T15 Overview

Home	About Us	Project Overview	Project Management	Project Documentation

Project Introduction

Sponsor Introduction

Khoo Teck Puat Hospital (KTPH) is a 590-bed general and acute care hospital, managed by Alexandra Health System. Alexandra Health, together with the School of Information Systems (SIS) at the Singapore Management University (SMU) have established a partnership to work together to demonstrate fresh and better ways to serve and satisfy patients whenever they are interacting with the Alexandra Health system. Through this partnership, a joint mechanism known as the “T-Lab” has been established that enables students, staff and faculty of SMU’s School of Information Systems (SIS) to team with professionals from Alexandra Health to work on a continuing series of projects to improve service delivery, quality, productivity and experience.

This partnership also provides for academic exchanges where SIS faculty will play an active role in research studies and consultancy for KTPH. Students will be able to tap the adjunct faculty’s extensive experience for insights into the healthcare operations and service delivery. In turn, Alexandra Health staff will benefit from interacting with SMU faculty on ways to improve processes and operations.

Project Introduction

The project is part of an ongoing effort to adopt data visualization to monitor public health by KTPH. Based on data from health screening and KTPH alignment programs, a dashboard can be constructed to assist health officers of KTPH to observe public health conditions, single out unhealthy individuals, and monitor their health progress. Additionally, the dashboard serves as a means to evaluate effectiveness of KTPH alignment programs to improve public health, and provide insights that can refine these programs to better target the population in future.

Motivation

KTPH manages a huge amount of data from public health screening. These data potentially contains valuable insights about individuals’ health relative to their lifestyles and medical background, but knowledge is not being extracted effectively from these data. At the same time, KTPH initiates many programs to promote public health, but there is room for improvement, especially when there is a lack of a data-driven method to make decisions during the execution process. Last but not least, KTPH currently has no means to measure the penetration rate of their past alignment programs, and likewise, no means to fine-tune future programs to achieve higher penetration rate. Data visualization is thus adopted by KTPH in collaboration with T-Lab as an ongoing effort to derive insights from their data-rich operations.

Project Objectives

This project is a follow-up of an IS480 project by team Cinquefoil. Our aim is to improve the KTPH dashboard by adopting a richer set of visualization techniques, so as to enable a more user-centric data querying and discovery process. KTPH users will be able to use the dashboard to identify unhealthy individuals of the population, the areas they are in, take appropriate actions and monitor the results of such actions.

As such, the objectives of our analytics project consist of the following:

• To visualize effectively the current health condition of the public across various regions of Singapore
• To allow health officers to track the health progress of individual at risks
• To assist health officers in monitoring the penetration rate of KTPH alignment programs targeted at the general public
• To allow users to interact with visualizations, thereby forming their own query and arriving at their own findings

Data

The data is provided by KTPH Health Population team, consisting of 6,744 patient records with the following attributes:

Demographics

• Gender
• Age/Age group
• Race
• Education level
• Occupation
• Home address

Health measurements

• Weight
• Height
• waist
• BMI
• Glucose measure
• Cholesterol level
• Blood pressure
• Systolic
• Diastolic
• Instances of strokes, heart attacks, diabetes
• Other health measurements

Lifestyle

• Smoking habit
• Stress level
• Exercise
• Diet

Intervention records

• Nurse intervention
• Doctor outcome
• Doctor revisit
• Follow up at clinics

Sample dataset

Methodology

Technology

As KTPH prefers a versatile tool that Health Population team can just use without the need for complex setup and installation, d3.js was used to develop a web application in Apache server. D3.js is a JavaScript library for developing visualizations on the web. D3.js is coded in Javascript and use SVG objects for visualization, which allows for more flexibility. SVG objects are also scalable and support visualization on mobile devices. It is convenient as a JavaScript library can run on all modern browsers without users having to install additional software.

In addition to that, we propose to explore dc.js library which is a closely related tool to d3.js. Dc.js allows effective cross-filtering across different charts and has improved performance compared to d3.js. This addition will boost the story-telling capability of the current dashboard and allow users to formulate their own queries in the process of data discovery.

Visualization

Treemap

Treemap is a powerful tool to simultaneously show the big picture, comparison of related items and allow navigation to the details. One important aspect of healthcare visual analytics is the ability to drill-down to details for further investigation. Using treemap to show the health indicators as the example below can provide a bird-eye’s view for users, such that they can observe patterns among the indicators before drilling down to study the details. This technique will be used in Screening Result module.

Parallel Coordinates

This technique can be used to analyze multiple clinical variables. Each axis represents one numerical clinical variable (eg. BMI, cholesterol level, systolic and diastolic levels). Users can look at the lines and quickly spot the sample line that is outside the normal range. A separate line representing national average could be used as a benchmark; alternatively, expert-defined healthy level for each indicators could also be used. This technique should be used in the intermediate level of drill-down so that the number of lines does not get too large and clutter the chart.

Chord Visualization

This chart is to study the association between clinical variables. More often than not, clinical variables are likely to have some relationship with one another, for example, a patient with overweight level of BMI is more likely to have high cholesterol level and higher risk of diabetes. Chord visualization allows data exploration that reveals such a pattern, and potentially helps to identify individuals at risk of diseases like diabetes based on their other health indicators.

Funnel Plot

Funnel plot is essentially a scatter plot with 2 sets of boundary lines: one set for 95% confidence and one for 99.8% confidence. The points that lie outside the boundaries will be highlighted as non-random variations that are extremely rare and should be examined more closely, compared to points that lie inside the boundaries that are random variations that happen by chance. In our case, the data points will represent households, x-axis is %population above a certain age and y-axis is %population above a certain age that responds to alignment program. Thus this chart can show penetration rate of KTPH health initiatives to improve public health.

Geospatial Intelligence

The current version does not show the percentage of households participating in KTPH health initiatives; instead it shows the absolute number of households reached out. We will modify the current OpenStreetMap view of the module to reflect the percentage and penetration rate by regions.

Scope of Work

The visualization should allow KTPH users in Health Population team to see an overview of public health condition, based on screening results, and then drill down to region and patient group level to further investigate the various factors that contribute to the status quo. Users can also examine the possible correlations between said factors.

The components to be examined and improved thus are:
Screening Result Module

• Stratification & Visual Presentation of Health Screening Results

Health Classification Module

• Health Classification
• Risk Analysis for Disease
• Summary of Unhealthy Screening Results

Geospatial Intelligence Module

• Public Health Screening Penetration Rate
• Public Health Status Ratio

Repeat Analysis Module (secondary)

• Flow Analysis of Population Health Screening Results
• Trend Analysis of Key Health Indicators

Patient Journey Module (secondary)

• Individual Resident Progress View
• Temporal Event Sequence Analysis

References

Reddy, C. (n.d.). Introduction to Visual Analytics and Medical Data Visualization. In Healthcare data analytics
Rowell, K. (2013, September 6). Category Archives: Design Basics. Retrieved January 10, 2015, from http://www.healthdataviz.com/category/design-basics/ (n.d.). Retrieved from http://vizhub.healthdata.org/gbd-compare/england