The Cancer Research Network's Virtual Data Warehouse
Hart GE, Hornbrook M, Field T, Wagner E, for the members of the CRN Scientific Data Resources Core
Mark C. Hornbrook, PhD, Kaiser Permanente Center for Health Research, 3800 North Interstate Avenue, Portland, OR 97227-1110. E-mail: email@example.com.
Background: The Cancer Research Network's (CRN) Virtual Data Warehouse (VDW) will encompass eligibility, clinical, utilization, and sociodemographic data across all participating health plans. We seek to reduce programming and data processing costs for multisite projects, to build on the lessons learned from previous studies, and to stimulate new proposals by reducing data extraction barriers.
Objective: To profile the early experiences of users of the CRN's VDW, inform HMO-affiliated researchers about the capabilities of the VDW, and alert potential users of some of the complexities involved in extracting standardized data for research purposes.
Methods: The CRN Project Leaders Forum (PLF) sets research priorities. The CRN Scientific Data Resources Core (SDRC) develops a standard set of variable names and structures that can be used across the health maintenance organizations (HMOs). The CRN Site Data Managers write programs to map their local data into the standardized names and structures. We have developed standard editing routines to detect problems in the data. Users of the VDW are expected to be any of the SAS programmers at the HMOs. The CRN Data Warehouse is "virtual" because the data stay at the individual HMOs and are not centralized until a specific research use has received Institutional Review Board/Health Insurance Portability and Accountability Act (IRB/HIPAA) approvals.
Results: Variability exists among HMOs in implementing the VDW. Some HMOs have disk space available and make permanent copies of their data using the new structures. Others write programs, such as SAS Data Views, that make transient copies of the standard data structures only when required. Some HMOs preselect their data to a population at risk before attempting to run a standard program.
Conclusions: Smaller HMOs create and run programs using the VDW for their entire memberships and can turn around simple requests running programs on the VDW within a matter of hours. Very large HMOs face substantial complexities in adopting the VDW model. The benefits of standardized datasets are clear, but turning around data requests quickly presents some risks. Previously, data processing required enough time for local IRB issues to be settled before programming was finished. The ability to turn around requests in a day means that users need to ensure that IRB/HIPAA requirements are met before local data are distributed to other research centers.
The CRN Cancer Counter–A Web-based Tool to Count Cancers Across Multiple CRN Sites
Hornbrook MC, Bachman DJ, Eubanks D, Beverly K, Hart G
Mark C. Hornbrook, PhD, Kaiser Permanente Center for Health Research, 3800 North Interstate Avenue, Portland, OR 97227-1110. E-mail: firstname.lastname@example.org.
Background: Cancer Research Network (CRN) researchers need accurate, rapid counts of cancers based on specific clinical, temporal, and demographic criteria to produce proposals for deadlines. Previously, researchers developing proposals would send a request to site investigators at some or all CRN sites for cancer count statistics. This prior process works, but it can be slow because it requires several persons per request to complete. This "Cancer Counter" was developed to allow researchers to analyze counts of cancers rapidly. It enables them to explore the data quickly, interactively and iteratively. The "Cancer Counter" can be used to 1) quickly provide information for proposal generation; 2) enable the CRN to rapidly respond to NCI and CRN queries; and 3) improve access to data for quality assurance purposes.
Objective: To describe the development of the CRN's Cancer Counter and profile the early experiences of users.
Methods: Each participating CRN site with a tumor registry sends a de-identified cancer case summary file from their virtual tumor registry to Kaiser Permanente Northwest where the data are loaded into the application. Access is restricted to CRN members. Users can select counts using the following variables: cancer site, morphology, stage, gender, age at diagnosis, vital status, year of diagnosis, and CRN health plan. Security considerations include: 1) that the information produced by this application is de-identified and anonymized, as defined under HIPAA guidelines; 2) requires UserID and password to gain access; 3) that participating HMOs are numbered and not named; 4) that frequency cells showing less than 5 cases will be shown as "less than 5;" 5) the reason for use, date, and all pages viewed by every user be recorded; 6) integrated into the CRN's national secure website; and 7) that the results are SSL encrypted.
Results: The CRN "Cancer Counter" has been used to develop proposals and answer queries. Users report they especially like the ability to query the data quickly and iteratively. CRN sites without tumor registries are presently excluded from the Cancer Counter, but the CRN is developing plans for including cancer counts from claims/encounter data files so that at least the prevalence of cancers can be measured consistently for all CRN sites.
Conclusions: The CRN "Cancer Counter" is a simple yet powerful tool that empowers researchers to quickly ascertain numbers of cancer cases in participating health plans using selection criteria. This tool has demonstrated value in developing proposals, running queries, and improving the quality of data.
