Dissertations / Theses on the topic 'Predictive Mental Health Analytics'

There is a marked trend of using information technologies to improve healthcare. Among all the health IT, electronic health record (EHR) systems hold great promises as they modernize the paradigm and practice of care provision. However, empirical studies in the literature found mixed evidence on whether EHRs improve quality of care. I posit two explanations for the mixed evidence. First, most prior studies failed to account for system use and only focused on EHR purchase or adoption. Second, most existing EHR systems provide inadequate clinical decision support and hence, fail to reveal the full potential of digital health. In this dissertation I address two broad research questions: a) Does meaningful use of EHRs improve quality of care? and b) How do we advance clinical decision making through innovative computational techniques of healthcare analytics? To these ends, the dissertation comprises four essays. The first essay examines whether meaningful use of EHRs improve quality of care through a natural experiment. I found that meaningful use significantly improve quality of care, and this effect is greater in historically disadvantaged hospitals such as small, non-teaching, or rural hospitals. These empirical findings present salient practical and policy implications about the role of health IT. On the other hand, in the other three essays I work with real-world EHR data sets and propose healthcare analytics frameworks and methods to better utilize clinical text (Essay II), integrate clinical guidelines and EHR data for risk prediction (Essay III), and develop a principled approach for multifaceted risk profiling (Essay IV). Models, frameworks, and design principles proposed in these essays advance not only health IT research, but also more broadly contribute to business analytics, design science, and predictive modeling research.

Business Administration/Management Information Systems<br>D.B.A.<br>One of the key determinants of an organization’s success is its ability to adapt to marketplace change. Given this reality, how do organizations survive or even thrive in today’s dynamic markets? The answer to this question is highly related to the adaptability of one of the organization’s key resource: its employees. Indeed, the central component of an organization’s success will depend on its ability to drive changes in the mental models of individual employees. Moreover, a critical facilitator of that will be the development of decision support tools that support change of those mental models. In response to this need there has been a tremendous growth in business analytic decision support tools, estimated to reach almost $200 billion in sales by 2019. The premise of this research is that these decision support tools are ill-suited to support true mental model change because they have focused on a feedback-enabled view and generally lack a predictive (feedforward-enabled) view of the likely outcomes of the decision. The purpose of this research is to study how changes in mental models can be facilitated through this feedforward mechanisms within the DSS tool. This research used a mixed method approach, leveraging the strengths of quantitative and qualitative research methodologies, to study this research question. The research showed that the feedforward-enabled DSS tool did create more mental model change and alignment (versus an ideal solution) compared to the control. The feedforward enabled tool also produced better alignment than the feedback-enabled decision support tool. In fact, the feedback-enabled decision support was shown to result in a poorer alignment with the ideal solution. This paper concludes by suggesting five areas for future research.<br>Temple University--Theses

Advanced information technologies promise a massive influx of individual-specific medical data. These rich sources offer great potential for an increased understanding of disease mechanisms and for providing evidence-based and personalized clinical decision support. However, the size, complexity, and biases of the data pose new challenges, which make it difficult to transform the data to useful and actionable knowledge using conventional statistical analysis. The so-called “Big Data” era has created an emerging and urgent need for scalable, computer-based data mining methods that can turn data into useful, personalized decision support knowledge in a flexible, cost-effective, and productive way. The goal of my Ph.D. research is to address some key challenges in current clinical deci-sion support, including (1) the lack of a flexible, evidence-based, and personalized data mining tool, (2) the need for interactive interfaces and visualization to deliver the decision support knowledge in an accurate and effective way, (3) the ability to generate temporal rules based on patient-centric chronological events, and (4) the need for quantitative and progressive clinical predictions to investigate the causality of targeted clinical outcomes. The problem statement of this dissertation is that the size, complexity, and biases of the current clinical data make it very difficult for current informatics technologies to extract individual-specific knowledge for clinical decision support. This dissertation addresses these challenges with four overall specific aims: Evidence-Based and Personalized Decision Support: To develop clinical decision support systems that can generate evidence-based rules based on personalized clinical conditions. The systems should also show flexibility by using data from different clinical settings. Interactive Knowledge Delivery: To develop an interactive graphical user interface that expedites the delivery of discovered decision support knowledge and to propose a new visualiza-tion technique to improve the accuracy and efficiency of knowledge search. Temporal Knowledge Discovery: To improve conventional rule mining techniques for the discovery of relationships among temporal clinical events and to use case-based reasoning to evaluate the quality of discovered rules. Clinical Casual Analysis: To expand temporal rules with casual and time-after-cause analyses to provide progressive clinical prognostications without prediction time constraints. The research of this dissertation was conducted with frequent collaboration with Children’s Healthcare of Atlanta, Emory Hospital, and Georgia Institute of Technology. It resulted in the development and adoption of concrete application deliverables in different medical settings, including: the neuroARM system in pediatric neuropsychology, the PHARM system in predictive health, and the icuARM, icuARM-II, and icuARM-KM systems in intensive care. The case studies for the evaluation of these systems and the discovered knowledge demonstrate the scope of this research and its potential for future evidence-based and personalized clinical decision support.

