Estimating medication adherence from Electronic Health Records: comparing methods for mining and processing asthma treatment prescriptions.

BACKGROUND: Medication adherence is usually defined as the extent of the agreement between the medication regimen agreed to by patients with their healthcare provider and the real-world implementation. Proactive identification of those with poor adherence may be useful to identify those with poor disease control and offers the opportunity for ameliorative action. Adherence can be estimated from Electronic Health Records (EHRs) by comparing medication dispensing records to the prescribed regimen. Several methods have been developed in the literature to infer adherence from EHRs, however there is no clear consensus on what should be considered the gold standard in each use case. Our objectives were to critically evaluate different measures of medication adherence in a large longitudinal Scottish EHR dataset. We used asthma, a chronic condition with high prevalence and high rates of non-adherence, as a case study. METHODS: Over 1.6 million asthma controllers were prescribed for our cohort of 91,334 individuals, between January 2009 and March 2017. Eight adherence measures were calculated, and different approaches to estimating the amount of medication supply available at any time were compared. RESULTS: Estimates from different measures of adherence varied substantially. Three of the main drivers of the differences between adherence measures were the expected duration (if taken as in accordance with the dose directions), whether there was overlapping supply between prescriptions, and whether treatment had been discontinued. However, there are also wider, study-related, factors which are crucial to consider when comparing the adherence measures. CONCLUSIONS: We evaluated the limitations of various medication adherence measures, and highlight key considerations about the underlying data, condition, and population to guide researchers choose appropriate adherence measures. This guidance will enable researchers to make more informed decisions about the methodology they employ, ensuring that adherence is captured in the most meaningful way for their particular application needs.

New Insights into Stroke from Continuous Passively Collected Temperature and Sleep Data Using Wrist-Worn Wearables.

Actigraphy may provide new insights into clinical outcomes and symptom management of patients through passive, continuous data collection. We used the GENEActiv smartwatch to passively collect actigraphy, wrist temperature, and ambient light data from 27 participants after stroke or probable brain transient ischemic attack (TIA) over 42 periods of device wear. We computed 323 features using established algorithms and proposed 25 novel features to characterize sleep and temperature. We investigated statistical associations between the extracted features and clinical outcomes evaluated using clinically validated questionnaires to gain insight into post-stroke recovery. We subsequently fitted logistic regression models to replicate clinical diagnosis (stroke or TIA) and disability due to stroke. The model generalization performance was assessed using a leave-one-subject-out cross validation method with the selected feature subsets, reporting the area under the curve (AUC). We found that several novel features were strongly correlated (|r|>0.3) with stroke symptoms and mental health measures. Using selected novel features, we obtained an AUC of 0.766 to estimate diagnosis and an AUC of 0.749 to estimate whether disability due to stroke was present. Collectively, these findings suggest that features extracted from the temperature smartwatch sensor may reveal additional clinically useful information over and above existing actigraphy-based features.

Language function following preterm birth: prediction using machine learning.

BACKGROUND: Preterm birth can lead to impaired language development. This study aimed to predict language outcomes at 2 years corrected gestational age (CGA) for children born preterm. METHODS: We analysed data from 89 preterm neonates (median GA 29 weeks) who underwent diffusion MRI (dMRI) at term-equivalent age and language assessment at 2 years CGA using the Bayley-III. Feature selection and a random forests classifier were used to differentiate typical versus delayed (Bayley-III language composite score <85) language development. RESULTS: The model achieved balanced accuracy: 91%, sensitivity: 86%, and specificity: 96%. The probability of language delay at 2 years CGA is increased with: increasing values of peak width of skeletonized fractional anisotropy (PSFA), radial diffusivity (PSRD), and axial diffusivity (PSAD) derived from dMRI; among twins; and after an incomplete course of, or no exposure to, antenatal corticosteroids. Female sex and breastfeeding during the neonatal period reduced the risk of language delay. CONCLUSIONS: The combination of perinatal clinical information and MRI features leads to accurate prediction of preterm infants who are likely to develop language deficits in early childhood. This model could potentially enable stratification of preterm children at risk of language dysfunction who may benefit from targeted early interventions. IMPACT: A combination of clinical perinatal factors and neonatal DTI measures of white matter microstructure leads to accurate prediction of language outcome at 2 years corrected gestational age following preterm birth. A model that comprises clinical and MRI features that has potential to be scalable across centres. It offers a basis for enhancing the power and generalizability of diagnostic and prognostic studies of neurodevelopmental disorders associated with language impairment. Early identification of infants who are at risk of language delay, facilitating targeted early interventions and support services, which could improve the quality of life for children born preterm.

