optima4BP Predictive Modeling

optima₄BP simulates the successive steps of the pharmacological intervention clinical reasoning process while engaging the provider and patient in shared HTN management.

European Society for Cardiology, 2018. Digital Health: Artificial Intelligence: Decision Support & Voice Interaction. Best Technology Award.

Abstract

1 Introduction

A novel integrated approach to hypertension (HTN) management has been deployed at University of California San Francisco to reduce the proportion of patients with uncontrolled HTN. The solution is comprised of: (a) a clinical reasoning analytic intelligence that identifies the on-demand need for and recommends an HTN pharmacological intervention optimization step; (b) an interoperability data exchange solution responsible for data collection and HTN treatment recommendation push to the treating physician; and (c) a patient web-service to monitor HTN control, complete periodic surveys, while engaging the patient in the management of HTN and reduction of cardiovascular (CVD) risk.

2 Methods

The first component of the integrated solution is the clinical reasoning analytic intelligence. It involves multiple computational frameworks used to convert the explicit steps of human clinical reasoning into a mathematical algorithm (optima4BP). The second component, the interoperability data solutions is developed to support the following functions: (a) collect → map → structure patient data from various sources: electronic health record (EHR; Epic^Ⓡ), BP remote monitoring (OMRON^Ⓡ), and survey management web-service; and (b) delivery of a pharmacological intervention recommendation into EHR In-Basket with integration into the clinical work flow. The third component, the patient web-service was established to support the following features: (a) progress of BP to goal; (b) CVD risk report on progress and steps to reduce the eventuality of an adverse event (heart attack, stroke or myocardial infarction); and (c) survey completion and survey answers access & review hub.

3 Results

The solution has been fully developed. It is presently used by the Division of General Internal Medicine at University of California San Francisco in an investigative clinical trial (n=300).

.

4 Conclusions

It is possible to build an integrated solution for HTN pharmacological intervention management that (a) is clinical reasoning based; (b) remotely collects data and delivers treatment recommendations within the existing clinical care flow; and (c) engages the patient in his/her HTN management.

optima4BP: HTN Management

In between office visits, optima4BP successfully optimizes pharmacological interventions responsible for blood pressure reduction towards clinical goal.

National Heart Lung and Blood Institute, 2018. Award # R44 HL132622-01A1.

Abstract

1 Introduction

Hypertension (HTN) is the leading risk factor for cardiovascular disease. A delay greater than 1.4 months in HTN medication [pharmacological intervention] optimization is associated with an increased risk of stroke, heart failure or myocardial infarction. Clinical trials have demonstrated that active HTN medication optimization to blood pressure (BP) goal reduces the incidence of stroke by 35-40%, myocardial infarction by 15-25%, and heart failure by up to 64%. optima4BP addresses pharmacological intervention management by identifying the on-demand need for a drug treatment change and recommending the most appropriate drug optimization for safe and effective BP lowering towards clinical goal. optima4BP’s integration into the clinical care flow and effectiveness in lowering BP to clinical goal will be studied in a population of 300 patients, managed by the Division of General Internal Medicine at University of California San Francisco.

2 Methods

Interoperability: optima4BP real-time monitoring and on-demand recommendation for optimization of pharmacological interventions requires interoperability with (a)electronic health record (EHR; Epic^Ⓡ), (b) BP remote monitoring (OMRON^Ⓡ); and (c) a survey management web-service. Effectiveness: 300 patients are being randomized into either the optima4BP interventional arm or usual care. The optima4BP patients are managed remotely and in real time, while patients assigned to the usual care arm follow the current in-office practice for HTN management. Primary Outcome: Reduction in the systolic blood pressure (BP) to a goal of ≤130 mmHg. The follow-up period is 12 months [ongoing].

3 Results

optima4BP was successfully integrated with the data storage platforms required for on-demand assessment and recommendation of a change in pharmacological intervention. A “how to” interoperability built-up protocol was developed to streamline future Epic^Ⓡ implementations.

The efficacy in BP lowering to clinical goal is currently being studied [ongoing investigation].

4 Conclusions

optima4BP demonstrates that it can interoperate with a multitude of data storage platforms supporting its real-time monitoring and optimization of pharmacological interventions for patients with uncontrolled HTN.

optim4BP efficacy in lowering BP to clinical goal is currently being studied.

Big Data-Driven Forecasting for HTN Management

A flexible multi-algorithm predictive modeling engine provides a highly personalized and on-demand pharmacological intervention optimization solution to hypertension (HTN) treatment management.

National Heart Lung and Blood Institute, 2018. Award # R43 HL140624-01.

Abstract

1 Introduction

Larger evidence is needed to translate segregated investigative outcomes into clinical practice. In- silico modeling relying on advanced statistical and cloud-computing approaches for distributed big data analysis are emerging. In the context of hypertension (HTN), wide validation across different independent cohorts and convincing metrics of performance are paramount to establish validity and efficacy of artificial intelligence (AI) predictive models.

The scope of this project is to understand and personalize the predictive efficacy of candidate AI solutions to improve the safety and efficacy of a pharmacological intervention optimization for HTN treatment management compared to usual care.

2 Methods

The approach is structured as a feasibility study composed of several training and validation stages. The selected big-data set for this investigation uses the Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial (ALLHAT) to investigate the strengths and weaknesses of various AI predictive models in the management of HTN through personalized optimization of pharmacological interventions.

3 Results

While each individual candidate AI predictive model had limitations in fully addressing the multi-stage and multi-faceted problem of optimizing HTN pharmacological management, each model performed optimally under a specific set of patient health status parameters. [ongoing investigation]

4 Conclusions

An AI selector engine allows for improving the on-demand predictive efficacy of a pharmacological treatment recommendation, based on changes in patient health status.

optima4heart: CVD Management

Integrating on-demand pharmacological interventions recommendations and associated transfer of care, optima4heart is positioned to improve the management of complex multi-disease cardiovascular (CVD) conditions.

National Heart Lung and Blood Institute, 2018. Award # [under review].

Abstract

1 Introduction

Uncontrolled hypertension (HTN), heart failure (HF), coronary artery disease (CAD), and atrial fibrillation (AFib) represent the primary drivers of the CVD “epidemic”. The American College of Cardiology (ACC) recommends two critical directions for improvement: (1) PCP/Cardiologist pharmacological interventions that drive effective treatment decisions, and (2) Standardization of Care Flows for cross-communication of treatment protocols amongst providers. Aim: Reduce the incidence of HF, stroke, myocardial infarction (MI)] or death.

2 Methods

optima4heart is being built to recommend: (1) a pharmacological intervention and (2) needed changes in team-based transfer of care [primary care provider (PCP) ⇔ Cardiologist] for clinicians managing patients with CVD. optima4heart is clinically applicable through a patient-tailored analytic intelligence (AI) augmented by a dashboard user-interface displayed into the electronic health record (EHR). This feasibility project focuses on expanding the disease-agnostic predictive modeling platform from managing HTN (optima4BP) to addressing the combination of HTN and HF.

3 Results

The integration of HF was performed by updating the same data and information modules specific to HTN management with the HF content, derived from national clinical guidelines and resources. A protocol for the transfer was developed and approved by the participating clinical divisions at University of California San Francisco. A selector module is being developed to prioritize the pharmacological intervention recommendation in response to on-demand patient needs. [ongoing development]

4 Conclusions

It is possible to develop a decision support analytic intelligence that seamlessly addresses pharmacological interventions as well as transfer of care for CVD management.