Data mining in data analytics

Assignment : answer Chapter 12 questions; at least one

Page; cite textbook

Please see chapter readings from textbook below

 

Chapter 12 HEALTHCARE  INFORMATION       

Chapter 12 questions

1.Justify the need for data mining in data analytics.

2.   Examine what it means to aggregate data. Identify some of the sources of data for aggregation. Determine how the interpretation and evaluation of aggregated data support the strategic uses of health information.

 

 

Aggregate Data

Aggregate data is when individual, comparative, or other multiple sources of data are compiled and analyzed to draw conclusions about a specific topic or area. For example, in a focus group study, data, observation, and interview data were compiled into an aggregate format so that none of the individuals in the multiple healthcare organizations that participated could be identified in any way. Varying methods and skills of leadership among HIM leaders and facilities were compared and contrasted in order to generate conclusions. However, since the focus group sample was small, not all the conclusions could be generalized (Sheridan et al. 2016). In fact, any data compiled from samples of data have limitations since the sample of data may not accurately reflect the characteristics across that entire population. One way to reduce this is to compare the sample’s demographic characteristics to the population’s demographics (if this information is available); if the characteristics prove similar, it increases the reliability of the sample data.

 

 

 

 

 

 

 

 

 

 

HITT 1301 CHAPTER 12

Health Information Management Technology,

An Applied Approach

Nanette Sayles, Leslie Gordon

 

Copyright ©2020 by the American Health Information Management Association. All rights reserved.

Except as permitted under the Copyright Act of 1976, no part of this publication may be reproduced,

stored in a retrieval system, or transmitted, in any form or by any means, electronic, photocopying,

recording, or otherwise, without the prior written permission of AHIMA, 233 North Michigan Avenue,

21st Floor, Chicago, Illinois 60601-5809 (http://www.ahima.org/reprint).

 

ISBN: 978-1-58426-720-1

AHIMA Product No.: AB103118

 

 

 

 

Healthcare information is used to monitor the quality of patient care, conduct medical research, and accurately reimburse healthcare organizations. Healthcare information is based on personal health data about individuals primarily for ­provider use in the management of patient care. Data collection techniques include traditional methods such as paper health records as well as eHealth tools such as templates. “A template is an EHR documentation tool utilized for the ­collection, presentation, and organization of clinical data elements” (Buttner et al. 2015). The sources of health information include the healthcare provider through documentation in the health record and the individual through the use of a personal health record. A personal health record (PHR) is a record created and managed by an individual in a private, secure, and confidential environment. The personal health record will be covered later in this chapter. In addition, the federal incentives for the adoption of the electronic heath record (EHR) have progressed healthcare information exchange, including returning a patient care summary to the patient. Databases of healthcare information collected or maintained by healthcare providers, institutions, payers, and government agencies are of great importance to those who use them; for example, researchers or public health agencies. These databases are used for administrative purposes, including determination of payment for services provided, measurement of quality performance indicators, and research.

 

Per the Federal Health IT Strategic Plan for 2015-2020, the benefits of electronic health information include lower healthcare cost, increased healthcare quality, improved population health, and an improvement in consumer engagement. The Federal Health IT Strategic Plan is illustrated in figure 12.1.

 

Figure 12.1 Strategies to achieve health IT goals

Source: ONC 2014a

 

With the implementation of the EHR and the changes that result, the roles and career options for health information management (HIM) professionals is growing. Some of the new roles include data analytics, consumer engagement, and health information exchange (HIE). This chapter discusses HIE information from the perspective of data analytics and explores the strategic uses of health information. In addition, the consumer’s link to healthcare information—specifically their needs for information, ease of access, navigational tools, telehealth, and PHRs—is described. The various aspects of sharing and exchanging healthcare information are also addressed.

 

Role of Data Analytics in Healthcare Information

Data are needed to arrive at information. Health data are not health information until they are interpreted, evaluated, and appropriately displayed (RWJF 2015). The difference between data and information is described in chapter 3, Health Information Functions, Purpose, and Users. Data analytics is the science of examining raw data with the purpose of drawing conclusions about that information. For example, data analytics can help hospitals with staffing by predicting the number of patients treated at a healthcare organization each month. The raw data examined in this example are admissions data, such as admissions records, rates, and patterns, which are analyzed over a period of time. Data analytics of admissions data can lead to the development of a web-based interface that enables physicians, nurses, and hospital administrators to forecast visits and admission rates for the future (Sreenivasan 2018).

