HCUP: Webinar Transcript
Select to access the full-length audio recording of the Web conference. (Audio File, 1 hour, 15 minutes)
Margie Shofer: Good morning or afternoon, depending on where you're located. I'm Margie Shofer in the Office of Communications and Knowledge Transfer at the Agency for Healthcare Research and Quality, otherwise known as AHRQ. Thank you for joining us for this Web conference on the Healthcare Costs and Utilization Project, or HCUP, and its tools and resources. HCUP is a family of health care databases. You can use HCUP and its tools to research a broad range of health policy issues at the national, State, and local levels.
In addition to an overview of HCUP, this Web conference will provide a live demonstration of HCUPnet, an online query system based on HCUP data. States can use HCUP data to identify, track, analyze, and compare hospital statistics on asthma, diabetes, and other conditions, either across the U.S. or by region. You can also assess utilization, access, charges, quality, and outcomes of hospital care provided in your State. Ultimately, HCUP and its tools can be used to pinpoint priority areas for health care quality improvements.
This Web conference is the first in a series of events highlighting several AHRQ tools created to help you identify and support areas for health care quality improvement. If, after learning about HCUP and its tools today, you are interested in learning more or in receiving technical assistance, please let us know. Our contact information will be provided on the last slide.
Today's Web conference will be presented by Lauren Wier. Lauren is a researcher in the Research Division in Thomson Reuters Healthcare. Currently, Lauren provides quantitative and qualitative research support to AHRQ by supporting HCUP. Lauren leads, coordinates, executes, and presents analysis on health care costs, utilization, and quality.
Lauren will begin by conducting an overview of HCUP's databases. She will indicate your State's level of involvement in HCUP, tell you how you can obtain HCUP data, and provide an example of how HCUP data has been used in numerous applied settings. We will take some questions after the background information has been presented and before the live demonstration of HCUPnet. After the demonstration of the tool, we will wrap up the presentation with a brief summary of HCUP products and support.
Finally, we will have a second open question and answer period, during which time we will address as many of your questions as possible. We would really appreciate your active participation, as the primary purposes of today's Web conference are to introduce you to HCUP and HCUPnet, explore practical applications for the tools, and address any initial questions you may have about the tools.
As I just mentioned, this Web conference will have two question and answer periods, one before the live demonstration of the tool and a second after the end of the presentation. There are two ways you can ask questions. You may submit questions at any time throughout this Web conference by typing your inquiry into the Q and A box located on the right-hand toolbar of your screen beneath the participant list. These questions will be entered into the queue and answered during the question and answer period, so feel free to submit them at any point during the presentation.
In order to ask verbal questions, you must have a telephone icon next to your name. If you do not see a telephone icon next to your name, please send us your question via the Q and A box. To ask a verbal question during the question and answer period, click the "raise hand" button located at the bottom of the box containing the participant list. This will place you in the queue of questions. We will notify you verbally when your line is unmuted so that you may ask your question.
While we encourage questions regardless of their format, asking a live question can be easier if you have any followup questions or if we need clarification from you. There can be a bit of lag time as information is typed into the Q and A box that doesn't exist in a live person-to-person exchange. We will try our best to get to all questions, but if we do not address your question, please E-mail it to us. Again, the E-mail address will be posted at the end of the Web conference, along with Web addresses for HCUP and HCUPnet. We will respond to all questions we receive after the Web conference. So without further ado, I'm now going to turn this over to Lauren.
Lauren Wier: Thanks for joining us today. As Margie said, we're going to be talking about the Healthcare Cost and Utilization Project, also known as HCUP. HCUP was developed as a unique partnership among States, the hospital industry, and the Federal government, and in particular, the Agency for Healthcare Research and Quality, or AHRQ.
HCUP is the largest collection of multiyear all-payer encounter-level inpatient, emergency department, and ambulatory surgery data. HCUP data are unique and powerful. As you can see from this slide, most States provide HCUP with data. We have strong representation of data on the inpatient side and we're working to increase the number of States that provide us with emergency department and ambulatory surgery data.
