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 a. We have discussed the impact of poor data quality on quality measurement. We have defined ten different attributes that are used to define data quality; we reviewed an example of each attribute in various clinical settings and highlighted key process recommendations to improve data quality. Let s summarize what we ve learned by reviewing common causes of insufficient data quality and review best practices that you can implement in your role to assure or improve the quality of health information. 1

3 The Objectives for Data Quality Improvement are to: Understand the different purposes of data. Discuss the impact of poor data quality on quality measurement. Identify ten attributes of data quality and key process recommendations. 2

4 In today s information age, data are increasingly driving healthcare decision making. Healthcare databases are filled with data that reflect clinical and clinically-related information. The data are usually collected through the routine processes and activities of patient care; however, its usefulness goes beyond the operational applications that generate the data. The documentation within the electronic health record is often used for quality improvement, payment, legal, research, and accreditation and licensing purposes. Quality aims such as eliminating duplicate or unnecessary tests, or screening and implementing preventive strategies for identified safety risks can be achieved by ensuring complete data collection and an effective health information exchange. Reimbursement can be enhanced by providing the appropriate prompts to substantiate medical necessity. Malpractice cases can often be more successfully defended if the content and quality of the record provides an accurate depiction of the events and jogs the memory of the provider. Research through public health and bio-surveillance agencies can be facilitated if the data collected meet identified definitions and standards for data quality. Last but not least, accreditation and licensing decisions often rest on whether or not the organization s documentation substantiates the standards set forth by the agency. Data are more precious than ever as their use and application expand. 3

5 It is important that a clear distinction is made regarding the use of data. When data are collected to improve care the collection characteristics are different than when it is collected to advance research and expand our knowledge. When the aim of the data collection is to improve care, you use observable data. Accept that there is a consistent bias and collect just enough data to make a decision regarding the outcome of the test in a small and sequential manner. When the aim is to acquire new knowledge, data are better collected blinded (when neither the researcher nor the subject are aware of the test being performed), attempt to eliminate biases when they exist and collect a wide variety of data points for a sample size determined through a power calculation. 4

6 National Quality Forum (NQF) joined with HL7, AHIMA and consulting firm Alschuler Associates to develop a draft standard called Health Quality Measure Format (HQMF). HQMF was developed in part to support, Meaningful Use, as described by the American Recovery and Reinvestment Act (ARRA). The main goal of this format is to standardize data to have consistency across vendors and regions and thus support the, Meaningful Use, of HIT technology. This standardization is accomplished through a number of domains such as document structure (e.g. sections), metadata (e.g. author, verifier), and definitions (e.g. numerator, initial patient population ). Although this standardization is an important first step, it does not enable full machine processing at this time. 5

7 As the previous scenario illustrates, many healthcare errors and adverse events occur as a result of poor data and information quality. Quality and safety issues can often be linked back to poor documentation, inaccurate data or insufficient communication between providers Operationally, poor-data quality leads to low satisfaction and increased cost. Even simple errors such as inaccurate/incorrect names, addresses, and insurance or benefit information can have a negative impact on the satisfaction of patients and staff alike. Patients have a right to expect that the details of their care are documented completely and correctly and that the quality and safety of their care is not compromised by inaccurate or ambiguous data. Operational costs can be increased because of unnecessary duplication of tests or procedures, inefficient care processes, or additional time and resources that must be directed toward detecting and correcting data problems. Strategically, poor data compromise individual and organizational decision making. Good decisions require an effective synthesis of multiple bits of datum that can be converted into meaningful information. Gaps in the data due to missing or incomplete detail or suspicious accuracy complicate the process of effective decision making. Teams can also get diverted by deliberating the quality of the data and never even get around to deliberating the decision at hand. Mistrust of data can spill over into mistrust of other team members and their motives. This can lead to duplicate-data collection resulting in further delays in analyses and ineffective decision-making. If decision making is hindered so is strategic planning, since it is a process that requires decision making. It requires thoughtful assessment of strengths, weaknesses, opportunities and threats and depends on high-quality internal and external data. Once developed and implemented, the plan must be evaluated to determine its effectiveness. If the reported results are of poor quality, knowing how to modify the strategic plan is made even more difficult. Data-quality problems have always persisted, sometimes to a larger degree in some data elements or processes than in others. In the past, the important data were edited or processes of data collection were corrected to ensure accuracy whenever the data were specifically identified as required for quality improvement or regulatory monitoring purposes. The other data elements were not deemed important for correction as these were assumed to be unnecessary. As we move into a new era of health information technology, we are finding new uses for the data and the accuracy and quality are increasingly important. For example, data within EHRs are increasingly used to detect errors. By applying queries, algorithms and decision rules, we can identify cases that represent potential adverse events. In order for these tools to be effective, the data contained in the EHR has to be completely and consistently entered by health care providers. Data-quality management must involve more than fixing problems after the data are entered; it involves preventing the issues from occurring. A fundamental shift to design with the end in mind requires the knowledge and skills of insightful clinical and IT professionals. The first step requires you to begin with quality data. 6

