Tag: statistics

J Vis. 2025 Apr 1;25(5):43. doi: 10.1167/jov.25.5.43.

ABSTRACT

Stable visual ensemble statistics can support performance on a visual search task (e.g. Corbett and Melcher, 2014). Such results may indicate improvements in the efficiency of visual cognition in response to a more predictable environment. Observations of improved cognition when immersed in nature (e.g. Berman et al., 2008) have been related to concepts of perceptual fluency and Attention Restoration. We investigated whether the statistical stability of natural scenes could be the underlying mechanism supporting enhancements in visual cognition related to natural images. In experiment 1, we replicated the first study from Corbett and Melcher (2014), showing that sequences of trials with a stable mean size of Gabor elements results in enhanced visual search for an orientation singleton target, compared to sequences with an unstable mean size. In experiment 2, we embedded visual search targets within a set of natural scene images. We leveraged existing variation in the image statistics between images to present sequences where the slope of the Fourier amplitude spectrum was relatively stable (gradually increasing/decreasing) or unstable (randomly ordered). The results have implications for our understanding of the effect of the visual environment on visuo-cognitive functions and the extraction of image statistics by the visual system. We discuss the likely role for eye movements in sampling natural scenes.

PMID:40215026 | DOI:10.1167/jov.25.5.43

J Vis. 2025 Apr 1;25(5):54. doi: 10.1167/jov.25.5.54.

ABSTRACT

Exposure to the colour statistics of natural scenes can both induce and counteract individual differences in colour perception. If different people inhabit different chromatic environments, calibration of the visual system to the colour statistics of those environments can cause individual differences in colour perception. Conversely, exposure to common colour statistics in a common visual environment can reduce individual differences in colour perception that would otherwise be caused by individual differences in physiological factors such as macular pigment density, lens density and cone spectral sensitivities. I will present some examples of our research on these themes, including a cross-environmental study on colour perception between participants living in remote rural versus urban environments in Ecuador (Skelton et al. 2023, Proc. Roy. Soc. B), and studies that explore how the visual systems of anomalous trichromats compensate for their altered cone spectral sensitivities.

PMID:40215015 | DOI:10.1167/jov.25.5.54

J Vis. 2025 Apr 1;25(5):57. doi: 10.1167/jov.25.5.57.

ABSTRACT

Infants start to interact with their visual environment during the first postnatal months. Immaturities in gross motor responses and spatial vision constrain their visual behavior during this rapid development. Analyses of first-person video and eye-tracking data from infants were performed to understand key components of visual experience during this period of visual learning.

METHODS: Infants wore head-mounted scene and binocular eye-tracking cameras (modified Pupil Labs Core) while engaging in naturalistic behavior in an 8ftx8ft home-like environment. Calibrated eye movements were identified using standard approaches (e.g. Engbert & Mergenthaler, 2006) and image statistics were extracted at fixation locations (>200ms).

RESULTS: Recordings (10.5 hours) at ages 2-3 (n=24) 5-6 (35) 8-9 (27) & 11-12 (11) months were analyzed. Eye position and saccade amplitude distributions relative to the head were tighter for younger infants. The distribution of RMS contrast around fixation was also highest at younger ages.

CONCLUSIONS: The youngest infants with limited head and trunk control exhibited the most restricted range of eye movements, suggesting no gaze shift compensation for limited mobility. This likely leads to less active sampling of the scene, slower rates of change in input, and a tight link between head- and eye-centered frames of reference. Early experience also provides a concentration of contrast serving the development of foveal and parafoveal function.

PMID:40215012 | DOI:10.1167/jov.25.5.57

Psychooncology. 2025 Apr;34(4):e70146. doi: 10.1002/pon.70146.

ABSTRACT

OBJECTIVE: To identify the contribution of psychosocial characteristics, supportive care needs, or quality of life on breast cancer survival outcomes.

METHODS: This study used data from a population-based longitudinal study involving women diagnosed with invasive breast cancer (n = 3326, response rate = 71%) in Queensland, Australia, 2010-2013, and followed up to 2020. Flexible parametric survival models were used to identify which factors were associated with survival outcomes. Model fit was assessed using D and $R_D^2$ statistics.

