Clinical Data PhD: Predicting Hospital-Acquired Acute Kidney Injury from Hospital Records

Clinical Data PhD: Predicting Hospital-Acquired Acute Kidney Injury from Hospital Records

Quick Decision

This PhD would build an admission-time HA-AKI risk model from NHS electronic records, then test how well it transports across hospital sites. The advertised work covers outcome and cohort definition, reproducible data preparation, statistical and machine-learning comparisons, internal and external validation, calibration and decision-curve analysis; an open-source risk tool is a proposed output. It fits applicants with strong quantitative or computational preparation and an interest in clinical prediction. The key dependency is whether the outcome code list and multi-site datasets are ready for the planned validation.

The EPSRC DLA award page lists up to 3.5 years full-time, a current 2026/27 stipend of £21,805, Home tuition and £5,000 total RTSG; the application closes at midday on 8 December 2026. Its project-specific page says references are due by 8 January 2027 and general interviews are likely 15–24 February 2027. The central EPSRC guidance instead gives 15 February 2027 for references and still lists a 2026 interview/outcome calendar despite describing 2027 entry. Confirm these dates with admissions; prepare references by the earlier 8 January date. Apply only through the official project description, and confirm the 2027/28 stipend, international-fee eligibility and data-access plan before relying on the headline package.

Research Problem and Doctoral Work

What Is Unsolved and Why It Matters

AKI can develop during a hospital admission and is associated with prolonged stays and substantial mortality. The 2025 UK Renal Registry report, covering England in 2024, records 710,532 AKI episodes in 612,254 people. Of episodes occurring in hospitalized patients, 44.9% arose after admission; because a further 32.9% of all episodes were recorded in people not hospitalized at the time, hospital-acquired episodes represent roughly 30% of the total, subject to the report's definitions and denominator. The project's stated 30–33% share is therefore broadly plausible, but it is not a cohort-size estimate for this study. NICE recognition criteria use changes in serum creatinine and urine output. The project description instead says HA-AKI outcomes will use “validated medicine code lists”; applicants should ask how those lists will be reconciled with the clinical and biochemical definition, and whether coded diagnoses, laboratory alerts or both establish the outcome.

The project's motivation is also consistent with a broader prediction-model problem: a model that ranks patients well may still be miscalibrated, miss high-risk people, or fail to improve care after deployment. A 2024 external validation of a proprietary US hospital model reported moderate performance and called for further clinical-impact evaluation. Other hospital and ICU models use different populations, AKI severity definitions and prediction horizons. They show why independent testing matters, but their performance cannot be transferred to this proposed UK admission-time model.

Work, Methods, and Potential Outputs

The advertisement describes a connected sequence of work:

1. Curate large NHS electronic health-record datasets into analysis-ready data and build reproducible preprocessing pipelines.
2. Define HA-AKI using the advertised validated medicine code lists. Candidate predictors include demographics, existing comorbidities, treatment history, laboratory biomarkers and prescribed medicines.
3. Predict risk using information available at hospital admission. Compare logistic and penalized regression, tree-based methods, random forests and gradient boosting; use feature-selection strategies to reduce dimensionality and support clinical interpretation.
4. Use cross-validation and bootstrapping for internal validation, then external validation across different NHS sites. Evaluate discrimination, calibration and decision-curve analysis. The project also proposes an accessible open-source individual risk tool, stakeholder engagement to assess workflow usability, reproducible pipelines and TRIPOD reporting.

These are advertised tasks, not confirmed results or a promise of clinical deployment. A potentially important research contribution would be a defensible outcome phenotype and leakage-safe admission-time design, followed by site-level transport tests and calibration or threshold analyses that inform whether a tool adds useful information to existing AKI alerts. That is an evidence-led interpretation, not a stated project guarantee. Before accepting an offer, ask for the AKI case definition, prediction horizon and landmark, handling of AKI present on admission, predictor timestamp windows, missing-data strategy, comparator and intended clinical action. Clarify which NHS sites and years of data are committed, the expected case count and approvals, and whether cross-site validation uses a fully held-out site. Ask how subgroup performance, recalibration and decision thresholds will be assessed, and whether prospective workflow or impact evaluation is in scope.

School, Programme and Field Context

The named degree-awarding institution is the University of Exeter; the project sits in its Department of Health and Community Sciences (HCS), at South Cloisters on St Luke's Campus. Exeter's official page reports QS World University Rankings 2027 at 136 overall and Times Higher Education 2027 at =157 overall. These are university-wide indicators, not rankings of HCS or this studentship. In REF 2021, Exeter's UoA 2 (Public Health, Health Services and Primary Care) profile was 35% four-star and 54% three-star; Clinical Medicine was 45% four-star and 48% three-star. REF profiles describe submitted subject portfolios, not an individual lab or project score.