Clinical intelligence about a patient's risk of future adverse health events can support clinical decision making in personalized and preventive care. Healthcare predictive analytics using electronic health records offers a promising direction to address the challenging tasks of risk profiling. Patients with chronic diseases often face risks of not just one, but an array of adverse health events. However, existing risk models typically focus on one specific event and do not predict multiple outcomes. To attain enhanced risk profiling, we adopt the design science paradigm and propose a principled approach called Bayesian multitask learning (BMTL). Considering the model development for an event as a single task, our BMTL approach is to coordinate a set of baseline models-one for each event-and communicate training information across the models. The BMTL approach allows healthcare providers to achieve multifaceted risk profiling and model an arbitrary number of events simultaneously. Our experimental evaluations demonstrate that the BMTL approach attains an improved predictive performance when compared with the alternatives that model multiple events separately. We also find that, in most cases, the BMTL approach significantly outperforms existing multitask learning techniques. More importantly, our analysis shows that the BMTL approach can create significant potential impacts on clinical practice in reducing the failures and delays in preventive interventions. We discuss several implications of this study for health IT, big data and predictive analytics, and design science research.

Clinical applications for heart rate variability (HRV) have become increasingly popular, gaining momentum and value as societies increased understanding of physiology reveals their true potential to reflect health. An additional reason for the rising popularity of HRV analysis, along with many other algorithmic based medical processes, is the relatively recent exponential increase of computing power and capabilities. Despite this many medical standards lag behind this booming increase in scientific knowledge, as the risks and precautions involved with healthcare necessarily take priority. Resultantly, the standards which pertain to the acceptable tolerance for accurate R-peak detection have remain unchanged for decades. For similar reasons, medical software is also prone to lag behind state-of-the-art developments. Yet, society is currently on the precipice of an age of high computational abilities, mass data storage, and capabilities to apply deep learning algorithms to reveal patterns that were previously inconceivable. So, when considering the needs of the future in relation to the place of HRV in healthcare, there is a distinct need for its accurate and precise collection, storage, and processing. In the work presented in this dissertation, the overarching aim was to increase the reliability of electrocardiogram (ECG) based HRV for use in predictive health analytics. To ensure both clarity and attainability, this project-level aim was broken down and addressed in a series of several works. The first a im w ork w as t o address the problems associated with the precision specified f or a ccurate p eak d etection, and thereby increase the reliability of predictive health analytics generated using HRV metrics. The study conducted around this initial aim investigates the specifics of match window requirements, clarifies the difference between fiducial marker and QRS complex detection, and makes recommendations on the precision required for accurate HRV metric extraction. In the second work, the aim was to ensure that there is a reliable foundation for the conduction of HRV-related research. Here, a thorough investigation of relevant literature revealed the lack of a suitable software, particularly for research requiring the analysis of large databases. Consequently, an improved HRV analysis platform was developed. Through use of both user-feedback and quantitative comparison to highly regarded software, the proposed platform is shown to offer a similar standard in estimated HRV metrics but requires significantly l ess manual e ffort (batch-processing approach) than the traditional single patient focused approach. The third work also addressed this aim, providing the base peak detection algorithm implemented within the HRV analysis platform. Experimentation undertaken here ensured that the developed algorithm performed precise fiducial marker detection, thereby increasing the reliability of the generated HRV metrics (measured against the framework presented in the first work). In the fourth work, the aim was to address the lack of published literature on the relationship between ECG sampling frequency (fs) and extracted HRV, in order to further ensure the reliability of predictive health analytics generated using HRV metrics. Here, a quantitative experimental approach was taken to evaluate the impact of ECG fs on subsequent estimations of HRV. This experimentation resulted in a recommendation for the minimum required ECG fs for reliable HRV extraction. The aim of the final work was to further improve the foundation for future predicative health analytics, by developing a robust pre-processing algorithm capable of autonomous detection of regions of valid ECG signal. This type of algorithm should be considered of critical importance to furthering machine learning (ML) based applications in the medical field. ML algorithms are heavily reliant on access to vast amounts of data, and without an automated pre-processing stage would require an unrealistic amount of hand-processing for implementation.<br>Thesis (PhD Doctorate)<br>Doctor of Philosophy (PhD)<br>School of Eng & Built Env<br>Science, Environment, Engineering and Technology<br>Full Text

Mental disorders have become a major concern in public health, since they are one of the main causes of the overall disease burden worldwide. Depressive disorders are the most common mental illnesses, and they constitute the leading cause of disability worldwide. Language is one of the main tools on which mental health professionals base their understanding of human beings and their feelings, as it provides essential information for diagnosing and monitoring patients suffering from mental disorders. In parallel, social media platforms such as Twitter, allow us to observe the activity, thoughts and feelings of people’s daily lives, including those of patients suffering from mental disorders such as depression. Based on the characteristics and linguistic features of the tweets, it is possible to identify signs of depression among Twitter users. Moreover, the effect of antidepressant treatments can be linked to changes in the features of the tweets posted by depressive users. The analysis of this huge volume and diversity of data, the so-called “Big Data”, can provide relevant information about the course of mental disorders and the treatments these patients are receiving, which allows us to detect, monitor and predict depressive disorders. This thesis presents different studies carried out on Twitter data in the Spanish language, with the aim of detecting behavioral and linguistic patterns associated to depression, which can constitute the basis of new and complementary tools for the diagnose and follow-up of patients suffering from this disease

