Bias in odds ratios from logistic regression methods with sparse data sets

The factors associated with the provision of public access defibrillation in Japan - A nationwide cohort study

AIM

The association between out-of-hospital cardiac arrest (OHCA) and the appropriate provision of public access defibrillation (PAD) remains unclear. This study aimed to evaluate the factors associated with whether or not PAD was provided.

METHODS

This retrospective cohort study utilized the All-Japan Utstein and Emergency Transport Registries in 2021. We included OHCA patients who were applied to automated external defibrillators (AEDs) by bystanders and were deemed eligible for defibrillation by an AED. We defined PAD provided or no PAD provided based on bystander defibrillation. Multivariable logistic regression analysis with the Firth bias adjustment method was employed to estimate the adjusted odds ratios (AORs) and 95% confidence intervals (CIs) for the exploratory evaluation of factors associated with PAD provided.

RESULTS

1949 patients were eligible for analysis (PAD provided, n = 1696 [87.0%]; no PAD provided, n = 253 [13.0%]). Factors positively associated with PAD provided were male (AOR [95% CI], 1.61 [1.17-2.21]; vs. female), other public place incidence (AOR [95% CI], 10.65 [1.40-1367.54]; vs. public place), non-family member witnessed (AOR [95% CI], 2.51 [1.86-3.42]; vs. unwitnessed) and conventional cardiopulmonary resuscitation (CPR), (AOR [95% CI], 1.75 [1.17-2.67]; vs. hands-only CPR). Conversely, factors negatively associated with no PAD provided were over 65 years old (AOR [95% CI], 0.48 [0.28-0.80]; vs. 19-64 yr), night-time onset (AOR [95% CI], 0.61 [0.45-0.83]; vs. daytime), non-cardiogenic (AOR [95% CI], 0.43 [0.31-0.61]; vs. cardiogenic), home setting (AOR [95% CI], 0.33 [0.14-0.83]; vs. public place), healthcare facility setting (AOR [95% CI], 0.40 [0.23-0.66]; vs. public place), no bystander CPR (AOR [95% CI], 0.31 [0.14-0.71]; vs. hands-only CPR), and dispatcher-assistance (AOR [95% CI], 0.72 [0.53-0.97]; vs. no dispatcher-assistance).

CONCLUSION

Male patients, other public place onset, witnessed by non-family and conventional CPR were associated with PAD provide. Therefore, training skilled first responders to use AEDs appropriately is necessary.

Uncertain about uncertainty in matching-adjusted indirect comparisons? A simulation study to compare methods for variance estimation

In health technology assessment, matching-adjusted indirect comparison (MAIC) is the most common method for pairwise comparisons that control for imbalances in baseline characteristics across trials. One of the primary challenges in MAIC is the need to properly account for the additional uncertainty introduced by the matching process. Limited evidence and guidance are available on variance estimation in MAICs. Therefore, we conducted a comprehensive Monte Carlo simulation study to evaluate the performance of different statistical methods across 108 scenarios. Four general approaches for variance estimation were compared in both anchored and unanchored MAICs of binary and time-to-event outcomes: (1) conventional estimators (CE) using raw weights; (2) CE using weights rescaled to the effective sample size (ESS); (3) robust sandwich estimators; and (4) bootstrapping. Several variants of sandwich estimators and bootstrap methods were tested. Performance was quantified on the basis of empirical coverage probabilities for 95% confidence intervals and variability ratios. Variability was underestimated by CE + raw weights when population overlap was poor or moderate. Despite several theoretical limitations, CE + ESS weights accurately estimated uncertainty across most scenarios. Original implementations of sandwich estimators had a downward bias in MAICs with a small ESS, and finite sample adjustments led to marked improvements. Bootstrapping was unstable if population overlap was poor and the sample size was limited. All methods produced valid coverage probabilities and standard errors in cases of strong population overlap. Our findings indicate that the sample size, population overlap, and outcome type are important considerations for variance estimation in MAICs.

