, data anonymisation methods, patient consent procedures, or data withdrawal mechanisms; (3) Commercial conflicts: funding sources (e.g., retail companies, loyalty card providers, pharmaceutical firms) and author affiliations with commercial entities. Studies that use transactional data without documented ethical approval or patient consent will be flagged in the synthesis. Any potential outcome reporting bias associated with commercial funding will be discussed. This information will be presented in a separate table and considered when interpreting the strength of evidence.

Project timeline

Phase 1 (cancer only) – database searches and screening (Month 1–3), data extraction (Month 4); Phase 2 (all diseases) – database searches (Month 4), AI-assisted title/abstract screening (Months 5–7), full-text screening (Month 8), data extraction (Months 9–10), analysis and synthesis (Months 11–12), manuscript submission (Month 13).

Discussion

Summary

This scoping review protocol presents a systematic approach to examining the use of transactional data in the early detection of diseases. It highlights key areas of interest, including the types of diseases studied, the nature of the data utilised, the methodologies employed, and the challenges encountered in this emerging field. The significance of this review lies in its potential to clarify the utility of transactional data—such as loyalty card purchases and OTC medication sales—as a tool for early diagnosis across a range of conditions.

Strengths and limitations

A major strength of this protocol is its comprehensive and methodical approach to literature search and data extraction, ensuring that the review will thoroughly explore the landscape of research in this area. The inclusion of various study types and the systematic categorisation of findings by disease type, data type, and methodology will provide a broad and detailed overview of the current state of the field.

However, several limitations must be acknowledged. First, there is the issue of publication bias, where studies with negative or inconclusive results may not have been published, potentially skewing the literature. Additionally, language bias could limit the scope of this review, as studies published in languages other than English may be excluded. Another challenge arises from the need to synthesise a large and diverse body of literature into a coherent summary, which may overlook some nuances of individual studies.

A significant limitation specific to this area of research is the restricted access to raw transactional data. For example, in studies like the Cancer Loyalty Card Study (CLOCS), the raw data cannot be shared with other researchers due to privacy and legal constraints. This limitation directly impacts our ability to fully answer the research question, particularly concerning the detailed analysis of transactional patterns associated with early disease detection. While we can report on the scope of the literature in terms of publication details, publication type, methods, location, cancer type, and general descriptions of the datasets used, the lack of access to raw data means that our ability to assess the robustness of the findings reported in primary studies is inherently constrained. However, most primary studies are expected to provide sufficient descriptions of the datasets and methodologies used, allowing us to still draw meaningful insights into the use of transactional data for early diagnosis.

Implications

The findings of this review will be instrumental in shaping future research on the use of transactional data in disease detection. By identifying gaps in the existing literature and highlighting under-researched areas, this review will guide researchers towards optimal study designs and methodologies. Moreover, it will provide healthcare professionals with the latest insights into innovative tools for early detection, potentially leading to advancements in screening and diagnosis practices that could improve patient outcomes.

In addition, this review aims to contribute to the ongoing dialogue on the ethical and privacy concerns associated with the use of transactional data in medical research. By addressing these concerns, we hope to influence the development of guidelines that balance the potential health benefits of this research with the need to protect patient privacy and data security.

Conclusion

In conclusion, the proposed scoping review will provide valuable contributions to our understanding of the potential role of transactional data in early diagnosis. By mapping the research landscape, evaluating methodologies, synthesising key findings, and identifying gaps, this review will inform both current practices and future directions in disease detection research. Despite the limitations posed by the inaccessibility of raw data, this review will offer a comprehensive overview of how transactional data is currently being used and its potential to enhance early detection strategies, ultimately aiming to improve patient outcomes.

Ethics and consent

Ethical approval and consent were not required.

Data availability

No data associated with this article since it is a research protocol and does not include associated datasets.

Datasets generated during research will be uploaded permanently on the Open Science Framework: Protocol of a Scoping Review on the Use of Transactional Data for Early Diagnosis (TRADED-ScR): https://doi.org/10.17605/OSF.IO/ZRJT5 under CC-By Attribution 4.0 International license.²⁰

Extended data

The full search strategy and PRISMA-P checklist are available via the Open Science Framework: Protocol of a Scoping Review on the Use of Transactional Data for Early Diagnosis (TRADED-ScR): https://doi.org/10.17605/OSF.IO/ZRJT5 under CC-By Attribution 4.0 International license.²⁰

Reporting guidelines

PRISMA-P (certain elements modified and added to reflect the specific PRISMA-ScR guideline, as outlined in PRISMA-P checklist) checklist is available via the Open Science Framework.

https://doi.org/10.17605/OSF.IO/ZRJT5 under CC-By Attribution 4.0 International license.²⁰

