Loneliness, Social Isolation, and Effects on Cognitive Decline in Patients with Dementia: A Retrospective Cohort Study Using Natural Language Processing

Loneliness, Social Isolation, and Effects on Cognitive Decline in Patients with Dementia: A Retrospective Cohort Study Using Natural Language Processing | medRxiv /* */ /* */ <!-- <!-- /*! * yepnope1.5.4 * (c) WTFPL, GPLv2 */ (function(a,b,c){function d(a){return"[object Function]"==o.call(a)}function e(a){return"string"==typeof a}function f(){}function g(a){return!a||"loaded"==a||"complete"==a||"uninitialized"==a}function h(){var a=p.shift();q=1,a?a.t?m(function(){("c"==a.t?B.injectCss:B.injectJs)(a.s,0,a.a,a.x,a.e,1)},0):(a(),h()):q=0}function i(a,c,d,e,f,i,j){function k(b){if(!o&&g(l.readyState)&&(u.r=o=1,!q&&h(),l.onload=l.onreadystatechange=null,b)){"img"!=a&&m(function(){t.removeChild(l)},50);for(var d in y[c])y[c].hasOwnProperty(d)&&y[c][d].onload()}}var j=j||B.errorTimeout,l=b.createElement(a),o=0,r=0,u={t:d,s:c,e:f,a:i,x:j};1===y[c]&&(r=1,y[c]=[]),"object"==a?l.data=c:(l.src=c,l.type=a),l.width=l.height="0",l.onerror=l.onload=l.onreadystatechange=function(){k.call(this,r)},p.splice(e,0,u),"img"!=a&&(r||2===y[c]?(t.insertBefore(l,s?null:n),m(k,j)):y[c].push(l))}function j(a,b,c,d,f){return q=0,b=b||"j",e(a)?i("c"==b?v:u,a,b,this.i++,c,d,f):(p.splice(this.i++,0,a),1==p.length&&h()),this}function k(){var a=B;return a.loader={load:j,i:0},a}var l=b.documentElement,m=a.setTimeout,n=b.getElementsByTagName("script")[0],o={}.toString,p=[],q=0,r="MozAppearance"in l.style,s=r&&!!b.createRange().compareNode,t=s?l:n.parentNode,l=a.opera&&"[object Opera]"==o.call(a.opera),l=!!b.attachEvent&&!l,u=r?"object":l?"script":"img",v=l?"script":u,w=Array.isArray||function(a){return"[object Array]"==o.call(a)},x=[],y={},z={timeout:function(a,b){return b.length&&(a.timeout=b[0]),a}},A,B;B=function(a){function b(a){var a=a.split("!"),b=x.length,c=a.pop(),d=a.length,c={url:c,origUrl:c,prefixes:a},e,f,g;for(f=0;f<d;f++)g=a[f].split("="),(e=z[g.shift()])&&(c=e(c,g));for(f=0;f<b;f++)c=x[f](c);return c}function g(a,e,f,g,h){var i=b(a),j=i.autoCallback;i.url.split(".").pop().split("?").shift(),i.bypass||(e&&(e=d(e)?e:e[a]||e[g]||e[a.split("/").pop().split("?")[0]]),i.instead?i.instead(a,e,f,g,h):(y[i.url]?i.noexec=!0:y[i.url]=1,f.load(i.url,i.forceCSS||!i.forceJS&&"css"==i.url.split(".").pop().split("?").shift()?"c":c,i.noexec,i.attrs,i.timeout),(d(e)||d(j))&&f.load(function(){k(),e&&e(i.origUrl,h,g),j&&j(i.origUrl,h,g),y[i.url]=2})))}function h(a,b){function c(a,c){if(a){if(e(a))c||(j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}),g(a,j,b,0,h);else if(Object(a)===a)for(n in m=function(){var b=0,c;for(c in a)a.hasOwnProperty(c)&&b++;return b}(),a)a.hasOwnProperty(n)&&(!c&&!--m&&(d(j)?j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}:j[n]=function(a){return function(){var b=[].slice.call(arguments);a&&a.apply(this,b),l()}}(k[n])),g(a[n],j,b,n,h))}else!c&&l()}var h=!!a.test,i=a.load||a.both,j=a.callback||f,k=j,l=a.complete||f,m,n;c(h?a.yep:a.nope,!!i),i&&c(i)}var i,j,l=this.yepnope.loader;if(e(a))g(a,0,l,0);else if(w(a))for(i=0;i (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0];var j=d.createElement(s);var dl=l!='dataLayer'?'&l='+l:'';j.src='//www.googletagmanager.com/gtm.js?id='+i+dl;j.type='text/javascript';j.async=true;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-P4HH5NV'); Skip to main content Home About Submit ALERTS / RSS Search for this keyword Advanced Search Loneliness, Social Isolation, and Effects on Cognitive Decline in Patients with Dementia: A Retrospective Cohort Study Using Natural Language Processing View ORCID Profile James A. C. Myers , View ORCID Profile Tom Stafford , View ORCID Profile Ivan Koychev , View ORCID Profile Robert Perneczky , View ORCID Profile Oliver Bandmann , View ORCID Profile Nemanja Vaci doi: https://doi.org/10.1101/2025.02.10.25321986 James A. C. Myers 1 School of Psychology, University of Sheffield , ICOSS, 219 Portobello, Sheffield S1 4DP, United Kingdom 2 Division of Neuroscience, University of Sheffield , Western Bank, Sheffield S10 2TN, United Kingdom 9 Oxford Health NHS Foundation Trust, Warneford Hospital , Warneford Lane, Headington, Oxford OX3 7JX, United Kingdom 10 Rotherham, Doncaster and South Humber NHS Foundation Trust , Woodfield House, Tickhill Road Site, Weston Road, Balby, Doncaster, DN4 8QN, UK MSc Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for James A. C. Myers Tom Stafford 1 School of Psychology, University of Sheffield , ICOSS, 219 Portobello, Sheffield S1 4DP, United Kingdom 2 Division of Neuroscience, University of Sheffield , Western Bank, Sheffield S10 2TN, United Kingdom PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Tom Stafford Ivan Koychev 3 Department of Psychiatry, University of Oxford, Warneford Hospital , Warneford Lane, Headington, Oxford OX3 7JX, United Kingdom 9 Oxford Health NHS Foundation Trust, Warneford Hospital , Warneford Lane, Headington, Oxford OX3 7JX, United Kingdom PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Ivan Koychev Robert Perneczky 2 Division of Neuroscience, University of Sheffield , Western Bank, Sheffield S10 2TN, United Kingdom 4 School of Medicine and Population Health, University of Sheffield , Beech Hill Road, Sheffield S10 2RX, United Kingdom 5 Department of Psychiatry and Psychotherapy, LMU Hospital , Nußbaumstraße 7, 80336 Munich, Germany 6 German Center for Neurodegenerative Diseases (DZNE) , Feodor-Lynen-Straße 17, 81377 Munich, Germany 7 Munich Cluster for Systems Neurology (SyNergy) , Feodor-Lynen-Straße 17, 81377 Munich, Germany 8 Ageing Epidemiology Research Unit (AGE), School of Public Health, Imperial College London , 90 Wood Lane, London W12 0BZ, United Kingdom PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Robert Perneczky Oliver Bandmann 2 Division of Neuroscience, University of Sheffield , Western Bank, Sheffield S10 2TN, United Kingdom 4 School of Medicine and Population Health, University of Sheffield , Beech Hill Road, Sheffield S10 2RX, United Kingdom PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Oliver Bandmann Nemanja Vaci 1 School of Psychology, University of Sheffield , ICOSS, 219 Portobello, Sheffield S1 4DP, United Kingdom 2 Division of Neuroscience, University of Sheffield , Western Bank, Sheffield S10 2TN, United Kingdom 9 Oxford Health NHS Foundation Trust, Warneford Hospital , Warneford Lane, Headington, Oxford OX3 7JX, United Kingdom 10 Rotherham, Doncaster and South Humber NHS Foundation Trust , Woodfield House, Tickhill Road Site, Weston Road, Balby, Doncaster, DN4 8QN, UK PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Nemanja Vaci For correspondence: n.vaci{at}sheffield.ac.uk Abstract Full Text Info/History Metrics Data/Code Preview PDF Abstract INTRODUCTION The study aimed to compare cognitive trajectories between patients with reports of social isolation and loneliness and those without. METHODS Reports of social isolation, loneliness, and Montreal Cognitive Assessment (MoCA) scores were extracted from dementia patients’ medical records using Natural Language Processing models and analysed using mixed-effects models. RESULTS Lonely patients (n = 382), compared to controls (n = 3912), showed an average MoCA score that was 0.83 points lower throughout the disease (p = 0.008). Socially isolated patients (n = 523) experienced a 0.21 MoCA points per year faster rate of cognitive decline in the six months before diagnosis (p = 0.029), but were comparable to controls before this period. This led to average MoCA scores that were 0.69 MoCA points lower at diagnosis (p = 0.011). DISCUSSION Lower cognitive levels in lonely and socially isolated patients suggest that these factors may contribute to dementia progression. Introduction Social isolation (SI) and loneliness are recognised as priority public health problems, 1 showing impact on physical and mental health, 2 – 7 with effects on mortality comparable to smoking and obesity. 8 Population attributable fraction, a measure that combines relative risk and prevalence of social isolation in a population, estimates that low social contact in older people explains up to 4·2% of the risk for dementia development. 9 – 12 While much of the research has focused on SI and loneliness as risk factors for dementia 13 , these factors may also manifest as symptoms of the disease itself 14 , particularly in its early stages. In this study, we examine how SI and loneliness, when reported in the presymptomatic or symptomatic stage of dementia disease, influence the cognition of patients and the progression of the disease. SI and loneliness are related but distinct concepts. SI is operationalised as an objective lack of social and support networks, while loneliness is seen as a negative subjective feeling resulting from the discrepancy between desired and actualised social connections and closeness to other people 1 . Participation in wider social network structures is associated with higher global cognitive function and moderates the association between cognitive functioning and amygdala volume, an indicator of neuropathological progression in Alzheimer’s disease (AD), in cognitively normal people as well as patients with Mild Cognitive Impairment (MCI) and early-stage AD 15 , 16 . These findings support the idea that SI influences cognitive reserve, by moderating cognitive function despite indicators of AD neuropathology. 17 , 18 Feelings of loneliness have been shown to predict dementia onset 19 and are associated with higher Amyloid Burden 7 in cognitively normal people, as well as with higher rates of cognitive decline 20 , 21 in patients with MCI, but not in participants with the diagnosis of AD 22 , 23 . Despite these studies, findings for the effect of SI and loneliness on cognitive trajectories are mixed 24 , 25 . The main challenge in reliably estimating the impact of SI and loneliness symptoms on the cognitive function of dementia patients lies in the need for large-scale longitudinal data 26 , which is essential to capture the progressive nature of the disease. In this study, we used Electronic Health Records (EHRs) to investigate the effect of SI and loneliness reports on the cognitive trajectories of patients with dementia diagnosis. Using textual records of patient—care provider interaction, we identify documents which discuss patients’ reports about SI and loneliness and combine them with longitudinal measures of cognitive functionality. Using data from more than 4800 patients, we estimate the cognitive trajectories of patients with reports of social isolation and loneliness and compare them to trajectories of patients without such reports, while testing cognitive changes after the first social isolation or loneliness reports. Given previous findings on the effect of SI and loneliness on the risk of dementia 9 , 12 , 20 , we expected that perceived SI and loneliness are associated with more severe cognitive decline throughout the disease. Methods Study design The study followed a retrospective cohort design defined through the extraction of information from EHRs collected by Oxford Health NHS Foundation Trust in the United Kingdom. The data is accessible through the UK-CRIS system, maintained by Akrivia Health ( https://akriviahealth.com/ ). The system allows access to structured information, e.g. demographic information and diagnosis code, as well as unstructured textual information, such as clinical records. These documents collect free-text information on the history of mental disorders under treatment, relevant cognitive assessments, and any other clinically relevant discussion between services and support that went on throughout the treatment. Cohort information Our cohort included data from all patients with a diagnosis of Alzheimer’s or other forms of dementia (ICD codes: F00 – F00.9, F01, F02, F03, G30, but excluding F06.7 as MCI is rarely followed up in Memory Clinic). The full cohort included 34,469 patients who collectively contributed 6,388,715 medical documents from 06/03/2008 to 25/06/2022. To use the information that resides in rich clinical texts we developed Natural Language Processing (NLP) models that extracted information on cognitive health assessments in dementia and reports of loneliness and social isolation made by patients, carers, and clinical staff. Cognitive Outcomes The main analysis used the Montreal Cognitive Assessment (MoCA) 27 measure, while in supporting materials we report analysis using Mini-Mental State Examination (MMSE) 28 scores. Both measures are widely used tools for assessing cognitive function, particularly in patients with dementia. MoCA detects mild cognitive impairments and early-stage dementia through its heavier emphasis on frontal and parietal function 29 , while MMSE, even