Fusing data mining, machine learning and traditional statistics to detect biomarkers associated with depression

BACKGROUND: Atheoretical large-scale data mining techniques using machine learning algorithms have promise in the analysis of large epidemiological datasets. This study illustrates the use of a hybrid methodology for variable selection that took account of missing data and complex survey design to identify key biomarkers associated with depression from a large epidemiological study.

METHODS: The study used a three-step methodology amalgamating multiple imputation, a machine learning boosted regression algorithm and logistic regression, to identify key biomarkers associated with depression in the National Health and Nutrition Examination Study (2009-2010). Depression was measured using the Patient Health Questionnaire-9 and 67 biomarkers were analysed. Covariates in this study included gender, age, race, smoking, food security, Poverty Income Ratio, Body Mass Index, physical activity, alcohol use, medical conditions and medications. The final imputed weighted multiple logistic regression model included possible confounders and moderators.

RESULTS: After the creation of 20 imputation data sets from multiple chained regression sequences, machine learning boosted regression initially identified 21 biomarkers associated with depression. Using traditional logistic regression methods, including controlling for possible confounders and moderators, a final set of three biomarkers were selected. The final three biomarkers from the novel hybrid variable selection methodology were red cell distribution width (OR 1.15; 95% CI 1.01, 1.30), serum glucose (OR 1.01; 95% CI 1.00, 1.01) and total bilirubin (OR 0.12; 95% CI 0.05, 0.28). Significant interactions were found between total bilirubin with Mexican American/Hispanic group (p = 0.016), and current smokers (p<0.001).

CONCLUSION: The systematic use of a hybrid methodology for variable selection, fusing data mining techniques using a machine learning algorithm with traditional statistical modelling, accounted for missing data and complex survey sampling methodology and was demonstrated to be a useful tool for detecting three biomarkers associated with depression for future hypothesis generation: red cell distribution width, serum glucose and total bilirubin.

Structural equation modelling of complex sample survey : an application to brand signalling data

The purpose of this paper is to present an empirical analysis of complex sample data with regard to the biasing effect of nonindependence of observations on standard error parameter estimates. In a two-factor confirmatory factor analysis model, using real data, we show how the bias in standard errors can be derived when the nonindependence is ignored. We demonstrate that the standard error bias produced by the nonindependence of observations can be considerable and we briefly discuss solutions to overcome the problem.

Effect of questionnaire length, personalisation and reminder type on response rate to a complex postal survey : randomised controlled trial

Background: Minimising participant non-response in postal surveys helps to maximise the generalisability of the inferences made from the data collected. The aim of this study was to examine the effect of questionnaire length, personalisation and reminder type on postal survey response rate and quality and to compare the costeffectiveness of the alternative survey strategies.
Methods: In a pilot study for a population study of travel behaviour, physical activity and the environment, 1000 participants sampled from the UK edited electoral register were randomly allocated using a 2 × 2 factorial design to receive one of four survey packs: a personally addressed long (24 page) questionnaire pack, a personally addressed short (15 page) questionnaire pack, a non-personally addressed long questionnaire pack or a nonpersonally addressed short questionnaire pack. Those who did not return a questionnaire were stratified by initial randomisation group and further randomised to receive either a full reminder pack or a reminder postcard. The effects of the survey design factors on response were examined using multivariate logistic regression.
Results: An overall response rate of 17% was achieved. Participants who received the short version of the questionnaire were more likely to respond (OR = 1.48, 95% CI 1.06 to 2.07). In those participants who received a reminder, personalisation of the survey pack and reminder also increased the odds of response (OR = 1.44, 95% CI 1.01 to 1.95). Item non-response was relatively low, but was significantly higher in the long questionnaire than the short (9.8% vs 5.8%; p = .04). The cost per additional usable questionnaire returned of issuing the reminder packs was £23.1 compared with £11.3 for the reminder postcards.
Conclusions: In contrast to some previous studies of shorter questionnaires, this trial found that shortening a relatively lengthy questionnaire significantly increased the response. Researchers should consider the trade off between the value of additional questions and a larger sample. If low response rates are expected, personalisation may be an important strategy to apply. Sending a full reminder pack to non-respondents appears a worthwhile, albeit more costly, strategy.

