1. An explicit mathematical description of PEP in terms of DBP was proposed, which in the first time quantitatively clarified the ventricular and arterial effects on PEP timing. / 2. A nonlinear pressure-volume relationship which reflected the natural arterial wall properties was introduced into the asymmetric T-tube arterial model, which effectively and quantitatively described the effect of pulsatile BP on arterial parameters, e.g., compliance, PTT etc. / 3. A mathematical relationship between PAT and BP was firstly proposed as a result of the heart-arterial interaction, which simulated a significantly strong and negative relationship between PAT and SBP and between PAT and MBP but a much weaker negative relationship between PAT and DBP during exercise. The hypothesis was supported by the experiment data. To our knowledge, it is the first study describing the quantitative relation of PAT and BP by both model-based study and experimental data. / 4. A novel wearable measurable CO parameter, PTRR, was proposed and it successfully showed a significantly high and positive correlation with CO during exercise both in model simulation and in the experiments. / 5. Linear prediction models using PAT to estimate MBP and using PTRR to estimate CO were proposed and evaluated in two exercise experiments conducted on 84 subjects with different ages and cardiovascular diseases. Results showed the proposed method could achieve the accuracy required for medical diagnosis. / 6. Taken the findings in 3, 4 and 5 together, this study in the first time provided both the theoretical basis and experimental verifications of developing a wearable and direct measurement technique of CPO in dynamic exercise using multiple physiological signals measured on body surface. / Cardiac power output (CPO) is defmed as the product of mean arterial blood pressure (MBP) and cardiac output (CO), and CPO measured during peak dynamic exercise (i.e. peak CPO) has been shown as a powerful predictor of death for heart failure patients. However, so far there has been no existing device which directly measures CPO, and CPO is acquired from simultaneous measurement of MBP and CO. Further, simultaneous MBP and CO measurement during dynamic exercise is a challenge for current BP and CO methods. Therefore, there is an urgent need to develop new devices which are fully wearable and unobtrusive for monitoring of CPO during dynamic exercise. Since the core problem in most wearable devices is how to estimate the target cardiovascular parameter, e.g. CPO in this study, through physiological signals measured from body surface, this thesis focus on developing a direct measurement technique of CPO in dynamic exercise using multiple physiological signals measured on body surface, specifically, electrocardiogram (ECG) and photoplehtysmogram (PPG). / Finally, based on the theoretical and experimental verifications, linear prediction models were proposed to estimate MBP from PAT and estimate CO from PTRR. The results showed that PAT can estimate MBP with a standard deviation of 7.42 mmHg, indicating PAT model has the potential to achieve the accuracy required by AMMI standard (mean error within +/- 5 mmHg and SD less than 8 mmHg). The results also showed that PTRR can estimate CO with a percent error of 22.57%, showing an accuracy which was considered as clinically acceptable (percent error less than 30%). / Heart failure is the end stage of many cardiovascular diseases, such as hypertension, coronary heart disease, diabetes mellitus, etc. Around 5.8 million people in the United States have heart failure and about 670,000 people are diagnosed with it each year. In 2010, heart failure will cost the United States $30.2 billion, and the cost of healthcare services is a major component of this total. With the resultant burden on health care resources it is imperative that heart failure patients with different risk stages are identified, ideally with objective indicators of cardiac dysfunction, in order that appropriate and effective treatment can be instituted. / In order to verify the theoretical findings, two experiments were carried out. One was incremental supine bicycle exercise conducted on 19 young healthy subjects and the other was incremental to maximum supine bicycle exercise conducted on 65 subjects, including heart failure patients, cardiovascular patients and healthy elderly. PAT showed significantly high and negative correlation with SBP and MBP, but lower correlation with DBP. PTRR showed significantly high and positive correlation with CO. / In this thesis, a model based study is conducted to address the above problem. Firstly, we deduced the mathematical expression of PEP as a function of DBP by introducing the arbitrary heart rate into the exponential mathematical description of a pressure-source model. Secondly, an asymmetric T-tube model was modified by introducing a nonlinear pressure-volume relationship where PTT was expressed as a dependent variant of BP. Thirdly, we proposed the mathematical equation between PAT and BP by coupling the modified ventricular and arterial models. Then, the relationships between PAT with systolic blood pressure (SBP), MBP and DBP were simulated under changing heart contractility, preload, heart rate, peripheral resistance, arterial stiffness and a mimic exercise condition. The simulation results indicated significantly high and negative correlations between PAT and SBP and between PAT and MBP whereas the correlation between DBP and PAT was low. / Next, we developed a novel CO index, namely pulse time reflection ratio (PTRR), expressed in terms of MBP and mean aortic reflection coefficient (Gamma(0)), from the modified asymmetric T-tube model. PTRR was further expressed in terms of PAT and inflection point area (IPA), a surrogate of Gamma(0) from the shape feature of PPG. The simulation results suggested significantly and positive relationship between PTRR and CO and between IPA and Gamma(0) during dynamic exercise. / Recently, a wearable measurable parameter, pulse arrival time or PAT, has been developed for BP measurement. PAT is the time delay from the R peak of ECG to the systolic foot of PPG. PAT consists of two timing components, the pre-ejection period (PEP) of the heart and pulse transit time (PTT). PTT is related to BP by an arterial elastic model and thus can be used to estimate beat-to-beat BP. However, PTT is difficult to be measured through a wearable device, and thus PAT is usually used as a surrogate of PTT for BP estimation, under the assumption of a constant PEP. However, PEP is not a constant but changing with physiological conditions, which may alter the PAT-BP relationship. Thus, it is important to clarify the PAT-BP relationship and address the feasibility of MBP estimation using PAT during dynamic exercise. / To summarize, the original contributions of this thesis are: / Wang, Ling. / Adviser: Y.T. Zhang. / Source: Dissertation Abstracts International, Volume: 73-03, Section: B, page: . / Thesis (Ph.D.)--Chinese University of Hong Kong, 2010. / Includes bibliographical references. / Electronic reproduction. Hong Kong : Chinese University of Hong Kong, [2012] System requirements: Adobe Acrobat Reader. Available via World Wide Web. / Electronic reproduction. [Ann Arbor, MI] : ProQuest Information and Learning, [201-] System requirements: Adobe Acrobat Reader. Available via World Wide Web. / Abstract also in Chinese.

Thesis (MPH.) -- University of Limpopo, 2014 / Objective: The purpose of this study was to determine the prevalence of impaired glucose tolerance, impaired fasting glucose, undiagnosed type 2 diabetes and its associated risk factors among adults patients attending the outpatient department of a level one hospital in a rural community of Barkley West, South Africa. Research methodology: This was a cross-sectional survey conducted by a simple random sampling of adults patients F 30 years old. Patients were screened using the American Diabetes Association and the World Health Organisation criteria. First, patients underwent the 75g oral glucose tolerance test and secondly, the 12-hours fasting plasma glucose tests after pre-test results of 5.5 mmol/L were obtained considered as positive for screening. To determine the prevalence of IGT, IFG, and undiagnosed type 2 diabetes; tests were conducted using both the capillary finger puncture and the laboratory methods. To ensure validity and reliability, each patient underwent two tests (fasting and random) by the capillary finger puncture method and two tests (fasting and random) by the laboratory method. Results: Eighty-five (85) questionnaires were distributed, supervised and returned by a research assistant, which brought the response rate to 100%. All patient known living with diabetes mellitus was not included in the study. The prevalence of IGT was 34.1% [34% for females and 9.4% for males] and that for IFG was 23.6% [25% for females and 6.0% for males]. The prevalence of undiagnosed type 2 diabetes discovered during the survey was 9.3% by 2-hours 75g glucose tolerance test [8.2% for females and 1.1% for males] and that by 12-hours fasting plasma glucose, the prevalence was 5.8% [4.7% for females and 1.1% for males].The associated risk factors were physical inactivity, overweight and obesity, unhealthy diet, alcohol consumption, hypertension, smoking habit, family history of diabetes, social deprivation and poverty. The prevalence of hyperglycaemia was also high among female patients due to a higher BMI with 25% overweight (females 18% overweight, males 7% overweight) and 75% obese (females 54% of obesity, males 21% of obesity); higher waist circumference with higher abdominal fat (females 71.7% had a W/C F 88 cm, males 28% had a W/C F 102 cm.); and a larger waist-to-hip ratio (females 61.1% had WHR > 0.85, males 7% had a WHR > 1.0). The sensitivity, specificity, positive and negative predictive values for IGT were 34%, 86%, 25%, and 86% and those for IFG were 24%, 86%, 19%, and 86% respectively. IGT sensitivity was greater than IFG sensitivity. xi Conclusion: There was a high prevalence of IGT, IFG and undiagnosed type 2 diabetes specifically among female patients. The ten percent difference of sensitivity between the two tests showed that the WHO diagnostic criteria produced more patients with the pathology than the ADA diagnostic criteria do. Patients attending the outpatient department of a level one hospital in Barkley West are at high risk of developing type 2 diabetes and remain unidentified, undetected, unscreened, undiagnosed and untreated. Obesity at primary health care level in the rural community of Barkley West needs to be addressed. . Keywords: Impaired glucose tolerance, prevalence, diabetes, screening, anthropometric measurements

