Reflection on modern methods: generalized linear models for prognosis and intervention-theory, practice and implications for machine learning

Use of Stacked Proportional Bar Graphs ("Grotta Bars") in Observational Neurology Research: A Meta-Research Study

BACKGROUND AND OBJECTIVES

Stacked proportional bar graphs (nicknamed "Grotta bars") are commonly used to visualize functional outcome scales in stroke research and are also used in other domains of neurology research. While lending themselves to a straightforward causal interpretation in ideal randomized controlled trials, in observational studies, Grotta bars cannot be generally interpreted causally if they show unadjusted, confounded comparisons. In a sample of recent observational neurology studies with confounding-adjusted effect estimates, we aimed to determine the frequency with which Grotta bars were used to visualize functional outcomes and how often unadjusted Grotta bars were presented without an accompanying adjusted version. We also assessed the methods used to generate adjusted Grotta bars.

METHODS

We identified the 15 top-ranked clinical neurology journals, according to journal impact factor, publishing full-length original research in English. Using PubMed, we retrieved all records published in these journals between 2020 and 2021 after applying a filter for observational studies. We included and systematically examined all observational studies aiming to identify a cause-and-effect relationship with an ordinal functional outcome and confounding-adjusted effect estimate. We determined whether at least 1 comparison using Grotta bars was present, whether the visualized comparisons were adjusted, and which adjustment strategies were applied to generate these graphs.

RESULTS

A total of 250 studies met all inclusion criteria. Of these, 93 (37.2%) used Grotta bars to depict functional outcome scale distributions, with 76 (81.7%) presenting only Grotta bars without model-based adjustment. These bars were most commonly presented in studies with stroke patient populations; 87 of 192 studies (45.3%) presented Grotta bars. Among the 17 studies that presented Grotta bars adjusted using a model, the adjustment strategies included propensity score matching (n = 10; 58.8%), regression (n = 6; 35.3%), and inverse probability weighting (n = 1; 5.9%).

DISCUSSION

Studies that presented adjusted associations for functional outcomes commonly showed only unadjusted Grotta bars, which alone have little value for causal questions. In observational research, Grotta bars are most informative if an adjusted version, aligning with adjusted effect estimates, is presented directly alongside the unadjusted version. Based on our findings, we offer recommendations to help authors generate more informative Grotta bars and to facilitate correct interpretation for readers.

Explainable artificial intelligence identifies an AQP4 polymorphism-based risk score associated with brain amyloid burden

Aquaporin-4 (AQP4) is hypothesized to be a component of the glymphatic system, a pathway for removing brain interstitial solutes like amyloid-β (Aβ). Evidence exists that genetic variation of AQP4 impacts Aβ clearance, clinical outcome in Alzheimer's disease as well as sleep measures. We examined whether a risk score calculated from several AQP4 single-nucleotide polymorphisms (SNPs) is related to Aβ neuropathology in older cognitively unimpaired white individuals. We used a machine learning approach and explainable artificial intelligence to extract information on synergistic effects of AQP4 SNPs on brain amyloid burden from the ADNI cohort. From this information, we formulated a sex-specific AQP4 SNP-based risk score and evaluated it using data from the screening process of the A4 study. We found in both cohorts significant associations of the risk score with brain amyloid burden. The results support the hypothesis of an involvement of the glymphatic system, and particularly AQP4, in brain amyloid aggregation pathology. They suggest also that different AQP4 SNPs exert a synergistic effect on the build-up of brain amyloid burden.

Guidance for unbiased predictive information for healthcare decision-making and equity (GUIDE): considerations when race may be a prognostic factor

Clinical prediction models (CPMs) are tools that compute the risk of an outcome given a set of patient characteristics and are routinely used to inform patients, guide treatment decision-making, and resource allocation. Although much hope has been placed on CPMs to mitigate human biases, CPMs may potentially contribute to racial disparities in decision-making and resource allocation. While some policymakers, professional organizations, and scholars have called for eliminating race as a variable from CPMs, others raise concerns that excluding race may exacerbate healthcare disparities and this controversy remains unresolved. The Guidance for Unbiased predictive Information for healthcare Decision-making and Equity (GUIDE) provides expert guidelines for model developers and health system administrators on the transparent use of race in CPMs and mitigation of algorithmic bias across contexts developed through a 5-round, modified Delphi process from a diverse 14-person technical expert panel (TEP). Deliberations affirmed that race is a social construct and that the goals of prediction are distinct from those of causal inference, and emphasized: the importance of decisional context (e.g., shared decision-making versus healthcare rationing); the conflicting nature of different anti-discrimination principles (e.g., anticlassification versus antisubordination principles); and the importance of identifying and balancing trade-offs in achieving equity-related goals with race-aware versus race-unaware CPMs for conditions where racial identity is prognostically informative. The GUIDE, comprising 31 key items in the development and use of CPMs in healthcare, outlines foundational principles, distinguishes between bias and fairness, and offers guidance for examining subgroup invalidity and using race as a variable in CPMs. This GUIDE presents a living document that supports appraisal and reporting of bias in CPMs to support best practice in CPM development and use.

Neuroma-to-Nerve Ratio: Does Size Matter?

BACKGROUND AND OBJECTIVES

Anatomic features of neuromas have been explored in imaging studies. However, there has been limited research into these features using resected, ex vivo human neuroma specimens. The aim of this study was to investigate the influence that time may have on neuroma growth and size, and the clinical significance of these parameters.

METHODS

Patients who underwent neuroma excision between 2022 through 2023 were prospectively included in this study. Neuroma specimens were obtained after operative resection. Standardized neuroma size measurements, expressed as a neuroma-to-nerve ratio (NNR), were conducted with ImageJ software. Pain data (numeric rating scale, 0-10) were prospectively recorded during preoperative evaluation, and patient factors were collected from chart reviews.

