Automated detection of patients with dementia whose symptoms have been identified in primary care but have no formal diagnosis: a retrospective case-control study using electronic primary care records.

OBJECTIVES: UK statistics suggest only two-thirds of patients with dementia get a diagnosis recorded in primary care. General practitioners (GPs) report barriers to formally diagnosing dementia, so some patients may be known by GPs to have dementia but may be missing a diagnosis in their patient record. We aimed to produce a method to identify these 'known but unlabelled' patients with dementia using data from primary care patient records. DESIGN: Retrospective case-control study using routinely collected primary care patient records from Clinical Practice Research Datalink. SETTING: UK general practice. PARTICIPANTS: English patients aged >65 years, with a coded diagnosis of dementia recorded in 2000-2012 (cases), matched 1:1 with patients with no diagnosis code for dementia (controls). INTERVENTIONS: Eight coded and nine keyword concepts indicating symptoms, screening tests, referrals and care for dementia recorded in the 5 years before diagnosis. We trialled machine learning classifiers to discriminate between cases and controls (logistic regression, naïve Bayes, random forest). PRIMARY AND SECONDARY OUTCOMES: The outcome variable was dementia diagnosis code; the accuracy of classifiers was assessed using area under the receiver operating characteristic curve (AUC); the order of features contributing to discrimination was examined. RESULTS: 93 426 patients were included; the median age was 83 years (64.8% women). Three classifiers achieved high discrimination and performed very similarly. AUCs were 0.87-0.90 with coded variables, rising to 0.90-0.94 with keywords added. Feature prioritisation was different for each classifier; commonly prioritised features were Alzheimer's prescription, dementia annual review, memory loss and dementia keywords. CONCLUSIONS: It is possible to detect patients with dementia who are known to GPs but unlabelled with a diagnostic code, with a high degree of accuracy in electronic primary care record data. Using keywords from clinic notes and letters improves accuracy compared with coded data alone. This approach could improve identification of dementia cases for record-keeping, service planning and delivery of good quality care.

Could dementia be detected from UK primary care patients' records by simple automated methods earlier than by the treating physician? A retrospective case-control study.

Background: Timely diagnosis of dementia is a policy priority in the United Kingdom (UK). Primary care physicians receive incentives to diagnose dementia; however, 33% of patients are still not receiving a diagnosis. We explored automating early detection of dementia using data from patients' electronic health records (EHRs). We investigated: a) how early a machine-learning model could accurately identify dementia before the physician; b) if models could be tuned for dementia subtype; and c) what the best clinical features were for achieving detection. Methods: Using EHRs from Clinical Practice Research Datalink in a case-control design, we selected patients aged >65y with a diagnosis of dementia recorded 2000-2012 (cases) and matched them 1:1 to controls; we also identified subsets of Alzheimer's and vascular dementia patients. Using 77 coded concepts recorded in the 5 years before diagnosis, we trained random forest classifiers, and evaluated models using Area Under the Receiver Operating Characteristic Curve (AUC). We examined models by year prior to diagnosis, subtype, and the most important features contributing to classification. Results: 95,202 patients (median age 83y; 64.8% female) were included (50% dementia cases). Classification of dementia cases and controls was poor 2-5 years prior to physician-recorded diagnosis (AUC range 0.55-0.65) but good in the year before (AUC: 0.84). Features indicating increasing cognitive and physical frailty dominated models 2-5 years before diagnosis; in the final year, initiation of the dementia diagnostic pathway (symptoms, screening and referral) explained the sudden increase in accuracy. No substantial differences were seen between all-cause dementia and subtypes. Conclusions: Automated detection of dementia earlier than the treating physician may be problematic, if using only primary care data. Future work should investigate more complex modelling, benefits of linking multiple sources of healthcare data and monitoring devices, or contextualising the algorithm to those cases that the GP would need to investigate.

Can the Use of Bayesian Analysis Methods Correct for Incompleteness in Electronic Health Records Diagnosis Data? Development of a Novel Method Using Simulated and Real-Life Clinical Data.

