Built-in Bias Research Articles

Non-collapsibility and built-in selection bias of period-specific and conventional hazard ratio in randomized controlled trials

BackgroundThe hazard ratio of the Cox proportional hazards model is widely used in randomized controlled trials to assess treatment effects. However, two properties of the hazard ratio including the non-collapsibility and built-in selection bias need to be further investigated.MethodsWe conduct simulations to differentiate the non-collapsibility effect and built-in selection bias from the difference between the marginal and the conditional hazard ratio. Meanwhile, we explore the performance of the Cox model with inverse probability of treatment weighting for covariate adjustment when estimating the marginal hazard ratio. The built-in selection bias is further assessed in the period-specific hazard ratio.ResultsThe conditional hazard ratio is a biased estimate of the marginal effect due to the non-collapsibility property. In contrast, the hazard ratio estimated from the inverse probability of treatment weighting Cox model provides an unbiased estimate of the true marginal hazard ratio. The built-in selection bias only manifests in the period-specific hazard ratios even when the proportional hazards assumption is satisfied. The Cox model with inverse probability of treatment weighting can be used to account for confounding bias and provide an unbiased effect under the randomized controlled trials setting when the parameter of interest is the marginal effect.ConclusionsWe propose that the period-specific hazard ratios should always be avoided due to the profound effects of built-in selection bias.

Foundational model aided automatic high-throughput drug screening using self-controlled cohort study.

Developing medicine from scratch to governmental authorization and detecting adverse drug reactions (ADR) have barely been economical, expeditious, and risk-averse investments. The availability of large-scale observational healthcare databases and the popularity of large language models offer an unparalleled opportunity to enable automatic high-throughput drug screening for both repurposing and pharmacovigilance. To demonstrate a general workflow for automatic high-throughput drug screening with the following advantages: (i) the association of various exposure on diseases can be estimated; (ii) both repurposing and pharmacovigilance are integrated; (iii) accurate exposure length for each prescription is parsed from clinical texts; (iv) intrinsic relationship between drugs and diseases are removed jointly by bioinformatic mapping and large language model - ChatGPT; (v) causal-wise interpretations for incidence rate contrasts are provided. Using a self-controlled cohort study design where subjects serve as their own control group, we tested the intention-to-treat association between medications on the incidence of diseases. Exposure length for each prescription is determined by parsing common dosages in English free text into a structured format. Exposure period starts from initial prescription to treatment discontinuation. A same exposure length preceding initial treatment is the control period. Clinical outcomes and categories are identified using existing phenotyping algorithms. Incident rate ratios (IRR) are tested using uniformly most powerful (UMP) unbiased tests. We assessed 3,444 medications on 276 diseases on 6,613,198 patients from the Clinical Practice Research Datalink (CPRD), an UK primary care electronic health records (EHR) spanning from 1987 to 2018. Due to the built-in selection bias of self-controlled cohort studies, ingredients-disease pairs confounded by deterministic medical relationships are removed by existing map from RxNorm and nonexistent maps by calling ChatGPT. A total of 16,901 drug-disease pairs reveals significant risk reduction, which can be considered as candidates for repurposing, while a total of 11,089 pairs showed significant risk increase, where drug safety might be of a concern instead. This work developed a data-driven, nonparametric, hypothesis generating, and automatic high-throughput workflow, which reveals the potential of natural language processing in pharmacoepidemiology. We demonstrate the paradigm to a large observational health dataset to help discover potential novel therapies and adverse drug effects. The framework of this study can be extended to other observational medical databases.

Value-driven attitude surveys: Lessons from the refugee crisis in Greece.

Community reaction to refugees and asylum-seekers is often gauged by attitude surveys that are not designed to overcome built-in bias. Questionnaires that do not account for context and background consequently yield results that misrepresent community attitudes and offer predictably negative responses to immigrant groups. Such surveys can alter public perception, fuel anti-refugee sentiment, and affect policy simply because of how they are constructed. This model survey among humanitarian aid-workers from nine Greek non-governmental organizations uses specific techniques designed to overcome these challenges by applying sample familiarity, non-inflammatory hypothesis-testing, educational question stems, intentional ordering of questions, and direct questioning rather than surrogate measures like statistical approximation. Respondents working in the refugee crisis in Greece demonstrate how empathy, education, and exposure to refugees serve to overcome the harmful stereotypes of outsiders as contributors to crime, terror, and social burden.

