Effective translation of research findings into clinical practice using Predictive Analytics will not only require the combination of expert domain-knowledge and data integration technology as outlined above. Effective translation will also need to address more general issues regarding the organization and structure of the emerging field. This will require joint efforts from all stakeholders including researchers, clinicians, patients, funding bodies, and policymakers. One such example is the Patient Centered Outcomes Research Network and its associated psychiatric networks, the Mood Network, the Interactive Autism Network, and the Community and Patient Partnered Centers of Excellence which focuses on behavior disorders in under served communities. 50 Given the often sensitive nature of the data needed to build predictive models – which might for example include electronic health records – an adequate level of security must be maintained at all times. Whether this speaks for decentralized infrastructure or outsourcing to specialized institutions is likely to remain a matter of intensive debate. As an example, PCORnet uses a federated datamart with a common data model infrastructure for multiple health care systems across the US that includes over 90 million people. Similar discussions will probably arise with regard to the predictive models themselves. Whereas only easy access to validated, pre-trained models will make them widespread, useful tools in the clinic, predictive models might also enable the prediction of sensitive personal data from the combination of seemingly harmless information an individual might readily provide. Thus,2 tier grow rack it is in the interest of all stakeholders to reach a public consensus regarding the regulation of access to pre-trained models before practically applicable models become available.

While some level of regulation is likely beneficial with regard to industry use, it will be essential for efficient model construction to encourage model sharing for research purposes. Especially for multi-modal models, sharing modality-specific, pretrained models will save substantial amounts of time and money. Finally, we need experts to consider the legal implications of deploying models which predict health-related information which potentially guides medical decisions. From a more applied perspective, we believe that technology will continue to simplify data acquisition and improve data quality in the years to come, thus bringing predictive Mobile Health applications within reach. While holding great promise, especially mHealth applications raise the question of whether it is generally better to rely on mechanistic predictors or instead on a pragmatic approach.23, 51 While we firmly believe that the identification of causal relationships provides the most robust and scientifically satisfying features for prediction, we expect a pragmatic approach to prevail in the years ahead for two reasons. First, while causal predictors might be most effective, they will often be inefficient. For example, measuring variables of brain metabolism causally linked to a disorder might enable the construction of highly accurate predictive models. If however, we can use cheaper and more readily obtainable measures not causal to the disorder with comparable or even slightly lower predictive power, those would probably be more efficient and thus more useful to clinicians in practice. Secondly, as decades of research have only begun to uncover causal links on single levels of observation, we think it highly unlikely that unified theoretical models across levels of observation will be established even in the mid-term. To promote the endeavor of creating individualized predictive models to improve patient care and maximize cost efficiency in psychiatry, concrete steps can to be taken by institutions, researchers and practitioners. For example, we have recently seen numerous educational efforts such as organizing workshops and seminars on the various technical topics.

Conferences such as the European College of Neuropsychopharmacology Congress or the Resting-State Conference and many others will continue to host sessions and satellite symposia dedicated to predictive analytics. Common in the field of machine learning, but currently scarce in psychiatry, predictive analytics competitions in which teams compete for the best predictive model performance bring together clinicians, researchers, and machine learners and may accelerate the availability of pre-trained, validated models in the midterm as well as make this research more visible to the public. Although patients, clinicians, and researchers share a common interest in improving mental health outcomes, there will need to be a thoughtful balancing of issues related to privacy, data security, and ethics in relation to the contrasting priorities and roles of various stakeholders. Currently, research and curation of shared data bases arise primarily from publicly funded, academic research groups, where data sharing is viewed as a common good to support greater utilization of large datasets to enhance predictive accuracy. A private business, on the other hand, could have the different role of using predictions to make decisions about reimbursing health care options or to advise on hiring practices or to identify potential customers for advertisements. Although these contrasting goals could lead to some tensions about the use of predictive analyses, there are examples where a public-private hybrid could be advantageous. For example, because intervention research is costly and complex, it tends to have limited numbers of subjects and relatively short durations . Public-private partnerships could take advantage of the ongoing administration of treatments to very large numbers of subjects over extended time periods. In summary, we believe that unimodal feature-engineering and model integration across levels of observation will be the key to highly accurate and efficient Predictive Analytics Models in mental health.