One of the most fundamental things to understand about HCUP, one of the things that really distinguishes it from most other data collection efforts, is that it's a partnership, a voluntary partnership. This partnership model is unique and it's an incredible strength. AHRQ leverages the very substantial investments that are made by each of the statewide data organizations that contribute their data to HCUP. So for a relatively small investment, we're able to work in unison to create a national data resource.
Without the support of the HCUP partners, the HCUP initiative just wouldn't be possible. Currently, 43 States participate in HCUP. All States provide inpatient data and selected States provide outpatient data as well. Contact information for each of these data organizations is listed on the HCUP user support Web site. Additionally, I would like to give a quick special welcome to our newest State partner, New Mexico.
Sometimes we're asked why all 50 States don't participate and the reasons for that vary. Some are prohibited from participating because of their State laws, some don't have the data systems in place to collect the data electronically, and some are currently working on it.
Now let's talk more about the HCUP data. Uniform billing codes like the UB-04 form are used by many hospitals and are the basis of HCUP data. These forms contain information used in the billing process. Basic demographic data such as patient age and gender are collected and more detailed information about the patient's hospital stay, such as the patient's diagnosis and the medical procedures performed, are also included. In addition, facility charges for patient care are included.
So when thinking about data, the HCUP data, it's important to remember that a large portion of these data are produced for billing purposes and not specifically to support research or policy. However, some States do add additional data elements, specifically to support research and public health. The addition of information on race is one example.
It's helpful to review an inpatient flow diagram from the patient's perspective and also from a data perspective. From a patient's perspective, a patient either makes an appointment for inpatient care, is admitted directly from a physician's office, or is transferred from another hospital or emergency department. The patient is then admitted to the hospital, receives inpatient care, and is discharged.
For the purposes of this presentation, it's also important to understand this process from a data perspective. While this perspective varies by hospital, generally a patient record is created that contains the demographic information about the patient, as well as the medical and clinical information about his or her inpatient services. From that patient record, a discharge summary is generated and given to a medical coder.
The medical coder then classifies the inpatient care into ICD-9-CM diagnosis and procedure codes. The billing department then uses the medical codes assigned by the coder to generate a hospital bill such as a UB-04 form. The foundation of HCUP data is based on this type of billing data, also known as administrative data.
How do we get the data? In general, the patient enters the hospital and a billing record is generated. The hospital may add additional data elements to the record. Then the hospital sends the data to their State-level data organization. These organizations store the data in varying formats. From State to State, there is some significant variation in the definition of data elements. For example, information as simple as gender might be coded in several different ways across the States. So one of the HCUP's most important jobs is the programming effort required to convert data coming in from 43 different sources into a uniform collection of information. The data from the 43 participating States includes millions upon millions of records.
What is HCUP and what is it not? HCUP is a discharge database, not a survey. HCUP is not specific to a single payer but includes all payers, even the uninsured. HCUP contains inpatient, ambulatory surgery, and emergency department data. It does not contain comprehensive physician office visits, pharmacy, laboratory, or radiology information. HCUP contains information on all hospital discharges, not just the sample. HCUP data is available as raw data files, in reports, and online, so in a variety of formats.
HCUP contains six types of databases, three inpatient databases and three outpatient databases. The State Inpatient Databases, or SID; the State Emergency Department Databases, or SEDD; the State Ambulatory Surgery Databases, or SASD; the Nationwide Inpatient Sample, or NIS; the Nationwide Emergency Department Sample, or NEDS; and the Kids' Inpatient Database, or KID. Let's talk first about the three State-level databases and then we'll move on to the three national samples.
The SID, the State Inpatient Databases, contains all inpatient hospital discharge data, including those admissions that started in the emergency department from participating HCUP States. The SEDD contains treat-and-release emergency department data from participating HCUP States, and the SASD contains hospital based and some freestanding ambulatory surgery data from participating HCUP States.
The State databases vary in size. Depending on the size of the State and the type of database, the files range from about 55,000 to nearly 9 million discharges per year. The file structure is similar for all three types of State databases. Data elements and the core file include variables such as patient demographics, diagnoses, procedures, and length of stay. Many States supplement information on the billing records with additional information.