RESULTS

Fifty terminal neuroma specimens from 31 patients were included, with 94.0% of the neuromas obtained from individuals with amputations. Most neuromas were excised from the lower extremities (n = 44, 88.0%). The neuromas had a median NNR of 2.45, and the median injury to neuroma excision interval was 6.3 years. Larger NNRs were associated with a longer injury to neuroma excision interval and with a smaller native nerve diameter. In addition, sensory nerves were associated with a larger NNR compared with mixed nerves. NNR was not associated with preoperative pain or with anatomical nerve distribution.

CONCLUSION

This study suggests that neuromas seem to continue to grow over time and that smaller nerves may form relatively larger neuromas. In addition, sensory nerves develop relatively larger neuromas compared with mixed nerves. Neuroma size does not appear to correlate with pain severity. These findings may stimulate future research efforts and contribute to a better understanding of symptomatic neuroma development.

Clinician perspectives and recommendations regarding design of clinical prediction models for deteriorating patients in acute care

BACKGROUND

Successful deployment of clinical prediction models for clinical deterioration relates not only to predictive performance but to integration into the decision making process. Models may demonstrate good discrimination and calibration, but fail to match the needs of practising acute care clinicians who receive, interpret, and act upon model outputs or alerts. We sought to understand how prediction models for clinical deterioration, also known as early warning scores (EWS), influence the decision-making of clinicians who regularly use them and elicit their perspectives on model design to guide future deterioration model development and implementation.

METHODS

Nurses and doctors who regularly receive or respond to EWS alerts in two digital metropolitan hospitals were interviewed for up to one hour between February 2022 and March 2023 using semi-structured formats. We grouped interview data into sub-themes and then into general themes using reflexive thematic analysis. Themes were then mapped to a model of clinical decision making using deductive framework mapping to develop a set of practical recommendations for future deterioration model development and deployment.

RESULTS

Fifteen nurses (n = 8) and doctors (n = 7) were interviewed for a mean duration of 42 min. Participants emphasised the importance of using predictive tools for supporting rather than supplanting critical thinking, avoiding over-protocolising care, incorporating important contextual information and focusing on how clinicians generate, test, and select diagnostic hypotheses when managing deteriorating patients. These themes were incorporated into a conceptual model which informed recommendations that clinical deterioration prediction models demonstrate transparency and interactivity, generate outputs tailored to the tasks and responsibilities of end-users, avoid priming clinicians with potential diagnoses before patients were physically assessed, and support the process of deciding upon subsequent management.

CONCLUSIONS

Prediction models for deteriorating inpatients may be more impactful if they are designed in accordance with the decision-making processes of acute care clinicians. Models should produce actionable outputs that assist with, rather than supplant, critical thinking.

Contesting the conventional wisdom of periodontal risk assessment

Over the years, several reviews of periodontal risk assessment tools have been published. However, major misunderstandings still prevail in repeated attempts to use these tools for prognostic risk prediction. Here we review the principles of risk prediction and discuss the value and the challenges of using prediction models in periodontology. Most periodontal risk prediction models have not been properly developed according to guidance given for the risk prediction model development. This shortcoming has led to several problems, including the creation of arbitrary risk scores. These scores are often labelled as 'high risk' without explicit boundaries or thresholds for the underlying continuous risk estimates of patient-important outcomes. Moreover, it is apparent that prediction models are often misinterpreted as causal models by clinicians and researchers although they cannot be used as such. Additional challenges like the critical assessment of transportability and applicability of these prediction models, as well as their impact on clinical practice and patient outcomes, are not considered in the literature. Nevertheless, these instruments are promoted with claims regarding their ability to deliver more individualized and precise periodontitis treatment and prevention, purportedly resulting in improved patient outcomes. However, people with or without periodontitis deserve proper information about their risk of developing patient-important outcomes such as tooth loss or pain. The primary objective of disseminating such information should not be to emphasize assumed treatment efficacy, hype individualization of care, or promote business interests. Instead, the focus should be on providing individuals with locally validated and regularly updated predictions of specific risks based on readily accessible and valid key predictors (e.g. age and smoking).

Human Substantia Nigra Neurons Encode Reward Expectations

Dopamine (DA) signals originating from substantia nigra (SN) neurons are centrally involved in the regulation of motor and reward processing. DA signals behaviorally relevant events where reward outcomes differ from expectations (reward prediction errors, RPEs). RPEs play a crucial role in learning optimal courses of action and in determining response vigor when an agent expects rewards. Nevertheless, how reward expectations, crucial for RPE calculations, are conveyed to and represented in the dopaminergic system is not fully understood, especially in the human brain where the activity of DA neurons is difficult to study. One possibility, suggested by evidence from animal models, is that DA neurons explicitly encode reward expectations. Alternatively, they may receive RPE information directly from upstream brain regions. To address whether SN neuron activity directly reflects reward expectation information, we directly examined the encoding of reward expectation signals in human putative DA neurons by performing single-unit recordings from the SN of patients undergoing neurosurgery. Patients played a two-armed bandit decision-making task in which they attempted to maximize reward. We show that neuronal firing rates (FR) of putative DA neurons during the reward expectation period explicitly encode reward expectations. First, activity in these neurons was modulated by previous trial outcomes, such that FR were greater after positive outcomes than after neutral or negative outcome trials. Second, this increase in FR was associated with shorter reaction times, consistent with an invigorating effect of DA neuron activity during expectation. These results suggest that human DA neurons explicitly encode reward expectations, providing a neurophysiological substrate for a signal critical for reward learning.

Leveraging multivariate approaches to advance the science of early-life adversity

Since the landmark Adverse Childhood Experiences (ACEs) study, adversity research has expanded to more precisely account for the multifaceted nature of adverse experiences. The complex data structures and interrelated nature of adversity data require robust multivariate statistical methods, and recent methodological and statistical innovations have facilitated advancements in research on childhood adversity. Here, we provide an overview of a subset of multivariate methods that we believe hold particular promise for advancing the field's understanding of early-life adversity, and discuss how these approaches can be practically applied to explore different research questions. This review covers data-driven or unsupervised approaches (including dimensionality reduction and person-centered clustering/subtype identification) as well as supervised/prediction-based approaches (including linear and tree-based models and neural networks). For each, we highlight studies that have effectively applied the method to provide novel insight into early-life adversity. Taken together, we hope this review serves as a resource to adversity researchers looking to expand upon the cumulative approach described in the original ACEs study, thereby advancing the field's understanding of the complexity of adversity and related developmental consequences.

