https://orcid.org/0000-0002-9905-3422
Figures
Abstract
Background and objectives
Cognitive impairment is common in haemodialysis patients and associated with adverse health outcomes. This may be due to cognitive impairments interfering with daily functioning and self-care, but evidence is limited. This cross-sectional study aims to explore the interrelationships between cognition and functional outcomes in haemodialysis patients.
Methods
Haemodialysis patients completed measures of objective cognitive function (Montreal Cognitive Assessment), everyday problem-solving skills (scenario-based task), and subjective cognitive complaints (self-report). Participants also self-reported sociodemographic information, functional interference, treatment nonadherence, and mood and fatigue symptoms. Patients’ clinical data including comorbidities and lab results were extracted from medical record. Structural equation modelling was performed.
Results
A total of 268 haemodialysis patients (mean age = 59.87 years; 42.5% female) participated. The final model showed satisfactory fit: CFI = 0.916, TLI = 0.905, RMSEA = 0.033 (90% confidence interval 0.024 to 0.041), SRMR = 0.066, χ2(493) = 618.573 (p < .001). There was a negative association between objective cognitive function and subjective cognitive complaints. Cognitive complaints were positively associated with both functional interference and treatment nonadherence, whereas objective performance was not. Everyday problem-solving skills emerged as a distinct aspect of cognition not associated with objective performance or subjective complaints, but had additive utility in predicting functional interference.
Conclusions
Subjective cognitive complaints and everyday problem-solving skills appear to be stronger predictors of functional variables compared to objective performance based on traditional tests. Routine screening of everyday cognitive difficulties may allow for early identification of dialysis patients at risk of cognitive impairment, functional interference, treatment nonadherence, and poor clinical outcomes.
Citation: Chan FHF, Sim P, Lim PXH, Zhu X, Lee J, Haroon S, et al. (2024) Structural equation modelling of the role of cognition in functional interference and treatment nonadherence among haemodialysis patients. PLoS ONE 19(10): e0312039. https://doi.org/10.1371/journal.pone.0312039
Editor: Henry H.L. Wu, Kolling Institute of Medical Research, The University of Sydney, AUSTRALIA
Received: May 6, 2024; Accepted: September 30, 2024; Published: October 17, 2024
Copyright: © 2024 Chan 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 author and source are credited.
Data Availability: All relevant data are within the paper and its Supporting information files.
Funding: This work was supported by the Venerable Yen Pei-National Kidney Foundation Research Fund, Singapore [grant number NKFRC/2021/01/02]. KG received research funding from National Kidney Foundation Singapore. The funding sources had no role in the study design, recruitment of patients, data collection, analysis, interpretation of the results, writing of the manuscript, or decision to submit the manuscript for publication.
Competing interests: The authors have declared that no competing interests exist.
Introduction
End-stage renal disease (ESRD) is the most advanced stage of chronic kidney disease where kidney function is irreversibly lost, necessitating dialysis or transplantation [1, 2]. Cognitive impairments (CIs) are common in ESRD patients receiving haemodialysis (HD) treatment, with more than 70% exhibiting at least mild impairments in one or more domains such as attention, memory, and executive function [3–6]. “Brain fog” has been a common complaint among dialysis patients [7] and a popular topic of discussion in online patient forums [8]. CIs in ESRD are associated with adverse health outcomes including dialysis withdrawal, hospitalisation, and mortality [3, 9–11]. These associations are assumed to be due to CIs interfering with daily functioning, decision-making, and self-management capabilities, however empirical evidence is scarce.
CIs are typically accompanied by functional interference because performance of everyday activities (e.g., personal hygiene, managing finances, etc.) is dependent upon the integrity of cognitive, motor, and sensory-perceptual skills [12]. In the context of ESRD, a specific aspect of daily functioning that is of particular relevance to patients is self-care and treatment adherence. HD patients are prescribed complex medical regimen that requires them to take daily medications, follow strict dietary guidelines, control fluid intake, and attend thrice-weekly dialysis sessions. These self-care activities are cognitively demanding and can be challenging for those with CIs. Medication taking, for instance, requires multiple cognitive processes including encoding and storage of health information (e.g., understanding the importance of taking medicine), executive function (e.g., developing a plan to adhere), prospective memory (e.g., remembering to take medicine on time), working memory (e.g., keeping the intention to take medicines active while preparing to take it), and source monitoring (e.g., remembering whether the medicine has been taken) [13].
Clearly cognition plays an essential role in HD patients’ daily functioning and ability to maintain and optimise their health by following the prescribed medical regimen. However, currently we lack a comprehensive picture of the interplay between aspects of cognition and aspects of daily functioning in this population. Studies using neuropsychological tests, the gold standard measure of cognition, have found positive associations between objective cognitive ability and functional independence in dialysis patients [14]. Other studies have used self-reported measures to assess subjective cognitive complaints (SCCs). More frequent SCCs in HD patients have been found to be associated with greater functional impairment [15], and worse treatment adherence indicated by self-reports and laboratory results [16, 17]. Furthermore, scenario-based tasks that assess everyday problem-solving skills have been used. These tasks require participants to generate solutions in response to problem scenarios, and are thought to reflect executive ability in real-world contexts [18]. Two studies found that everyday problem-solving skills positively predicted medication adherence in kidney transplant recipients [19, 20].