Background: The prevalence of mental ill-health and problem behaviour within the intellectually disabled population is reported to range from 30 to 50%. However, the longer term outcomes of mental ill-health and problem behaviour, such as persistence, new onset, remission and resilience, are unknown. Accordingly, the factors predictive of such outcomes are also unknown. Aims: To determine the long term outcomes of mental ill-health and problem behaviour, and the factors predictive of and associated with such outcomes, over a 10 year time-period in a cohort of adults with mild to profound intellectual disabilities. Method: A population-based cohort of adults with intellectual disabilities (n=100) was investigated at three time points over a 10 year period. Data were collected using a range of measures. Descriptive statistics were derived and regression analyses performed to determine factors predictive of outcomes. Results: The rate of psychopathology was found to have increased in the cohort over the 10 year period. Factors predictive of this increase were experiencing an angry interaction and trusting to share a secret with only one person, or anyone. The majority of the cohort experienced episodic mental ill-health, with relapse being predicted by being female and experiencing life events. New onset of mental ill-health was predicted by experiencing life events, and resilience was predicted by not experiencing life events and having urinary continence. Problem behaviours were persistent in 50%, with 50% remitting. New onset of problem behaviours was predicted by not experiencing life events, and resilience was predicted by having mild intellectual disabilities, not experiencing an angry interaction and having more than one close friend. Small but significant negative correlations were found between psychopathology and participation in social, leisure, and peer activities. Findings should be interpreted with caution due to the small sample size. Conclusions: The present study is the only existing longitudinal investigation following an adult cohort with mild to profound intellectual disabilities, at several time points over a 10 year period. Therefore, future research is needed to confirm findings. Given the increase in psychopathology, more effective monitoring, treatment and intervention is needed.

Focal epilepsy in childhood is associated with increased risk for developing behavioral, emotional, cognitive and social–adaptive impairments. The present thesis focused on mental health difficulties in paediatric refractory focal epilepsy. It undertook a detailed evaluation of the predictive power of several demographic (gender, age at assessment), clinical (age at onset and duration of epilepsy, seizure frequency), localization (lobe and lateralization of pathology) and cognitive variables (performance in intellectual, memory and academic attainment measures) for mood, conduct, inattention/hyperactivity and peer relationship difficulties, as assessed by parental report. Data from a population of 282 children and adolescents, previously collected for clinical purposes, were examined, using a series of univariate and multivariate analyses. Mental health difficulties were found to be highly prevalent, with peer relationships the most frequently reported area of difficulty, followed by inattention/hyperactivity and emotional difficulties. Different patterns of associations between the variables examined here and individual emotional/behavioural difficulties were revealed, partially confirming and extending previous findings in the literature. Longer duration of epilepsy was found to increase the risk for developing emotional difficulties; male gender and earlier age at onset the risk for conduct difficulties; male gender, earlier age at onset, longer duration and frontal lobe localization the risk for attention/hyperactivity difficulties; and finally longer duration, higher seizure frequency and right hemisphere lateralization the risk for peer difficulties. Lower cognitive functioning was found associated with overall increased mental health difficulties and a lower VIQ was predictive of all types of difficulties. Developing a firm understanding of the risk factors that contribute to mental health comorbidities in focal paediatric epilepsy can help identify and provide assessment and intervention to children who are at higher risk earlier, thus significantly improving quality of life.

Context. Foreign Accent Syndrome is a rare neurological disorder, where among other symptoms of the patient’s emotional speech is affected. As FAS is one of the mildest speech disorders, there has not been much research done on the cost-effective biomarkers which reflect recovery of competences speech. Objectives. In this pilot study, we implement the Emotional Temperature biomarker and check its validity for assessing the FAS. We compare the results of implemented biomarker with another biomarker based on the global distances for FAS and identify the better one. Methods. To reach the objective, the emotional speech data of two patients at different phases of the treatment are considered. After preprocessing, experiments are performed on various window sizes and the observed correctly classified instances in automatic recognition are used to calculate Emotional temperature. Further, we use the better biomarker for tracking the recovery in the patient’s speech. Results. The Emotional temperature of the patient is calculated and compared with the ground truth and with that of the other biomarker. The Emotional temperature is calculated to track the emergence of compensatory skills in speech. Conclusions. A biomarker based on the frame-view of speech signal has been implemented. The implementation has used the state of art feature set and thus is an unproved version of the classical Emotional Temperature. The biomarker has been used to automatically assess the recovery of two patients diagnosed with FAS. The biomarker has been compared against the global view biomarker and has advantages over it. It also has been compared to human evaluations and captures the same dynamics.