ABSTRACT PATIENT MONITORING VIA MOBILE AD HOC NETWORK - MAXIMIZING RELIABILITY WHILE MINIMIZING POWER USAGE AND DELAYS BY SWETA SNEHA May 22nd, 2008 Committee Chair: Dr. Upkar Varshney Major Department: Computer Information Systems Comprehensive monitoring of patients based on wireless and mobile technologies has been proposed for early detection of anomalies, provision of prompt medical attention, and corresponding reduction in healthcare expenses associated with unnecessary hospitalizations and treatment. However the quality and reliability of patient monitoring applications have not been satisfactory, primarily due to their sole dependence on infrastructure-oriented wireless networks such as wide-area cellular networks and wireless LANs with unpredictable and spotty coverage. The current research is exploratory in nature and seeks to investigate the feasibility of leveraging mobile ad hoc network for extending the coverage of infrastructure oriented networks when the coverage from the latter is limited/non-existent. Although exciting, there are several challenges associated with leveraging mobile ad hoc network in the context of patient monitoring. The current research focuses on power management of the low-powered monitoring devices with the goal to maximize reliability and minimize delays. The PRD protocols leveraging variable-rate transmit power and the PM-PRD scheme are designed to achieve the aforementioned objective. The PRD protocols manage power transmitted by the source and intermediate routing devices in end to end signal transmission with the obejective to maximize end to end reliability. The PM-PRD scheme operationalizes an appropriate PRD protocol in end to end signal transmission for diverse patient monitoring scenarios with the objective to maximize reliability, optimize power usage, and minimize delays in end to end signal transmission. Analytical modeling technique is utilized for modeling diverse monitoring scenarios in terms of the independent variables and assessing the performance of the research artifacts in terms of the dependent variables. The evaluation criterion of the research artifacts is maximization of reliability and minimization of power usage and delays for diverse monitoring scenarios. The performance evaluation of the PRD protocols is based on maximization of end to end reliability in signal transmission. The utility of the PM-PRD scheme is associated with operationalizing an appropriate protocol for a given monitoring scenario. Appropriateness of a protocol for a given scenario is based on the performance of the PRD protocols with respect to the dependent variables (i.e., end to end reliability, end to end power usage, and end to end delays). Hence the performance evaluation of the PRD protocols in terms of the dependent variables is utilized to (a) discover the best protocol and (b) validate the accuracy and utility of the PM-PRD scheme in allocating the best protocol for diverse monitoring scenarios. The results validate the effectiveness of the research artifacts in maximizing reliability while minimizing power usage and delays in end to end signal transmission via a multi-hop mobile ad hoc network. Consequently the research establishes the feasibility of multi-hop mobile ad hoc network in supplementing the spotty network coverage of infrastructure oriented networks thereby enhancing the quality and dependability of the process of signal transmission associated with patient monitoring applications.

In this dissertation is proposed a methodology to identify patient's recovery patterns after cardiovascular surgery based on laboratory tests results. The main purpose is to enhance the understanding of the manifestations of postsurgical complications in patients who underwent cardiovascular surgery. The analysis of patients' recovery process is based on the relationship between plasma calcium, ionized calcium and platelet count over time. Laboratory results from the James A. Haley Veterans' Hospital databases, related to patients admitted to the Surgical Intensive Care Unit (SICU) after cardiac surgery (coronary artery bypass, aortic value replacement and mitral valve replacement), are used. These databases contain information regarding commonly ordered tests such as Complete Blood Count tests (CBC) and Basic Metabolic Panel (BMP) for a large group of patients over time. Physicians usually order these tests as a component of screening, routine evaluation, or serial assessment. These test results, contain a large amount of information used by most physicians during the diagnosis process and patient monitoring. This study creates time series of some components of the aforementioned tests to analyze their behavior during the perioperative and postoperative period. Time series based clusters are developed to determine the similarities among tests results from four different types of patients: patients who had a satisfactory recovery process without any manifestation of complications, patients who experienced complications but survived, viii patients who experienced complications and then died during their recovery and patients who died during the perioperative period. As a conclusion, the time series based clustering techniques were able to identify whether a patient is likely to fully recover from the surgery, but it does not have the power to detect effectively results corresponding to a patient experiencing complications. The development of this methodology provides statistical evidence of the differences among different patterns on patient recovery. It is clear that patients experiencing complications have a steeper drop of test results after surgery, and also a non-stable trend towards normal levels. The appropriate use of the proposed methodology could help to timely anticipate complications in patient condition, improve the comprehensiveness of the assessment of patient condition based on laboratory test results and enhance the utilization of laboratory results databases.