Despite these promising findings, a study that comprehensively examines the complex associations between cognition and real-world functioning in HD patients is lacking. As such, the current cross-sectional study aims to adopt structural equation modelling (SEM) analysis to disentangle the interrelationships among multiple aspects of cognition (i.e., objective performance, subjective complaint, and everyday problem-solving) and key outcomes (i.e., functional interference, treatment nonadherence, and clinical endpoints) in HD patients. SEM is a powerful and flexible statistical technique that integrates factor analysis, path analysis, and regression into a single framework [21]. It simultaneously accounts for multiple direct and indirect associations among a range of variables, hence allowing for the validation of complex theoretical models using a unified approach [21]. With this advanced statistical technique, this study will map key cognitive indicators of functional interference and nonadherence so that future interventions and support strategies could be developed to target the cognitive challenges that interfere with these key endpoints.
The hypothesised model to be tested in the current study is shown in Fig 1. We hypothesised that objective cognitive function and everyday problem-solving skills would covary (hence the bidirectional arrow) since they are both considered indicators of distinct aspects of cognitive abilities [18]. While objective tests reflect individuals’ cognitive performance in optimal conditions, problem-solving tasks reflect individuals’ problem-solving ability in everyday contexts. We also hypothesised that these two variables would directly contribute to SCCs since individuals with worse cognitive performance may experience more cognitive difficulties in daily lives, hence reporting more frequent SCCs [7, 22]. Furthermore, the three cognitive indicators were each hypothesised to have a direct effect on functional interference and treatment nonadherence based on previous evidence [14–17, 19, 20].
Materials and methods
Participants
A convenience sample of HD patients was recruited from the National Kidney Foundation Singapore (NKF) between May 19 and November 4 2022. Ten NKF dialysis centres were selected to ensure geographical representation of dialysis centres across Singapore. The inclusion criteria were: (1) 21 years of age or older, (2) an estimated glomerular filtration rate lower than 15mL/min/1.73m2, (3) having undergone HD treatment for at least 3 months, and (4) fluent in either English or Mandarin. The exclusion criteria were: (1) only fluent in dialects, (2) unable to give consent due to psychiatric diagnoses or established diagnosis of dementia, or (3) unable to complete survey due to visual or hearing impairments.
Procedure
The study protocol was approved by the Institutional Review Board of the Nanyang Technological University (NTU-IRB-2021-025). A list of eligible patients was provided by the nurse managers of each dialysis centre. Study team members fluent in the patients’ preferred language approached each patient during their dialysis sessions. After obtaining written consent, the following instruments were administered. Upon completion, patients were given a cash compensation.
Measures
Objective cognitive function.
The Montreal Cognitive Assessment (MoCA) was used to assess objective cognitive function [23]. The MoCA is a cognitive screening test that assesses visuospatial and executive functions (i.e., Trail-Making Test part B, cube copy, clock drawing, abstraction), attention (i.e., digit span forward and backward, vigilance, serial-7 subtraction), short-term memory (i.e., delayed recall), language (i.e., naming, sentence repetition, verbal fluency), and orientation (i.e., awareness of time and place) [23].
Everyday problem-solving skills.
The Everyday Problem-Solving (EPS) task consists of real-world problem scenarios where participants were asked to generate solutions. Six scenarios that have been used in previously studies were used in the current study [19, 24]. Three scenarios described general daily problems and the other three described health-specific problems. For each scenario participants were asked to generate as many solutions as possible. The number of safe and effective solutions generated by each patient was used as an indicator of problem-solving skills [20].
Subjective cognitive complaints.
SCCs were measured using the 33-item Patient’s Assessment of Own Functioning Inventory (PAOFI) [25]. This measure assesses SCCs in four domains: memory (10 items), language (nine items), motor/sensory-perceptual ability (five items), and higher-level cognitive functions (nine items). Participants rated on a six-point Likert scale from “almost never” to “almost always” [25]. Mean scores of the four subscales were calculated, with a higher score indicating more frequent SCCs. Cronbach’s alpha was 0.87 for the memory subscale, 0.89 for the language subscale, 0.73 for the motor/sensory-perceptual subscale, and 0.87 for the higher-level cognitive function subscale.
Functional interference.
We assessed functional interference as a key dependent variable of the current study. The Work and Social Adjustment Scale (WSAS) is a measure of self-perceived functional interference in five domains (i.e., work, home management, social leisure activities, private leisure activities, and social relationships) attributable to an identified problem [26]. The original five items were used but preface was reworded to be cognition-specific. Participants rated on a nine-point Likert scale ranging from 0 (i.e., “not at all impaired”) to 8 (i.e., “very severely impaired”). Cronbach’s alpha was 0.93 for WSAS.
Treatment nonadherence.
Patients’ self-reported medication nonadherence was measured by the five-item Medication Adherence Report Scale (MARS-5 ©Professor Rob Horne), which was rated on a five-point scale ranging from “never” to “always” [27]. A higher total score indicates poorer medication adherence. Cronbach’s alpha was 0.77. The Dialysis Diet and Fluid non-adherence Questionnaire (DDFQ) was also assessed. DDFQ is a four-item scale that assesses frequency and degree of dietary and fluid nonadherence in dialysis patients [28]. Moreover, we collected patients’ interdialytic weight gain (IDWG) as a clinical indicator of fluid adherence. Relative IDWG (i.e., the ratio of absolute IDWG to a patient’s dry weight) was assessed prior to each dialysis session in the week of and the week before the survey date. The IDWG values were averaged across the two weeks. Additionally, the latest lab assay results (i.e., sodium [Na], potassium [K], calcium [Ca], phosphorus [PO4], and calcium-phosphorus product [Ca×PO4]) were collected and used as indicators of dietary and medication adherence.
