Kidney failure risk in multiple long-term conditions – establishing validation in a population with multiple longterm conditions and frailty, and exploration of patients’ and healthcare professionals’ perspectives of kidney failure risk

Walker, Heather (2026) Kidney failure risk in multiple long-term conditions – establishing validation in a population with multiple longterm conditions and frailty, and exploration of patients’ and healthcare professionals’ perspectives of kidney failure risk. PhD thesis, University of Glasgow.

Abstract

Background
Chronic kidney disease (CKD) is increasingly common in adults and is associated with the presence of multiple long-term health conditions (multimorbidity) and/or frailty. Individuals with CKD and multimorbidity and/or frailty are at increased risk of their kidney disease progressing to kidney failure and of mortality compared to those without multimorbidity or frailty. A large number of kidney failure prognostic models have been developed to predict the risk of progression of CKD to kidney failure. The kidney failure risk equation (KFRE) is the most widely used and validated of these models and is recommended in clinical guidance. However, it is unclear if kidney failure prognostic models, including KFRE, have been developed with consideration for individuals with
multimorbidity or frailty and if the presence of these impact how models perform or how they might be used in clinical practice.
Aims
The overall hypothesis was that the presence of multimorbidity and frailty in CKD negatively impacts the performance of the kidney failure risk equation and influences how it is used in the clinical care of individuals with multimorbidity and/or frailty.
This thesis aimed to:
1) Assess the prevalence of multimorbidity or frailty in studies that developed or validated kidney failure prognostic models for CKD.
2) Ascertain if the presence of multimorbidity or frailty in CKD impacted kidney failure prognostic model performance (focusing on the kidney failure risk equation) and if performance is influenced by the competing risk of mortality.
3) Evaluate if the performance of the kidney failure risk equation is impacted by the use of the biomarkers creatinine or cystatin-C in the estimation of glomerular filtration rate (GFR) utilised in the model, in the context of multimorbidity or frailty.
4) Explore how multimorbidity and frailty influence patients’ and healthcare professionals’ understanding of CKD and perceptions of risk of kidney failure, and the potential use of the kidney failure risk equation in shared decision-making in these groups.
Methods
A systematic review of existing kidney failure prognostic models in CKD was performed, followed by two quantitative validation studies utilising data from a prospective research based cohort (UK Biobank) and a routine care population based cohort (Stockholm CREAtinine Measurements Project: SCREAM) and finally a qualitative descriptive study involving patients with CKD and multimorbidity and/or frailty and healthcare professionals involved in their care from primary and secondary care settings. Following completion of the described studies, triangulation of the findings was performed to establish findings of convergence (agreement), complementarity (added insight), and dissonance (discrepancy).
Results
There was a lack of reporting of multimorbidity and frailty in a systematic review of ninety-seven studies reporting 121 different kidney failure prognostic models. Only two studies reported prevalence of multimorbidity and one prevalence of frailty. Individual comorbidities were reported more frequently (diabetes: 67.0%, hypertension 54.6% and cardiovascular disease 39.2% of included studies). No studies assessed the impact of multimorbidity on prognostic model performance and a single study explored associations between frailty and risk of kidney failure and death but not the direct impacts on model performance.
Quantitative analysis confirmed that multimorbidity was common in CKD (14,998 of 24,489 individuals in UK Biobank: 61.2% and 30,147 of 42,902 individuals in SCREAM: 70.3%). Discrimination of KFRE was good in individuals with and without multimorbidity (area under receiver operating characteristic curve ≥0.86).
However, there was miscalibration, with underestimation of kidney failure risk in people with multimorbidity in UK Biobank (observed/expected (O/E) ratio 1.75 at 5-years; estimated GFR [eGFR] creatinine). Conversely, in SCREAM calibration-in-the-large (O/E ratio) at 5-years was 1.05 in the multimorbidity group (eGFR creatinine). Substituting creatinine with cystatin C to estimate GFR did not improve the performance of KFRE. The cumulative incidence of death was higher with multimorbidity groups compared to no multimorbidity. An updated model (using the same covariates as KFRE) considering competing mortality improved calibration performance over KFRE, O/E ratio 0.98 in multimorbidity group of the validation cohort (UK Biobank) at 5-years.
