No full-text available
To read the full-text of this research,
you can request a copy directly from the author.
Bias Reduction of Maximum Likelihood Estimates
Abstract
It is shown how, in regular parametric problems, the first-order term is removed from the asymptotic bias of maximum likelihood estimates by a suitable modification of the score function. In exponential families with canonical parameterization the effect is to penalize the likelihood by the Jeffreys invariant prior. In binomial logistic models, Poisson log linear models and certain other generalized linear models, the Jeffreys prior penalty function can be imposed in standard regression software using a scheme of iterative adjustments to the data.
... , µ p ) ⊤ is well known to be problematic, in part because the estimate of µ i diverges if team i wins or loses all of its matches (e.g., Kosmidis and Firth, 2021). Bias-reduced maximum likelihood (Firth, 1993) overcomes the limitations of maximum likelihood, guaranteeing finiteness and also better frequentist properties of the strength estimates (Kosmidis and Firth, 2021). Finite estimates can also be obtained through other forms of penalized maximum likelihood (Mease, 2003) or Bayesian methods (Caron and Doucet, 2012). ...
... The simulations of Section 4 and the application to English Premier League show that the amount of shrinkage needed for good predictive performance when the training set and the number of teams are both small is appreciably greater than the shrinkage inherent in the bias-reduced maximum likelihood method of Firth (1993). This is unsurprising: it is well known that unbiasedness typically is not the main consideration for predictive performance. ...
... The boxplots show that the predictive performance of pairwise empirical Bayes is substantially better than that of maximum likelihood estimation, in particular when the training data are match results from only the first 10 or 15 weeks of the season. At those early stages of the tournament, just as we saw in the simulation studies of Section 4, maximum likelihood very often produces strength-parameter estimates that predict future matches worse than does the naive forecast.The results here also show that the bias-reduction method ofFirth (1993) improves upon maximum likelihood estimation in terms of prediction: in all 28 seasons and for all sizes of the training data, the logarithmic skill scores based on bias-reduced maximum likelihood are better than those of standard maximum likelihood. However, pairwise empirical Bayes typically performs even better than bias-reduced maximum likelihood. ...
Paired comparison models, such as Bradley-Terry and Thurstone-Mosteller, are commonly used to estimate relative strengths of pairwise compared items in tournament-style datasets. With predictive performance as primary criterion, we discuss estimation of paired comparison models with a ridge penalty. A new approach is derived which combines empirical Bayes and composite likelihoods without any need to re-fit the model, as a convenient alternative to cross-validation of the ridge tuning parameter. Simulation studies, together with application to 28 seasons of English Premier League football, demonstrate much better predictive accuracy of the new approach relative to ordinary maximum likelihood. While the application of a standard bias-reducing penalty was found to improve appreciably the performance of maximum likelihood, the ridge penalty with tuning as developed here yields greater accuracy still.
... Various methods of estimation have been proposed for contexts relevant to CJ. For example, one stream of literature considers bias reduction in relevant model families (Firth, 1993;Firth, 2009, 2011;Kosmidis, 2014;Kenne Pagui et al., 2017;Kosmidis et al., 2020;Kosmidis and Firth, 2021). Another, independent, stream of investigation considers estimation in the context of Rasch models, typically considering both bias and predictive accuracy (Molenaar, 1995;Linacre, 2004;Haberman, 2004;Robitzsch, 2021). ...
... Robitzsch (2021) found the ϵ-adjustment to be one of the best-performing of a wide variety of proposed estimation methods when considering the wider context of Rasch models, looking at bias and root mean squared error. None of the other estimation methods discussed here was considered in that study, though the method of Warm (1989), which was considered and rejected, is closely related to that of Firth (1993). Davidson and Solomon (1973) propose a Bayesian approach with a conjugate Beta(γ, γ) prior on each of the pairwise preference probabilities. ...
... These give the densities shown in Figure 1 For the dummy item method, c 0 will be taken to be 0.25. In testing based on a roundrobin tournament, where the expected penalty was calculated analytically, this value was found to provide a good approximation to the method of Firth (1993). For the ϵ-adjustment approach, ϵ will be taken to be 0.3. ...
Comparative Judgement is an assessment method where item ratings are estimated based on rankings of subsets of the items. These rankings are typically pairwise, with ratings taken to be the estimated parameters from fitting a Bradley-Terry model. Likelihood penalization is often employed. Adaptive scheduling of the comparisons can increase the efficiency of the assessment. We show that the most commonly used penalty is not the best-performing penalty under adaptive scheduling and can lead to substantial bias in parameter estimates. We demonstrate this using simulated and real data and provide a theoretical explanation for the relative performance of the penalties considered. Further, we propose a superior approach based on bootstrapping. It is shown to produce better parameter estimates for adaptive schedules and to be robust to variations in underlying strength distributions and initial penalization method.
... org/ econo mic-freed om/ datas et). Standard logit (Panel A) and Firth's (1993) logit (Panel B) estimation methods were used. The results in Panel B were produced using the -firthlogit-Stata command (Coveney 2008). ...
... All models were estimated using the standard maximum likelihood method (Table 5, Panel A) as well as the penalized maximum likelihood methodology (Table 5, Panel B) proposed by Firth (1993). Firth's (1993) method is appropriate for small samples and overcomes the 'separation' problem (Heinze and Schemper 2002). ...
... All models were estimated using the standard maximum likelihood method (Table 5, Panel A) as well as the penalized maximum likelihood methodology (Table 5, Panel B) proposed by Firth (1993). Firth's (1993) method is appropriate for small samples and overcomes the 'separation' problem (Heinze and Schemper 2002). ...
This paper aims to investigate the convergence pattern of sovereign bond yields in European Union (EU) countries during the Covid-19 pandemic. To this purpose we employ the Phillips and Sul convergence methodology. We further test whether the above pattern was affected by the intensity of government responses against the Covid-19 dispersion. Our findings indicate the existence of two convergence clubs. Both clubs are composed of countries that have previously been classified as either ‘core’ or ‘periphery’ countries, indicating that in this particular health crisis, this classification is too simplistic. Moreover, our results suggest that the intensity of Covid-19 containment policies had little to no effect in shaping the convergence pattern of sovereign bond yields in the EU. These conclusions are robust even after accounting for the size of the fiscal sector of each country.
... Additionally, we evaluated estimates of the treatment effects by biomarker level. Results of a standard Cox model were compared with a Firth-corrected Cox model, i.e., a Cox model with a modified score function 15 . Profile likelihood (PL) and Wald confidence intervals (CI) were also compared. ...
... Here, exp β TM low = HR TM low and exp β TM high = HR TM high were the HRs for experimental vs. standard treatment in subgroups of low and high marker levels, respectively. All datasets were also analyzed by a bias-eliminating approach originally developed by Firth 15 for generalized linear models and later implemented in Cox regression 16 . In contrast to the standard Cox model, a Firth-corrected Cox model provides finite HR estimates for monotone likelihoods, i.e., when the likelihood function does not have a unique maximum and the parameter estimate of a Cox model diverges with infinite standard error. ...
Predictive biomarkers are essential for personalized medicine since they select the best treatment for a specific patient. However, of all biomarkers that are evaluated, only few are eventually used in clinical practice. Many promising biomarkers may be erroneously abandoned because they are investigated in small studies using standard statistical techniques which can cause small sample bias or lack of power. The standard technique for failure time endpoints is Cox proportional hazards regression with a multiplicative interaction term between binary variables of biomarker and treatment. Properties of this model in small studies have not been evaluated so far, therefore we performed a simulation study to understand its small sample behavior. As a remedy, we applied a Firth correction to the score function of the Cox model and obtained confidence intervals (CI) using a profile likelihood (PL) approach. These methods are generally recommended for small studies of different design. Our results show that a Cox model estimates the biomarker-treatment interaction term and the treatment effect in one of the biomarker subgroups with bias, and overestimates their standard errors. Bias is however reduced and power is increased with Firth correction and PL CIs. Hence, the modified Cox model and PL CI should be used instead of a standard Cox model with Wald based CI in small studies of predictive biomarkers.
... To identify genes associated with familial PD, we performed gene-based analyses for both low-frequency (minor allele frequency (MAF) <0.05) and ultrarare (≤5 carriers) nonsynonymous variants using an omnibus test that combines burden, sequencing kernel association test (SKAT) and the aggregated Cauchy association variant test (ACAT-V) (Methods) [20][21][22] . No significant disease associations were observed within the ultrarare variant category (Extended Data Fig. 5a,b); however, analyses of low-frequency variants (MAF <0.05) revealed disease associations for LRRK2, GBA and RAB32 at exome-wide significance (P < 2.7 × 10 −6 ; Fig. 1a analysis indicated that carriers shared a ~300-kb haplotype surrounding the p.S71R variant (Extended Data Fig. 7). ...
... For low-frequency variants (MAF <0.05), we additionally performed SKAT and ACAT-V tests 22,52 . To account for the unbalanced case-control ratio, burden tests were performed using Firth's logistic regression 20 , the robust version of SKAT was employed 21 and we adapted ACAT-V so that Firth's logistic regression is used to perform the burden and single-variant tests that underlie ACAT-V 53 . Test statistics across the three statistical tests were combined using ACAT, which is designed to combine potentially correlated test statistics 22 . ...