The presence of certain data elements varies by State. This information can sometimes be different from what is included in the billing data. For example, some States provide information on patient's race or ethnicity or encrypted patient and/or physician identifiers. These variables allow researchers to conduct detailed analyses of health or health care disparities. It can also allow researchers to link patients across settings or over time to investigate hospital readmissions, which we'll talk a little bit more about later. Even though some of these data elements are State specific, the HCUP files include uniformly formatted versions of these variables.
The State-level files include many detailed variables such as payer categories. This slide shows the details that one of our States provides for their private insurers. These categories are standardized to facilitate comparisons across States and years. As you can see here, the original variable, which is retained in the data file, pay one underscore X, lists three different payers for private insurance. The standardized data elements, pay one, collapses these into a single value.
Some State-level files also contain detailed race categories; however, you can see that we also provide a standardized version of these variables on the central distributor file to allow for cross-State comparisons. If researchers wanted to go through each State's application process, they could obtain their version of the State data.
How do HCUP files differ from data files available directly from the State partners? There are several key differences. First, the HCUP files contain a subset of data elements. Second, during HCUP processing, AHRQ creates and includes many value-added data elements. Another difference lies in the area of uniformity. Unlike the data files available from the State, the HCUP files are uniformly coded to make it easier for researchers to conduct State-by-State comparison.
AHRQ also conducts numerous standard data quality checks across all State data. For example, checks are performed to ensure that the diagnosis codes are valid. Another routine check is the gender check; for example, AHRQ obstetric diagnosis is exclusively coded for women and prostate surgery is expressly coded for men. Fortunately, the data that States send AHRQ are of high quality, but at times, anomalies are discovered. Finally, because of the processing involved, there is a greater delay in availability of HCUP State files relative to the partner State files.
You can obtain HCUP formatted uniform files through the HCUP central distributor, or you can contact other States directly to request their data. The HCUP central distributor provides one-stop shopping for purchasing many of the States' inpatient data, as well as the national data. But your own State data organization can provide more recent data, as well as data elements not released outside, such as patient identifiers, physician identifiers, and certain linkage variables.
The 26 States noted on this slide release State-level files through the HCUP central distributor. State-level databases for the 2008 data year are now being released through the central distributor. HCUP resources have been informing health care at the national, regional, and State level for more than 20 years. In this time of health care reform, HCUP can play a key role in informing policy decisions with data and evidence.
One example of using HCUP data to inform health policy is the use of the Maryland HCUP SID file to look at the impact of insurance coverage on utilization of C-sections and vaginal birth after C-section deliveries. Few data are available on the impact of mandatory enrollment and managed care organizations, so this investigator used the HCUP 1995 and 2000 State Inpatient Databases in Maryland to analyze the numbers of cesarean sections and vaginal births by Medicaid mothers before and after mandatory Health Choice managed care enrollment.
AHRQ's inpatient quality indicators suggest that a lower rate of cesarean section and a higher rate of vaginal birth after C-section are desirable. The analysis revealed that the increase in C-sections in Maryland was limited more effectively by Health Choice than by care organizations and other providers for the privately insured. Shortly after publication of the results, the Maryland Center for Health Program Development and Management expressed interest in using the analysis as part of its evaluation of the State's Public Health Insurance Program in an effort to ensure access to high-quality health care for Medicaid beneficiaries.
As I mentioned before, there are three national databases and these are derived from the HCUP State-level databases. The NIS, the Nationwide Inpatient Sample, contains inpatient hospital discharge data, including admissions that start in the emergency department, from a sample of hospitals and participating HCUP States. The KID contains pediatric inpatient hospital admission data, including admissions that start in the emergency department, from a sample of pediatric discharges in HCUP participating States. The NEDS contains emergency department data, both treat and release and those that resulted in admission to a hospital, from a sample of hospitals participating in HCUP States.
