Loneliness, social isolation and incident chronic kidney disease among patients with diabetes

Study design and population

The UK Biobank study is a large population-based cohort, which enrolled over 500 000 participants representative of the general population. Participants were recruited at 22 assessment centres across the UK between 2006 and 2010. Comprehensive information about the UK Biobank’s design and study population has been described previously.14

In the current study, our analysis focused solely on participants with diabetes at baseline, which was determined either by the onset time of diabetes (on or before the date of attending the assessment centre) or through a self-reported history of doctor-diagnosed diabetes (n=24 593) (online supplemental table S1). We identified a total of 18 972 patients with diabetes in our study after excluding 5552 participants with prevalent CKD and 69 participants with incomplete loneliness or isolation data (figure 1).

Request permission

Loneliness and Isolation Scales

We developed the Scales of Loneliness and Social Isolation using baseline self-reported questionnaire according to previous UK Biobank studies.15 First, the Loneliness Scale was evaluated by asking two questions: ‘Do you often feel lonely?’ (1 point for yes) and ‘How often can you confide in someone close to you?’ (1 point for no more than once a month). Second, the Social Isolation Scale was assessed by asking three questions: (1) ‘Including yourself, how many people are living together in your household?’ (1 point for living alone); (2) ‘How often do you visit friends or family or have them visit you?’ (1 point for having friends and family visit less than once a month); (3) ‘Which of the following leisure/social activities do you engage in once a week or more often?’ (1 point for not participating in social activity at least once per week). Those answers with low-risk factors were coded as 0. The sum of the responses to loneliness questions resulted in a scale ranging from 0 to 2 and social isolation questions in a scale ranging from 0 to 3. The detailed information on the scoring assessment for loneliness and social isolation can be found in online supplemental table S2.

Ascertainment of CKD

We defined CKD to encompass all stages (1–5) and used the International Statistical Classification of Diseases, Tenth Revision (ICD-10) codes (N18, N18.1, N18.2, N18.3, N18.4, N18.5 and N18.9) to ascertain incident CKD cases.16 Hospital inpatient records were obtained through linkages with the Hospital Episode Statistics for England, Scottish Morbidity Records for Scotland and Patient Episode Database for Wales. CKD-related deaths were identified by connecting to the death registry. The follow-up time in this study was calculated from the baseline date to the diagnosis of CKD, the censoring date or death (until May 2021), whichever occurred first. Detailed information on the ascertainment of outcomes is available online (https://biobank.ctsu.ox.ac.uk/showcase/label.cgi?id=2000).

To identify prevalent CKD, we used the ICD-10 codes mentioned before, or albuminuria above 3 mg/mmol, or the estimated glomerular filtration rate (eGFR) below 60 mL/min/1.73 m² at baseline.

Assessment of covariates

Self-reported information was collected by a touch-screen questionnaire, including age, sex, race, smoking status (never, past and current), alcohol consumption status (never, past and current) and medication use for cholesterol, blood pressure or diabetes. The Townsend deprivation index was gathered from local National Health Service Primary Care Trust registries and the name of the recruitment centre based on an individual’s postcode, a composite measure of deprivation considering factors including unemployment, non-car ownership, non-home ownership and household overcrowding. A lower value represents a higher socioeconomic status. During the assessment visit, both height and weight were recorded, with body mass index (BMI) being determined by dividing the weight in kilograms by the square of height in metres (kg/m²). eGFR was calculated by the CKD-EPI Creatinine Equation (2021) using the blood sample collected at baseline. Physical activity was measured by minutes of moderate or vigorous physical activity or an equivalent combination per week, and physical activity was defined according to the guideline as >150 min of moderate-intensity activity per week or >75 min of vigorous activity per week or an equivalent combination per week. Healthy diet scores were computed using a 0–5 scale, with 1 point assigned for each of these criteria: daily intake of fruits and vegetables, with at least 3 servings of each; weekly consumption of fish, with 2 or more servings; processed meat intake limited to less than 1 serving per week and unprocessed meat consumption restricted to 1.5 servings or fewer per week. Depressive symptoms were evaluated using two questions from the Patient Health Questionnaire-2, which addressed the frequency of depressed mood and disinterest or absence of enthusiasm in the previous 2 weeks. Participants’ responses of ‘not at all’, ‘several days’, ‘more than half the days’ or ‘nearly every day’ were assigned scores of 0, 1, 2 and 3, respectively. By summing the scores for both items, the overall Depression Score was determined, ranging from 0 to 6. Diabetes duration was defined via the health records. We assigned a missing indicator category for categorical variables with missing covariates, and we used mean values for continuous variables with missing data. More in-depth information on these measurements can be found on the UK Biobank’s website (https://biobank.ctsu.ox.ac.uk/showcase).

