Search PubMed⌕ Search

SEARCH · Search PubMed

Results for “Explainability”

Search indexed PubMed citations on genomics, clinical trials, systematic reviews and public health. Explore titles, authors and supplied subject terms, then open the PubMed record.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 73 records · Page 4Linked to original sources
PubMed1998

Behavioral antagonism in the moth Helicoverpa zea in response to pheromone blends of three sympatric heliothine moth species is explained by one type of antennal neuron.

We have discovered a type of sensillum on the antennae of male corn earworm moths, Helicoverpa zea, that houses two types of receptor neuron (RN) that explains both the upwind flight attraction of males to their own species' pheromone blend and arrestment to the quite similar blends emitted by females of three other sympatric North American heliothine species. The first RN type is a large-spiking neuron that is most sensitive to (Z)-9-hexadecenal (Z9-16:Ald), the secondary H. zea pheromone component that along with the major component, (Z)-11-hexadecenal, causes attraction to the female. This RN is also responsive to (Z)-9-tetradecenal (Z9-14:Ald; not a H. zea pheromone component) at higher dosages. The sensitivity of this RN thus explains the attraction that has been observed in other studies when small proportions of Z9-14:Ald are added to Z11-16:Ald to mimic the conspecific blend. The second type of RN in this sensillum is a small-spiking neuron that is again responsive to Z9-14:Ald (which in larger proportions acts as a strong antagonist to upwind flight), but this RN is actually more sensitive to two other strong behavioral antagonists, (Z)-11-hexadecenyl acetate and (Z)-11-hexadecenol. Thus, activation of this single broadly tuned 'antagonist' RN could explain why H. zea males will orient only to their conspecific females. These three compounds are emitted by females of three other North American species, H. subflexa, H. phloxiphaga, and H. virescens, as agonists in their blends, which also contain the H. zea components Z11-16:Ald and Z9-16:Ald. This antagonist RN may also explain why a blend of Z11-16:Ald and a small amount of Z9-14:Ald is never as attractive to H. zea males as the conspecific blend. Enhanced specificity for the conspecific blend arises because the antagonist RN is never stimulated to fire, even when large proportions of the pheromone component, Z9-16:Ald, are added to Z11-16:Ald. When Z9-14:Ald is used instead of Z9-16:Ald, however, and the proportion of Z9-14:Ald becomes too great, the threshold of the antagonist neuron as well as that of the agonist neuron is exceeded, and the upwind flight response begins to be suppressed.

Aldehydes↗
PubMed1994

Using patient and general practice characteristics to explain variations in cervical smear uptake rates.

OBJECTIVES: To produce practice and patient variables for general practices from census and family health services authority data, and to determine the importance of these variables in explaining variation in cervical smear uptake rates between practices. DESIGN: Population based study examining variations in cervical smear uptake rates among 126 general practices using routine data. SETTING: Merton, Sutton, and Wandsworth Family Health Services Authority, which covers parts of inner and outer London. MAIN OUTCOME MEASURE: Percentage of women aged 25-64 years registered with a general practitioner who had undergone a cervical smear test during the five and a half years preceding 31 March 1992. RESULTS: Cervical smear uptake rates varied from 16.5% to 94.1%. The estimated percentage of practice population from ethnic minority groups correlated negatively with uptake rates (r = -0.42), as did variables associated with social deprivation such as overcrowding (r = -0.42), not owning a car (r = -0.41), and unemployment (r = -0.40). Percentage of practice population under 5 years of age correlated positively with uptake rate (r = 0.42). Rates were higher in practices with a female partner than in those without (66.6% v 49.1%; difference 17.5% (95% confidence interval 10.5% to 24.5%)), and in computerised than in non-computerised practices (64.5% v 50.5%; 14.0% (6.4% to 21.6%)). Rates were higher in larger practices. In a stepwise multiple regression model that explained 52% of variation, five factors were significant predictors of uptake rates: presence of a female partner; children under 5; overcrowding; number of women aged 35-44 as percentage of all women aged 25-64; change of address in past year. CONCLUSIONS: Over half of variation in cervical smear uptake rates can be explained by patient and practice variables derived from census and family health services authority data; these variables may have a role in explaining variations in performance of general practices and in producing adjusted measures of practice performance. Practices with a female partner had substantially higher uptake rates.