The role of data analytics depends on the type of data being captured, reviewed, and used for the purpose of turning them into healthcare information. Multiple types of data exist, two of which—administrative and clinical—are further explained in the next section. If the data are of a clinical nature, then the analytics revolve around the contents of the health record. Clinical data could include elements such as lab values, number of patients with pneumonia, and so on. Administrative data are focused on other components such as financial data. A type of data analytics that uses clinical data is a clinical decision support (CDS) system. A CDS is a type of data analysis since it takes information from more than one source and provides an avenue for clinicians to make observations and decisions. “Clinical decision support provides clinicians, staff, patients or other individuals with knowledge and person-specific information, intelligently filtered or presented at appropriate times, to enhance health and healthcare” (ONC 2013).

 

Clinical data about an individual can also be combined with clinical data from other individuals to form population-based healthcare data. The resulting information may be used to improve the health of the public. For example, the occurrence of measles in one town could be combined with measles occurrence in a state or a region and that information could then be communicated on a ­national level if the rate of measles in children has increased from previous years. Analytics has the potential to play a role in leveraging data to improve healthcare quality and patient outcomes. For example, the data compare the health of a group from one region or state to another. The following is an introduction to analytics, its tools, and the knowledge areas for HIM professionals in data analytics.

 

Introduction to Analytics

There are different types of analytics. Descriptive analytics answers the question “what happened,” diagnostic analytics answers the question “why did it happen,” predictive analytics answers “what will happen,” and prescriptive analytics answers “how can we make it happen” (Laney et al. 2012). To further illustrate for clinical data analytics, descriptive analytics could be centered on the increase in the incidence of Legionnaires’ disease in individuals 65 years and older in a specific state over a five-year period of time. Diagnostic analytics would review the why of increased rates of Legionnaires’ disease. For predictive analytics, once the why is found, it could be extrapolated that an increase will be seen in other states if certain conditions are found. Using this same situation, prescriptive analytics would examine ways to reduce the potential rate of increase of Legionnaires’ disease in individuals over age 65 even if certain conditions (as found in the diagnostic phase) occur.

Analytics involves acquiring, managing, studying, interpreting, and transforming data into useful information. Types of data include clinical, financial, and operational data and the types of analytics include healthcare data analytics and clinical data analytics. Healthcare data analytics is the practice of using data to make business decisions in healthcare, whereas clinical data analytics is the process by which health information is captured, reviewed, and used to measure quality of care provided. What data are involved, the consumer of the information, and the decision the analysis supports influences the analytic process and choice of tools. However, there are certain steps that occur to prepare healthcare data for data analysis. The first step is data ­capture, which helps ensure the data needed are available and that the data are correct. Data collection is discussed later in this chapter. The second is data provisioning, which ensures that the data are in a format that can be manipulated for data analysis. For example, in the data field gender, male might be “1” and female “2.” Data analysis, where data are interpreted, is the final stage of transforming raw data into meaningful analytics.

 

Analytics Tools

The amount and types of data available for analysis have increased as more data are available electronically. In addition, as technology advances, the various tools available to perform analytics allow for new ways to study and present the data. A few of the more common tools are those used for visualization, to report on process measures, to capture the data, and for extracting and examining data from a database.

 

Data Visualization

Data visualization is the presentation of data using a graph, diagram, or chart. The graphic display of data can help the viewer understand the data trend. For example, it can identify areas that need action, such as addressing a decline in the number of patients or an increase in the infection rate. Types of data visualization tools include tables, charts, and graphs. Choosing one visualization method over another can mean the difference between correct or incorrect data representation and drawing an accurate or erroneous conclusion. For example, tables display exact values whereas graphs show trends.