Population-Based Influenza Vaccine Effectiveness Against Laboratory-Confirmed Influenza Infection in Southern China, 2023-2024 Season

Background

In China, the 2022-2023 influenza season began earlier and was characterized by higher levels of influenza activity and co-circulation of various respiratory pathogens compared with seasons before the coronavirus disease 2019 (COVID-19) pandemic. Timely and precise estimates of influenza vaccine effectiveness (IVE) against infections can be used to guide public health measures.

Methods

A test-negative study was conducted to estimate IVE against laboratory-confirmed influenza using data from the CHinese Electronic health Records Research in Yinzhou (CHERRY) study that prospectively integrated laboratory, vaccination, and health administrative data in Yinzhou, southern China. We included patients who presented influenza-like illness and received nucleic acid tests and/or antigen tests between October 2023 and March 2024. Estimates of IVE were adjusted for age, gender, month of specimen submitted, chronic comorbidities, and hospitalization status.

Results

A total of 205 028 participants, including 96 298 influenza cases (7.6% vaccinated) and 108 730 influenza-negative controls (13.4% vaccinated), were eligible for this analysis. The estimates of IVE were 49.4% (95% CI, 47.8%-50.9%), 41.9% (95% CI, 39.8%-44.0%), and 59.9% (95% CI, 57.9%-61.9%) against overall influenza, influenza A, and influenza B, respectively. A lower IVE was observed for individuals aged 7-17 years (38.6%), vs 45.8% for 6 months-6 years, 46.7% for 18-64 years, and 46.1% for ≥65 years. Vaccination reduced the risk of infection by 44.4% among patients with chronic comorbidities. IVEs varied by epidemic weeks with the changes in influenza activity levels and the switch of dominant influenza strains.

Conclusions

Influenza vaccination in the 2023-2024 season was protective against infection for the entire population.

A Simple MRI Score Predicts Pathological General Movements in Very Preterm Infants with Brain Injury-Retrospective Cohort Study

BACKGROUND/OBJECTIVES

Very preterm infants are at increased risk of brain injury and impaired brain development. The Total Abnormality Score and biometric parameters, such as biparietal width, interhemispheric distance and transcerebellar diameter, are simple measures to evaluate brain injury, development and growth using cerebral magnetic resonance imaging data at term-equivalent age. The aim of this study was to evaluate the association between the Total Abnormality Score and biometric parameters with general movements in very preterm infants with brain injury.

METHODS

This single-center retrospective cohort study included 70 very preterm infants (≤32 weeks' gestation and/or <1500 g birth weight) born between January 2017 and June 2021 in a level-three neonatal intensive care unit with brain injury-identified using cerebral magnetic resonance imaging data at term-equivalent age. General movements analysis was carried out at corrected age of 8-16 weeks. Binary logistic regression and Spearman correlation were used to examine the associations between the Total Abnormality Score and biometric parameters with general movements.

RESULTS

There was a significant association between the Total Abnormality Score and the absence of fidgety movements [OR: 1.19, 95% CI = 1.38-1.03] as well as a significant association between the transcerebellar diameter and fidgety movements (Spearman ρ = -0.269, p < 0.05).

CONCLUSIONS

Among very preterm infants with brain injury, the Total Abnormality Score can be used to predict the absence of fidgety movements and may be an easily accessible tool for identifying high-risk very preterm infants and planning early interventions accordingly.

Caveats of Covariate Adjustment in Disproportionality Analysis for Best Practices

Spontaneous reporting systems (SRS) provide valuable data for detecting unidentified adverse events not observed in clinical trials and for conducting safety assessments that accurately reflect real-world clinical practice. With the increasing number of publications using the SRS for disproportionality analysis (DA), there is an increasing demand for a comprehensive understanding of the research limitations associated with the SRS. However, there is a lack of understanding of the caveats associated with adjusting covariates in DA of the SRS. Herein, we summarized the use of covariate adjustment and its caveats in DA. The Council for International Organizations of Medical Sciences VIII suggests considering adjustments such as stratification when they can enhance the sensitivity and/or specificity of statistical analysis. However, several database-specific and statistical caveats have been identified when adjusting for covariates derived from the SRS. Disproportionality analysis may be affected not only by reporting bias at the time of enrollment but also by sparse-data bias due to variations in the number of enrollment reports. Statistical evidence is needed to determine in which cases and to what extent sensitivity and/or specificity are affected. Nevertheless, it is important for researchers to acknowledge that certain limitations discussed in this context may be inherent and cannot be rectified. Based on this understanding, they can then make an informed decision on whether to perform a covariate adjustment.