Let's talk a little bit more about the design of these databases. The purpose of the NIS is to allow researchers to access a comprehensive database for conducting national and regional analyses of inpatient hospital utilization, costs, and overall quality of inpatient stays. This slide shows how the NIS is created. Note that each SID contains a census of hospitals in the State with all of their discharges. In contrast, the NIS is created from a stratified random sample of hospitals in the SID. This approximates a 20 percent sample of roughly 5,000 community hospitals in the United States, even though it is drawn from a sampling frame of less than all the community hospitals.
The NIS is a stratified probability sample of hospitals in the frame with sampling probabilities proportional to the number of U.S. community hospitals in each stratum. The frame is limited by the availability of inpatient data from the data sources currently participating in HCUP. The NIS is stratified on the basis of five hospital characteristics: geographic regions: Northeast, Midwest, South, or West; location: urban or rural; teaching status: teaching or nonteaching; hospital control: government non-Federal, private not for profit, or private investor owned; and bed size: small, medium, or large.
Within each stratum, enough hospitals are selected to approximate a 20 percent sample of all U.S. community hospitals of that type. About a thousand hospitals are included in the NIS. From each selected hospital, all of the discharges for that data year are included. In the 2007 NIS, there are over 8 million discharge records, representing more than 39 million discharges nationwide. Note that because State is not included as a stratum, the NIS cannot be used for State-level analyses. As you can see from this slide, State participation in the NIS has grown since the beginning in 1988.
Like the NIS, the SID is the starting point for the creation of the KID. However, unlike the NIS, which is a sample of hospitals, the KID is a sample of pediatric discharges. To create the KID, pediatric discharges are stratified by three strata: uncomplicated in-hospital births, complicated in-hospital births, and pediatric nonbirths. Using a systematic random sampling design, 10 percent of uncomplicated in-hospital births and 80 percent of other pediatric discharges are sampled from the stratum. Again, because State isn't included as a stratum, the KID cannot be used for State-level analyses.
Similar to the design of the NIS, the NEDS is built using a 20 percent stratified sample of institutions. For the NIS it's a sample of U.S. hospitals. For the NEDS it's a sample of U.S. hospital-based emergency departments. Note that the five NEDS strata are the same as the NIS, with one exception. In the NEDS sample design, the hospital bedside stratum is replaced by the trauma stratum. By stratifying on these important hospital characteristics, the NEDS represents a microcosm of emergency departments in the United States. For example, by including trauma center designation in the sampling strategy, the NEDS has the same percentage of trauma hospitals as the entire United States.
The HCUP national databases can be used to support a variety of research topics, including utilization and costs of hospital inpatient, emergency department, and ambulatory care; trends in health care utilization and cost; quality of care analyses, impact of health policy changes; diffusion of medical technology; medical practice variation; and medical treatment effectiveness.
HCUP is much more than databases. The HCUP project provides myriad research products, including software tools, supplemental files, online tools, methods reports, publications, and a publication search feature. Most of the AHRQ HCUP tools can be applied to any health care administrative database.
The clinical classification software tool (CCS) was developed by AHRQ for clustering patient diagnoses and procedures into a manageable number of clinically meaningful categories. The CCS is a diagnosis and procedure categorization scheme that can be used in projects to analyze data on diagnoses and procedures. The CCS for ICD-9-CM collapses a multitude of ICD-9-CM codes into a smaller number of clinically meaningful categories. The CCS is sometimes more useful for presenting descriptive statistics than individual ICD-9 codes.
AHRQ has also developed a mental health and substance abuse clinical classification software tool that specifically generates categories based on mental health and substance abuse-related ICD-9-CM diagnoses in hospital discharge records. In addition, a version of the CCS is available for use of ICD-10 diagnoses. Lastly, the CCS services and procedures tool provides a method for classifying current procedural terminology, or CPT codes, and Healthcare Common Procedure Coding System, or HCPCS code, into clinically meaningful procedure categories.
The procedure classes are created to facilitate health services research on hospital procedures using administrative data. This classification tool allows researchers to readily determine if a procedure is diagnostic or therapeutic and whether a procedure is minor or major in terms of invasiveness and/or research utilization. The procedure classes assign all ICD-9-CM procedure codes into one of four categories: minor diagnostic, such as EKG; minor therapeutic, such as a pacemaker; major diagnostic, such as pericardial biopsy; and major therapeutic, like a coronary artery bypass graft (CABG) procedure. The maps of procedure classes are available for download on the HCUP user support Web site in a text format and are updated on an annual basis.