Adult↗
PubMed1998

Do changes in cardiovascular risk factors explain the increasing socioeconomic difference in mortality from ischaemic heart disease in Finland?

STUDY OBJECTIVE: To estimate the extent to which changes in blood pressure, smoking, and serum cholesterol concentration explain the observed increase in socioeconomic differences in mortality from ischaemic heart disease (IHD) in Finland during the past 20 years. DESIGN: Predicted changes in mortality from IHD were calculated using logistic regression models with the risk factor levels assessed by cross sectional population surveys conducted in 1972, 1977, 1982, and 1987. The subjects included white collar and blue collar workers and farmers. The predicted changes were compared with the observed mortality changes in the same socioeconomic groups in the total population of the same geographical area. SETTING: North Karelia and Kuopio provinces, eastern Finland. PARTICIPANTS: 16,741 men and 16,389 women aged 30-59 randomly drawn from the population registers of the study areas. Mortality data were obtained from the total population in the same areas. MAIN RESULTS: In men, the changes in diastolic blood pressure, total serum cholesterol, and smoking predicted a 28% decline in the mortality from IHD among white collar workers, a 30% decline among blue collar workers, and a 33% decline in farmers. Observed declines in the same socioeconomic groups were 61%, 40%, and 37%, respectively. In women, the predicted decline was 41% among white collar workers, 35% among blue collar workers, and 39% among farmers. The respective observed declines were 57%, 43%, and 20%. CONCLUSIONS: Less than half of the decline in IHD mortality among white collar men was explained by the risk factor changes, while they explained 75% of the decline among blue collar men and 89% of the decline among male farmers. Changes in risk factors did not explain the increasing difference in IHD mortality between the socioeconomic groups, especially among men.

Adult↗
PubMed1999

Do morphology and stage explain the inferior lung cancer survival in Denmark?

Danish lung cancer patients diagnosed during 1983-1987 experienced 5-yr relative survival rates 2-7% inferior to patients in the other Nordic countries, despite the similarity of cancer registration and healthcare systems in the Nordic countries. Is the inferior relative survival in Denmark due to differences in morphology or stage of lung cancers? The present study compared in detail the survival of 92,719 patients diagnosed with lung cancer during 1978-1992 in Denmark, Finland, and Norway. In particular, differences in morphology and extent of disease were studied. A poor survival rate for small cell anaplastic lung carcinoma compared with all other morphologies was confirmed. However, this could not explain the relative survival differences observed between countries. Extent of disease was the most important predictor of survival. Part of the observed survival differences could be explained by a less favourable stage distribution in Denmark, combined with a slightly lower relative survival rate for those with metastatic disease. Differences in treatment are unlikely to explain the findings, although delays in diagnosing and treating patients in Denmark compared with neighbouring countries could partially explain the lower patient survival in Denmark. In conclusion, the main factor in the lower survival rate in Denmark is unfavourable stage distribution.

Aged↗
PubMed2002

Factors explaining variance in perceived pain in women with fibromyalgia.

BACKGROUND: We hypothesized that a substantial proportion of the subjectively experienced variance in pain in fibromyalgia patients would be explained by psychological factors alone, but that a combined model, including neuroendocrine and autonomic factors, would give the most parsimonious explanation of variance in pain. METHODS: Psychometric assessment included McGill Pain Questionnaire, General Health Questionnaire, Hospital Anxiety and Depression Rating Scale, Eysenck personality Inventory, Neuroticism and Lie subscales, Toronto Alexithymia Scale, and Multidimensional Health Locus of Control Scale and was performed in 42 female patients with fibromyalgia and 48 female age matched random sample population controls. A subgroup of the original sample (22 fibromyalgia patients and 13 controls) underwent a pharmacological challenge test with buspirone to assess autonomic and adrenocortical reactivity to serotonergic challenge. RESULTS: Although fibromyalgia patients scored high on neuroticism, anxiety, depression and general distress, only a minor part of variance in pain was explained by psychological factors alone. High pain score was associated with high neuroticism, low baseline cortisol level and small drop in systolic blood pressure after buspirone challenge test. This model explained 41.5% of total pain in fibromyalgia patients. In population controls, psychological factors alone were significant predictors for variance in pain. CONCLUSION: Fibromyalgia patients may have reduced reactivity in the central sympathetic system or perturbations in the sympathetic-parasympathetic balance. This study shows that a biopsychosocial model, including psychological factors as well as factors related to perturbations of the autonomic nervous system and hypothalamic-pituitary-adrenal axis, is needed to explain perceived pain in fibromyalgia patients.