though not as sensitive as MoCA, captures moderate to severe cognitive impairment. Both measures are frequently used in clinical practice and research 30 . MoCA scores below the cut-off point of 26 points are taken as suggestive of MCI, below 17 as suggestive of moderate impairment, and under 10 as suggestive of severe cognitive impairment. The minimum clinically important difference the smallest change in the outcome that patients would find significant, is reported to be between 0.01 and 2 points depending on the severity of the disease 31 . Procedure We used structured and unstructured data from EHRs in this study. The unstructured data covered all textual records for the defined patient cohort, while structured data included information about the sex, ethnicity, and date of birth of patients, their marital and accommodation status, and ICD-10 codes for dementia and depression diagnosis (F32.0 to F34.1). To extract information about the cognition of patients, we used the previously published NLP model 32 (for previous work on methodological considerations and description of the mental health EHR data see references 33 – 37 ). NLP model for reports of SI and Loneliness A novel NLP model was developed for the SI and loneliness reports. The model was implemented in Python and processed textual records for reports of SI and loneliness in two stages: pattern matching and classification stage. In the pattern matching stage, we used a statistical model for word processing from the Spacy library to identify words that describe SI and loneliness. This allowed us to find all documents including expressions such as “loneliness”, “social isolation”, “living alone”, etc. In the next, classification, stage, we used sentence transformer models from Huggingface’s Spacy-Setfit library to process and classify sentences with SI and loneliness mentions. Sentence transformers 38 are types of neural network models that produce numerical representations of sentence and paragraph-level linguistic content. This vector space encodes semantic relationships, allowing us to identify and categorise semantically similar sentences. We trained sentence transformers to classify sentences with reports of SI and loneliness into four different categories: a) SI, b) loneliness, c) non-informative isolation, and d) non-informative sentences. Reports that mention lack of social contact, living alone, and being away from family, or that mention barriers in receiving support from family, were used as an indication of social isolation 3 . Loneliness was operationalized as consisting of reports on emotional aspects of feeling lonely and suffering due to the lack of social connections 39 . The non-informative isolation category included reports of temporary and physical isolation (e.g. “isolated fall” or “isolating in the tv-room”), while the non-informative sentences category covered all incorrectly included sentences from the pattern matching stage forwarded to the sentence classification stage (see Table 1 for examples of sentences and NLP categories). View this table: View inline View popup Download powerpoint Table 1: Example of sentences that express the social isolation and loneliness of patients and used categorisation when training the NLP model The full model was trained on a randomly selected subset of 11,000 medical documents from the corpus. The terms for the pattern matching stage were derived from a combination of the UCLA Loneliness Scale 40 and linguistic phrases observed in the training set, specifically those referring to reports of SI and loneliness. Sentences identified by the pattern matching in the training data were then used to train the classifiers. NLP accuracy To evaluate the model’s performance in identifying reports of SI and loneliness, we conducted a manual annotation process using an unseen sample of 5,000 documents. The annotation was performed independently by the first and last authors. In instances of disagreement, the annotations were discussed collaboratively, and a consensus was reached to ensure consistent labelling. The annotated data served as the ground truth for evaluating the model’s classification performance. Standard classifier performance metrics were employed, including Sensitivity, Specificity, and Balanced or F1 Accuracy. Across four different categories, the NLP model achieved average F1 accuracy of .74, reaching .83 accuracy for SI sentences (Sensitivity: .73 and Specificity: .93) and .91 accuracy for sentences reporting loneliness (Sensitivity: .88 and Specificity: .95). Full Python and R code, detailed measure of model performance, and Sensitivity analysis are reported in Supplemental materials ( https://osf.io/2d5jg/ ). Once trained, the NLP models were deployed on the data from the full cohort, effectively processing over 6 million medical records. To focus on the symptomatic interpretation of SI and loneliness, we only considered reports that were made 5 years before or after the initial dementia diagnosis. Our data indicates that most of these reports occur several months before and at the time of diagnosis, and that there was a substantial increase in both types of reports during the first year of the COVID-19 pandemic. Participants The final data used in this study combined patients with measures of cognitive performance, as measured by MoCA scores, clinical diagnosis information, and loneliness or SI reports, consisting of 4817 patients with 9298 observations. The patient flow chart in Figure 1 outlines the procedure used to derive the sample. Our procedure identified 382 patients (851 observations) with loneliness reports and 523 patients (1185 observations) with SI reports in comparison to the 3912 patients without such mentions, which were defined as a control group in our study (see Table 2 for a split between the groups on main variables). As the retrospective cohort data were collated from EHRs, diversity, equality and inclusion could not be directly addressed during data collation. Figure 1. Patient Flow Chart Illustrating Sample Derivation View this table: View inline View popup Table 2: Descriptive statistics for variables of interest (Mean and SD for quantitative variables and Percentage and Number of cases for categorical variables) Statistical analysis The data were analysed using a combination of generalised additive mixed-effect modelling (GAMMs) 41 and linear