Smart computer algorithms and signal processing techniques have led to rapid development in the field of patient monitoring. Accelerated growth in the field of medical science has made data analysis more demanding and thus the complexity of decision-making procedures. Anaesthetists working in the operating theatre are responsible for carrying out a multitude of tasks which requires constant vigilance and thus a need for a smart decision support system has arisen. It is anticipated that such an automated decision support tool, capable of detecting pathological events can enhance the anaesthetist’s performance by providing the diagnostic information to the anaesthetist in an interactive and ergonomic display format. The main goal of this research was to develop a clinically useful diagnostic alarm system prototype for monitoring pathological events during anaesthesia. Several intelligent techniques, fuzzy logic, artificial neural networks, a probabilistic alarms and logistic regression were explored for developing the optimum diagnostic modules in detecting these events. New real-time diagnostic algorithms were developed and implemented in the form of a prototype system called real time – smart alarms for anaesthesia monitoring (RT-SAAM). Three diagnostic modules based on, fuzzy logic (Fuzzy Module), probabilistic alarms (Probabilistic Module) and respiration induced systolic pressure variations (SPV Module) were developed using MATLABTM and LabVIEWTM. In addition, a new data collection protocol was developed for acquiring data from the existing S/5 Datex-Ohmeda anaesthesia monitor in the operating theatre without disturbing the original setup. The raw physiological patient data acquired from the S/5 monitor were filtered, pre-processed and analysed for detecting anaesthesia related events like absolute hypovolemia (AHV) and fall in cardiac output (FCO) using SAAM. The accuracy of diagnoses generated by SAAM was validated by comparing its diagnostic information with the one provided by the anaesthetist for each patient. Kappa-analysis was used for measuring the level of agreement between the anaesthetist’s and RT-SAAM’s diagnoses. In retrospective (offline) analysis, RT-SAAM that was tested with data from 18 patients gave an overall agreement level of 81% (which implies substantial agreement between SAAM and anaesthetist). RT-SAAM was further tested in real-time with 6-patients giving an agreement level of 71% (which implies fair level of agreement). More real-time tests are required to complete the real-time validation and development of RT-SAAM. This diagnostic alarm system prototype (RT-SAAM) has shown that evidence based expert diagnostic systems can accurately diagnose AHV and FCO events in anaesthetized patients and can be useful in providing decision support to the anaesthetists.

Man has a limited ability to accurately and continuously analyse large amounts of data. Observers are typically required to monitor displays over extended periods and to execute overt detection responses to the appearance of low probability critical signals. The signals are usually clearly perceivable when observers are alerted to them, but they can be missed in the operating environment. The challenge is to develop a computer application that will accumulate information on a variable, or several variables, over time and identify when the trend in observations has changed. In recent years, there has been a rapid growth in patient monitoring and medical data analysis using decision support systems, smart alarm monitoring systems, expert systems and many other computer aided protocols. The expert systems have the potential to improve clinician performance by accurately executing repetitive tasks, to which humans are ill-suited. Anaesthetists working in the operating theatre are responsible for carrying out a multitude of tasks which requires constant vigilance and thus a need for a smart decision support system has arisen. The decision support tools capable of detecting pathological events can enhance the anaesthetist’s performance by providing alternative diagnostic information. The main goal of this research was to develop a clinically useful diagnostic alarm system using two different techniques for monitoring a pathological event during anaesthesia. Several techniques including fuzzy logic, artificial neural networks, control and monitoring techniques were explored. Firstly, an industrial monitoring system called Supervisory Control and Data Acquisition (SCADA) software is used and implemented in the form of a prototype system called SCADA monitoring system (SMS). The output of the system in detecting hypovolaemia was classified into three levels; mild, moderate and severe using SCADA’s InTouch software. In addition, a new GUI display was developed for direct interaction with the anaesthetists. Secondly, a fuzzy logic monitoring system (FLMS) was developed using the fuzzy logic technique. New diagnostic rules and membership functions (MF) were developed using MATLAB. In addition, fuzzy inference system FIS, adaptive neuro fuzzy inference system ANFIS and clustering techniques were explored for developing the FLMS’s diagnostic modules. The raw physiological patient data acquired from an S/5 monitor were converted to a readable format using the DOMonitor application. The data was filtered, preprocessed, and analysed for detecting anaesthesia related events like hypovolaemia. The accuracy of diagnoses generated by SMS and FLMS was validated by comparing their diagnostic information with the one provided by the anaesthetist for each patient. Kappa-analysis was used for measuring the level of agreement between the anaesthetist’s, SMS’s, and FLMS’s diagnoses. In offline analysis both systems were tested with data from 15 patients. The SMS and FLMS achieved an overall agreement level of 87 and 88 percent respectively. It implies substantial level of agreement between SMS or FLMS and the anaesthetists. These diagnostic alarm systems (SMS and FLMS) have shown that evidence-based expert diagnostic systems can diagnose hypovolaemia, with a substantial degree of accuracy, in anaesthetized patients and could be useful in providing decision support to anaesthetists.