References

1. Yuan X, Chen S, Sun C, et al.:A novel early diagnostic framework for chronic diseases with class imbalance. Sci. Rep. 2022; 12(1): 8614. PubMed Abstract | Publisher Full Text | Free Full Text
2. Whitaker K:Earlier diagnosis: the importance of cancer symptoms. Lancet Oncol. 2020; 21(1): 6–8. PubMed Abstract | Publisher Full Text
3. Varghese SE, El Otol RHM, Al Olama FS, et al.:The importance of early detection of genetic diseases. Dubai Med. J. 2021; 4(2): 133–141. Publisher Full Text
4. Cohn JN, Hoke L, Whitwam W, et al.:Screening for early detection of cardiovascular disease in asymptomatic individuals. Am. Heart J. 2003; 146(4): 679–685. PubMed Abstract | Publisher Full Text
5. McClellan WM, Knight DF, Karp H, et al.:Early detection and treatment of renal disease in hospitalized diabetic and hypertensive patients: important differences between practise and published guidelines. Am. J. Kidney Dis. 1997; 29(3): 368–375. PubMed Abstract | Publisher Full Text
6. Whitaker KL, Winstanley K, Macleod U, et al.:Low cancer suspicion following experience of a cancer ‘warning sign’. Eur. J. Cancer. 2015; 51(16): 2473–2479. PubMed Abstract | Publisher Full Text | Free Full Text
7. Davies A, Green MA, Singleton AD: Using machine learning to investigate self-medication purchasing in England via high street retailer loyalty card data.Rao P, editor. PLoS One. 2018;13(11): e0207523. PubMed Abstract | Publisher Full Text | Free Full Text
8. Brewer HR, Chadeau-Hyam M, Johnson E, et al.:Cancer Loyalty Card Study (CLOCS): feasibility outcomes for an observational case-control study focusing on the patient interval in ovarian cancer. BMJ Open. 2023; 13(6): e066022. Publisher Full Text
9. Brewer HR, Hirst Y, Chadeau-Hyam M, et al.:Association between purchase of over-the-counter medications and ovarian cancer diagnosis in the Cancer Loyalty Card Study (CLOCS): Observational Case-Control Study. JMIR Public Health Surveill. 2023; 9: e41762. PubMed Abstract | Publisher Full Text | Free Full Text
10. Das D, Metzger K, Heffernan R, et al.:Monitoring over-the-counter medication sales for early detection of disease outbreaks—New York City. MMWR Suppl. 2005; 54: 41–46. PubMed Abstract
11. Qin Q, Cheng L, Tan L, et al.:Over-the-counter medication sales surveillance for early detection of respiratory epidemics in rural China. Online J. Public Health Inform. 2014; 6(1): e61404. Publisher Full Text
12. Dapkevičiūtė A, Šapoka V, Martynova E, et al.:Time from symptom onset to diagnosis and treatment among haematological malignancies: influencing factors and associated negative outcomes. Medicina (Kaunas). 2019; 55: 31163661. Publisher Full Text
13. Dulal S, Paudel BD, Wood LA, et al.:Reliance on self-medication increase delays in diagnosis and management of GI cancers: results from Nepal. JCO Glob. Oncologia. 2020; 6: 1258–1263. PubMed Abstract | Publisher Full Text | Free Full Text
14. Goff BA, Mandel L, Muntz HG, et al.:Ovarian carcinoma diagnosis: results of a national ovarian cancer survey. Cancer. 2000; 89(10): 2068–2075. PubMed Abstract | Publisher Full Text
15. Flanagan JM, Skrobanski H, Shi X, et al.:Self-care behaviors of ovarian cancer patients before their diagnosis: proof-of-concept study. JMIR Cancer. 2019; 5(1): e10447. PubMed Abstract | Publisher Full Text | Free Full Text
16. Dolan EH, Goulding J, Tata LJ, et al.:Using shopping data to improve the diagnosis of ovarian cancer: computational analysis of a web-based survey. JMIR Cancer. 2023; 9: e37141. PubMed Abstract | Publisher Full Text | Free Full Text
17. Levac D, Colquhoun H, O’Brien KK:Scoping studies: advancing the methodology. Implement. Sci. 2010; 5(1): 69. PubMed Abstract | Publisher Full Text | Free Full Text
18. Page MJ, McKenzie JE, Bossuyt PM, et al.:The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ. 2021; 372: n71. PubMed Abstract | Publisher Full Text | Free Full Text
19. Foster ED, Deardorff A:Open Science Framework (OSF). JMLA. 2017; 105(2). Publisher Full Text
20. Jacob BM, Araz K: Protocol of a Scoping Review on the Use of Transactional Data for Early Diagnosis (TRADED-ScR).OSF Registries;2024; [cited 2024 Oct 21]. Reference Source
21. Ouzzani M, Hammady H, Fedorowicz Z, et al.:Rayyan—a web and mobile app for systematic reviews. Syst. Rev. 2016; 5(1): 210. PubMed Abstract | Publisher Full Text | Free Full Text
22. Chai KEK, Lines RLJ, Gucciardi DF, et al.:Research Screener: a machine learning tool to semi-automate abstract screening for systematic reviews. Syst. Rev. 2021; 10(1): 93. PubMed Abstract | Publisher Full Text | Free Full Text
23. Bero L, Chartres N, Diong J, et al.:The Risk of Bias in Observational Studies of Exposures (ROBINS-E) tool: concerns arising from application to observational studies of exposures. Syst. Rev. 2018; 7(1): 242. PubMed Abstract | Publisher Full Text | Free Full Text
24. Sterne JA, Hernán MA, Reeves BC, et al.:ROBINS-I: a tool for assessing risk of bias in non-randomised studies of interventions. BMJ. 2016; 355: i4919. PubMed Abstract | Publisher Full Text | Free Full Text

Comments on this article Comments (0)

Version 2

VERSION 2 PUBLISHED 10 Jan 2025

Author details Author details

Kurdo Araz
Roles: Conceptualization, Methodology, Visualization, Writing – Original Draft Preparation, Writing – Review & Editing

Benjamin M Jacob
Roles: Conceptualization, Writing – Original Draft Preparation, Writing – Review & Editing

James Flanagan
Roles: Conceptualization, Writing – Review & Editing

Patrick Redmond
Roles: Conceptualization, Writing – Original Draft Preparation, Writing – Review & Editing

Competing interests

No competing interests were disclosed.

Grant information

This research was supported by a grant from the HSE National Cancer Control Programme. SDAP-2023-033 Tailoring Lung Cancer Screening for Ireland: Using Existing Datasets to Generate Evidence-Based Recommendations for Health Policy.
The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Article Versions (2)

version 2

Revised

Published: 05 May 2026, 8:3

https://doi.org/10.12688/hrbopenres.14015.2

version 1

Published: 10 Jan 2025, 8:3

https://doi.org/10.12688/hrbopenres.14015.1

© 2026 Araz K et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Download

Export To

metrics

VIEWS

673

downloads

Citations

SEE MORE DETAILS

CITE

how to cite this article

Araz K, Jacob BM, Flanagan J and Redmond P. Protocol of a Scoping Review on the Use of Transactional Data for Early Diagnosis (TRADED-ScR) [version 2; peer review: 2 approved with reservations]. HRB Open Res 2026, 8:3 (https://doi.org/10.12688/hrbopenres.14015.2)

NOTE: If applicable, it is important to ensure the information in square brackets after the title is included in all citations of this article.

track

receive updates on this article

Track an article to receive email alerts on any updates to this article.

Open Peer Review

Version 2

VERSION 2

PUBLISHED 05 May 2026

Revised

Views

Reviewer Report 16 Jul 2026

Adekola George Adepoju, Public Sector Informatics, Essentrify, Pittsburgh, Pennsylvania, USA

Daniel Adepoju, Product Architecture, Essentrify, Pittsburgh, Pennsylvania, USA

Daniel Cheruiyot, Evidence Lab, Essentrify, Pittsburgh, Pennsylvania, USA

Approved with Reservations

https://doi.org/10.21956/hrbopenres.15915.r56122

SUMMARY:

This article presents a protocol for a scoping review on the use of transactional data, including loyalty card records, retail purchases and over-the-counter medication sales, for the early detection of disease. The review will be completed in two phases: Phase 1 will focus on cancer, while Phase 2 will include all diseases.

The topic is important and suitable for a scoping review because the available evidence is likely to vary in disease area, data source, study design and outcome. Version 2 includes useful additions, including clearer definitions, an AI-assisted screening plan, critical appraisal, ethical considerations and a project timeline. However, several areas require further revision to make the methods clear and reproducible.

Major revisions

1. Correct and clarify the inclusion rule in Table 1

Step 3 states: “If negative to all the above, then include.” This appears inconsistent with Steps 1 and 2. Taken literally, it suggests that a paper should be included if it does not involve transactional data and does not examine disease diagnosis or risk.

The authors may mean that a record should be included if it has not received an exclusion label, but this should be stated clearly. For example:
Include if the answers to Steps 1 and 2 are both positive. If eligibility is unclear from the title and abstract, retain the record for full-text review.
This is important because the table will guide both human and AI-assisted screening.

2. Expand Table 2 to match the planned data extraction

Table 2 includes only publication details, publication type, study method, location, disease type, disease data and transactional-data type. However, the Methods section states that the authors will also examine population characteristics, lookback period, operational definitions, items purchased, predictive measures, key findings, limitations, ethical approval, privacy arrangements, funding and commercial conflicts.

These variables are not included in the displayed form. The amendments also state that a field for each study’s “definition of early detection” and a code legend were added, but the table does not appear to contain this field or a complete explanation of all codes.

The authors should expand Table 2 or provide the complete extraction form through the Open Science Framework. All codes, including M-CCC, D-CPRD, D-THIN and L-NA, should be defined. The coding approach for non-cancer diseases in Phase 2 should also be explained.

3. Clarify the AI-assisted screening procedure

The protocol states that either Research Screener or Rayyan’s AI function will be used. The authors should select and name the exact tool, as the tools may operate differently.

They should also clarify:

whether AI will only prioritise records or will make exclusion decisions;
what the calibration checks every 500 abstracts will involve;
what action will be taken if an eligible paper is found in the 10% sample of AI-excluded records; and
how performance will be checked when moving from cancer literature to literature covering all diseases.