Despite substantial progress, causal variants are identified only for a minority of familial Parkinson’s disease (PD) cases, leaving high-risk pathogenic variants unidentified1,2. To identify such variants, we uniformly processed exome sequencing data of 2,184 index familial PD cases and 69,775 controls. Exome-wide analyses converged on RAB32 as a novel PD gene identifying c.213C > G/p.S71R as a high-risk variant presenting in ~0.7% of familial PD cases while observed in only 0.004% of controls (odds ratio of 65.5). This variant was confirmed in all cases via Sanger sequencing and segregated with PD in three families. RAB32 encodes a small GTPase known to interact with LRRK2 (refs. 3,4). Functional analyses showed that RAB32 S71R increases LRRK2 kinase activity, as indicated by increased autophosphorylation of LRRK2 S1292. Here our results implicate mutant RAB32 in a key pathological mechanism in PD—LRRK2 kinase activity5–7—and thus provide novel insights into the mechanistic connections between RAB family biology, LRRK2 and PD risk.
... Here, I(β) is the Fisher information matrix for β. 39 The modified score equation will be ...
... In other words, this method produces finite estimates of the parameter using the average penalized ML estimate.42 It should be mentioned that Firth's method can be applied not only for logistic regression models but also for other generalized linear models, especially Cox regression.39,43 Heinze and colleagues have demonstrated that in the presence of sparse data, the penalized Firth's method performs better than the standard ML method in terms of mitigating the bias when estimating the OR.34,42,43 3 | RESULTS3.1 | Characteristics of study participantsThis study includes 2636 women who experience menopause naturally. ...
Background and Aims
Declines in estradiol levels after menopause have been reported to be associated with several health outcomes. This study aimed to determine the effect of age at natural menopause (ANM) on some of the most common chronic diseases.
Methods
This historical cohort study was performed on 2636 postmenopausal women aged 40–70 years participating in phase one of the PERSIAN cohort study in Kharameh, Iran, during 2015–2017. The effect of early (<45 years), intermediate (45–53 years), and late menopause (>53 years) on chronic diseases such as hypertension, diabetes, ischemic heart diseases, stroke, thyroid diseases, and depression was assessed using classic logistic regression for diseases with an incidence rate of more than 10% and Firth's logistic regression for diseases with an incidence of less than this amount.
Results
The mean age of women was 53.48 ± 8.59. Respectively, early and intermediate menopause was associated with ischemic heart disease (odds ratio [OR = 1.61, 95% confidence interval [CI]: 1.08–2.42; p = 0.020), (OR = 1.57, 95% CI: 1.13–2.21; p = 0.008) and thyroid diseases (OR = 3.10, 95% CI: 1.64–6.24; p < 0.001), (OR = 1.83, 95% CI: 1.02–3.57; p = 0.042). furthermore, early menopause was a risk factor for diabetes (OR = 1.46, 95% CI: 1.07–2.00; p = 0.018), depression (OR = 4.79, 95% CI: 2.20–11.79; p = <0.001) and stroke (OR = 3.00, 95% CI: 1.08–9.32; p = 0.034).
Conclusions
In this study, women with diabetes, ischemic heart diseases, stroke, thyroid disorders, and depression had a younger ANM compared to their healthy counterparts. Therefore, applying appropriate strategies to postpone the age of menopause, can reduce the incidence of these types of chronic diseases.
... For each simulation, we included three models for relative risk comparison, biascorrected (Firth [8]) Poisson, Logistic regression with exclusion criteria and Logistic regression without exclusion criteria. For the logistic regression models, the odds ratios were converted to relative risk with the conversion formula presented in Theoretical Background section. ...
... Another well-known issue for binary outcomes is the small cell count issue that leads to infinite estimate. The infinite estimates usually arise when one of the cells in a 2 × 2 table is 0. This can be corrected with implementing the Firth correction (Firth [8]). ...
The Phenome-wide association studies (PheWAS) have become widely used for efficient, high-throughput evaluation of relationship between a genetic factor and a large number of disease phenotypes, typically extracted from a DNA biobank linked with electronic medical records (EMR). Phecodes, billing code-derived disease case-control status, are usually used as outcome variables in PheWAS and logistic regression has been the standard choice of analysis method. Since the clinical diagnoses in EMR are often inaccurate with errors which can lead to biases in the odds ratio estimates, much effort has been put to accurately define the cases and controls to ensure an accurate analysis. Specifically in order to correctly classify controls in the population, an exclusion criteria list for each Phecode was manually compiled to obtain unbiased odds ratios. However, the accuracy of the list cannot be guaranteed without extensive data curation process. The costly curation process limits the efficiency of large-scale analyses that take full advantage of all structured phenotypic information available in EMR. Here, we proposed to estimate relative risks (RR) instead. We first demonstrated the desired nature of RR that overcomes the inaccuracy in the controls via theoretical formula. With simulation and real data application, we further confirmed that RR is unbiased without compiling exclusion criteria lists. With RR as estimates, we are able to efficiently extend PheWAS to a larger-scale, phenome construction agnostic analysis of phenotypes, using ICD 9/10 codes, which preserve much more disease-related clinical information than Phecodes.
... In addition, these authors have also verified that there may be substantial bias in the ML estimator of that persists even at large sample sizes. In order to overcome this shortcoming in the class of von Mises regression models, we shall derive and evaluate the bias reduction approach developed by Firth [21], which removes the first-order term of the asymptotic bias of the ML estimator by a suitable modification in the original score function. In short, the modified score function ensures an estimator with smaller asymptotic bias than the original ML estimator. ...
... Firth [21]'s method delivers estimators with ( −2 ) biases which may be obtained through the solution of the following modified score function: ...
... Since the binary dependent variable has a very skewed distributionchange of capital occurs in less than half of a percent of the country-year units in the samplewe employ a logit estimator for our benchmark. We also run robustness tests using a rare-events logit specification, specifically constructed to handle rare outcomes (King & Zeng, 2001) and penalized logit (Firth, 1993), in addition to probit and OLS. The sign and significance of the relationship is robust to choice of estimator (Appendix Tables A2-A3). ...
Why do some countries make the costly decision to relocate their capital city? Existing research offers four general explanations for this momentous action: administrative functionality, economic development, environmental degradation, and national integration. We offer a less sanguine, political explanation: capital relocations offer autocratic leaders a way to mitigate different security threats, including coups, popular protests, and foreign interventions. Using original data on capital relocations in 202 polities after 1789, we test several implications of our argument, at different levels of analysis. First, we show that autocracies are much more likely to relocate their capitals than democracies. Second, using different indicators of internal and external threats, we find that autocracies more likely relocate their capitals when breakdown is looming. Third, running subnational analyses, we find evidence that capitals are relocated to smaller cities and areas less susceptible to urban development.
... For binary outcomes, a log-linear model will be fitted to estimate risk ratios (RRs). If significant issues with model convergence are encountered, a logistic regression model will be fitted instead, using the Firth method to accommodate rare outcomes [31]. Time-to-event outcomes (e.g. ...
Background
Due to increasing life expectancy, almost half of people with type 2 diabetes are aged 65 years or over worldwide. When metformin alone does not control blood sugar, the choice of which second-line therapy to prescribe next is not clear from currently available evidence. The existence of frailty and comorbidities in older adults further increases the complexity of medical decision-making. As only a relatively small proportion of trials report results separately for older adults, the relative efficacy and safety of second-line therapies in older adults with type 2 diabetes mellitus are unknown and require further investigation. This individual participant data (IPD) network meta-analysis evaluates the relative efficacy and safety of second-line therapies on their own or in combination in older adults with type 2 diabetes mellitus.
Methods
All relevant published and unpublished trials will be identified. Studies published prior to 2015 will be identified from two previous comprehensive aggregate data network meta-analyses. Searches will be conducted in CENTRAL, MEDLINE, and EMBASE from 1st January 2015 onwards, and in clinicaltrials.gov from inception. Randomised controlled trials with at least 100 estimated older adults (≥ 65 years) receiving at least 24 weeks of intervention that assess the effects of glucose-lowering drugs on mortality, glycemia, vascular and other comorbidities outcomes, and quality of life will be eligible. The screening and data extraction process will be conducted independently by two researchers. The quality of studies will be assessed using the Cochrane risk of bias tool 2. Anonymised IPD of all eligible trials will be requested via clinical trial portals or by contacting the principal investigators or sponsors. Received data will be reanalysed where necessary to standardise outcome metrics. Network meta-analyses will be performed to determine the relative effectiveness of therapies.
Discussion
With the increasing number of older adults with type 2 diabetes worldwide, an IPD network meta-analysis using data from all eligible trials will provide new insights into the optimal choices of second-line antidiabetic drugs to improve patient management and reduce unnecessary adverse events and the subsequent risk of comorbidities in older adults.
Systematic review registration
PROSPERO CRD42021272686.
... Firth's logistic regression (penalized log-likelihood) was used to reduce small-sample bias (i.e. small arising from the small number of ART interruptions) 28 . To identify factors associated with ART interruption, a univariable analysis was performed. ...
With a national prevalence of 0.9%, Burundi is close to achieving UNAIDS’ 2025 targets. Despite this, different types of crises periodically disrupt its HIV health services. The community-based program EPIC measured the impact of the COVID-19 health crisis on people living with HIV (PLHIV) in Burundi in 2021. Specifically, it assessed ART interruption and associated factors since the beginning of the pandemic. The study questionnaire was administered to PLHIV in three cities between October and November 2021. Participants were recruited using convenience sampling. Logistic regression models helped identify factors associated with ART interruption. Of the 317 respondents, 37 (11.7%) reported interruption. The majority (79.2%) self-identified as belonging to key populations. Interruption was significantly associated with: fewer HIV medical follow-up visits (adjusted Odds Ratio, aOR = 7.80, p = 0.001) and forced HIV status disclosure (aOR = 4.10, p = 0.004). It was inversely associated with multi-month ART dispensing (aOR = 0.36, p = 0.017) since the beginning of the pandemic and the perception of not having been sufficiently informed by the HIV medical team about the risk of COVID-19 infection (aOR = 0.11, p < 0.001). Our results highlight the importance of multi-month ART dispensing, enhanced communication, and voluntary disclosure of one’s HIV status in preventing ART interruption in times of crises in Burundi.