Derivation of asthma severity from electronic prescription records using British thoracic society treatment steps.

BACKGROUND: Asthma severity is typically assessed through a retrospective assessment of the treatment required to control symptoms and to prevent exacerbations. The joint British Thoracic Society and Scottish Intercollegiate Guidelines Network (BTS/SIGN) guidelines encourage a stepwise approach to pharmacotherapy, and as such, current treatment step can be considered as a severity categorisation proxy. Briefly, the steps for adults can be summarised as: no controller therapy (Step 0), low-strength Inhaled Corticosteroids (ICS; Step 1), ICS plus Long-Acting Beta-2 Agonist (LABA; Step 2), medium-dose ICS + LABA (Step 3), and finally either an increase in strength or additional therapies (Step 4). This study aimed to investigate how BTS/SIGN Steps can be estimated from across a large cohort using electronic prescription records, and to describe the incidence of each BTS/SIGN Step in a general population. METHODS: There were 41,433,707 prescriptions, for 671,304 individuals, in the Asthma Learning Health System Scottish cohort, between 1/2009 and 3/2017. Days on which an individual had a prescription for at least one asthma controller (preventer) medication were labelled prescription events. A rule-based algorithm was developed for extracting the strength and volume of medication instructed to be taken daily from free-text data fields. Asthma treatment regimens were categorised by the combination of medications prescribed in the 120 days preceding any prescription event and categorised into BTS/SIGN treatment steps. RESULTS: Almost 4.5 million ALHS prescriptions were for asthma controllers. 26% of prescription events had no inhaled corticosteroid prescriptions in the preceding 120 days (Step 0), 16% were assigned to BTS/SIGN Step 1, 7% to Step 2, 21% to Step 3, and 30% to Step 4. The median days spent on a treatment step before a step-down in treatment was 297 days, whereas a step-up only took a median of 134 days. CONCLUSION: We developed a reproducible methodology enabling researchers to estimate BTS/SIGN asthma treatment steps in population health studies, providing valuable insights into population and patient-specific trajectories, towards improving the management of asthma.

Investigating Wrist-Based Acceleration Summary Measures across Different Sample Rates towards 24-Hour Physical Activity and Sleep Profile Assessment.

Wrist-worn wearable sensors have attracted considerable research interest because of their potential in providing continuous, longitudinal, non-invasive measurements, leading to insights into Physical Activity (PA), sleep, and circadian variability. Three key practical considerations for research-grade wearables are as follows: (a) choosing an appropriate sample rate, (b) summarizing raw three-dimensional accelerometry data for further processing (accelerometry summary measures), and (c) accurately estimating PA levels and sleep towards understanding participants' 24-hour profiles. We used the CAPTURE-24 dataset, where 148 participants concurrently wore a wrist-worn three-dimensional accelerometer and a wearable camera over approximately 24 h to obtain minute-by-minute labels: sleep; and sedentary light, moderate, and vigorous PA. We propose a new acceleration summary measure, the Rate of Change Acceleration Movement (ROCAM), and compare its performance against three established approaches summarizing three-dimensional acceleration data towards replicating the minute-by-minute labels. Moreover, we compare findings where the acceleration data was sampled at 10, 25, 50, and 100 Hz. We demonstrate the competitive advantage of ROCAM towards estimating the five labels (80.2% accuracy) and building 24-hour profiles where the sample rate of 10 Hz is fully sufficient. Collectively, these findings provide insights facilitating the deployment of large-scale longitudinal actigraphy data processing towards 24-hour PA and sleep-profile assessment.

Sex Differences in Cardiac Troponin I and T and the Prediction of Cardiovascular Events in the General Population.

