HCUP: Webinar Transcript (continued)

Webinar Transcript (continued)

Transcript from the 2010 webinar describing the Healthcare Cost and Utilization Project (HCUP).

Margie Shofer: We've now entered our first Q and A period, and as a reminder, there are two ways you can ask a question. You can submit a question by typing your inquiry into the Q and A box located on the right-hand toolbar of your screen, or you can click on the "raise hand" button located at the bottom of the box containing the participant list. We would love to hear your voice, so please consider asking a verbal question.

Question: What is the difference between charges and costs?

Lauren Wier: Hospital charges are actually the amount that the hospital charges for the entire hospital stay. It's important to remember when using HCUP data that the charge isn't specific to a procedure or a condition but to the entire hospital stay. The charges are the amount that the hospital charged for the entire hospital stay, whereas costs tend to reflect the actual cost of production. Total charges can be converted to cost using the cost-to-charge ratio file that I mentioned earlier. In general, costs are less than charges. For each hospital, a hospitalwide cost-to-charge ratio can be used to determine costs.

Question: Since you mentioned the cost-to-charge ratio files, are they adjusted for cost of living in the various regions?

Lauren Wier: In a way they are. The cost-to-charge ratio files are hospital-specific files that take into account the hospital's location. So cost of living is indirectly taken into account based on where the hospital is located. There's no need to further adjust for that.

Question: Can HCUP be used to look at readmissions, and if so, what would be involved in the analysis?

Lauren Wier: As I mentioned, the revisit analyses files allow HCUP researchers to link together the SID, the SASD, and the SEDD in order to identify sequential visits for an individual. You can also use the available clinical information to determine if the visits are unrelated and expect followup, complications from a previous treatment, or an unexpected revisit or hospitalization.

There is a recent example that was published in the Journal of the American Medical Association (JAMA) looking at readmissions for sickle cell disease in the emergency department and hospitalization from selected States in the 2005 and 2006 SID and SEDD. That study found that among patients with sickle cell disease, acute care encounters and rehospitalizations were frequent, particularly for 18- to 30-year-olds and people who were publicly insured. So the revisit files are a great tool to look at readmission. It's simply a matter of linking together the relevant databases or data years for those States that release that information.

Question: When will the MONAHRQ application be released?

Margie Shofer: This is something that we are very eager to get released. We know people are really excited about it. We are doing some final testing on it and hope to have it out pretty soon. I wish I could tell people an exact date, but we expect it to come out really soon. I promise that when it comes out, we will let all of you know.

Why don't we move on to the HCUPnet demo? That way, we can make sure we have enough time at the end for more Qs and As.

Lauren Wier: I hope that everyone is now seeing HCUPnet up on their screen. HCUPnet provides national and regional information on U.S. community hospital stays based on the Nationwide Inpatient Sample. It also captures details on pediatric care from the Kids' Inpatient Database and provides access to State-level hospital encounters from the State Inpatient Databases. Data on national and State-level emergency department care is also available. In addition, a variety of pathways enable specialized queries of mental health-related hospitalizations, health care statistics by expected payer, and hospital-type comparisons.

HCUPnet also provides ready-to-use national benchmarks for health care quality indicators. Finally, the quick or national State statistics pathways that you can see here on the right provide ready-to-use tables on commonly requested information from the NIS, KID, and NEDS, as well as selected SID and SEDD databases for those States that have agreed to have their data included on HCUPnet. It's important to note that the SASD is not available through HCUPnet.

An HCUPnet tutorial and overview are available in the upper right-hand corner for first-time users, and a link to definitions of commonly used terms is also available. Below this, HCUPnet news items and links are regularly updated with important information.

Let's take some time and run through some sample queries. Let's first take a look at how a user could obtain a national perspective of hospitalization for Clostridium difficile or C. diff., which is a bacterial stomach infection in the United States. Because we're interested in a national data perspective on C. diff. hospitalization, we'll click on national statistics on all stays.

Then let's select researcher/medical professional. The layperson/data novice link removes some of the options and was created for the media. Since everyone on this call is pretty familiar with health services research, we'll strongly recommend selecting researcher or medical professional.

As you can see, HCUPnet takes you through a step-by-step process to select the information that you want. We have several options. We can get information on specific diagnoses and procedures by year, statistics on stays by year not broken down by procedure or diagnosis, or national trends from 1993 to 2007, or we can rank order diagnoses or procedures by factors such as number of discharges, charges, or mortality rates.