mixed-effect modelling (LME) 42 . GAMM is a non-parametric data-driven method that estimates a non-linear relationship between predictors of interest and outcome variables. Using this model, we estimated and visualised average changes in cognitive function throughout the disease and tested how trajectories differed between patients who reported loneliness and SI in comparison to the control group of patients without such reports in their EHRs. To investigate the parametric effects of individual predictors on the slope of cognitive decline, we used LME. The primary outcome in the analysis was cognitive function as measured by MoCA and MMSE tests, while we tested the effect of SI and loneliness, and controlled for the effects of Age, Type of Dementia, Sex, Marital Status, Accommodation Status, and whether patients had a diagnosis of Depression in their medical history. In case of missing information for marital and accommodation status, these patients were included in the analysis as Missing categories, while models without these predictors or cases with missing data are reported in the Sensitivity analysis. The random by-patient intercept effects were adjusted in all models, which allowed intercepts to vary for each patient 32 . The adjustment of the random-effect structure allowed us to model repeated measures of cognitive scores and guards against overfitting of the model. Specifically, the random structure in GAMM and LME models enables us to weigh intra-individual changes over time and inter-individual differences, allowing us to estimate cognitive trajectories over time. Patients with multiple observations provide more information about the rate and direction of change, while those with fewer observations provide more information about the variability at the group level. In addition to modelling how SI and loneliness are associated with the cognitive ability of patients throughout their disease, we also explored the rate of cognitive change after patients’ first reports of loneliness and SI. The sensitivity analysis included looking at reports of SI and loneliness in shorter time windows around the diagnosis date, different methods for dealing with missing observations in predictor variables, and MMSE score analysis. We also report analysis focusing on patients with both sets of reports, SI and loneliness, in their EHRs. In comparison to the other three groups, these patients are seen four times more frequently by the health and social services but have fewer measures of cognition, potentially indicating a more complex disease phenotype. Ethical approvals We state that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008. The study was approved by the local UK-CRIS oversight committees and the University of Sheffield Ethics application review board (ID: 045869). Individual patient consent was not required for the use of anonymised data. The R and Python code used to analyse the data and develop NLP models is reported in supporting materials. Results The full cohort consisted of 4817 dementia patients with a mean age of 80.79 years, of whom 57% (n = 2765) were female and 26% (n = 1240) were married. Of these, 8% (n = 382) reported loneliness, 11% (n = 523) reported social isolation, and 81% (n = 3912) had no reports of either. Patients with loneliness were more likely to be women (75%) and widowed (26%), while those with SI were older (mean age 81.71 years). Controls had the highest percentage of married individuals (27%) and fewer cases of depression (4%). Table 2 reports full descriptive statistics for all variables of interest. The effect of loneliness and SI on cognitive trajectories The results show a significant difference in average MoCA scores between patients with loneliness and SI reports in their EHRs and those without such reports. In the case of patients with loneliness reports, MoCA scores (see Figure 2A ) were lower throughout the disease, as illustrated by Figure 2C . When estimated using LME, we see that the scores of patients who reported being lonely are lower by 0.83 MoCA points at the time of dementia diagnosis (see Table 3 ). Figure 2: The effect of loneliness and social isolation on the nonlinear changes in cognition as measured by MoCA. A. Cognitive trajectories of the Control group (full green line) and patients with Loneliness reports (purple dashed line). B. Cognitive trajectories of the Control group (full green line) and patients with social isolation reports (purple dashed line). C. Differences in the cognitive functionality between the two groups (Average difference in MoCA scores between the control group and loneliness group) where disease periods estimated as statistically different are highlighted by the red line. D. Differences in the cognitive functionality between two groups (Average difference in MoCA scores between the control group and social isolation group) where disease periods estimated as statistically different are highlighted by the red line. View this table: View inline View popup Table 3: Linear mixed-effect model coefficients Patients with SI reports have comparable MoCA scores to control patients before being diagnosed. Approximately six months before diagnosis, the slope of MoCA changes for SI patients becomes more severe and they start declining faster than control patients (See Figure 2B and 2D ). This pattern is illustrated in the parametric model where the reference group of patients with no loneliness or SI reports show an average decline of –0.38 MoCA points per year (p < 0.001), whereas the significant interaction between SI reports and slope of diagnosis duration shows SI patients decline faster by a further –0.21 MoCA points per year relative to controls (p = 0.029). The interaction between loneliness reports and diagnosis duration was not significant, suggesting rates of decline were similar to that of patients with no SI or loneliness reports (see Table 3 ). Cognitive change after first report of SI or loneliness The change in MoCA scores after the first SI and loneliness report behaves differently between the two groups of patients. In the case of patients reporting loneliness, the cognitive function continues to decline at the same rate as before the report of loneliness (see Figure 3A ). The cognitive function of patients experiencing SI improves on average after the first mention of SI (See Figure 3B ). Looking at differences in MoCA