 

Following established guidelines for data visualization results in the delivery of a clear message. Those overall guidelines for creating any visual presentation, including the following:

 

Understand the data

Evaluate the information to communicate and the way it should be visualized

Define your audience and examine how they process visual information

Display the intended information to the appropriate audience in the clearest, simplest form (SAS 2018)

Tables are used to organize quantitative data or data expressed as numbers. Charts (such as pie charts and bar charts) and graphs (such as line graphs) are appropriate when presenting relationships. For example, in figure 12.2 the first pie chart shows percentages that add up to more than 100 percent, while percentages in the second chart are a part of the whole and add up to 100 percent. Each tool has specific features to keep in mind when depicting the data. For more information on presenting statistical data using tables, charts, and graphs, see chapter 13, Research and Data Analysis.

 

Figure 12.2 Poor and improved data display

Source: ©AHIMA.

 

Figure 12.2 provides an example of a poor and an improved pie chart display.

 

Dashboard

The dashboard is a data analytics tool that is a computerized visual display of specific data points. Typically, a dashboard focuses on a process and the rate of achievement. A dashboard is different from a scorecard. A scorecard, which can also be a computerized visual display, focuses on outcome or goal achieved, such as money raised for an event or cause. Both a dashboard and a scorecard can involve key indicators. A key indicator is a quantifiable measure used over time to determine whether some structure, process, or outcome in the provision of care to a patient supports high-quality performance measured against best practice criteria. For example, a key indicator could monitor death rates or infections. Chapter 18, Performance Improvement, discusses scorecards in more detail.

 

Health information management professionals use dashboards to monitor a number of indicators to improve performance and meet quality goals such as reducing the infection rate. To track the process measure over time, metrics (way to measure something) or benchmarks are established. Information is displayed on a dashboard to show the status of predetermined benchmarks. Often dashboards use color such as red, yellow, and green in a stoplight scheme. Similar to a traffic light, red means stop and go back, yellow means caution, and green means all good. Dashboards provide early warning signals and alert the manager to areas in need of attention.

 

For example, a recent HIM trend is instituting a clinical documentation integrity (CDI) program. Since this is not a small undertaking, dashboards can assist in measuring whether the program is successful. A monthly dashboard might show the number of clarifications requested by a CDI specialist that impacted a diagnosis-related group based on a benchmark. The dashboard would show green if the metric is met, yellow if it is in progress or halfway met, and red if the metric is below standard.

Dashboards are also used to manage revenue cycle management performance. For example, the Healthcare Financial Management Association (HFMA) has a web-based application called MAP App for use by healthcare providers to check revenue cycle performance and evaluate against provider peer groups (HFMA 2019). The HFMA’s key performance indicators can be used to track, monitor, and improve revenue cycle performance.

 

Data Capture Tools

Data capture is the process of recording data in a health record system or database. A database is an organized collection of data, text, references, or pictures in a standardized format, typically stored in an information system for multiple applications. A database contains a large amount of data, often from multiple sources. Additionally, a database can provide comparisons using tools from within the database software. One of the most common healthcare databases is the relational database, which stores data in predefined tables consisting of rows and columns. Healthcare providers as well as patients may be the source of the data. There are several tools available for acquiring health-related data. Historically, data capture into a health record was via written notes or traditional voice dictation that was transcribed and typed into a paper report. Another method for data capture is scanning documents into electronic document management systems that create a picture of the scanned document, making it accessible electronically. Devices also include traditional keyboard or touch screen handheld computers or patient-generated health data devices (discussed later in this chapter). When the software application is run on a mobile platform such as a tablet or cellular phone, system and application software (often referred to as apps) is needed for the device to function and perform the desired tasks.

Electronic healthcare data capture is a fundamental function of the EHR (HealthIT 2018). The EHR is an information system with several components and data capture is an element in each component. The components include source systems (such as the laboratory information system), core clinical EHR systems (such as point-of-care charting), supporting infrastructure such as ­human–computer interfaces, and connectivity systems such as personal health records (Amatayakul 2013, 16–19). In point-of-care charting, the ­information is entered into the health record at the time and location of service. Nurses entering data using a tablet as they conduct patient assessments at the bedside is an example of point-of-care charting.

A human–computer interface is the device used by humans to access and enter data into an information system. A number of mobile devices are used for data entry into point-of-care charting systems. These handheld devices include tablet computers, laptop computers, and smartphones. These devices often contain built-in methods to facilitate the capture of structured data such as predefined or custom-built templates or forms with drop-down menus and point and click fields and word macros. These devices exist to make data collection easier.