Let's first look at the statistics on C. diff. for a single year. We'll select statistics on specific diagnoses or procedures and let's select 2007, the most recent data year. Now we can select what kind of diagnosis and procedure information we want. For this, we'll use diagnoses grouped by ICD-9-CM code.

Now we have the option to choose between obtaining health diagnosis. For this query, let's select principal diagnosis to indicate that C. diff. is chiefly responsible for admission. Notice that the definitions for all of these key terms are provided in the box on the right-hand side

The navigation bar tells you exactly where you are in your query and what remains to be selected. You can also move backward by just selecting these links. The back button works as well.

Let's select principal diagnosis. Now we can enter the ICD-9-CM code of interest. I'll enter it in this box here. If you were interested in looking at multiple codes, you could ask to see these separately or all codes combined. We just have one code here. Now we can select the outcomes and measure in which we are interested. For this, let's select the number of discharges. Now we can select patient and hospital characteristics. We'll keep all patients and all hospitals selected, and let's also select to view the data by region of the United States.

HCUPnet is based on aggregated statistics tables to speed up data transfer and also to protect individual records. As you can see, we get the results nearly instantaneously. This table shows us that there were an estimated 104,000 hospital stays in the United States for C. diff. in 2007, and you should note that unlike with when you purchase the national databases, this data has already been weighted for you. We can see that States differ by region. Let's look a little bit more closely at this data.

Based on the data from HCUPnet, I just created this quick little graph. We can see from this graph that the South has the highest percentage of C. diff. stays, with about one in three C. diff. stays occurring in the South. The Northeast and Midwest each have about one in four stays, and the West only has about 16 percent of all stays. The question is: Are C. diff. stays more prevalent in the southern population, or does the large percentage of stays reflect a large population size of the South?

You can take a closer look at the population variation by certain characteristics. You can see that there are big population differences by age group, region, and location. Adjusting for population differences can make a big difference for these characteristics. Similarly, it might not be as important to adjust for gender at the national level, since males and females each roughly make up about 50 percent of the population. As you can see, once the data is population adjusted, the rate of stays is highest in the Northeast, followed the by the Midwest, South, and lastly, the West. The table on the top right of this slide lists the population count by region. You can see that the South has a far larger population than any of the other regions. We bring this up to show you that while the information you get from HCUPnet is great, sometimes you do need to consider other factors and adjust the data appropriately.

Let's go back to HCUPnet and continue to do some queries. Back on the results screen, I'll point out a few additional features that can help you with your work. At the top of the page, there is an option to save the data as an Excel spreadsheet. You can also check statistical significance. To do so, use the Z-Test Calculator provided here and click on the link. A small screen will open up, and you can just copy and paste the information.

If you would like to repeat the query for a different year, you can easily rerun your data on another database or another year with just a few clicks. We'll select the repeat query on another database link and then select the 2006 data year. We instantly get the results for the same diagnosis, outcomes, and patient hospital characteristics that we previously selected. If you're interested in running an entirely new query, you can click "Run a new query" or you can also just return to the home page.

Now that we've seen some information for 2006 and 2007, let's look at C. diff. hospitalizations over a longer time period. You can either go back, like I showed you, to the home page, or you can just click on the type of query option and it will bring you back to this selection page.

Let's click on "Trends." Again, we'll use specific diagnoses by ICD-9-CM codes. We'll stick with principal diagnosis. Notice that the code that I had previously entered was retained. We'll hit "Next." For this query, let's look at the number of discharges, and we'll also look at discharge status. You can see that the trend data is quickly generated for 1993 through 2007, and the standard errors are provided below.

Also notice that on the screen we have some cells with asterisks. These statistics are suppressed because the standard errors aren't reliable or the cell size is too small. HCUP prohibits reporting cell sizes less than 11. In the event that you're conducting a query and see an asterisk, an explanation is here below. This is a lot of data to take in, and HCUPnet has provided an option to show this information graphically, so let's take a quick look. Looking at the graph, we can see that hospitalizations for C. diff. have increased, particularly in the last 10 years. Scrolling down, we can get additional information on discharge status. We can see that the percentage of routine discharges for stays with C. diff. also declined. All of that information is available here.

Let's change topics and look at the emergency department setting and asthma diagnoses. I'm going to go down to "National statistics on all stays," on "All ED visits." We'll choose "All ED visits," which includes both encounters where the patient is treated and released and those where the patient was admitted for inpatient treatment. We'll select "Statistics on specific diagnoses" and the most current data year, 2007.