Testing the tool on Phase 1 is reasonable, but performance on cancer studies may not automatically apply to other disease areas. The essential screening procedure should be described in this protocol rather than deferred to a later protocol.

4. Clarify the calculation of positive predictive value

Calculating PPV from sensitivity, specificity and an appropriate prevalence estimate may be reasonable. However, the authors should explain how external prevalence estimates will be selected and matched to the population and disease examined in each study.

The statement that odds ratios and hazard ratios will be converted into approximate PPVs needs more detail. These measures cannot generally be converted directly without further information, such as baseline risk, follow-up period and a defined prediction threshold.

The authors should either provide a referenced method and clearly state the assumptions required or report ORs, HRs and AUCs in their original forms when PPV cannot be reliably calculated.

5. Clarify the scope and treatment of different evidence types

The eligibility criteria include primary studies, qualitative studies, reviews, theses, conference abstracts, letters and editorials. The authors should explain how each source type will be used and how double-counting will be avoided when several publications report the same study.

The protocol should also distinguish individual-level early diagnosis from population-level outbreak surveillance. These uses of transactional data have different units of analysis, outcomes and measures of performance.

ROBINS-E and ROBINS-I do not cover every eligible evidence type. The authors should state which studies will undergo formal appraisal and which tool or approach will be used for each main study design.

Minor revisions

State whether data extraction will be completed by two reviewers independently or checked by a second reviewer.
Explain how duplicate records and multiple reports from the same study will be handled.
Clearly state any language and publication-date restrictions. The Discussion mentions possible exclusion of non-English studies, but this is not specified in the Methods.
Distinguish between ethical approval not obtained, not required and not reported. A lack of reporting does not necessarily mean that approval was not obtained.
Change “use transactional data” to “use of transactional data” in Objective 1.
Change “a software that designed” to “software designed.”
Correct the unmatched bracket in the sentence defining the lookback period.
Replace “event horizon” with “lookback period” in Objective 2.
Use “Appendix A” or “Appendix 1” consistently.
Cite the specific PRISMA-P and PRISMA-ScR guideline papers. Reference 18 currently cites the general PRISMA 2020 statement.

Overall, the protocol addresses a relevant and emerging area. Addressing these points would improve the clarity, reproducibility and scientific strength of the proposed review.

Is the rationale for, and objectives of, the study clearly described?

Yes
Is the study design appropriate for the research question?

Yes
Are sufficient details of the methods provided to allow replication by others?

Partly
Are the datasets clearly presented in a useable and accessible format?

Not applicable

Competing Interests: No competing interests were disclosed.

Reviewer Expertise: Clinical and public health informatics, health data analytics, early disease detection and screening, epidemiology, and evidence synthesis. I am able to assess the health-informatics relevance, study design, data methods, ethical use of health data, and public-health implications of this protocol.

We confirm that we have read this submission and believe that we have an appropriate level of expertise to confirm that it is of an acceptable scientific standard, however we have significant reservations, as outlined above.

CITE

Report a concern

Respond or Comment

Version 1

VERSION 1

PUBLISHED 10 Jan 2025

Views

Reviewer Report 30 Jun 2025

Paul A. Townsend, University of Stirling, Stirling, UK

Approved with Reservations

https://doi.org/10.21956/hrbopenres.15383.r47866

This protocol describes a scoping review looking at how transactional data (like loyalty card purchases and over-the-counter medicine sales) might help detect diseases early.

The idea is that people's shopping patterns could reveal when they are self-medicating, which might indicate underlying health problems, particularly cancer.

Strengths

Clear reasoning: The authors explain convincingly why this approach could be useful - many people delay seeking medical help and try to manage symptoms themselves with shop-bought medicines.

Sensible methodology: A scoping review is the right approach for this emerging field where research is likely to be patchy and varied. Using the established Arksey and O'Malley framework shows good practice.

Thorough search plan: They have planned to search multiple databases comprehensively, with a sensible two-phase approach (cancer first, then other diseases).

Proper guidelines: The protocol follows recognised reporting standards (PRISMA-P and PRISMA-ScR), which is reassuring.

Main Concerns

1. Missing methodological details
Search strategy: Whilst they say full search terms are available online, they should include at least one example in the paper itself so readers can assess the approach.

AI screening: They mention possibly using AI to help screen papers in Phase 2, but do not explain which tool, how it would be validated, or what safeguards would be in place.

Screening criteria: Table 1 is rather basic. They need clearer definitions of what counts as "transactional data" and how they will handle borderline cases.

2. Data handling issues
Extraction form: Table 2 uses confusing codes ("P-Journal," "M-Review") without explanation. They need clearer categories and better definitions of what constitutes "early detection."

Analysis approach: They mention calculating predictive values but do not explain how they'll do this when studies might not provide complete data, or how they will handle different study designs.

3. Quality assessment gaps
The authors plan to use standard risk-of-bias tools but do not explain how these will work for transactional data studies, which are quite different from typical medical research.

4. Ethics and practicalities
There' i limited discussion of how they will handle studies with questionable ethics, privacy issues, or potential commercial conflicts of interest - important given the commercial nature of much transactional data.

Minor points

The term "event horizon" is confusing in this context
No timeline provided for completing the work
Some terminology could be clearer throughout

Essential changes needed:

Include search strategy examples in the main paper
Expand screening criteria with clearer definitions
Clarify AI screening plans with proper validation details
Improve data extraction forms with better coding systems
Explain analysis methods more clearly, especially for calculating predictive values

Would be helpful:

Add a realistic timeline
Strengthen the quality assessment section
Address ethical considerations more thoroughly

This is interesting and potentially important research tackling a novel area. The overall approach is sound, but several key methodological details need adding/clarifying to make the review reproducible and reliable. The authors deserve credit for taking on this innovative topic, but the protocol needs tightening up to meet proper systematic review standards.
Once these issues are addressed, this could provide a valuable foundation for understanding how shopping data might help with early disease detection.

Is the rationale for, and objectives of, the study clearly described?

Yes
Is the study design appropriate for the research question?

Yes
Are sufficient details of the methods provided to allow replication by others?

Partly
Are the datasets clearly presented in a useable and accessible format?

Not applicable

Competing Interests: No competing interests were disclosed.

Reviewer Expertise: cancer early detection, machine learning, multiple omics and mass spec

I confirm that I have read this submission and believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard, however I have significant reservations, as outlined above.

CITE

Report a concern

Author Response 05 May 2026

Kurdo Araz, Royal College of Surgeons in Ireland Department of Medicine, County Dublin, Ireland

05 May 2026

Author Response
Dear Dr. Townsend,

Re: Response to Review of Protocol: "Protocol of a Scoping Review on the Use of Transactional Data for Early Diagnosis (TRADED-ScR)" [version 1; peer review: 1 ... Continue reading
Dear Dr. Townsend,

Re: Response to Review of Protocol: "Protocol of a Scoping Review on the Use of Transactional Data for Early Diagnosis (TRADED-ScR)" [version 1; peer review: 1 approved with reservations]

Thank you very much for your thoughtful and detailed review of our protocol. We greatly appreciate your recognition of the study’s strengths, including its clear rationale and sensible methodological framework. Your constructive feedback on the methodological details, data handling, quality assessment, and ethical considerations is invaluable, and we have carefully revised the protocol to address each concern.
Below, we provide a point-by-point response to your comments. All corresponding changes have been made in the revised manuscript (marked in yellow highlight for your convenience).