Statistical analysis

Continuous variables are presented as mean (SD) while categorical variables are expressed as counts with percentages according to the Scale of Loneliness and Social Isolation. Statistical significance was determined using χ² tests for categorical variables and general linear models after adjusting for age (except age) for continuous variables. Cox proportional hazards models were employed to assess the association between the Loneliness Scale or Isolation Scale or individual loneliness factors or individual isolation factors and the risk of CKD separately, with follow-up years serving as the underlying time metric. The proportional hazards assumption was evaluated using Schoenfeld residuals, and all analyses met the required criteria. Confounders were selected based on the association with outcome and exposure, as well as the review of literature.9 We adjusted several potential confounders in these models. Model 1 was adjusted for basic demographic and socioeconomic factors: age, sex, race and Townsend deprivation index. Model 2 was built on Model 1 by further adjusting for lifestyle factors, including physical activity, healthy diet score, smoking status, alcohol use and BMI. Model 3, based on Model 2, included additional adjustments for medical and psychological factors: eGFR, diabetes duration, the use of diabetes medication, cholesterol-lowering medication, antihypertensive medication and depression score. This model aimed to provide a comprehensive adjustment for medical and psychological factors that could further confound the relationship between our variables of interest and CKD risk. By progressively adjusting for different types of variables, we aimed to disentangle the complex interplay of demographic, lifestyle, medical and psychological factors influencing CKD risk in patients with diabetes. To assess the importance of loneliness or isolation in predicting CKD, we analysed the relative importance of loneliness or isolation alongside some traditional risk factors by calculating the R² values of the Cox models,17 and we tested the results’ consistency by calculating the explainable log-likelihood that was attributable to each risk factor.

Furthermore, we performed stratified analyses by sex (women or men), race (white or non-white), Townsend deprivation index (quintiles 1, quintiles 2–4, quintiles 5), obesity (no or yes), smoking status (current or non-current smoker), alcohol use (current or non-current drinker), eGFR (<90 or ≥90 mL/min/1.73 m²), healthy diet score (<median or ≥median), depression score (<3 or ≥3), regular physical activity (no or yes), duration of diabetes (<5, 5–10 and ≥10 years), antihypertensive drug use (no or yes), diabetes drug use (no or yes) and cholesterol-lowering drug use (no or yes). To assess interactions between the Loneliness Scale and these factors, multiplicative interaction was evaluated by adding interaction terms to the original Cox models. In the sensitivity analyses, we additionally adjusted for feeling lonely and confiding in someone close mutually in the model. And we further adjusted for Social Isolation Scale in this model to assess the association between individual loneliness factors and the risk of CKD among patients with diabetes at baseline.

All statistical analyses were conducted using SAS V.9.4 TS Level 1M7 (SAS Institute) and R V.3.6.1 (R Foundation for Statistical Computing, Vienna, Austria). A two-tail p value less than 0.05 was considered statistically significant.

Study population

The baseline information for patients with diabetes without pre-existing CKD, categorised by the Loneliness Scale or Social Isolation Scale is displayed in table 1 and online supplemental table S3. Of the 18 972 patients with diabetes, 61.9% had a loneliness score of 0, 29.1% scored 1 and 9.0% had a score of 2 (table 1). For participants with a social isolation score of 0, 1, 2 or 3, the respective percentages were 45.5%, 41.7%, 11.6% and 1.2% (online supplemental table S3). Individuals with lower scores on the Loneliness or Social Isolation Scales were generally older, less likely to be white, exhibited a lower Townsend deprivation index and were more likely to be non-smokers, current drinkers and maintain a healthy diet compared with those with higher scores on these scales. Furthermore, patients with lower loneliness or social isolation scores demonstrated a tendency towards lower BMI, depression scores and higher eGFR and reduced likelihood of insulin and cholesterol-lowering medication use (all p<0.05).