The chronic condition indicator is another software tool produced by AHRQ and it provides an easy way for users to categorize ICD-9-CM diagnosis codes into one of two categories: chronic or not chronic. For the purpose of this tool, a chronic condition is defined as a condition that lasts 12 month or longer and either places limitations on self- care, independent living, and social interaction or results in the need for ongoing intervention with medical products, services, and special equipment. Again, the mappings of chronic conditions are available for download on the HCUP user support Web site and are updated annually.
Researchers might be interested in knowing what types of comorbidities individuals readmitted with asthma experience. The comorbidity software is based on the ICD-9-CM coding scheme and creates 29 variables that identify major comorbidities like obesity, congestive heart failure, and HIV/AIDS in hospital discharge records. Again, these are updated on an annual basis and are available for download on the HCUP Web site.
Another tool, the utilization flag software, creates 30 data elements that reveal additional information about the use of health care services. By combining information from ICD-9-CM procedure codes and uniform billing revenue codes, it is possible to obtain a more complete picture of utilization of services in different settings. The utilization flags can be used to study issues such as the use of intensive care units and to more reliably examine utilization of diagnostic and therapeutic services beyond the information that can be gleaned from ICD-9-CM procedure codes alone.
In administrative health care data based on claims, use of services might be reported using ICD-9-CM procedure codes, charges to the revenue centers of the hospital in which services were performed, or both. The ICD-9-CM procedure codes provide information on health care services, but there is concern about underreporting certain procedures such as noninvasive diagnostic services. In addition, ICD-9-CM codes don't provide information on some services, such as stays in intensive care units and psychiatric units. UB-04 revenue codes identify services and accommodations using revenue. Detailed charges based on revenue codes provide additional evidence that the patient received a specific service.
The AHRQ quality indicators were created in response to a request from HCUP partners for a tool that made better use of administrative data, more than the information they were getting from their data currently. The AHRQ quality indicators are measures of health care quality that make use of readily available hospital inpatient administrative data.
AHRQ has developed four separate quality indicators: prevention quality indicators, or PQIs; inpatient patient quality indicators, or IQIs; patient safety indicators, PSIs; and pediatric indicators, PDIs. The quality indicators were developed using HCUP data and they were improved and are annually maintained using all HCUP data from all States. Every year the quality indicators and all HCUP data are used for the national quality and disparities reports.
The prevention quality indicators identify hospital admissions that evidence suggests could have been avoided, at least in part, through high quality outpatient care. Some examples of PQI measures include pediatric asthma and diabetes admission rate. Inpatient quality indicators reflect quality of care inside hospitals, including inpatient mortality for medical conditions and surgical procedures. CABG and aneurysm repair volumes are examples of IQI measures. The patient safety indicators also reflect quality of care inside hospitals, but they focus on potentially avoidable complications and adverse events; for instance, foreign body left during procedures. Pediatric indicators are specific to quality of care for children inside hospitals and include things like pediatric heart surgery mortality and postoperative sepsis.
AHRQ also supports myriad supplemental files. It is important to note that these HCUP supplemental files are only applicable to HCUP databases. Many researchers are interested in the costs of hospital stays. The data that we get from our HCUP partners includes what the hospitals charge for a hospital stay but not actually what the hospital stay costs. An economist at AHRQ developed a method to estimate costs based on charges. These cost-to-charge ratio files are designed to supplement data elements on the NIS, KID, and SID. These files are useful to users who are interested in learning how hospital charges translate into actual costs. These files are also unique by year.
The hospital market structure file contains various measures of hospital market competition. These measures are aggregate and are meant to broadly characterize the intensity of competition that hospitals may be facing under various definitions of market area. Again, these are available free of charge from the HCUP central distributor.
