Welcome back. With some knowledge of clinical data, we're in a good position now to compare these complex forms of data within administrative data. As you recall, clinical data are collected for managing patient care. In contrast, administrative data are collected for a variety of purposes that involve public health and medical care. Although administrative data is related to health, it is often collected for the purpose of managing population health, reimbursing providers, or documenting services for operational monitoring or quality improvement. At the end of the lesson, you will be able to trace why various types of administrative data are collected and then recognize the value of administrative data for analytics, even though it might not always be as complete or as rich as clinical data. Let's get started in this lesson by reviewing some important features about administrative health data. First, the objective of creating clinical data is usually to manage patient care. As a result, the data structure has some notes, but a lot of natural language is created by providers who use notes and comments to document the progress of the patient. In contrast, the purpose of creating administrative data differs. First, administrative data created for organizational monitoring. As an example, a hospital may create data systems to monitor patient admissions, discharges, or other movements so they can improve the efficiency of care. Second, huge amount of administrative data is created for finance. In the US, the providers and payers often come from different organizations, thus, specific data structures and codes need to be created to facilitate reimbursement. As discussed below, these files are often referred to as claims data, given that providers are making a financial claim to the insurance company. Finally, public health organizations often run by state and national governments, collect health information to monitor diseases. All of these data types tend to have much less natural language, given the large files involved and the need to use computers to make processing fast, efficient, and error-free. As a result, the data often involves codes, payment amounts, and categories of services. As an administrative data analyst, you will have to deal with many categories of services, but you will likely not have to process free text written by providers. It is worth clarifying the distinction between three types of administrative data that are used by health plans or insurance companies. These are eligibility, claims, and encounter data. First, eligibility data contain records for the time periods that a potential insurance member is eligible for services. Often, eligibility is determined monthly and that ensure is talk about per member per month calculations. Fee-for-service claims are submitted by providers to insurers or health plans to be reimbursed for services. Third, managed care encounter data are collected to identify visits and services. Managed care plans are paid on a per member per month basis. Although managed care plans are not paid for individual services, they are often required to submit insurance companies or government agencies encounter data for each visit. Fee-for-service claims data are known to be of higher quality in comparison to managed care encounter data, given that financial reimbursement is associated with the former. Programs have recently been started however to improve the quality of encounter data, to ensure that all the data are submitted, data elements are correctly coded, and that the data represent real healthcare visits. Although important achievements have been made to improve quality of encounter data, information derived from these data should be used with caution. With imperfect data, identifying medical conditions takes effort. It can also lead to imperfect processes. As I just mentioned, administrative data is collected for administrative functions, such as payments, documentation, or public health. In contrast, clinical data is collected to manage patient care. Thus, clinical data is tied very closely to providers managing a specific patient to remember. Clinical data are unlikely to be perfect, but we expect that in comparison to administrative health data, it might be better at capturing specific information. For example, a patient diagnosis might be more likely to be captured, albeit possibly in a complicated way within the EHR. Claims and encounter data are administrative data, and thus, specific concepts such as medical diagnosis are often imperfectly captured. Both clinical data and administrative data are heterogeneous and complex. In the latter part of this course series, you may work with Medicare claims data for your assignments. When you look at the data, you will realize that administrative healthcare data often covers numerous domains and it is often complicated. The biggest difference is the purpose of collecting data in the first place. Given the high costs and variable quality of healthcare, there has been an accelerating interest to use both clinical and administrative data to measure various dimensions of medical delivery, and then use the information for quality improvement or payment incentive plans. For example, health services researchers in the United States have used administrative data to construct performance measures to evaluate medical errors, mortality rates, and utilization patterns. Analysts and clinicians construct metrics for specific medical conditions and procedures and then compare the performance of hospitals or physicians with respect to their outcomes. For example, administrative hospital data can be used to define specific congestive heart failure patients that enter hospitals, and then after adjusting for other patient comorbidities profile, which hospitals have higher or lower in patient mortality rates. Once made transparent, it is hoped that low-performing hospitals will improve their quality through greater use of evidence-based practices. In addition, consumers and payers can transfer their money and loyalty to higher performing providers. In recent years, the Center for Medicare and Medicaid Services has been spearheading the pay-for-performance movement in which providers doing better on performance metrics will be rewarded with greater reimbursement. This movement in medicine has become more important as more healthcare stakeholders recognize that the fragmented American system is too expensive in variable with respect to quality. Now, let us ask the question; are administrative data reliable enough? Creators of healthcare performance measures are becoming increasingly aware of the need to focus more attention to the area of data quality. This is especially true when there is an interest to use administrative data for quality measurement. Although these data have proven useful for the purposes of quality improvement and public reporting, there is often a need to understand biases and errors in the data. For example, some facilities may under-report adverse medical events to avoid legal threats, whereas other facilities might over-report the same types of events to gain greater reimbursements from payers. Possibly of greater significance is likely that hospitals differ with respect to the processes associated with coding clinical information for billing or how the clinical data are coded and submitted to payers and government agencies. Of course, validation studies show that some data fields and elements are more reliably and consistently coded than others. Thus, if used cautiously, administrative data can be effectively used for research and quality improvement. Next, let's consider the question, is administrative data thing of the past? This is an important question because EHRs are now storing a huge amount of clinical data. An observer might guess that the industry will rapidly take advantage of this richer and possibly more accurate source of information. I cannot foresee the future, but I suspect that administrative data will remain important for many years to come. Here are my few guesses about why I think administrative data will remain important. EHR implementations are incomplete and clinical registries that transform unstructured and complex data into analytical files will take time and resources to create. Next, administrative functions such as reimbursement require a simpler view of the data. For example, payment systems such as the Hierarchical Condition Categories for Medicare Advantage are based on encounters and an entire industry has evolved to support processes associated with risk-adjusted payments. Finally, government laws mandate collection of administrative data. These are used for public health and are often based on a simplified structure of the clinical encounter. Overall, expected administrative data will be here for a while in the future. We started to look at data associated with clinical care and then we moved to data collected to manage healthcare operations. In the next lesson, we will go into the micro world of genomic data.