Psychological Capital (PsyCap) is a theory of human behaviour that refers to a suite of four positive psychological resources comprising hope, (self-) efficacy, resilience and optimism (HERO). PsyCap theory postulates that due to shared commonalities of the HERO constructs, omnibus PsyCap (i.e., combined HERO) more powerfully predicts a range of mental health, wellbeing and occupational outcomes than any individual component part. PsyCap theory and research developed within organisational psychology as a contextual construct associated with attitudes, behaviours and performance which has been evaluated extensively with employees within workplaces and adult learners within tertiary education settings. There is extensive empirical support for PsyCap across these settings and samples, with studies consistently demonstrating associations of PsyCap with several mental health symptoms (e.g., stress, anxiety, depression), subjective wellbeing (e.g., life satisfaction) and vocational outcomes (e.g., work/study engagement, performance). Furthermore, adult PsyCap interventions have demonstrated that PsyCap can be developed in adults and may be associated with buffering mental health symptoms in the face of adverse challenges, as well as increasing wellbeing. The preventative and promotive qualities of PsyCap on outcomes of mental health and wellbeing has drawn attention from researchers seeking to assess if these effects might apply to adolescents, a population with high prevalence rates of mental health concerns, particularly anxiety and depression. Certainly, replicability of PsyCap inquiry from places of work to places of tertiary study has provided a framework for this research to be extended to a school setting. Schools, where young people have close proximal and frequent access throughout childhood and adolescence, provide a unique and powerful entry point to reach children at risk of mental illness, or with symptoms of mental ill health, via school-based interventions that might result in improvements in student wellbeing, along with greater student engagement, attendance and success at school and throughout life. Despite this, PsyCap research in young people to date has been fragmented and sparse. Some studies, seeking to determine whether PsyCap might have a therapeutic impetus, have explored PsyCap with clinically depressed teenagers. Other studies, following PsyCap’s theoretical underpinnings have tested PsyCap in adolescent students in a school setting. However, a common problem in the current literature is the absence of a developmentally sensitive and contextually grounded conceptualisation of PsyCap in children and adolescents. This crucial shortfall arguably leads to issues with reliable and valid measurement of PsyCap, drawing study conclusions into question. The overarching aim of the current program of research was to examine HERO constructs through a developmental lens, to provide a developmentally sensitive conceptualisation of PsyCap from which a PsyCap measure could be derived, and assessment and intervention could be tailored for students. The first study of this PhD explored the associations between PsyCap and mental health and subjective wellbeing outcomes, and the predictive role of PsyCap on these outcomes, in a cross-sectional sample of students aged 9 to 14 years (n = 456). The findings indicated significant negative relationships between PsyCap and mental health symptoms and significant positive relationships between PsyCap and subjective wellbeing. Further, optimism was found to be the most influential predictor of all outcome measures, although the combination of all HERO constructs was a stronger predictor on outcomes than any individual HERO construct alone. The second study was a naturalistic longitudinal observation of the impact of time and gender on mental health symptoms and subjective wellbeing in a cohort of Year 10 students aged 14 to 17 years (n = 56), prior to the COVID-19 pandemic and 3 months later during the pandemic. The predictive utility of baseline PsyCap was also examined on follow-up mental health and subjective wellbeing outcomes at 3-months assessment. Findings from this study demonstrated that there were no significant changes in mental health symptoms from time 1 to time 2; however, subjective wellbeing significantly declined between the two timepoints. There were no gender differences in the degree of change across time; however, girls had significantly higher levels of mental health symptoms than boys at both time points. Of the HERO constructs, baseline efficacy was the strongest unique predictor of mental health symptoms at time 2, and baseline hope was the strongest unique predictor of subjective wellbeing at time 2, following the onset of the pandemic, with overall PsyCap being a stronger predictor of subjective wellbeing, though not anxiety and depression, than any of the individual HERO components, after controlling for time 1 effects. The third and final study aimed to test the preliminary effectiveness and acceptability of a PsyCap intervention on a cohort of Year 12 female students aged 17 to 18 years (n = 82) in an open trial. The brief (4 modules) school-based intervention was aimed at increasing HERO capabilities and reducing perfectionism. Secondary outcomes of mental health symptoms and subjective wellbeing were also assessed. The findings of this final study indicated significant increases in levels of efficacy, optimism, and omnibus PsyCap (combined HERO), and a significant decline in perfectionism, from pre-intervention to post-intervention. There were no significant changes in hope, resilience or secondary outcomes. Informed by a conceptually sound, theoretical PsyCap framework, the findings of this current research program demonstrate the concordant and predictive relationships of student PsyCap on outcomes of mental health and wellbeing, and the potential for PsyCap to be cultivated via a school-based intervention. Taken together these findings provide an empirical foundation upon which future research can be built.<br>Thesis (Professional Doctorate)<br>Doctor of Philosophy in Clinical Psychology (PhD ClinPsych)<br>School of Applied Psychology<br>Griffith Health<br>Full Text

As a consequence of increased awareness and the current scholarly debate regarding women's differential predictors of recidivism, criminal justice agencies are working with researchers in the field to expand their knowledge in this area. In 2007, Portland State University researchers in collaboration with the Oregon Department of Corrections conducted an investigation of factors emerging in the pathways and gender responsive literature as predictive of women's recidivism in a randomly selected sample of female (n=150) and male (n=150) inmates. This study used information gathered from that investigation for two purposes: (1) to assess the prevalence rates of victimization experiences (childhood, adolescent and adulthood), substance abuse and mental health diagnosis across male and female ODOC inmates, and (2) to assess the predictive nature of victimization experiences, substance abuse and mental health diagnoses on recidivism across gender after a three year period. Findings suggest that females suffered from higher rates of victimization experiences throughout their lifetime than male ODOC inmates and higher rates of DSM-IV-TR mood and anxiety diagnosis. Similar rates were found across gender when assessing substance abuse and diagnosis of co-occurring disorders. When assessing the predictive impact of victimization, substance abuse and mental health diagnosis on recidivism this study found support for both gender neutral and gender responsive perspectives.