The outcome of point-of-care charting can be unstructured or structured data. Unstructured data are nonbinary, human-readable data, whereas structured data are binary, machine-readable data in discrete fields. An example of unstructured data is free text that describes the patient’s description of his or her condition. An example of structured data is using checkboxes to indicate patient symptoms. Structured data has many advantages over unstructured data when it comes to data analytics and health information exchange. Structured and unstructured data are covered in more detail in chapter 6, Data Management.

The structured data’s entry fields and the potential entries in those fields are controlled, defined, and limited, resulting in discrete data. Discrete data represent separate and distinct values or observations; that is, data that contain only finite numbers and have only specified values. Stored in databases and data warehouses, these standardized data are available in a usable and accessible form. However, physicians and other healthcare providers may express frustration when limited to recording only certain data in specific fields. While a set format ensures consistency and provides standard meaning, it may limit details considered important by clinicians.

 

When considering methods for EHR data capture, follow these best practices:

Collect data at the point of care directly from the patient

· Facilitate data accuracy using guidelines for documentation per governmental and other stakeholder standards

· Create and evaluate data integrity policies

· Establish information governance guidelines (AHIMA 2019)

Additionally, key areas such as patient identification, the use of documentation templates, copy and paste functionality, making amendments and corrections, and the incorporation of data captured in other areas of a healthcare organization not networked to the EHR such as outpatient services should be part of the role of HIM (AHIMA 2019).

 

Data capture may also occur with word processing software. The word processing copy and paste functionality in an EHR system must be carefully monitored and limited or prohibited to prevent data quality issues. Examples of data quality issues include copying outdated information or copying content from one patient to another that does not apply. Measures for preventing data quality problems include the following:

· Clearly label the information as copied from another source

· Limit the ability for data to be copied and pasted from other information systems

· Limit the ability of one author to copy from another author’s documentation

· Allow a provider to mark specific results as reviewed

· Allow only key, predefined elements of reports and results to be copied or imported

· Monitor a clinician’s use of copy and paste (AHIMA Work Group 2015)

For additional information on the copy and paste function and risks associated with it, refer to chapter 3, Health Information Functions, Purpose, and Users.

 

Two other technologies—speech recognition (speech-to-text) and natural language processing (NLP)—provide yet another way to acquire health data. NLP is a technology that converts human language (structured or unstructured) into data that can be translated and then manipulated by computer systems. Integration of these technologies within the EHR can result in the provision of clinical information needed by providers to inform decision-making.

Back-end speech recognition (BESR) is a specific use of speech recognition technology (SRT) in an environment where the recognition process occurs after the completion of dictation by sending voice files through a server. In BESR, an employee edits or corrects the dictation. Front-end speech recognition (FESR) is a process where the provider speaks into a microphone or headset attached to a PC and upon speaking, the words are displayed as they are recognized. The physician corrects misrecognitions at the time of dictation. Use of FESR integrated with an EHR provides the best outcome, as the provider is able to respond to prompts from the EHR resulting in more complete, accurate, and timely documentation (AHIMA 2013). Templates and macros are also tools used with SRT to capture data. Macros are used by transcriptionists to insert content into a transcribed document with just a few keystrokes. For example, the transcriptionist might create shortcuts to insert commonly used phrases or other content. As the output of SRT is digital text, combining it with NLP results in the conversion of the text or any free text narrative into data that can be translated and then manipulated by computer systems. Once transformed, it becomes searchable along with other structured data.

 

Data Mining

Data mining is the process of extracting and analyzing large volumes of data from a database for the purpose of identifying hidden and sometimes subtle relationships or patterns and using those relationships to predict behaviors. It is a key piece of analytics and of the knowledge discovery process. There are several knowledge discovery process models such as the Knowledge Discovery in Databases (KDD), Sample, Explore, Modify, Model, Assess (SEMMA), and Cross-Industry Standard Process for Data Mining (CRISP-DM) as well as hybrid models. Each has defined steps, with data mining being one of them.