Journal Article↗
PubMed2002

Altered GH elimination kinetics in type 1 diabetes mellitus can explain the elevation in circulating levels: bicompartmental approach.

Type 1 diabetes mellitus (DM 1) is associated with elevated circulating GH concentrations. Because these high GH levels could be explained either by an augmented pituitary secretion and/or delayed elimination clearance or distribution, we sought to evaluate GH pharmacokinetics to propose a model that better explains the elimination kinetics in patients with DM 1 and assess possible differences with normal volunteers that could justify elevation in GH circulating levels in these patients. A multicompartmental analysis was applied to serum GH concentrations measured at different times for 150 min in six patients with DM 1 and six age-, sex-, and body mass index-matched normal subjects after the administration of an iv bolus of recombinant human GH (200 microg), previous suppression of endogenous GH release with octreotide. The best fitting to the GH disappearance profiles was obtained with the biexponential equation in both groups. From it, we propose a bicompartmental model to explain GH kinetics in normal and diabetic patients. The mean transit time in both compartments and the mean residence time in patients with DM 1 were more than twice the values from control group. So in DM 1 elevated circulating GH concentrations are, at least partially, caused by a delayed GH plasmatic clearance. The DM 1 patients included in this study had a normal renal function; thus, our results agree with the hypothesis that DM 1 constitutes a GH-insensitivity state because a reduced GH clearance by its receptor-mediated mechanism might explain the delayed GH elimination kinetics shown in patients with DM 1. However, the possibility of additional factors contributing to the slowed GH removal from circulation is not completely excluded.

Adult↗
PubMed2006

Does clinical presentation explain practice variability in the treatment of febrile infants?

BACKGROUND: Previous studies documented considerable variability in the treatment of febrile infants, despite the existence of practice guidelines for this condition. None of those studies documented the extent to which this variability is accounted for by differences in clinical severity. OBJECTIVE: To quantify the individual effects of the patient's clinical presentation, demographic, provider, and practice characteristics, and regional variables on practice variability in the evaluation and treatment of febrile infants. METHODS: With data collected through the Pediatric Research in Office Settings network, we analyzed data on the treatment of 2712 febrile infants examined by 484 pediatricians located in 194 practices. We analyzed hospitalization, lumbar puncture, urinalysis and/or urine culture, blood work, and initial antibiotic administration. We obtained a summary score for evaluation and treatment intensity (ranging from no tests or treatments to comprehensive testing, hospitalization, and antibiotic therapy) by performing principal-components analysis with these 5 variables. This summary score was regressed with respect to patients' clinical presentation, demographic and practice/practitioner features, and geographic region. Provider fixed effects were also included in the model. RESULTS: Although the overall model explained 46.5% of the variance, the clinical characteristics of the patient alone explained 29.7% of the overall variance. Practice site fixed effects explained nearly 15% of the overall variance. Provider and practitioner characteristics and geographic region had minimal explanatory power. CONCLUSIONS: Our results show that measures of the patient's clinical presentation account for nearly one third of the variability that our model explains. This suggests that differences in clinical presentation and severity of illness underlie much of the observed practice variability among pediatricians evaluating and treating febrile infants. These findings demonstrate that the management of this common and potentially serious condition depends more on the clinical presentation of the patient than on the characteristics of the provider/practice and the residential region.

Fever↗
PubMed2006

Which factors could explain the low birth weight paradox?