The built-in selection bias of hazard ratios formalized using structural causal models

It is known that the hazard ratio lacks a useful causal interpretation. Even for data from a randomized controlled trial, the hazard ratio suffers from so-called built-in selection bias as, over time, the individuals at risk among the exposed and unexposed are no longer exchangeable. In this paper, we formalize how the expectation of the observed hazard ratio evolves and deviates from the causal effect of interest in the presence of heterogeneity of the hazard rate of unexposed individuals (frailty) and heterogeneity in effect (individual modification). For the case of effect heterogeneity, we define the causal hazard ratio. We show that the expected observed hazard ratio equals the ratio of expectations of the latent variables (frailty and modifier) conditionally on survival in the world with and without exposure, respectively. Examples with gamma, inverse Gaussian and compound Poisson distributed frailty and categorical (harming, beneficial or neutral) distributed effect modifiers are presented for illustration. This set of examples shows that an observed hazard ratio with a particular value can arise for all values of the causal hazard ratio. Therefore, the hazard ratio cannot be used as a measure of the causal effect without making untestable assumptions, stressing the importance of using more appropriate estimands, such as contrasts of the survival probabilities.

A Novel Spinel Ferrite-Hexagonal Ferrite Composite for Enhanced Magneto-Electric Coupling in a Bilayer with PZT.

The magnetoelectric effect (ME) is an important strain mediated-phenomenon in a ferromagnetic-piezoelectric composite for a variety of sensors and signal processing devices. A bias magnetic field, in general, is essential to realize a strong ME coupling in most composites. Magnetic phases with (i) high magnetostriction for strong piezomagnetic coupling and (ii) large anisotropy field that acts as a built-in bias field are preferred so that miniature, ME composite-based devices can operate without the need for an external magnetic field. We are able to realize such a magnetic phase with a composite of (i) barium hexaferrite (BaM) with high magnetocrystalline anisotropy field and (ii) nickel ferrite (NFO) with high magnetostriction. The BNx composites, with (100 - x) wt.% of BaM and x wt.% NFO, for x = 0-100, were prepared. X-ray diffraction analysis shows that the composites did not contain any impurity phases. Scanning electron microscopy images revealed that, with an increase in NFO content, hexagonal BaM grains become prominent, leading to a large anisotropy field. The room temperature saturation magnetization showed a general increase with increasing BaM content in the composites. NFO rich composites with x ≥ 60 were found to have a large magnetostriction value of around -23 ppm, comparable to pure NFO. The anisotropy field HA of the composites, determined from magnetization and ferromagnetic resonance (FMR) measurements, increased with increasing NFO content and reached a maximum of 7.77 kOe for x = 75. The BNx composite was cut into rectangular platelets and bonded with PZT to form the bilayers. ME voltage coefficient (MEVC) measurements at low frequencies and at mechanical resonance showed strong coupling at zero bias for samples with x ≥ 33. This large in-plane HA acted as a built-in field for strong ME effects under zero external bias in the bilayers. The highest zero-bias MEVC of ~22 mV/cm Oe was obtained for BN75-PZT bilayers wherein BN75 also has the highest HA. The Bilayer of BN95-PZT showed a maximum MEVC ~992 mV/cm Oe at electromechanical resonance at 59 kHz. The use of hexaferrite-spinel ferrite composite to achieve strong zero-bias ME coupling in bilayers with PZT is significant for applications related to energy harvesting, sensors, and high frequency devices.