The present thesis was carried out in part fulfilment of the Doctorate in Clinical Psychology at the University of Edinburgh. It is presented in portfolio format, comprising of two individual papers although a total thesis abstract provides an overview of the entire thesis. The first paper is a systematic review of existing empirical research. It explores the predictive validity of risk assessment tools for imminent (short-term) violence and aggression in forensic psychiatric settings. The second paper is an empirical study exploring how frontline nursing staff both predict and emotionally cope with experiencing violence and aggression in a high-security setting. Paper one was prepared for Aggression and Violent Behavior and paper two for The International Journal of Forensic Mental Health; so, follow their respective author guidelines. Mental health, and forensic mental health nurses have been identified as being at particular risk of experiencing patient perpetrated violence and aggression (PPVA). There is relatively little research investigating how nursing staff predict and cope with more immediate, imminent inpatient violence and aggression, specifically within secure (forensic) settings. Negative outcomes of PPVA are widely accepted and demonstrated within empirical literature, including increased anxiety and stress for staff, fractures to the therapeutic relationship between patients and staff, and difficulties with staff retention and absenteeism for the organization. Due to the extensive negative outcomes associated with PPVA, a wealth of research has focused on developing the area of violence risk assessment. Despite this, there remains limited understanding regarding the utility of existing risk assessment tools for predicting and assessing violence risk over brief time frames (i.e. days to weeks). Therefore, a systematic review was conducted to explore the predictive validity of violence risk assessment tools for imminent, short-term risk in inpatient forensic psychiatric settings. Findings demonstrated that multiple tools had decent predictive validity, however quality scores were impacted by small sample sizes. The Dynamic Appraisal of Situational Aggression- Inpatient Version was the most effective tool with the highest mean quality score. The main limitations were the small number of studies assessing some of the included tools and the level of ambiguity between studies regarding the definition of imminent, short-term violence. Developing a shared understanding of what constitutes short-term risk and improving the number and quality of studies on the largely neglected tools, should therefore be research priorities. How nurses actually recognize and predict inpatient violence and aggression in forensic psychiatric settings, and how they emotionally cope with the aftermath, are poorly explored and understood processes. A social constructivist grounded theory approach was used to analyze the transcripts from 12 interviews with frontline nursing staff from an inpatient high-security setting. A model was constructed integrating nurses' beliefs and assumptions about subtypes of violence, their efforts to use observation skills in order to aid risk prediction, and their resultant emotional experiences following PPVA. Nurses emotional coping seemed to be affected by several factors relating to the culture of the organization and the accessibility of support. Seemingly, knowing the patient helped nurses to better identify underlying needs leading to violent behavior. This understanding helped nurses to implement targeted, needs-led interventions to address these unmet needs, and so reduce recurrent and cyclical violence. Recommendations are made to build upon, and utilize nursing skills in risk prediction and management, and to help better support the emotional impact of experiencing PPVA within forensic psychiatric settings.

Na atual conjuntura assistencial em saúde mental, em que se prioriza modalidades de tratamento comunitário, o efetivo funcionamento da rede de assistência ao portador de transtorno mental é de vital importância. A região de Ribeirão Preto-SP, durante o processo de desinstitucionalização de pacientes com transtorno mental das últimas décadas, assistiu à diminuição substancial do número de leitos psiquiátricos e à criação de novos serviços comunitários, que mais recentemente parecem não ser suficientes para atender à demanda de internação existente. Nesse contexto, o objetivo deste estudo foi analisar fatores sociodemográficos e clínicos associados à má aderência ao tratamento, às readmissões hospitalares e às necessidades de cuidados não atendidas de pacientes com transtorno mental, pois essas informações juntas podem ser um indicador de qualidade da assistência oferecida. Inovou-se, ao utilizar em um estudo brasileiro, na área da saúde mental, o índice de vulnerabilidade social como uma variável sociodemográfica, e também por se avaliarem as necessidades de cuidados dos pacientes da rede de assistência utilizando a Camberwell Assessment of Needs (CAN), instrumento mundialmente utilizado para esse fim. Foi realizada busca de todos os 933 pacientes psiquiátricos que apresentaram um transtorno mental que foi grave o suficiente para justificar suas primeiras internações nos anos de 2006 e 2007, nos hospitais com leitos psiquiátricos em Ribeirão Preto. Como a busca foi do indivíduo na comunidade, independente do seu vínculo com serviços de saúde, esse desenho conferiu ao estudo uma característica naturalística do que aconteceu aos pacientes na rede assistencial em um período mínimo de quatro anos após a primeira internação. Por meio de entrevista com 333 pacientes, identificou-se taxa de aderência de 59,6%, e os fatores associados à melhor aderência ao tratamento foram: maior faixa etária, inatividade profissional e internação em hospital geral; enquanto o diagnóstico de transtorno relacionado ao uso de substâncias psicoativas associou-se a pior aderência. Em relação à taxa de reinternação, 22,2% dos pacientes foram reinternados no período. A inatividade profissional, a internação índice com 31 dias ou mais e diagnósticos de mania e transtornos psicóticos do tipo esquizofrênico associaram-se positivamente à readmissão hospitalar. Finalmente, a média de necessidades de cuidados não atingidas encontrada foi de 4,5, similar aos resultados de países desenvolvidos. Nesta amostra, as variáveis significativamente associadas com maiores escores na CAN foram: menos anos de estudo, inatividade laboral e o fato de morar com o companheiro. Os dados do presente estudo alcançaram os objetivos propostos e trouxeram informações sobre o perfil dos pacientes que necessitam de mais assistência e devem ser priorizados pelas políticas públicas de saúde.<br>In the current context of mental health where community mental services are a priority, the effective operation of the health care network for people with mental disorders is essential. During the process of deinstitutionalization of patients with mental disorders over the past decades, the region of Ribeirão Preto SP witnessed a substantial decrease in the number of psychiatric beds and the creation of new community services, which, however, cannot meet the current demands for hospitalization. Therefore, we aimed to analyze the sociodemographic and clinical factors associated to poor treatment adherence, readmissions and unmet needs of patients with mental disorders. Such information, taken together, can provide an indicator of the quality of health care delivered. The use of an index of social vulnerability as a socio-demographic variable in a Brazilian study in the field of mental health is an innovation, as well as the assessment of care needs by patients admitted in the local mental health network through the Camberwell Assessment of Needs (CAN), a tool used worldwide for this purpose. In our search, we identified all the 933 psychiatric patients who had a sufficiently serious mental disorder that justified their first admissions in 2006 and 2007, in hospitals with psychiatric beds in Ribeirão Preto. Since the individuals were contacted in their community, regardless of their attachment to health services, this design conferred to the study a naturalistic feature of what happened to the patients in the health network for a minimum period of 4 years after the first admission. In the interviews of the 333 patients, we identified a rate of adherence to treatment of 59.6%, and the factors associated to better adherence to treatment were older age, professional inactivity and hospitalization in a general hospital, while a diagnosis of disorder related to the use of psychoactive substances was associated to the worst treatment adherence. Regarding the rate of hospital readmission, 22.2% of the patients were readmitted in the period. Inactivity, hospitalization for 31 days or more and diagnoses of mania and psychotic disorders were positively associated with hospital readmission. Finally, the average rate of unmet care needs was 4.5, similar to the results from developed countries. In our sample, the variables significantly associated to higher scores in the CAN were less years of schooling, inactivity and living with a partner. Our data achieved the proposed goals and brought information on the profile of the patients who require more care and should be a priority of the public health policies.