The available data for analytics strategy and mining can come from EHRs and various databases such as a clinical data repository and clinical data warehouse. A clinical data repository is a central database that focuses on clinical information. The clinical data warehouse allows access to data from multiple databases and combines the results into a single query and reporting interface. Specific applications of data mining methods are customized for certain uses of the extracted data. For example, data mining may be used to extract clinical data directly from the EHR for the purpose of compiling content for reporting clinical quality measures. The clinical data warehouse lends itself to data mining as it encompasses multiple sources of data. The varying sources of data that feed a clinical data warehouse may include data sets, clinical data repositories, a case-mix system, laboratory information systems, or a health plans database. The data in the clinical data warehouse depends on how they will be used. For example, if the clinical data warehouse is going to be used to determine what treatment is most effective, then data would need to include data that would support that research. In this case, the clinical data warehouse might include blood pressure, test results, symptoms, treatments, and more. In the clinical data warehouse, the data from these sources can be “mined” to identify and implement better evidence-based solutions.

 

Systematically analyzing the data uncovers hidden patterns or trends for use in predicting behaviors. The information discovered from data mining databases aids clinical research. For example, data mining could be used to detect early signals of potential adverse drug events. Other data mining applications are used for the evaluation of treatment effectiveness, management of healthcare, customer relationship management, and detection of fraud and abuse (Koh and Tan 2005).

 

HIM Professionals and Analytics

Analytics start with data and HIM professionals, with their understanding of healthcare data, help ensure correct and accurate data are captured. HIM professionals are also proficient in business operations and clinical processes. However, data analytics require going beyond these into competencies such as business intelligence (see chapter 6, Data Management), database administration, inferential and descriptive statistics (see chapter 13, ­Research and Data Analysis), health information technology (see chapter 11, Health Information Systems), and project management (see chapter 17, Management) (Sandefer et al. 2015).

 

AHIMA lists the following knowledge topics as important for data analytics:

· Clinical, financial, and operational data

· Understanding of database queries (such as structured query language [SQL])

· Understanding statistical software

· Data mining

· Quality standards, processes, and outcome measures

· Risk adjustment

· Business practices (for example, workflow or payer guidelines)

· Medical terminology

· Healthcare reimbursement methodologies

· Classification systems

· Source data

· Qualitative and quantitative analysis (AHIMA 2015a)

 

Strategic Uses of Healthcare Information

There are many reasons to collect data and turn it into information, including administrative uses such as claims submission, revenue cycle management, meeting quality measurement reporting requirements, assessing health status and outcomes, and performing clinical research. As health information technology (IT) systems evolve, the ability to aggregate the collected data improves and the information from it better supports strategic analytics and organizational decision-making. Through interpretation and evaluation of aggregated data from a variety of sources, development of strategies to improve patient care outcomes, reduce costs, and plan the future are possible through decision support, quality measurement, and clinical research, which are addressed in the following sections.

 

Decision Support

Information systems in healthcare are adopted for a variety of reasons. One of these is to improve the outcome in decision-making tasks. A decision support system (DSS) is an information system that gathers data from a variety of sources and assists in providing structure to the data by using various analytical models and visual tools to facilitate and improve the ultimate outcome in decision-making tasks associated with nonroutine and nonrepetitive problems. For example, the DSS can help administration decide whether to add an additional operating room. Management is the primary user of a DSS for operational as well as strategic decisions. It is not used for day-to-day decisions such as scheduling staff. A clinical decision support system (CDSS) is a “special subcategory of clinical information systems designated to help healthcare providers make knowledge-based clinical decisions” (Fenton and Biedermann 2014, 39). (Clinical information systems are discussed in more detail in chapter 11, Health Information Systems.) In DSS and CDSS, typically the problem in need of solving is unstructured or the circumstances are unknown. A CDSS could deliver targeted clinical decision support by supplying clinical reminders and alerts impacting the quality and efficiency of care. For example, within an EHR the clinician may receive a reminder that it is time for the patient’s annual gynecological exam.

With data, analytical models, and visual tools at their disposal, the user can perform simulations of patterns based on various assumptions, monitor and assess key indicators, or perform data comparisons to look for trends. For example, to evaluate the success or failure of interventions, track trends, and identify opportunities for improvement, a manager may monitor readmission rates using a scorecard generated by the DSS.

An executive information system (EIS), a type of DSS, facilitates and supports senior managerial decisions. Given that information is an en