Man has a limited ability to accurately and continuously analyse large amounts of data. Observers are typically required to monitor displays over extended periods and to execute overt detection responses to the appearance of low probability critical signals. The signals are usually clearly perceivable when observers are alerted to them, but they can be missed in the operating environment. The challenge is to develop a computer application that will accumulate information on a variable, or several variables, over time and identify when the trend in observations has changed. In recent years, there has been a rapid growth in patient monitoring and medical data analysis using decision support systems, smart alarm monitoring systems, expert systems and many other computer aided protocols. The expert systems have the potential to improve clinician performance by accurately executing repetitive tasks, to which humans are ill-suited. Anaesthetists working in the operating theatre are responsible for carrying out a multitude of tasks which requires constant vigilance and thus a need for a smart decision support system has arisen. The decision support tools capable of detecting pathological events can enhance the anaesthetist’s performance by providing alternative diagnostic information. The main goal of this research was to develop a clinically useful diagnostic alarm system using two different techniques for monitoring a pathological event during anaesthesia. Several techniques including fuzzy logic, artificial neural networks, control and monitoring techniques were explored. Firstly, an industrial monitoring system called Supervisory Control and Data Acquisition (SCADA) software is used and implemented in the form of a prototype system called SCADA monitoring system (SMS). The output of the system in detecting hypovolaemia was classified into three levels; mild, moderate and severe using SCADA’s InTouch software. In addition, a new GUI display was developed for direct interaction with the anaesthetists. Secondly, a fuzzy logic monitoring system (FLMS) was developed using the fuzzy logic technique. New diagnostic rules and membership functions (MF) were developed using MATLAB. In addition, fuzzy inference system FIS, adaptive neuro fuzzy inference system ANFIS and clustering techniques were explored for developing the FLMS’s diagnostic modules. The raw physiological patient data acquired from an S/5 monitor were converted to a readable format using the DOMonitor application. The data was filtered, preprocessed, and analysed for detecting anaesthesia related events like hypovolaemia. The accuracy of diagnoses generated by SMS and FLMS was validated by comparing their diagnostic information with the one provided by the anaesthetist for each patient. Kappa-analysis was used for measuring the level of agreement between the anaesthetist’s, SMS’s, and FLMS’s diagnoses. In offline analysis both systems were tested with data from 15 patients. The SMS and FLMS achieved an overall agreement level of 87 and 88 percent respectively. It implies substantial level of agreement between SMS or FLMS and the anaesthetists. These diagnostic alarm systems (SMS and FLMS) have shown that evidence-based expert diagnostic systems can diagnose hypovolaemia, with a substantial degree of accuracy, in anaesthetized patients and could be useful in providing decision support to anaesthetists.

Thesis (M.Eng. (Hons.)) -- University of Western Sydney, 2001. / Bibliography : leaves [185]-195 (v. 1).