OBJECTIVE: Low birth weight children are unusual among well-off families. However, in Brazil, low birth weight rate was higher in a more developed city than in a less developed one. The study objective was to find out the reasons to explain this paradox. METHODS: A study was carried out in two municipalities, Ribeirão Preto (Southeastern Brazil) and São Luís (Northeastern Brazil), which low birth weight rates were 10.7% and 7.6% respectively. Data from two birth cohorts were analyzed: 2,839 newborns in Ribeirão Preto in 1994 and 2,439 births in São Luís in 1997-1998. Multiple logistic regression analysis was performed, adjusted for confounders. RESULTS: Low birth weight risk factors in São Luís were primiparity, maternal smoking and maternal age less than 18 years. In Ribeirão Preto, the associated variables were family income between one and three minimum wages, maternal age less than 18 and equal to or more than 35 years, maternal smoking and cesarean section. In a combined model including both cohorts, Ribeirão Preto presented a 45% higher risk of low birth weight than São Luís. When adjusted for maternal smoking habit, the excess risk for low birth weight in Ribeirão Preto compared to São Luís was reduced by 49%, but the confidence interval was marginally significant. Differences in cesarean section rates between both cities contributed to partially explain the paradox. CONCLUSIONS: Maternal smoking was the most important risk factor for explaining the difference in low birth weight between both cities. The other factors contributed little to explain the difference in low birth weight rates.

Brazil↗
PubMed1999

Explaining educational differences in mortality: the role of behavioral and material factors.

OBJECTIVES: This study examined the role of behavioral and material factors in explaining educational differences in all-cause mortality, taking into account the overlap between both types of factors. METHODS: Prospective data were used on 15,451 participants in a Dutch longitudinal study. Relative hazards of all-cause mortality by educational level were calculated before and after adjustment for behavioral factors (alcohol intake, smoking, body mass index, physical activity, dietary habits) and material factors (financial problems, neighborhood conditions, housing conditions, crowding, employment status, a proxy of income). RESULTS: Mortality was higher in lower educational groups. Four behavioral factors (alcohol, smoking, body mass index, physical activity) and 3 material factors (financial problems, employment status, income proxy) explained part of the educational differences in mortality. With the overlap between both types of factors accounted for, material factors were more important than behavioral factors in explaining mortality differences by educational level. CONCLUSIONS: The association between educational level and mortality can be largely explained by material factors. Thus, improving the material situation of people might substantially reduce educational differences in mortality.

Adolescent↗
PubMedDate not supplied

Factors which explain amount of participation in rural adolescent alcohol use prevention task forces.

Project Northland task force members are typically female, have children, and belong to a number of other community or social groups. Members were likely to participate more in the task forces if they were relative newcomers to the community and if they got satisfaction from their participation. While explaining only a modest amount of the variance (15.5%), the explanatory power of the results are comparable to other studies of this nature, where 23% of explained variance predicting participation has been considered informative. Future models will need to consider other factors to increase the explained variance. Satisfaction with being a task force member was associated with the amount of control and ownership each member felt and the extent of agreement with the direction of the task force. These variables explained 26% of the variance. It may be that relative newcomers to the community, who tended to participate more, have lived elsewhere and been exposed to environments where alcohol was less frequently used, and this contrast contributed to their motivation and preparedness to work towards change. They may not be as integrated into the community as task force members who have lived in their communities for their entire lives, and may not be part of the dominant "alcohol culture" that exists in the region. If membership in community or social groups is an indicator of integration into the community, then the results suggest less integrated task force members were more active participants.(ABSTRACT TRUNCATED AT 250 WORDS)

Adolescent↗
PubMed2005

Different mechanisms to explain the reversed effects of mental health on work characteristics.

OBJECTIVES: The number of longitudinal studies reporting evidence for reversed effects of strain on work is growing, but evidence regarding the mechanisms underlying such effects is scarce. In this study, earlier longitudinal findings were reviewed, and the following four mechanisms for reversed effects were proposed that reflect within-person or environmental changes: (i) the rosy perception mechanism, (ii) the gloomy perception mechanism, (iii) the upward selection mechanism, and (iv) the drift mechanism. METHODS: These mechanisms were tested using structural equation modeling and longitudinal data from a Dutch four-phase study (N=1588 participants). RESULTS: The results revealed that work characteristics and mental health influenced each other reciprocally and longitudinally. The reversed effects were examined in more detail, and it was found that these could be accounted for by both within-person and environmental change mechanisms. The rosy perception mechanism was found to explain the positive effects from health on job demands; the upward selection mechanism explained the positive (environmental) effects from health on job control; the gloomy perception mechanism explained the reversed (evaluation) effects from health on supervisory social support. No support was found for the drift mechanism. CONCLUSIONS: Mechanisms that may explain reversed causation are yet poorly understood. The main contribution of the present study lies in the fact that it proposes (i) a conceptual framework with which to analyze the effects of health on work characteristics and (ii) methods for testing these mechanisms. The study revealed that there is good reason to pursue research on reversed causality.

