Institute: ONC | Component: 2 | Unit: 7 | Lecture: c | Slide: 10
Institute:Office of National Coordinator (ONC) Workforce Training Curriculum
Component:The Culture of Health Care
Unit:Quality Measurement and Improvement
Lecture:Role of IT and informatics Results of current approaches to quality assessment
Slide content:EHRs Can Augment Data Used in Quality Measures Coded information in EHR Improves ability to assess diabetes quality measures (Tang, 2007) Administrative (or claims ) data insufficient to calculate HEDIS measuresEHR data can improve accuracy of calculating HEDIS measures ( Pawlson , Scholle , & Powers, 2007) But some measures are in narrative text that is harder to access In heart failure, important data inaccessible in clinical notes, especially exclusion data for medications (Baker et al., 2007) Some data can be extracted by natural language processing (NLP) as effectively as manual abstractors in areas such as smoking cessation advice (Hazlehurst et al., 2005), diabetic foot exam ( Pakhomov et al., 2008), and congestive heart failure (CHF) ( Pakhomov et al., 2008) Overall, EHR data quality is mixed for quality measurement; important attributes to improve include are granularity, timeliness, and comparability (Chan, Fowles , & Weiner, 2010) 10
Slide notes:Whether or not EHRs are associated with improved quality, its clear that they can augment the data that is used in quality measures. In fact, today its really a requirement. There certainly is great value to the coded information contained in an EHR. One analysis by Tang and colleagues found that an EHR significantly improved the ability to assess diabetes quality measures. In addition, administrative data (sometimes called claims data) alone is not sufficient for calculating, for example, health care effectiveness data and information set (or HEDIS [ hee- diss ]), measures. Data from the EHR can improve the accuracy of calculating HEDIS measures as well as calculating metrics such as disease-specific mortality; however, the existence of an EHR system doesnt necessarily mean the EHRs contain quality data. A lot of data in the EHR is narrative text that is difficult to access and process. It has been shown in heart failure, for example, that some important data needed to assess quality becomes inaccessible because its in the form of clinical notes. One example of this is exclusion data for medications that patients should be on, such as a beta blocker or an angiotensin [an- jee -oh- ten -sin]-converting enzyme inhibitor. On the other hand, some data can be extracted by natural language processing techniques as effectively as manual abstraction in particular areas. Theres no general ability to do natural language processing in every area, but in some areas, such as smoking cessation advice, the diabetic foot exam, and congestive heart failure, its possible to create natural language processing systems that recognize data that can be used to feed quality measures. The EHR may be a mix of discrete data and text information. System applications exist that can be applied to text for data mining, resulting in data that can be reportable and ultimately used in quality reporting. These systems are not perfect, and they require human oversight; however, they provide another tool to support data retrieval that can be used in quality reporting activities. 10