1. Missing methodological details

Search strategy example: You rightly note that an example search strategy should be included in the main paper. We have now added Appendix A (presented within the manuscript) showing the full MEDLINE search strategy as a representative example. The complete strategies for all databases remain available on OSF.

AI screening: We have expanded this section to specify that we will use the Research Screener tool (Chai et al., 2021 – already cited) or Rayyan’s built-in AI functionality. We now explicitly describe the validation process: (1) The AI will be run on the Phase 1 results (where we have human reviewer decisions); (2) We will calculate sensitivity (proportion of included articles correctly identified) and specificity; (3) Only if sensitivity ≥95% will we rely on AI-assisted screening for Phase 2; (4) Two human reviewers will still verify all AI-excluded articles flagged as potentially relevant. Safeguards, including regular calibration checks and resolution of disagreements by a third reviewer, are now described.

Screening criteria (Table 1): We have substantially expanded Table 1. It now includes explicit operational definitions for "transactional data" (e.g., loyalty card records, OTC medication sales, retail purchase histories, supermarket scanner data, pharmacy transaction logs) and "early detection" (diagnosis at a stage preceding typical clinical presentation, including pre-symptomatic or early-symptomatic phases). We have also added a fourth step to handle borderline cases (e.g., studies using prescription data only or those that do not link purchases to an incident diagnosis), with clear inclusion/exclusion rules.

2. Data handling issues

Extraction form (Table 2): We apologize for the lack of clarity. Table 2 has been completely revised. Codes are now explained in a legend directly beneath the table (e.g., "P-Journal = peer-reviewed journal article; M-Review = systematic/scoping/narrative review; M-Cohort = cohort study"). We have also added a new column for "Definition of early detection used in study" and a free-text field for "Operational definition of the diagnostic window."

Analysis approach for predictive values: We now provide a detailed explanation of how we will calculate implied positive predictive value (PPV). Specifically: (1) For studies reporting sensitivity and specificity, we will calculate PPV using a range of plausible disease prevalence estimates (derived from population statistics for the relevant disease and demographic group); (2) For studies reporting odds ratios or hazard ratios, we will convert these to approximate PPVs using the method described by Moons et al. (2012); (3) Where data are insufficient, we will narratively report the available measures of association and explicitly state that PPV could not be calculated. We will handle different study designs by stratifying the analysis (case-control vs. cohort vs. cross-sectional) and discussing the limitations of each for predictive inference.

3. Quality assessment gaps

You raise an important point that standard risk-of-bias tools may not map perfectly onto transactional data studies. We have added a new subsection to the Methods: "Adaptation of risk-of-bias tools for transactional data studies." Here, we explain how we will apply ROBINS-E and ROBINS-I with domain-specific modifications:

For confounding (a major concern in loyalty card studies), we will assess whether studies adjusted for socioeconomic status, healthcare access, and comorbidities.

For selection bias, we will evaluate whether the transactional data source is representative of the target population (e.g., loyalty card holders may differ from non-holders).

For information bias, we will assess the validity of using purchase data as a proxy for medication consumption (e.g., purchased does not equal ingested).
We will not calculate an overall "quality score" but will present domain-level assessments in a narrative summary and a traffic-light table.

4. Ethics and practicalities

We agree that this is a critical omission. We have added a new subsection: "Handling of ethical and commercial considerations." This specifies:

Ethical approval of primary studies: We will note whether studies report obtaining ethics approval; where approval is absent, we will flag this in the synthesis and discuss its implications.

Privacy and data governance: We will extract any reported GDPR compliance, data anonymization methods, and patient consent procedures. Studies using transactional data without explicit consent will be highlighted as a potential limitation.

Commercial conflicts of interest: We will systematically record funding sources (e.g., retail companies, loyalty card providers, pharmaceutical firms) and any author affiliations with commercial entities. This information will be presented in a separate table and discussed in relation to potential bias in outcome reporting.

Response to Minor Points

"Event horizon": We have replaced this term with the clearer phrase "time window between data collection and diagnosis (the ‘lookback period’)" throughout the manuscript.

Timeline: We have added a Project Timeline table to the Methods section, showing: Phase 1 completion (Month 3), Phase 2 screening (Months 4–8), data extraction (Months 9–10), analysis (Months 11–12), and manuscript submission (Month 13).

Terminology clarity: We have reviewed the entire manuscript to ensure consistent use of terms. Notably, we now distinguish clearly between "transactional data" (raw purchase records), "loyalty card data" (a subset linked to individual consumers), and "OTC medication sales data" (aggregated or anonymized pharmacy data). A glossary has been added as an appendix.

Kindest Regards,

Kurdo Araz
Dear Dr. Townsend,

Re: Response to Review of Protocol: "Protocol of a Scoping Review on the Use of Transactional Data for Early Diagnosis (TRADED-ScR)" [version 1; peer review: 1 approved with reservations]

Thank you very much for your thoughtful and detailed review of our protocol. We greatly appreciate your recognition of the study’s strengths, including its clear rationale and sensible methodological framework. Your constructive feedback on the methodological details, data handling, quality assessment, and ethical considerations is invaluable, and we have carefully revised the protocol to address each concern.
Below, we provide a point-by-point response to your comments. All corresponding changes have been made in the revised manuscript (marked in yellow highlight for your convenience).

1. Missing methodological details

Search strategy example: You rightly note that an example search strategy should be included in the main paper. We have now added Appendix A (presented within the manuscript) showing the full MEDLINE search strategy as a representative example. The complete strategies for all databases remain available on OSF.

AI screening: We have expanded this section to specify that we will use the Research Screener tool (Chai et al., 2021 – already cited) or Rayyan’s built-in AI functionality. We now explicitly describe the validation process: (1) The AI will be run on the Phase 1 results (where we have human reviewer decisions); (2) We will calculate sensitivity (proportion of included articles correctly identified) and specificity; (3) Only if sensitivity ≥95% will we rely on AI-assisted screening for Phase 2; (4) Two human reviewers will still verify all AI-excluded articles flagged as potentially relevant. Safeguards, including regular calibration checks and resolution of disagreements by a third reviewer, are now described.

Screening criteria (Table 1): We have substantially expanded Table 1. It now includes explicit operational definitions for "transactional data" (e.g., loyalty card records, OTC medication sales, retail purchase histories, supermarket scanner data, pharmacy transaction logs) and "early detection" (diagnosis at a stage preceding typical clinical presentation, including pre-symptomatic or early-symptomatic phases). We have also added a fourth step to handle borderline cases (e.g., studies using prescription data only or those that do not link purchases to an incident diagnosis), with clear inclusion/exclusion rules.

2. Data handling issues

Extraction form (Table 2): We apologize for the lack of clarity. Table 2 has been completely revised. Codes are now explained in a legend directly beneath the table (e.g., "P-Journal = peer-reviewed journal article; M-Review = systematic/scoping/narrative review; M-Cohort = cohort study"). We have also added a new column for "Definition of early detection used in study" and a free-text field for "Operational definition of the diagnostic window."

Analysis approach for predictive values: We now provide a detailed explanation of how we will calculate implied positive predictive value (PPV). Specifically: (1) For studies reporting sensitivity and specificity, we will calculate PPV using a range of plausible disease prevalence estimates (derived from population statistics for the relevant disease and demographic group); (2) For studies reporting odds ratios or hazard ratios, we will convert these to approximate PPVs using the method described by Moons et al. (2012); (3) Where data are insufficient, we will narratively report the available measures of association and explicitly state that PPV could not be calculated. We will handle different study designs by stratifying the analysis (case-control vs. cohort vs. cross-sectional) and discussing the limitations of each for predictive inference.