Mood and fatigue symptoms.
Patients’ mood symptoms were measured by the two-item Patient Health Questionnaire (PHQ-2; α = 0.65) and the two-item Generalised Anxiety Disorder (GAD-2; α = 0.77). These two measures are brief screening tools of depression and anxiety that have been used in dialysis patients [29, 30]. Fatigue was measured using the one-item vitality subscale from the Kidney Disease Quality of Life questionnaire [31–33]. Higher scores indicate greater mood or fatigue symptoms.
Sociodemographic and clinical information.
Self-reported demographic information was collected, including age, gender, ethnicity, education, relationship status, and employment status. Clinical information including primary kidney disease diagnosis, comorbidities, duration on HD, dialysis dose (Kt/V), and medication count, were extracted from patients’ medical record.
Statistical analyses
The steps of conducting the SEM analysis involved (1) testing of the baseline measurement model with confirmatory factor analysis (CFA), (2) specification of the structural model containing all hypothesised paths, and (3) modification of the structural model [34–36]. These steps were performed using the “lavaan” package with the WLSMV estimator in R 4.2.2 [37, 38]. To assess model fit, we used the Comparative Fit Index (CFI) [39], Tucker—Lewis Index (TLI) [40], the Root Mean Square Error of Approximation (RMSEA) [41], the Standardised Root Mean Square (SRMR) [21], and the Chi-square [21]. For CFI and TLI, values above 0.90 were considered to indicate adequate model fit, and values above 0.95 were considered to indicate excellent fit [39, 42–45]. RMSEA values lower than 0.06 and SRMR values lower than 0.08 were considered to indicate good fit [21].
Model development
Measurement model.
The measurement model was first constructed, with a total of eight latent variables, each measured by multiple indicator variables. Three latent variables were constructed to reflect different aspects of cognition, namely “objective cognitive function”, “everyday problem-solving skills”, and “subjective cognitive complaints”. These three latent variables were indicated by the subscale/subdomain scores in the MoCA, EPS, and PAOFI, respectively. Two latent variables were constructed for the dependent variables, including “functional interference” measured by the five WSAS items, and “treatment nonadherence” measured by self-report (i.e., MARS-5 and DDFQ), and physiological and biochemical parameters (i.e., IDWG, Na, K, Ca, PO4, and Ca×PO4).
Following previous SEM studies of cognitive function in patients with chronic disease [20, 34, 46], we constructed three additional latent variables in order to account for the confounding effects sociodemographic, clinical, and psychological factors in the hypothesised relationships. Specifically, a “sociodemographic” latent variable (indicated by age and years of education) and a “comorbidity” latent variable (indicated by presence of diabetes, hypertension, hyperlipidaemia, and cardiovascular disease) were constructed because these are established risk or protective factors of CIs in HD patients [47–49]. A “mood and fatigue symptoms” latent variable (indicated by PHQ-2, GAD-2, and KDQOL vitality item) was also specified because these symptoms can exacerbate SCCs [22] and are associated with behavioural and clinical outcomes in ESRD patients [50].
A CFA was first conducted to verify the measurement quality of the eight latent variables. The fit indices for the baseline measurement model were as follows: CFI = 0.839, TLI = 0.820, RMSEA = 0.043 (90% confidence interval 0.036 to 0.049), SRMR = 0.078, χ2(566) = 810.166 (p < .001), which suggested unsatisfactory model fit. All indicator variables had significant factor loadings on their corresponding latent variables except for two adherence indicators (i.e., Na and Ca). These two indicators were therefore removed from the model. However, fit indices for the revised measurement model were still in the unacceptable range, CFI = 0.853, TLI = 0.835, RMSEA = 0.043 (90% confidence interval 0.036 to 0.050), SRMR = 0.078, χ2(499) = 719.322 (p < .001).
Modification indices suggested inclusion of error covariance between PO4 and Ca×PO4 in the model. These two variables were indeed highly correlated with each other, r = 0.95, p < .001. There is evidence that calcium and phosphorus are controlled by similar regulatory mechanisms [51], and high levels of serum phosphorus can combine with calcium to form calcium-phosphorus product [52]. It may therefore be theoretically justifiable to allow the residual terms of these two variables to freely covary [21]. We also added error covariance among DDFQ items due to the similar wordings of these questions, and we added error covariance between relative IDWG and the two fluid items in DDFQ as they are both thought to reflect fluid adherence. Following this modification, the final measurement model showed acceptable fit with the exception of Chi-square, CFI = 0.917, TLI = 0.905, RMSEA = 0.033 (90% confidence interval 0.024 to 0.041), SRMR = 0.065, χ2(492) = 616.607 (p < .001). However, Chi-square test is sensitive to sample size with larger samples decreasing the p value; the significant Chi-square was therefore not used as a basis for model rejection.
In the final measurement model, all factor loadings of the latent variables were at a significance level of p < .05. Detailed description of the final measurement model, including the latent and indicator variables, as well as their measurement, interpretation, mean values, percentage, and factor loadings, are presented in Table 1.