RESULTS: Unmet physical and daily living needs, social support, age, stage at diagnosis, tumour grade, clinical subtype and mode of detection explained 39% of survival variability ( $R_D^2$ 0.39; 95% CI 0.33-0.44), with a Harrell’s C statistic of 0.84 (95% CI 0.81-0.86). Unmet physical and daily living needs and social support, which fall under the categories of supportive care needs and psychosocial characteristics respectively, were identified as key factors that predict breast cancer survival, explaining 3% of survival variability. When compared to women who had less unmet physical needs and adequate social support (5-year survival: 96.6%, 95% CI 92%-99%), those who had more unmet physical needs and limited social support had poorer breast cancer-specific survival (5-year survival: 86.8%, 95% CI 72%-95%).

CONCLUSION: The study found that unmet physical and daily living needs and social support play a marginal but significant role in influencing breast cancer outcomes. The findings enhance the current literature regarding the impact of psychosocial characteristics and supportive care needs on breast cancer survival and suggest that integrating psychosocial support and interventions alongside medical treatment may further improve the survival outcomes for women diagnosed with breast cancer.

PMID:40215001 | DOI:10.1002/pon.70146

Pediatr Blood Cancer. 2025 Apr 11:e31700. doi: 10.1002/pbc.31700. Online ahead of print.

ABSTRACT

BACKGROUND: Children with cancer face an increased risk of complications and death beyond the 5-year survival mark. National surveillance efforts facilitate the systematic tracking of long-term health outcomes, including treatment-related complications and late mortality, among childhood cancer survivors. We aimed to describe the population of 5-year childhood cancer survivors in Canada, quantify the risk of death among survivors relative to the general population, and identify characteristics associated with late mortality.

METHODS: This retrospective cohort study used the Canadian Cancer Registry linked to the Canadian Vital Statistics-Death database (excludes Quebec). Survivors were diagnosed with cancer before 15 years old (1992-2012) and still alive five years after diagnosis. We approximated the risk of late mortality relative to the general population using standardized mortality ratios (SMRs) and absolute excess ratios (AERs). Cumulative all-cause and cause-specific mortality and time-to-event models identified characteristics associated with late mortality.

RESULTS: Of the 10,800 5-year survivors, 405 (4%) had a late death by 2017 (median follow-up: 9.1 years). Cancer recurrence or progression caused most late deaths (64%), followed by subsequent primary neoplasms (11%) and other health-related causes (15%). Survivors had a higher risk of all-cause mortality than the general population (SMR = 9.4; 95% CI = 8.5-10.4; AER = 34.8, 95% CI = 30.8-38.8). Risk was highest in the first 5-9 years of follow-up. Cumulative mortality differed significantly by age at diagnosis, sex and cancer type.

INTERPRETATION: Our results underline the importance of long-term surveillance of childhood cancer survivors, as mortality rates remain higher than the general population for at least two decades after diagnosis.

PMID:40214998 | DOI:10.1002/pbc.31700

JAMA Netw Open. 2025 Apr 1;8(4):e254516. doi: 10.1001/jamanetworkopen.2025.4516.

ABSTRACT

IMPORTANCE: Following decades-long declines, significant increases in abortion rates have been reported in some jurisdictions from 2020 to 2023, but it is not yet known whether these trends are occurring in Canada.

OBJECTIVE: To assess abortion rates in Ontario from 2012 to 2022 and to examine trend changes associated with mifepristone availability, the COVID-19 pandemic, and postpandemic periods.

DESIGN, SETTING, AND PARTICIPANTS: This population-based interrupted time series cohort study examined all medication and procedural abortions provided in Ontario from January 1, 2012, to December 31, 2022, to females aged 15 to 44 years with provincial insurance coverage, identified using linked health administrative data that included records from practitioner billings, inpatient and outpatient hospital services, same-day surgeries, and outpatient prescription dispensations.

EXPOSURE: Availability of mifepristone regulated as a normal (ie, prescribed by an authorized prescriber without additional certification or registration and dispensed by a pharmacist) prescription medication (in November 2017) and the COVID-19 pandemic period (from March 2020 to December 2021).

MAIN OUTCOMES AND MEASURES: The main outcome was the abortion rate (number of abortions per 1000 females per year) overall and within age strata, using an interrupted time series design.