scores before and after the reports, results show that MoCA scores are on average higher after the report in the SI group (Before = 17.32 versus After = 18.08, t = 2.11, df = 1048.6, p = 0.034), but not in the loneliness group (Before = 17.36 versus After = 17.08, t = –0.67, df = 770.62, p = 0.49). However, when we identify patients with at least one MoCA measure before and after the first report and calculate repeated measure t-tests, we see that both groups of patients decline in their MoCA scores (Loneliness reports: Before = 17.58 versus After = 16.62, t = 2.31, df = 85, p = 0.023 and SI reports: Before = 18.03 versus After = 16.87, df = 115, t = 3.35, p = 0.001). Figure 3: The nonlinear cognitive change, as measured by MoCA, before and after the first report for loneliness and social isolation group. A. Cognitive trajectories of patients with report(s) of loneliness. B. Cognitive trajectories of patients with report(s) of social isolation. Discussion The effect of SI on the risk of dementia development is well established 9 , 15 , but the status of SI and loneliness when presented in presymptomatic and symptomatic stages of the disease and their effect on the progression of the disease is less explored. We examined the effect of reported SI and loneliness on the rate of cognitive change of patients with dementia using EHRs from a UK NHS mental health trust. We showed that patients with evidence of loneliness have worse cognition throughout their disease course. Patients whose EHRs mention SI have comparable cognition, before diagnosis, to patients without mentions of such symptoms. However, six months before being diagnosed, the cognitive ability of socially isolated patients starts declining at a higher rate, resulting in worse cognition at the point of dementia diagnosis and later in the disease course. Mechanisms of SI and loneliness The conceptualisation of the two concepts, SI and loneliness, likely underlies the estimated differences between the three groups of patients in our study 1 . SI represents more of a physical barrier to receiving social support and/or maintaining social connections, while loneliness reflects emotional aspects of this feeling. We found that socially isolated patients have stronger rates of cognitive deterioration several months before a dementia diagnosis while being comparable with controls in the preceding period. In the context of a life-changing and stigmatising diagnosis such as dementia, the lack of social contact prevents patients from receiving needed support, and their cognition declines at higher rates 43 . When reported for the first time, we show that average values of MoCA scores increase for this subgroup of patients. This improvement may reflect positive action taken by healthcare and/or social services in response to the report. In contrast, patients who experience loneliness have worse cognition throughout the disease course. Such results indicate that loneliness might be an intermediary for depressive symptomatology or is caused by common origins 44 – 46 . This interpretation is also supported by the demographic differences split by group, where we see that patients with reports of SI and loneliness also have higher rates of ICD-10 codes for depression. These patients also do not observe any improvement in their cognition after the first report of loneliness, which might be expected given that improvement in feeling of loneliness requires a change in the availability of social networks, psychological well-being, life satisfaction, activities, and other psychiatric symptoms 47 . Strengths and limitations Our study illustrates the potential of medical records from mental health institutions to provide evidence-based results on the effect of symptoms in dementia diseases 9 , 48 . Using large data and advanced statistical modelling, our study shows that SI can be seen as a disease progression factor. Not only is this a novel finding in the domain, contrasting with some previous studies which showed limited effects or their complete absence 20 , 22 , 23 but we show differing effects of SI and loneliness on the cognitive trajectories of patients with dementia. There are several limitations when using large observational datasets 49 . We cannot allocate patients’ membership in the group and can only control for a limited number of factors that could have moderating effects on SI and loneliness, such as depression. Lack of control over the allocation of patients and barriers when accessing healthcare may have led to inadequacies regarding diversity, equality and inclusion, which may reduce the wider generalisability of findings. Similarly, patients with SI/loneliness reports recorded considerably before diagnosis may be a sign of additional health care needs and could additionally limit the generalisability of findings. The correlational nature of the data limits the causal interpretation of our findings, and even though we see improvement in cognition after the first report of SI, we cannot ascertain what change to patients’ social circumstances followed. There are multiple sources of support that patients can receive after SI is identified, such as closer family connections, social services provision of care or change to their living conditions (e.g. admittance to residential care). Our interpretation of SI relies on a subjective perception from the patient, carer or clinician, rather than an objective measure e.g. of frequency of social contact. Automated and trained NLP model architectures are probabilistic, and even though they achieve high levels of accuracy, they introduce an additional layer of noise to the later data analysis 50 . Clinical implications While acknowledging limitations, we show that SI and loneliness could be seen as a disease progression factor for dementia patients, given their effect on cognitive trajectories, even before a formal diagnosis. This means patients experiencing SI or loneliness might benefit from closer monitoring of their cognitive health. While cognitive decline is expected in dementia, the rate of decline can have significant impacts on both patient care and quality of life. By recognising SI and loneliness as potential factors influencing the speed of decline, these findings could have direct implications for clinical practice, such as identifying potential avenues for intervention, informing treatment