3. Quality assessment gaps

You raise an important point that standard risk-of-bias tools may not map perfectly onto transactional data studies. We have added a new subsection to the Methods: "Adaptation of risk-of-bias tools for transactional data studies." Here, we explain how we will apply ROBINS-E and ROBINS-I with domain-specific modifications:

For confounding (a major concern in loyalty card studies), we will assess whether studies adjusted for socioeconomic status, healthcare access, and comorbidities.

For selection bias, we will evaluate whether the transactional data source is representative of the target population (e.g., loyalty card holders may differ from non-holders).

For information bias, we will assess the validity of using purchase data as a proxy for medication consumption (e.g., purchased does not equal ingested).
We will not calculate an overall "quality score" but will present domain-level assessments in a narrative summary and a traffic-light table.

4. Ethics and practicalities

We agree that this is a critical omission. We have added a new subsection: "Handling of ethical and commercial considerations." This specifies:

Ethical approval of primary studies: We will note whether studies report obtaining ethics approval; where approval is absent, we will flag this in the synthesis and discuss its implications.

Privacy and data governance: We will extract any reported GDPR compliance, data anonymization methods, and patient consent procedures. Studies using transactional data without explicit consent will be highlighted as a potential limitation.

Commercial conflicts of interest: We will systematically record funding sources (e.g., retail companies, loyalty card providers, pharmaceutical firms) and any author affiliations with commercial entities. This information will be presented in a separate table and discussed in relation to potential bias in outcome reporting.

Response to Minor Points

"Event horizon": We have replaced this term with the clearer phrase "time window between data collection and diagnosis (the ‘lookback period’)" throughout the manuscript.

Timeline: We have added a Project Timeline table to the Methods section, showing: Phase 1 completion (Month 3), Phase 2 screening (Months 4–8), data extraction (Months 9–10), analysis (Months 11–12), and manuscript submission (Month 13).

Terminology clarity: We have reviewed the entire manuscript to ensure consistent use of terms. Notably, we now distinguish clearly between "transactional data" (raw purchase records), "loyalty card data" (a subset linked to individual consumers), and "OTC medication sales data" (aggregated or anonymized pharmacy data). A glossary has been added as an appendix.

Kindest Regards,

Kurdo Araz
Competing Interests: NA Close
Report a concern
Respond or Comment

COMMENTS ON THIS REPORT

Author Response 05 May 2026

Kurdo Araz, Royal College of Surgeons in Ireland Department of Medicine, County Dublin, Ireland

05 May 2026

Author Response
Dear Dr. Townsend,

Re: Response to Review of Protocol: "Protocol of a Scoping Review on the Use of Transactional Data for Early Diagnosis (TRADED-ScR)" [version 1; peer review: 1 ... Continue reading
Dear Dr. Townsend,

Re: Response to Review of Protocol: "Protocol of a Scoping Review on the Use of Transactional Data for Early Diagnosis (TRADED-ScR)" [version 1; peer review: 1 approved with reservations]

Thank you very much for your thoughtful and detailed review of our protocol. We greatly appreciate your recognition of the study’s strengths, including its clear rationale and sensible methodological framework. Your constructive feedback on the methodological details, data handling, quality assessment, and ethical considerations is invaluable, and we have carefully revised the protocol to address each concern.
Below, we provide a point-by-point response to your comments. All corresponding changes have been made in the revised manuscript (marked in yellow highlight for your convenience).

1. Missing methodological details

Search strategy example: You rightly note that an example search strategy should be included in the main paper. We have now added Appendix A (presented within the manuscript) showing the full MEDLINE search strategy as a representative example. The complete strategies for all databases remain available on OSF.

AI screening: We have expanded this section to specify that we will use the Research Screener tool (Chai et al., 2021 – already cited) or Rayyan’s built-in AI functionality. We now explicitly describe the validation process: (1) The AI will be run on the Phase 1 results (where we have human reviewer decisions); (2) We will calculate sensitivity (proportion of included articles correctly identified) and specificity; (3) Only if sensitivity ≥95% will we rely on AI-assisted screening for Phase 2; (4) Two human reviewers will still verify all AI-excluded articles flagged as potentially relevant. Safeguards, including regular calibration checks and resolution of disagreements by a third reviewer, are now described.

Screening criteria (Table 1): We have substantially expanded Table 1. It now includes explicit operational definitions for "transactional data" (e.g., loyalty card records, OTC medication sales, retail purchase histories, supermarket scanner data, pharmacy transaction logs) and "early detection" (diagnosis at a stage preceding typical clinical presentation, including pre-symptomatic or early-symptomatic phases). We have also added a fourth step to handle borderline cases (e.g., studies using prescription data only or those that do not link purchases to an incident diagnosis), with clear inclusion/exclusion rules.

2. Data handling issues

Extraction form (Table 2): We apologize for the lack of clarity. Table 2 has been completely revised. Codes are now explained in a legend directly beneath the table (e.g., "P-Journal = peer-reviewed journal article; M-Review = systematic/scoping/narrative review; M-Cohort = cohort study"). We have also added a new column for "Definition of early detection used in study" and a free-text field for "Operational definition of the diagnostic window."

Analysis approach for predictive values: We now provide a detailed explanation of how we will calculate implied positive predictive value (PPV). Specifically: (1) For studies reporting sensitivity and specificity, we will calculate PPV using a range of plausible disease prevalence estimates (derived from population statistics for the relevant disease and demographic group); (2) For studies reporting odds ratios or hazard ratios, we will convert these to approximate PPVs using the method described by Moons et al. (2012); (3) Where data are insufficient, we will narratively report the available measures of association and explicitly state that PPV could not be calculated. We will handle different study designs by stratifying the analysis (case-control vs. cohort vs. cross-sectional) and discussing the limitations of each for predictive inference.

3. Quality assessment gaps

You raise an important point that standard risk-of-bias tools may not map perfectly onto transactional data studies. We have added a new subsection to the Methods: "Adaptation of risk-of-bias tools for transactional data studies." Here, we explain how we will apply ROBINS-E and ROBINS-I with domain-specific modifications:

For confounding (a major concern in loyalty card studies), we will assess whether studies adjusted for socioeconomic status, healthcare access, and comorbidities.

For selection bias, we will evaluate whether the transactional data source is representative of the target population (e.g., loyalty card holders may differ from non-holders).

For information bias, we will assess the validity of using purchase data as a proxy for medication consumption (e.g., purchased does not equal ingested).
We will not calculate an overall "quality score" but will present domain-level assessments in a narrative summary and a traffic-light table.

4. Ethics and practicalities

We agree that this is a critical omission. We have added a new subsection: "Handling of ethical and commercial considerations." This specifies:

Ethical approval of primary studies: We will note whether studies report obtaining ethics approval; where approval is absent, we will flag this in the synthesis and discuss its implications.

Privacy and data governance: We will extract any reported GDPR compliance, data anonymization methods, and patient consent procedures. Studies using transactional data without explicit consent will be highlighted as a potential limitation.

Commercial conflicts of interest: We will systematically record funding sources (e.g., retail companies, loyalty card providers, pharmaceutical firms) and any author affiliations with commercial entities. This information will be presented in a separate table and discussed in relation to potential bias in outcome reporting.

Response to Minor Points

"Event horizon": We have replaced this term with the clearer phrase "time window between data collection and diagnosis (the ‘lookback period’)" throughout the manuscript.