Trends in Lower Extremity Artery Disease Repair Incidence, Comorbidity, and Mortality: A Danish Nationwide Cohort Study, 1996-2018

Background

The prevalence of occlusive lower extremity artery disease (LEAD) is rising worldwide while European epidemiology data are scarce. We report incidence and mortality of LEAD repair in Denmark from 1996 through 2018, stratified on open aorto-iliac, open peripheral, and endovascular repair.

Methods

A nationwide cohort study of prospective data from population-based Danish registers covering 1996 to 2018. Comorbidity was assessed by Charlson's Comorbidity Index (CCI). Incidence rate (IR) ratios and mortality rate ratios (MRR) were estimated by multivariable Poisson and Cox regression, respectively.

Results

We identified 41,438 unique patients undergoing 46,236 incident first-time LEAD repairs by either aorto-iliac- (n=5213), peripheral surgery (n=18,665) or percutaneous transluminal angioplasty (PTA, n=22,358). From 1996 to 2018, the age- and sex-standardized IR for primary revascularization declined from 71.8 to 50.2 per 100,000 person-years (IRR, 0.70; 95% CI, 0.66-0.75). Following a 2.5-fold IR increase of PTA from 1996 to 2010, all three repair techniques showed a declining trend after 2010. The declining IR was driven by decreasing LEAD repair due to claudication, and by persons aged below 80 years, while the IR increased in persons aged above 80 years (p interaction<0.001). LEAD repair was more frequent in men (IRRfemale vs male, 0.78; 95% CI, 0.77-0.80), which was consistent over calendar time (p interaction=0.41). Crude mortality decreased following open/surgical repair, and increased following PTA, but all three techniques trended towards lower adjusted mortality comparing the start and the end of the study period (MRRaorto-iliac, 0.71; 95% CI, 0.54-0.93 vs MRRperipheral, 0.76; 95% CI, 0.69-0.83 vs MRRPTA, 0.96; 95% CI, 0.86-1.07). Increasing age and CCI, male sex, smoking, and care dependency associated with increased mortality.

Conclusion

The incidence rate of LEAD repair decreased in Denmark from 1996 to 2018, especially in persons younger than 80 years, and primarily due to reduced revascularization for claudication. Adjusted mortality rates decreased following open surgery, but seemed unaltered following PTA.

Documenting and modeling the acoustic variability of intervocalic alveolar taps in conversational Peninsular Spanish

This study constitutes an investigation into the acoustic variability of intervocalic alveolar taps in a corpus of spontaneous speech from Madrid, Spain. Substantial variability was documented in this segment, with highly reduced variants constituting roughly half of all tokens during spectrographic inspection. In addition to qualitative documentation, the intensity difference between the tap and surrounding vowels was measured. Changes in this intensity difference were statistically modeled using Bayesian finite mixture models containing lexical and phonetic predictors. Model comparisons indicate predictive performance is improved when we assume two latent categories, interpreted as two pronunciation variants for the Spanish tap. In interpreting the model, predictors were more often related to categorical changes in which pronunciation variant was produced than to gradient intensity changes within each tap type. Variability in tap production was found according to lexical frequency, speech rate, and phonetic environment. These results underscore the importance of evaluating model fit to the data as well as what researchers modeling phonetic variability can gain in moving past linear models when they do not adequately fit the observed data.

Real-world causal evidence for planned predictive enrichment in critical care trials: A scoping review

BACKGROUND

Randomised clinical trials in critical care are prone to inconclusiveness due, in part, to undue optimism about effect sizes and suboptimal accounting for heterogeneous treatment effects. Although causal evidence from rich real-world critical care can help overcome these challenges by informing predictive enrichment, no overview exists.

METHODS

We conducted a scoping review, systematically searching 10 general and speciality journals for reports published on or after 1 January 2018, of randomised clinical trials enrolling adult critically ill patients. We collected trial metadata on 22 variables including recruitment period, intervention type and early stopping (including reasons) as well as data on the use of causal evidence from secondary data for planned predictive enrichment.

RESULTS

We screened 9020 records and included 316 unique RCTs with a total of 268,563 randomised participants. One hundred seventy-three (55%) trials tested drug interventions, 101 (32%) management strategies and 42 (13%) devices. The median duration of enrolment was 2.2 (IQR: 1.3-3.4) years, and 83% of trials randomised less than 1000 participants. Thirty-six trials (11%) were restricted to COVID-19 patients. Of the 55 (17%) trials that stopped early, 23 (42%) used predefined rules; futility, slow enrolment and safety concerns were the commonest stopping reasons. None of the included RCTs had used causal evidence from secondary data for planned predictive enrichment.

CONCLUSION

Work is needed to harness the rich multiverse of critical care data and establish its utility in critical care RCTs. Such work will likely need to leverage methodology from interventional and analytical epidemiology as well as data science.

Impact of special educational needs provision on hospital utilisation, school attainment and absences for children in English primary schools stratified by gestational age at birth: A target trial emulation study protocol

Introduction

One third of children in English primary schools have additional learning support called special educational needs (SEN) provision, but children born preterm are more likely to have SEN than those born at term. We aim to assess the impact of SEN provision on health and education outcomes in children grouped by gestational age at birth.

Methods

We will analyse linked administrative data for England using the Education and Child Health Insights from Linked Data (ECHILD) database. A target trial emulation approach will be used to specify data extraction from ECHILD, comparisons of interest and our analysis plan. Our target population is all children enrolled in year one of state-funded primary school in England who were born in an NHS hospital in England between 2003 and 2008, grouped by gestational age at birth (extremely preterm (24-<28 weeks), very preterm (28-<32 weeks), moderately preterm (32-<34 weeks), late preterm (34-<37 weeks) and full term (37-<42 weeks). The intervention of interest will comprise categories of SEN provision (including none) during year one (age five/six). The outcomes of interest are rates of unplanned hospital utilisation, educational attainment, and absences by the end of primary school education (year six, age 11). We will triangulate results from complementary estimation methods including the naïve estimator, multivariable regression, g-formula, inverse probability weighting, inverse probability weighting with regression adjustment and instrumental variables, along with a variety for a variety of causal contrasts (average treatment effect, overall, and on the treated/not treated).