Association of loneliness or social isolation with CKD risk in patients with diabetes

During a median follow-up of 10.8 years, a total of 1127 incidents with CKD events were observed among patients with diabetes at baseline. Our results indicated that the Loneliness Scale, rather than the Isolation Scale, had a significant association with an increased risk of CKD in patients with diabetes (table 2). In the sex, age, race and Townsend deprivation index adjusted model, a Higher Loneliness Scale was significantly associated with a higher risk of CKD, with hazard ratios (HRs) (95% CI) of 0 (reference), 1.21 (1.06 to 1.38) and 1.42 (1.17 to 1.73) across the Loneliness Scales of 0, 1 and 2 (P-trend <0.001). After further adjustment for physical activity, healthy diet score, smoking, alcohol use, BMI, the association was slightly attenuated but still significant with HRs (95% CI) of 1.16 (1.02 to 1.33) and 1.34 (1.09 to 1.63) for participants with a Loneliness Scale of 1 and 2 (p-trend <0.001). The observed association showed a modest attenuation if we further included eGFR, diabetes duration, the use of diabetes medication, cholesterol-lowering medication, antihypertensive medication and depression score into the model with HRs (95% CI) of 1.11 (0.95 to 1.28) and 1.25 (1.00 to 1.57) for participants with a Loneliness Scale of 1 and 2 (P-trend=0.019) (table 2). In addition, the associations between individual loneliness or social isolation factors and the risk of CKD are shown in table 3 and online supplemental table S4. Regarding the individual loneliness factors, we found that the feeling-lonely aspect, rather than confiding in someone close, was significantly associated with an increased CKD risk. In the multivariable-adjusted model, when comparing participants without feeling lonely, those who felt lonely exhibited an HR (95% CI) of 1.22 (1.04 to 1.43) for CKD risk (table 3). In the sensitivity analyses, similar associations were observed when we adjusted for feeling lonely and confiding in someone close mutually in the model. The results were not changed when we further adjusted Social Isolation Scale in this model (online supplemental table S5). However, regarding the individual social isolation factors, we did not observe a significant association between any individual factors of social isolation and the risk of CKD (online supplemental table S4).

Furthermore, we conducted stratified analyses according to potential risk factors (online supplemental table S6) to evaluate whether these covariables modified the relationship between the Loneliness Scale and the risk of CKD. However, we did not find significant interactions between these covariates and the Scale of Loneliness on the risk of CKD.

The relative importance of loneliness and social isolation factors compared with traditional risk factors in predicting CKD events among patients with diabetes

We evaluated the relative importance of each loneliness factor and social isolation factor in predicting CKD incidence among patients with diabetes in comparison to other traditional risk factors (figure 2). In our investigation of specific loneliness factors, we found that feeling lonely held the 3rd position in relation to CKD risk among individuals with diabetes, while the other factor of loneliness (confiding in someone else) was the 10th position. The predictive capacity of feeling lonely for CKD was surpassed by HbA1c and systolic blood pressure. Nevertheless, feeling lonely ranked higher than several traditional risk factors including BMI, smoking, physical activity, alcohol, dietary scores and depression scores, Townsend deprivation index and low-density lipoprotein-cholesterol in predicting CKD. As for the exploration of particular social isolation elements, the frequency of social visits ranked 5th, though the association was not significant. The remaining two factors ranked even lower, specifically 11th and 15th, with non-significant associations. The results with the use of explained log-likelihood were consistent with the results obtained using the explained relative risk (R²) models (online supplemental figure S1).