Now we can select "Diagnoses grouped by clinical classification, or CCS." This is the coding system that we reviewed earlier that clusters ICD-9-CM codes into a smaller number of clinically meaningful categories. Let's select that. Diagnoses on the ED path are designated by first listed and all listed. As you can see from the definition box on the right-hand side, the first listed diagnosis is the diagnosis that appears first on the record. For ED visits that result in hospital admission, this is the principal diagnosis. For ED visits that result in discharge, it may not be the principal diagnosis, but it might just be the diagnosis that happens to appear first on the record. For this query, let's choose "All listed diagnoses."

We're going to look at asthma here. You can scroll through to find asthma, or we can just type it into the box and hit the search button, and it appears. We'll highlight it and hit "Next." Now let's select some characteristics. We'll look at all patients in all hospitals, as well as patient age. And we have the results pretty quickly.

The first column shows the weighted national estimates for all ED visits combined. The second column shows ED visits that resulted in hospital admission. And the third column shows treat-and-release ED visits with an all-listed asthma diagnosis. As you can see, there were an estimated 5.8 million ED encounters involving asthma. Of these, about a quarter of the encounters resulted in the patient being admitted to the hospital. You can also see that nearly 60 percent of ED visits for asthma among individuals ages 65 to 84 resulted in admission to the hospital. In contrast, only about 11 percent of ED visits for asthma among 1- to 17-year-olds resulted in admission.

You can also see the ICD-9-CM codes that fall into each category by clicking on this link here below. All you need to do is click "Find" to see which ICD-9-CM codes cover the diagnoses. We'll see those here in the CCS category for asthma.

Now that we've performed some national queries, let's conduct a query at the State level. Let's look at hospital stays for diabetes. So, again, select researcher. We'll look at State statistics on specific diagnoses. Let's select Florida for 2008. And let's look at diagnoses grouped by Clinical Classification Software. We'll look at principal diagnosis. Again, we can search using the search box function. We'll select diabetes with complications. You could also select "Multiple" if you're interested. Let's look at the number of discharges and the mean hospital charges. And let's look at all patients in all hospitals, and we'll look at it by primary expected payer.

Here are the results. We can see that in the State of Florida, there were about 34,000 discharges with a principal diagnosis of diabetes with complications. We can see that about 60 percent of those discharges were billed to the public payers, Medicare and Medicaid. You can also see that the mean charges were highest for those payers as well. Again, we could repeat this query on another database very easily by just clicking here.

Going back to HCUPnet's home page, let's look at the type of information available from the QI summary tables. Let's look at information for diabetes complications. We can get national-level data on quality of care based on the quality indicators using this path. To get the information on diabetes stays, select the "Prevention Quality Indicators" or "PQI" indicator path for 2007. Now let's select the "Diabetes with long-term complications" indicator.

We get instant results for the Nation. These are national benchmarks. We can see that in 2007, there were about 124 hospital stays per 100,000 population with diabetes with long-term complications as the principal diagnosis among adults age 18 and over. As you can see, rates were also broken down by patient and hospital characteristics. You can see that the rate of hospitalization increased with age and that rates were generally higher among males than females.

That concludes this portion of the HCUPnet demonstration and we'll return to the presentation. I encourage you to take some time to explore HCUPnet.

As you've seen, HCUPnet is a versatile tool that can produce simple statistics, sample size calculations, trends information, rank ordering of diagnoses and procedures, and significance testing. In addition, tables and graphs can easily be saved in Excel and exported to a workbook. However, HCUPnet isn't useful for certain more complex research questions or statistics.

Now that we've reviewed the research tools and supplemental styles produced by the project, let's talk about the research publications and other publication resources. HCUP statistical briefs are short, focused reports on topics related to specific conditions, procedures, or populations. Statistical briefs are useful to a wide variety of audiences, including policy analysts, decisionmakers, media personnel, and others in need of summary facts and figures on current health care issues.

To date, AHRQ has released 89 statistical briefs spanning a range of topics. HCUP also produces fact books and an annual report. The annual HCUP Facts and Figures Report highlights the rich potential of HCUP data by providing targeted analysis of important trends in hospital care organized around high-interest topics, such as hospital and discharge characteristics, diagnoses, procedures, and costs.

In addition, this year's report presents a special section that details trends in hospital care by expected payer, including Medicare, Medicaid, private insurance, and the uninsured. This year's report, along with previous facts and figures reports and the fact books and statistical briefs can be downloaded from the HCUP Web site.