RESULTS: Of 422 867 medication and procedural abortions identified using data from health records of 225 540 reproductive-aged females (mean [SD] age, 28.5 [6.6] years), the abortion rate declined steadily from 15.6 abortions per year per 1000 females, aged 15 to 44 years, in 2012 to 12.3 in 2021 and then increased to 14.1 in 2022. When mifepristone was introduced in 2017 as a normal prescription medication, no immediate change in the abortion rate (-0.1 [95% CI, -0.7 to 0.8]) and a nonsignificant slope increase (0.6 [95% CI, -0.5 to 0.7]) were found. However, this trend resulted in an additional 1.5 (95% CI, 0.3-2.6) abortions per 1000 females by the first quarter of 2020 compared with premifepristone trends; rates increased more among those aged 15 to 19 years, less among those aged 35 to 44 years, and did not increase for those aged 25 to 29 years. During the pandemic period, abortion rates decreased by 1.2 (95% CI, -2.5 to -0.8), most pronounced among those aged 20 to 34 years. Compared with expected rates based on premifepristone trends, 5-year availability of normally prescribed mifepristone was associated with a rate difference of 1.9 (95% CI, 0.7-5.4) in 2022, with a greater increase among those aged 20 to 24 years (4.2 [95% CI, 1.5-9.0]) and no change among those aged 25 to 29 years (1.0 [95% CI, -1.7 to 6.2]). The increased abortion rate in 2022 was consistent with 5-year trends following normally prescribed mifepristone, although social forces potentially impacting international rates may have contributed.

CONCLUSIONS AND RELEVANCE: This study found that, following longstanding declines, abortion rates in Ontario gradually increased with mifepristone availability in 2017 in Ontario. Following a pandemic-related decrease in rates (in 2020 and 2021), substantial increases in abortion rates reported elsewhere from 2020 to 2023 did not occur in Ontario as of 2022, suggesting that Ontario’s health services environment and Canada’s regulatory and policy approach to preserving reproductive health services may have helped stabilize abortion rates. Future research is needed to understand how sociocultural changes affecting abortion service use elsewhere may be affecting contraception access and use and thus abortion rates in Canada.

PMID:40214991 | DOI:10.1001/jamanetworkopen.2025.4516

JAMA Netw Open. 2025 Apr 1;8(4):e254666. doi: 10.1001/jamanetworkopen.2025.4666.

ABSTRACT

IMPORTANCE: A resurgence of respiratory syncytial virus (RSV)-associated acute respiratory tract infection (ARI) was observed in 2022 and 2023 after the COVID-19 pandemic. Changes in the demographic characteristics, disease severity, and outcomes of patients were observed, which could impact the identification of risk groups for interventions aimed at reducing the severity of RSV disease.

OBJECTIVES: To identify factors associated with severe clinical outcomes among children hospitalized with RSV-associated ARIs in 2022 and 2023.

DESIGN, SETTING, AND PARTICIPANTS: This observational cohort study, conducted at 2 large, Canadian, tertiary-level pediatric hospitals, comprised all 709 cases of RSV-associated ARI among children younger than 18 years who were admitted to the hospital or intensive care unit (ICU) from July 1, 2022, to June 30, 2023.

EXPOSURE: Diagnosis of RSV-associated ARI.

MAIN OUTCOMES AND MEASURES: The primary outcome of severe disease was defined as requiring noninvasive or invasive ventilation or death. Risk factors for severe disease and ICU admission (secondary outcome) were assessed using multivariable Poisson regression, and results were reported as adjusted risk ratios (ARRs) with 95% CIs, with age-stratified models (<2 years and ≥2 years).

RESULTS: A total of 709 cases (median age, 13.1 months [IQR, 2.0-36.6 months]; 442 boys [62.3%]) were admitted with RSV-associated ARI; 452 (63.8%) were younger than 2 years, and 257 (36.2%) were aged 2 years or older. Severe disease was documented for 204 cases (28.8%). Patients with severe disease were younger than those with nonsevere disease (median age, 2.6 months [IQR, 1.3-16.0 months] vs 18.6 months [IQR, 4.5-39.1 months]; P < .001). Pulmonary disease and use of home oxygen (ARR, 2.47 [95% CI, 1.30-4.68]) and neurologic, neuromuscular, and developmental conditions (ARR, 1.89 [95% CI, 1.03-3.49]) were associated with severe disease among children aged 2 years or older. Among children younger than 2 years, age younger than 3 months (ARR, 2.34 [95% CI, 1.43-3.84]), age 3 to less than 6 months (ARR, 2.79 [95% CI, 1.65-4.70]), and prematurity (ARR, 1.40 [95% CI, 1.03-1.89]) were associated with severe disease.