Structural model.
A structural model was then constructed to examine the hypothesised regression paths between the latent variables specified in Fig 1. Objective cognitive function and EPS skills were hypothesised to covary with each other, and to have negative effects on SCCs. These three cognitive indicators were also hypothesised to have direct effects on functional interference and treatment nonadherence. However, the regression path from objective cognitive function to treatment nonadherence was considered optional because, despite theoretical assumptions, extant empirical evidence in the context of kidney disease suggest no association between these two variables [19, 20].
In addition, regression paths were added from sociodemographic factors, comorbidity, and mood and fatigue symptoms, to each of the five latent variables in Fig 1 (i.e., objective cognitive function, everyday problem-solving skills, subjective cognitive complaints, functional interference, treatment nonadherence), in order to account for their potential confounding effects [22, 47, 53–55]. The structural model was tested and respecified until a final model was determined by weighing model fit indices and model parsimony (i.e., a simpler model with fewer parameters is favoured over more complex models provided the models fit the data similarly well).
Results
Sample characteristics
We approached a total of 459 HD patients in NKF dialysis centres. Ninety patients were excluded due to ineligibility. Within the remaining 369 eligible patients, 268 consented to participate (response rate 72.6%). The main reasons for rejection were lack of interest and feeling unwell. Therefore, 268 patients were included in final analyses. Table 2 reports the sociodemographic and clinical profiles of the sample. The mean age of the sample was 59.87 (SD = 11.72). Patients on average had been on HD for 78.85 (SD = 62.80) months.
Final structural model
The fit indices for the full structural model were as follows: CFI = 0.917, TLI = 0.905, RMSEA = 0.033 (90% confidence interval 0.024 to 0.041), SRMR = 0.065, χ2(492) = 616.607 (p < .001). The path between objective cognitive function and treatment nonadherence was indeed not statistically significant, consistent with prior work [19, 20]. We removed this path while retaining all other paths and reran the structural model, with fit indices as follows: CFI = 0.916, TLI = 0.905, RMSEA = 0.033 (90% confidence interval 0.024 to 0.041), SRMR = 0.066, χ2(493) = 618.573 (p < .001). A chi-squared difference test showed that this trimmed model, which was also the more parsimonious model, could explain the observed data equally well compared to the full structural model without a significant loss in data-model fit (p = .101).
The final trimmed model is presented in Fig 2. For conciseness, only regression paths that were statistically significant were presented in the figure. The model revealed that objective cognitive performance negatively predicted SCCs, but was not associated with EPS skills, functional interference, or treatment nonadherence. SCCs were positively associated with both functional interference and treatment nonadherence, suggesting indirect effects of objective performance on these dependent variables through SCCs. On the other hand, EPS skills were found to be unrelated to either objective performance or SCCs, but had a negative association with functional interference, suggesting that individuals with worse EPS skills had more severe functional interference. Sociodemographic factors (i.e., age and years of education) were associated with objective cognitive function, EPS skills, and treatment nonadherence, whereas mood and fatigue symptoms were predictive of SCCs and functional interference. Comorbidity, however, was not associated with any other latent variables, and was therefore not presented in Fig 2. The final model explained 65.7% of the variance in functional interference (indicated by an R2 of 0.657), and 63.3% of the variance in treatment nonadherence (indicated by an R2 of 0.633).
Standardised coefficients with standard errors are reported for regression paths. Standardised factor loadings are reported for paths between indicator variables and latent variables. Nonsignificant paths are not presented in the figure.
Discussion
To the best of our knowledge, this is the first SEM study to explore the complex interrelationships between multiple cognitive indicators and key functional and clinical variables in dialysis patients. The final model revealed interesting pathways through which cognition was directly and indirectly associated with functional interference and treatment nonadherence, highlighting cognitive difficulties as an important barrier in this population to functional capacity, life participation, treatment adherence, and disease management.
A key finding was that patients with more frequent SCCs also experienced more severe daily interference and exhibited poorer treatment adherence. In contrast, objective cognitive performance was only indirectly associated with these outcomes through SCCs. Indeed, it has been suggested that objective CIs based on neuropsychological tests do not necessarily translate to impairments in real-world functioning outside of test environment because some patients may adopt strategies to compensate for everyday cognitive failure, which is typically not allowed in standardised testing [56]. For some other patients, cognitive difficulties may be too mild to be detected by objective tests, but are nevertheless problematic for everyday life [57]. Consistent with our findings, Song et al. found that SCCs, but not objective performance, predicted self-reported difficulties in performing activities of daily living in dialysis patients [15].
Importantly, treatment nonadherence in the present study was measured by both self-report and biochemical and physiological parameters. Inter-dialytic weight gain is typically thought to be an indicator of fluid adherence and sodium intake, whereas potassium and phosphorus may indicate dietary adherence [55]. Serum phosphorus can also indicate patient adherence to phosphate binders. Higher values of these clinical markers have been associated with poor survival [55]. Our findings are consistent with a recent longitudinal study where a reduction in SCCs over time was accompanied by a significant improvement in serum levels of PO4 and Ca×PO4 in HD patients [17]. Notably, the observed association between SCCs and nonadherence was significant even taking into account sociodemographic factors, comorbidity, and mood and fatigue symptoms, suggesting that this effect cannot be explained by other well-established determinants of nonadherence such as age and depression. Taken together, SCCs appear to be a useful measure that can simultaneously red-flag functional interference, nonadherence, and poor clinical outcomes in dialysis patients.