Submitted by Emanuel Varela Cardoso (emanuel.varela@ufcg.edu.br) on 2018-05-04T17:30:47Z No. of bitstreams: 1 ALDENOR FALCÃO MARTINS – DISSERTAÇÃO (PPGEE) 2015.pdf: 15602466 bytes, checksum: 608173ca67ff68da8ae45b321aa82204 (MD5) / Made available in DSpace on 2018-05-04T17:30:47Z (GMT). No. of bitstreams: 1 ALDENOR FALCÃO MARTINS – DISSERTAÇÃO (PPGEE) 2015.pdf: 15602466 bytes, checksum: 608173ca67ff68da8ae45b321aa82204 (MD5) Previous issue date: 2015-04-24 / Segundo o IBGE em 2013, o Brasil apresentava 13% de sua população composta por pessoas acima de 65 anos, somado a isto, o estilo de vida das sociedades ocidentais tem facilitado o aparecimento de doenças crônicas cada vez mais cedo. A premissa é que tornemos mais eficiente a utilização do nosso sistema de saúde, pois este é um recurso escasso. Uma forma de melhorar esta eficiência é assegurar que os tratamentos prescritos serão devidamente seguidos. Quando o paciente se encontra no hospital uma gama de recursos monitora a saúde do paciente oferecendo acompanhamento seguro na eventualidade de um desvio, alertando e armazenando as informações do paciente no decorrer de suas atividades. Um recurso que ajuda no acompanhamento deste paciente é a monitoração remota do paciente, que possibilita que sensores enviem a informação da condição de saúde do paciente e permitam o acompanhamento do mesmo. Sistemas Físicos Cibernéticos (SFC) são entidades computacionais ligadas em rede que operam entidades no mundo físico de maneira cooperativa. Tais sistemas podem ser utilizados em redes de monitoramento remoto de pacientes com o fim de apresentar e ajustar o tratamento de acordo com as recomendações do médico. Este trabalho propõe um passo na direção da autonomia, que permita uma melhor qualidade de vida ao paciente crônico, permitindo que situações conhecidas e dentro de um regime de segurança previamente determinado pelo médico sejam ajustadas. Este trabalho apresenta uma proposta de um Sistema Físico Cibernético (SFC), que permite que adequações ao tratamento previamente elaboradas sejam colocadas em planos de tratamento por meio de agentes inteligentes e de planejadores SAT e sejam disponibilizadas de acordo com a mudança da condição do paciente, através de uma rede monitoramento do paciente, seguindo padrões estabelecidos para dispositivos médicos utilizados em casa que disponibiliza o tratamento ao paciente. O modelo proposto é indicado para o acompanhamento em casa de doenças crônicas através de um coletor central responsável pela coordenação do acompanhamento do paciente. / According to IBGE in 2013 13% of the population had 60 or more years old. As the national population ages, we have to move towards more efficient use of SUS. A way to improve is the closer followup of patient’s evolution by the healthcare professional. At the hospital the patient has access to a set of equipments and expert knowledge capable to correct the treatment path. From this scenario it is easy to imply the need for a change, the current status quo is unbearable financially and cumbersome for patient and doctor routines. A resource that helps is the remote patient monitoring (RPM) , where sensors provide the latest information about patient’s health status and are able to suggest a course correction on the treatment path. A Cyber-Physical System (CPS) is a network of interacting computational entities with physical inputs and outputs that work together towards a goal. A CPS can be part of a RPM in order to present and adjust the treatment according to the healthcare professional recommendations. This work offers a framework for situations where the medical expert knowledge is complete allowing changes on the treatment path be adjusted with minimum risk. Our proposal to deal with the problem is a CPS based remote patient monitoring network where a model for the system is developed based on Multiagent Agent System (MAS) and automatic planning system based on SAT, allowing safe and minimal course correction on treatment paths already set for a patient. This proposal operates through a central hub element responsible to coordinate the followup of the patient.

Complex patient are becoming more and more of a challenge to the health care system given the amount of care they require and the amount of documentation needed to keep track of their state of health and treatment. Record keeping using the EHR makes this easier but mounting amounts of patient data also means that clinicians are faced with information overload. Information overload has been shown to have deleterious effects on care, with increased safety concerns due to missed information. Patient record summarization has been a promising mitigator for information overload. Subsequently, a lot of research has been dedicated to record summarization since the introduction of EHRs. In this dissertation we examine whether unsupervised inference methods can derive patient problem-oriented summaries, that are robust to different patients. By grounding our experiments with HIV patients we leverage the data of a group of patients that are similar in that they share one common disease (HIV) but also exhibit complex histories of diverse comorbidities. Using a user-centered, iterative design process, we design an interactive, longitudinal patient record summarization tool, that leverages automated inferences about the patient's problems. We find that unsupervised, joint learning of problems using correlated topic models, adapted to handle the multiple data types (structured and unstructured) of the EHR, is successful in identifying the salient problems of complex patients. Utilizing interactive visualization that exposes inference results to users enables them to make sense of a patient's problems over time and to answer questions about a patient more accurately and faster than using the EHR alone.