... We used univariable logistic regression to examine associations between radiographic characteristics and need for THA. Because 'subluxation' is a known risk factor for early hip failure, we tested the effect of osteonecrosis adjusted for subluxation using Firth's penalised likelihood estimates [21]. We estimated the cumulative occurrence of THA over time using Kaplan-Meier survival analysis. ...
Background
Patients with osteonecrosis of the femoral head secondary to DDH frequently require total hip arthroplasty (THA), but it is not well understood which factors necessitate this requirement. We determined the incidence of THA in patients who have osteonecrosis secondary to DDH and factors associated with need for THA.
Methods
We included patients who received closed or open reductions between 1995 and 2005 with subsequent development of osteonecrosis. We determined osteonecrosis according to Bucholz and Ogden; osteoarthritis severity (Kellgren-Lawrence), subluxation (Shenton’s line); neck-shaft angle; and acetabular dysplasia (centre-edge and Sharp angles). We also recorded the number of operations of the hip in childhood and reviewed case notes of patients who received THA to describe clinical findings prior to THA. We assessed the association between radiographic variables and the need for THA using univariate logistic regression.
Results
Of 140 patients (169 hips), 22 patients received 24 THA (14%) at a mean age of 21.3 ± 3.7 years. Associated with the need for THA were grade III osteonecrosis (OR 4.25; 95% CI 1.70-10.77; p = 0.0019), grade IV osteoarthritis (21.8; 7.55–68.11; p < 0.0001) and subluxation (8.22; 2.91–29.53; p = 0.0003). All patients who required THA reported at least 2 of: severe pain including at night, stiffness, and reduced mobility. Acetabular dysplasia and number of previous operations were not associated with the need for THA.
Conclusions
We identified a 14% incidence of THA by age 34 years in patients with osteonecrosis secondary to DDH. Grade III osteonecrosis (global involvement femoral head and neck) was strongly associated with THA, emphasising the importance to avoid osteonecrosis when treating DDH.
... From the variables selected with LASSO in each NSES category, variables with a nonzero coefficient, considering a 95% confidence interval (CI), were kept; Firth penalised logistic regression was used to estimate the coefficients and CI (Firth, 1993). Finally, a logistic regression with maximum likelihood estimation (MLE) was performed to estimate OR of spontaneous PTB in each category, using the same variables in both NSES categories. ...
Background
Preterm birth (PTB) is the main condition related to perinatal morbimortality worldwide. The aim of this study was to determine the indirect effects of neighbourhood socioeconomic status (NSES) on the risk of spontaneous PTB.
Methods
We carried out a retrospective case-control study including sociodemographic and obstetric data of multigravid women who gave birth at a maternity hospital in Tucumán, Argentina, between 2005 and 2010: 949 women without previous PTB nor pregnancy loss who delivered at term and 552 who had spontaneous PTB. NSES was estimated from the Unsatisfied Basic Needs index of census data. Variables selected through penalised regressions were used to create a data-driven Bayesian network; then, pathways were identified and mediation analyses performed.
Results
Maternal age less than 20 years mediated part of the protective effect of high NSES on spontaneous PTB [natural indirect effect (NIE) -0.0125, 95% confidence interval (CI) (-0.0208, -0.0041)] and on few prenatal visits (< 5) [NIE − 0.0095, 95% CI (-0.0166, -0.0025)]. These pathways showed greater sensitivity to unobserved confounders that affect the variables mediator-outcome in the same direction, and exposure-mediator in the opposite direction. They did not show sensitivity to observed potential confounders, nor to the parameterization used to define NSES. Meanwhile, urinary tract infections showed a trend in mediating the effect of low NSES on spontaneous PTB [NIE 0.0044, 95% CI (-0.0006, 0.0093), P 0.0834].
Conclusions
High NSES has protective indirect effects on spontaneous PTB risk, mainly associated with a lower frequency of teenage pregnancy.
... Logistic regression was run with the reoffending outcome as the dependent variable and intervention types by the IASU as independent variables. Due to the modest sample size (n = 71), we ran two versions of the logistic regression model; one standard and one with Firth (1993) method, that acts in a similar way to penalized likelihood to correct for biases associated with small samples. Prior to running the models, we first dichotomized the independent variables so that they were mutually exclusive. ...
... To ensure that our sample was reflective of the population and correct for sampling-related biases, we incorporated replicate weights that were computed using the delete one Jackknife replication method. Associations between levels of segregation and healthcare utilization among informal caregivers (n = 583) were assessed using Firth's penalized logistic regression analyses to resolve issues related to using a general logistic regression model, i.e., a quasi-separation or non-convergence on the estimate [59,60]. In this instance, given that the dependent variable "healthcare visits" was dichotomous, where most response categories were "yes," it was plausible that the maximum likelihood estimation did not exist when using a general logistic regression approach. ...
This study aimed to investigate the impact of racial residential segregation on healthcare utilization and perceived quality of care among informal caregivers in the US. It further assessed potential variations in the estimated impact across caregivers' race and socioeconomic status. We used data from the Health Information National Trends Survey Data Linkage Project (fielded in 2020) for a sample of 583 self-identified informal caregivers in the US. Fitting a series of regression models with the maximum likelihood estimation, we computed the beta coefficients (β) of interest and their associated Wald 95% confidence limits (CI). Caregivers who resided in areas with higher segregation, compared to those living in lower segregated areas, were less likely to visit a healthcare professional [β = − 2.08; Wald 95%CI − 2.093, − 2.067] (moderate); [β = − 2.53; Wald 95%CI − 2.549, − 2.523] (high)]. Further, caregivers residing in moderate [β = − 0.766; Wald 95%CI − 0.770, − 0.761] and high [β = − 0.936; Wald 95%CI − 0.941, − 0.932] segregation regions were less likely to perceive a better quality of care compared to those located in low segregation areas. Moreover, as segregation level increased, Black caregivers were less likely to see a health professional, less frequently used healthcare services, and had poorer perceived healthcare quality when compared to Whites. Our findings indicate that higher residential segregation is associated with lower healthcare utilization, such as visiting a healthcare professional, and poorer perceived healthcare quality among informal caregivers. Given the essential role of informal caregivers in the current healthcare system, it is vital to investigate and address challenges associated with access to and quality of essential healthcare services to improve caregivers' health and well-being, specifically for caregivers of minority backgrounds.
... Prespecified secondary clinical efficacy outcomes and technical efficacy outcomes are shown in table 2 and in the appendix (pp [19][20][21]. At 180 days, 16 (17%) of 96 participants assigned to decompressive craniectomy plus best medical treatment, and 27 (27%) of 101 assigned to best medical treatment alone had died (aRR 0·61, 95% CI 0·36 to 1·01, aRD −11%, 95% CI −21 to 0; for a Kaplan-Meier plot see appendix p 37). Secondary ordinal analysis showed lower mRS following decompressive craniectomy plus best medical treatment compared with best medical treatment alone (common odds ratio for a worse outcome 0·57, 95% CI 0·34 to 0·97; figure 2). ...
... Therefore, it is necessary to use a rare event-adjusted logistic regression. Similar to Kanuri and Andrews (2019) and Salisbury and Zhao (2020), we use a penalized likelihood-based logistic regression: the Firth logit model (Firth, 1993). 5 ...
This study assesses the value of excess and deficit patronization intentions toward a service provider in predicting future customer behavior and its financial consequences for the provider in a continuous service context. The excess and deficit patronization measures employ widely available customer feedback data and can be used by managers to identify at-risk customers and those unlikely to defect. We argue that a customer's satisfaction provides a baseline level of patronization intentions and that excess patronization intentions-intentions greater than those that can be explained by a customer's satisfaction with a firm's offerings (i.e., the residuals in a model that regresses patronization intentions on satisfaction)-are generated in part by the presence of customer-level switching costs. Conversely, any deficit patronization intentions are generated in part by a customer's variety seeking. Using data from the financial services industry, we find that these residuals serve as indicators of the presence and extent of customer-level switching cost and variety seeking. In addition to providing measures of interesting and under-researched phenomena , this suggests that the measures may serve as proxies to test existing theories concerning switching costs and variety seeking in situations where measurement and data availability have previously limited such research.
... (4) Sample size: The study used eight pairs of birds for observations, which is a relatively small sample size and may not capture the full diversity of behaviors across the species. (5) Statistical considerations: They have taken care to use appropriate statistical methods, including Firth's [25] bias reduction for maximum likelihood estimates due to complete separation, and checking for multicollinearity with variance inflation factors. However, any statistical analysis is based on the assumption that the model correctly represents the underlying biological processes. ...
A recent study found evidence of symbolic gesture use in Japanese tits ( Parus minor ). The study reveals how these birds use wing-fluttering movements to transmit an “after you” to their partners, implying a degree of cognitive skill previously thought to be unique to humans and great apes. If confirmed, this research contradicts long-held notions about animal communication by proving that Japanese tits not only participate in intricate vocal communications, which can comprise phrases with specific grammatical rules, but also use body language as a form of engagement. Here, we evaluate this claim by inspecting the data and utilizing bootstrapping to expand the sample size. We find a large variation in how frequently the “after you” gesture is employed in different bootstrap samples, suggesting that while the gesture is a consistent behavior, its frequency can vary significantly. Moreover, the timing of the male’s response to the female’s gesture can fluctuate, though the response itself appears to be a stable phenomenon. Because the frequency of the “after you” gesture varied greatly, especially in light of potential cognitive and other biases influencing the study, this bold claim should be taken with caution.