Another interesting finding was that patients with worse EPS skills were more likely to experience functional interference. The EPS task used in this study required participants to generate safe and effective solutions to the given scenarios. There is evidence in the cognitive development literature that the ability to generate a number of alternative solutions is a good indicator of problem-solving ability [58]. Patients who were unable to generate multiple solutions in response to the EPS task scenarios may also be more likely to experience failure in solving problems that arise in real-world contexts, which in the long term can impair independence in various aspects of life.
Unexpectedly though, EPS task performance was not associated with treatment nonadherence. This is inconsistent with two other studies where EPS skills were found to predict medication adherence in kidney transplantation recipients [19, 20]. These two studies, however, only assessed cognition using neuropsychological tests and the EPS task. It may be that SCCs have a stronger association with adherence and therefore by accounting for its effect in our model, the predictive value of EPS skills on adherence diminished. Indeed, the SCC measure used in our study is comprehensive and covers multiple cognitive domains, whereas EPS skills are dependent not just upon executive function, but also life experience, knowledge, and environmental factors [59].
We found that patients with worse cognitive performance on objective tests also subjectively reported more cognitive complaints. This observed association was modest, which is not unexpected based on previous work in other populations [22, 60]. Although objective tests and subjective reports both intend to capture the same underlying construct (i.e., cognition), they measure it using completely different methods and are influenced by different sets of predictors. While objective tests assess performance at a single time point in distraction-free environments, self-reports are based on accumulative daily experience that may be more reflective of longitudinal cognitive changes [60]. Objective and subjective cognition have therefore been considered as distinct constructs that complement each other with different utility in clinical and research settings. Assessing objective CIs can help establishing diagnosis of CIs and allow for advance care planning, whereas assessing SCCs can help identifying aspects of CIs that have the greatest impact on patients, hence allowing for a more patient-centred approach to managing this debilitating complication [61].
The current study has several important clinical implications. First, cognitive indicators that incorporate everyday scenarios (i.e., EPS and SCCs) appeared to be better predictors of real-world functional and clinical outcomes in HD patients, compared to traditional cognitive tests. The associations of SCCs with underlying cognitive deficits and worse outcomes highlight the potential of SCCs as a stand-alone patient-reported outcome measure with clinical utility in dialysis settings that simultaneously signifies risks in multiple aspects of patients’ health and well-being. Indeed, SCCs are increasingly considered as a prodromal marker of progression to dementia in Alzheimer’s disease research [62, 63], and as a core patient-reported outcome in populations such as cancer [64] and HIV patients [65]. To date, research and clinical practice in the field of nephrology have been predominantly focusing on objective cognition, with SCCs being understudied and underrecognised. The emphasis on objective cognition is essential for diagnostic purposes, but hinders a comprehensive understanding of cognitive well-being in this population. Previous studies determined the utility of SCC measures solely based on their ability to distinguish patients with and without objective CIs [66]. We propose that SCCs should be treated as an outcome equally important as objective CIs, and should be assessed in combination with neuropsychological tests where possible.
Second, everyday cognitive difficulties reflected by SCCs and the EPS task emerged as potential modifiable risk factors for functional interference and treatment nonadherence. There is evidence that functional impairment is present in 21 to 85% of ESRD patients [11], and that optimal adherence is not achieved in as many as 80% of dialysis patients [55]. Our study showed that these high rates may in part be explained by the cognitive burden experienced by this population. It is therefore pivotal to develop and implement interventions that improve everyday cognitive abilities and compensate for cognitive lapses in this population so that the impacts of CIs on daily functioning and self-care can be mitigated. To date, there is a lack of research on the feasibility and effectiveness of cognitive interventions for ERSD patients. Future work in this area is needed to improve patient-centred care and to optimise adaptation to dialysis initiation.
Several limitations warrant acknowledgement. First, it should be noted that this is a cross-sectional observational study. Also, SEM is not inherently a causal method [21]. Caution is therefore needed when interpreting the significant paths in our final model. Although the directions of these associations have been hypothesised and tested, the study design and statistical method do not permit the interpretation of a cause-and-effect sequence. Future longitudinal studies and intervention research are needed to explore potential causal mechanisms and further validate our observations. Second, the MoCA used in this study is only a brief screening tool for global cognition. No comprehensive neuropsychological battery or standard diagnostic test was conducted to determine objective cognitive function. This is because the tools administered in this study were already very lengthy and required about 30–40 minutes for each patient to complete. A more comprehensive neuropsychological assessment was therefore not carried out considering patients’ response burden and potential fatigue. Third, SCCs and functional interference were measured by self-report, which may be susceptible to recall bias. Future studies that incorporate informant-reports of SCCs and more objective measures of functional capacity may be needed. Finally, the study was conducted in a sample of HD patients, which limited the generalisability to other kidney disease subgroups such as peritoneal dialysis and kidney transplantation.
Conclusion
In summary, this study adopted the SEM technique to disentangle the complex associations between cognition and key functional and clinical parameters in HD patients. Results revealed the central role of SCCs in indicating not just underlying cognitive deficits but also functional interference, treatment nonadherence, and suboptimal clinical outcomes. In contrast, EPS skills were found to be associated with only functional interference but not nonadherence, whereas performance on traditional tests did not associate with any of these variables. It may be important to screen for SCCs in dialysis patients which would allow for early identification of the high-risk population, and subsequently early prevention or intervention strategies to mitigate the consequences of CIs.