strategies, and strengthening rationale for social prescribing. We hope that these effects could steer the debate concerning modifiable symptoms, as part of a holistic assessment, that could be used to support the care of patients and outline a research approach that could provide us with more evidence-based studies of modifiable disease progression factors. Data Availability All patient data used in the present study are accessible through the UK-CRIS system. The source data for this work is owned by Oxford Health NHS Foundation Trust using anonymised patient records via CRIS Powered by Akrivia Health. The data cannot be made publicly available but can be accessed with permissions from Oxford Health NHS Foundation Trust for UK NHS staff and UK academics within a secure firewall, in the same manner as the authors. The R and Python code used to analyse the data and develop NLP models is reported in supporting materials. Conflicts of Interest IK is a paid medical advisor to Five Lives, a digital healthcare company developing a platform for addressing preventable risk factors for dementia in ageing adults. All other authors declare no conflicts of interest. Funding This study was supported by CRIS Powered by Akrivia Health, using data, systems and support from the NIHR Oxford Health Biomedical Research Centre (NIHR203316) Research Informatics Team. IK is funded through MRC, NIHR Oxford Health Biomedical Research Facility and an investigator-initiated grant by Novo Nordisk. Consent Statement Informed consent was not required for the study as we utilised anonymised patients’ electronic health records. The data were accessed and analysed in accordance with applicable legal and ethical guidelines to ensure the privacy and confidentiality of all subjects. Contributions JM and NV were granted access to the Electronic Health Records. NV developed Natural Language Processing models, and JM supported their training and evaluation. JM cleaned and processed extracted data, JM and NV jointly analysed data, and All authors worked on the interpretation of the results. IK, RP, and OB supervised the clinical interpretation of the findings, while TS supervised analytical procedures. JM and NV wrote the manuscript, and all authors revised and reviewed the manuscript. NV, as the corresponding author, had final responsibility for the decision to submit the manuscript. Data Sharing The source data for this work is owned by Oxford Health NHS Foundation Trust using anonymised patient records via CRIS Powered by Akrivia Health. The data cannot be made publicly available but can be accessed with permissions from Oxford Health NHS Foundation Trust for UK NHS staff and UK academics within a secure firewall, in the same manner as the authors. Acknowledgements We acknowledge the work and support of the Oxford Research Informatics Team, Tanya Smith (Head of Research Informatics), Adam Pill and Suzanne Fisher (CRIS Academic Support and Information Analyst), Suzanne Fisher (Research Informatics Systems Analysts) and Lulu Kane (Research Informatics Administrator). Matthew Bladen’s contribution to preparing the article for publication is greatly appreciated. The interpretation and conclusions contained in this report are those of the authors alone. The views expressed are those of the authors and not necessarily those of the NHS, the National Institute for Health Research or the Department of Health. Footnotes This version has been updated following reviewer comments. The main updates include: clarification of findings in the abstract to improve interpretation of results; additional full cohort details in the text and Table 2; minor changes to Table 3 to improve interpretability; added additional discussion regarding real-world context and impact of study findings; added additional discussion regarding limitations to using secondary data; updated research in context document to reflect changes. The title has also been amended to better reflect the focus of the study. References 1. ↵ Organization WH . Social isolation and loneliness among older people: advocacy brief . 2021 ; 2. ↵ Evans IE , Martyr A , Collins R , Brayne C , Clare L . Social isolation and cognitive function in later life: a systematic review and meta-analysis . Journal of Alzheimer’s disease . 2019 ; 70 ( s1 ): S119 – S144 . OpenUrl 3. ↵ Smith KJ , Victor C . Typologies of loneliness, living alone and social isolation, and their associations with physical and mental health . Ageing & Society . 2019 ; 39 ( 8 ): 1709 – 1730 . OpenUrl CrossRef 4. Zhou J , Tang R , Wang X , Li X , Heianza Y , Qi L . Improvement of Social Isolation and Loneliness and Excess Mortality Risk in People With Obesity . JAMA Network Open . 2024 ; 7 ( 1 ): e2352824 – e2352824 . OpenUrl 5. Lyu C , Siu K , Xu I , Osman I , Zhong J . Social isolation changes and long-term outcomes among older adults . JAMA Network Open . 2024 ; 7 ( 7 ): e2424519 – e2424519 . OpenUrl PubMed 6. Gerlach LB , Solway ES , Malani PN . Social Isolation and Loneliness in Older Adults . JAMA . 2024 ; 7. ↵ Donovan NJ , Okereke OI , Vannini P , et al. Association of higher cortical amyloid burden with loneliness in cognitively normal older adults . JAMA psychiatry . 2016 ; 73 ( 12 ): 1230 – 1237 . OpenUrl PubMed 8. ↵ Holt-Lunstad J , Smith TB , Baker M , Harris T , Stephenson D . Loneliness and social isolation as risk factors for mortality: a meta-analytic review . Perspectives on psychological science . 2015 ; 10 ( 2 ): 227 – 237 . OpenUrl CrossRef PubMed 9. ↵ Livingston G , Huntley J , Sommerlad A , et al. Dementia prevention, intervention, and care: 2020 report of the Lancet Commission . The lancet . 2020 ; 396 ( 10248 ): 413 – 446 . OpenUrl CrossRef 10. Kelly ME , Duff H , Kelly S , et al. The impact of social activities, social networks, social support and social relationships on the cognitive functioning of healthy older adults: A systematic review . Systematic Reviews , 6 ( 1 ), 1 – 18 . 2017 . OpenUrl PubMed 11. Penninkilampi R , Casey A-N , Singh MF , Brodaty H . The association between social engagement, loneliness, and risk of dementia: a systematic review and meta-analysis . Journal of Alzheimer’s Disease . 2018 ; 66 ( 4 ): 1619 – 1633 . OpenUrl 12. ↵ Sommerlad A , Sabia S , Singh-Manoux A , Lewis G , Livingston G . Association of social contact with dementia and cognition: 28-year follow-up of the Whitehall II cohort study . PLoS medicine . 