Search

[This corrects the article DOI: 10.1371/journal.pone.0271610.].

BackgroundApproaching epidemiological data with flexible machine learning algorithms is of great value for understanding disease-specific association patterns. However, it can be difficult to correctly extract and understand those patterns due to the lack of model interpretability.MethodWe here propose a machine learning workflow that combines random forests with Bayesian network surrogate models to allow for a deeper level of interpretation of complex association patterns. We first evaluate the proposed workflow on synthetic data. We then apply it to data from the large population-based Study of Health in Pomerania (SHIP). Based on this combination, we discover and interpret broad patterns of individual serum TSH concentrations, an important marker of thyroid functionality.ResultsEvaluations using simulated data show that feature associations can be correctly recovered by combining random forests and Bayesian networks. The presented model achieves predictive accuracy that is similar to state-of-the-art models (root mean square error of 0.66, mean absolute error of 0.55, coefficient of determination of R2 = 0.15). We identify 62 relevant features from the final random forest model, ranging from general health variables over dietary and genetic factors to physiological, hematological and hemostasis parameters. The Bayesian network model is used to put these features into context and make the black-box random forest model more understandable.ConclusionWe demonstrate that the combination of random forest and Bayesian network analysis is helpful to reveal and interpret broad association patterns of individual TSH concentrations. The discovered patterns are in line with state-of-the-art literature. They may be useful for future thyroid research and improved dosing of therapeutics.

BackgroundAs healthcare-related data proliferate, there is need to annotate them expertly for the purposes of personalized medicine. Crowdworking is an alternative to expensive expert labour. Annotation corresponds to diagnosis, so comparing unlabeled records to labeled ones seems more appropriate for crowdworkers without medical expertise. We modeled the comparison of a record to two other records as a triplet annotation task, and we conducted an experiment to investigate to what extend sensor-measured stress, task duration, uncertainty of the annotators and agreement among the annotators could predict annotation correctness.Materials and methodsWe conducted an annotation experiment on health data from a population-based study. The triplet annotation task was to decide whether an individual was more similar to a healthy one or to one with a given disorder. We used hepatic steatosis as example disorder, and described the individuals with 10 pre-selected characteristics related to this disorder. We recorded task duration, electro-dermal activity as stress indicator, and uncertainty as stated by the experiment participants (n = 29 non-experts and three experts) for 30 triplets. We built an Artificial Similarity-Based Annotator (ASBA) and compared its correctness and uncertainty to that of the experiment participants.ResultsWe found no correlation between correctness and either of stated uncertainty, stress and task duration. Annotator agreement has not been predictive either. Notably, for some tasks, annotators agreed unanimously on an incorrect annotation. When controlling for Triplet ID, we identified significant correlations, indicating that correctness, stress levels and annotation duration depend on the task itself. Average correctness among the experiment participants was slightly lower than achieved by ASBA. Triplet annotation turned to be similarly difficult for experts as for non-experts.ConclusionOur lab experiment indicates that the task of triplet annotation must be prepared cautiously if delegated to crowdworkers. Neither certainty nor agreement among annotators should be assumed to imply correct annotation, because annotators may misjudge difficult tasks as easy and agree on incorrect annotations. Further research is needed to improve visualizations for complex tasks, to judiciously decide how much information to provide, Out-of-the-lab experiments in crowdworker setting are needed to identify appropriate designs of a human-annotation task, and to assess under what circumstances non-human annotation should be preferred.

PURPOSE:Exophthalmometry is a common examination in ophthalmology. For example it is relevant for diagnosis or follow-up of thyroid eye disease. However, exophthalmometry is affected by several factors such as ethnicity, sex and age. The purpose of this study was to determine the globe position by magnetic resonance imaging (MRI) and to investigate its correlates among the general Northeast German adult population. METHODS:A total of 3030 subjects aged between 20 and 89 from the population-based Study of Health in Pomerania (SHIP) underwent a standardised whole-body MRI. Axial length and globe position were determined in axial T1-weighted images of the orbit. The image had to include the corneal apex as well as the optic nerve head. Study participants were excluded from imaging analysis if there was no plane available that included both structures. Further exclusion criterion was a lateral deviation of the subject's viewing direction. Images with inadequate quality due to motion artefacts or other technical reasons were excluded as well. Globe position was defined as the perpendicular distance between the interzygomatic line and the posterior surface of the cornea (exophthalmometric value). The distance between the posterior surface of the cornea and the posterior pole of the eyeball, at the boundary with orbital fat, was defined as axial length. We used posterior surface of the cornea for our measurements, because it seemed to be less vulnerable for motion artefacts than the anterior one. Moreover body measurements including body mass index (BMI), waist and hip circumferences were determined. Associations between anthropometric measurements with exophthalmometric outcomes were analysed by linear regressions adjusted for age and stratified by sex. P-values