... В частности, в работе (King, Zeng, 2001) предложен двухшаговый алгоритм, состоящий из переоценки коэффициентов логистической регрессии с учетом поправки на смещение и введения корректирующей функции в логит-модель (подход relogit). Кроме того, в исследовании (Firth, 1993) для решения проблемы смещения, возникающей при применении метода максимального правдоподобия, предлагается включение в логарифмическую функцию правдоподобия штрафного множителя (penalization term), чувствительного к уменьшению выборки и сокращению числа событий (подход firthlogit). Результаты переоценки моделей с использованием двух модификаций (табл. ...
Настоящая статья посвящена анализу факторов, влияющих на участие субъектов малого и среднего предпринимательства в двухуровневом механизме поддержки, в рамках которого госбанки предоставляли конечным получателям льготные кредиты, частичные гарантии и гибридный инструмент (100%-ная гарантия со льготным фондированием от Банка России). В существующей исследовательской литературе преимущественно рассматривается одноуровневый механизм государственной поддержки, в то время как вопросы функционирования двухуровневой системы остаются малоизученными. С учетом специфики изучаемого механизма поддержки в значительной степени был пересмотрен набор факторов, влияющих на участие компаний в программах поддержки. Для количественной оценки использовалась классическая логит-модель, однако, с учетом того что в структуре данных выявлены «редкие события», регрессии были переоценены с использованием двух модификаций (relogit, firthlogit). Анализ показал, что государственные банки в целом обладают низкими стимулами к обслуживанию новых ниш и использованию альтернативных кредитных технологий. При этом сам же механизм государственной поддержки требует значительной корректировки.
This article analyzes the factors influencing the participation of SMEs in the second-tier support mechanism, under which state banks provided end recipients with soft loans, partial guarantees and a hybrid instrument (100% guarantee with preferential funding from the Bank of Russia). The existing research literature predominantly considers the one-tier mechanism of state support, while the issues of functioning of the two-tier system remains understudied. Taking into account the peculiarities of the examines’ support mechanism, the set of factors influencing the participation of companies in support programs was largely reconsidered. For quantitative assessment a classical logit model was used, however, taking into account the presence of "rare events" in the data structure, regressions were re-estimated using two modifications (relogit, firthlogit). The analysis showed that state banks in general have low incentives to serve new clients and use alternative lending technologies. At the same time, the mechanism of state support itself requires significant adjustments.
... All models were performed first on all infants, and then stratified according to SGA or AGA status, using the same covariates. Due to the lower number of SGA infants, incidence of infections was calculated with a penalized logistic regression [32] and cumulative infection risk using the log-rank test. P values less than 0.05 were considered statistically significant, while values less than 0.1 were considered as a tendency to a significant effect. ...
Background
For very preterm infants, human milk is often fortified with formula products based on processed bovine milk. Intact bovine colostrum (BC), rich in anti-inflammatory milk factors, is considered an alternative. We investigated if BC affects anti-inflammatory/TH2 immunity and infection risk in very preterm infants.
Methods
For a secondary analysis of a multicenter, randomized controlled trial (NCT03537365), very preterm infants (26–31 weeks gestation, 23% small for gestational age, SGA) were randomized to receive BC (ColoDan, Biofiber, Denmark, n = 113) or conventional fortifier (PreNAN, Nestlé, Switzerland, n = 116). Infection was defined as antibiotic treatment for five or more consecutive days and 29 cytokines/chemokines were measured in plasma before and after start of fortification.
Results
In general, infection risk after start of fortification was associated with low gestational age, SGA status and antibiotics use prior to fortification. Adjusted for confounders, infants fortified with BC showed more infection episodes (20 vs 12%, P < 0.05) and higher cumulative infection risk (hazard ratio, HR 1.9, P = 0.06), particularly for SGA infants (HR 3.6, P < 0.05). Additionally, BC-fortified infants had higher levels of TH2-related cytokines/chemokines (IL-10, MDC, MCP4) and reduced levels of cytokines related to TH1/TH17-responses (IL-15, IL-17, GM-CSF). The differences were most pronounced in SGA infants, displaying higher levels of TH2-related IL-4, IL-6, and IL-13, and lower interferon-γ and IL-1α levels in the BC group.
Conclusion
Infants fortified with BC displayed a delayed shift from TH2- to TH1-biased systemic immunity, notably in SGA infants, possibly influenced by multiple confounding factors, alongside elevated antibiotic use, suggesting increased susceptibility to infection.
... In the subgroup of patients receiving DMT other than anti-CD20 agents or S1PRMs, Firth logistic regression modelling was used (firthlogist 0.5.0 toolbox, https:// pypi. org/ proje ct/ firth logist/ ) to solve the statistical issue of strict separation (see Table 2), as previously described elsewhere [22,23]. ...
Background and purpose
Coronavirus disease 2019 (COVID‐19) vaccination has been associated with a dampened humoral and/or cellular immune response in patients with multiple sclerosis (MS) who were concurrently on disease‐modifying treatment (DMT) with B‐cell depleting agents or sphingosine‐1‐phosphate receptor modulators (S1PRMs). Our main goal was to investigate the impact of these DMT classes on the clinical effectiveness of COVID‐19 vaccination.
Methods
Since March 2020, demographics and clinical data of patients with MS who developed COVID‐19 have been collected at the Belgian National MS Centre in Melsbroek. Patients were considered to be ‘protected by vaccination’ if they were (i) fully vaccinated and (ii) tested positive for COVID‐19 in the period ranging from 14 days to 6 months after the last administered vaccine.
Results
On 19 December 2022, 418 COVID‐19 cases were retrospectively identified in 389 individual patients. Hospitalization and mortality rates resulting from the infection were 10.8% and 2.4%, respectively. Being ‘unprotected by vaccination’ was significantly associated with a worse COVID‐19 outcome (i.e., hospitalization and/or death) in the total cohort (N = 418, odds ratio [OR] 3.96), in patients on ongoing DMT other than anti‐CD20 agents or S1PRMs (N = 123, OR 31.75) and in patients without DMT (N = 182, OR 5.60), but not in those receiving anti‐CD20 agents (N = 91, OR 0.39); the S1PRMs subgroup was considered too small (22 infections) for any meaningful analysis.
Conclusions
Coronavirus disease 2019 vaccination protects against severe infection in patients with MS but it was not possible to confirm this effect in those on DMT with B‐cell depleting agents.
... Our first estimation strategy could be to follow Bromley-Trujillo, Holman, and Sandoval (2019) and use a multilevel model with random effects, wherein states are the second level of our analysis. Alternatively, Cook et al. (2020) classify this problem as a special case of separation and demonstrate that a PML-FE (penalized maximum likelihood estimator with fixed effects) estimator (Firth 1993) produces more accurate estimates than random effects estimators, given the assumption of orthogonality between predictors and unit effects is likely violated. Finally, Timoneda (2021) demonstrates that the linear probability model with fixed effects outperforms maximum likelihood estimation strategies when outcomes are truly rare (e.g. less than 1 % of observations). ...
Despite substantial progress on understanding the influence of nonprofits in the policy process, the role of nonprofits in local agenda setting remains under-examined. Policy for land conservation is a key plank in climate change mitigation strategies making the role of land trusts in agenda setting for land conservation votes an important topic. Prior studies demonstrate that temperature anomalies increase the likelihood that climate change policy appears on state political agendas. We propose that land trusts play a critical role in facilitating this connection at the local level. Temperature anomalies create a policy window which contributes to the likelihood that conservation measures make it onto the political agenda. Combining temperature, land trust presence, and local conservation vote data from U.S. counties from 2009 to 2019, we find that both temperature anomalies and the presence of land trusts increase the probability that a land conservation measure will appear on the local political agenda. Further, we find that the effect of temperature anomalies on the probability of a vote taking place is conditional on land trust presence within the county. The findings advance existing nonprofit theory on ways in which nonprofits engage in the policy process and hint towards promising avenues of future research. The findings also provide practical evidence for practitioners and hold implications for how local policymakers can leverage existing organizations to implement climate change policy.
... Clinicopathological and CT imaging features of the PD-L1-positive and -negative groups were compared using the chisquare test or Fisher's exact test for categorical variables and the independent t-test (variables with normality) or Mann-Whitney U test (variables without normality) for continuous variables. Firth's logistic regression analysis was performed to determine the association between PD-L1 expression and clinicopathological and CT features to overcome the issues of separation (31,32). To develop a predictive model of PD-L1 expression, features with p < 0.2 in univariate analysis and/or clinical significance based on our clinical experience were selected as 'total significant features' for PD-L1 expression status. ...
Purpose:
To develop models to predict programmed death ligand 1 (PD-L1) expression in pulmonary squamous cell carcinoma (SCC) using CT.
Materials and methods:
A total of 97 patients diagnosed with SCC who underwent PD-L1 expression assay were included in this study. We performed a CT analysis of the tumors using pretreatment CT images. Multiple logistic regression models were constructed to predict PD-L1 positivity in the total patient group and in the 40 advanced-stage (≥ stage IIIB) patients. The area under the receiver operating characteristic curve (AUC) was calculated for each model.
Results:
For the total patient group, the AUC of the 'total significant features model' (tumor stage, tumor size, pleural nodularity, and lung metastasis) was 0.652, and that of the 'selected feature model' (pleural nodularity) was 0.556. For advanced-stage patients, the AUC of the 'selected feature model' (tumor size, pleural nodularity, pulmonary oligometastases, and absence of interstitial lung disease) was 0.897. Among these factors, pleural nodularity and pulmonary oligometastases had the highest odds ratios (8.78 and 16.35, respectively).
Conclusion:
Our model could predict PD-L1 expression in patients with lung SCC, and pleural nodularity and pulmonary oligometastases were notable predictive CT features of PD-L1.