Ethics and dissemination

We have existing research ethics approval for analyses of the ECHILD database described in this protocol. We will disseminate our findings to diverse audiences (academics, relevant government departments, service users and providers) through seminars, peer-reviewed publications, short briefing reports and infographics for non-academics (published on the study website).

Polygenic Propensity for Longevity, APOE-ε4 Status, Dementia Diagnosis, and Risk for Cause-Specific Mortality: A Large Population-Based Longitudinal Study of Older Adults

To deepen the understanding of genetic mechanisms influencing mortality risk, we investigated the impact of genetic predisposition to longevity and APOE-ε4, on all-cause mortality and specific causes of mortality. We further investigated the mediating effects of dementia on these relationships. Using data on 7 131 adults aged ≥50 years (mean = 64.7 years, standard deviation [SD] = 9.5) from the English Longitudinal Study of Aging, genetic predisposition to longevity was calculated using the polygenic score approach (PGSlongevity). APOE-ε4 status was defined according to the absence or presence of ε4 alleles. The causes of death were ascertained from the National Health Service central register, which was classified into cardiovascular diseases, cancers, respiratory illness, and all other causes of mortality. Of the entire sample, 1 234 (17.3%) died during an average 10-year follow-up. One-SD increase in PGSlongevity was associated with a reduced risk for all-cause mortality (hazard ratio [HR] = 0.93, 95% confidence interval [CI]: 0.88-0.98, p = .010) and mortalities due to other causes (HR = 0.81, 95% CI: 0.71-0.93, p = .002) in the following 10 years. In gender-stratified analyses, APOE-ε4 status was associated with a reduced risk for all-cause mortality and mortalities related to cancers in women. Mediation analyses estimated that the percent excess risk of APOE-ε4 on other causes of mortality risk explained by the dementia diagnosis was 24%, which increased to 34% when the sample was restricted to adults who were aged ≤75 years old. To reduce the mortality rate in adults who are aged ≥50 years old, it is essential to prevent dementia onset in the general population.

Periodontitis and risk of immune-mediated systemic conditions: A systematic review and meta-analysis

INTRODUCTION

The aim of this review is to examine and quantify the long-term risk of immune-mediated systemic conditions in people with periodontitis compared to people without periodontitis.

METHODS

Medline, EMBASE and Cochrane databases were searched up to June 2022 using keywords and MeSH headings. The 'Risk of Bias in Non-Randomised Studies of Interventions' tool was used to assess bias. Cohort studies comparing incident metabolic/autoimmune/inflammatory diseases in periodontitis to healthy controls were included. Meta-analysis and meta-regression quantified risks and showed impact of periodontitis diagnosis type and severity.

RESULTS

The search retrieved 3354 studies; 166 studies were eligible for full-text screening, and 30 studies were included for review. Twenty-seven studies were eligible for meta-analysis. The risks of diabetes, rheumatoid arthritis (RA) and osteoporosis were increased in people with periodontitis compared to without periodontitis (diabetes-relative risk [RR]: 1.22, 95% CI: 1.13-1.33; RA-RR: 1.27, 95% CI: 1.07-1.52; osteoporosis-RR: 1.40, 95% CI: 1.12-1.75). Risk of diabetes showed gradient increase by periodontitis severity (moderate-RR = 1.20, 95% CI = 1.11-1.31; severe-RR = 1.34, 95% CI = 1.10-1.63).

CONCLUSION

People with moderate-to-severe cases of periodontitis have the highest risk of developing diabetes, while the effect of periodontal severity on risk of other immune-mediated systemic conditions requires further investigation. More homologous evidence is required to form robust conclusions regarding periodontitis-multimorbidity associations.

So many variables, but what causes what?

Paradis' keynote article provides a comprehensive overview of factors influencing bilingual children's dual language abilities. It includes the 'usual suspects', such as input quantity, and also highlights areas requiring further investigation, such as cognitive abilities. As such, it will no doubt serve as a valuable basis for the field as we move forward. Paradis quite rightly points out that whilst some of these factors may be unidirectionally related to language abilities, suggesting causality, for many others such relations are bi- or multidirectional and as such, caution is required in interpreting them. In order to pinpoint the nature and direction of these relations (currently absent from Figure 1 in the keynote), more complex analytic techniques are needed, as Paradis herself notes: "The relations among attitudes/identity, input and interaction, and perhaps social adjustment and wellbeing, are likely to be complex; therefore, more complex analytic techniques are needed to understand the path(s) between family attitudes about the HL on one hand, and children's HL outcomes on the other." (Paradis, 2023: 19). In this commentary, we provide an illustration of how the complex relationships between the variables discussed in Paradis's keynote article could be conceptualised within a causal inference approach. We offer a modest starting point by summarising key features of causal inference modelling and by illustrating how it might help us better understand what causes what.

The shared genetic architecture of smoking behaviours and psychiatric disorders: evidence from a population-based longitudinal study in England

BACKGROUND

Considering the co-morbidity of major psychiatric disorders and intelligence with smoking, to increase our understanding of why some people take up smoking or continue to smoke, while others stop smoking without progressing to nicotine dependence, we investigated the genetic propensities to psychiatric disorders and intelligence as determinants of smoking initiation, heaviness of smoking and smoking cessation in older adults from the general population.

RESULTS

Having utilised data from the English Longitudinal Study of Ageing (ELSA), our results showed that one standard deviation increase in MDD-PGS was associated with increased odds of being a moderate-heavy smoker (odds ratio [OR] = 1.11, SE = 0.04, 95%CI = 1.00-1.24, p = 0.028). There were no other significant associations between SZ-PGS, BD-PGS, or IQ-PGS and smoking initiation, heaviness of smoking and smoking cessation in older adults from the general population in the UK.