BACKGROUND: Cardiac troponin concentrations differ in women and men, but how this influences risk prediction and whether a sex-specific approach is required is unclear. We evaluated whether sex influences the predictive ability of cardiac troponin I and T for cardiovascular events in the general population. METHODS: High-sensitivity cardiac troponin (hs-cTn) I and T were measured in the Generation Scotland Scottish Family Health Study of randomly selected volunteers drawn from the general population between 2006 and 2011. Cox-regression models evaluated associations between hs-cTnI and hs-cTnT and the primary outcome of cardiovascular death, myocardial infarction, or stroke. RESULTS: In 19 501 (58% women, mean age 47 years) participants, the primary outcome occurred in 2.7% (306/11 375) of women and 5.1% (411/8126) of men during the median follow-up period of 7.9 (IQR, 7.1-9.2) years. Cardiac troponin I and T concentrations were lower in women than men (P < 0.001 for both), and both were more strongly associated with cardiovascular events in women than men. For example, at a hs-cTnI concentration of 10 ng/L, the hazard ratio relative to the limit of blank was 9.7 (95% CI 7.6-12.4) and 5.6 (95% CI 4.7-6.6) for women and men, respectively. The hazard ratio for hs-cTnT at a concentration of 10 ng/L relative to the limit of blank was 3.7 (95% CI 3.1-4.3) and 2.2 (95% CI 2.0-2.5) for women and men, respectively. CONCLUSIONS: Cardiac troponin concentrations differ in women and men and are stronger predictors of cardiovascular events in women. Sex-specific approaches are required to provide equivalent risk prediction.

Eye-tracking for longitudinal assessment of social cognition in children born preterm.

BACKGROUND AND OBJECTIVES: Preterm birth is associated with atypical social cognition in infancy, and cognitive impairment and social difficulties in childhood. Little is known about the stability of social cognition through childhood, and its relationship with neurodevelopment. We used eye-tracking in preterm and term-born infants to investigate social attentional preference in infancy and at 5 years, its relationship with neurodevelopment and the influence of socioeconomic deprivation. METHODS: A cohort of 81 preterm and 66 term infants with mean (range) gestational age at birth 28+5 (23+2 -33+0 ) and 40+0 (37+0 -42+1 ) respectively, completed eye-tracking at 7-9 months, with a subset re-assessed at 5 years. Three free-viewing social tasks of increasing stimulus complexity were presented, and a social preference score was derived from looking time to socially informative areas. Socioeconomic data and the Mullen Scales of Early Learning at 5 years were collected. RESULTS: Preterm children had lower social preference scores at 7-9 months compared with term-born controls. Term-born children's scores were stable between time points, whereas preterm children showed a significant increase, reaching equivalent scores by 5 years. Low gestational age and socioeconomic deprivation were associated with reduced social preference scores at 7-9 months. At 5 years, preterm infants had lower Early Learning Composite scores than controls, but this was not associated with social attentional preference in infancy or at 5 years. CONCLUSIONS: Preterm children have reduced social attentional preference at 7-9 months compared with term-born controls, but catch up by 5 years. Infant social cognition is influenced by socioeconomic deprivation and gestational age. Social cognition and neurodevelopment have different trajectories following preterm birth.

Measuring and reporting treatment adherence: What can we learn by comparing two respiratory conditions?

Medication non-adherence, defined as any deviation from the regimen recommended by their healthcare provider, can increase morbidity, mortality and side effects, while reducing effectiveness. Through studying two respiratory conditions, asthma and tuberculosis (TB), we thoroughly review the current understanding of the measurement and reporting of medication adherence. In this paper, we identify major methodological issues in the standard ways that adherence has been conceptualised, defined and studied in asthma and TB. Between and within the two diseases there are substantial variations in adherence reporting, linked to differences in dosing intervals and treatment duration. Critically, the communicable nature of TB has resulted in dose-by-dose monitoring becoming a recommended treatment standard. Through the lens of these similarities and contrasts, we highlight contemporary shortcomings in the generalised conceptualisation of medication adherence. Furthermore, we outline elements in which knowledge could be directly transferred from one condition to the other, such as the application of large-scale cost-effective monitoring methods in TB to resource-poor settings in asthma. To develop a more robust evidence-based approach, we recommend the use of standard taxonomies detailed in the ABC taxonomy when measuring and discussing adherence. Regimen and intervention development and use should be based on sufficient evidence of the commonality and type of adherence behaviours displayed by patients with the relevant condition. A systematic approach to the measurement and reporting of adherence could improve the value and generalisability of research across all health conditions.