Predicting spatio-temporal distribution for eastern Australian birds using birds Australia's atlas data : survey method, habitat and seasonal effects

Faunal atlases are landscape-level survey collections that can be used for describing spatial and temporal patterns of distribution and densities. They can also serve as a basis for quantitative analysis of factors that may influence the distributions of species. We used a subset of Birds Australia’s Atlas of Australian Birds data (January 1998 to December 2002) to examine the spatio-temporal distribution patterns of 280 selected species in eastern Australia (17–37°S and 136–152°E). Using geographical information systems, this dataset was converted into point coverage and overlaid with a vegetation polygon layer and a half-degree grid. The exploratory data analysis involved calculating species-specific reporting rates spatially, per grid and per vegetation unit, and also temporally, by month and year. We found high spatio-temporal variability in the sampling effort. Using generalised linear models on unaggregated point data, the influences of four factors – survey method and month, geographical location and habitat type – were analysed for each species. When counts of point data were attributed to grid-cells, the total number of species correlated with the total number of surveys, while the number of records per species was highly variable. Surveys had high interannual location fidelity. The predictive values of each of the four factors were species-dependent. Location and habitat were correlated and highly predictive for species with restricted distribution and strong habitat preference. Month was only of importance for migratory species. The proportion of incidental sightings was important for extremely common or extremely rare species. In conclusion, behaviour of species differed sufficiently to require building a customized model for each species to predict distribution. Simple models were effective for habitat specialists with restricted ranges, but for generalists with wide distributions even complex models gave poor predictions.

Nonpharmacologic interventions need outcomes for evaluating complex interventions in rheumatic diseases

Introduction. Nonpharmacologic interventions are themselves complex and are often combined with drugs and other interventions in the treatment of rheumatic diseases. Therefore, overall strategies for treatment are complex interventions. These should be evaluated regarding their processes and outcomes.
Methods. The CARE network, an international organization of health professionals (physicians and nonphysicians) and patients conducted a survey in 2008 to identify core outcomes in the ICF perspective, completed with a second survey (2009–2010) with patients in routine practice. These surveys have provided new information about domains to investigate as a basis for evaluating complex interventions.
Results. Outcome Measures in Rheumatology Clinical Trials (OMERACT) participants in this Special Interest Group agreed that current outcomes used in pharmacological research are not sufficient if the nonpharmacologic independent or combined contributions are to be assessed; other domains need to be addressed. This is an area of interest for further development.
Conclusion. Recommendations are proposed to develop research in the area of outcome for evaluation of complex interventions in rheumatic diseases.

Associations between fruit and vegetable consumption and depressive symptoms: evidence from a national Canadian longitudinal survey

BACKGROUND: Several cross-sectional studies have demonstrated associations between diet quality, including fruit and vegetable consumption, and mental health. However, research examining these associations longitudinally, while accounting for related lifestyle factors (eg, smoking, physical activity) is scarce. METHODS: This study used data from the National Population Health Survey (NPHS), a large, national longitudinal survey of Canadians. The sample included 8353 participants aged 18 and older. Every 2 years from 2002/2003 to 2010/2011, participants completed self-reports of daily fruit and vegetable consumption, physical activity, smoking and symptoms of depression and psychological distress. Using generalised estimating equations, we modelled the associations between fruit and vegetable consumption at each timepoint and depression at the next timepoint, adjusting for relevant covariates. RESULTS: Fruit and vegetable consumption at each cycle was inversely associated with next-cycle depression (β=-0.03, 95% CI -0.05 to -0.01, p<0.01) and psychological distress (β=-0.03, 95% CI -0.05 to -0.02, p<0.0001). However, once models were adjusted for other health-related factors, these associations were attenuated (β=-0.01, 95% CI -0.04 to 0.02, p=0.55; β=-0.00, 95% CI -0.03 to 0.02, p=0.78 for models predicting depression and distress, respectively). CONCLUSIONS: These findings suggest that relations between fruit and vegetable intake, other health-related behaviours and depression are complex. Behaviours such as smoking and physical activity may have a more important impact on depression than fruit and vegetable intake. Randomised control trials of diet are necessary to disentangle the effects of multiple health behaviours on mental health.