Acknowledgments
We thank the nurse managers and staff of the dialysis centres of National Kidney Foundation Singapore for facilitating the patient recruitment procedures and providing patients’ medical data. We would also like to thank Kevin Tan, Shane Tan, Sze Ing Tan, and Nicole Tan for their assistance in patient recruitment and data entry.
References
- 1. Levey AS, Coresh J. Chronic kidney disease. Lancet. 2012;379: 165–180.
- 2. Himmelfarb J, Vanholder R, Mehrotra R, Tonelli M. The current and future landscape of dialysis. Nat Rev Nephrol. 2020;16: 573–585. pmid:32733095
- 3. van Zwieten A, Wong G, Ruospo M, Palmer SC, Teixeira-Pinto A, Barulli MR, et al. Associations of Cognitive Function and Education Level With All-Cause Mortality in Adults on Hemodialysis: Findings From the COGNITIVE-HD Study. Am J Kidney Dis. 2019;74: 452–462. pmid:31160141
- 4. Drew DA, Tighiouart H, Rollins J, Duncan S, Babroudi S, Scott T, et al. Evaluation of screening tests for cognitive impairment in patients receiving maintenance hemodialysis. J Am Soc Nephrol. 2020;31: 855–864. pmid:32132197
- 5. Erken E, Altunoren O, Senel ME, Tuncel D, Yilmaz T, Ganidagli SE, et al. Impaired cognition in hemodialysis patients: The Montreal Cognitive Assessment (MoCA) and important clues for testing. Clin Nephrol. 2019;91: 275–283. pmid:30686287
- 6. Murray AM, Tupper DE, Knopman DS, Gilbertson DT, Pederson SL, Li S, et al. Cognitive impairment in hemodialysis patients is common. Neurology. 2006;67: 216–223.
- 7. Chan FHF, Goh ZZS, Zhu X, Tudor Car L, Newman S, Khan BA, et al. Subjective cognitive complaints in end-stage renal disease: a systematic review and meta-analysis. Health Psychol Rev. 2023;17: 614–640.
- 8. Crowe K, Quinn TJ, Mark PB, Findlay MD. “Is It Removed During Dialysis?”—Cognitive Dysfunction in Advanced Kidney Failure—A Review Article. Front Neurol. 2021;12: 787370. pmid:34925220
- 9. Kurella M, Mapes DL, Port FK, Chertow GM. Correlates and outcomes of dementia among dialysis patients: The dialysis outcomes and practice patterns study. Nephrol Dial Transplant. 2006;21: 2543–2548. pmid:16751655
- 10. Griva K, Stygall J, Hankins M, Davenport A, Harrison M, Newman SP. Cognitive Impairment and 7-Year Mortality in Dialysis Patients. Am J Kidney Dis. 2010;56: 693–703. pmid:20800327
- 11. Kallenberg MH, Kleinveld HA, Dekker FW, van Munster BC, Rabelink TJ, van Buren M, et al. Functional and cognitive impairment, frailty, and adverse health outcomes in older patients reaching ESRD-a systematic review. Clin J Am Soc Nephrol. 2016;11: 1624–1639. pmid:27342598
- 12. Cipriani G, Danti S, Picchi L, Nuti A, Di Fiorino M. Daily functioning and dementia. Dement e Neuropsychol. 2020;14: 93–102. pmid:32595877
- 13. Insel K, Morrow D, Brewer B, Figueredo A. Executive function, working memory, and medication adherence among older adults. Journals Gerontol—Ser B Psychol Sci Soc Sci. 2006;61B: 102–107. pmid:16497953
- 14. Kavanagh NT, Schiller B, Saxena AB, Thomas IC, Kurella Tamura M. Prevalence and correlates of functional dependence among maintenance dialysis patients. Hemodial Int. 2015;19: 593–600. pmid:25731070
- 15. Song M-K, Ward SE, Bair E, Weiner LJ, Bridgman JC, Hladik GA, et al. Patient-reported cognitive functioning and daily functioning in chronic dialysis patients. Hemodial Int. 2015;19: 90–99. pmid:25110172
- 16. Chan FHF, Newman S, Khan BA, Griva K. The role of subjective cognitive complaints in self‑management among haemodialysis patients: a cross‑sectional study. BMC Nephrol. 2022;23: 363. pmid:36376848
- 17. Chan FHF, Newman S, Khan BA, Griva K. Prevalence and trajectories of subjective cognitive complaints and implications for patient outcomes: a prospective study of haemodialysis patients. Br J Health Psychol. 2023;28: 651–671. pmid:36720474
- 18. Thornton WL, Dumke HA. Age differences in everyday problem-solving and decision-making effectiveness: A meta-analytic review. Psychol Aging. 2005;20: 85–99. pmid:15769216
- 19. Gelb SR, Shapiro RJ, Thornton WL. Predicting Medication Adherence and Employment Status Following Kidney Transplant: The Relative Utility of Traditional and Everyday Cognitive Approaches. Neuropsychology. 2010;24: 514–526. pmid:20604625
- 20. Paterson TSE, O’Rourke N, Jean Shapiro R, Thornton WL. Medication adherence in renal transplant recipients: A latent variable model of psychosocial and neurocognitive predictors. PLoS One. 2018;13: e0204219. pmid:30265697
- 21.