Epidemiological features of suicidal ideation among the elderly in China based meta-analysis

BACKGROUND

Studies on the prevalence of suicidal ideation (SI) and its associated factors among the elderly in China show considerable variability. This meta-analysis aims to clarify the epidemiological features of SI in this population.

METHODS

We systematically searched English and Chinese databases for relevant literature up to September 15, 2022. The extracted data facilitated the calculation of prevalence and odds ratios (ORs) for factors associated with SI among China's elderly.

RESULTS

We analyzed 31 cross-sectional studies, comprising a total of 79,861 participants from over 20 provinces and municipalities. The pooled prevalence of SI was found to be 11.47% [95% confidence interval (CI): 7.82-15.71%]. Significant variations in prevalence were influenced by residence, physical health (including chronic diseases and daily living capabilities), mental health (depressive symptoms and life satisfaction), economic status, and time-specific assessment tools. Notably, the prevalence from 2011-2020 (15.59%, 95% CI: 9.08-23.44%) was almost double that of 2001-2010 (7.85%, 95% CI: 5.08-11.16%). The SI prevalence in the eastern region (8.06%, 95% CI 5.59-10.94%) was significantly lower than in the central and western regions (16.97%, 95% CI 12.04-22.53%). Fourteen factors exhibited a significant pooled OR greater than 1 (p < 0.05), and two factors had ORs less than 1 (p < 0.05), indicating notable association with SI among the elderly.

CONCLUSION

SI among China's elderly showed relatively high prevalence and considerable heterogeneity across different characteristics and associated factors. This underscores the need for targeted intervention strategies and standardized temporal assessments of SI to effectively address suicide risk in this population.

Advancing unanchored simulated treatment comparisons: A novel implementation and simulation study

Population-adjusted indirect comparisons, developed in the 2010s, enable comparisons between two treatments in different studies by balancing patient characteristics in the case where individual patient-level data (IPD) are available for only one study. Health technology assessment (HTA) bodies increasingly rely on these methods to inform funding decisions, typically using unanchored indirect comparisons (i.e., without a common comparator), due to the need to evaluate comparative efficacy and safety for single-arm trials. Unanchored matching-adjusted indirect comparison (MAIC) and unanchored simulated treatment comparison (STC) are currently the only two approaches available for population-adjusted indirect comparisons based on single-arm trials. However, there is a notable underutilisation of unanchored STC in HTA, largely due to a lack of understanding of its implementation. We therefore develop a novel way to implement unanchored STC by incorporating standardisation/marginalisation and the NORmal To Anything (NORTA) algorithm for sampling covariates. This methodology aims to derive a suitable marginal treatment effect without aggregation bias for HTA evaluations. We use a non-parametric bootstrap and propose separately calculating the standard error for the IPD study and the comparator study to ensure the appropriate quantification of the uncertainty associated with the estimated treatment effect. The performance of our proposed unanchored STC approach is evaluated through a comprehensive simulation study focused on binary outcomes. Our findings demonstrate that the proposed approach is asymptotically unbiased. We argue that unanchored STC should be considered when conducting unanchored indirect comparisons with single-arm studies, presenting a robust approach for HTA decision-making.

Delays in Accessing Healthcare Across the Gender Spectrum in the All of Us Research Program

BACKGROUND

Gender minorities and cisgender women face barriers to healthcare access. Prior work suggests cost may represent a particular barrier to accessing care for transgender and gender diverse (TGD) individuals.

OBJECTIVE

To examine odds of delaying care for any reason and, secondarily, for 7 specific reasons among TGD individuals and cisgender women compared with cisgender men in the All of Us Research Program.