Adult↗
PubMed2001

Associations between adverse events in childhood and chronic widespread pain in adulthood: are they explained by differential recall?

OBJECTIVE: Clinic based studies suggest that adverse events in childhood may predispose to chronic pain in adult life. These have been conducted on highly selected groups, and it is unknown whether these relationships hold in the general population and to what extent the increased rate of adverse childhood events in persons with pain is an artefact of differential reporting. We examined the hypothesis that chronic widespread pain was associated with reports of adverse experiences in childhood and whether any observed relationships could be explained by differential recall. METHODS: A cross sectional population based screening survey was conducted. Subjects completed a questionnaire that included assessments of pain and psychological state. In total, 296 subjects who had demonstrated psychological distress were randomly selected and had a detailed interview, which included an assessment of 14 adverse childhood experiences. Medical records relating to childhood were also examined for those subjects. RESULTS: The prevalence of self-reported adverse childhood experiences was greatest in adult subjects with current chronic widespread pain. Exposure to illness in family members, parental loss, operations, and abuse were all associated with increased, but nonsignificant, odds of having chronic widespread pain versus those without such exposures. However the only statistically significant association was with childhood hospitalizations. From medical record information the associations of hospitalizations (OR 5.1, 95% CI 2.0-13.0) and operations (OR 3.0, 95% CI 1.2-7.2) with pain previously noted were partly explained by differential recall between subjects with and without pain: hospitalizations, OR 2.2, 95% CI 0.9-5.5; operations, OR 1.2, 95% CI 0.5-3.4. CONCLUSION: Although several reported adverse events in childhood were observed to be associated with chronic widespread pain in adulthood, only reports of hospitalizations were significantly associated. Validation of self-reported exposures suggests that there was differential recall of past events among those with and without pain, and this differential recall explained the association between hospitalizations and current chronic pain. Such differential recall may explain other observations of an association between reports of adverse childhood events and chronic pain in adulthood.

Adult↗
PubMed2004

Improving the diagnosis related grouping model's ability to explain length of stay of elderly medical inpatients by incorporating function-linked variables.

INTRODUCTION: This study first aimed to determine the adequacy of the Diagnosis Related Grouping (DRG) model's ability to explain (1) the variance in the actual length of stay (LOS) of elderly medical inpatients and (2) the LOS difference in the same cohort between the departments of Geriatric Medicine (GRM) and General Medicine (GM). We then looked at how these explanatory abilities of the DRG changed when patients' function-linked variables (ignored by DRG) were incorporated into the model. MATERIALS AND METHODS: Basic demographic data of a consecutively hospitalised cohort of elderly medical inpatients from GRM and GM, as well as their actual LOS, discharge DRG codes [with their corresponding trimmed average length of stay (ALOS)] and selected function-linked variables (including premorbid functional status, change in functional profile during hospitalisation and number of therapists seen) were recorded. Beginning with ALOS, function-linked variables that were significantly associated with LOS were then added into two multiple liner regression models so as to quantify how the functional dimension improved the DRGs' abilities to explain LOS variances and interdepartmental LOS differences. Forward selection procedure was employed to determine the final models. For the interdepartmental analysis, the study sample was restricted to patients who shared common DRG codes. RESULTS: 114 GRM and 118 GM patients were studied. Trimmed ALOS alone explained 8% of the actual LOS variance. With the addition of function-linked variables, the adjusted R2 of the final model increased to 28%. Due to common code restrictions, the data of 79 GRM and 78 GM patients were available for the analysis of interdepartmental LOS differences. At the unadjusted stage, the median stay of GRM patients was 4.3 days longer than GM's and with adjustments made for the DRGs, this difference was reduced to 3.9 days. Additionally adjusting for the patients' functional features diminished the interdepartmental LOS discrepancy even further, to 2.1 days. CONCLUSION: This study demonstrates that for elderly medical inpatients, the incorporation of patients' functional status significantly improves the DRG model's ability to predict the patients' actual LOS as well as to explain interdepartmental LOS differences between GRM and GM.