Request permission

Main findings

In this large prospective cohort, the associations of loneliness and social isolation with the risk of incident CKD were investigated in individuals with diabetes. We observed that a higher loneliness scale, but not social isolation, was significantly associated with a 25% higher risk of CKD among patients with diabetes. Among the individual loneliness factors, the sense of feeling lonely, as opposed to having someone to confide in, emerged as the primary factor contributing to the elevated risk of CKD. Compared with individuals who did not feel lonely, those experiencing feelings of loneliness demonstrated a 22% higher probability of CKD development. In our analysis, we observed a sequential decrease in the effect sizes from Model 1 to Model 3. This trend can be attributed to the progressive inclusion of various confounding factors across the models. As we adjusted for more variables that could independently affect CKD risk, the isolated impact of loneliness factors became more refined and potentially more representative of their true effect. In addition, feeling lonely demonstrated greater relative importance in predicting CKD compared with traditional risk factors such as BMI, smoking, physical activity and diet.

Our study, for the first time, showed that loneliness was significantly related to a higher risk of CKD in patients with diabetes. Our findings are partly supported by previous studies which have primarily focused on the relationship between loneliness and cardiovascular health. For instance, studies in the general population have demonstrated that loneliness is associated with higher risks of CVD events.9 18 We recently demonstrated a significant association between the Loneliness Scale and increased risks of total CVD and coronary heart disease among individuals with diabetes, as evidenced by higher HRs of 1.15 (95% CI 1.07 to 1.25) and 1.38 (95% CI 1.23 to 1.54) for participants with loneliness scores of 1 and 2, respectively, compared with those with the lowest loneliness score (0).9 Moreover, in an English longitudinal study, loneliness has been linked to an increased risk of developing both coronary heart disease and stroke, irrespective of traditional risk factors associated with CVDs.18 Another cohort study among older women from the USA revealed that a high loneliness score corresponded to a 14% increase in the hazard of CVD compared with a low loneliness score. The mechanisms linking loneliness to CVD, such as dysregulation of the autonomic nervous system and increased systemic inflammation, are also pertinent in the context of CKD.19 This overlap in pathophysiological pathways suggests that factors influencing CVD might similarly impact CKD, especially in patients with diabetes where vascular health is already compromised.

Given the established relationship between loneliness and CVD, it’s crucial to explore how these psychosocial stressors extend their influence on CKD. Both CVD and CKD are vascular complications often observed in diabetes, sharing common risk factors such as hypertension and poor glycaemic control. Our investigation into the impact of loneliness on CKD stems from this interconnection, hypothesising that the psychosocial stressors contributing to CVD might similarly influence CKD progression. In this study, we recognised the dual role of loneliness in CKD risk among patients with diabetes. As a confounder, loneliness, independently linked with both diabetes and CKD, emerges as a significant risk factor for type 2 diabetes.20 This emphasises its importance in CKD risk analysis. Additionally, loneliness might also act as an effect mediator: diabetes-related factors like reduced mobility can increase loneliness, which in turn may exacerbate CKD progression.21 In addition, we investigated the associations between different aspects of loneliness (ie, feeling lonely and the frequency of confiding in someone close) and the risk of CKD among patients with diabetes. Intriguingly, the subjective experience of feeling lonely, but not the frequency of confiding in someone close, showed a significant association with the risk of CKD. The subjective experience of feeling lonely encompasses emotional and cognitive aspects, reflecting an individual’s perception of social isolation and a lack of connection with others.22 Moreover, we compared the relative importance of each aspect of loneliness with traditional risk factors in predicting CKD among patients with diabetes. We found that the relative strength of feeling lonely was greater than several traditional risk factors such as BMI, smoking, physical activity and diet, highlighting the importance of considering psychosocial factors in preventing CKD complications in patients with diabetes, beyond traditional risk factors. Our findings support the recent statement by the AHA that SDOH is important among patients with diabetes in preventing vascular complications.