... We employed multiple regression techniques (OLS, logistic, and multinomial logistic) to examine associations between exposure to police violence on digital media and sleep problems, net of covariates. Given the sample size and the small number of positive cases on the outcomes in binary models, the penalized maximum likelihood estimation proposed by Firth is employed to reduce bias [24]. We also employed the Karlson-Holm-Breen (KHB) method to examine how police-related mental health factors attenuate those associations [25]. ...
... Therefore, the full datasets were analyzed with Poisson regressions as count data and the subset data at each residue age were analyzed with logistic regressions as presence-absence data. We used the logistf function from package "logistf" for logistic regressions with complete separation issues because it utilizes a bias-reduced penalized likelihood logistic regression (Firth 1993, Heinze et al. 2022). ...
The Douglas-fir twig weevil (Cylindrocopturus furnissi Buchanan) (Coleoptera: Curculionidae) has recently emerged as a significant pest of Christmas trees grown in the Pacific Northwest United States. The larvae girdle and disfigure twigs, which adversely affects tree marketability. Trees produced for export are also routinely destroyed for phytosanitary reasons when C. furnissi is discovered at border crossings. Due to historically being a sporadic and benign pest on planted and natural Douglas-fir (Psuedotsuga menziesii), there is a lack of chemical management options. In laboratory experiments, we assessed the knockdown effects (ability to kill or incapacitate) of 4 insecticides commonly used on Christmas trees: one assay tested knockdown after direct contact for 24 h, and the other assay tested knockdown after being allowed to feed on treated twigs with 2 days, 7 days, and 14 days residuals. Concurrently, we monitored temperature and adult C. furnissi emergence at a noble fir bough farm for 2 years to estimate the ideal degree-day window for applying insecticides. Bifenthrin and esfenvalerate knocked down all weevils on contact within just 4 h, whereas chlorpyrifos and acephate failed to achieve 100% knockdown within 24 h. Only acephate failed to knock down more weevils than the control (water) after feeding on treated twigs, regardless of the insecticide residue age. Degree-day modeling revealed a variable emergence window between the 2 years but 50% of adult emergence occurred between approximately 1,000–1,100 degree days (1st January, 50 °F (10 °C), single sine). Future work should assess the resulting management recommendation: apply bifenthrin or esfenvalerate once annually just after 1,000 growing degree days for 2 or more years prior to harvest.
... Since the infant mortality rate was low, the Fisher's logistic regression model was used. This model is suitable for handling rare events (14). The stepwise method was used to select variables for the multiple logistic regression analysis, and the significance level was set to 0.15. ...
Background
A single umbilical artery (SUA) may coexist with a single anomaly or multiple congenital anomalies. Although anomalies associated with SUA can primarily cause high perinatal mortality, their clinical significance has not been evaluated.
Objective
We investigated the relationship between the clinical features and the type or number of concurrent anomalies in neonates with SUA.
Materials and Methods
In this cross-sectional study, 104 neonates with SUA were enrolled from January 2000 to December 2020 at Dongsan hospital, Daegu, South Korea. Data on the maternal history and the neonates demographic characteristics, clinical course, chromosomal analysis, and congenital anomalies, were collected.
Results
Among the neonates with SUA included, 77 (74.0%) had one or more congenital anomalies; 66 (63.5%) were cardiac, 20 (19.2%) were genitourinary, 12 (11.5%) were gastrointestinal, 5 (4.8%) were central nervous system, 12 (11.5%) were skeletal, and 5 (4.8%) were facial anomalies. The number of concurrent anomalies ranged from 0–4. Neonates with SUA and concurrent gastrointestinal anomaly had a high incidence of initial positive ventilation, intubation, and inotropic drug use and lower Apgar score at 1 min and 5 min. 7 (6.7%) neonates with SUA died. Low birth weight (odds ratio = 6.16, p = 0.05), maternal multiparity (2.41, p = 0.13), gastrointestinal anomaly (5.06, p = 0.11), and initial cardiac resuscitation (7.77, p = 0.11) were risk factors for mortality in neonates with SUA.
Conclusion
Neonates with SUA and concurrent gastrointestinal anomaly, low birth weight, maternal multiparity, and initial cardiac resuscitation had poor outcomes.
... Univariate logistic regression analysis was used to initially identify possible prognostic factors for delayed neurocognitive recovery at 1 week, and the variables with a P value of < 0.05 were included in subsequent Firth logistic regression analyses. Firth's penalization reduced the small sample bias of maximum likelihood coefficients [17]. For all comparisons, P < 0.05 was considered statistically significant. ...
Background
The effect of two lung ventilation (TLV) with carbon dioxide artificial pneumothorax on cerebral desaturation and postoperative neurocognitive changes in elderly patients undergoing elective minimally invasive esophagectomy (MIE) is unclear.
Objectives
The first aim of this study was to compare the effect of TLV and one lung ventilation (OLV) on cerebral desaturation. The second aim was to assess changes in early postoperative cognitive outcomes of two ventilation methods.
Methods
This prospective, randomized, controlled trial enrolled patients 65 and older scheduled for MIE. Patients were randomly assigned (1:1) to TLV group or OLV group. The primary outcome was the incidence of cerebral desaturation events (CDE). Secondary outcomes were the cumulative area under the curve of desaturation for decreases in regional cerebral oxygen saturation (rSO2) values below 20% relative to the baseline value (AUC.20) and the incidence of delayed neurocognitive recovery.
Results
Fifty-six patients were recruited between November 2019 and August 2020. TLV group had a lower incidence of CDE than OLV group [3 (10.71%) vs. 13 (48.14%), P = 0.002]. TLV group had a lower AUC.20 [0 (0–35.86) % min vs. 0 (0–0) % min, P = 0.007], and the incidence of delayed neurocognitive recovery [2 (7.4%) vs. 11 (40.7%), P = 0.009] than OLV group. Predictors of delayed neurocognitive recovery on postoperative day 7 were age (OR 1.676, 95% CI 1.122 to 2.505, P = 0.006) and AUC.20 (OR 1.059, 95% CI 1.025 to 1.094, P < 0.001).
Conclusion
Compared to OLV, TLV had a lower incidence of CDE and delayed neurocognitive recovery in elderly patients undergoing MIE. The method of TLV combined with carbon dioxide artificial pneumothorax may be an option for these elderly patients.
Chinese Clinical Trial Registry (identifier: ChiCTR1900027454).
... For analytic categories with n < 5, logistic regression with Firth's bias correction was applied, with a penalized regression model, in which a penalty term is placed on the standard ML function to generate parameter estimates [23]. To correct for multiple comparisons, p-values were adjusted using the false discovery rate (FDR) [24]. ...
... The advantage of relying on matched control groups of patents is that it also decreases the influence of potential confounding factors, as we control not only for the linear terms of the covariates, but also for any arbitrary combination of them. Third, we estimate rare events logit models with the Firth logit approach, which is a penalized likelihood method taking into account the low shares of 1 in the outcome variable (Firth, 1993). 26 Finally, since in our analysis we focus only on transacted patent applications, our econometric results might suffer from selection bias if the probability of being transacted differs between patents attractive to NPEs and those attractive to PEs. ...
This paper delves into the proliferation of non-practicing entities (NPEs), a hot topic in academia and public policy, especially in the United States. The common belief is that Europe is less exposed to NPEs due to a robust patent system, higher enforcement costs, and smaller damage awards. Yet, using a new database of NPE patent applications at the European Patent Office (EPO), the study uncovers that NPEs activity in Europe is arising: they own nearly 20,000 EPO patents, primarily in Electrical Engineering. Moreover, we contribute the literature investigating the heterogeneity of the NPE business model and its relationship with the characteristics and use of the patents they target. Our econometric analysis provides threefold original evidence. First, NPEs with higher propensity for litigation (i.e., “Litigation” NPEs) acquire patents with higher infringement risk but similar technological quality than practicing entities. Second, patent aggregators (i.e., “Portfolio” NPEs) and technology companies (i.e., “Technology” NPEs) acquire higher-quality patents compared to those acquired by practicing entities. Third, patent acquisitions by “Litigation” NPEs and “Portfolio” NPEs reduce the subsequent use of protected technologies.
... aroxysmal nocturnal hemoglobinuria (PNH) is a rare disease characterized by hemolysis, thrombosis, and bone marrow failure. 1 PNH is caused by the expansion of hematopoietic stem cells carrying a somatic mutation in PIGA, which encodes phosphatidylinositol N-acetylglucosaminyltransferase, subunit A. 2 Affected red cells lack surface glycosylphosphatidylinositollinked complement regulators CD55 3 and CD59, 4 which renders them susceptible to complement activation leading to intravascular hemolysis (Fig. S1A in the Supplementary Appendix, available with the full text of this article at NEJM.org). 5,6 Intravenous anti-C5 monoclonal antibodies became the standard treatment for hemolytic PNH in 2007 after eculizumab led to intravascular hemolysis control (with hemoglobin stabilization in approximately half the patients), 7-9 reduced thromboembolic risk, 10 and improved long-term survival. ...
Background:
Persistent hemolytic anemia and a lack of oral treatments are challenges for patients with paroxysmal nocturnal hemoglobinuria who have received anti-C5 therapy or have not received complement inhibitors. Iptacopan, a first-in-class oral factor B inhibitor, has been shown to improve hemoglobin levels in these patients.
Methods:
In two phase 3 trials, we assessed iptacopan monotherapy over a 24-week period in patients with hemoglobin levels of less than 10 g per deciliter. In the first, anti-C5-treated patients were randomly assigned to switch to iptacopan or to continue anti-C5 therapy. In the second, single-group trial, patients who had not received complement inhibitors and who had lactate dehydrogenase (LDH) levels more than 1.5 times the upper limit of the normal range received iptacopan. The two primary end points in the first trial were an increase in the hemoglobin level of at least 2 g per deciliter from baseline and a hemoglobin level of at least 12 g per deciliter, each without red-cell transfusion; the primary end point for the second trial was an increase in hemoglobin level of at least 2 g per deciliter from baseline without red-cell transfusion.