Mueller RO, Hancock GR. Best Practices in Structural Equation Modeling. In: Osborne J, editor. Best Practices in Quantitative Methods. Thousand Oaks, CA: SAGE Publications, Inc.; 2008. pp. 488–508.
- 22. Rabin LA, Smart CM, Amariglio RE. Subjective Cognitive Decline in Preclinical Alzheimer’s Disease. Annu Rev Clin Psychol. 2017;13: 369–396. pmid:28482688
- 23. Nasreddine ZS, Phillips NA, Bedirian V, Charbonneau S, Whitehead V, Collin I, et al. The Montreal Cognitive Assessment, MoCA: A Brief Screening Tool For Mild Cognitive Impairment. J Am Geriatr Soc. 2005;53: 695–699. pmid:15817019
- 24. Allaire JC, Marsiske M. Well- and ill-defined measures of everyday cognition: relationship to older adults’ intellectual ability and functional status. Psychol Aging. 2002;17: 101–115. pmid:11931279
- 25.
Chelune GJ, Heaton RK, Lehman RAW. Neuropsychological and Personality Correlates of Patients’ Complaints of Disability. In: Goldstein G, Tarter RE, editors. Advances in Clinical Neuropsychology. Boston, MA.: Springer; 1986. pp. 95–126.
- 26. Mundt JC, Marks IM, Shear MK, Greist JH. The Work and Social Adjustment Scale: A simple measure of impairment in functioning. Br J Psychiatry. 2002;180: 461–464. pmid:11983645
- 27. Horne R, Weinman J. Self-regulation and self-management in asthma: Exploring the role of illness perceptions and treatment beliefs in explaining non-adherence to preventer medication. Psychol Heal. 2002;17: 17–32.
- 28. Vlaminck H, Maes B, Jacobs A, Reyntjens S, Evers G. The dialysis diet and fluid non-adherence questionnaire: Validity testing of a self-report instrument for clinical practice. J Clin Nurs. 2001;10: 707–715. pmid:11822521
- 29. Arroll B, Goodyear-Smith F, Crengle S, Gunn J, Kerse N, Fishman T, et al. Validation of PHQ-2 and PHQ-9 to screen for major depression in the primary care population. Ann Fam Med. 2010;8: 348–353. pmid:20644190
- 30. Plummer F, Manea L, Trepel D, McMillan D. Screening for anxiety disorders with the GAD-7 and GAD-2: A systematic review and diagnostic metaanalysis. Gen Hosp Psychiatry. 2016;39: 24–31. pmid:26719105
- 31. Yang F, Wang VW, Joshi VD, Lau TWL, Luo N. Validation of the english version of the kidney disease quality of life questionnaire (KDQOL-36) in haemodialysis patients in Singapore. Patient. 2013;6: 135–141. pmid:23613285
- 32. Rao S, Carter WB, Mapes DL, Kallich JD, Kamberg CJ, Spritzer KL, et al. Development of subscales from the symptoms/problems and effects of kidney disease scales of the Kidney Disease Quality of Life Instrument. Clin Ther. 2000;22: 1099–1111. pmid:11048907
- 33. Hays RD, Kallich JD, Mapes DL, Coons SJ, Carter WB. Development of the Kidney Disease Quality of Life (KDQOL) Instrument. Qual Life Res. 1994;3: 329–338. pmid:7841967
- 34. Vance DE, Okonkwo O, Antia L, Smith BA, Blanshan SA, Hiers KM, et al. Factors of cognitive complaints in adults with HIV: A structural equation model analysis. Occup Ther Ment Heal. 2009;25: 4–25.
- 35.
Ramlall I. Model Modification. Applied Structural Equation Modelling for Researchers and Practitioners. Bingley: Emerald Group Publishing Limited; 2016. pp. 75–77.
- 36.
Kremelberg D. Structural Equation Modeling. Practical Statistics: A Quick and Easy Guide to IBM® SPSS® Statistics, STATA, and Other Statistical Software. SAGE Publications, Inc.; 2014.
- 37. Rosseel Y. lavaan: An R Package for Structural Equation Modeling. J Stat Softw. 2012;48: 1–36.
- 38.
R Core Team. R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing; 2018. https://www.r-project.org/
- 39. Hu L, Bentler PM. Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Struct Equ Model. 1999;6: 1–55.
- 40. McDonald RP, Marsh HW. Choosing a multivariate model: Noncentrality and goodness of fit. Psychol Bull. 1990;107: 247.
- 41.
Browne MW, Cudeck R. Alternative ways of assessing model fit. In: Bollen KA, Long JS, editors. Testing structural equation models. Beverly Hills, CA: Sage; 1993. pp. 136–162.
- 42. Hopwood CJ, Donnellan MB. How should the internal structure of personality inventories be evaluated? Personal Soc Psychol Rev. 2010;14: 332–346. pmid:20435808
- 43. Marsh HW, Hau KT, Wen Z. In search of golden rules: Comment on hypothesis-testing approaches to setting cutoff values for fit indexes and dangers in overgeneralizing Hu and Bentler’s (1999) findings. Struct Equ Model. 2004;11: 320–341.
- 44. Clark DA, Bowles RP. Model Fit and Item Factor Analysis: Overfactoring, Underfactoring, and a Program to Guide Interpretation. Multivariate Behav Res. 2018;53: 544–558. pmid:29683723
- 45.