Developing a Virtual Data Warehouse: KPCO's Experience
Ellis JL, Newsom B, Mouchawar J
Jennifer L. Ellis, MSPH, Kaiser Permanente Colorado, PO Box 378066, Denver, CO 80237-8066. E-mail: email@example.com.
Background: Virtual Data Warehouses (VDWs), as defined relative to the Cancer Research Network (CRN), are algorithms designed to produce standardized data elements across health plans. VDWs provide a mechanism for quick response to data requests relative to multisite cancer research. The warehouses are virtual both in the sense that data from diverse health plans can be combined consistently and that data may or may not be stored in final form until a request for data is made. VDWs will be developed along content lines (eg, tumor registry, demographics). Information on whether a person has insurance at a given time is basic to much research; therefore the enrollment VDW was given high priority. Methods of developing the enrollment VDW varied by site.
Objective: To profile the experiences of Kaiser Permanente Colorado (KPCO) in developing its Virtual Data Warehouse.
Methods: The CRN's Scientific and Data Resources Core (SDRC) reviewed various definitions of enrollment used in CRN studies and identified elements necessary to most research. Challenges faced at KPCO included determining sources of enrollment data, changing the format of source data to that prescribed by the SDRC, and deciding how best to store the data. Data were extracted from DB2 tables and manipulated using SAS V8.02. KPCO made the decision to store data in datasets defined by date ranges; these data can be combined via a macro according to specific data requests.
Results: Once the sites' VDWs were developed, 2 SAS programs were distributed in November 2003. One program required slight modification to allow for data to be subset prior to pulling from KPCO's mainframe. Both programs, 1 used in conjunction with the demographics VDW and 1 in conjunction with the virtual tumor registry, ran smoothly in a matter of minutes and produced credible output.
Conclusion: Preliminary results indicate that implementation of the enrollment VDW at KPCO was successful. Virtual data warehouses allow health plans to create site-specific algorithms to produce consistently defined variables and to store data in the manner most appropriate to the site. Although implementation of a VDW requires extensive preliminary work, the end product is enhanced ability to respond quickly to diverse research requests.
Distributed Data Model for the Vaccine Safety Link (VSD) Project
Baggs J, Lewis E, Amundson JH, et al.
James M. Baggs, PhD, Centers for Disease Control and Prevention, 1600 Clifton Rd, NE, MS E-61, Atlanta, GA 30333. E-mail: firstname.lastname@example.org.
Background: To monitor and evaluate vaccine safety, the Centers for Disease Control and Prevention (CDC) established the Vaccine Safety Datalink (VSD) project in 1990. As part of this research, the VSD collects automated medical care data on a large population currently or previously enrolled at 8 health maintenance organizations (HMOs) located across the United States. Previous to the year 2001, annual automated data formatted to CDC specifications as SAS datasets were sent as a cycle tape to the CDC. Because of recent heightened confidentiality concerns, this centralized data model has been replaced by a distributed data model.
Objective: To develop and implement a distributed data model for the VSD in which annual automated datasets will reside at the HMO but will remain accessible to CDC data analysts.
Methods: In the distributed data model, the HMOs no longer provide annual automated data sets to CDC. Rather, the data are formatted as SAS datasets and reside at the HMO sites. Two different methods of data access have been developed to compile data sets for specific studies. In the first method, the HMO computer at specified intervals retrieves SAS program code for a specific study from a secure, prespecified location.Four SAS macros have been developed to facilitate access of the data and retrieval of SAS logs and output. In the second method, CDC researchers access the data files required for a specific study through a secure SAS remote session using SAS Connect. Both methods will use encryption technology to insure security of data transmissions.
Results: Distributed data models have been developed and successfully tested at all 8 HMO sites. Six sites developed distributed data models using the first method described above while 2 chose the latter method.
Conclusions: The application of the distributed data model with some site-specific components will permit VSD research to continue in a timely manner while providing state-of-the-art protection of the confidentiality of individual medical information used in VSD studies.
Effectiveness of Decision Support in an Electronic Medical Record on Prescribing in the Elderly
Smith D, Perrin N, Feldstein A, et al.
David Smith, PhD, Kaiser Permanente Center for Health Research, 3800 N Interstate Avenue, Portland, OR 97239. E-mail: email@example.com.
Background: While considerable efforts have focused on disseminating evidence-based guidelines, desired changes in physician behavior have not been fully realized. Systems analysis suggests that computerized physician order entry with decision support systems (eg, alerts and reminders) can reduce medication errors. Data support the effectiveness of these methods in reducing medication errors in inpatients; however, there is little research on their effectiveness in outpatients. Kaiser Permanente Northwest (KPNW) uses an electronic medical record and medication alerts at clinician order entry.