Machine Learning to Predict the Likelihood of Acute Myocardial Infarction.

BACKGROUND: Variations in cardiac troponin concentrations by age, sex, and time between samples in patients with suspected myocardial infarction are not currently accounted for in diagnostic approaches. We aimed to combine these variables through machine learning to improve the assessment of risk for individual patients. METHODS: A machine learning algorithm (myocardial-ischemic-injury-index [MI3]) incorporating age, sex, and paired high-sensitivity cardiac troponin I concentrations, was trained on 3013 patients and tested on 7998 patients with suspected myocardial infarction. MI3 uses gradient boosting to compute a value (0-100) reflecting an individual's likelihood of a diagnosis of type 1 myocardial infarction and estimates the sensitivity, negative predictive value, specificity and positive predictive value for that individual. Assessment was by calibration and area under the receiver operating characteristic curve. Secondary analysis evaluated example MI3 thresholds from the training set that identified patients as low risk (99% sensitivity) and high risk (75% positive predictive value), and performance at these thresholds was compared in the test set to the 99th percentile and European Society of Cardiology rule-out pathways. RESULTS: Myocardial infarction occurred in 404 (13.4%) patients in the training set and 849 (10.6%) patients in the test set. MI3 was well calibrated with a very high area under the receiver operating characteristic curve of 0.963 [0.956-0.971] in the test set and similar performance in early and late presenters. Example MI3 thresholds identifying low- and high-risk patients in the training set were 1.6 and 49.7, respectively. In the test set, MI3 values were <1.6 in 69.5% with a negative predictive value of 99.7% (99.5-99.8%) and sensitivity of 97.8% (96.7-98.7%), and were ≥49.7 in 10.6% with a positive predictive value of 71.8% (68.9-75.0%) and specificity of 96.7% (96.3-97.1%). Using these thresholds, MI3 performed better than the European Society of Cardiology 0/3-hour pathway (sensitivity, 82.5% [74.5-88.8%]; specificity, 92.2% [90.7-93.5%]) and the 99th percentile at any time point (sensitivity, 89.6% [87.4-91.6%]); specificity, 89.3% [88.6-90.0%]). CONCLUSIONS: Using machine learning, MI3 provides an individualized and objective assessment of the likelihood of myocardial infarction, which can be used to identify low- and high-risk patients who may benefit from earlier clinical decisions. CLINICAL TRIAL REGISTRATION: URL: https://www.anzctr.org.au. Unique identifier: ACTRN12616001441404.

Telemedicine Cognitive Behavioral Therapy for Anxiety After Stroke: Proof-of-Concept Randomized Controlled Trial.

BACKGROUND AND PURPOSE: Disabling anxiety affects a quarter of stroke survivors but access to treatment is poor. We developed a telemedicine model for delivering guided self-help cognitive behavioral therapy (CBT) for anxiety after stroke (TASK-CBT). We aimed to evaluate the feasibility of TASK-CBT in a randomized controlled trial workflow that enabled all trial procedures to be carried out remotely. In addition, we explored the feasibility of wrist-worn actigraphy sensor as a way of measuring objective outcomes in this clinical trial. METHODS: We recruited adult community-based stroke patients (n=27) and randomly allocated them to TASK-CBT (n=14) or relaxation therapy (TASK-Relax), an active comparator (n=13). RESULTS: In our sample (mean age 65 [±10]; 56% men; 63% stroke, 37% transient ischemic attacks), remote self-enrolment, electronic signature, intervention delivery, and automated follow-up were feasible. All participants completed all TASK-CBT sessions (14/14). Lower levels of anxiety were observed in TASK-CBT when compared with TASK-Relax at both weeks 6 and 20. Mean actigraphy sensor wearing-time was 33 days (±15). CONCLUSIONS: Our preliminary feasibility data from the current study support a larger definitive clinical trial and the use of wrist-worn actigraphy sensor in anxious stroke survivors. Registration: URL: https://www.clinicaltrials.gov. Unique identifier: NCT03439813.