Perspectives of clinical handover processes: a multi-site survey across different health professionals

AIMS AND OBJECTIVES: To examine the perspectives of health professionals of different disciplines about clinical handover. BACKGROUND: Ineffective handovers can cause major problems relating to the lack of delivery of appropriate care. DESIGN: A prospective, cross-sectional design was conducted using a survey about clinical handover practices. METHODS: Health professionals employed in public metropolitan hospitals, public rural hospitals and community health centres were involved. The sample comprised doctors, nurses and allied health professionals, including physiotherapists, social workers, pharmacists, dieticians and midwives employed in Western Australia, New South Wales, South Australia and the Australian Capital Territory. The survey sought information about health professionals' experiences about clinical handover; their perceived effectiveness of clinical handover; involvement of patients and family members; health professionals' ability to confirm understanding and to clarify clinical information; role modelling behaviour of health professionals; training needs; adverse events encountered and possibilities for improvements. RESULTS: In all, 707 health professionals participated (response rate = 14%). Represented professions were nursing (60%), medicine (22%) and allied health (18%). Many health professionals reported being aware of adverse events where they noticed poor handover was a significant cause. Differences existed between health professions in terms of how effectively they gave handover, perceived effectiveness of bedside handover vs. nonbedside handover, patient and family involvement in handover, respondents' confirmation of understanding handover from their perspective, their observation of senior health professionals giving feedback to junior health professionals, awareness of adverse events and severity of adverse events relating to poor handovers. CONCLUSIONS: Complex barriers impeded the conduct of effective handovers, including insufficient opportunities for training, lack of role modelling, and lack of confidence and understanding about handover processes. RELEVANCE TO CLINICAL PRACTICE: Greater focus should be placed on creating opportunities for senior health professionals to act as role models. Sophisticated approaches should be implemented in training and education.

Self-management support and training for patients with chronic and complex conditions improves health-related behaviour and health outcomes

The Sharing Health Care SA chronic disease self-management (CDSM) project in rural South Australia was designed to assist patients with chronic and complex conditions (diabetes, cardiovascular disease and arthritis) to learn how to participate more effectively in the management of their condition and to improve their self-management skills. Participants with chronic and complex conditions were recruited into the Sharing Health Care SA program and offered a range of education and support options (including a 6-week peer-led chronic disease self-management program) as part of the Enhanced Primary Care care planning process. Patient self-reported data were collected at baseline and subsequent 6-month intervals using the Partners in Health (PIH) scale to assess self-management skill and ability for 175 patients across four data collection points. Health providers also scored patient knowledge and self-management skills using the same scale over the same intervals. Patients also completed a modified Stanford 2000 Health Survey for the same time intervals to assess service utilisation and health-related lifestyle factors. Results show that both mean patient self-reported PIH scores and mean health provider PIH scores for patients improved significantly over time, indicating that patients demonstrated improved understanding of their condition and improved their ability to manage and deal with their symptoms. These results suggest that involvement in peer-led self-management education programs has a positive effect on patient self-management skill, confidence and health-related behaviour.

PhysarumSpreader: a new bio-Inspired methodology for identifying influential spreaders in complex networks

Identifying influential spreaders in networks, which contributes to optimizing the use of available resources and efficient spreading of information, is of great theoretical significance and practical value. A random-walk-based algorithm LeaderRank has been shown as an effective and efficient method in recognizing leaders in social network, which even outperforms the well-known PageRank method. As LeaderRank is initially developed for binary directed networks, further extensions should be studied in weighted networks. In this paper, a generalized algorithm PhysarumSpreader is proposed by combining LeaderRank with a positive feedback mechanism inspired from an amoeboid organism called Physarum Polycephalum. By taking edge weights into consideration and adding the positive feedback mechanism, PhysarumSpreader is applicable in both directed and undirected networks with weights. By taking two real networks for examples, the effectiveness of the proposed method is demonstrated by comparing with other standard centrality measures.