Field-free magnetization switching through modulation of zero-field spin–orbit torque efficacy

To make spin–orbit torque magnetic random access memory (SOT-MRAM) practical, current-induced magnetization switching without an external bias field is essential. Given that the CoFeB/MgO structure has already been used in typical spin-transfer torque-MRAM for its high tunneling magnetoresistance, leveraging a similar material system to achieve field-free SOT switching is of great importance. In this work, we systematically investigate the field-free switching mechanism in CoFeB/W/CoFeB T-type structures, where the two CoFeB layers are in-plane and perpendicularly magnetized, respectively. Initial SOT characterization shows a sizable zero-field SOT efficacy (χHx=0) for such T-type devices. Furthermore, field-free angle-dependent SOT measurement confirms the parallel relationship between the built-in bias field and the magnetization of the in-plane magnetized CoFeB. Based on thorough verification and exclusion of other potential mechanisms, the Néel orange-peel effect emerges as the dominant origin for such a built-in bias field, where a positive correlation between the deposited film surface roughness and χHx=0 is found. Based on this discovery, the field-free switching efficacy in T-type structures is further optimized via film roughness tuning and examined with pillar-shaped devices. Our results provide insights into the tentative approach to improve field-free switching using T-type devices and the feasibility of downscaling.

How hazardous are hazard ratios? An empirical investigation of individual patient data from 27 large randomized clinical trials.

The use of hazard ratios as the standard treatment effect estimators for randomized trials with time-to-event outcomes has been the subject of repeated criticisms in recent years, e.g., for its non-collapsibility or with respect to (causal) interpretation. Another important issue is the built-in selection bias, which arises when the treatment is effective and when there are unobserved or not included prognostic factors that influence time-to-event. In these cases, the hazard ratio has even been termed "hazardous" because it is estimated from groups that increasingly differ in their (unobserved or omitted) baseline characteristics, yielding biased treatment estimates. We therefore adapt the Landmarking approach to assess the effect of ignoring a gradually increasing proportion of early events on the estimated hazard ratio. We propose an extension called "Dynamic Landmarking". This approach is based on successive deletion of observations, refitting Cox models and balance checking of omitted but observed prognostic factors, to obtain a visualization that can indicate built-in selection bias. In a small proof-of-concept simulation, we show that our approach is valid under the given assumptions. We further use "Dynamic Landmarking" to assess the suspected selection bias in the individual patient data sets of 27 large randomized clinical trials (RCTs). Surprisingly, we find no empirical evidence of selection bias in these RCTs and thus conclude that the supposed bias of the hazard ratio is of little practical relevance in most cases. This is mainly due to treatment effects in RCTs being small and the patient populations being homogeneous, e.g., due to inclusion and exclusion criteria.

Effect of a niobium-doped PZT interfacial layer thickness on the properties of epitaxial PMN-PT thin films

We are reporting on high quality epitaxial thin films of [Pb(Mg1/3Nb2/3)O3]0.67-(PbTiO3)0.33 [PMN-PT (67/33)]. These films were deposited on (001) oriented, vicinal SrTiO3 single crystal substrates, using 1 mol. % niobium-doped Pb(Zr0.52,Ti0.48)O3 (Nb-PZT) as an interfacial layer. The functional properties of the epitaxial PMN-PT (67/33) thin films were investigated as a function of the layer thickness of the Nb-PZT layer. The deposited hetero-structures are perovskite phase pure and fully (001)-oriented. The variation in Nb-PZT interfacial layer thickness results in an increasing trend change of the in-plane lattice parameter of that layer, which in turn causes a decrease in the c/a ratio of the PMN-PT film on top. The most noticeable effect related to this is a decrease in built-in-bias (imprint) voltage. Thus, the built-in bias can be tuned by changing the interfacial layer thickness. The ferroelectric capacitor properties are found to be most stable for the thinnest interfacial layers under a high number (108) of switching cycles.

Invited Perspective: The Potential of Potential Outcomes in Air Pollution Epidemiology.

Invited Perspective: The Potential of Potential Outcomes in Air Pollution Epidemiology.