Linkage of primary care prescribing records and pharmacy dispensing Records in the Salford Lung Study: application in asthma.

BACKGROUND: Records of medication prescriptions can be used in conjunction with pharmacy dispensing records to investigate the incidence of adherence, which is defined as observing the treatment plans agreed between a patient and their clinician. Using prescribing records alone fails to identify primary non-adherence; medications not being collected from the dispensary. Using dispensing records alone means that cases of conditions that resolve and/or treatments that are discontinued will be unaccounted for. While using a linked prescribing and dispensing dataset to measure medication non-adherence is optimal, this linkage is not routinely conducted. Furthermore, without a unique common event identifier, linkage between these two datasets is not straightforward. METHODS: We undertook a secondary analysis of the Salford Lung Study dataset. A novel probabilistic record linkage methodology was developed matching asthma medication pharmacy dispensing records and primary care prescribing records, using semantic (meaning) and syntactic (structure) harmonization, domain knowledge integration, and natural language feature extraction. Cox survival analysis was conducted to assess factors associated with the time to medication dispensing after the prescription was written. Finally, we used a simplified record linkage algorithm in which only identical records were matched, for a naïve benchmarking to compare against the results of our proposed methodology. RESULTS: We matched 83% of pharmacy dispensing records to primary care prescribing records. Missing data were prevalent in the dispensing records which were not matched - approximately 60% for both medication strength and quantity. A naïve benchmarking approach, requiring perfect matching, identified one-quarter as many matching prescribing records as our methodology. Factors associated with delay (or failure) to collect the prescribed medication from a pharmacy included season, quantity of medication prescribed, previous dispensing history and class of medication. Our findings indicate that over 30% of prescriptions issued were not collected from a dispensary (primary non-adherence). CONCLUSIONS: We have developed a probabilistic record linkage methodology matching a large percentage of pharmacy dispensing records with primary care prescribing records for asthma medications. This will allow researchers to link datasets in order to extract information about asthma medication non-adherence.

Applications of Machine Learning in Real-Life Digital Health Interventions: Review of the Literature.

BACKGROUND: Machine learning has attracted considerable research interest toward developing smart digital health interventions. These interventions have the potential to revolutionize health care and lead to substantial outcomes for patients and medical professionals. OBJECTIVE: Our objective was to review the literature on applications of machine learning in real-life digital health interventions, aiming to improve the understanding of researchers, clinicians, engineers, and policy makers in developing robust and impactful data-driven interventions in the health care domain. METHODS: We searched the PubMed and Scopus bibliographic databases with terms related to machine learning, to identify real-life studies of digital health interventions incorporating machine learning algorithms. We grouped those interventions according to their target (ie, target condition), study design, number of enrolled participants, follow-up duration, primary outcome and whether this had been statistically significant, machine learning algorithms used in the intervention, and outcome of the algorithms (eg, prediction). RESULTS: Our literature search identified 8 interventions incorporating machine learning in a real-life research setting, of which 3 (37%) were evaluated in a randomized controlled trial and 5 (63%) in a pilot or experimental single-group study. The interventions targeted depression prediction and management, speech recognition for people with speech disabilities, self-efficacy for weight loss, detection of changes in biopsychosocial condition of patients with multiple morbidity, stress management, treatment of phantom limb pain, smoking cessation, and personalized nutrition based on glycemic response. The average number of enrolled participants in the studies was 71 (range 8-214), and the average follow-up study duration was 69 days (range 3-180). Of the 8 interventions, 6 (75%) showed statistical significance (at the P=.05 level) in health outcomes. CONCLUSIONS: This review found that digital health interventions incorporating machine learning algorithms in real-life studies can be useful and effective. Given the low number of studies identified in this review and that they did not follow a rigorous machine learning evaluation methodology, we urge the research community to conduct further studies in intervention settings following evaluation principles and demonstrating the potential of machine learning in clinical practice.

Developing a large scale population screening tool for the assessment of Parkinson's disease using telephone-quality voice.