Incorporating psychosocial factors into the prevention of vascular complications in patients with diabetes offers several potential benefits. First, unlike traditional risk factors that predominantly affect the biological aspect, psychosocial factors can impact health behaviours and self-care practices.23 Feeling lonely may lead to decreased motivation and poor adherence to diabetes management strategies. Second, psychosocial factors have been linked to cardiovascular health.13 18 Loneliness and other psychosocial stressors have been associated with dysregulation of the autonomic nervous system and the hypothalamic–pituitary–adrenal axis, impaired immune functioning24 and increased systemic inflammation. These mechanisms, in conjunction with traditional risk factors, can contribute to the development and progression of vascular complications, such as CKD, in patients with diabetes. Even though the precise mechanisms between loneliness and CKD remain unclear, several other potential explanations are worthy of consideration. Loneliness is related to impaired self-regulation, such as physical inactivity, alcohol abuse, sleep disturbances and physiological changes in cardiovascular health, such as elevated blood pressure.25 All these mechanisms may play a role in the development of CKD in patients with diabetes. Therefore, while traditional risk factors remain critical, the multifaceted impact of loneliness highlights its unique and potent role in CKD development. Recognising this, it becomes imperative to incorporate psychosocial health management as part of comprehensive CKD prevention strategies in diabetic care.

We did not find a significant association between social isolation and CKD among patients with diabetes. This finding diverges from several studies investigating social isolation/support and kidney health outcomes.26 27 For example, a nationally representative longitudinal survey, China Health and Retirement Longitudinal Study, reported a significant association between social isolation and increased risks of rapid eGFR decline and CKD onset in middle-aged and older adults with normal kidney function in mainland China.26 Additionally, an update about social support and CKD mentioned that social support emerges as an important modifiable risk factor across chronic diseases, including end-stage renal disease.27 One possible explanation for this discrepancy is that patients with diabetes may have distinct social support needs or coping mechanisms compared with the broader population, thereby attenuating the impact of social isolation on CKD risk. Alternatively, the measures used to assess social isolation in our study may not have fully captured the complexity of this construct in relation to CKD risk among patients with diabetes. However, our results align with several prior investigations. For example, a systematic review and meta-analysis found no significant link between low structural social support and the prevalence of myocardial infarction in healthy populations.28 In addition, our findings are also in line with our recent findings that social isolation is not related to CVDs among individuals with diabetes.9 Loneliness and social isolation, though inter-related, are separate notions. Loneliness embodies the subjective sense of solitude, disconnection or absence of companionship, often arising from a perceived deficiency in one’s social connections. In contrast, social isolation denotes an objective condition characterised by a limited number of social ties, a constrained social network or sporadic social engagement. Loneliness is an emotional state, while social isolation can be quantified as a facet of an individual’s social existence.29 Notably, a person may experience feeling lonely without social isolation. Our results indicated that the qualitative aspects of the social environment (emotional) might hold greater significance than the quantitative dimensions (behavioural) when assessing the risk of CKD among individuals with diabetes. The major strengths include its prospective design, the sizeable cohort of patients with diabetes with available data on loneliness and isolation, and the extensive, detailed information on covariates in this population.

Limitations

However, we recognise several potential limitations. First, the loneliness and social isolation measures were based on simple questions, which may not fully capture the complexities of social construction and interaction. Nonetheless, these measures have been employed in numerous prior studies, implying their effectiveness in population research.18 Second, generalisability is another limitation, as over 90% of the UK Biobank cohort is composed of people of white ancestry. Further investigations are needed for other racial/ethnic groups. Third, UK Biobank participants are more likely to exhibit healthier behaviours compared with the overall UK population. However, this would not compromise the internal validity of this study. Fourth, as an observational study, we cannot infer causality between loneliness and CKD risk. Fifth, our analysis might not fully capture the complex interplay between the specific biological mechanisms of type 1 and type 2 diabetes, and how these distinct mechanisms might differentially interact with psychosocial factors. Their subsequent influence on CKD risk represents an area that warrants further, more detailed investigation. Finally, despite accounting for several potential confounders, the possibility of residual confounding cannot be entirely excluded.

Implications

Our findings indicate that a higher loneliness scale, rather than social isolation, is significantly associated with a higher CKD risk among individuals with diabetes. Furthermore, the sensation of feeling lonely exhibited higher relative strength in predicting CKD in comparison to traditional risk factors such as BMI, smoking, physical activity and diet, highlighting the importance of considering psychosocial factors in preventing vascular complications in patients with diabetes.