CONCLUSIONS

Smoking is a behaviour that does not appear to share common genetic ground with schizophrenia, bipolar disorders, and intelligence in older adults, which may suggest that it is more likely to be modifiable by smoking cessation interventions. Once started to smoke, older adults with a higher polygenic predisposition to major depressive disorders are more likely to be moderate to heavy smokers, implying that these adults may require targeted smoking cessation services.

The long-term effects of a polygenetic predisposition to general cognition on healthy cognitive ageing: evidence from the English Longitudinal Study of Ageing

BACKGROUND

As an accelerated cognitive decline frequently heralds onset of severe neuropathological disorders, understanding the source of individual differences in withstanding the onslaught of cognitive ageing may highlight how best cognitive abilities may be retained into advanced age.

METHODS

Using a population representative sample of 5088 adults aged •50 years from the English Longitudinal Study of Ageing, we investigated relationships of polygenic predisposition to general cognition with a rate of change in cognition during a 10-year follow-up period. Polygenic predisposition was measured with polygenic scores for general cognition (GC-PGS). Cognition was measured employing tests for verbal memory and semantic fluency.

RESULTS

The average baseline memory score was 11.1 (s.d. = 2.9) and executive function score was 21.5 (s.d. = 5.8). An increase in GC-PGS by one standard deviation (1-s.d.) was associated with a higher baseline verbal memory by an average 0.27 points (95% CI 0.19-0.34, p < 0.001). Similarly, 1-s.d. increase in GC-PGS was associated with a higher semantic fluency score at baseline in the entire sample (β = 0.45, 95% CI 0.27-0.64, p < 0.001). These associations were significant for women and men, and all age groups. Nonetheless, 1-s.d. increase in GC-PGS was not associated with decreases in verbal memory nor semantic fluency during follow-up in the entire sample, as well stratified models by sex and age.

CONCLUSION

Although common genetic variants associated with general cognition additively are associated with a stable surplus to cognition in adults, a polygenic predisposition to general cognition is not associated with age-related cognitive decline during a 10-year follow-up.

Quantitative prediction of radiographic progression in patients with axial spondyloarthritis using neural network model in a real-world setting

BACKGROUND

Predicting radiographic progression in axial spondyloarthritis (axSpA) remains limited because of the complex interaction between multiple associated factors and individual variability in real-world settings. Hence, we tested the feasibility of artificial neural network (ANN) models to predict radiographic progression in axSpA.

METHODS

In total, 555 patients with axSpA were split into training and testing datasets at a 3:1 ratio. A generalized linear model (GLM) and ANN models were fitted based on the baseline clinical characteristics and treatment-dependent variables for the modified Stoke Ankylosing Spondylitis Spine Score (mSASSS) of the radiographs at follow-up time points. The mSASSS prediction was evaluated, and explainable machine learning methods were used to provide insights into the model outcome or prediction.

RESULTS

The R² values of the fitted models were in the range of 0.90-0.95 and ANN with an input of mSASSS as the number of each score performed better (root mean squared error (RMSE) = 2.83) than GLM or input of mSASSS as a total score (RMSE = 2.99-3.57). The ANN also effectively captured complex interactions among variables and their contributions to the transition of mSASSS over time in the fitted models. Structural changes constituting the mSASSS scoring systems were the most important contributing factors, and no detectable structural abnormalities at baseline were the most significant factors suppressing mSASSS change.

CONCLUSIONS

Clinical and radiographic data-driven ANN allows precise mSASSS prediction in real-world settings. Correct evaluation and prediction of spinal structural changes could be beneficial for monitoring patients with axSpA and developing a treatment plan.

Assessing spatiotemporal variability in SARS-CoV-2 infection risk for hospital workers using routinely-collected data

The COVID-19 pandemic has emphasised the need to rapidly assess infection risks for healthcare workers within the hospital environment. Using data from the first year of the pandemic, we investigated whether an individual's COVID-19 test result was associated with behavioural markers derived from routinely collected hospital data two weeks prior to a test. The temporal and spatial context of behaviours were important, with the highest risks of infection during the first wave, for staff in contact with a greater number of patients and those with greater levels of activity on floors handling the majority of COVID-19 patients. Infection risks were higher for BAME staff and individuals working more shifts. Night shifts presented higher risks of infection between waves of COVID-19 patients. Our results demonstrate the epidemiological relevance of deriving markers of staff behaviour from electronic records, which extend beyond COVID-19 with applications for other communicable diseases and in supporting pandemic preparedness.

The healthy donor effect and survey participation, becoming a donor and donor career

BACKGROUND

The healthy donor effect (HDE) is a selection bias caused by the health criteria blood donors must meet. It obscures investigations of beneficial/adverse health effects of blood donation and complicates the generalizability of findings from blood donor cohorts. To further characterize the HDE we investigated how self-reported health and lifestyle are associated with becoming a blood donor, lapsing, and donation intensity. Furthermore, we examined differences in mortality based on donor status.

STUDY DESIGN AND METHODS

The Danish National Health Survey was linked to the Scandinavian Donations and Transfusions (SCANDAT) database and Danish register data. Logistic- and normal regression was used to compare baseline characteristics and participation. Poisson regression was used to investigate future donation choices. Donation intensity was explored by the Anderson-Gill model and Poisson regression. Mortality was investigated using Poisson regression.

RESULTS

Blood donors were more likely to participate in the surveys, OR = 2.45 95% confidence interval (2.40-2.49) than non-donors. Among survey participants, better self-reported health and healthier lifestyle were associated with being or becoming a blood donor, donor retention, and to some extent donation intensity, for example, current smoking conveyed lower likelihood of becoming a donor, OR = 0.70 (0.66-0.75). We observed lower mortality for donors and survey participants, respectively, compared with non-participating non-donors.

CONCLUSION

We provide evidence that blood donation is associated with increased likelihood to participate in health surveys, possibly a manifestation of the HDE. Furthermore, becoming a blood donor, donor retention, and donation intensity was associated with better self-reported health and healthier lifestyles.