University of Technology Sydney. Faculty of Engineering and Information Technology.<br>In recent years, advances in Artificial Intelligence (AI) are opening the door for intelligent health care data prediction and decision making. Machine learning, as an increasingly popular approach to AI, has been widely used to learn directly from data, adapt independently, and produce predictive outcomes, which support doctors when encountering complex health care predictive analytics. However, traditional machine learning methods are not always perfectly working in the health field, intrinsically due to little consideration for characteristic problems within health care data. For example, the small sample size problem is common due to complex data collection procedures and privacy concerns. Missing data is also widely encountered since most data are collected as a second-product of patient-care activities instead of following systematic research protocols. The class imbalance is another inevitable problem in the medical data as the normal class always predominates over the disease class. To solve aforementioned issues in health care predictive analytics, this study stands on the principles of machine learning and transfer learning to develop five advanced prediction models. The first model is an output-based transfer least squares support vector machines (LS-SVMs) model which can leverage knowledge from the existing prediction model or on-line tool to facilitate the learning process on the current domain of interest with insufficient data. This model overcomes the small sample size problem and improves the health care data prediction by learning knowledge from the other domain. The second model is a novel additive LS-SVMs model which can make predictions simultaneously considering the influences on the classification error caused by missing features in a dataset. This model can generate valuable explanations regarding the influence levels of missing features for health professionals to improve the future data collection process. The third model is a transfer-based additive LS-SVMs model which can deal with missing data from a transfer learning perspective. It can leverage the model knowledge learned from the complete portion of the dataset to help the learning process on the whole dataset with missing data. The proposed model can provide supplementary information for health professionals to improve the data quality via data cleaning. The forth model is a deep transfer additive LS-SVMs model called DTA-LS-SVMs and its imbalanced version called iDTA-LS-SVMs to enhance the prediction performance on the balanced and imblanced datasets. Inspired by the stacked architecture and transfer learning mechanism, the model stacks multiple additive LS-SVMs based modules layer-by-layer and embeds model transfer between adjacent modules to guarantee their consistency. The fifth model is a deep cross-output transfer LS-SVMs model called DCOT-LS-SVMs and its imbalanced version called IDCOT-LS-SVMs to improve the prediction performance on the balanced and imbalanced datasets. The cross-output transfer is used to transfer the predictive outcome from the previous module to the adjacent higher layer to achieve a better learning. Moreover, modules’ parameters can be randomly assigned in the proposed model which significantly reduces the time for model selection. The proposed models are verified using experiments on the public UCI datasets. Moreover, case studies are conducted to validate and integrate the proposed models with real world applications, including bladder cancer prognosis, prostate cancer diagnosis, and predictions of elderly quality of life (QOL). The results have demonstrated that the models in this study can enhance the prediction performance while taking the characteristic problems within health care data into account, thus exhibiting promising potential for use in different health applications in future.

The U.S. health care system is considered costly and highly inefficient, devoting substantial resources to the treatment of acute conditions in a hospital setting rather than focusing on prevention and keeping patients out of the hospital. The potential for cost savings is large; in the U.S. more than $30 billion are spent each year on hospitalizations deemed preventable, 31% of which is attributed to heart diseases and 20% to diabetes. Motivated by this, our work focuses on developing centralized and distributed learning methods to predict future heart- or diabetes- related hospitalizations based on patient Electronic Health Records (EHRs). We explore a variety of supervised classification methods and we present a novel likelihood ratio based method (K-LRT) that predicts hospitalizations and offers interpretability by identifying the K most significant features that lead to a positive prediction for each patient. Next, assuming that the positive class consists of multiple clusters (hospitalized patients due to different reasons), while the negative class is drawn from a single cluster (non-hospitalized patients healthy in every aspect), we present an alternating optimization approach, which jointly discovers the clusters in the positive class and optimizes the classifiers that separate each positive cluster from the negative samples. We establish the convergence of the method and characterize its VC dimension. Last, we develop a decentralized cluster Primal-Dual Splitting (cPDS) method for large-scale problems, that is computationally efficient and privacy-aware. Such a distributed learning scheme is relevant for multi-institutional collaborations or peer-to-peer applications, allowing the agents to collaborate, while keeping every participant's data private. cPDS is proved to have an improved convergence rate compared to existing centralized and decentralized methods. We test all methods on real EHR data from the Boston Medical Center and compare results in terms of prediction accuracy and interpretability.