CONCLUSIONS AND RELEVANCE: In this cohort study of children hospitalized with RSV in 2022 and 2023, severe RSV disease was more likely among those aged 2 years or older with pulmonary and neurologic, neuromuscular, or developmental conditions. For children younger than 2 years, age younger than 6 months and prematurity were the main risk factors. These findings support prevention strategies for all younger children, including premature infants, with potential benefit for children aged 2 years or older in specific high-risk groups.

PMID:40214990 | DOI:10.1001/jamanetworkopen.2025.4666

JAMA Netw Open. 2025 Apr 1;8(4):e254689. doi: 10.1001/jamanetworkopen.2025.4689.

ABSTRACT

IMPORTANCE: Variations in perioperative dexamethasone dosing are common in brain metastasis resection, but their impact on patient outcomes remains unclear.

OBJECTIVE: To evaluate the association between perioperative dexamethasone dosing and patient outcomes, focusing on overall survival (OS) and progression-free survival (PFS).

DESIGN, SETTING, AND PARTICIPANTS: This retrospective multicenter comparative effectiveness study used data collected from January 2010 to December 2023. Patients with symptomatic brain metastases undergoing primary surgical resection at 7 neurological centers in Germany and 1 in Austria and who had complete records of perioperative dexamethasone dosing were included. Propensity score matching (PSM) was used to control for confounders. Analysis was conducted from March to June 2024.

EXPOSURES: Cumulative perioperative dexamethasone administration over 27 days, dichotomized at 122 mg using maximally selected rank statistics.

MAIN OUTCOMES AND MEASURES: The primary outcome was OS. Secondary outcomes included extracranial PFS (ecPFS) and intracranial PFS (icPFS) as well as incidence of wound revision surgery after brain metastasis resection. Hazard ratios (HRs) were calculated using Cox proportional hazards models.

RESULTS: A total of 1064 patients were included in the analysis. The median (IQR) age was 64 (56-72) years, with 489 female patients (49%) and 541 male patients (51%). Non-small cell lung cancer (NSCLC) was the most common tumor entity (564 patients [53%]), followed by breast cancer (146 patients [14%]) and melanoma (138 patients [13%]). After PSM, patients receiving cumulative dexamethasone doses less than 122 mg had a median OS of 19.1 (95% CI, 15.2-22.4) months compared with 12.0 (95% CI, 9.1-14.7) months for those receiving 122 mg or more (P = .002). Multivariable analysis showed an independent association between higher cumulative dexamethasone doses and reduced OS (HR, 1.40; 95% CI, 1.18-1.66; P < .001). Secondary analyses demonstrated consistent findings with icPFS and ecPFS and a dose-response association between cumulative dexamethasone and hazard for death.

CONCLUSIONS AND RELEVANCE: In this study, higher cumulative perioperative dexamethasone was associated with reduced OS, icPFS, and ecPFS in patients undergoing brain metastasis resection. These findings suggest that stricter dosing protocols could improve outcomes. Prospective trials are warranted to confirm these associations and guide evidence-based practice.

PMID:40214989 | DOI:10.1001/jamanetworkopen.2025.4689

JAMA Netw Open. 2025 Apr 1;8(4):e254729. doi: 10.1001/jamanetworkopen.2025.4729.

ABSTRACT

IMPORTANCE: Adults with functional disabilities require more medical care, but it remains unclear whether they use more health services, including high- and low-value services.

OBJECTIVES: To examine health care utilization by functional disability among US adults.

DESIGN, SETTING, AND PARTICIPANTS: This cross-sectional study analyzed data from the 2013 to 2021 Medical Expenditure Panel Survey. The sample comprised noninstitutionalized US civilians aged 18 years or older. Statistical analysis was conducted between May and October 2024.

EXPOSURES: Self-reported functional disability. Functional disability was assessed through 6 questions on difficulties (with vision, hearing, memory or concentration, walking, self-care, and performing errands) and categorized as no (0 difficulties), moderate (1-2 difficulties), and severe (≥3 difficulties).