The applications of HCUP data are endless. HCUP data support cutting edge health services research and, to date, more than 1,200 journal publications feature HCUP data, software products, and tools. Often, this research affects health care policy, practice, and future research. Here is a sampling of the journals that have published studies using HCUP data.

HCUP data also supports congressionally mandated reports such as the National Healthcare Disparities Report and the National Healthcare Quality Report. Additionally, studies that use HCUP data have been featured in magazines such as Newsweek.

The HCUP Web site now features a page where users can search for HCUP publications based on data year, database, and tool products. The publication search page allows two different ways to search for articles. The simple search option allows users to type in keywords and select from peer review journals, other publications, or all publications. The advanced search feature allows users to search within specific fields, such as author, title, periodical, publication, and abstract. To access the publication search page, you can access the HCUP Web site.

In addition to the databases, software tools, and research products, HCUP provides a wealth of resources and support. Substantial information about HCUP can be accessed from the HCUP user support Web site, including how to obtain HCUP data, review extensive documentation on HCUP databases, software tools, research projects, and supplemental files, and how to access HCUPnet. You can find a comprehensive listing of HCUP-related publications and reports on the HCUP user support Web site and you can access technical assistance.

For future reference, you may want to take the online HCUP overview course. It contains information similar to what we've presented here today. AHRQ is also pleased to announce a new HCUP online tutorial series. These online trainings are designed to provide data users with information about HCUP data and tools, as well as training on technical methods for conducting research with HCUP databases.

The first technical course in the series, HCUP sample design tutorial, has recently been released. This tutorial assists users in understanding the sampling strategy of the three nationwide databases, the NIS, the KID, and the NEDS. Additional modules will be released later this year. Topics will include loading and checking HCUP data and producing national and regional estimates.

In addition to the resources provided on the HCUP user support Web site, users can also E-mail or call HCUP technical assistance staff with questions about HCUP data, tools, or products. The technical assistance staff consists of senior research personnel who are available to answer technical questions you may have regarding the application of an HCUP product or tool. The researchers are trained in epidemiology, health services research, statistics, and medicine and can provide expertise regarding HCUP databases and tools. The technical assistance team will review the information provided and typically respond within 3 business days.

We'd be happy to take any additional questions that you may have at this time.

Margie Shofer: We have entered our last Q and A period. Again, there are two ways you can ask a question. You can type your question into the Q and A box located on the right-hand toolbar of your screen or you can click on the "raise hand" button located at the bottom of the box containing the participant list.

Question: I have a question particularly on a market share segment that's available. Has AHRQ received feedback from either hospital associations or the hospitals themselves about being competitively at a disadvantage because physicians or ambulatory surgical centers would be able to access information that typically most other businesses wouldn't have access to in other industries? How does AHRQ go about addressing that? Is there any literature to support agencies who wish to make that information more available for public consumption, that there is no harm in making it available? Thank you.

Lauren Wier: It brings up a good point, which is that the data that the project produces are intended for research and policy purposes. However, everyone who uses the data needs to sign a data use agreement in which they agree not to identify physicians, individuals, patients, or hospitals. To that end, the data is not intended or allowed to be used to identify and separate out individual hospitals.

Margie Shofer: O.k. We have a hand raised.

Participant: Hi. I actually have three questions.

Margie Shofer: Oh, go for it.

Question: My first question: I heard throughout the presentation various mentions of free tools in addition to tools where cost is not mentioned. Could you review what is free and what is not on HCUP and what is the cost to access the tools?

Question: The second question is: Is there individual hospital data available, or is it only aggregated by State?

Question: And the last question is: Is payment data available? You talked about charges and costs, but I didn't hear anything about payments. Thank you.

Lauren Wier: Great, thank you. To address the first question, what's free and what's not. HCUPnet is totally free. That can be used by anyone and you simply go online and access that. The databases vary in price. Those are available for purchase through the HCUP central distributor or the State databases can be purchased through the States themselves. You can find information about the exact costs of all of the databases on the HCUP user support Web site by clicking on the central distributor link. It will show you exactly what data years and States are available for purchase and how much it costs.

The database is available through the HCUP central distributor at the national level. Those prices are set by AHRQ. There is also a student fee, which is a reduced rate. The State-level databases really vary in how much they cost. It can range anywhere from 20 dollars up to 3,000 dollars. That's the cost for State-level databases, so there's some variability in the costs.

In terms of your second question about individual hospital data, if you access information through HCUPnet, all of the information publicly available on HCUPnet is in aggregated form. You won't be able to identify any individual hospitals or discharges and they have taken precautions to suppress information that might potentially identify. That's why you saw stars in some of the boxes in the presentation.