Recent studies have demonstrated that analysis of laboratory-quality voice recordings can be used to accurately differentiate people diagnosed with Parkinson's disease (PD) from healthy controls (HCs). These findings could help facilitate the development of remote screening and monitoring tools for PD. In this study, 2759 telephone-quality voice recordings from 1483 PD and 15 321 recordings from 8300 HC participants were analyzed. To account for variations in phonetic backgrounds, data were acquired from seven countries. A statistical framework for analyzing voice was developed, whereby 307 dysphonia measures that quantify different properties of voice impairment, such as breathiness, roughness, monopitch, hoarse voice quality, and exaggerated vocal tremor, were computed. Feature selection algorithms were used to identify robust parsimonious feature subsets, which were used in combination with a random forests (RFs) classifier to accurately distinguish PD from HC. The best tenfold cross-validation performance was obtained using Gram-Schmidt orthogonalization and RF, leading to mean sensitivity of 64.90% (standard deviation, SD, 2.90%) and mean specificity of 67.96% (SD 2.90%). This large scale study is a step forward toward assessing the development of a reliable, cost-effective, and practical clinical decision support tool for screening the population at large for PD using telephone-quality voice.

Robust fundamental frequency estimation in sustained vowels: detailed algorithmic comparisons and information fusion with adaptive Kalman filtering.

There has been consistent interest among speech signal processing researchers in the accurate estimation of the fundamental frequency (F(0)) of speech signals. This study examines ten F(0) estimation algorithms (some well-established and some proposed more recently) to determine which of these algorithms is, on average, better able to estimate F(0) in the sustained vowel /a/. Moreover, a robust method for adaptively weighting the estimates of individual F(0) estimation algorithms based on quality and performance measures is proposed, using an adaptive Kalman filter (KF) framework. The accuracy of the algorithms is validated using (a) a database of 117 synthetic realistic phonations obtained using a sophisticated physiological model of speech production and (b) a database of 65 recordings of human phonations where the glottal cycles are calculated from electroglottograph signals. On average, the sawtooth waveform inspired pitch estimator and the nearly defect-free algorithms provided the best individual F(0) estimates, and the proposed KF approach resulted in a ∼16% improvement in accuracy over the best single F(0) estimation algorithm. These findings may be useful in speech signal processing applications where sustained vowels are used to assess vocal quality, when very accurate F(0) estimation is required.

Symptom tracking in endometriosis using digital technologies: Knowns, unknowns, and future prospects.

Endometriosis is a common chronic pain condition with no known cure and limited treatment options. Digital technologies, ranging from smartphone apps to wearable sensors, have shown potential toward facilitating chronic pain assessment and management; however, to date, many of these tools have not been specifically deployed or evaluated in patients with endometriosis-associated pain. Informed by previous studies in related chronic pain conditions, we discuss how digital technologies may be used in endometriosis to facilitate objective, continuous, and holistic symptom tracking. We postulate that these pervasive and increasingly affordable technologies present promising opportunities toward developing decision-support tools assisting healthcare professionals and empowering patients with endometriosis to make better-informed choices about symptom management.

Quantifying Parkinson's disease severity using mobile wearable devices and machine learning: the ParkApp pilot study protocol.

INTRODUCTION: The clinical assessment of Parkinson's disease (PD) symptoms can present reliability issues and, with visits typically spaced apart 6 months, can hardly capture their frequent variability. Smartphones and smartwatches along with signal processing and machine learning can facilitate frequent, remote, reliable and objective assessments of PD from patients' homes. AIM: To investigate the feasibility, compliance and user experience of passively and actively measuring symptoms from home environments using data from sensors embedded in smartphones and a wrist-wearable device. METHODS AND ANALYSIS: In an ongoing clinical feasibility study, participants with a confirmed PD diagnosis are being recruited. Participants perform activity tests, including Timed Up and Go (TUG), tremor, finger tapping, drawing and vocalisation, once a week for 2 months using the Mobistudy smartphone app in their homes. Concurrently, participants wear the GENEActiv wrist device for 28 days to measure actigraphy continuously. In addition to using sensors, participants complete the Beck's Depression Inventory, Non-Motor Symptoms Questionnaire (NMSQuest) and Parkinson's Disease Questionnaire (PDQ-8) questionnaires at baseline, at 1 month and at the end of the study. Sleep disorders are assessed through the Parkinson's Disease Sleep Scale-2 questionnaire (weekly) and a custom sleep quality daily questionnaire. User experience questionnaires, Technology Acceptance Model and User Version of the Mobile Application Rating Scale, are delivered at 1 month. Clinical assessment (Movement Disorder Society-Unified Parkinson Disease Rating Scale (MDS-UPDRS)) is performed at enrollment and the 2-month follow-up visit. During visits, a TUG test is performed using the smartphone and the G-Walk motion sensor as reference device. Signal processing and machine learning techniques will be employed to analyse the data collected from Mobistudy app and the GENEActiv and correlate them with the MDS-UPDRS. Compliance and user aspects will be informing the long-term feasibility. ETHICS AND DISSEMINATION: The study received ethical approval by the Swedish Ethical Review Authority (Etikprövningsmyndigheten), with application number 2022-02885-01. Results will be reported in peer-reviewed journals and conferences. Results will be shared with the study participants.