Stability of ferroelectric bubble domains

Nanoscale ferroelectric topologies such as vortices, antivortices, bubble patterns, etc., are stabilized in thin films by a delicate balance of both mechanical and electrical boundary conditions. A systematic understanding of the phase stability of bubble domains, particularly when the above factors act simultaneously, remains elusive. Here we present first-principle-based simulations in combination with scanning probe microscopy of ultrathin epitaxial (001) $\mathrm{Pb}{\mathrm{Zr}}_{0.4}{\mathrm{Ti}}_{0.6}{\mathrm{O}}_{3}$ heterostructures to address this gap. The simulations predict that as-grown labyrinthine domains will transform to bubbles under combinations of reduced film thickness, increased mechanical pressure, and/or improved electrical screening. These topological transitions are explained by a common fundamental mechanism. Namely, we argue that, independently of the nature of the driving force, the evolution of the domain morphology allows the system to conserve its original residual depolarization field. Thereby, the latter remains pinned to a value determined by an external or built-in electric bias. To verify our predictions, we then exploit tomographic atomic force microscopy to achieve the concurrent effect of reducing film thickness and increased mechanical stimulus. The results provide a systematic understanding of phase stability and demonstrate controlled manipulation of nanoscale ferroelectric bubble domains.

Video nadzor i veštačka inteligencija - uticaj na pravo na privatnost i prava intelektualne svojine

The protection of individual privacy is increasingly questioned in the context of today's high-tech video surveillance by using superior technologies, such as Artificial Intelligence (smart cameras and video surveillance systems, biometrics and facial recognition) which is capable of analyzing a huge amount of data, identifying links between that data and de-anonymizing them. Technology is synonymous with evolution. Yet, the advantages of using new technology are combined with great risks. The use of video cameras for surveillance raises important privacy issues. Biometric remote identification can only be performed under certain safeguards, in the context of a justified interest and with respect for the principle of proportionality. In recent years, facial recognition technology has become increasingly widespread, and highly controversial, as it is omnipresent (at airport check-in lines, police departments, pharmacies, etc.). While it may add a sense of security and comfort for businesses implementing it, such technology has been widely criticized by privacy advocates, especially for its built-in racial bias and potential for abusive use. "Real-time" biometric identification of individuals in public-accessible spaces for law enforcement purposes is seen as highly intrusive on the rights and freedoms of the individuals concerned, but it also evokes a sense of constant surveillance and indirectly discourages the exercise of freedom of assembly and other fundamental rights. Another important issue is the use of video surveillance in a context where it can cause problems in terms of intellectual property. The unauthorized use of recordings may affect confidential business operations. Trade secrets and confidential information are often an essential part of a company's intellectual property portfolio. Thus, companies have to take extra steps to ensure that images captured by cameras are stored securely and accessible only to authorized personnel. Cameras can capture images of artistic works, performances, exhibitions, and uncontrolled access can lead to infringement of copyright and related rights. As captured videos can be uploaded to platforms, such use of intellectual property can lead to copyright infringement, especially when the video entails the use of copyrighted material without the owner's permission, which constitutes an infringement of intellectual property rights.

A sensitivity analysis approach for the causal hazard ratio in randomized and observational studies.

The hazard ratio (HR) is often reported as the main causal effect when studying survival data. Despite its popularity, the HR suffers from an unclear causal interpretation. As already pointed out in the literature, there is a built-in selection bias in the HR, because similarly to the truncation by death problem, the HR conditions on post-treatment survival. A recently proposed alternative, inspired by the Survivor Average Causal Effect, is the causal HR, defined as the ratio between hazards across treatment groups among the study participants that would have survived regardless of their treatment assignment. We discuss the challenge in identifying the causal HR and present a sensitivity analysis identification approach in randomized controlled trials utilizing a working frailty model. We further extend our framework to adjust for potential confounders using inverse probability of treatment weighting. We present a Cox-based and a flexible non-parametric kernel-based estimation under right censoring. We study the finite-sample properties of the proposed estimation methods through simulations. We illustrate the utility of our framework using two real-dataexamples.