Causal inference for planning randomised critical care trials: Protocol for a scoping review

BACKGROUND

Randomised clinical trials in critical care are prone to inconclusiveness owing, in part, to undue optimism about effect sizes and suboptimal accounting for heterogeneous treatment effects. Planned predictive enrichment based on secondary critical care data (often very rich with respect to both data types and temporal granularity) and causal inference methods may help overcome these challenges, but no overview exists about their use to this end.

METHODS

We will conduct a scoping review to assess the extent and nature of the use of causal inference from secondary data for planned predictive enrichment of randomised clinical trials in critical care. We will systematically search 10 general and specialty journals for reports published on or after 1 January 2018, of randomised clinical trials enrolling adult critically ill patients. We will collect trial metadata (e.g., recruitment period and phase) and, when available, information pertaining to the focus of the review (predictive enrichment based on causal inference estimates from secondary data): causal inference methods, estimation techniques and software used; types of patient populations; data provenance, types and models; and the availability of the data (public or not). The results will be reported in a descriptive manner.

DISCUSSION

The outlined scoping review aims to assess the use of causal inference methods and secondary data for planned predictive enrichment in randomised critical care trials. This will help guide methodological improvements to increase the utility, and facilitate the use, of causal inference estimates when planning such trials in the future.

Interactive Visualization and Computation of 2D and 3D Probability Distributions

Purpose

Mathematical modeling, probability estimation, and statistical inference represent core elements of modern artificial intelligence (AI) approaches for data-driven prediction, forecasting, classification, risk-estimation, and prognosis. Currently there are many tools that help calculate and visualize univariate probability distributions, however, very few resources venture beyond into multivariate distributions, which are commonly used in advanced statistical inference and AI decision-making. This article presents a new web-calculator that enables some calculation and visualization of bivariate and trivariate probability distributions.

Methods

Several methods are explored to compute the joint bivariate and trivariate probability densities, including the optimal multivariate modeling using Gaussian copula. We developed an interactive webapp to visually illustrate the parallels between the mathematical formulation, computational implementation, and graphical depiction of multivariate probability density and cumulative distribution functions. To ensure the interface and functionality are hardware platform independent, scalable, and functional, the app and its component widgets are implemented using HTML5 and JavaScript.

Results

We validated the webapp by testing the multivariate copula models under different experimental conditions and inspecting the performance in terms of accuracy and reliability of the estimated multivariate probability densities and distribution function values.

Conclusion

This article demonstrates the construction, implementation, and utilization of multivariate probability calculators. The proposed webapp implementation is freely available online (https://socr.umich.edu/HTML5/BivariateNormal/BVN2/) and can be used to assist with education and research of a diverse array of data scientists, STEM instructors, and AI learners.

High polygenic predisposition for ADHD and a greater risk of all-cause mortality: a large population-based longitudinal study

BACKGROUND

Attention deficit hyperactivity disorder (ADHD) is a highly heritable, neurodevelopmental disorder known to associate with more than double the risk of death compared with people without ADHD. Because most research on ADHD has focused on children and adolescents, among whom death rates are relatively low, the impact of a high polygenic predisposition to ADHD on accelerating mortality risk in older adults is unknown. Thus, the aim of the study was to investigate if a high polygenetic predisposition to ADHD exacerbates the risk of all-cause mortality in older adults from the general population in the UK.

METHODS

Utilising data from the English Longitudinal Study of Ageing, which is an ongoing multidisciplinary study of the English population aged ≥ 50 years, polygenetic scores for ADHD were calculated using summary statistics for (1) ADHD (PGS-ADHD_single) and (2) chronic obstructive pulmonary disease and younger age of giving first birth, which were shown to have a strong genetic correlation with ADHD using the multi-trait analysis of genome-wide association summary statistics; this polygenic score was referred to as PGS-ADHD_multi-trait. All-cause mortality was ascertained from the National Health Service central register that captures all deaths occurring in the UK.

RESULTS

The sample comprised 7133 participants with a mean age of 64.7 years (SD = 9.5, range = 50-101); of these, 1778 (24.9%) died during a period of 11.2 years. PGS-ADHD_single was associated with a greater risk of all-cause mortality (hazard ratio [HR] = 1.06, 95% CI = 1.02-1.12, p = 0.010); further analyses showed this relationship was significant in men (HR = 1.07, 95% CI = 1.00-1.14, p = 0.043). Risk of all-cause mortality increased by an approximate 11% for one standard deviation increase in PGS-ADHD_multi-trait (HR = 1.11, 95% CI = 1.06-1.16, p < 0.001). When the model was run separately for men and women, the association between PGS-ADHD_multi-trait and an increased risk of all-cause mortality was significant in men (HR = 1.10, 95% CI = 1.03-1.18, p = 0.003) and women (HR = 1.11, 95% CI = 1.04-1.19, p = 0.003).

CONCLUSIONS

A high polygenetic predisposition to ADHD is a risk factor for all-cause mortality in older adults. This risk is better captured when incorporating genetic information from correlated traits.

Interplay between polygenic propensity for ageing-related traits and the consumption of fruits and vegetables on future dementia diagnosis

BACKGROUND

Understanding how polygenic scores for ageing-related traits interact with diet in determining a future dementia including Alzheimer's diagnosis (AD) would increase our understanding of mechanisms underlying dementia onset.

METHODS

Using 6784 population representative adults aged ≥50 years from the English Longitudinal Study of Ageing, we employed accelerated failure time survival model to investigate interactions between polygenic scores for AD (AD-PGS), schizophrenia (SZ-PGS) and general cognition (GC-PGS) and the baseline daily fruit and vegetable intake in association with dementia diagnosis during a 10-year follow-up. The baseline sample was obtained from waves 3-4 (2006-2009); follow-up data came from wave 5 (2010-2011) to wave 8 (2016-2017).