Defining clinical subtypes of adult asthma using electronic health records: Analysis of a large UK primary care database with external validation.

INTRODUCTION: Asthma is one of the commonest chronic conditions in the world. Subtypes of asthma have been defined, typically from clinical datasets on small, well-characterised subpopulations of asthma patients. We sought to define asthma subtypes from large longitudinal primary care electronic health records (EHRs) using cluster analysis. METHODS: In this retrospective cohort study, we extracted asthma subpopulations from the Optimum Patient Care Research Database (OPCRD) to robustly train and test algorithms, and externally validated findings in the Secure Anonymised Information Linkage (SAIL) Databank. In both databases, we identified adults with an asthma diagnosis code recorded in the three years prior to an index date. Train and test datasets were selected from OPCRD using an index date of Jan 1, 2016. Two internal validation datasets were selected from OPCRD using index dates of Jan 1, 2017 and 2018. Three external validation datasets were selected from SAIL using index dates of Jan 1, 2016, 2017 and 2018. Each dataset comprised 50,000 randomly selected non-overlapping patients. Subtypes were defined by applying multiple correspondence analysis and k-means cluster analysis to the train dataset, and were validated in the internal and external validation datasets. RESULTS: We defined six asthma subtypes with clear clinical interpretability: low inhaled corticosteroid (ICS) use and low healthcare utilisation (30% of patients); low-to-medium ICS use (36%); low-to-medium ICS use and comorbidities (12%); varied ICS use and comorbid chronic obstructive pulmonary disease (4%); high (10%) and very high ICS use (7%). The subtypes were replicated with high accuracy in internal (91-92%) and external (84-86%) datasets. CONCLUSION: Asthma subtypes derived and validated in large independent EHR databases were primarily defined by level of ICS use, level of healthcare use, and presence of comorbidities. This has important clinical implications towards defining asthma subtypes, facilitating patient stratification, and developing more personalised monitoring and treatment strategies.

Association of Preterm Birth and Socioeconomic Status With Neonatal Brain Structure.