2019 ; 16 ( 8 ): e1002862 . OpenUrl 13. ↵ Lara E , Martín-María N , De la Torre-Luque A , et al. Does loneliness contribute to mild cognitive impairment and dementia? A systematic review and meta-analysis of longitudinal studies . Ageing research reviews . 2019 ; 52 : 7 – 16 . OpenUrl CrossRef PubMed 14. ↵ Victor CR . Is loneliness a cause or consequence of dementia? A public health analysis of the literature . Frontiers in psychology . 2021 ; 11 : 612771 . OpenUrl PubMed 15. ↵ Perry BL , Roth AR , Peng S , Risacher SL , Saykin AJ , Apostolova LG . Social networks and cognitive reserve: Network structure moderates the association between amygdalar volume and cognitive outcomes . The Journals of Gerontology: Series B . 2022 ; 77 ( 8 ): 1490 – 1500 . OpenUrl PubMed 16. ↵ Bennett DA , Schneider JA , Tang Y , Arnold SE , Wilson RS . The effect of social networks on the relation between Alzheimer’s disease pathology and level of cognitive function in old people: a longitudinal cohort study . The Lancet Neurology . 2006 ; 5 ( 5 ): 406 – 412 . OpenUrl PubMed 17. ↵ Mathys H , Peng Z , Boix CA , et al. Single-cell atlas reveals correlates of high cognitive function, dementia, and resilience to Alzheimer’s disease pathology . Cell . 2023 ; 186 ( 20 ): 4365 – 4385 . e27. OpenUrl CrossRef PubMed 18. ↵ Stern Y , Albert M , Barnes CA , Cabeza R , Pascual-Leone A , Rapp PR . A framework for concepts of reserve and resilience in aging . Neurobiology of aging . 2023 ; 124 : 100 – 103 . OpenUrl CrossRef PubMed 19. ↵ Holwerda TJ , Deeg DJ , Beekman AT , et al. Feelings of loneliness, but not social isolation, predict dementia onset: results from the Amsterdam Study of the Elderly (AMSTEL). Journal of Neurology , Neurosurgery & Psychiatry . 2014 ; 85 ( 2 ): 135 – 142 . OpenUrl 20. ↵ Zhang Y , Tatewaki Y , Liu Y , et al. Perceived social isolation is correlated with brain structure and cognitive trajectory in Alzheimer’s disease . GeroScience . 2022 ; 44 ( 3 ): 1563 - 1574 . OpenUrl PubMed 21. ↵ Freak-Poli R , Wagemaker N , Wang R , et al. Loneliness, not social support, is associated with cognitive decline and dementia across two longitudinal population-based cohorts . Journal of Alzheimer’s disease . 2022 ; 85 ( 1 ): 295 – 308 . OpenUrl 22. ↵ Sun W , Matsuoka T , Oba H , Narumoto J . Importance of loneliness in behavioral and psychological symptoms of dementia . International Journal of Geriatric Psychiatry . 2021 ; 36 ( 4 ): 540 – 546 . OpenUrl PubMed 23. ↵ Victor CR , Rippon I , Nelis SM , et al. Prevalence and determinants of loneliness in people living with dementia: Findings from the IDEAL programme . International Journal of Geriatric Psychiatry . 2020 ; 35 ( 8 ): 851 – 858 . OpenUrl CrossRef PubMed 24. ↵ Joyce J , Ryan J , Owen A , et al. Social isolation, social support, and loneliness and their relationship with cognitive health and dementia . International Journal of Geriatric Psychiatry . 2022 ; 37 (1) 25. ↵ Boss L , Kang D-H , Branson S . Loneliness and cognitive function in the older adult: a systematic review . International psychogeriatrics . 2015 ; 27 ( 4 ): 541 – 553 . OpenUrl CrossRef PubMed 26. ↵ Janssen O , Vos SJ , García-Negredo G , et al. Real-world evidence in Alzheimer’s disease: the ROADMAP Data Cube . Alzheimer’s & Dementia . 2020 ; 16 ( 3 ): 461 – 471 . OpenUrl 27. ↵ Nasreddine ZS , Phillips NA , Bédirian V , et al. The Montreal Cognitive Assessment, MoCA: a brief screening tool for mild cognitive impairment . Journal of the American Geriatrics Society . 2005 ; 53 ( 4 ): 695 – 699 . OpenUrl CrossRef PubMed Web of Science 28. ↵ Folstein MF , Folstein SE , McHugh PR . “Mini-mental state”: a practical method for grading the cognitive state of patients for the clinician . Journal of psychiatric research . 1975 ; 12 ( 3 ): 189 – 198 . OpenUrl CrossRef PubMed Web of Science 29. ↵ Pinto TC , Machado L , Bulgacov TM , et al. Is the Montreal Cognitive Assessment (MoCA) screening superior to the Mini-Mental State Examination (MMSE) in the detection of mild cognitive impairment (MCI) and Alzheimer’s Disease (AD) in the elderly? International psychogeriatrics . 2019 ; 31 ( 4 ): 491 – 504 . OpenUrl CrossRef PubMed 30. ↵ Dementia: assessment, management and support for people living with dementia and their carers . National Institute for Health and Care Excellence ( 2021 ). 31. ↵ Stojanovic M , Mikula C , John S , Kiselica A . Clinical importance in Alzheimer’s disease: effects of anchor agreement and disease severity . Aging Clinical and Experimental Research . 2024 ; 36 ( 1 ): 5 . OpenUrl PubMed 32. ↵ Vaci N , Koychev I , Kim C-H , et al. Real-world effectiveness, its predictors and onset of action of cholinesterase inhibitors and memantine in dementia: retrospective health record study . The British Journal of Psychiatry . 2021 ; 218 ( 5 ): 261 – 267 . OpenUrl PubMed 33. ↵ Kormilitzin A , Vaci N , Liu Q , Nevado-Holgado A . Med7: A transferable clinical natural language processing model for electronic health records . Artificial Intelligence in Medicine . 2021 ; 118 : 102086 . OpenUrl CrossRef PubMed 34. Senior M , Burghart M , Yu R , et al. Identifying Predictors of Suicide in Severe Mental Illness: A Feasibility Study of a Clinical Prediction Rule (Ox ford M ental I llness and S uicide Tool or OxMIS) . Frontiers in psychiatry . 2020 ; 11 : 268 . OpenUrl PubMed 35. Goodday SM , Kormilitzin A , Vaci N , et al. Maximizing the use of social and behavioural information from secondary care mental health electronic health records . Journal of biomedical informatics . 2020 ; 107 : 103429 . OpenUrl PubMed 36. Li Z , Kormilitzin A , Fernandes M , et al. Validation of UK Biobank data for mental health outcomes: A pilot study using secondary care electronic health records . International Journal of Medical Informatics . 2022 ; 160 : 104704 . OpenUrl PubMed 37. ↵ Vaci N , Liu Q , Kormilitzin A , et al. Natural language processing for structuring clinical text data on depression using UK-CRIS . BMJ Ment Health . 2020 ; 23 ( 1 ): 21 – 26 . OpenUrl Abstract / FREE Full Text 38. ↵ Tunstall L , Reimers N , Jo UES , et al. Efficient few-shot learning without prompts . arXiv preprint arXiv:220911055 . 2022 ; 39. ↵ De Jong-Gierveld J , van Tilburg TG . A 6-item scale for overall, emotional, and social loneliness: Confirmatory tests on survey data . Research on aging . 