8 Technology is a critical tool in achieving high-quality data in an electronic health record and realizing the benefits of health information exchange. However, technology alone is not sufficient. It is imperative for organizations to hard-wire patient safety and quality of care measures into their electronic processes and systems. Because data quality will positively impact the efficiency, quality and safety of care it follows then that a robust and high quality EHR is required. Such an EHR then becomes an important adjunct to quality, and takes its place as an evidence-based decision-making tool. In 1998, the American Health Information Management Association s e-him workgroup developed the Data Quality Management (or DQM) Model for implementing an EHR documentation improvement process. The model was reviewed and adopted again in 2006 and includes continuous quality improvement in the domains of data application, collection, analysis and warehousing. In this model, the application is the purpose for which the data are collected. The collection includes the processes by which data elements are accumulated. Translating the data into a form that can be used for the designated purpose is part of the analysis phase. You may hear this referred to as transforming data into information. And, warehousing describes the processes and systems used to archive data and data journals. The model includes a number of data quality attributes that can be applied to each domain. The model is generic. It can be adapted to any care setting, used with any application, and can be used in any role that you, as an HIT professional, choose. 7

9 Data quality is a complex topic and it is affected by more than just the accuracy of the data. A review of the literature yields a number of terms that can be used to describe data quality attributes. The DQM model attributes include definition, accuracy, accessibility, comprehensiveness, consistency, currency, timeliness, granularity, precision, and relevancy. Each of these attributes will be described, and an example will be provided as well as key process issues that HIT professionals should consider for effective health information exchange. 8

10 We know and understand that data are often used for purposes other than that for which they were originally collected. It has been said that one man s junk is another man s treasure! Therefore, to support the multi-use of data that is collected, clear definitions for each data element should be provided so current and future users will know what the data mean. For example, does the word football mean the same thing to people all around the world? No and without a clear definition that lets the user know what type of football we are referring to errors in interpretation are likely. In addition, standard definitions are necessary in order to compare data with data stored in other databases, for example external registries or quality databases, or to compare data over time, such as trend data for quality purposes. Each element should have clear meaning and acceptable values. For example, in addition to clear definitions, a data dictionary should provide data type, length restrictions, and other rules including uniqueness, consecutiveness or calculated data and acceptable ranges. For example, should an EHR allow a temperature with 4 digits (without a decimal point) to be entered? Should numeric data be allowed where only text is expected? Inaccurate data can result from mistakes that are made when data are extracted, transformed or transferred to secondary data sources. Well documented data definitions and rules that govern accepted values can protect against inappropriate use of data. The federal Health IT legislation described earlier requires that, Meaningful Use, of health information technology includes electronic reporting of data on the quality of care. When the Centers for Medicare and Medicaid Services were preparing to announce their final rules for the meaningful use standards and finalizing the list of core quality measures, one hospital had examined their opportunities for outcome reporting through a nationally recognized database for quality reporting. They decided to participate in the National Health Safety Network (NHSN), a voluntary, secure internet-based surveillance system that is managed by the Division of Healthcare Quality Promotion at the Centers for Disease Control and Prevention. Exchange of data through NHSN requires adherence to the guidelines and procedures for data 9