HCS describes applied health and care research with NHS, public-sector, patient and community partnerships. Its postgraduate research themes include research methods, data science and bioinformatics. PenCHORD provides a nearby applied environment spanning operational research, data science, AI, linked health data, prediction and service stakeholders. These pages indicate relevant institutional capability; they do not confirm that this PhD has access to a particular database or facility. Professor Yinghui Wei is the named Exeter lead and Professor of Data Science/Exeter BRC Academic Lead for Data Science. The PDF also names Nick Kennedy of Royal Devon University Healthcare NHS Foundation Trust as a co-supervisor, which suggests a clinical-provider link but does not establish a joint degree, employment arrangement or a specific data-sharing agreement. The general award page says supervisory teams include at least two Exeter academics, while the project PDF names only one Exeter academic; confirm who the second Exeter academic will be.

The project PDF lists “Health Research; Statistics” as potential programme areas, but the formal doctoral degree title, registration route, training modules and milestones are not identified. HCS advises funded applicants to confirm a lead supervisor and apply through the relevant course; ask admissions which programme code and degree title would appear on the award and how methods, clinical and responsible-data training are organized. The listed duration is up to 3.5 years full-time, with part-time funding pro-rated; the exact start date is not specified beyond the September 2027 intake.

Career context is transferable but not project-specific: the UK National Careers Service data-scientist profile lists a broad £32,000–£83,000 salary range and employers including the NHS, government and technology firms. The 2026 UK Digital and Technologies Jobs Plan estimates around 800,000 roles to fill over a decade across sectors backed by the Industrial Strategy. Neither figure predicts health-data vacancies or the prospects of a particular PhD graduate in 2031.

Supervisor and Lab Assessment

Supervisor Research Directions

Professor Yinghui Wei's official profile describes research in clinical prediction, epidemiological design and analysis, evidence synthesis and complex health data. It identifies work with NHS records, linked electronic health records, national registries and longitudinal cohorts, including renal and hospital-acquired infection topics. That is a strong methods and data-domain match for the advertised project. PenCHORD's applied health analytics and stakeholder-facing work is also relevant to the proposed workflow assessment. The available public evidence does not state the day-to-day division of supervision with Nick Kennedy, the second Exeter academic required by the generic award guidance, or which partner organization holds the proposed dataset.

Recent Important Publications

• In a 2025 OpenSAFELY study of 17.99 million adults, Wei and colleagues examined characteristics associated with a clinically coded long-COVID diagnosis. Only 0.21% had such a code, and the authors note likely under-recording. This is useful precedent for working with large linked EHRs and thinking critically about code-based phenotypes; it is not an AKI prediction study and does not settle this project's outcome definition.
• A 2025 scoping review coauthored by Wei examined 16 studies using mobile-health patient-generated data in outpatient management and clinical decisions. It found potential decision-support value but limited evidence of long-term impact and barriers around accuracy, standardization, workflow integration and professional acceptance. That context supports asking how stakeholder work will test usability; it does not validate the proposed risk tool.
• A 2023 methods paper by Wei and colleagues developed bivariate copula regression for semi-competing risks and applied it to kidney-transplant graft failure and death. It illustrates statistical-methods and renal-data expertise, but it is neither the advertised modelling approach nor an AKI prediction result.

Lab, Resources, and Collaboration Environment

HCS and PenCHORD describe interdisciplinary applied-health research, linked health data and engagement with services. Exeter's NIHR Biomedical Research Centre profile connects Wei's work to clinical data-science capacity. NHS health data are commonly accessed through approved, controlled environments and governance agreements; NHS England's Secure Data Environment is an example of those requirements, not evidence that it is the route for this project. The advertisement does not name the dataset custodian, a secure environment, data-access approval, NHS sites, data years, sample size or timeline. These determine whether external validation is a practical project deliverable or an aspiration dependent on approvals and partner access.

Funding, Application, and Living Cost

The Exeter EPSRC Doctoral Landscape Award listing (reference 6017) describes up to 3.5 years full-time, the current 2026/27 stipend of £21,805 per year, Home tuition and £5,000 total Research Training Support Grant. The award page's summary line labels £21,805 as a “2027/28 rate,” but its detailed terms identify it as the current 2026/27 rate; the 2027/28 amount has not yet been published. UKRI does not cover visa costs, the immigration health surcharge or relocation. Eligible international full-award holders may receive an Exeter waiver for the international-fee difference, but the scheme limits international places to up to 30% of the cohort and residence criteria are complex. Confirm eligibility and the exact fee waiver in writing.

The project-specific award page sets the application deadline at midday Tuesday 8 December 2026, references due by 8 January 2027, and general interviews likely 15–24 February 2027, with a possible technical interview first. The central EPSRC DLA guidance repeats the 8 December application deadline but gives a later reference date (15 February 2027); it also retains interview and outcome dates in 2026 despite describing 2027 entry. These pages conflict, and the central calendar appears stale. Confirm with Exeter admissions; until then, use 8 January 2027 as the reference deadline and do not rely on the central page's 2026 interview/outcome dates. Required documents include a personal statement, CV, transcript(s), two references and English evidence where required; applications must be in English, with certified translated academic qualifications. Submit through the link in the project description and follow the listed project-specific academic requirement: an appropriate subject with a strong quantitative or computational element; a relevant Master's is preferred. The generic award page also permits consideration of some Lower Second applicants with a Master's or significant relevant non-academic experience, subject to project criteria.