When you purchase the full database, identifiable information is available. You won't have patient names or Social Security numbers or anything like that, but you will be able to identify hospitals and link to other files, such as the AHA linkage files. But again, there is a data use agreement not to identify specific individuals, hospitals, patients, physicians, and so on.

Then with respect to your third question about payment data, we don't have information on payments. We have charge information from which we extrapolate costs and AHRQ is currently working on a price-to-charge ratio from which they would be able to extract the price for services. We don't have information on what was reimbursed to the hospitals.

Question: How often is the HCUP data refreshed?

Lauren Wier: We collect data from the States on an annual basis, so they provide a year's worth of data to us. Then typically there's about an 18-month lag between the end of the calendar year and the availability of HCUP data. Currently, we have 2008 State-level databases released, as you saw during our HCUPnet demo. We're currently working on the 2008 national database. There is about an 18-month lag between the end of the calendar year and the availability of HCUP data.

Question: What sort of quality checks are performed on HCUP data?

Lauren Wier: Most of the validation that we do for HCUP is looking at the completeness of specific data fields. While we do limited checks for accuracy, like gross checks to see whether males are being coded with giving birth or females are being coded with male-specific procedures, we don't do detailed checks on the accuracy of coding. We do compare the HCUP NIS counts with other national data sources of hospital care, such as the National Hospital Discharge Survey files. Additional information about the quality checks can be found in methods reports on the HCUP Web site.

Question: Can I look at patient admissions rather than discharges?

Lauren Wier: The NIS record is created only after the patient is discharged from the hospital, so it's really more appropriate to use the NIS to look at care from the discharge perspective rather than from the admission perspective. It also brings up another great point, which is that these are not patients but these are discharges. This is discharge-level information. When you're looking at revisit analyses, you can consider looking at patient-level information, but each line in the file represents an individual discharge. It's not an admission and it's not a patient; it's a discharge.

Question: Are older annual databases updated to account for a claims lag?

Lauren Wier: I don't exactly understand the question, but I'll answer it the best I can. How it works is that we get all of the data from a State, so there's lag time involved in that. For instance, we don't ask a State to send us their data until they feel like they have complete data for a given data year. After they send it to us, we process and format it. We try to capture as much information as we can, but that's why there is an 18-month or so lag time between when the actual data calendar year ended and when the data becomes available.

Question: When will the NIS, KID, and NEDS be released?

Lauren Wier: Again, there's typically about an 18-month lag time, so we're currently working on the 2008 NIS and we expect that that will be released later this summer. The KID and NEDS are also in the works. The KID is produced every 3 years, starting in 1997, so we've got '97, 2000, 2003, 2006, and 2009. The KID will probably be released in 2011. We're currently working on the NEDS and we expect that to be released later this year.

Question: We have purchased HCUP data in the past. Do we need to fill out the application kit every time we order a new HCUP yearly dataset?

Lauren Wier: Yes, a new application kit is needed each time a new dataset is purchased and a new data use agreement will also need to be signed. It's not necessary to take the HCUP data use agreement training course more than once. There's a data use agreement training course online. It takes about 15 minutes to complete. You complete that and then you receive a code, a completion code or a certificate code. Then you'll put that on your application when you first apply. After the first submission, you don't need to repeat the course, but you will need to submit a new form. Similarly, if you have a colleague or student who is going to be using the database that you've purchased, it's your responsibility to have that person also take the online data use agreement training course and submit a data use agreement.

Margie Shofer: That seems to be all the questions we have right now, so we're going to wrap up. I'd like to thank you all for the thoughtful questions and your participation in this Web conference today. We hope this event was helpful for you. If you have any questions about follow-on technical assistance opportunities, please do not hesitate to submit them to the quality tools E-mail address on this slide. If you have any other questions or comments, including a request for slides from today's presentation in case you didn't have a chance to print it off, please send an E-mail to the same address.

This outreach program will run through October 2010, so please notify us of any further information or technical assistance needs before then. Thanks again, and this concludes today's Web conference, and we look forward to hearing from you. Goodbye.

Current as of May 2010
Internet Citation: HCUP: Webinar Transcript (continued): Webinar Transcript (continued). May 2010. Agency for Healthcare Research and Quality, Rockville, MD. http://www.ahrq.gov/cpi/centers/ockt/kt/webinars/hcuptrans/hcupwebtrans2.html