Age Factors↗
PubMed1995

Hospitalization for congestive heart failure. Explaining racial differences.

OBJECTIVE: To determine whether the higher rate of hospitalization among African Americans for congestive heart failure (CHF) could be explained by racial differences in the prevalence of clinical risk factors for CHF. DESIGN: Retrospective cohort study. SETTING: A large health maintenance organization (HMO). PATIENTS: A sample of 64,877 enrollees (27% African American and 73% white) of the Northern California Kaiser Permanente Medical Care Program who took at least one multiphasic health checkup (MHC) at or after the age of 40 years and were free of CHF at that time. MAIN OUTCOME MEASURES: First hospitalization with a principal diagnosis of CHF. RESULTS: Among cohort members younger than 60 years at baseline MHC, the age-adjusted risk ratio (RR) (African American/white) for CHF hospitalization was 2.14 for men and 2.73 for women, while for persons 60 years of age and older at MHC, the age-adjusted RR was 1.48 for both sexes. Cox proportional hazards models were used to adjust for risk factors and length of follow-up. In persons aged 60 years and older, the race difference was explained by greater prevalence of hypertension and diabetes in African Americans (RR = 1.12; 95% confidence interval [CI], 0.94 to 1.34 after adjustment for hypertension and diabetes). In those younger than 60 years, findings differed by sex. For men, African-American race was no longer a significant predictor of CHF after adjusting for hypertension, diabetes, left ventricular hypertrophy on electrocardiogram, and body mass index (adjusted RR = 1.16; 95% CI, 0.86 to 1.56). However, among younger women, African Americans continued at increased risk despite adjustment for these variables as well as smoking, plasma cholesterol, renal function, alcohol use, and myocardial infarction (adjusted RR = 1.49; 95% CI, 1.00 to 2.21). CONCLUSIONS: In this HMO population, the race differences in first hospitalization for CHF are largely explained by known clinical and behavioral risk factors, although in younger women these risk factors do not completely explain the excess risk among African Americans. These findings highlight the role of hypertension and diabetes in the development of CHF, particularly among African Americans.

Black People↗
PubMed1996

Abdominal fat distribution and insulin levels only partially explain adverse cardiovascular risk profile in Asian Indians.

BACKGROUND: Asian Indians show an increased risk of non-insulin-dependent diabetes mellitus (NIDDM) and coronary heart disease, together with adverse fat distribution and hyperinsulinaemia relative to other ethnic groups. Using population-based data, we investigated the question of whether the adverse fat distribution observed in Indians can be explained by differences in behavioural risk factor levels. We have examined the question of whether ethnic differences in fat distribution are responsible for the unfavourable risk factor profile of Indians. SUBJECTS AND METHODS: Fat distribution (waist: hip ratio) was compared in population-based samples of Asian Indian (n = 4394), Creole (n = 1746), and Chinese (n = 425) Mauritians, after controlling for body mass index and other factors. The contribution of this ratio to ethnic differences in cardiovascular disease risk factors and the role of fasting insulin concentrations were also determined. RESULTS: Indian men had the highest mean waist: hip ratio, despite having the lowest body mass index. In Indian women the mean waist: hip ratio and body mass index were intermediate between those of Chinese and Creole women. Indians of both sexes had low levels of high-density lipoprotein (HDL) cholesterol and HDL: total cholesterol compared with Creoles or Chinese, whereas triglycerides levels were highest in Indian men but intermediate in Indian women. Mean fasting and 2 h insulin concentrations were not consistently highest in the Indian subgroups. Blood pressure and serum urate levels were lowest in Indians of both sexes and Indian women also had lower total cholesterol concentrations than either Creoles or Chinese. The elevated waist: hip ratio in Indians was not explained by differences in physical activity, cigarette smoking, or alcohol consumption and the differences in this ratio (and insulin levels) did not explain the observed ethnic differences in metabolic parameters. CONCLUSIONS: The susceptibility of Indians to abdominal obesity contributed to the less desirable levels of some, but not all, cardiovascular risk factors. Ethnic differences in cardiovascular risk factors in Mauritians were not explained by differences in abdominal obesity, serum insulin or behavioural risk factors. These data suggest that it is over-simplistic to ascribe the adverse cardiovascular risk factor profile commonly observed in Asian Indians to a tendency to abdominal obesity.