Patterning of large area nanoscale domains in as-grown epitaxial ferroelectric PbTiO3 films

Effective tuning of nanoscale domain structures provides fundamental basis for controlling and engineering of various functionalities in ferroelectric materials. In this work, we demonstrate the precise patterning of nanoscopic domain structures in as-grown epitaxial PbTiO3 (PTO) films by merely introducing an ultrathin pre-patterned doping layer (e.g., Fe-doped PTO). The doping layer can effectively reverse the interfacial built-in bias, consequent to a reversed initial polarization reorientation in the as-grown film, which makes it possible to transfer the nano-patterns in the doping layer into the domain structure of ferroelectric films. For instance, we have successfully fabricated large area ordered array of nanoscale cylindrical domains (downward polarization) embedded in the matrix domain with opposite polarization (upward polarization) in PTO film. These nanoscale cylinder domains also allow deterministic and reversible erasure and creation induced by biased tip scanning. The results provide an effective pathway for on-demand patterning of large area nanoscale domains in the as-grown films, which may find applications in a wide range of nanoelectronic devices.

Implications of Mortality Displacement for Effect Modification and Selection Bias

Mortality displacement is the concept that deaths are moved forward in time (e.g., a few days, several months, and years) by exposure from when they would occur without the exposure, which is common in environmental time-series studies. Using concepts of a frail population and loss of life expectancy, it is understood that mortality displacement may decrease rate ratio (RR). Such decreases are thought to be minimal or substantial depending on study populations. Environmental epidemiologists have interpreted RR considering mortality displacement. This theoretical paper reveals that mortality displacement can be formulated as a built-in selection bias of RR in Cox models due to unmeasured risk factors independent from exposure of interest, and mortality displacement can also be viewed as an effect modifier by integrating the concepts of rate and loss of life expectancy. Thus, depending on the framework through which we view bias, mortality displacement can be categorized as selection bias in the bias taxonomy of epidemiology, and simultaneously mortality displacement can be seen as an effect modifier. This dichotomy provides useful implications regarding policy, effect modification, exposure time-windows selection, and generalizability, specifically why research in epidemiology may produce unexpected and heterogeneous RR over different studies and sub-populations. Keywords: Selection bias, Effect modification, Mediation, Directed Acyclic Graph, Frailty, Depletion of susceptible, Non-collapsibility, Generalizability, Hazard ratio (HR), Rate Ratio (RR),

A Relational Model for One-Shot Classification of Images and Pen Strokes

We show that a deep learning model with built-in relational inductive bias can bring benefits to sample-efficient learning, without relying on extensive data augmentation. Our study shows that excellent results can be achieved with a model in which the relational inductive bias is applied to images, while building an efficient one-shot classifier on top of raw strokes is more challenging. The proposed one-shot classification model performs relational matching of a pair of inputs in the form of local and pairwise attention. Our approach solves with almost perfect accuracy the one-shot image classification Omniglot challenge when combined with a Hungarian matching algorithm and attains competitive results on the same task on characters represented as rotation-augmented strokes.

Cohort, signaling, and early-career dynamics: The hidden significance of class in black-white earnings inequality

A central tenet in the now classic Wilson hypothesis surrounding racial earnings inequality emphasizes the elevated labor-market challenges for black workers of limited productive assets, yet the empirical evidence on this issue remains inconclusive. In this article, drawing on the Panel Study of Income Dynamics (PSID), I uncover three mechanisms that tend to underestimate the difficulty facing lower segments of the black labor force: (1) the built-in bias of cross-sectional data that conflate career stages, (2) the cohort bias that concentrates on labor-market dynamics of a conservative era, and (3) the interplay between discrimination and productivity signaling that delivers heterogenous outcomes among black job seekers. When these mechanisms are accounted for, a pattern that is consistent with the Wilson hypothesis emerges – well-equipped African Americans see narrowed gaps in early-career earnings with Whites. These findings reconcile conflicting evidence in existence and provide guidance for future work.

Community Newsletter: Autism training for Black parents, Google’s built-in gender bias

Social media in the autism research space this week talked about making training for Black parents of autistic children “culturally relevant” and how — right on time for #InternationalWomensDay — Google auto-demotes famous women scientists.