2006 ; 28 ( 5 ): 582 – 598 . OpenUrl CrossRef Web of Science 40. ↵ Russell DW . UCLA Loneliness Scale (Version 3): Reliability, validity, and factor structure . Journal of personality assessment . 1996 ; 66 ( 1 ): 20 – 40 . OpenUrl CrossRef PubMed Web of Science 41. ↵ van Rij J , Vaci N , Wurm LH , Feldman LB . Alternative quantitative methods in psycholinguistics: Implications for theory and design . Word Knowl Word Usage . 2020 ; 83 42. ↵ Pinheiro JC , Bates DM . Linear mixed-effects models: basic concepts and examples . Mixed-effects models in S and S-Plus . 2000 : 3 – 56 . 43. ↵ Murata C , Saito T , Saito M , Kondo K . The association between social support and incident dementia: a 10-year follow-up study in Japan . International journal of environmental research and public health . 2019 ; 16 ( 2 ): 239 . OpenUrl 44. ↵ Erzen E , Çikrikci Ö . The effect of loneliness on depression: A meta-analysis . International Journal of Social Psychiatry . 2018 ; 64 ( 5 ): 427 – 435 . OpenUrl CrossRef PubMed 45. Weeks DG , Michela JL , Peplau LA , Bragg ME . Relation between loneliness and depression: a structural equation analysis . Journal of personality and social psychology . 1980 ; 39 ( 6 ): 1238 . OpenUrl CrossRef PubMed Web of Science 46. ↵ Shen C , Rolls ET , Cheng W , et al. Associations of social isolation and loneliness with later dementia . Neurology . 2022 ; 99 ( 2 ): e164 – e175 . OpenUrl PubMed 47. ↵ Åkerlind I , Hörnquist JO . Stability and change in feelings of loneliness: A two-year prospective longitudinal study of advanced alcohol abusers . Scandinavian Journal of Psychology . 1989 /06/01; 30 ( 2 ) doi: 10.1111/j.1467-9450.1989.tb01073.x OpenUrl CrossRef 48. ↵ Oliver D , Arribas M , Perry BI , et al. Using Electronic Health Records To Facilitate Precision Psychiatry . Biological psychiatry . 2024 ; 49. ↵ Meystre SM , Savova GK , Kipper-Schuler KC , Hurdle JF . Extracting information from textual documents in the electronic health record: a review of recent research . Yearb Med Inform . 2008 : 128 – 44 . 50. ↵ Perera S , Sheth A , Thirunarayan K , Nair S , Shah N . Challenges in understanding clinical notes: why NLP engines fall short and where background knowledge can help . presented at: Proceedings of the 2013 international workshop on Data management & analytics for healthcare ; 2013 ; San Francisco, California, USA. View the discussion thread. Back to top Previous Next Posted May 19, 2025. Download PDF Data/Code Email Thank you for your interest in spreading the word about medRxiv. NOTE: Your email address is requested solely to identify you as the sender of this article. Your Email * Your Name * Send To * Enter multiple addresses on separate lines or separate them with commas. You are going to email the following Loneliness, Social Isolation, and Effects on Cognitive Decline in Patients with Dementia: A Retrospective Cohort Study Using Natural Language Processing Message Subject (Your Name) has forwarded a page to you from medRxiv Message Body (Your Name) thought you would like to see this page from the medRxiv website. Your Personal Message CAPTCHA This question is for testing whether or not you are a human visitor and to prevent automated spam submissions. Share Loneliness, Social Isolation, and Effects on Cognitive Decline in Patients with Dementia: A Retrospective Cohort Study Using Natural Language Processing James A. C. Myers , Tom Stafford , Ivan Koychev , Robert Perneczky , Oliver Bandmann , Nemanja Vaci medRxiv 2025.02.10.25321986; doi: https://doi.org/10.1101/2025.02.10.25321986 Share This Article: Copy Citation Tools Loneliness, Social Isolation, and Effects on Cognitive Decline in Patients with Dementia: A Retrospective Cohort Study Using Natural Language Processing James A. C. Myers , Tom Stafford , Ivan Koychev , Robert Perneczky , Oliver Bandmann , Nemanja Vaci medRxiv 2025.02.10.25321986; doi: https://doi.org/10.1101/2025.02.10.25321986 Citation Manager Formats BibTeX Bookends EasyBib EndNote (tagged) EndNote 8 (xml) Medlars Mendeley Papers RefWorks Tagged Ref Manager RIS Zotero Tweet Widget Facebook Like Google Plus One Subject Area Psychiatry and Clinical Psychology Subject Areas All Articles Addiction Medicine (570) Allergy and Immunology (864) Anesthesia (302) Cardiovascular Medicine (4445) Dentistry and Oral Medicine (444) Dermatology (383) Emergency Medicine (609) Endocrinology (including Diabetes Mellitus and Metabolic Disease) (1515) Epidemiology (15234) Forensic Medicine (30) Gastroenterology (1127) Genetic and Genomic Medicine (6610) Geriatric Medicine (669) Health Economics (1000) Health Informatics (4548) Health Policy (1370) Health Systems and Quality Improvement (1613) Hematology (543) HIV/AIDS (1266) Infectious Diseases (except HIV/AIDS) (15925) Intensive Care and Critical Care Medicine (1104) Medical Education (623) Medical Ethics (147) Nephrology (668) Neurology (6612) Nursing (346) Nutrition (999) Obstetrics and Gynecology (1147) Occupational and Environmental Health (957) Oncology (3340) Ophthalmology (975) Orthopedics (369) Otolaryngology (420) Pain Medicine (436) Palliative Medicine (130) Pathology (665) Pediatrics (1694) Pharmacology and Therapeutics (693) Primary Care Research (714) Psychiatry and Clinical Psychology (5458) Public and Global Health (9243) Radiology and Imaging (2204) Rehabilitation Medicine and Physical Therapy (1370) Respiratory Medicine (1197) Rheumatology (596) Sexual and Reproductive Health (715) Sports Medicine (530) Surgery (713) Toxicology (99) Transplantation (289) Urology (265) (function(){function c(){var b=a.contentDocument||a.contentWindow.document;if(b){var d=b.createElement('script');d.innerHTML="window.__CF$cv$params={r:'a02175bfdbb23fe2',t:'MTc3OTg0NzU4Mg=='};var a=document.createElement('script');a.src='/cdn-cgi/challenge-platform/scripts/jsd/main.js';document.getElementsByTagName('head')[0].appendChild(a);";b.getElementsByTagName('head')[0].appendChild(d)}}if(document.body){var a=document.createElement('iframe');a.height=1;a.width=1;a.style.position='absolute';a.style.top=0;a.style.left=0;a.style.border='none';a.style.visibility='hidden';document.body.appendChild(a);if('loading'!==document.readyState)c();else if(window.addEventListener)document.addEventListener('DOMContentLoaded',c);else{var e=document.onreadystatechange||function(){};document.onreadystatechange=function(b){e(b);'loading'!==document.readyState&&(document.onreadystatechange=e,c())}}}})();