PURPOSE AND METHODS:A meta-analysis using data from seven German population-based cohorts was performed by the German Epidemiological consortium of Peripheral Arterial Disease (GEPArD) to investigate whether one question about claudication is more efficient for PAD screening than established questionnaires. Claudication was defined on the basis of the answer to one question asking for pain in the leg during normal walking. This simple question was compared with established questionnaires, including the Edinburgh questionnaire. The associations of claudication with continuous ABI values and decreased ABI were analyzed by linear and logistic regression analysis, respectively. The results of the studies were pooled in a random effect meta-analysis, which included data from 27,945 individuals (14,052 women, age range 20-84 years). RESULTS:Meta-analysis revealed a significant negative association between claudication and ABI, which was stronger in men (β = -0.07; 95%CI -0.10, -0.04) than in women (β = -0.02; 95%CI -0.02, -0.01). Likewise, the presence of claudication symptoms was related to an increased odds of a decreased ABI in both men (Odds ratio = 5.40; 95%CI 4.20, 6.96) and women (Odds ratio = 1.99; 95%CI 1.58, 2.51). CONCLUSIONS:Asking only one question about claudication was able to identify many individuals with a high likelihood of a reduced ABI with markedly higher sensitivity and only slightly reduced specificity compared to more complex questionnaires. At least in men, this question should be established as first screening step.

ObjectivesIt is still not clear why impaction of third molars occurs. Craniofacial morphology and facial parameters have been discussed to be strong predictors for third molar impaction. Thus, this study aimed to investigate the effect of craniofacial morphology on erupted or impacted third molars in a German population sample.Materials and methodsErupted and impacted third molars in 2,484 participants from the Study of Health in Pomerania were assessed by whole-body magnetic resonance imaging. Markers of facial morphology were determined in 619 individuals of those participants in whose 421 participants (16.7%) had at least one impacted third molar. Craniofacial morphology was estimated as linear measurements and was associated in a cross-sectional study design with impacted and erupted third molars by multinomial logistic regression models. Erupted third molars were used as reference outcome category and regression models were adjusted for age and sex.ResultsMaximum Cranial Width (Eurion-Eurion distance) was significantly associated with impacted third molars (RR: 1.079; 95% confidence interval 1.028-1.132). This association was even more pronounced in the mandible. Individuals with a lower total anterior facial height (Nasion-Menton distance) and a lower facial index also have an increased risk for impacted third molars in the mandible (RR 0.953; 95% confidence interval 0.913-0.996 and RR: 0.943; 95% confidence interval 0.894-0.995). No significant associations of third molar status with facial width (Zygion-Zygion distance), and sagittal cranial dimension (Nasion-Sella distance; Sella-Basion distance) were observed.ConclusionIndividuals with an increased maximal cranial width have a higher risk for impaction of third molars in the mandible and in the maxilla. Individuals with a lower anterior total anterior facial height and lower facial index also have an increased risk for third molars impaction in the mandible. These findings could help orthodontic dentists, oral surgeons and oral and maxillofacial surgeons in decision-making for third molars removal in their treatment. These findings highlight the necessity of an additional analysis of the maximal cranial width by the Eurion- Eurion distance.

BackgroundLittle is known about outpatient health services use following critical illness and intensive care. We examined the association of intensive care with outpatient consultations and quality of life in a population-based sample.MethodsCross-sectional analysis of data from 6,686 participants of the Study of Health in Pomerania (SHIP), which consists of two independent population-based cohorts. Statistical modeling was done using Poisson regression, negative binomial and generalized linear models for consultations, and a fractional response model for quality of life (EQ-5D-3L index value), with results expressed as prevalence ratios (PR) or percent change (PC). Entropy balancing was used to adjust for observed confounding.ResultsICU treatment in the previous year was reported by 139 of 6,686 (2,1%) participants, and was associated with a higher probability (PR 1.05 [CI:1.03;1.07]), number (PC +58.0% [CI:22.8;103.2]) and costs (PC +64.1% [CI:32.0;103.9]) of annual outpatient consultations, as well as with a higher number of medications (PC +37.8% [CI:17.7;61.5]). Participants with ICU treatment were more likely to visit a specialist (PR 1.13 [CI:1.09; 1.16]), specifically internal medicine (PR 1.67 [CI:1.45;1.92]), surgery (PR 2.42 [CI:1.92;3.05]), psychiatry (PR 2.25 [CI:1.30;3.90]), and orthopedics (PR 1.54 [CI:1.11;2.14]). There was no significant effect regarding general practitioner consultations. ICU treatment was also associated with lower health-related quality of life (EQ-5D index value: PC -13.7% [CI:-27.0;-0.3]). Furthermore, quality of life was inversely associated with outpatient consultations in the previous month, more so for participants with ICU treatment.ConclusionsOur findings suggest that ICU treatment is associated with an increased utilization of outpatient specialist services, higher medication intake, and impaired quality of life.