Mortality displacement as a type of selection bias: consequences and causal perspectives

BACKGROUND AND AIM: Mortality displacement (MD) is a phenomenon in which the observed mortality rate decreases temporarily after exposure, resulting in deaths being moved forward a time period from when they could occur without exposure. However, little is known about whether MD biases rate ratios (RR) in different study designs, the resulting type of bias, and what consequences MD has on observational epidemiology. METHODS: Here, we present theory to show that: 1) MD is a built-in selection bias under various study settings including survival analysis, time-series analysis, and case-crossover analysis; 2) MD distorts observation of effective exposure periods and effect modifications; and 3) how biased RR can be causally interpreted. To illustrate the consequences, we provide simulation analyses of open and closed cohorts and real-world data analyses of ambient particulate matter and mortality in South Korea and in the US. RESULTS:We found that: 1) MD is a bias toward the null for effective exposure period and RR; 2) in the distributed lag model framework as a generalized version of time-window methods, MD is a bias decreasing RR of distributed lags; 3) modification of RR may be severely biased due to differential MD across population strata; 4) RRs in population-based studies may be strongly biased by MD; 5) nevertheless, biased RR may be viewed as unbiased RR with a newly defined causal effect with advanced timing of outcomes by exposure; and 6) biased modification of RR may be seen as a new type of effect modification under this newly defined causal effect. CONCLUSIONS:MD is a selection bias inherently occurring in various study designs. Researchers should consider consequences of MD in analyzing and interpreting associations. Future studies should develop methods to adjust for MD in epidemiologic research; however, adjustment for MD may not be necessary depending on what types of causal effect researchers wish to target. KEYWORDS: Mortality displacement, selection bias, causal interpretations, effect modification, study designs

Void Galaxies Follow a Distinct Evolutionary Path in the Environmental COntext Catalog

Abstract We measure the environmental dependence, where environment is defined by the distance to the third nearest neighbor, of multiple galaxy properties inside the Environmental COntext (ECO) catalog. We focus primarily on void galaxies, which we define as the 10% of galaxies having the lowest local density. We compare the properties of void and non-void galaxies: baryonic mass, color, fractional stellar mass growth rate (FSMGR), morphology, and gas-to-stellar-mass ratio (estimated from a combination of H i data and photometric gas fractions calibrated with the REsolved Spectroscopy Of a Local VolumE survey). Our void galaxies typically have lower baryonic masses than galaxies in denser environments, and they display the properties expected of a lower mass population: they have more late types, are bluer, have a higher FSMGR, and are more gas-rich. We control for baryonic mass and investigate the extent to which void galaxies are different at fixed mass. Void galaxies are bluer, more gas-rich, and more star-forming at fixed mass than non-void galaxies, which is a possible signature of galaxy assembly bias. Furthermore, we show that these trends persist even at fixed mass and morphology, and we find that voids host a distinct population of early types that are bluer and more star-forming than the typical red and quenched early types. In addition to these empirical observational results, we also present theoretical results from mock catalogs with built-in galaxy assembly bias. We show that a simple matching of galaxy properties to (sub)halo properties, such as mass and age, can recover the observed environmental trends in ECO galaxies.

Mass ratio, the hills mechanism, and the Galactic Centre S-stars

ABSTRACT The Galactic Centre contains several young populations within its central parsec: a disc between ∼0.05 and 0.5 pc from the centre, and the isotropic S-star cluster extending an order of magnitude further inwards in radius. Recent observations (i.e. spectroscopy and hypervelocity stars) suggest that some S-stars originate in the disc. In particular, the S-stars may be remnants of tidally disrupted disc binaries. However, there is an apparent inconsistency in this scenario: the disc contains massive O and Wolf–Rayet stars while the S-stars are lower mass, B stars. We explore two different explanations for this apparent discrepancy: (i) a built-in bias in binary disruptions, where the primary star remains closer in energy to the centre-of-mass orbit than the secondary and (ii) selective tidal disruption of massive stars within the S-star cluster. The first explanation is plausible. On the other hand, tidal disruptions have not strongly affected the mass distribution of the S-stars over the last several Myr.