Skip to content

Methods and limitations

Evidence boundary: This dossier explains the evidence and reasoning behind an Analysis article. It does not establish that DMICP caused harm, contains the civilian defects described, is unsafe, or fails DCB0129/DCB0160. Evidence cut-off: 15 July 2026.

Publication design

The manuscript is an Analysis, not Original Research or a systematic review. It develops an evidence-supported argument from selected peer-reviewed literature, official Defence and NHS publications, bounded CQC inspection observations and seven illustrative PFD mechanisms. It does not present the project's targeted 71-report corpus, mechanism counts or cross-case coding as new journal results.

That decision avoids implying a level of search closure, independent screening, causal classification and reproducibility that the present targeted review does not yet support. The broader corpus, screening decisions and extraction trail remain available through SRC-013, the search protocol, review counts and method.

Source-selection logic

The article uses four evidence layers:

  1. Conceptual and empirical literature to establish sociotechnical and interoperability mechanisms.
  2. Official Defence, NHS, parliamentary, procurement and inspection sources to describe the bounded UK Defence context.
  3. Selected PFD reports to provide concrete prospective failure scenarios.
  4. DCB0129 and DCB0160 as an NHS clinical-risk-management benchmark for lifecycle assurance.

The project follows the authority chain primary source → report-specific analysis → controlled CLM- claim → mechanism → candidate REQ- requirement → designed AST- test. The source types and reliability rules govern how each layer may be used.

Method guidance informed structure and reporting without changing the review's status: JBI evidence-synthesis structure (SRC-030), PRISMA reporting principles (SRC-031), NHS thematic-review guidance (SRC-032) and Judiciary PFD procedural guidance (SRC-033). The work is not labelled JBI-conformant or a PRISMA systematic review.

Public and internal knowledge boundary

The manuscript's factual support is public and independently identifiable. SRC-005 records author-supplied operational context and SRC-142 records a sibling-project source-routing audit. They guide evidence requests and wording boundaries but are not used as independent authorities in the journal article.

Raw sibling-project paths, working extracts and any operationally sensitive detail remain outside the rendered dossier. The curated Defence and DMICP baseline exposes only the status, permitted formulation and outstanding evidence requirement.

Causal and transferability limits

  • The manuscript does not estimate incidence, prevalence, trend or comparative risk.
  • PFDs describe individual deaths and future-risk concerns; some do not establish digital causation.
  • Recipient responses remain distinct from coronial findings and do not prove remediation effectiveness.
  • CQC findings are bounded by site, inspection date and the invited DMSR programme; they are not a DCB assessment.
  • External civilian mechanisms and international military literature support plausibility, not a DMICP defect attribution.
  • Historical platform, hosting and access material is not treated as the exact current architecture.
  • CORTISONE programme and procurement statements establish intent or status, not delivered capability or safety acceptance.
  • DCB0129/DCB0160 are used as an NHS benchmark. Their formal Defence applicability and DMICP's present clinical-safety regime are unknown, not absent or non-conformant.
  • Candidate Defence requirements are unvalidated and designed assurance tests are unexecuted.

Important evidence gaps

The main limitation is the absence of a controlled contemporary DMICP architecture, configuration, interface and clinical-safety evidence pack. The outstanding questions are maintained under VAL-019 and summarised in the Defence evidence baseline.

The article's seven PFD cases were selected for explanatory coverage of the principal seams, not through a new independent article-specific systematic search. The wider targeted corpus remains useful for hazard discovery but lacks a complete denominator for claims about digital-PFD prevalence or trend.

AI use and human accountability

OpenAI Codex assisted with literature triage, source-linked drafting, document production and language refinement. It is not an author. BMJ's AI policy is registered as SRC-161, and the manuscript contains a transparent use statement.

Human authors must verify every claim and reference, decide the argument and wording, revise the manuscript, obtain all required clearances and accept responsibility for the submitted text.

Update protocol before submission

  1. Re-run literature and programme-status searches to the submission date.
  2. Recheck every time-sensitive official source and publication status.
  3. Recheck the journal and ScholarOne requirements against SRC-159 to SRC-161.
  4. Update the source register and argument map before changing a factual proposition.
  5. Rebuild and visually inspect the Word manuscript.
  6. Record human factual review, authorship approval and MOD/institutional clearance in the submission checklist.