Results:
In the first trial, 51 of the 60 patients who received iptacopan had an increase in the hemoglobin level of at least 2 g per deciliter from baseline, and 42 had a hemoglobin level of at least 12 g per deciliter, each without transfusion; none of the 35 anti-C5-treated patients attained the end-point levels. In the second trial, 31 of 33 patients had an increase in the hemoglobin level of at least 2 g per deciliter from baseline without red-cell transfusion. In the first trial, 59 of the 62 patients who received iptacopan and 14 of the 35 anti-C5-treated patients did not require or receive transfusion; in the second trial, no patients required or received transfusion. Treatment with iptacopan increased hemoglobin levels, reduced fatigue, reduced reticulocyte and bilirubin levels, and resulted in mean LDH levels that were less than 1.5 times the upper limit of the normal range. Headache was the most frequent adverse event with iptacopan.
Conclusions:
Iptacopan treatment improved hematologic and clinical outcomes in anti-C5-treated patients with persistent anemia - in whom iptacopan showed superiority to anti-C5 therapy - and in patients who had not received complement inhibitors. (Funded by Novartis; APPLY-PNH ClinicalTrials.gov number, NCT04558918; APPOINT-PNH ClinicalTrials.gov number, NCT04820530.).
... In [4], the author proposed to modify the score function to reduce the bias of the MLE. His proposal is an alternative to the usual approach of subtracting from the MLE its estimated bias. ...
The issues of performing inference on the parameters of quasi-Lindley (QL) distribution and Nadarajah-Haghighi exponential distribution (N-H) is addressed. It is shown graphically that the likelihood function of the quasi-Lindley distribution and Nadarajah-Haghighi exponential distribution is configured with a flat monotone shape. Which makes it very difficult to pick the values of the parameters that maximize the likelihood function. A penalization scheme is used to improve maximum likelihood point estimation. A penalization scheme based on the Jeffreys prior.
... For the toxin tolerance assay, we coded the survival of each individual larvae in a vial using a binary strategy (0 = deceased, 1 = adult successfully emerged). To assess the contribution of genotype (stock), treatment (diet), and the interaction between the two factors on the probability of survival, we performed a logistic regression analysis using a generalized linear model with binomial distribution, logit link, and Firth bias-adjusted estimates (Firth, 1993). ...
Understanding the often antagonistic plant–herbivore interactions and how host defenses can influence herbivore dietary breadth is an area of ongoing study in ecology and evolutionary biology. Typically, host plants/fungi that produce highly noxious chemical defenses are only fed on by specialists. We know very little about generalist species that can feed and develop on a noxious host. One such example of generalists feeding on toxic host occurs in the mushroom‐feeding Drosophila found in the immigrans‐tripunctata radiation. Although these species are classified as generalists, their acceptable hosts include deadly Amanita species. In this study, we used behavioral assays to assess associations between one mushroom‐feeding species, Drosophila guttifera , and the deadly Amanita phalloides . We conducted feeding assays to confirm the presence of cyclopeptide toxin tolerance. We then completed host preference assays in female flies and larvae and did not find a preference for toxic mushrooms in either. Finally, we assessed the effect of competition on oviposition preference. We found that the presence of a competitor's eggs on the preferred host was associated with the flies increasing the number of eggs laid on the toxic mushrooms. Our results highlight how access to a low competition host resource may help to maintain associations between a generalist species and a highly toxic host.
... When outcomes were rare events (i.e., the outcome under consideration had a small incident probability), for example, in the case of all-cause mortality outcomes, Firth's bias-reduced multivariable logistic regression model was fitted. 24 Multivariable ordered logistic regression was used to model the proportional odds of severity of hospitalization, which was an ordinal outcome. While treatment status was the main predictor, all regression models were adjusted for all covariates of interest, which were included in the PSM analysis. ...
Objectives
To evaluate whether subcutaneous neutralizing monoclonal antibody (mAb) treatment given in the emergency department (ED) setting was associated with reduced hospitalizations, mortality, and severity of disease when compared to nontreatment among mAb‐eligible patients with coronavirus disease 2019 (COVID‐19).
Methods
This retrospective observational cohort study of ED patients utilized a propensity score‐matched analysis to compare patients who received subcutaneous casirivimab and imdevimab mAb to nontreated COVID‐19 control patients in November–December 2021. The primary outcome was all‐cause hospitalization within 28 days, and secondary outcomes were 90‐day hospitalization, 28‐ and 90‐day mortality, and ED length of stay (LOS).
Results
Of 1340 patients included in the analysis, 490 received subcutaneous casirivimab and imdevimab, and 850 did not received them. There was no difference observed for 28‐day hospitalization (8.4% vs. 10.6%; adjusted odds ratio [aOR] 0.79, 95% confidence intervals [CI] 0.53–1.17) or 90‐day hospitalization (11.6% vs. 12.5%; aOR 0.93, 95% CI 0.65–1.31). However, mortality at both the 28‐day and 90‐day timepoints was substantially lower in the treated group (28‐day 0.6% vs. 3.1%; aOR 0.18, 95% CI 0.08–0.41; 90‐day 0.6% vs. 3.9%; aOR 0.14, 95% CI 0.06–0.36). Among hospitalized patients, treated patients had shorter hospital LOS (5.7 vs. 11.4 days; adjusted rate ratio [aRR] 0.47, 95% CI 0.33–0.69), shorter intensive care unit LOS (3.8 vs. 10.2 days; aRR 0.22, 95% CI 0.14–0.35), and the severity of hospitalization was lower (aOR 0.45, 95% CI 0.21–0.97) compared to untreated.
Conclusions
Among ED patients who presented for symptomatic COVID‐19 during the Delta variant phase, ED subcutaneous casirivimab/imdevimab treatment was not associated with a decrease in hospitalizations. However, treatment was associated with lower mortality at 28 and 90 days, hospital LOS, and overall severity of illness.
... Adjusted mortality hazard ratios (HRs) were calculated using Cox models that included age group, sex, race, and CKD stage (for patients with non-ESRD CKD) or ESRD treatment (i.e., dialysis and transplant for patients with ESRD). When a category had zero events, Firth penalized maximum likelihood estimation was used to reduce bias in the estimates [25]. All adjusted HRs were reported as point estimates with 95% CIs. ...
Background
Autosomal dominant polycystic kidney disease (ADPKD) is the leading inheritable cause of end-stage renal disease (ESRD). Mortality data specific to patients with ADPKD is currently lacking; thus, the aim of this study was to estimate mortality in patients with ADPKD.
Methods
We analyzed data from the United States Renal Data System (USRDS) for patients with ADPKD available during the study period of 01/01/2014–12/31/2016, which included a cohort of patients with non-ESRD chronic kidney disease (CKD) and a cohort of patients with ESRD. Mortality rates with 95% confidence intervals (CIs) were calculated overall and by age group, sex, and race for the full dataset and for a subset of patients aged ≥ 65 years. Adjusted mortality hazard ratios (HRs) were calculated using Cox regression modeling by age group, sex, race, and CKD stage (i.e., non-ESRD CKD stages 1–5) or ESRD treatment (i.e., dialysis and transplant).
Results
A total of 1,936 patients with ADPKD and non-ESRD CKD and 37,461 patients with ADPKD and ESRD were included in the analysis. Age-adjusted mortality was 18.4 deaths per 1,000 patient-years in the non-ESRD CKD cohort and 37.4 deaths per 1,000 patient-years in the ESRD cohort. As expected, among the non-ESRD CKD cohort, patients in CKD stages 4 and 5 had a higher risk of death than patients in stage 3 (HR = 1.59 for stage 4 and HR = 2.71 for stage 5). Among the ESRD cohort, patients receiving dialysis were more likely to experience death than patients who received transplant (HR = 2.36). Age-adjusted mortality among patients aged ≥ 65 years in the non-ESRD CKD cohort was highest for Black patients (82.7 deaths per 1,000 patient-years), whereas age-adjusted mortality among patients aged ≥ 65 years in the ESRD cohort was highest for White patients (136.1 deaths per 1,000 patient-years).
Conclusions
Mortality rates specific to patients aged ≥ 65 years suggest racial differences in mortality among these patients in both non-ESRD CKD and ESRD cohorts. These data fill an important knowledge gap in mortality estimates for patients with ADPKD in the United States.
... In addition, variables were also investigated with the complete dataset using logistic regression. As the squalor prevalence values are low, Firth's Bias-Reduced logistic regression [64] was used to account for the rarity of squalor events in the datasets [65]. To analyse squalor prevalence over time, fixed prevalence data was calculated for each variable and for each year and displayed in values and as line graphs. ...
Background
Severe domestic squalor occurs when a person lives in a dwelling that is significantly unclean, disorganised and unhygienic. The limited previous research has primarily focused on the characteristics of those who live in squalor and the associated risk factors. Robust and reliable studies of squalor prevalence have not been conducted. This study sought to produce a reliable estimate of the point prevalence of squalor.
Methods
Using data from 13-years of the English Housing Survey, N = 85,681 households were included in a prevalence meta-analysis. Squalor prevalence over time, subgroup analysis and logistic regression investigated the role played by household and community characteristics.
Results
The point prevalence of squalor was estimated to be 0.85% and squalor was seen to decrease significantly over time. More significant community deprivation, a rented dwelling, lower income and high numbers of people in the home was associated with a greater risk of squalor.
