Working with Health IT Systems is available under a Creative Commons Attribution-NonCommercial- ShareAlike 3.0 Unported license.

Transcription

1 Working with Health IT Systems is available under a Creative Commons Attribution-NonCommercial- ShareAlike 3.0 Unported license. Johns Hopkins University.

2 Welcome to Quality Improvement: Data Quality Improvement. This is Lecture c. 1

3 The Objectives for Data Quality Improvement are to: Discuss common causes of data insufficiency. Describe how Health Information Technology (HIT) design can enhance quality. 2

4 A case study of data quality in medical registries published by Arts, De Keizer, and Scheffer, in 2002, offers a good summary of many of the data quality issues previously presented in this module. In this article, they discuss common causes of insufficient data quality as either systematic or random. The systematic causes (or what is statistically referred to as Type I errors) are those that can be attributed to some bias or flaw in the measurement process that is not due to chance. Systematic causes, if not corrected, will cause repeated flaws or errors with a predictable pattern or a high degree of uncertainty. Some frequent systematic causes of insufficient data quality are: Unclear or ambiguous definitions Incomplete or unsuitable format Violations in the collection, processing or analysis Poor design in the tools or forms for data entry And a lack of quality auditing or control processes. 3

5 Random causes occur with less predictability and can include Inaccurate transcription or typing (as in free-text entries), Sheer data overload and the possibility of ambiguity or selection of irrelevant data, Inattention or poor understanding on the part of the individual completing the entry, analysis, or data warehousing procedures. 4

6 The team examined planned and systematic procedures that take place before, during and after the data collection to identify causes of insufficient data quality. As an HIT professional, you will be in the best position to consider what can be done to prevent, detect, and facilitate improvement efforts. You will seek to create the best possible quality through the design of the application, the data collection process and subsequent reports or analyses. You will implement activities to detect potential or real flaws that can pose threats to data quality and you will take corrective actions to improve data quality. 5

7 Let s recap what some of those activities include under each of the three areas. Identify the required data elements for the task. A data dictionary with standard definitions and data formats will be essential to promoting data quality. Seek terminology harmonization with accepted standards and avoid local naming conventions or glossaries. Data capture and completeness will improve when the required data location is limited to fewer locations within the EHR. Optimizing the use of structured data fields over free-text will reduce the likelihood of missing data and improve data retrieval. While data extraction programs, such as natural language processing programs can be used to extract data, these programs work best when the variables are narrowly and consistently defined. Standard guidelines for data collection, analysis, and storage should also be documented and should note clear inclusion and exclusion criteria. 6

8 Privacy and security policies should govern the controlled access of the data, and responsibility and accountability for data management must be delineated and observed. Attention must be paid to how clinical workflow and system design can affect data quality. The placement and design of structured data should facilitate the work of the clinician. Excessive navigation and requiring multiple "clicks" will decrease documentation compliance and promote the unintended use of free-text, or lead to missing or inaccurate data. The clinical specificity and number of option choices for structured data, if designed correctly, can facilitate data quality. Synonym and acronym recognition should be used wisely. It can speed data entry, but can also lead to inappropriate entries if the wrong choice is selected. Recognition and correction of data flaws requires a thoughtful monitoring plan. Priority data elements, such as those used for quality measurement, should be identified and targeted for monitoring and improvement as indicated. These data should be reviewed for data granularity, precision, currency, timeliness, completion, and accuracy. Data documentation quality can be improved through staff training and education. Content can include system use, screen navigation, use of references such as the data dictionary and collection guidelines and placement and completion of priority data elements. Standard content and delivery methods should be identified to minimize localized work-arounds. A plan for ongoing education of new users and new content or upgrades should be identified to prevent deterioration in the data quality due to lack of knowledge. 7

9 Automatic domain or consistency checks, such as out-of-range data or inconsistencies between two data fields, at data entry, extraction or transfer can detect potential flaws or errors. Data errors, such as incorrect patient location, which can be undetectable through programming checks, can still occur. Manual processes for auditing or data checking should be developed. Regular review of data collection protocols and report logic should be conducted with care to correct sources of ambiguity or a lack of currency with other changes in data definitions or catalog updates. Review of priority data items, such as frequency analysis or cross tabulations, can be conducted to detect flaws or unacceptable deviations in the data. 8

10 Users will continue to assume that quality is present in the absence of data to inform them otherwise. Regular reports about data quality should be made available to them. Once inaccuracies or flaws in the data are known, corrections and edits to the data must be made. Documentation and correction of the flaws or errors is essential and may provide justification for future design modification. As discussed in the beginning of this module, data are often shared across multiple interfaces and often used for applications other than the originally intended purpose. Systematic study of the source of error and inadequacies should be conducted so current and future corrections can be made across all applicable systems. 9

11 HIT solutions offer a number of opportunities to improve the quality of data. Standardization of terminology has multiple effects in the improvement of data quality. In their study, Thede et al, described some of the benefits of standardization in the nursing field as including: better communication among healthcare providers, improved patient care enhanced data collection to evaluate outcomes greater adherence to standards of care and facilitation of assessment of competency. Some advantages can be assigned to the use of standardized terminology by other professionals. Another significant aspect of the improvement of data quality relies on the use of structured data fields. Structured data fields are fields that contain information that has a pre-defined data model. This allows this data to fit into relational tables thus permitting the easy extraction into reports. The use of structured data fields is somewhat challenging, particularly for clinicians. During their training and their previous experience with paper charts clinicians have used a narrative form to document the ailments of their patients. This story telling format is difficult to translate into the more dry structured data format and requires some retraining for all clinicians. However, in the current EMRs there are areas where clinicians can document in free text or story telling format and yet preserve the integrity of the structured data fields establishing a balance that preserves data quality and the needs for narrative documentation. Structured data fields can be accomplished through pick lists, check buttons or radio buttons. Finally, there is another format of data capture based on voice recognition that will enable the user to capture more comprehensive data. However the majority of this data capture albeit comprehensive happens in a free text format. 10

12 Finally, there are options that although not currently in wide use can enhance the quality of data collected through an HIT solution. The use of natural-language processing machines capable of learning can eventually prove a valuable tool to enhance quality of data. The use of biometric tools that enter the measurements directly into the EHR can be an additional tool for data quality improvement. 11

13 This concludes Lecture c of Data Quality Improvement. Clinical data are increasingly used to drive healthcare decisions. The data that are captured to document patient care are also used for billing, risk management, accreditation, quality, and health care research. Poor data quality can threaten patient safety and quality; decrease satisfaction; increase cost; and compromise strategic planning. Data quality is a complex, multi-dimensional concept and a number of attributes should be considered at all phases of HIT development and use. An HIT professional who is aware of the common systematic and/or random causes of data insufficiency can skillfully employ best practices in the areas of prevention, detection, and quality improvement to enhance the overall quality and usefulness of healthcare data. 12

14 No audio. 13