MAIN OUTCOMES AND MEASURES: Outpatient visits, prescription drug fills, 10 high-value services, and 12 low-value services.

RESULTS: The sample comprised 188 954 adults (mean [SD] age, 48.1 [17.9] years; 101 706 females [53.8%]). Of these adults, 151 562 (80.2%) had no, 28 518 (15.0%) had moderate, and 8874 (4.6%) had severe functional disabilities. Adults with functional disabilities, especially those with severe disabilities, had a higher percentage of outpatient visits (86.2% vs 74.9%) and prescriptions filled (81.2% vs 64.2%) compared with those with no disabilities. The mean number of outpatient visits and prescription drug fills was significantly higher among those with severe vs no or moderate functional disabilities (outpatient visits: 17.5 [95% CI, 16.5-18.4] vs 8.6 [95% CI, 8.6-8.7] or 14.0 [95% CI, 13.8-14.3]; prescription drug fills: 27.8 [95% CI, 25.7-29.9] vs 10.6 [95% CI, 10.5-10.7] or 18.0 [95% CI, 17.6-18.4], respectively). Compared with adults with no functional disabilities, those with moderate and severe disabilities had higher rates of services that could be performed during an appointment, both high value (eg, adjusted differences, blood pressure measurement: 3.4 [95% CI, 2.9-3.9] percentage points and 3.6 [95% CI, 2.9-4.2] percentage points; cholesterol measurement: 3.6 [95% CI, 2.6-4.5] percentage points and 4.7 [95% CI, 3.6-5.7] percentage points, respectively) and low value (eg, adjusted differences, benzodiazepine for depression: 4.5 [95% CI, 2.5-6.4] percentage points and 8.1 [95% CI, 6.3-9.8] percentage points; opioid for back pain: 4.5 [95% CI, 3.5-5.5] percentage points and 6.7 [95% CI, 6.5-6.9] percentage points, respectively). Conversely, those with moderate and severe disabilities used fewer services that typically required a separate appointment, such as high-value cancer screenings (eg, adjusted differences, breast: -1.1 [95% CI, -1.3 to -0.9] percentage points and -9.9 [95% CI, -12.1 to -7.7] percentage points; cervical: -3.3 [95% CI, -4.9 to -1.7] percentage points and -17.3 [95% CI, -20.3 to -14.4] percentage points, respectively) and low-value cancer screenings (eg, adjusted differences, cervical: -4.9 [95% CI, -7.7 to -2.1] percentage points and -8.1 [95% CI, -12.1 to -4.0] percentage points, respectively).

CONCLUSIONS AND RELEVANCE: In this cross-sectional study, adults with functional disabilities used higher rates of health services than adults with no functional disabilities. However, the ease of access to services-independent of clinical value-plays an important role in utilization for those with functional disabilities.

PMID:40214987 | DOI:10.1001/jamanetworkopen.2025.4729

PNAS Nexus. 2025 Apr 2;4(4):pgaf107. doi: 10.1093/pnasnexus/pgaf107. eCollection 2025 Apr.

ABSTRACT

Is AI disrupting jobs and creating unemployment? This question has stirred public concern for job stability and motivated studies assessing occupations’ automation risk. These studies used readily available employment and wage statistics to quantify occupational changes for employed workers. However, they did not directly examine unemployment dynamics primarily due to the lack of data across occupations, geography, and time. Here, we overcome this barrier using monthly occupation-level unemployment data from each US state’s unemployment insurance office from 2010 to 2020 to assess AI exposure models, job separations, and unemployment through a new measure called unemployment risk. We demonstrate that standard employment statistics are inadequate proxies for occupations’ unemployment risk and find that individual AI exposure models are poor predictors of occupations’ unemployment risk states’ total unemployment rates, and states’ total job separation rates. However, an ensemble approach exhibits substantial predictive power, accounting for an additional 18% of variation in unemployment risk across occupations, states, and time compared to a baseline model that controls for education, occupations’ skills, seasonality, and regional effects. These results suggest that competing models may capture different aspects of AI exposure and that automation shapes US unemployment. Our results demonstrate the power of occupation-specific job disruption data and that efforts using only one AI exposure score will misrepresent AI’s impact on the future of work.

PMID:40213807 | PMC:PMC11983276 | DOI:10.1093/pnasnexus/pgaf107