Conclusions
Squalor prevalence was higher than previous estimates and supports community care services in associated service planning. The results regarding household characteristics help to inform which households and individuals may be at a higher risk of living in squalid conditions.
... Due to the rarity of cohabitation, conventional logistic regression may lead to biased estimates of probabilities of cohabitation (36). We apply Firth logistic regression to correct the potential biases due to the rarity of cohabitation (37,38). Different from regular logistic regression, Firth Cohabitation of Chinese Muslims, Page 20 logistic regression adds the inverse of the amount of information the data contain to the log likelihood function as the penalty so that the coefficients will be adjusted in proportion to the level of ignorance due to rare observations. ...
While modern family-related ideas and behaviors have become more widely accepted in contemporary China, Chinese Muslim minorities continue to hold on to traditional religious practices. Surprisingly, data from our survey conducted in Gansu province in China’s northwestern borderlands reveal that Muslims of the Hui and Dongxiang ethnicities reported much higher rates of cohabitation experience than the secular majority Han. Based on follow-up qualitative interviews, we found the answer to lie in the interplay between the highly interventionist Chinese state and the robust cultural resilience of local Islamic communities. While the state maintains a high minimum legal age of marriage, the early marriage norm remains strong in Chinese Muslim communities, where religion constitutes an alternative and often more powerful source of legitimacy—at least in the private sphere of life. Using the 2000 census data, we further show that women in almost all 10 Muslim ethnic groups have higher percentages of underage births and premarital births than Han women, both nationally and in the northwest where most Chinese Muslims live. As the once-outlawed behavior of cohabitation became more socially acceptable during the reform and opening-up era, young Muslim Chinese often found themselves in “arranged cohabitations” as de facto marriages formed at younger-than-legal ages. In doing so, Chinese Muslim communities have reinvented the meaning of cohabitation. Rather than liberal intimate relationship based on individual autonomy, cohabitation has served as a coping strategy by which Islamic patriarchs circumvent the Chinese state’s aggressive regulations aimed at “modernizing” the Muslim family.
... For PHVL, missing data for day 7 SARS-CoV-2 viral load were imputed using relevance sequence imputation (RSI) as follows: imputed using the first available non-missing data for day 5, day 3, day 11, or day 1. For categorical measures, statistical hypotheses testing was conducted using logistic regression with a regression model with a Firth penalized likelihood [12]. The differences in the proportion of PHVL among the treatment groups, along with the relative risk and odds ratio, were estimated using data with RSI. ...
Treatment with monoclonal antibodies provides rapid, passive immunity and may stop COVID-19 disease progression. The study evaluated the effect of bamlanivimab (BAM) or BAM + etesevimab (ETE)/sotrovimab compared to placebo on SARS-CoV-2 viral load in patients with COVID-19.
The phase 2, randomized, single-dose study included patients aged between ≥ 18 and < 65 years, not hospitalized at the time of randomization, and had ≥ 1 mild or moderate COVID-19 symptoms. Study included arms 1–6 (placebo, BAM 175 mg + ETE 350 mg, BAM 700 mg + ETE 1400 mg, BAM 2800 mg + ETE 2800 mg, BAM 700 mg alone, and BAM 350 mg + ETE 700 mg, respectively), BAM 700 mg + ETE 700 mg unintentional dosing; and arms 7 and 8 (BAM 700 mg + sotrovimab 500 mg and placebo, respectively). The primary endpoint was proportion of patients with SARS-CoV-2 log viral load > 5.27 on day 7 (persistently high viral load [PHVL]) who received BAM or BAM + (ETE or sotrovimab).
A total of 725 patients, mean age 39.6 years (range 18–75 years), 50.2% male were randomized and infused with study drug in arms 1–6; and a total 202 patients, mean age 38 years (range 18–63 years), 53.5% female were randomized and infused with study drug in arms 7 and 8. A significantly lower proportion of patients in arms 2–6 and arm 7 experienced PHVL on day 7 compared to placebo. On day 7, patients in arms 2, 3, and 6 consistently experienced significantly greater reduction in viral load than placebo. Significant improvement was observed in time to viral load clearance and time to symptom improvement by day 29 in some arms compared to placebo. No new safety concerns were observed with drug combinations.
The study demonstrated that a significantly lower proportion of patients with mild-to-moderate COVID-19 treated with BAM or BAM + (ETE or sotrovimab) experienced a PHVL at day 7.
ClinicalTrials.gov identifier, NCT04634409.
... Survival rates in the second and eighth years postplanting were assessed through GLMs with binomial distributions using the "brglm2" package's glm function (Kosmidis and Firth 2021), based on Firth's bias correction for complete separation (Firth 1993). Survival in the second year was analyzed to determine the effect of summer planting, whereas eighth-year survival was analyzed to determine the effects of silvicultural practices. ...
In Hokkaido, northern Japan, bareroot Sakhalin fir seedlings are conventionally planted in short periods during spring and fall following strip site preparation that alternates managed and unmanaged strips to reduce labor requirements. Postplanting, seven years of weeding are imperative owing to the relentless encroachment of evergreen dwarf bamboo from unmanaged strips. Given diminishing forest labor availability, this protocol poses increasing challenges. We hypothesized that comprehensive mechanical site preparation (MSP) to eradicate dwarf bamboo across the entire site could reduce weeding frequency by delaying the regeneration of competing vegetation. Furthermore, using container seedlings could extend the planting season to summer. Over an eight-year period, we examined the survival and growth of summer-planted bareroot and container seedlings in a fully MSP-treated site, with a single weeding operation conducted in the fifth year. Whole-site MSP induced a shift from dwarf bamboo to deciduous woody and herbaceous species, fostering consistently high Sakhalin fir seedling survival and growth rates despite the lone weeding operation. Container seedlings exhibited superior establishment after summer planting and maintained higher survival rates over eight years relative to bareroot seedlings. Planting in the forwarder trail detrimentally impacted overall seedling growth and survival. These findings advocate for container seedling summer planting alongside comprehensive MSP as an effective Sakhalin fir reforestation technique, mitigating weeding frequency and broadening the planting timeframe.
In a high dimensional regression setting in which the number of variables ($p$) is much larger than the sample size ($n$), the number of possible two-way interactions between the variables is immense. If the number of variables is in the order of one million, which is usually the case in e.g., genetics, the number of two-way interactions is of the order one million squared. In the pursuit of detecting two-way interactions, testing all pairs for interactions one-by-one is computational unfeasible and the multiple testing correction will be severe. In this paper we describe a two-stage testing procedure consisting of a screening and an evaluation stage. It is proven that, under some assumptions, the tests-statistics in the two stages are asymptotically independent. As a result, multiplicity correction in the second stage is only needed for the number of statistical tests that are actually performed in that stage. This increases the power of the testing procedure. Also, since the testing procedure in the first stage is computational simple, the computational burden is lowered. Simulations have been performed for multiple settings and regression models (generalized linear models and Cox PH model) to study the performance of the two-stage testing procedure. The results show type I error control and an increase in power compared to the procedure in which the pairs are tested one-by-one.
Due to their low frequency, estimating the effect of rare variants is challenging. Here, we propose RareEffect, a method that first estimates gene or region-based heritability and then each variant effect size using an empirical Bayesian approach. Our method uses a variance component model, popular in rare variant tests, and is designed to provide two levels of effect sizes, gene/region-level and variant-level, which can provide better interpretation. To adjust for the case-control imbalance in phenotypes, our approach uses a fast implementation of the Firth bias correction. We demonstrate the accuracy and computational efficiency of our method through extensive simulations and the analysis of UK Biobank whole exome sequencing data for five continuous traits and five binary disease phenotypes. Additionally, we show that the effect sizes obtained from our model can be leveraged to improve the performance of polygenic scores.
Focus groups have become increasingly popular in political science alongside the growth in field experimental and other causal inference-oriented work in comparative politics. Yet, scholars rarely provide details about recruitment processes and descriptive statistics on focus-group participants. This situation is problematic given the likelihood of self-selection and the fact that scholars often use focus groups to pretest or refine experimental treatments or survey questionnaires. By leveraging a series of focus groups that were recruited from a pool of large-N survey respondents, this article demonstrates a method for assessing which variables drive the decision to participate. I recommend that scholars diagnose self-selection into focus groups whenever possible; that they compare participants to relevant baselines when working with samples of convenience; and that they always provide descriptive statistics and details on how focus-group members were recruited.
Stark intergroup variation in prosocial behavior, as elicited with economic experiments, is evident even though humans are highly mobile. Conformity to local norms has been posited to play an integral role in the maintenance of this variation. Experiments suggest that adults indeed rapidly infer pro- and antisocial norms in a new or changed social environment and adjust their behavior to reflect the inferred norms. Studies of the ontogeny of prosocial behavior show that by middle childhood, children’s prosocial behavior conforms to that of local adults. Furthermore, by this stage, children are susceptible to the manipulation of explicit normative information. However, as yet unknown is whether children concomitantly have the propensity to 1) rapidly infer local cooperative norms in a novel, realistic social environment, 2) extend these inferences to norms for unobserved behaviors, and 3) apply the inferred norms in the same environment. Here, we used a slideshow to introduce children (age nine to eleven) to a novel social environment, Neighborhood X, which differed by condition (Prosocial or Antisocial). We measured perceived cooperative norms in children’s Own Neighborhood and in Neighborhood X via questionnaires; norms for Neighborhood X diverged drastically dependent upon condition, a result robust to the exclusion of questions about norms for unobserved behaviors. Children’s perceived cooperative norms in Own Neighborhood predicted their prosocial behavior (Dictator Game) in Own Neighborhood. Moreover, even though information about giving behavior was not presented in the slideshow, inferred norms for Neighborhood X predicted children’s prosocial behavior in the same milieu. Changes from baseline prosocial behavior, as measured with a separate helping task, did not extend beyond Neighborhood X. Our results are consistent with the hypothesis that humans have a propensity to rapidly infer and conform to local cooperative norms, thus maintaining group differences in prosocial behavior, and further indicate that this propensity is in operation by middle childhood.