BackgroundPlatelet count is known to be associated with sex, age and mean platelet volume (MPV). Sex and age were proposed for adjustment of platelet count reference intervals, but MPV is currently not used for further adjustment. We investigated the association of MPV, age and sex with platelet counts and established individualized reference ranges respecting MPV.MethodsThe association of platelet count with age, sex and MPV was assessed in healthy participants (n = 3,033 individuals; 1,542 women) in the cross-sectional population-based cohort Study of Health in Pomerania. Reference intervals respecting age, sex, and MPV were estimated using quantile regressions for the 2.5th and 97.5th percentile.ResultsWomen had higher platelet counts than men (239 vs. 207 x109/L, pConclusionMPV and sex have a stronger association with platelet count than age. MPV should be considered to adjust platelet count reference intervals and needs to be respected as confounder for platelet counts in epidemiological studies and clinical practice.

Pneumothoraces are the most frequently occurring complications of CT-guided percutaneous transthoracic pulmonary biopsies (PTPB). The aim of this study was to evaluate the influence of pre-diagnostic lung emphysema on the incidence and extent of pneumothoraces and to establish a risk stratification for the evaluation of the pre-procedure complication probability.CT-guided PTPB of 100 pre-selected patients (mean age 67.1±12.8 years) were retrospectively enrolled from a single center database of 235 PTPB performed between 2012-2014. Patients were grouped according to pneumothorax appearance directly after PTPB (group I: without pneumothorax, n = 50; group II: with pneumothorax, n = 50). Group II was further divided according to post-interventional treatment (group IIa: chest tube placement, n = 24; group IIb: conservative therapy, n = 26). For each patient pre-diagnostic percentage of emphysema was quantified using CT density analysis. Emphysema stages were compared between groups using bivariate analyses and multinomial logistic regression analyses.Emphysema percentage was significantly associated with the occurrence of post-interventional pneumothorax (p = 0.006). Adjusted for potential confounders (age, gender, lesion size and length of interventional pathway) the study yielded an OR of 1.07 (p = 0.042). Absolute risk of pneumothorax increased from 43.4% at an emphysema rate of 5% to 73.8% at 25%. No differences could be seen in patients with pneumothorax between percentage of emphysema and mode of therapy (p = 0.721).The rate of lung emphysema is proportionally related to the incidence of pneumothorax after CT-guided PTPB and allows pre-interventional risk stratification. There is no association between stage of emphysema and post-interventional requirement of chest tube placement.

Exercise and statins reduce cardiovascular disease (CVD). Exercise capacity may be assessed using cardiopulmonary exercise testing (CPET). Whether statin medication is associated with CPET parameters is unclear. We investigated if statins are related with exercise capacity during CPET in the general population.Cross-sectional data of two independent cohorts of the Study of Health in Pomerania (SHIP) were merged (n = 3,500; 50% males). Oxygen consumption (VO2) at peak exercise (VO2peak) and anaerobic threshold (VO2@AT) was assessed during symptom-limited CPET. Two linear regression models related VO2peak with statin usage were calculated. Model 1 adjusted for age, sex, previous myocardial infarction, and physical inactivity and model 2 additionally for body mass index, smoking, hypertension, diabetes and estimated glomerular filtration rate. Propensity score matching was used for validation.Statin usage was associated with lower VO2peak (no statin: 2336; 95%-confidence interval [CI]: 2287-2,385 vs. statin 2090; 95%-CI: 2,031-2149 ml/min; P < .0001) and VO2@AT (no statin: 1,172; 95%-CI: 1,142-1,202 vs. statin: 1,111; 95%-CI: 1,075-1,147 ml/min; P = .0061) in males but not females (VO2peak: no statin: 1,467; 95%-CI: 1,417-1,517 vs. statin: 1,503; 95%-CI: 1,426-1,579 ml/min; P = 1.00 and VO2@AT: no statin: 854; 95%-CI: 824-885 vs. statin 864; 95%-CI: 817-911 ml/min; P = 1.00). Model 2 revealed similar results. Propensity scores analysis confirmed the results.In the general population present statin medication was related with impaired exercise capacity in males but not females. Sex specific effects of statins on cardiopulmonary exercise capacity deserve further research.