DESIGN

We calculated the odds of delayed care by gender identity relative to cisgender men using multivariable-adjusted logistic regression, with adjustment for age, race, income, education, and Charlson comorbidity index.

PARTICIPANTS

We examined 117,806 All of Us participants who completed the healthcare access and utilization survey.

MAIN MEASURES

The primary outcome was self-reported delayed care in the past 12 months for any of 7 potential reasons: cost (out-of-pocket cost, co-payment costs, and/or high deductible), lack of childcare, lack of eldercare, nervousness associated with visiting the healthcare provider, rurality, inability to take time off work, and lack of transportation.

KEY RESULTS

Compared with cisgender men, the multivariable-adjusted odds ratio (OR) for delaying care for any reason was 1.48 (95% CI, 1.44-1.53; P < 0.001) among cisgender women, 1.65 (95% CI, 1.24-2.21; P < 0.001) among TGD individuals assigned male at birth, and 2.76 (95% CI, 2.26-3.39; P < 0.001) among TGD individuals assigned female at birth. Results were consistent across multiple sensitivity analyses. TGD individuals were substantially more likely to cite nervousness with visiting a healthcare provider as a barrier, whereas cisgender women were more likely to delay care due to lack of childcare coverage.

CONCLUSIONS

Cisgender women and TGD individuals were more likely to delay seeking heath care compared with cisgender men, and for partially different reasons. These findings highlight the need to address common and distinct barriers to care access among marginalized groups.

Measuring susceptibility to use tobacco in an increasingly complex consumer marketplace: How many questions do we really need?

OBJECTIVE

Predicting which young people are likely to use tobacco in the future is critical for prevention and intervention. Although measures for assessing susceptibility to using tobacco have fulfilled this goal for decades, there is almost no standard for the number of items that should be administered, or which items should be administered for which products. This study explored whether brief but psychometrically sound versions of commonly used susceptibility measures can adequately capture the construct relative to longer measures.

METHOD

A sample of young people (N = 451; Mage = 16.5 years; 64% females; 65% White) completed 33 susceptibility items, which are designed to assess susceptibility to use different types of tobacco products (cigarette, smokeless tobacco, vaping products, and little cigars/cigarillos) of various flavors (tobacco, menthol, and sweet).

RESULTS

Analysis of these 33 items indicated that asking about the likelihood of using each tobacco product class when a best friend offers it (four items in all) captures 98.5% of information that is captured using the longer set of items; asking the best friend question for each product by each flavor category (11 items in all) captures 99.7% of the information.

CONCLUSIONS

Depending on research needs, tobacco use susceptibility can be measured with little loss of information by administering a limited set of items assessing the likelihood that a young person will use a tobacco product if a friend offers it for any product-flavor combination. (PsycInfo Database Record (c) 2024 APA, all rights reserved).

Maternal Prenatal Infections and Biliary Atresia in Offspring

Importance

Investigations into the association of antepartum maternal infections with the pathogenesis of biliary atresia (BA) in human offspring are insufficient.

Objective

To examine the association between prenatal infections in mothers and the development of BA in their offspring.

Design, Setting, and Participants

This population-based case-control study obtained administrative data from the Taiwan National Health Insurance Research Database with linkage to the Taiwan Maternal and Child Health Database, capturing demographic and medical information on nearly all 23 million of the Taiwan population. The cohort comprised 2 905 978 singleton live births among mother-infant dyads between January 1, 2004, and December 31, 2020, in Taiwan. The case group of infants with BA was identified from use of International Classification of Diseases diagnostic codes for BA and subsequent Kasai procedure or liver transplant. The control group was randomly selected from infants without BA, representing approximately 1 in 1000 study population. Data analyses were performed from May 1 to October 31, 2023.

Exposure

Prenatal maternal infections, including intestinal infection, influenza, upper airway infection, pneumonia, soft-tissue infection, and genitourinary tract infection.

Main Outcomes and Measures

The main outcome was exposure to prenatal maternal infections. Inverse probability weighting analysis was performed by building a logistic regression model to estimate the probability of the exposure observed for a particular infant and using the estimated probability as a weight in subsequent analyses. The weighted odds ratio (OR) estimated by logistic regressions was then used to assess the risk of BA in offspring after prenatal maternal infections.