The supplemental files for revisit analyses are the newest addition to the supplemental files created under HCUP. The revisit files are designed to facilitate analyses for tracking patients, both across time and hospital settings, in the HCUP State Inpatient Databases [SID], State Ambulatory Surgery Databases [SASD], and State Emergency Department Databases [SEDD], while adhering to strict privacy guidelines.
Each record in an HCUP database represents one discharge abstract from a hospital setting, inpatient, emergency department, or ambulatory surgery. For example, if an individual visited the hospital three times in a given year, the HCUP databases would include three separate records in their respective database.
After being combined with the corresponding SID, SASD, or SEDD, the HCUP supplemental files for revisit analyses allow researchers to uniformly identify sequential visits for an individual and use the available clinical information to determine if the visits are unrelated, an unexpected revisit or rehospitalization, or an expected followup visit. Total charge information from each visit can be combined to determine the total charge or cost for an episode of hospital care. The files enhance the value of the HCUP State databases by allowing a variety of analyses, such as evaluating repeat emergency department use, readmissions to the hospital, admissions to the hospital after ambulatory surgery, or patterns of emergency department and hospital utilization for chronic conditions.
The HCUP supplemental files for revisit analyses are designed to be used exclusively with the HCUP State databases and are unique by State and data year. Currently, 12 HCUP partners have authorized the release of their State revisit files through the HCUP central distributor. The files are available free of charge from the central distributor.
HCUP produces several additional supplemental files, including the trends files, the NIS hospital ownership files, and AHA linkage files. The NIS trends files are discharge-level files that provide the NIS data user with both the trends wait and data elements that are consistently defined across data years. The purpose of the NIS trends files is to ease the burden on researchers conducting longitudinal analyses.
The KID trends file is the discharge-level file that provides KID data users with trends consistently defined between 1997 and later years. The NIS hospital ownership files are hospital-level files designed to facilitate analysis of the NIS by hospital ownership categories. These HCUP supplemental files allow the user to identify in the 1998 through 2007 NIS the following three types of hospital: government non-Federal, private nonprofit, and private investor owned.
Finally, the AHA linkage files are hospital-level files that contain a small number of data elements that allow researchers to link hospital identifiers on the HCUP State databases and the American Hospital Association annual survey databases. Linkage is only possible in States that allow the release of hospital identifiers.
AHRQ also produces two online tools: MONAHRQ and HCUPnet. MONAHRQ, My Own Network, powered by AHRQ, is a Web-based software tool that enables organizations, such as State and local data organizations, chartered value exchanges, hospitals, health plans, and providers, to input their own hospital administrative data and generate a data-driven Web site. Any organization with hospital administrative data can use MONAHRQ to generate their own Web site and choose how to make the Web site available and at what level; internally to be used by their own staff, externally to be used by member organizations, say, through a password-protected site, or publicly to provide information to the community.
The Web site generated by MONAHRQ will be an interactive querying tool that users can navigate to learn about health care in their area. MONAHRQ software will analyze, summarize, and present information on quality of care, preventable hospitalizations, rates of conditions and procedures, and health care utilization in a format ready to use by consumers and other decisionmakers. We expect to publicly launch MONAHRQ later this year, and you can sign up on the MONAHRQ mailing list to be notified when the software becomes available. You can sign up by going to the HCUP user support Web site.
HCUP data can be accessed via a free online Web query system known as "HCUPnet." HCUPnet provides instant access to health care statistics from the largest set of publicly available all-payer hospital based inpatient and outpatient data in the United States. HCUPnet can provide fast reliable data to a wide range of users, from patients inquiring about medical procedure outcomes, students preparing for class projects, and researchers investigating pressing health care issues to policymakers seeking evidence for legislation.
HCUPnet can be used to quickly and easily study a wide range of topics, including trends and inpatient and outpatient access, charges and outcomes, utilization by special population, the most common and most expensive conditions and procedures, quality of care in patient safety, and differences in outcome between hospital types. HCUPnet also enables users with more complex analytical needs to determine whether there will be enough cases in the data to perform their analyses and also to preview the data included in the full HCUP databases, which are available through purchase through the HCUP central distributor.
We'll now open it up for some questions.