Brown TA. Confirmatory factor analysis for applied research. New York: Guilford Press; 2006.
- 46. Carter SL, Murji S, Rourke SB, Shore D, Rourke BP. Cognitive complaints, depression, medical symptoms, and their association with neuropsychological functioning in HIV infection: A structural equation model analysis. Neuropsychology. 2003;17: 410–419. pmid:12959507
- 47. Olczyk P, Kusztal M, Gołębiowski T, Letachowicz K, Krajewska M. Cognitive Impairment in End Stage Renal Disease Patients Undergoing Hemodialysis: Markers and Risk Factors. Int J Environ Res Public Health. 2022;19: 2389. pmid:35206577
- 48. Drew DA, Weiner DE, Sarnak MJ. Cognitive Impairment in CKD: Pathophysiology, Management, and Prevention. Am J Kidney Dis. 2019;74: 782–790. pmid:31378643
- 49. Viggiano D, Wagner CA, Martino G, Nedergaard M, Zoccali C, Unwin R, et al. Mechanisms of cognitive dysfunction in CKD. Nat Rev Nephrol. 2020;16: 452–469. pmid:32235904
- 50. Goh ZS, Griva K. Anxiety and depression in patients with end-stage renal disease: impact and management challenges—a narrative review. Int J Nephrol Renov Diseajse. 2018;11: 93–102. pmid:29559806
- 51.
Lamberg-Allardt C, Kemi V. Interaction Between Calcium and Phosphorus and the Relationship to Bone Health. 1st ed. In: Gutiérrez OM, Kalantar-Zadeh K, Mehrotra R, editors. Clinical Aspects of Natural and Added Phosphorus in Foods. 1st ed. New York, NY: Springer; 2017. pp. 145–157.
- 52.
National Kidney Foundation. K/DOQI Clinical Practice Guidelines for Bone Metabolism and Disease in Chronic Kidney Disease. American Journal of Kidney Diseases. 2003.
- 53. Kardas P, Lewek P, Matyjaszczyk M. Determinants of patient adherence: A review of systematic reviews. Front Pharmacol. 2013;4: 91. pmid:23898295
- 54. Molinuevo JL, Rabin LA, Amariglio R, Buckley R, Dubois B, Ellis KA, et al. Implementation of subjective cognitive decline criteria in research studies. Alzheimer’s Dement. 2017;13: 296–311. pmid:27825022
- 55. Clark S, Farrington K, Chilcot J. Nonadherence in dialysis patients: Prevalence, measurement, outcome, and psychological determinants. Semin Dial. 2014;27: 42–49. pmid:24164416
- 56. Domensino AF, Evans J, van Heugten C. From word list learning to successful shopping: The neuropsychological assessment continuum from cognitive tests to cognition in everyday life. Appl Neuropsychol. 2022. pmid:35654744
- 57. Horowitz TS, Suls J, Treviño M. A Call for a Neuroscience Approach to Cancer-Related Cognitive Impairment. Trends Neurosci. 2018;41: 493–496. pmid:29907436
- 58. Denney NW, Pearce KA. A developmental study of practical problem solving in adults. Psychol Aging. 1989;4: 438–442. pmid:2619950
- 59. Crawford S, Channon S. Dissociation between performance on abstract tests of executive function and problem solving in real-life-type situations in normal aging. Aging Ment Heal. 2002;6: 12–21. pmid:11827618
- 60. Costa DSJ, Fardell JE. Why are objective and perceived cognitive function weakly correlated in patients with cancer? J Clin Oncol. 2019;37: 1154–1158. pmid:30920881
- 61. Van Dyk K, Ganz PA, Ercoli L, Petersen L, Crespi CM. Measuring cognitive complaints in breast cancer survivors: psychometric properties of the patient’s assessment of own functioning inventory. Support Care Cancer. 2016;24: 4939–4949. pmid:27432402
- 62. Li H, Tan C-C, Tan L, Xu W. Predictors of cognitive deterioration in subjective cognitive decline: evidence from longitudinal studies and implications for SCD-plus criteria. J Neurol Neurosurg Psychiatry. 2023. pmid:36868847
- 63. Chapman S, Rentería MA, Dworkin JD, Garriga SM, Barker MS, Avila-Rieger J, et al. Association of Subjective Cognitive Decline With Progression to Dementia in a Cognitively Unimpaired Multiracial Community Sample. Neurology. 2023;100: e1020–e1027. pmid:36450605
- 64. Henneghan AM, Van Dyk K, Kaufmann T, Harrison R, Gibbons C, Kesler SR. Measuring Self-Reported Cancer-Related Cognitive Impairment: Recommendations From the Cancer Neuroscience Initiative Working Group. J Natl Cancer Inst. 2021;113: djab027. pmid:33638633
- 65. Waldrop D, Irwin C, Nicholson C, Lee CA, Webel A, Fazeli PL, et al. The Intersection of Cognitive Ability and HIV: A Review of the State of the Nursing Science. J Assoc Nurses AIDS Care. 2021;32: 306–321. pmid:33449578
- 66. Kurella M, Luan J, Yaffe K, Chertow GM. Validation of the Kidney Disease Quality of Life (KDQOL) Cognitive Function subscale. Kidney Int. 2004;66: 2361–2367. pmid:15569327