Timeline: We have added a Project Timeline table to the Methods section, showing: Phase 1 completion (Month 3), Phase 2 screening (Months 4–8), data extraction (Months 9–10), analysis (Months 11–12), and manuscript submission (Month 13).

Terminology clarity: We have reviewed the entire manuscript to ensure consistent use of terms. Notably, we now distinguish clearly between "transactional data" (raw purchase records), "loyalty card data" (a subset linked to individual consumers), and "OTC medication sales data" (aggregated or anonymized pharmacy data). A glossary has been added as an appendix.

Kindest Regards,

Kurdo Araz
Dear Dr. Townsend,

Re: Response to Review of Protocol: "Protocol of a Scoping Review on the Use of Transactional Data for Early Diagnosis (TRADED-ScR)" [version 1; peer review: 1 approved with reservations]

Thank you very much for your thoughtful and detailed review of our protocol. We greatly appreciate your recognition of the study’s strengths, including its clear rationale and sensible methodological framework. Your constructive feedback on the methodological details, data handling, quality assessment, and ethical considerations is invaluable, and we have carefully revised the protocol to address each concern.
Below, we provide a point-by-point response to your comments. All corresponding changes have been made in the revised manuscript (marked in yellow highlight for your convenience).

1. Missing methodological details

Search strategy example: You rightly note that an example search strategy should be included in the main paper. We have now added Appendix A (presented within the manuscript) showing the full MEDLINE search strategy as a representative example. The complete strategies for all databases remain available on OSF.

AI screening: We have expanded this section to specify that we will use the Research Screener tool (Chai et al., 2021 – already cited) or Rayyan’s built-in AI functionality. We now explicitly describe the validation process: (1) The AI will be run on the Phase 1 results (where we have human reviewer decisions); (2) We will calculate sensitivity (proportion of included articles correctly identified) and specificity; (3) Only if sensitivity ≥95% will we rely on AI-assisted screening for Phase 2; (4) Two human reviewers will still verify all AI-excluded articles flagged as potentially relevant. Safeguards, including regular calibration checks and resolution of disagreements by a third reviewer, are now described.

Screening criteria (Table 1): We have substantially expanded Table 1. It now includes explicit operational definitions for "transactional data" (e.g., loyalty card records, OTC medication sales, retail purchase histories, supermarket scanner data, pharmacy transaction logs) and "early detection" (diagnosis at a stage preceding typical clinical presentation, including pre-symptomatic or early-symptomatic phases). We have also added a fourth step to handle borderline cases (e.g., studies using prescription data only or those that do not link purchases to an incident diagnosis), with clear inclusion/exclusion rules.

2. Data handling issues

Extraction form (Table 2): We apologize for the lack of clarity. Table 2 has been completely revised. Codes are now explained in a legend directly beneath the table (e.g., "P-Journal = peer-reviewed journal article; M-Review = systematic/scoping/narrative review; M-Cohort = cohort study"). We have also added a new column for "Definition of early detection used in study" and a free-text field for "Operational definition of the diagnostic window."

Analysis approach for predictive values: We now provide a detailed explanation of how we will calculate implied positive predictive value (PPV). Specifically: (1) For studies reporting sensitivity and specificity, we will calculate PPV using a range of plausible disease prevalence estimates (derived from population statistics for the relevant disease and demographic group); (2) For studies reporting odds ratios or hazard ratios, we will convert these to approximate PPVs using the method described by Moons et al. (2012); (3) Where data are insufficient, we will narratively report the available measures of association and explicitly state that PPV could not be calculated. We will handle different study designs by stratifying the analysis (case-control vs. cohort vs. cross-sectional) and discussing the limitations of each for predictive inference.

3. Quality assessment gaps

You raise an important point that standard risk-of-bias tools may not map perfectly onto transactional data studies. We have added a new subsection to the Methods: "Adaptation of risk-of-bias tools for transactional data studies." Here, we explain how we will apply ROBINS-E and ROBINS-I with domain-specific modifications:

For confounding (a major concern in loyalty card studies), we will assess whether studies adjusted for socioeconomic status, healthcare access, and comorbidities.

For selection bias, we will evaluate whether the transactional data source is representative of the target population (e.g., loyalty card holders may differ from non-holders).

For information bias, we will assess the validity of using purchase data as a proxy for medication consumption (e.g., purchased does not equal ingested).
We will not calculate an overall "quality score" but will present domain-level assessments in a narrative summary and a traffic-light table.

4. Ethics and practicalities

We agree that this is a critical omission. We have added a new subsection: "Handling of ethical and commercial considerations." This specifies:

Ethical approval of primary studies: We will note whether studies report obtaining ethics approval; where approval is absent, we will flag this in the synthesis and discuss its implications.

Privacy and data governance: We will extract any reported GDPR compliance, data anonymization methods, and patient consent procedures. Studies using transactional data without explicit consent will be highlighted as a potential limitation.

Commercial conflicts of interest: We will systematically record funding sources (e.g., retail companies, loyalty card providers, pharmaceutical firms) and any author affiliations with commercial entities. This information will be presented in a separate table and discussed in relation to potential bias in outcome reporting.

Response to Minor Points

"Event horizon": We have replaced this term with the clearer phrase "time window between data collection and diagnosis (the ‘lookback period’)" throughout the manuscript.

Timeline: We have added a Project Timeline table to the Methods section, showing: Phase 1 completion (Month 3), Phase 2 screening (Months 4–8), data extraction (Months 9–10), analysis (Months 11–12), and manuscript submission (Month 13).

Terminology clarity: We have reviewed the entire manuscript to ensure consistent use of terms. Notably, we now distinguish clearly between "transactional data" (raw purchase records), "loyalty card data" (a subset linked to individual consumers), and "OTC medication sales data" (aggregated or anonymized pharmacy data). A glossary has been added as an appendix.

Kindest Regards,

Kurdo Araz
Competing Interests: NA Close
Report a concern

Comments on this article Comments (0)

Version 2

VERSION 2 PUBLISHED 10 Jan 2025

Open Peer Review

Reviewer Status

Reviewer Reports

	Invited Reviewers
	1	2
Version 2 (revision) 05 May 26		read
Version 1 10 Jan 25	read

Paul A. Townsend, University of Stirling, Stirling, UK
Adekola George Adepoju, Essentrify, Pittsburgh, USA

Daniel Adepoju, Essentrify, Pittsburgh, USA

Daniel Cheruiyot, Essentrify, Pittsburgh, USA

Comments on this article

All Comments(0)

Add a comment

Back to all reports

Reviewer Report

9 Views

16 Jul 2026 | for Version 2

Adekola George Adepoju, Public Sector Informatics, Essentrify, Pittsburgh, Pennsylvania, USA

Daniel Adepoju, Product Architecture, Essentrify, Pittsburgh, Pennsylvania, USA

Daniel Cheruiyot, Evidence Lab, Essentrify, Pittsburgh, Pennsylvania, USA

9 Views Cite this report Responses(0)

Approved With Reservations

whether AI will only prioritise records or will make exclusion decisions;
what the calibration checks every 500 abstracts will involve;
what action will be taken if an eligible paper is found in the 10% sample of AI-excluded records; and
how performance will be checked when moving from cancer literature to literature covering all diseases.

