Hospitalisation events in people with chronic kidney disease as a component of multimorbidity: parallel cohort studies in research and routine care settings

Chronic kidney disease (CKD) is a global health problem and is closely linked to adverse outcomes [1]. Compared to those without CKD, people with CKD are more likely to be hospitalised [2], develop complications while in hospital [3] and be re-admitted [4]. They have frequent contacts with health care services: clinic visits, blood tests, procedures and in the case of advanced CKD, the need for dialysis and/or kidney transplantation. Unplanned hospitalisations are additional, undesirable events with heightened anxiety, particularly when admissions are via emergency services. CKD is typically accompanied by multimorbidity (the co-occurrence of two or more long-term conditions: LTCs), which may have caused CKD, developed as direct or indirect complications of CKD or are unrelated [5]. Multimorbidity has been identified by the medical community as a major challenge which should be made a research priority [6]. Polypharmacy and high treatment burden are frequently experienced by these people, which reduce their quality of life [7, 8]. Hospital admissions may be directly linked to CKD (fluid overload, vascular access surgery) or other illnesses which occur in excess in CKD (infections, cardiovascular events [2]). In addition to therapeutic intervention, people are exposed to the risk of the healthcare environment (e.g. nosocomial infection and isolation). However, there is a paucity of evidence about the implications of multimorbidity and CKD in those with mild to moderate CKD nor do we know the relationship between different types of LTCs and CKD [9].

In this study, we sought to fill this evidence gap and to examine the associations between CKD, multimorbidity and emergency admissions to hospital. We hypothesised that people with multimorbidity would have high rates of emergency hospitalisation and that the rates would be higher when CKD was one of the LTCs. We also hypothesised that amongst those with CKD, subgroups with proven susceptibilities to adverse outcomes would be high risk: those with advanced CKD [2], those living in socioeconomically deprived areas [10] and those with low body weight [11]. A 2021 National Institute for Health Research policy paper on multimorbidity states that improving our understanding of combinations of conditions, or clusters, may help develop strategies to prevent ill health [12]. We therefore explored the associations between hospitalisation and combinations of conditions which have been shown to be associated with increased risk of adverse outcomes: multiple cardiometabolic conditions [13] (i.e. heart failure, hypertension, coronary heart disease, peripheral vascular disease, atrial fibrillation, diabetes and stroke), complex LTCs [14] (three or more conditions from three or more body systems) and mixed physical and mental conditions [15].

Two different types of cohort were studied: first, a prospective cohort study was used because it has extensive clinical phenotyping and there is extensive published data demonstrating its utility for studying multimorbidity [16‐20]. Second, because healthy volunteer bias can occur in research studies, a nationally representative primary care cohort generated from routine care records was used. This approach allowed us to confirm the generalisability of our findings to the general population.

We have studied emergency hospitalisations in a combined 2.1 million individuals from a prospective research study and a routine care database. Those with more LTCs had high rates of emergency hospitalisation and the risk was substantially increased by two to threefold in those with CKD (depending on the cohort). We also showed that the type of LTCs was important: those with CKD plus multiple cardiometabolic conditions, complex LTCs and physical and mental health LTCs were at heightened risk of hospitalisation.

Previous studies have identified a relationship between reduced eGFR and hospitalisation [2, 40]. Others have examined cohorts of patients with CKD and demonstrated that patients with LTCs are at high risk of hospitalisation [41]. What has not been studied before is how CKD relates to hospitalisation compared to, or in combination with, other LTCs. We have demonstrated that CKD is not equivalent to other LTCs as part of a multimorbidity count, but rather that individuals with CKD as one of the LTCs are particularly vulnerable to hospitalisation. In our study, people with CKD were frequently admitted with cardiovascular problems. This vulnerability to cardiovascular problems amongst people with CKD is well known, and it undoubtedly contributed to the overall high rates of hospitalisation in our study. The use of sodium-glucose cotransporter-2 inhibitors may prevent a proportion of these admissions in future [42]. Because numerous multimorbidity measures exist, researchers are encouraged to use a measure which suits their purpose [43]. Our study supports this message, emphasising that CKD is a critical condition in these measures and its significance should not be overlooked.