Importance: Preterm birth and socioeconomic status (SES) are associated with brain structure in childhood, but the relative contributions of each during the neonatal period are unknown. Objective: To investigate associations of birth gestational age (GA) and SES with neonatal brain morphology by testing 3 hypotheses: GA and SES are associated with brain morphology; associations between SES and brain morphology vary with GA; and associations between SES and brain structure and morphology depend on how SES is operationalized. Design, Setting, and Participants: This cohort study recruited participants from November 2016 to September 2021 at a single center in the United Kingdom. Participants were 170 extremely and very preterm infants and 91 full-term or near-term infants. Exclusion criteria were major congenital malformation, chromosomal abnormality, congenital infection, cystic periventricular leukomalacia, hemorrhagic parenchymal infarction, and posthemorrhagic ventricular dilatation. Exposures: Birth GA and SES, operationalized at the neighborhood level (using the Scottish Index of Multiple Deprivation), the family level (using parental education and occupation), and subjectively (World Health Organization Quality of Life measure). Main Outcomes and Measures: Brain volume (85 parcels) and 5 whole-brain cortical morphology measures (gyrification index, thickness, sulcal depth, curvature, surface area) at term-equivalent age (median [range] age, 40 weeks, 5 days [36 weeks, 2 days to 45 weeks, 6 days] and 42 weeks [38 weeks, 2 days to 46 weeks, 1 day] for preterm and full-term infants, respectively). Results: Participants were 170 extremely and very preterm infants (95 [55.9%] male; 4 of 166 [2.4%] Asian, 145 of 166 [87.3%] White) and 91 full-term or near-term infants (50 [54.9%] male; 3 of 86 [3.5%] Asian, 78 of 86 [90.7%] White infants) with median (range) birth GAs of 30 weeks, 0 days (22 weeks, 1 day, to 32 weeks, 6 days) and 39 weeks, 4 days (36 weeks, 3 days, to 42 weeks, 1 day), respectively. In fully adjusted models, birth GA was associated with a higher proportion of brain volumes (27 of 85 parcels [31.8%]; ß range, -0.20 to 0.24) than neighborhood-level SES (1 of 85 parcels [1.2%]; ß = 0.17 [95% CI, -0.16 to 0.50]) or family-level SES (maternal education: 4 of 85 parcels [4.7%]; ß range, 0.09 to 0.15; maternal occupation: 1 of 85 parcels [1.2%]; ß = 0.06 [95% CI, 0.02 to 0.11] respectively). There were interactions between GA and both family-level and subjective SES measures on regional brain volumes. Birth GA was associated with cortical surface area (ß = 0.10 [95% CI, 0.02 to 0.18]) and gyrification index (ß = 0.16 [95% CI, 0.07 to 0.25]); no SES measure was associated with cortical measures. Conclusions and Relevance: In this cohort study of UK infants, birth GA and SES were associated with neonatal brain morphology, but low GA had more widely distributed associations with neonatal brain structure than SES. Further work is warranted to elucidate the mechanisms underlying the association of both GA and SES with early brain development.

Relevance, redundancy, and complementarity trade-off (RRCT): A principled, generic, robust feature-selection tool.

We present a new heuristic feature-selection (FS) algorithm that integrates in a principled algorithmic framework the three key FS components: relevance, redundancy, and complementarity. Thus, we call it relevance, redundancy, and complementarity trade-off (RRCT). The association strength between each feature and the response and between feature pairs is quantified via an information theoretic transformation of rank correlation coefficients, and the feature complementarity is quantified using partial correlation coefficients. We empirically benchmark the performance of RRCT against 19 FS algorithms across four synthetic and eight real-world datasets in indicative challenging settings evaluating the following: (1) matching the true feature set and (2) out-of-sample performance in binary and multi-class classification problems when presenting selected features into a random forest. RRCT is very competitive in both tasks, and we tentatively make suggestions on the generalizability and application of the best-performing FS algorithms across settings where they may operate effectively.

Estimating Medication Adherence from Electronic Health Records Using Rolling Averages of Single Refill-based Estimates.

Medication adherence is usually defined as the manner in which a patient takes their medication, in relation to the regimen agreed to with their healthcare provider. Electronic Health Records (EHRs) can be used to estimate adherence in a cost-effective and non-invasive manner across large-scale populations, although there is no universally agreed optimal approach to doing so. We sought to explore patterns of asthma ICS prescription refills in a large EHR dataset, and to evaluate the use of rolling-average based measures towards short-term adherence estimation. Over 1.6 million asthma controllers were prescribed for our cohort of 91,332 individuals, between January 2009 and March 2017. The Continuous Single interval measures of medication Availability (CSA) and Gaps (CSG) were calculated for individual prescriptions, as well as rolling-average adherence measures of the CSA over 3, 5, or 10 past prescription intervals. 16.7% of the study population had only a single prescription during their follow-up (a median duration of 7.1 years). 51% of prescriptions were refilled before (or on the day that) supply was exhausted, and for 19% of prescription refills, the amount of medication dispensed should have lasted at least twice as long as the duration before the next refill was filled. The rolling average measures had statistically strong associations (Spearman |R|>0.7) with the estimate for the subsequent prescription refill. Rolling averages of multiple individual refill-level adherence estimates provide a novel and simple way to crudely smoothen estimates from individual prescription refills, which are strongly influenced by common (and adherent) real-world behaviors, for more meaningful and effective trend detection. Clinical Relevance- This demonstrates a novel methodology for estimating medication adherence which can detect recent changes in trends.