11 collection that includes specific definitions for all variables collected and reported. 9

12 One of the most important functions you, as an HIT professional, can perform is to assist the team in the development of a thorough data dictionary. Let s discuss how the attribute of data definitions can be applied to the Data Quality Management Model domains of application, collection, analysis, and warehousing. Appropriate use of the data requires an understanding of the purpose and data definitions. The data collection process should guide the user to enter only acceptable values and minimize or eliminate any ambiguity. Meaningful analysis relies on clear understanding of the data and making appropriate relationships among the variables. Warehousing requires assigning responsibility for the ownership and maintenance of the data and documentation over time along with corresponding policies and procedures for data and information management. 10

13 Accuracy is a term used to refer to the extent that the data properly represent the real-life objects they are intended to represent. Accuracy implies that the value is valid and correct, and the person who the value is related to is properly assigned. Inaccuracy can result from deficiencies in other attributes that we will be discussing later on. A lack of precision or completeness can also influence the accuracy of the data and the answers to the questions you intend to find through the use of that data. Remember, inaccuracy can lead to an understanding of a valid real-world state, but NOT the one intended. Payment for care rendered is critical to the survival of any provider in all care settings. During the registration process, insurance information must be gathered and validated. You, as an HIT professional, can assist in improving the data quality in this process. By working with business office personnel, you can establish a process to develop and maintain a reference table with codes for all approved insurance providers. An automated process can be instituted to verify the accuracy of the insurance information by limiting entry to only those codes that are available in a drop down menu that is linked to the reference table. Entry of any other data would require a process of verification and pre-approval for any manual entries for this field of data. 11

14 A key process recommendation for you to enhance accuracy in the application of the Data Quality Management Model is to collaborate with users of the data to establish a policy or process to identify how data used in EHRs are generated and how data content will be determined and standardized. In an effort to improve data accuracy, you can prompt the users to think about purpose and application of the data and how choices in the data entry may be limited to improve the accuracy, maintain integrity and improve the reliability and validity of the data. Accuracy in data collection can be improved by educating and communicating data definitions to those who collect the data. For example, if the time of initiation and discontinuation is critical for payment, the staff need to know exact, not approximate, times that must be collected. You can also assist in accurate analysis of the data by ensuring that the algorithms, formulas and translation software are correct. For example, one of the alternate quality measures to demonstrate meaningful use of electronic health records is childhood immunization status. You would meet with the pediatric providers to discuss the purpose of this metric, the data elements that are collected, who will collect them, how the data will be applied to decisionmaking, and how the data will be transmitted to the immunization registry without losing data accuracy. Appropriate edits must be made to ensure accuracy prior to warehousing the data for future use. Exception and errors reports should be developed so that corrections to the data can be made. For example, some diagnoses or patient locations may be incorrect for the age or gender of the patient. Screening for these types of problems and making corrections will improve the accuracy. 12

15 Accessibility is the extent to which data is available or easily obtainable for use. But easily obtainable does not mean that unauthorized individuals should be able to gain entry into protected personal health information. Accessibility incorporates ease of gaining entry with the safeguards that are absolutely required to assure confidentiality and privacy of patient data. These safeguards should be built into the process and automatically deploy without any special effort by the user. The Health Insurance Portability and Accountability Act (HIPAA) includes rules, standards and guidelines to guide you in establishing the appropriate procedures for health data access. The burden of data collection can often derail safety, quality improvement, and research efforts. Often the data that are needed already exist someplace within the scope of the electronic health record. A typical example of data that are often needed, but shouldn t have to be collected again by clinicians, are the demographics of a selected population of patients, such as home care patients with known congestive heart failure, who will be included in a quality improvement or research study. However, the detail and the use of this data must be evaluated for the patient s protection under HIPAA and other regulations. You can guide the team to select the best, least-costly, and legally-appropriate way to access and collect the data that are needed. The amount and accessibility of the necessary data can be increased through system interfaces. 13