RESULTS

Consuming < 5 portions of fruit and vegetables a day was associated with 33-37% greater risk for dementia in the following 10 years depending on an individual polygenic propensity. One standard deviation (1-SD) increase in AD-PGS was associated with 24% higher risk of dementia and 47% higher risk for AD diagnosis. 1-SD increase in SZ-PGS was associated with an increased risk of AD diagnosis by 66%(95%CI = 1.05-2.64) in participants who consumed < 5 portions of fruit or vegetables. There was a significant additive interaction between GC-PGS and < 5 portions of the baseline daily intake of fruit and vegetables in association with AD diagnosis during the 10-year follow-up (RERI = 0.70, 95%CI = 0.09-4.82; AP = 0.36, 95%CI = 0.17-0.66).

CONCLUSION

A diet rich in fruit and vegetables is an important factor influencing the subsequent risk of dementia in the 10 years follow-up, especially in the context of polygenetic predisposition to AD, schizophrenia, and general cognition.

Is the Way Forward to Step Back? Documenting the Frequency with which Study Goals are Misaligned with Study Methods and Interpretations in the Epidemiologic Literature

In any research study, there is an underlying research process that should begin with a clear articulation of the study's goal. The study's goal drives this process; it determines many study features including the estimand of interest, the analytic approaches that can be used to estimate it, and which coefficients, if any, should be interpreted. "Misalignment" can occur in this process when analytic approaches and/or interpretations do not match the study's goal; misalignment is potentially more likely to arise when study goals are ambiguously framed. This study documented misalignment in the observational epidemiologic literature and explored how the framing of study goals contributes to its occurrence. The following misalignments were examined: 1) use of an inappropriate variable selection approach for the goal (a "goal-methods" misalignment) and 2) interpretation of coefficients of variables for which causal considerations were not made (e.g., Table 2 Fallacy, a "goal-interpretation" misalignment). A random sample of 100 articles published 2014-2018 in the top 5 general epidemiology journals were reviewed. Most reviewed studies were causal, with either explicitly stated (13/103, 13%) or associationally-framed (71/103, 69%) aims. Full alignment of goal-methods-interpretations was infrequent (9/103, 9%), although clearly causal studies (5/13, 38%) were more often fully aligned than seemingly causal ones (3/71, 4%). Goal-methods misalignments were common (34/103, 33%), but most frequently, methods were insufficiently reported to draw conclusions (47/103, 46%). Goal-interpretations misalignments occurred in 31% (32/103) of studies and occurred less often when the methods were aligned (2/103, 2%) compared with when the methods were misaligned (13/103, 13%).

Use of directed acyclic graphs (DAGs) to identify confounders in applied health research: review and recommendations

Background

Directed acyclic graphs (DAGs) are an increasingly popular approach for identifying confounding variables that require conditioning when estimating causal effects. This review examined the use of DAGs in applied health research to inform recommendations for improving their transparency and utility in future research.

Methods

Original health research articles published during 1999–2017 mentioning ‘directed acyclic graphs’ (or similar) or citing DAGitty were identified from Scopus, Web of Science, Medline and Embase. Data were extracted on the reporting of: estimands, DAGs and adjustment sets, alongside the characteristics of each article’s largest DAG.

Results

A total of 234 articles were identified that reported using DAGs. A fifth (n = 48, 21%) reported their target estimand(s) and half (n = 115, 48%) reported the adjustment set(s) implied by their DAG(s). Two-thirds of the articles (n = 144, 62%) made at least one DAG available. DAGs varied in size but averaged 12 nodes [interquartile range (IQR): 9–16, range: 3–28] and 29 arcs (IQR: 19–42, range: 3–99). The median saturation (i.e. percentage of total possible arcs) was 46% (IQR: 31–67, range: 12–100). 37% (n = 53) of the DAGs included unobserved variables, 17% (n = 25) included ‘super-nodes’ (i.e. nodes containing more than one variable) and 34% (n = 49) were visually arranged so that the constituent arcs flowed in the same direction (e.g. top-to-bottom).

Conclusion

There is substantial variation in the use and reporting of DAGs in applied health research. Although this partly reflects their flexibility, it also highlights some potential areas for improvement. This review hence offers several recommendations to improve the reporting and use of DAGs in future research.

Latent class regression improves the predictive acuity and clinical utility of survival prognostication amongst chronic heart failure patients

The present study aimed to compare the predictive acuity of latent class regression (LCR) modelling with: standard generalised linear modelling (GLM); and GLMs that include the membership of subgroups/classes (identified through prior latent class analysis; LCA) as alternative or additional candidate predictors. Using real world demographic and clinical data from 1,802 heart failure patients enrolled in the UK-HEART2 cohort, the study found that univariable GLMs using LCA-generated subgroup/class membership as the sole candidate predictor of survival were inferior to standard multivariable GLMs using the same four covariates as those used in the LCA. The inclusion of the LCA subgroup/class membership together with these four covariates as candidate predictors in a multivariable GLM showed no improvement in predictive acuity. In contrast, LCR modelling resulted in a 18-22% improvement in predictive acuity and provided a range of alternative models from which it would be possible to balance predictive acuity against entropy to select models that were optimally suited to improve the efficient allocation of clinical resources to address the differential risk of the outcome (in this instance, survival). These findings provide proof-of-principle that LCR modelling can improve the predictive acuity of GLMs and enhance the clinical utility of their predictions. These improvements warrant further attention and exploration, including the use of alternative techniques (including machine learning algorithms) that are also capable of generating latent class structure while determining outcome predictions, particularly for use with large and routinely collected clinical datasets, and with binary, count and continuous variables.

A scoping review of causal methods enabling predictions under hypothetical interventions

BACKGROUND

The methods with which prediction models are usually developed mean that neither the parameters nor the predictions should be interpreted causally. For many applications, this is perfectly acceptable. However, when prediction models are used to support decision making, there is often a need for predicting outcomes under hypothetical interventions.

AIMS

We aimed to identify published methods for developing and validating prediction models that enable risk estimation of outcomes under hypothetical interventions, utilizing causal inference. We aimed to identify the main methodological approaches, their underlying assumptions, targeted estimands, and potential pitfalls and challenges with using the method. Finally, we aimed to highlight unresolved methodological challenges.

