Selecting Learning Algorithms for Simultaneous Identification of Depression and Comorbid Disorders

Machine Learning in Mental Health

“…Noisy/ ambiguous signals Ambiguous words/lexical variations Chancellor [ 28 ]; Nobles et al [ 134 ]; Saha and De Choudhury [ 165 ]; Yazdavar et al [ 211 ] Ambiguity in signals (e.g . , for audio: robust speaker detection; distinguish personal speaking style from symptoms) Chang et al [ 31 ]; Mallol-Ragolta et al [ 112 ]; Rabbi et al [ 153 ]; Salekin et al [ 168 ]; Spathis et al [ 176 , 177 ]; Zhou et al [ 222 ] Managing irrelevant, redundant information Ojeme and Mbogho [ 136 ] Dataset limitations Restrictions due to data subjects/scale/study context Too small or restricted study sample/need for larger (more diverse) datasets Adamou et al [ 2 ]; Diederich et al [ 45 ]; Feng et al [ 61 ], Kavuluru et al [ 89 ]; Morshed et al [ 128 ], Nobles et al [ 134 ]; Ojeme and Mbogho [ 136 ]; Parades et al [ 140 ]; Park et al [ 141 ], Pestian et al [ 145 ]; Quisel et al [ 152 ]; Ray et al [ 155 ]; Salekin et al [ 168 ]; Spathis et al [ 177 ], Yazdavar et al [ 211 ]; Zhou et al [ 222 ] Unknown confounding variables + limitations of study context Fatima et al [ 57 ]; Saha and De Choudhury [ 165 ]; Salekin et al [ 168 ] Reference dataset not explicitly designed for mental health-related analysis Alam et al [ 6 ] Biased, missing, incomplete data General acknowledgement of biases inherent to model design and datasset used for training Ernala et al [ 53 ]; Hirsch et al [ 78 ]; Park et al [ 141 ] Difficulties due to missing data values/sparse data Alam et al [ 6 ]; Spathis et al…”

“…In a few instances of regression tasks 9 [ 27 , 112 , 122 , 128 , 155 , 193 ], metrics of mean error (e.g., MSE, MAE, RMSE, SMAPE) were applied [ 50 , 62 , 112 , 155 , 179 ] to reveal any unexpected values, sensitivities towards outliers, and risks of over-or underestimating false predictions [ 51 ]. Individual works also applied more specific metrics to evaluate multi-dimensional classification (i.e., using Hamming score, Hamming Loss, Exact-match [ 136 ]); or the confidence [ 139 ], coherence [ 64 ], and completeness [ 45 ] of clustering outcomes (e.g . , using WCSS, Dunn Index, DB Index, or Silhouette Index to assess similarity within, and separation between, clusters [ 179 ]).…”

“…Ojeme and Mbogho [ 136 ], Srividya et al [ 179 ] Statement of study and data being exempt from ethics review Park et al [ 141 ] Privacy protection Public data access + user anonymization Park et al [ 141 ]; Saha and De Choudhury [ 165 ]; Yazdevar et al [ 211 ] No recording of person identifiable data (e.g . , text, speech, low-level interactions)…”

Causal Modelling for Supporting Planning and Management of Mental Health Services and Systems: A Systematic Review

“…Although causal modelling has been widely applied in general health and mental health care for supporting decision-making (patient level, e.g., for supporting the diagnosis) [35,37,38], it is not the case for the other levels of analysis (micro, meso, and macro). Despite its utility and increased interest in the last years [22,58,61,62], the development of causal modelling is still scarce due to the complexity (the number of variables –sometimes grouped in imprecise domains or constructs- and their causal relationships –sometimes difficult to explain- are very high) and the uncertainty (the statistical nature of the variables are unknown –unreliable or imprecise- and there are missing variables) of real environments.…”

“…In health care, BNs have been applied for decision-making [28,29,30,31] and case assessment: analyzing new diagnosis strategies [32,33,34] and diagnosing social anxiety [35], depression [36,37], and Alzheimer’s disease [38,39]. Despite its reported utility in formalising the explicit knowledge about the structure of a system, in assessing potential responses: XMxji(u) where Xi and Xj are two subsets in the set of variables –endogenous or exogenous (U)- of the BN and Mxj is the action: boldnormaldo(Xj=xj,∀boldnormalj) on it (system) and in evaluating counterfactual sentences (in situation u, X…”