Exeter estimates basic living costs at about £1,300–£2,000 per month, based on local average spending in April 2026. The current stipend is about £1,817 per month before tax and individual circumstances, leaving a simple monthly margin from about £517 to a shortfall of about £183 across that estimate. This excludes one-off costs and future changes. Current 2026/27 postgraduate room examples range from £124.74 per week for a Garden Hill House standard room to £257.04 for a Rowancroft studio near St Luke's; neither price nor availability is guaranteed for 2027/28. Budget for visa/IHS if applicable, travel, deposits and setup costs separately.

Applicant Fit and Preparation

Required Background and Skills

The project-specific requirement is a First or Upper Second degree in an appropriate subject with a strong quantitative or computational element; a relevant Master's is preferred. The generic EPSRC scheme requires an appropriate science or technology qualification and allows Lower Second consideration when accompanied by a Master's or significant relevant non-academic experience. Any English-language evidence requirement depends on the applicant's circumstances. The project methods do not mean applicants must already be expert in every listed algorithm.

Preferred Experience and Suggested Preparation

The most relevant preparation is statistical modelling, programming, data cleaning and reproducible analysis, with an appreciation of clinical outcome definitions and the limits of routine health records. Useful application evidence could include a dissertation or portfolio showing a transparent train/test design, calibration as well as discrimination, missing-data handling, documentation, and communication of results to non-technical collaborators. Applicants with clinical or epidemiological experience can make the case for translating care pathways into usable variables; applicants from mathematics, statistics, computer science or engineering can show motivation to learn clinical context and governance. These are preparation suggestions, not additional formal entry criteria. Before an interview, review AKI diagnostic criteria and existing alert pathways, and prepare to explain how to prevent temporal leakage when predicting risk at admission.

Next Actions and Questions to Verify

Before Applying

• Confirm the second Exeter academic supervisor, formal registration route and programme title; the published project PDF names only one Exeter academic, while the award page requires at least two.
• Ask whether the dataset, NHS sites, data custodian and governance approvals are secured, and what information on data period and case count can be shared with applicants.
• Request clarification of “validated medicine code lists,” the outcome reference standard and whether laboratory alert criteria will be used.
• Ask Exeter admissions to resolve the reference-date conflict (8 January on the project award page; 15 February on the central EPSRC page) and confirm the interview schedule, because the central page still lists 2026 interview/outcome dates for 2027 entry. Until confirmed, prepare references by 8 January. Also confirm international-residency eligibility, fee-waiver coverage and application documents; do not rely on the inconsistent summary-year label for the stipend.

At Interview

• Ask how the admission-time landmark and prediction horizon are defined; how present-on-admission cases, variable timestamps, missingness and leakage are handled.
• Ask whether external validation will use separate NHS sites and how calibration, subgroup fairness, threshold choice and decision-curve results will affect the proposed tool.
• Clarify the expected split between methods development, clinical interpretation and stakeholder engagement, and whether the open-source tool is a research prototype or intended for implementation testing.

Before Accepting an Offer

• Obtain written confirmation of the 2027/28 stipend, international tuition treatment, award duration and any conditions on continuation.
• Confirm the named supervisory team, degree registration, data-access milestones, secure working environment and fallback plan if partner-site approvals are delayed.

Evidence and Sources

Project, award and institutional context

• Official HA-AKI project description
• Exeter EPSRC DLA award page, reference 6017
• EPSRC DLA eligibility and funding guidance
• UKRI support for funded students
• Department of Health and Community Sciences
• Health and Community Sciences postgraduate research
• PenCHORD: Operational Research and Data Science
• Professor Yinghui Wei, NIHR Exeter Biomedical Research Centre
• Exeter rankings and league tables
• Exeter REF 2021 results

Supervisor publications and field evidence

• OpenSAFELY EHR study of clinically coded long COVID (2025)
• Scoping review of mobile-health data and clinical decision-making (2025)
• Copula regression for semi-competing risks, including kidney transplantation (2023)
• UK Renal Registry report on AKI in England, 2024 data (published 2025)
• NICE guidance NG148: Acute kidney injury
• NICE historical AKI costing statement
• External validation of a commercial US acute-kidney-injury model (2024)
• Multicentre validation of an AKI model in critically ill patients (2023)
• Real-time inpatient AKI prediction model across five hospitals (2024)
• Review of AI at the bedside for AKI (2025)
• TRIPOD+AI reporting guidance (2024)
• PROBAST+AI quality and applicability guidance (2025)
• NHS England Secure Data Environment service
• UK secure data environment policy guidelines

Costs and careers

• Exeter budgeting guidance
• Exeter accommodation room finder
• UK National Careers Service: data scientist
• UK Digital and Technologies Jobs Plan