16 Inaccessibility of data can be a frustration to clinicians who need data to generate information about ways to improve care. However, a lack of data stewardship also has serious risks and consequences associated with unauthorized access or inappropriate use of health information. Proper observation of the domains in the Data Quality Management Model requires you to work with clinicians to define and agree on the types of data and the minimum amount of data that needs to be available to support the team in achieving its mission and objectives. The intended application or use of data, and the legal, regulatory and financial boundaries often determine which data should be accessible. Collection of accessible data should be assigned based on the expertise and scope of practice of team members; with registration staff collecting demographics, clinicians documenting physiologic findings such as symptoms or scale ratings, and coders assigning medical record coding. Data analyses should be supported by timely access to the required data. For example, if there is a recall on a lot of vaccines, care providers in a primary-care office need rapid access to vaccine administration data to be aware of potential patients to be alerted to the recall. Policies should define the process, restrictions and rights for retrieval of data from database systems and warehouses. The accountability and chain of trust within HIPAA should be delineated. Organizations should be specific in their internal policies and business associate contracts about what identifiable health data may be used and for what purpose, by both the business associate and its agents; also what HIPAA de-identified data may be used and to whom they are applied; the requirement that business associates have contracts with their agents that are equivalent to business associate contracts; and the use of HIPAA definitions for any deidentification of protected health information. Methods to regularly monitor and audit access to data should be in place. 14

17 Comprehensiveness is the ability of an information system to reflect every possible state in the real world. Intentional limitations of the data should be documented, and every effort to include all of the data elements that are required is made. It is understood, of course, that not every piece of data can be captured, but many projects have suffered from lack of deep forethought about what and how to measure. A high level of missing data will reduce the reliability and validity of your analysis. In order to minimize missing data, rules can be assigned to a data set to define mandatory elements that require a value, optional elements that may have a value assigned based on some set of conditions, or inapplicable attributes that may not have a value. Comprehensiveness is illustrated in the following scenario. In October 2008, the Center for Medicare and Medicaid Services began requiring hospitals that receive federal funding from Medicare and Medicaid to begin disclosing, never events. Never events are conditions that CMS defines as preventable, and serious in their consequences for patients, and that indicate a real problem in the safety and credibility of a healthcare facility. Included in this list of conditions are pressure ulcers or what is often referred to in layman terms as bed sores. CMS has stated that they will no longer reimburse hospitals for any costs associated with never events, and hospitals are prohibited from passing the costs onto patients. The ability to differentiate conditions that were present when the patient was admitted, versus those that were acquired during the hospital stay, requires a comprehensive assessment and documentation as a means to avoid potential penalty and quality concerns. Increasingly, clinicians are turning to HIT professionals to assist them in defining data elements and rules for their completion for clinical, financial and risk management needs. 15

18 Key process recommendations to apply the attribute of comprehensiveness within the context of the Data Quality Management Model is for you to seek clarity from the team about how the data will be used and how end-users can assist to ensure that complete data will be collected. Opportunities to create interfaces with other automated systems should be pursued when doing so can enhance the comprehensiveness and quality of the data collection. An example of this might be to link the skin assessment completed in the emergency department s electronic health record from one vendor to the skin assessment completed in inpatient electronic health records made by a different vendor. The goal is to make the collection of the necessary data elements as comprehensive and as seamless as possible across care-settings. Be alert to the multiple places that the same data element might be recorded and attempt to reduce the variation in data completeness. Whenever possible, provide structured response choices and reduce the number of free-text entries to facilitate complete data entry and extraction. You should recommend that all relevant data are collected and analyzed in concert. For example, in addition to assessing whether or not pressure ulcers were present on admission, the team may also want to know about risk factors to aid in a comprehensive assessment of the quality problem if a number of patients later develop ulcers. In warehousing data, be aware of and educate all data stakeholders of the data that are available to prevent redundancy and conflicting data collection. 16

19 This concludes Lecture a of Data Quality Improvement. In summary, data use for research purposes is collected under different conditions than that used for QI. Poor data quality contributes to error. The ten attributes of data quality are: Definition Accuracy Accessibility Comprehensiveness Consistency Currency Timeliness Granularity Precision Relevancy 17

20 No audio. End. 18