Background: Patient health information is collected routinely in electronic health records (EHRs) and used for research purposes, however, many health conditions are known to be under-diagnosed or under-recorded in EHRs. In research, missing diagnoses result in under-ascertainment of true cases, which attenuates estimated associations between variables and results in a bias toward the null. Bayesian approaches allow the specification of prior information to the model, such as the likely rates of missingness in the data. This paper describes a Bayesian analysis approach which aimed to reduce attenuation of associations in EHR studies focussed on conditions characterized by under-diagnosis. Methods: Study 1: We created synthetic data, produced to mimic structured EHR data where diagnoses were under-recorded. We fitted logistic regression (LR) models with and without Bayesian priors representing rates of misclassification in the data. We examined the LR parameters estimated by models with and without priors. Study 2: We used EHR data from UK primary care in a case-control design with dementia as the outcome. We fitted LR models examining risk factors for dementia, with and without generic prior information on misclassification rates. We examined LR parameters estimated by models with and without the priors, and estimated classification accuracy using Area Under the Receiver Operating Characteristic. Results: Study 1: In synthetic data, estimates of LR parameters were much closer to the true parameter values when Bayesian priors were added to the model; with no priors, parameters were substantially attenuated by under-diagnosis. Study 2: The Bayesian approach ran well on real life clinic data from UK primary care, with the addition of prior information increasing LR parameter values in all cases. In multivariate regression models, Bayesian methods showed no improvement in classification accuracy over traditional LR. Conclusions: The Bayesian approach showed promise but had implementation challenges in real clinical data: prior information on rates of misclassification was difficult to find. Our simple model made a number of assumptions, such as diagnoses being missing at random. Further development is needed to integrate the method into studies using real-life EHR data. Our findings nevertheless highlight the importance of developing methods to address missing diagnoses in EHR data.

Identifying undetected dementia in UK primary care patients: a retrospective case-control study comparing machine-learning and standard epidemiological approaches.

BACKGROUND: Identifying dementia early in time, using real world data, is a public health challenge. As only two-thirds of people with dementia now ultimately receive a formal diagnosis in United Kingdom health systems and many receive it late in the disease process, there is ample room for improvement. The policy of the UK government and National Health Service (NHS) is to increase rates of timely dementia diagnosis. We used data from general practice (GP) patient records to create a machine-learning model to identify patients who have or who are developing dementia, but are currently undetected as having the condition by the GP. METHODS: We used electronic patient records from Clinical Practice Research Datalink (CPRD). Using a case-control design, we selected patients aged >65y with a diagnosis of dementia (cases) and matched them 1:1 by sex and age to patients with no evidence of dementia (controls). We developed a list of 70 clinical entities related to the onset of dementia and recorded in the 5 years before diagnosis. After creating binary features, we trialled machine learning classifiers to discriminate between cases and controls (logistic regression, naïve Bayes, support vector machines, random forest and neural networks). We examined the most important features contributing to discrimination. RESULTS: The final analysis included data on 93,120 patients, with a median age of 82.6 years; 64.8% were female. The naïve Bayes model performed least well. The logistic regression, support vector machine, neural network and random forest performed very similarly with an AUROC of 0.74. The top features retained in the logistic regression model were disorientation and wandering, behaviour change, schizophrenia, self-neglect, and difficulty managing. CONCLUSIONS: Our model could aid GPs or health service planners with the early detection of dementia. Future work could improve the model by exploring the longitudinal nature of patient data and modelling decline in function over time.

Creating longitudinal datasets and cleaning existing data identifiers in a cystic fibrosis registry using a novel Bayesian probabilistic approach from astronomy.

Patient registry data are commonly collected as annual snapshots that need to be amalgamated to understand the longitudinal progress of each patient. However, patient identifiers can either change or may not be available for legal reasons when longitudinal data are collated from patients living in different countries. Here, we apply astronomical statistical matching techniques to link individual patient records that can be used where identifiers are absent or to validate uncertain identifiers. We adopt a Bayesian model framework used for probabilistically linking records in astronomy. We adapt this and validate it across blinded, annually collected data. This is a high-quality (Danish) sub-set of data held in the European Cystic Fibrosis Society Patient Registry (ECFSPR). Our initial experiments achieved a precision of 0.990 at a recall value of 0.987. However, detailed investigation of the discrepancies uncovered typing errors in 27 of the identifiers in the original Danish sub-set. After fixing these errors to create a new gold standard our algorithm correctly linked individual records across years achieving a precision of 0.997 at a recall value of 0.987 without recourse to identifiers. Our Bayesian framework provides the probability of whether a pair of records belong to the same patient. Unlike other record linkage approaches, our algorithm can also use physical models, such as body mass index curves, as prior information for record linkage. We have shown our framework can create longitudinal samples where none existed and validate pre-existing patient identifiers. We have demonstrated that in this specific case this automated approach is better than the existing identifiers.

Predicting dementia from primary care records: A systematic review and meta-analysis.