Successful Predictive Analytics projects will thus require 1) substantial domain knowledge to enable optimal feature-engineering for the often massively multivariate datasets obtained on each level of observation and 2) profound machine learning expertise with a focus on model integration techniques. With technology rapidly simplifying data acquisition and model construction, we urge all stakeholders including researchers, clinicians, patients, funding bodies, and policymakers to initiate an open discussion regarding key-issues such as data-sharing and model access-regulations to enable Predictive Analytics technology to close the gap between bench and bedside. Neuroimaging in patients with major depression has revealed abnormal activation patterns in multiple brain networks, including the default mode , cognitive control, and affective networks. The DMN, anchored in the medial prefrontal cortex and posterior cingulate cortex , is suppressed in healthy adults during tasks that demand external attention, but does not show the typical pattern of task-induced deactivation in adults and adolescents with MDD . The cognitive control network, including dorsal lateral prefrontal cortex , which is typically activated during cognitively demanding tasks, has shown decreased activations in adults with MDD . The affective network includes the amygdala and other limbic-region structures , and most saliently for MDD, the sub-genual anterior cingulate cortex , which is considered a core region in the functional and structural pathophysiology of MDD . The affective network exhibits abnormal activation patterns during emotion processing in adults with MDD . These abnormal activations in distributed networks may account for corticolimbic dysregulation in MDD .Mirroring these brain activation abnormalities, patients of different ages with MDD have shown abnormal intrinsic functional connectivity of the brain measured via resting-state fMRI . First,commercial grow set up increased resting-state connectivity within the DMN and between the DMN and sgACC has been reported in adults and adolescents with MDD. Hyper connectivity of sgACC correlated with duration of current depressive episodes in adults and with emotional dysregulation in pediatric depression . These results support the possibility that DMN-sgACC hyperconnectivity might underlie depressive rumination . Second, several studies reported decreased resting-state connectivity within the cognitive control network in adult patients with MDD . In line with this evidence, MDD has been conceptualized as an imbalance between the DMN and the cognitive control network . Third, atypical connectivity between the amygdala and cortical structures has been found in adults and children with MDD and is thought to reflect deficits in emotion regulation. Despite evidence of abnormal functional connectivity across distributed brain networks in patients with MDD, it is unclear whether these differences reflect the state of current depression versus neurobiological traits that predispose individuals to be at risk for MDD. One approach to distinguishing between current state and predisposing traits is the study of unaffected individuals at heightened risk for MDD, such as unaffected children at familial risk for MDD by virtue of having a parent with MDD. Such familial history increases the risk of MDD in offspring by three to five fold , and increases the risk of a broader spectrum of mood and anxiety disorders . Understanding whether rs-fMRI findings represent trait or state markers of MDD in the young can lead to the identification of informative neural biomarkers of risk for mood and anxiety disorders and help develop early intervention strategies to mitigate this risk. Rs-fMRI also possesses significant translational strengths in its short duration of scanning, and the lack of task performance demands that can complicate interpretation of activations. In the present study, we examined rs fMRI in unaffected children at familial risk for MDD and other mood and anxiety disorders by virtue of being offspring of parents with MDD and compared them with age matched children who were offspring of parents with no lifetime history of any mood disorder . Two previous studies examining at-risk children and adolescents found decreased connectivity between amygdala and frontal-parietal network in unaffected children of depressed mother and in children with early onset depression , and decreased connectivity within the frontal-parietal cognitive control network in unaffected adolescent girls with parental depression .

Based on previous functional connectivity results in patients with MDD, we focused on functional connectivity differences between at-risk and control children in the DMN, the cognitive control network, and the affective network, using a seed-based functional connectivity approach. We examined connectivity differences from the two midline anchor regions of the DMN , which are associated with self-referential processing and self-focused rumination in MDD , and from seed regions in left and right DLPFC and amygdala. We tested: 1) whether unaffected at-risk children exhibit patterns of abnormal functional connectivity similar to those reported in patients with MDD, and 2) whether connectivity of DMN-sgACC is related to symptom scores in at-risk children. To further test whether resting-state connectivity can be a useful neural biomarker for risk for MDD, we built classification models based on resting-state d ata to discriminate at-risk versus control children.Data were acquired on a 3T TrioTim Siemens scanner using a 32-channel head coil. T1- weighted whole brain anatomical images were acquired. After the anatomical scan, participants underwent a resting fMRI scan in which participants were instructed to keep their eyes open and the screen was blanked. Resting scan images were obtained in 67 2-mm thick transverse slices, covering the entire brain . The resting scan lasted 6.2 min .Two dummy scans were included at the start of the sequence. Functional connectivity analysis Rs-fMRI data were first preprocessed in SPM8, using standard spatial preprocessing steps. Images were slice-time corrected, realigned to the first image of the resting scan, resampled such that they matched the first image of the resting scan voxel-for-voxel, normalized in MNI space, and smoothed with a 6-mm kernel . Functional connectivity analysis was performed using a seed-driven approach with in-house, custom software “CONN” . We performed seed-voxel correlations by estimating maps showing temporal correlations between the BOLD signal from our a priori regions of interest and that at every brain voxel. We performed resting-state connectivity analysis from the DMN seeds , cognitive control network seeds , and bilateral amygdala seeds . The DMN and DLPFC seeds were defined as 6-mm spheres around peak coordinates from . The amygdala seeds were defined from the WFU Pick Atlas . Physiological and other spurious sources of noise were estimated and regressed out using the anatomical CompCor method . Global signal regression , a widely used preprocessing method was not used because it artificially creates negative correlations which prevents the interpretation of anticorrelation and can contribute to spurious group differences in positive correlations . Instead, aCompCor allows for interpretation of anticorrelations and yields higher specificity and sensitivity compared to GSR . See Supplementary Information for details on the aCompCor. A temporal band-pass filter of0.008 Hz to 0.083 Hz was applied simultaneously to all regressors in the model. Residual head motion parameters were regressed out. Artifact/outlier scans were also regressed out. Head displacement across the resting scan did not differ significantly between the two groups for either frame-to-frame translations in x, y, z directions or frame-to-frame rotations . The number of outliers also did not differ significantly between the groups .