METHODS

We systematically reviewed literature published by December 2019, considering papers in the health domain that used causal considerations to enable prediction models to be used for predictions under hypothetical interventions. We included both methodologies proposed in statistical/machine learning literature and methodologies used in applied studies.

RESULTS

We identified 4919 papers through database searches and a further 115 papers through manual searches. Of these, 87 papers were retained for full-text screening, of which 13 were selected for inclusion. We found papers from both the statistical and the machine learning literature. Most of the identified methods for causal inference from observational data were based on marginal structural models and g-estimation.

CONCLUSIONS

There exist two broad methodological approaches for allowing prediction under hypothetical intervention into clinical prediction models: (1) enriching prediction models derived from observational studies with estimated causal effects from clinical trials and meta-analyses and (2) estimating prediction models and causal effects directly from observational data. These methods require extending to dynamic treatment regimes, and consideration of multiple interventions to operationalise a clinical decision support system. Techniques for validating 'causal prediction models' are still in their infancy.

Time to reality check the promises of machine learning-powered precision medicine

Machine learning methods, combined with large electronic health databases, could enable a personalised approach to medicine through improved diagnosis and prediction of individual responses to therapies. If successful, this strategy would represent a revolution in clinical research and practice. However, although the vision of individually tailored medicine is alluring, there is a need to distinguish genuine potential from hype. We argue that the goal of personalised medical care faces serious challenges, many of which cannot be addressed through algorithmic complexity, and call for collaboration between traditional methodologists and experts in medical machine learning to avoid extensive research waste.

COVID-19 and the epistemology of epidemiological models at the dawn of AI

The models used to estimate disease transmission, susceptibility and severity determine what epidemiology can (and cannot tell) us about COVID-19. These include: 'model organisms' chosen for their phylogenetic/aetiological similarities; multivariable statistical models to estimate the strength/direction of (potentially causal) relationships between variables (through 'causal inference'), and the (past/future) value of unmeasured variables (through 'classification/prediction'); and a range of modelling techniques to predict beyond the available data (through 'extrapolation'), compare different hypothetical scenarios (through 'simulation'), and estimate key features of dynamic processes (through 'projection'). Each of these models: address different questions using different techniques; involve assumptions that require careful assessment; and are vulnerable to generic and specific biases that can undermine the validity and interpretation of their findings. It is therefore necessary that the models used: can actually address the questions posed; and have been competently applied. In this regard, it is important to stress that extrapolation, simulation and projection cannot offer accurate predictions of future events when the underlying mechanisms (and the contexts involved) are poorly understood and subject to change. Given the importance of understanding such mechanisms/contexts, and the limited opportunity for experimentation during outbreaks of novel diseases, the use of multivariable statistical models to estimate the strength/direction of potentially causal relationships between two variables (and the biases incurred through their misapplication/misinterpretation) warrant particular attention. Such models must be carefully designed to address: 'selection-collider bias', 'unadjusted confounding bias' and 'inferential mediator adjustment bias' - all of which can introduce effects capable of enhancing, masking or reversing the estimated (true) causal relationship between the two variables examined.1 Selection-collider bias occurs when these two variables independently cause a third (the 'collider'), and when this collider determines/reflects the basis for selection in the analysis. It is likely to affect all incompletely representative samples, although its effects will be most pronounced wherever selection is constrained (e.g. analyses focusing on infected/hospitalised individuals). Unadjusted confounding bias disrupts the estimated (true) causal relationship between two variables when: these share one (or more) common cause(s); and when the effects of these causes have not been adjusted for in the analyses (e.g. whenever confounders are unknown/unmeasured). Inferentially similar biases can occur when: one (or more) variable(s) (or 'mediators') fall on the causal path between the two variables examined (i.e. when such mediators are caused by one of the variables and are causes of the other); and when these mediators are adjusted for in the analysis. Such adjustment is commonplace when: mediators are mistaken for confounders; prediction models are mistakenly repurposed for causal inference; or mediator adjustment is used to estimate direct and indirect causal relationships (in a mistaken attempt at 'mediation analysis'). These three biases are central to ongoing and unresolved epistemological tensions within epidemiology. All have substantive implications for our understanding of COVID-19, and the future application of artificial intelligence to 'data-driven' modelling of similar phenomena. Nonetheless, competently applied and carefully interpreted, multivariable statistical models may yet provide sufficient insight into mechanisms and contexts to permit more accurate projections of future disease outbreaks.

Using Bayesian networks to clarify interpretation of exposure-response regression coefficients: blood lead-mortality association as an example

We examine how Bayesian network (BN) learning and analysis methods can help to meet several methodological challenges that arise in interpreting significant regression coefficients in exposure-response regression modeling. As a motivating example, we consider the challenge of interpreting positive regression coefficients for blood lead level (BLL) as a predictor of mortality risk for nonsmoking men. We first note that practices such as dichotomizing or categorizing continuous confounders (e.g. income), omitting potentially important socioeconomic confounders (e.g. education), and assuming specific parametric regression model forms leave unclear to what extent a positive regression coefficient reflects these modeling choices, rather than a direct dependence of mortality risk on exposure. Therefore, significant exposure-response coefficients in parametric regression models do not necessarily reveal the extent to which reducing exposure-related variables (e.g. BLL) alone, while leaving fixed other correlates of exposure and mortality risks (e.g. education, income, etc.) would reduce adverse outcome risks (e.g. mortality risks). We then consider how BN structure-learning and inference algorithms and nonparametric estimation methods (partial dependence plots) can be used to clarify dependencies between variables, variable selection, confounding, and quantification of joint effects of multiple factors on risk, including possible high-order interactions and nonlinearities. We conclude that these details must be carefully modeled to determine whether a data set provides evidence that exposure itself directly affects risks; and that BN and nonparametric effect estimation and uncertainty quantification methods can complement regression modeling and help to improve the scientific basis for risk management decisions and policy-making by addressing these issues.