Background
Leptospirosis is a zoonotic disease. It is particularly prevalent in tropical countries and has major consequences for human and animal health. In Benin, the disease's epidemiology remains poorly understood, especially in livestock, for which data are lacking.
Objectives
To characterise Leptospira seroprevalence and locally circulating serogroups in livestock from Cotonou and to estimate the prevalence of Leptospira renal carriage in cattle.
Methods
We conducted a cross‐sectional study in February 2020 during which livestock were sampled at an abattoir and in an impoverished city district. We analysed blood samples from 279 livestock animals (i.e. cattle, sheep, goats and pigs) using the microscopic agglutination test. Additionally, samples of renal tissue from 100 cattle underwent 16s rRNA (rrs) real‐time PCR analysis.
Results
For the 131 cattle, 85 sheep, and 50 goats tested, seroprevalence was 18% (95% confidence interval [CI] [12%, 26%]), 9% (95% CI [4%, 17%] and 2% (95% CI [0%, 9%]), respectively, and most of the seropositive animals were associated with 1:100 titres. All 13 pigs were seronegative. Leptospira DNA was found in the renal tissue of 10% (95% CI [5%, 18%]) of the cattle tested (n = 100). Leptospira borgpetersenii was the main species present (n = 7), but Leptospira interrogans (n = 2) and Leptospira kirschneri (n = 1) were also detected. Various serogroups (Canicola, Grippotyphosa, Sejroe, Icterohaemorrhagiae, Pomona, Pyrogenes, Australis and Autumnalis) were detected using microscopic agglutination test without a clear predominance of any of them.
Conclusions
These results suggest that abattoir workers and people living in close contact with livestock in poor urban areas are exposed to the risk of Leptospira infection.
Non-linguistic pattern learning uses distinct implicit and explicit processes, which differ in behavioural signatures, inductive biases and proposed model architectures. This study asked whether both processes are available in phonotactic learning in the lab. Five Internet experiments collected generalisation, learning curves, response times and detailed debriefings from 671 valid participants. Implicit and explicit learners were found in all conditions and experiments. Objective measures of implicit vs. explicit learning were correlated with introspective self-report. Participants spontaneously discovered and named phonetic features. These findings contradict the common (usually tacit) assumption that ‘artificial-language’ participants learn only implicitly. Learning mode also affected inductive bias: Implicit learning improved performance on family-resemblance patterns relative to biconditionals (if-and-only-if, exclusive-or) in two experiments. The direction of this effect is unexpected under many current theories of how implicit and explicit concept learning differ, and is consistent with models of explicit learning which take pattern-irrelevant features into account.
Objectives
Hospital‐acquired catheter‐associated urinary tract infections (UTIs) have been regarded as preventable adverse events, yet their risk in endoscopic skull base surgery (ESBS) has not been well described despite common use. We determine the incidence of UTI following ESBS and identify contributing clinical factors.
Methods
Retrospective review was conducted for a cohort of 229 consecutive adult patients who underwent endoscopic endonasal surgery for treatment of any skull base pathology between July 2018 and June 2022 at a tertiary academic skull base surgery program. Postoperative UTI comprised the primary outcome. Independent variables included patient demographics, use and length of urinary catheterization, and pre‐existing genitourinary conditions.
Results
Nosocomial UTIs were identified in 1.3% (3/229) of patients, occurring on postoperative days 2, 9, and 14, respectively; all were catheter‐associated. Overall, 86.0% (197/229) of patients received urinary catheters (mean duration 2.2 ± 1.8 days). Compared to those without, patients with UTI were older (70.0 ± 15.4 vs. 52.2 ± 16.8 years, p = 0.034), had lengthier stays (94.7 ± 126.8 vs. 5.9 ± 8.4 days, p < 0.001), and had prolonged catheterizations (9.3 ± 5.5 vs. 2.1 ± 1.5 days, p < 0.001). Preoperative genitourinary conditions were also associated with UTI development, namely, chronic urinary retention/obstruction (66.7% vs. 4.0%, p = 0.006), urinary incontinence (66.7% vs. 6.2%, p = 0.013), prostate disease (100.0% vs. 17.8%, p = 0.035), and renal dysfunction (100.0% vs. 9.7%, p = 0.001). Among intraoperative cerebrospinal fluid leak patients, postoperative CSF leak incidence was not associated with catheter use versus nonuse (3.3% vs. 12.5%, p = 0.276).
Conclusion
Although UTIs are uncommon in ESBS patients, advanced age, length of stay, duration of indwelling urinary catheterization, and comorbid genitourinary conditions may elevate risk.
Shigella spp. are a leading bacterial cause of diarrhea. No widely licensed vaccines are available and there is no generally accepted correlate of protection. We tested a S. sonnei Generalized Modules for Membrane Antigen (GMMA)-based vaccine (1790GAHB) in a phase 2b, placebo-controlled, randomized, controlled human infection model study (NCT03527173) enrolling healthy United States adults aged 18–50 years. We report analyses evaluating immune responses to vaccination, with the aim to identify correlates of risk for shigellosis among assessed immunomarkers. We found that 1790GAHB elicited S. sonnei lipopolysaccharide specific α4β7+ immunoglobulin (Ig) G and IgA secreting B cells which are likely homing to the gut, indicating the ability to induce a mucosal in addition to a systemic response, despite parenteral delivery. We were unable to establish or confirm threshold levels that predict vaccine efficacy facilitating the evaluation of vaccine candidates. However, serum anti-lipopolysaccharide IgG and bactericidal activity were identified as potential correlates of risk for shigellosis.
The Peace‐Athabasca Delta in Alberta, Canada has numerous perched basins that are primarily recharged after large ice jams cause floods (an ecological benefit). Previous studies have estimated that such large floods are likely to decrease in frequency under various climate projections. However, there is a sizable uncertainty range in these predicted flood probabilities, in part due to the short 60‐year systematic record that contained few large ice jam floods. An additional 50 years of historical data are available from various sources, with expert‐interpreted flood categories; however, these categorizations are uncertain in magnitude and occurrence. We developed a Bayesian framework that considers magnitude and occurrence uncertainties within a logistic regression model that predicts the annual probability of a large flood. The Bayesian regression estimates the joint distribution of parameters describing the effects of climatic factors and parameters that describe the probability that historical flood magnitudes were recorded as large (or not) when a truly large (or not) flood occurred. We compare four models for hindcasting and projecting large ice jam flood probabilities in future climates. The models consider: (a) historical data uncertainty, (b) no historical data uncertainty, (c) only the systematic record, and (d) the systematic record with a different model. Neglecting historical data uncertainty provides inaccurate estimates, while using only the systematic record provides wider prediction intervals than considering the full record with uncertain historical data. Thus, we demonstrate that including uncertain historical information can effectively extend the record length and make flood frequency analyses more accurate and precise.
We propose and evaluate an automated pipeline for discovering significant topics from legal decision texts by passing features synthesized with topic models through penalized regressions and post-selection significance tests. The method identifies case topics significantly correlated with outcomes, topic-word distributions which can be manually interpreted to gain insights about significant topics, and case-topic weights which can be used to identify representative cases for each topic. We demonstrate the method on a new dataset of domain name disputes and a canonical dataset of European Court of Human Rights violation cases. Topic models based on latent semantic analysis as well as language model embeddings are evaluated. We show that topics derived by the pipeline are consistent with legal doctrines in both areas and can be useful in other related legal analysis tasks.
This article is part of the theme issue ‘A complexity science approach to law and governance’.
Objective and Design Forced anal examinations are used to prosecute sexual and gender minorities (SGM) in at least seven countries under the presumption that decreased sphincter tone, estimated by a finger inserted into the anal canal, can detect persons practicing receptive anal intercourse. In a cross-sectional analysis of the baseline data from a longitudinal study, we aimed to determine factors associated with sphincter tone and the accuracy of sphincter tonality to detect persons engaging in receptive anal intercourse.
Setting Clinicians in Chicago, Houston, and Milwaukee, USA conducted digital anal rectal examinations (DARE) on 838 participants, 94.0% of whom were cisgendered males. Clinicians used the Digital Rectal Examination Scoring System to score sphincter resting tone (RT) and squeeze tone (ST). On a separate survey, individuals reported their preferred position for anal intercourse: i.e., either always/mostly insertive anal intercourse, always/mostly receptive anal intercourse, or both receptive and insertive anal intercourse. Multivariable regression assessed factors associated with decreased sphincter tone while area under the Receiver Operating Characteristic curves (AUC) estimated the accuracy of sphincter tonality to detect receptive anal intercourse.
Results 11.3% had decreased RT (95/838) and 6.3% had decreased ST (53/838). The accuracy of DARE to detect any receptive anal intercourse was little better than random guessing (AUC 0.53, 95% CI 0.51 to 0.55, and AUC 0.51, 95% CI 0.49 to 0.53, respectively. RT and ST decreased with age regardless of sexual behavior (ptrend<0.01 for both). Compared to individuals having always/mostly insertive anal intercourse, individuals having always/mostly receptive anal intercourse was associated with decreased RT, but not ST, while those equally preferring both insertive and receptive anal intercourse were not associated with decreased RT or ST.
Conclusions Decreased sphincter tone is uncommon among SGM who prefer receptive anal intercourse. Given virtually no accuracy, a finger inserted into the anus has no utility to detect individuals practicing receptive anal intercourse and thus should not be used as such.
Trial registration NCT04090060
ResearchGate has not been able to resolve any references for this publication.