Results

Among the mother-infant dyads included, 447 infants with BA were cases (232 females [51.9%]) and 2912 infants without BA were controls (1514 males [52.0%]). The mean (SD) maternal age at childbirth was 30.7 (4.9) years. Offspring exposed to prenatal intestinal infection (weighted OR, 1.46; 95% CI, 1.17-1.82) and genitourinary tract infection (weighted OR, 1.22; 95% CI, 1.05-1.41) in mothers exhibited a significantly higher risk of BA. Furthermore, maternal intestinal infection (weighted OR, 6.05; 95% CI, 3.80-9.63) and genitourinary tract infection (weighted OR, 1.55; 95% CI, 1.13-2.11) that occurred during the third trimester were associated with an increased risk of BA in offspring.

Conclusions and Relevance

Results of this case-control study indicate an association between prenatal intestinal infection and genitourinary tract infection in mothers and BA occurrence in their offspring. Further studies are warranted to explore the underlying mechanisms of this association.

A nomogram for predicting the recurrence of small bowel obstruction after gastrectomy in patients with gastric cancer

BACKGROUND

This study aimed to create a nomogram for predicting the recurrence of small bowel obstruction (SBO) after gastrectomy in patients with gastric cancer (GC) in order to provide better guidance for its diagnosis and treatment.

METHODS

A total of 173 patients undergone gastrectomy and developed SBO from January 2015 to October 2022 were admitted into this case-control study. The risk factors of postoperative recurrent SBO were analyzed by univariate and multivariate regression, and a nomogram for predicting the recurrent SBO after gastrectomy was developed using R Studio.

RESULTS

Thirty-nine cases of postoperative recurrent SBO occurred among the 173 GC patients who underwent radical gastrectomy, and the percentage of recurrent SBO was 22.54% (39/173). Age [odds ratio (OR) = 0.938, p = 0.026], WBC count (OR = 1.547, p < 0.001), tumor size (OR = 1.383, p = 0.024), postoperative metastasis (OR = 11.792, p = 0.030), and the interval from gastrectomy to first SBO (OR = 1.057, p < 0.001) were all identified as independent risk factors for postoperative recurrent SBO by logistic regression analysis. The receiver operating characteristic curve, the calibration curve, the model consistency index, and the decision curve analysis showed that the nomogram had good predictive performance.

CONCLUSION

Based on these factors, we created a nomogram to predict the occurrence of postoperative recurrent SBO. This novel nomogram could serve as a crucial early warning indicator that would guide doctors to make informed decisions while managing patients with gastric cancer.

A comparison of bias-adjusted generalized estimating equations for sparse binary data in small-sample longitudinal studies

Using a generalized estimating equation (GEE) can lead to a bias in regression coefficients for a small sample or sparse data. The bias-corrected GEE (BCGEE) and penalized GEE (PGEE) were proposed to resolve the small-sample bias. Moreover, the standard sandwich covariance estimator leads to a bias of standard error for small samples; several modified covariance estimators have been proposed to address this issue. We review the modified GEEs and modified covariance estimators, and evaluate their performance in sparse binary data from small-sample longitudinal studies. The simulation results showed that GEE and BCGEE often failed to achieve convergence, whereas the convergence proportion for PGEE was quite high. The bias for the regression coefficients was generally in the ascending order of PGEE < $$ < $$ BCGEE < $$ < $$ GEE. However, PGEE and BCGEE did not sufficiently remove the bias involving 20-30 subjects with unequal exposure levels with a 5% response rate. The coverage probability (CP) of the confidence interval for BCGEE was relatively poor compared with GEE and PGEE. The CP with the sandwich covariance estimator deteriorated regardless of the GEE methods under the small sample size and low response rate, whereas the CP with the modified covariance estimators-such as Morel's method-was relatively acceptable. PGEE will be the reasonable way for analyzing sparse binary data in small-sample studies. Instead of using the standard sandwich covariance estimator, one should always apply the modified covariance estimators for analyzing these data.