State whether data extraction will be completed by two reviewers independently or checked by a second reviewer.
Explain how duplicate records and multiple reports from the same study will be handled.
Clearly state any language and publication-date restrictions. The Discussion mentions possible exclusion of non-English studies, but this is not specified in the Methods.
Distinguish between ethical approval not obtained, not required and not reported. A lack of reporting does not necessarily mean that approval was not obtained.
Change “use transactional data” to “use of transactional data” in Objective 1.
Change “a software that designed” to “software designed.”
Correct the unmatched bracket in the sentence defining the lookback period.
Replace “event horizon” with “lookback period” in Objective 2.
Use “Appendix A” or “Appendix 1” consistently.
Cite the specific PRISMA-P and PRISMA-ScR guideline papers. Reference 18 currently cites the general PRISMA 2020 statement.

Overall, the protocol addresses a relevant and emerging area. Addressing these points would improve the clarity, reproducibility and scientific strength of the proposed review.

Is the rationale for, and objectives of, the study clearly described?

Yes
Is the study design appropriate for the research question?

Yes
Are sufficient details of the methods provided to allow replication by others?

Partly
Are the datasets clearly presented in a useable and accessible format?

Not applicable

Competing Interests

No competing interests were disclosed.

Reviewer Expertise

Clinical and public health informatics, health data analytics, early disease detection and screening, epidemiology, and evidence synthesis. I am able to assess the health-informatics relevance, study design, data methods, ethical use of health data, and public-health implications of this protocol.

Respond to this report

Responses (0)

Back to all reports

Reviewer Report

32 Views

30 Jun 2025 | for Version 1

Paul A. Townsend, University of Stirling, Stirling, UK

32 Views Cite this report Responses(1)

Approved With Reservations

The term "event horizon" is confusing in this context
No timeline provided for completing the work
Some terminology could be clearer throughout

Essential changes needed:

Include search strategy examples in the main paper
Expand screening criteria with clearer definitions
Clarify AI screening plans with proper validation details
Improve data extraction forms with better coding systems
Explain analysis methods more clearly, especially for calculating predictive values

Would be helpful:

Add a realistic timeline
Strengthen the quality assessment section
Address ethical considerations more thoroughly

Is the rationale for, and objectives of, the study clearly described?

Yes
Is the study design appropriate for the research question?

Yes
Are sufficient details of the methods provided to allow replication by others?

Partly
Are the datasets clearly presented in a useable and accessible format?

Not applicable

Competing Interests

No competing interests were disclosed.

Reviewer Expertise

cancer early detection, machine learning, multiple omics and mass spec

Respond to this report

Responses (1)

Author Response

05 May 2026

Kurdo Araz, Royal College of Surgeons in Ireland Department of Medicine, County Dublin, Ireland

Dear Dr. Townsend,

Re: Response to Review of Protocol: "Protocol of a Scoping Review on the Use of Transactional Data for Early Diagnosis (TRADED-ScR)" [version 1; peer review: 1 approved with reservations]

Thank you very much for your thoughtful and detailed review of our protocol. We greatly appreciate your recognition of the study’s strengths, including its clear rationale and sensible methodological framework. Your constructive feedback on the methodological details, data handling, quality assessment, and ethical considerations is invaluable, and we have carefully revised the protocol to address each concern.
Below, we provide a point-by-point response to your comments. All corresponding changes have been made in the revised manuscript (marked in yellow highlight for your convenience).

1. Missing methodological details

Search strategy example: You rightly note that an example search strategy should be included in the main paper. We have now added Appendix A (presented within the manuscript) showing the full MEDLINE search strategy as a representative example. The complete strategies for all databases remain available on OSF.
AI screening: We have expanded this section to specify that we will use the Research Screener tool (Chai et al., 2021 – already cited) or Rayyan’s built-in AI functionality. We now explicitly describe the validation process: (1) The AI will be run on the Phase 1 results (where we have human reviewer decisions); (2) We will calculate sensitivity (proportion of included articles correctly identified) and specificity; (3) Only if sensitivity ≥95% will we rely on AI-assisted screening for Phase 2; (4) Two human reviewers will still verify all AI-excluded articles flagged as potentially relevant. Safeguards, including regular calibration checks and resolution of disagreements by a third reviewer, are now described.
Screening criteria (Table 1): We have substantially expanded Table 1. It now includes explicit operational definitions for "transactional data" (e.g., loyalty card records, OTC medication sales, retail purchase histories, supermarket scanner data, pharmacy transaction logs) and "early detection" (diagnosis at a stage preceding typical clinical presentation, including pre-symptomatic or early-symptomatic phases). We have also added a fourth step to handle borderline cases (e.g., studies using prescription data only or those that do not link purchases to an incident diagnosis), with clear inclusion/exclusion rules.

2. Data handling issues

Extraction form (Table 2): We apologize for the lack of clarity. Table 2 has been completely revised. Codes are now explained in a legend directly beneath the table (e.g., "P-Journal = peer-reviewed journal article; M-Review = systematic/scoping/narrative review; M-Cohort = cohort study"). We have also added a new column for "Definition of early detection used in study" and a free-text field for "Operational definition of the diagnostic window."
Analysis approach for predictive values: We now provide a detailed explanation of how we will calculate implied positive predictive value (PPV). Specifically: (1) For studies reporting sensitivity and specificity, we will calculate PPV using a range of plausible disease prevalence estimates (derived from population statistics for the relevant disease and demographic group); (2) For studies reporting odds ratios or hazard ratios, we will convert these to approximate PPVs using the method described by Moons et al. (2012); (3) Where data are insufficient, we will narratively report the available measures of association and explicitly state that PPV could not be calculated. We will handle different study designs by stratifying the analysis (case-control vs. cohort vs. cross-sectional) and discussing the limitations of each for predictive inference.

3. Quality assessment gaps

You raise an important point that standard risk-of-bias tools may not map perfectly onto transactional data studies. We have added a new subsection to the Methods: "Adaptation of risk-of-bias tools for transactional data studies." Here, we explain how we will apply ROBINS-E and ROBINS-I with domain-specific modifications:

For confounding (a major concern in loyalty card studies), we will assess whether studies adjusted for socioeconomic status, healthcare access, and comorbidities.
For selection bias, we will evaluate whether the transactional data source is representative of the target population (e.g., loyalty card holders may differ from non-holders).
For information bias, we will assess the validity of using purchase data as a proxy for medication consumption (e.g., purchased does not equal ingested).
We will not calculate an overall "quality score" but will present domain-level assessments in a narrative summary and a traffic-light table.

4. Ethics and practicalities

We agree that this is a critical omission. We have added a new subsection: "Handling of ethical and commercial considerations." This specifies:

Ethical approval of primary studies: We will note whether studies report obtaining ethics approval; where approval is absent, we will flag this in the synthesis and discuss its implications.
Privacy and data governance: We will extract any reported GDPR compliance, data anonymization methods, and patient consent procedures. Studies using transactional data without explicit consent will be highlighted as a potential limitation.
Commercial conflicts of interest: We will systematically record funding sources (e.g., retail companies, loyalty card providers, pharmaceutical firms) and any author affiliations with commercial entities. This information will be presented in a separate table and discussed in relation to potential bias in outcome reporting.

Response to Minor Points

"Event horizon": We have replaced this term with the clearer phrase "time window between data collection and diagnosis (the ‘lookback period’)" throughout the manuscript.
Timeline: We have added a Project Timeline table to the Methods section, showing: Phase 1 completion (Month 3), Phase 2 screening (Months 4–8), data extraction (Months 9–10), analysis (Months 11–12), and manuscript submission (Month 13).
Terminology clarity: We have reviewed the entire manuscript to ensure consistent use of terms. Notably, we now distinguish clearly between "transactional data" (raw purchase records), "loyalty card data" (a subset linked to individual consumers), and "OTC medication sales data" (aggregated or anonymized pharmacy data). A glossary has been added as an appendix.