Amongst those with CKD, we found low eGFR and advanced age to be associated with high hospitalisation rates. However, regression analyses showed there was a disproportionately strong association between the number of LTCs and hospitalisation in those under the age of 50. We had hypothesised that the link between LTCs and hospitalisation would be strongest in those with more advanced CKD. We were surprised to find that the association between the number of LTCs and hospitalisation was less strong than the same association in those with mild to moderate CKD. It may be that because people with advanced CKD are primarily elderly, most of them have multiple LTCs and they have such a high baseline rate of hospitalisation, the influence of additional LTCs is attenuated. A previous study in 530,771 Canadians with CKD studied the link between the type of LTC and adverse outcomes [41]. As in our study, they found associations between LTCs and hospitalisation. In their study, this relationship was not unique to concordant LTCs, with associations also seen for discordant and mental health conditions. We have meaningfully extended this subject area by finding that certain combinations of LTCs were associated with heightened risk of hospitalisation (cardiometabolic, complex and physical/mental LTCs).

These findings are important for patients, carers, healthcare professionals and policy makers. As people with CKD and LTCs are known to be high-risk, clinicians caring for them should provide targeted monitoring. This does not mean monitoring of blood tests in isolation, as people with CKD and multimorbidity should have regular, thorough reviews of their clinical status, medications and preferences. CKD is common in the general population and although asymptomatic in the early stages, knowing which people have CKD may be helpful for healthcare planning. Combined physical and mental health conditions were a risk factor for hospitalisation in our study. Although we cannot assume that mental health support would prevent hospitalisations, people with mental and physical conditions are in need of psychological support [44], which has been proven to reduce depression and improve self-management [45].

Alternative strategies to hospitalisation exist, with improvements in quality of life for patients and cost reductions for healthcare systems [46]. Safe and effective care can be provided for outpatient management of illnesses such as pneumonia [47]. Alternatively, some people may not wish to be hospitalised and “Hospital at Home” services [48] and/or anticipatory care planning [49] may be better for some people. Clinicians should be mindful of these strategies when seeing people with multiple health conditions, and they should discuss the options during routine appointments, so their patients know what alternatives to emergency admission exist. Care models like these are not appropriate for all people or all illnesses, but when they are used, they can be beneficial for patients and less costly for healthcare systems. Structured interventions are, however, not always successful [50], and incentivisation may be necessary to reduce admissions [51].

Our study has several strengths. Using two large cohorts, we have expanded from a research setting with healthy volunteer bias [52] to a routine care database to confirm the generalisability of our findings in the general population. The use of linked healthcare records with universal coverage in the UK ensures we have identified most hospitalisations [53]. UK Biobank has been used extensively to study risk factors for health outcomes, but it sometimes draws criticism for not being representative of the general population [52]. Although event rates were higher in SAIL and the general population were older with more LTCs, the trends identified in UK Biobank were similar in SAIL.

Our study has some limitations. Although we have adjusted for age in our regression models, there is still a possibility of residual confounding. Some risk factors are undoubtedly on the causal pathways to our exposures and our outcome (e.g. obesity, alcohol use). It has not been possible in this study to unravel the complex relationships between all risk factors, exposures and the outcome. The relative lack of ethnic diversity in these particular UK cohorts means that the study should be replicated in other contexts. The eGFR equation we used (CKD-EPI) incorporates ethnicity, and there is not yet consensus in the medical community about whether this is appropriate [54]. LTCs in UK Biobank were self-reported and this risks the introduction of error. Although self-report may be less accurate for some conditions such as heart failure [55], it has been found to be a valid approach [56, 57]. CKD status, LTCs and covariates were only taken at baseline and we have not taken into account changes during follow-up. Data about severity of conditions would have been informative (particularly for some conditions such as heart failure and chronic obstructive pulmonary disease), but this was not available. Regardless, it would have been difficult to synthesise such information for 42 conditions. We employed counts of conditions rather than an index which assigns scores to conditions associated with greater morbidity. The evidence regarding whether simple counts or weighted measures are preferable is mixed [58] with some systematic reviews concluding that both are equally effective at predicting most outcomes [59]. A meta-review of six systematic reviews on this topic concluded there is a lack of a clear consensus, and it suggested selection of measures should depend on the purpose of any given study [60]. Our finding that CKD is linked to a heightened risk of hospitalisation may be transferable to other specific conditions, but we have not repeated it for each condition. We excluded 37.9% of the SAIL population without biochemistry data, who tended to be younger with fewer additional LTCs than those included. Sensitivity analysis showed that the difference in hospitalisation rates between non-CKD and CKD groups widened when participants without biochemistry were categorised as non-CKD. SAIL participants were lost to follow-up if they move away from Wales, which means some hospitalisation events may have not been recorded. The rates of missing data were high for some variables in SAIL. Multiple imputation was used, with similar results obtained in complete case analysis.

People with increasing multimorbidity count are therefore at high risk of emergency hospitalisation, and the rates are two to threefold higher when CKD is present. People with CKD at heightened risk of hospitalisation should be targeted by research aimed at addressing emergency hospital admissions.