The primary aim of this thesis was to investigate the feasibility and robustness of predictive machine-learning models in the context of improving hospital resources’ utilisation with data- driven approaches and predicting hospitalisation with hospital quality assessment metrics such as length of stay. The length of stay predictions includes the validity of the proposed methodological predictive framework on each hospital’s electronic health records data source. In this thesis, we relied on electronic health records (EHRs) to drive a data-driven predictive inpatient length of stay (LOS) research framework that suits the most demanding hospital facilities for hospital resources’ utilisation context. The thesis focused on the viability of the methodological predictive length of stay approaches on dynamic and demanding healthcare facilities and hospital settings such as the intensive care units and the emergency departments. While the hospital length of stay predictions are (internal) healthcare inpatients outcomes assessment at the time of admission to discharge, the thesis also considered (external) factors outside hospital control, such as forecasting future hospitalisations from the spread of infectious communicable disease during pandemics. The internal and external splits are the thesis’ main contributions. Therefore, the thesis evaluated the public health measures during events of uncertainty (e.g. pandemics) and measured the effect of non-pharmaceutical intervention during outbreaks on future hospitalised cases. This approach is the first contribution in the literature to examine the epidemiological curves’ effect using simulation models to project the future hospitalisations on their strong potential to impact hospital beds’ availability and stress hospital workflow and workers, to the best of our knowledge. The main research commonalities between chapters are the usefulness of ensembles learning models in the context of LOS for hospital resources utilisation. The ensembles learning models anticipate better predictive performance by combining several base models to produce an optimal predictive model. These predictive models explored the internal LOS for various chronic and acute conditions using data-driven approaches to determine the most accurate and powerful predicted outcomes. This eventually helps to achieve desired outcomes for hospital professionals who are working in hospital settings.

碩士<br>國立臺灣科技大學<br>工業管理系<br>107<br>With the evolution of the information and communication technology, following the development of Industry 4.0, Artificial Intelligence (AI), Big Data, and Cloud Computing play important roles in the smart manufacturing factory. In this thesis, we focused on the big data predictive analytics, having three properties: velocity, variety and volume in healthcare management. In this thesis, we proposed a real time intelligent medical forecasting system, which was divided into two phases. In the first phase, a Big Data approach for Medical Demand Forecasting, including several time series forecasting methods, such as weighted moving average method, exponential smoothing method and simple linear regression, to compensate the missing values. Then, applying ARIMA and BPNN to forecast the medical demand. In the second phase called the, Real-Time Big Data Predictive Analytics for Medical Referral Strategy, we focused on the patients who contracted the cardiovascular diseases and deployed the BPNN to fit the historical data to forecast that the original health center should refer patients to the designated health center according to the type of cardiovascular diseases. Furthermore, we used the data set from the ABC medical group as a case study in the field of healthcare management and this forecasting system not only used for this case data but also it could apply to other relatively data sets.

Indiana University-Purdue University Indianapolis (IUPUI)<br>Community Health Centers (CHCs) play a pivotal role in delivery of primary healthcare to the underserved, yet have not benefited from a modern data analytics platform that can support clinical, operational and financial decision making across the continuum of care. This research is based on a systems redesign collaborative of seven CHC organizations spread across Indiana to improve efficiency and access to care. Three research questions (RQs) formed the basis of this research, each of which seeks to address known knowledge gaps in the literature and identify areas for future research in health informatics. The first RQ seeks to understand the information needs to support operations at CHCs and implement an information architecture to support those needs. The second RQ leverages the implemented data infrastructure to evaluate how advanced analytics can guide open access scheduling – a specific use case of this research. Finally, the third RQ seeks to understand how the data can be visualized to support decision making among varying roles in CHCs. Based on the unique work and information flow needs uncovered at these CHCs, an end to-end analytics solution was designed, developed and validated within the framework of a rapid learning health system. The solution comprised of a novel heterogeneous longitudinal clinic data warehouse augmented with big data technologies and dashboard visualizations to inform CHCs regarding operational priorities and to support engagement in the systems redesign initiative. Application of predictive analytics on the health center data guided the implementation of open access scheduling and up to a 15% reduction in the missed appointment rates. Performance measures of importance to specific job profiles within the CHCs were uncovered. This was followed by a user-centered design of an online interactive dashboard to support rapid assessments of care delivery. The impact of the dashboard was assessed over time and formally validated through a usability study involving cognitive task analysis and a system usability scale questionnaire. Wider scale implementation of the data aggregation and analytics platform through regional health information networks could better support a range of health system redesign initiatives in order to address the national ‘triple aim’ of healthcare.

abstract: The purpose of this study was to evaluate whether five select scales of the MMPI-A (F, Scale 2, A-dep, A-lse, and A-aln) are predictive of a diagnosis of a major depressive episode according to the current DSM-IV-TR criteria. Participants were 90 girls and 58 boys in a clinical psychiatric setting. The study examined two separate hypotheses across the five scales. The first set of hypotheses tested whether a significant T-score on each of the five scales would predict a diagnosis of a major depressive episode in clinical adolescents. The second set of hypotheses attempted to step away from the constraints of diagnostic and statistical cut-off criteria and evaluated the ability of discrete T-scores of the MMPI-A in predicting the number of symptoms of a major depressive episode in clinical adolescents. Results indicated that none of the five scales were predictive of a diagnosis of a major depressive disorder in clinical adolescents. All but one scale (Scale 2) was significant in its ability to predict the number of depressive symptoms in clinical adolescents. Implications of this study include the need for a better diagnostic criteria for adolescent depression as well as re-evaluating the cut-off criteria of scales on the MMPI-A. Directions for future research are also discussed.<br>Dissertation/Thesis<br>Ph.D. Counseling Psychology 2010