Kindest Regards,

Kurdo Araz

View more View less

Competing Interests

Alongside their report, reviewers assign a status to the article:

Approved - the paper is scientifically sound in its current form and only minor, if any, improvements are suggested

Approved with reservations - A number of small changes, sometimes more significant revisions are required to address specific details and improve the papers academic merit.

Not approved - fundamental flaws in the paper seriously undermine the findings and conclusions

[1] 1. Yuan X, Chen S, Sun C, et al.:A novel early diagnostic framework for chronic diseases with class imbalance. Sci. Rep. 2022; 12(1): 8614. PubMed Abstract | Publisher Full Text | Free Full Text

[2] 2. Whitaker K:Earlier diagnosis: the importance of cancer symptoms. Lancet Oncol. 2020; 21(1): 6–8. PubMed Abstract | Publisher Full Text

[3] 3. Varghese SE, El Otol RHM, Al Olama FS, et al.:The importance of early detection of genetic diseases. Dubai Med. J. 2021; 4(2): 133–141. Publisher Full Text

[4] 4. Cohn JN, Hoke L, Whitwam W, et al.:Screening for early detection of cardiovascular disease in asymptomatic individuals. Am. Heart J. 2003; 146(4): 679–685. PubMed Abstract | Publisher Full Text

[5] 5. McClellan WM, Knight DF, Karp H, et al.:Early detection and treatment of renal disease in hospitalized diabetic and hypertensive patients: important differences between practise and published guidelines. Am. J. Kidney Dis. 1997; 29(3): 368–375. PubMed Abstract | Publisher Full Text

[6] 6. Whitaker KL, Winstanley K, Macleod U, et al.:Low cancer suspicion following experience of a cancer ‘warning sign’. Eur. J. Cancer. 2015; 51(16): 2473–2479. PubMed Abstract | Publisher Full Text | Free Full Text

[7] 7. Davies A, Green MA, Singleton AD: Using machine learning to investigate self-medication purchasing in England via high street retailer loyalty card data.Rao P, editor. PLoS One. 2018;13(11): e0207523. PubMed Abstract | Publisher Full Text | Free Full Text

[8] 8. Brewer HR, Chadeau-Hyam M, Johnson E, et al.:Cancer Loyalty Card Study (CLOCS): feasibility outcomes for an observational case-control study focusing on the patient interval in ovarian cancer. BMJ Open. 2023; 13(6): e066022. Publisher Full Text

[9] 9. Brewer HR, Hirst Y, Chadeau-Hyam M, et al.:Association between purchase of over-the-counter medications and ovarian cancer diagnosis in the Cancer Loyalty Card Study (CLOCS): Observational Case-Control Study. JMIR Public Health Surveill. 2023; 9: e41762. PubMed Abstract | Publisher Full Text | Free Full Text

[10] 10. Das D, Metzger K, Heffernan R, et al.:Monitoring over-the-counter medication sales for early detection of disease outbreaks—New York City. MMWR Suppl. 2005; 54: 41–46. PubMed Abstract

[11] 11. Qin Q, Cheng L, Tan L, et al.:Over-the-counter medication sales surveillance for early detection of respiratory epidemics in rural China. Online J. Public Health Inform. 2014; 6(1): e61404. Publisher Full Text

[12] 12. Dapkevičiūtė A, Šapoka V, Martynova E, et al.:Time from symptom onset to diagnosis and treatment among haematological malignancies: influencing factors and associated negative outcomes. Medicina (Kaunas). 2019; 55: 31163661. Publisher Full Text

[13] 13. Dulal S, Paudel BD, Wood LA, et al.:Reliance on self-medication increase delays in diagnosis and management of GI cancers: results from Nepal. JCO Glob. Oncologia. 2020; 6: 1258–1263. PubMed Abstract | Publisher Full Text | Free Full Text

[14] 14. Goff BA, Mandel L, Muntz HG, et al.:Ovarian carcinoma diagnosis: results of a national ovarian cancer survey. Cancer. 2000; 89(10): 2068–2075. PubMed Abstract | Publisher Full Text

[15] 15. Flanagan JM, Skrobanski H, Shi X, et al.:Self-care behaviors of ovarian cancer patients before their diagnosis: proof-of-concept study. JMIR Cancer. 2019; 5(1): e10447. PubMed Abstract | Publisher Full Text | Free Full Text

[16] 16. Dolan EH, Goulding J, Tata LJ, et al.:Using shopping data to improve the diagnosis of ovarian cancer: computational analysis of a web-based survey. JMIR Cancer. 2023; 9: e37141. PubMed Abstract | Publisher Full Text | Free Full Text

[17] 17. Levac D, Colquhoun H, O’Brien KK:Scoping studies: advancing the methodology. Implement. Sci. 2010; 5(1): 69. PubMed Abstract | Publisher Full Text | Free Full Text

[18] 18. Page MJ, McKenzie JE, Bossuyt PM, et al.:The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ. 2021; 372: n71. PubMed Abstract | Publisher Full Text | Free Full Text

[19] 19. Foster ED, Deardorff A:Open Science Framework (OSF). JMLA. 2017; 105(2). Publisher Full Text

[20] 20. Jacob BM, Araz K: Protocol of a Scoping Review on the Use of Transactional Data for Early Diagnosis (TRADED-ScR).OSF Registries;2024; [cited 2024 Oct 21]. Reference Source

[21] 21. Ouzzani M, Hammady H, Fedorowicz Z, et al.:Rayyan—a web and mobile app for systematic reviews. Syst. Rev. 2016; 5(1): 210. PubMed Abstract | Publisher Full Text | Free Full Text

[22] 22. Chai KEK, Lines RLJ, Gucciardi DF, et al.:Research Screener: a machine learning tool to semi-automate abstract screening for systematic reviews. Syst. Rev. 2021; 10(1): 93. PubMed Abstract | Publisher Full Text | Free Full Text

[23] 23. Bero L, Chartres N, Diong J, et al.:The Risk of Bias in Observational Studies of Exposures (ROBINS-E) tool: concerns arising from application to observational studies of exposures. Syst. Rev. 2018; 7(1): 242. PubMed Abstract | Publisher Full Text | Free Full Text

[24] 24. Sterne JA, Hernán MA, Reeves BC, et al.:ROBINS-I: a tool for assessing risk of bias in non-randomised studies of interventions. BMJ. 2016; 355: i4919. PubMed Abstract | Publisher Full Text | Free Full Text

Protocol of a Scoping Review on the Use of Transactional Data for Early Diagnosis (TRADED-ScR)

Abstract

Background

Aim

Methods

Implications

Keywords

Revised Amendments from Version 1

Introduction

Aims and objectives

Methods

Reporting and registration

Eligibility criteria

Information sources

Search strategy

Selection of sources of evidence

Table 1. Screening algorithm.

Data collection and data management

Table 2. Data charting pro-forma.

Data items

Critical appraisal

Data synthesis

Project timeline

Discussion

Summary

Strengths and limitations

Implications

Conclusion

Ethics and consent

Data availability

Extended data

Reporting guidelines

References

Comments on this article Comments (0)

Open Peer Review

Comments on this article Comments (0)

Open Peer Review

Reviewer Status

Reviewer Reports

Comments on this article

Competing Interests Policy

Stay Updated

Are you a HRB-funded researcher?

Thank you!