INTRODUCTION: Possible dementia is usually identified in primary care by general practitioners (GPs) who refer to specialists for diagnosis. Only two-thirds of dementia cases are currently recorded in primary care, so increasing the proportion of cases diagnosed is a strategic priority for the UK and internationally. Variables in the primary care record may indicate risk of developing dementia, and could be combined in a predictive model to help find patients who are missing a diagnosis. We conducted a meta-analysis to identify clinical entities with potential for use in such a predictive model for dementia in primary care. METHODS AND FINDINGS: We conducted a systematic search in PubMed, Web of Science and primary care database bibliographies. We included cohort or case-control studies which used routinely collected primary care data, to measure the association between any clinical entity and dementia. Meta-analyses were performed to pool odds ratios. A sensitivity analysis assessed the impact of non-independence of cases between studies. From a sift of 3836 papers, 20 studies, all European, were eligible for inclusion, comprising >1 million patients. 75 clinical entities were assessed as risk factors for all cause dementia, Alzheimer's (AD) and Vascular dementia (VaD). Data included were unexpectedly heterogeneous, and assumptions were made about definitions of clinical entities and timing as these were not all well described. Meta-analysis showed that neuropsychiatric symptoms including depression, anxiety, and seizures, cognitive symptoms, and history of stroke, were positively associated with dementia. Cardiovascular risk factors such as hypertension, heart disease, dyslipidaemia and diabetes were positively associated with VaD and negatively with AD. Sensitivity analyses showed similar results. CONCLUSIONS: These findings are of potential value in guiding feature selection for a risk prediction tool for dementia in primary care. Limitations include findings being UK-focussed. Further predictive entities ascertainable from primary care data, such as changes in consulting patterns, were absent from the literature and should also be explored in future studies.

Gaussian process classification of Alzheimer's disease and mild cognitive impairment from resting-state fMRI.

Multivariate pattern analysis and statistical machine learning techniques are attracting increasing interest from the neuroimaging community. Researchers and clinicians are also increasingly interested in the study of functional-connectivity patterns of brains at rest and how these relations might change in conditions like Alzheimer's disease or clinical depression. In this study we investigate the efficacy of a specific multivariate statistical machine learning technique to perform patient stratification from functional-connectivity patterns of brains at rest. Whilst the majority of previous approaches to this problem have employed support vector machines (SVMs) we investigate the performance of Bayesian Gaussian process logistic regression (GP-LR) models with linear and non-linear covariance functions. GP-LR models can be interpreted as a Bayesian probabilistic analogue to kernel SVM classifiers. However, GP-LR methods confer a number of benefits over kernel SVMs. Whilst SVMs only return a binary class label prediction, GP-LR, being a probabilistic model, provides a principled estimate of the probability of class membership. Class probability estimates are a measure of the confidence the model has in its predictions, such a confidence score may be extremely useful in the clinical setting. Additionally, if miss-classification costs are not symmetric, thresholds can be set to achieve either strong specificity or sensitivity scores. Since GP-LR models are Bayesian, computationally expensive cross-validation hyper-parameter grid-search methods can be avoided. We apply these methods to a sample of 77 subjects; 27 with a diagnosis of probable AD, 50 with a diagnosis of a-MCI and a control sample of 39. All subjects underwent a MRI examination at 3T to obtain a 7minute and 20second resting state scan. Our results support the hypothesis that GP-LR models can be effective at performing patient stratification: the implemented model achieves 75% accuracy disambiguating healthy subjects from subjects with amnesic mild cognitive impairment and 97% accuracy disambiguating amnesic mild cognitive impairment subjects from those with Alzheimer's disease, accuracies are estimated using a held-out test set. Both results are significant at the 1% level.

Engineering nanoscale order into a designed protein fiber.

We have established a designed system comprising two peptides that coassemble to form long, thickened protein fibers in water. This system can be rationally engineered to alter fiber assembly, stability, and morphology. Here, we show that rational mutations to our original peptide designs lead to structures with a remarkable level of order on the nanoscale that mimics certain natural fibrous assemblies. In the engineered system, the peptides assemble into two-stranded alpha-helical coiled-coil rods, which pack in axial register in a 3D hexagonal lattice of size 1.824 nm, and with a periodicity of 4.2 nm along the fiber axis. This model is supported by both electron microscopy and x-ray diffraction. Specifically, the fibers display surface striations separated by nanoscale distances that precisely match the 4.2-nm length expected for peptides configured as alpha-helices as designed. These patterns extend unbroken across the widths (>/=50 nm) and lengths (>10 microm) of the fibers. Furthermore, the spacing of the striations can be altered predictably by changing the length of the peptides. These features reflect a high level of internal order within the fibers introduced by the peptide-design process. To our knowledge, this exceptional order, and its persistence along and across the fibers, is unique in a biomimetic system. This work represents a step toward rational bottom-up assembly of nanostructured fibrous biomaterials for potential applications in synthetic biology and nanobiotechnology.