The prevalence of frailty in UK Biobank was 35% (8,533 of 24,489 individuals). Discrimination of KFRE was good (area under receiver operating characteristic curve ≥0.88 across all frailty sub-groups). Calibration was sub-optimal, with kidney failure at 5-years under-estimated in individuals with frailty (observed/expected (O/E) ratio 1.70). Substitution of creatinine with cystatin in the eGFR equation improved calibration-in-the-large (5-years O/E ratio 1.20). There was overestimation of kidney failure risk in analyses that do not account for competing mortality risk compared to those that do, particularly in individuals with frailty and at higher KFRE predicted risk levels.
Thirty-one patients and nineteen healthcare professionals participated in interviews or focus groups in the qualitative study. Patients were aged 46-94 years old, reported 3-10 long-term health conditions and had a clinical frailty scale of 1-7. The key factors that were found to influence understanding and perception of CKD and kidney failure risk were: 1) Variability in patients’ knowledge and understanding of CKD, with many unaware of having received a diagnosis of CKD; 2) Work required to live with CKD alongside multimorbidity and/or frailty, including healthcare-associated work relating to complexity; 3) Relationships and interactions, where shared decision-making was facilitated by continuity and trust; 4) Context and priorities, including the level of prioritisation of CKD; and 5) Uncertainty.
Triangulation of findings identified that there were results from the four studies that agreed and complemented one another, but also some that were dissonant. There was agreement that frailty and multimorbidity impact KFRE performance and its clinical applicability/use. The findings from the qualitative study explored the challenges that can arise from the uncertainties of risk prediction in people with frailty and multimorbidity and how risk information may be used informatively and relevantly in clinical practice. Dissonance arose because although the quantitative studies found the KFRE to be statistically adequate, the qualitative work uncovered its limitations in real-world use, and the systematic review showed minimal reporting of frailty and multimorbidity or evaluation these groups in the existing literature. Application of KFRE scores in clinical management of CKD patients with multimorbidity and frailty may have sub-optimal performance with a degree of uncertainty which should inform the discussion and shared decision-making with these group of patients.
Conclusions
Multimorbidity and frailty are common in CKD and impact the performance of KFRE. Risk prediction is improved for those with frailty when cystatin C-based eGFR equations are utilised in KFRE but not for those with multimorbidity. Accounting for competing risk of mortality is also important for kidney failure risk prognostication in both multimorbidity and frailty. An updated model (utilising the same covariates as KFRE) accounting for competing mortality risk improves calibration in multimorbidity.
Qualitative perspectives identified that CKD was often not a priority, compared to other long-term conditions, both for patients and HCPs. There was a desire from many patients to increase knowledge of CKD and kidney failure risk alongside other health problems. The potential value of KFRE was highlighted in reducing uncertainty, although optimal ways to communicate risk and uncertainty require further research.
Overall, these findings highlight that consideration of statistical model performance is important in ensuring that risk prediction is optimal for individuals with multimorbidity and frailty. However, implementation and clinical use of KFRE and other kidney failure risk prediction models is shaped by how CKD is understood and prioritised alongside the complexity and potential burden of multimorbidity and frailty.

• Item Type: Thesis (PhD)
• Qualification Level: Doctoral
• Additional Information: Supported in part by the Wellcome Trust.
• Subjects: R Medicine > RC Internal medicine
• Colleges/Schools: College of Medical Veterinary and Life Sciences > School of Cardiovascular & Metabolic Health
• Funder's Name: Wellcome Trust (WELLCOTR)
• Supervisor's Name: Mark, Professor Patrick B., Sullivan, Dr. Michael, Jani, Dr. Bhautesh Dinesh and Gallacher, Dr. Katie
• Date of Award: 2026
• Depositing User: Theses Team
• Unique ID: glathesis:2026-86145
• Copyright: Copyright of this thesis is held by the author.
• Date Deposited: 21 Jul 2026 12:47
• Last Modified: 21 Jul 2026 12:49
• URI: https://theses.gla.ac.uk/id/eprint/86145

Related URLs

• Article DOI
• Article DOI
• Article DOI
• Article DOI
• Article DOI
• Article DOI
• Article DOI
• Article DOI
• Article DOI
• Article DOI
• Conference item
• Conference item
• Conference item
• Conference item
• Conference item
• Conference item
• Conference item
• Article DOI
• Article DOI