Child trauma exposure (CTE) is an important public health concern in the U.S.; more than two-thirds of children report experiencing a traumatic event by the age of 16. CTE may have important acute and long-term physiological, developmental, behavioral, and psychological implications if not addressed. Trauma-focused cognitive behavioral therapy (TF-CBT) is the gold standard for treatment of child trauma and is well-supported for resulting in significant decreases in negative mental health outcomes associated with CTE. Despite the efficacy of evidence-based interventions such as TF-CBT, many children do not receive treatment due to a variety of contextual, logistical, and interpersonal barriers. This mixed-methods exploratory study examines possible predictors of enrollment and completion of TF-CBT, specifically parental trauma exposure, at a community organization that serves abused and traumatized children in the metro Atlanta area. Data were collected during individual assessments consisting of a computer survey and semi-structured interview (n=41). Data analysis focused on parental trauma exposure, and qualitative interviews were examined for common themes regarding intentions for their child’s enrollment and completion of services. Results indicated that caregivers of children referred to services had relatively high (56.1%) rates of trauma exposure. Results from logistical regression indicate that parents with a trauma history were 10.5 times more likely to have a child enroll in therapy. These results indicate that parents with personal trauma histories may be more committed to their child receiving services, therefore public health efforts aimed towards educating parents without trauma histories may be beneficial.

RESUMO: Introdução: A Comissão Nacional para a reestruturação dos Serviços de Saúde Mental em 2004, fez uma proposta de âmbito regional, ao nível da região de Saúde do Norte, levando a uma alteração da rede de referenciação hospitalar dos internamentos em psiquiatria. Métodos: Realizou-se um estudo epidemiológico longitudinal para avaliar factores preditivos de internamento através de um serviço de Urgência de Segunda Linha (SII), que de algum modo reflectisse o funcionamento dos serviços de internamento na região de Saúde do Norte, ao longo de 12 anos, nomeadamente relacionando com factores organizacionais contemplados no Plano Nacional de Saúde Mental e na reorganização da rede de referenciação hospitalar. Resultados: Durante os 12 anos do estudo, verificou-se um aumento estatisticamente significativo do número e duração de internamentos através do SII, com ponto de partida no ano 2008-2009, e de novo a partir do ano 2010-2011 (nº de internamentos), para o qual contribuiu a alteração da rede hospitalar na região de saúde do Norte, nomeadamente pelo facto do HMLemos, assumir a responsabilidade de novo, dos internamentos das áreas de Famalicão, Gondomar e Santa Maria da Feira. Em relação ao número de internamentos, e na análise exploratória, encontramos nas áreas hospitalares fora da área de influência do HMLemos, uma contribuição positiva significativa para o aumento do nº de internamentos ao longo dos anos com os Dx (290, 296, 297, 291, 309). Em relação à área do HMLemos restrita (PVC, STT, Matosinhos, Porto Ocidental), de referir a contribuição positiva significativa dos Dx 309 e 301, para o aumento do número de internamentos ao longo do tempo, sendo que a prevalência maior se mantém relacionada às Psicoses (Dx 295, 296 e 297). Não se concluiu por uma contribuição estatisticamente significativa ( positiva ou negativa), das variáveis independentes idade, sexo ou natureza do internamento em relação à variável dependente ( duração de internamentos/ano). Em relação á variável dependente (nº de internamentos/ano), relativamente aos doentes fora de área de influência do HMLemos, concluiu-se uma contribuição positiva estatisticamente significativa da variável independente idade. Conclusões: Através da análise exploratória foi possível perceber o esforço realizado pelos hospitais no sentido de melhorar a equidade e acessibilidade dos doentes à Saúde Mental, a par da reorganização da rede hospitalar. De destacar a necessidade de encontrar alternativas às situações de internamento, com menos critérios de gravidade diagnóstica, nomeadamente reforçar a importância da criação de consultas de crises nos respectivos Hospitais de Dia dos DPSM.----------------ABSTRACT:Introduction : The National Commission for the restructuring of mental health services in 2004 , has proposed at a regional level ( North Health Region), a change in the network of hospital referrals of admissions in psychiatry. Methods: We conducted a longitudinal epidemiological study to assess predictors of hospitalization through a Second Line Emergency Service ( SII) , that somehow reflect the operation of inpatient services in North Health Region, over 12 years, particularly relating to organizational factors included in the National Mental Health Plan and reorganization of the hospital referral network. Results: During the period of the study, there was a statistically significant increase in the number and duration of hospitalizations through the SII, with starting point in the year 2008-2009 and again from 2010-2011 (number of admissions) , for which counted the change of the hospital network referral in Northern health region , in particular because Hospital Magalhães Lemos (HMLemos) , took the new responsibility of admissions from areas of Famalicão, Gondomar and Santa Maria da Feira . Regarding the number of hospitalizations, in the exploratory analysis , we found in hospital areas outside the area of influence of HMLemos , a significant positive contribution to the increase in number of admissions over the years with Diagnosis of 290, 296, 297, 291 , 309 in the ICD-9. With respect to the restricted area of HMLemos (PVC, STT , Porto Ocidental and Matosinhos) , we found a significant positive contribution of Diagnosis 309 and 301, to increase the number of hospitalizations over time, with higher prevalence rates remaining the psychoses ( Dx 295, 296 and 297 ) . Did not conclude for any statistically significant contribution (positive or negative) of the independent variables age, sex and nature of admission to the dependent variable (duration of hospitalization / year). In relation to the dependent variable (number of admissions / year) relative to patients outside the area of influence of HMLemos, it was found a statistically significant positive contribution of the independent variable age . Conclusions: Through the exploratory analysis, it was possible to see the efforts made by hospitals to improve the accessibility of patients to Mental Health, throughout the hospital network reorganization. Its important to highlight the need to find alternatives to inpatient admissions in those with less gravity diagnostic criteria, reinforcing the importance of creating specific crisis consultations in Day Hospital regime.