Abdomen↗
PubMed2026

Data-centric, robust, and explainable multimodal deep learning for clinical decision support: A systematic review.

PURPOSE: Multimodal deep learning is increasingly proposed for clinical decision support (CDS) under a "data-centric" framing that prioritizes label quality, missing-modality robustness, distribution shift, calibration, and explainability. Prior reviews have examined multimodal medical AI, CDS, and data-centric methods separately, but none address their intersection. We mapped the modalities, fusion strategies, and data-centric and explainability techniques used in this recent literature, quantified how often each is implemented rather than merely mentioned, assessed deployment-relevant evidence (external validation, clinical-outcome measurement, equity), and formally appraised study-level risk of bias. METHODS: Following the PRISMA 2020 statement (PROSPERO CRD420261427815; registered retrospectively), we screened 150 records and included primary, clinical, multimodal studies that applied machine or deep learning to a decision-support task and reported at least one quantitative result. Two reviewers screened and extracted data with consensus adjudication. Each study was coded against pre-specified operational definitions, separating implemented or empirically evaluated techniques from those only mentioned. Study-level risk of bias was assessed with PROBAST + AI. Synthesis was narrative. RESULTS: Thirty-one studies met inclusion; 30 (97%) were published between 2024 and 2026, with a median of three modalities (range 2-6), most commonly structured EHR (71%) and imaging (39%). Data-centric techniques were frequently reported (74-84% across label-noise, distribution-shift, calibration, missing-modality and class-imbalance handling; equity 61%). However, external validation was reported in only 4/31 studies (13%), a clinical or provider outcome in 3/31 (10%), and no study reported routine deployment. Overall risk of bias was high in 27/31 studies (87%), driven by the analysis domain. CONCLUSION: Within this recent, self-selected slice of the field, technical robustness and explainability techniques are widely reported but rarely validated out-of-distribution or against clinical outcomes, and the underlying evidence is at high risk of bias. Progress requires external multi-site validation, clinical-outcome measurement, formal bias appraisal, and adherence to AI reporting standards (e.g., TRIPOD + AI) before deployment can be justified.

Deep Learning↗
PubMed1996

Explained variation in survival analysis.

Several measures of explained variation have been suggested for the Cox proportional hazards regression model. We have categorized these measures into three classes which correspond to three different definitions of multiple R2 of the general linear model. In an empirical study we compared the performance of these measures and classified them by their adherence to a set of criteria which we think should be met by a measure of explained variation for survival data. We suggest that currently there is no uniformly superior measure, particularly as the concepts of either uncensored or censored populations may lead to different choices. For uncensored populations, a measure by Kent and O'Quigley and the squared rank correlation between survival time and the predictor from a Cox regression model appear recommendable choices. For the latter, censored survival times are terminated using a very recent data augmentation algorithm for multiple imputation under proportional hazards. With censored populations, Schemper's measure, V2, could be considered. We give an introductory example, discuss aspects of application and stress the desirability of routinely evaluating explained variation in studies of survival.

Humans↗
PubMed2004

Fertility pattern does not explain social gradient in breast cancer in denmark.

The present study was undertaken to assess the impact of reproductive behavior on the social class gradient in breast cancer occurrence in Denmark. Objectives were to study whether the gradient across socioeconomic groups could be explained by fertility differences, whether the gradient across educational groups could be explained by fertility differences and whether the effect of socioeconomic group on breast cancer incidence and mortality could be explained by education and vice versa. We studied 674,084 women aged 20-39 at the census on 9 November 1970 for whom we had complete data on fertility history. The cohort was followed up for breast cancer incidence and mortality until 8 November 1998. Fertility history varied considerably across socioeconomic group, where 38% of the academics were childless at the age of 30, in contrast to only 8% of women in agriculture. The academics had the highest risk of breast cancer and women in agriculture had the lowest risk. For incidence, the gradient in the relative risks was 1.74, which changed to 1.49 when fertility history was incorporated and to 1.29 when school education was also taken into account. For school education, women with > or = 12 years of schooling had the highest risk and women with < or = 7 years of schooling had the lowest risk. For incidence, the gradient in the relative risk was 1.38, which changed to 1.26 when fertility history was incorporated and to 1.22 when socioeconomic group was also taken into account.

Adult↗