Objective: To study the effect of a computerized physician order entry system with medication alerts and reminders on reducing prescribing errors in the elderly.
Methods: The alerts studied targeted diazepam, flurazepam, imipramine, amitriptyline, chlordiazepoxide, doxepin, and suggested the shorter-acting/less sedating alternatives of oxazepam, buspirone, paroxetine, temazepam, trazodone, nortriptyline, and desipramine. These drug-specific alerts presented upon prescribing, regardless of patient age. We examined the rate of dispensing for these drugs in KPNW members aged ≥65 years, 1 year before and 1 year after the alerts, using an interrupted time series analysis (ARIMA). The main outcome measure was monthly dispensing for target drugs as a fraction of dispensing for all 13 drugs (targeted plus desired alternative drugs).
Results: Prior to the alerts, the raw rate of dispensing for target drugs was 31% for those ≥65 years and 21% for those ≥75 years: these rates fell to 25% and 18% after the alerts. Applying the ARIMA modeling showed an absolute percentage drop of 4.6% ( < .01) for those ≥65 years and 3.5% ( < .05) for those ≥75 years, representing a 15% to 16% relative reduction from baseline. The slope coefficients were also significant (and negative) in the time-series model, suggesting that not only was the interruption (immediate level change) significant, but that the continued downward trend was steeper than before the intervention.
Conclusion: In the outpatient setting, alerts at the time of clinician medication order entry can increase the appropriateness of drug therapy and thus may enhance the safe use of medications, however, some level of potentially inappropriate prescribing remains. Important next steps involve work to assess the generalizability of these findings to other safety issues, and methods to increase clinician acceptance, including studying alerts that are more patient specific.
Electronic Data Collection from Patients on Breast Cancer Risk Factor Information in a Mammography Setting
Aiello E, Taplin S, Reid R, et al.
Erin Aiello, MPH, Center for Health Studies, Group Health Cooperative, 1730 Minor Ave Suite 1600, Seattle, WA 98101. E-mail: firstname.lastname@example.org.
Background: Information on breast cancer risk factors is important to identifying high-risk groups who may be eligible for prevention activities, but the data collection process is time consuming. Data collection currently occurs at Group Health Cooperative (GHC) using a paper survey at the time of the mammogram.
Objective: To evaluate patient acceptance and feasibility of using an electronic data collection questionnaire that pre-fills some answers in an effort to reduce the amount of time spent completing the questionnaire and to improve the accuracy of data collected.
Methods: The Health Insurance Portability and Accountability Act compliant survey software was developed on a Fujitsu Tablet PC and incorporated pre-filled answers from each woman's previous survey. We piloted the prototype in one GHC clinic over a 3-month period. One hundred sixty women were randomized to use the electronic survey (n = 86) or the paper survey (n = 74 controls) and complete an evaluation form. We compared the distribution of Likert scale responses between the intervention and control groups, and between age groups (<60 vs. ≥60 years old).
Results: Overall, 90% of women in the intervention group preferred using the tablet compared to the paper questionnaire. Preference for the Tablet did not differ by age; however, women ≥60 years did not find the Tablet as easy to use as women <60 years. Every woman liked seeing her pre-filled answers; 97% stated that their pre-filled answers were accurate. The majority (65%) did not think that the Tablet was very easy to carry throughout their appointment.
Conclusions: Electronic questionnaires are feasible to use in a mammography setting and are preferred by nearly all women, even older women. Although the Tablet PC was feasible to use as a prototype, this configuration may not be suitable for full deployment in a setting where patient mobility and high volumes are necessary. Clinics elsewhere may have different technology requirements thus requiring further evaluation of different hardware and software options.
Go Codeless…Explore Your SAS Data With a Point and a Click
David M. Rubanowice, Group Health Cooperative-Center for Health Studies, 1730 Minor Avenue, Suite 1600, Seattle, WA 98101. E-mail: email@example.com.
Background: Producing useful output from SAS datasets often requires programmer time, a scarce resource. A method for producing basic descriptive output of study data without SAS programming is needed to improve research team efficiency and efficacy.
Objective: To create a web-based application that will allow researchers to access SAS datasets without needing SAS programming.
Methods: A secure ASP.NET web application was constructed using Microsoft® Visual Basic .NET and Microsoft® Visual Studio® .NET. The web-based user interface facilitates production of dataset and variable listings, one-way variable frequencies, two-way cross tabulations, univariate statistics, and intertable cross tabulations using live research study data stored as SAS datasets. Outputs are easily printed, saved, and distributed as standalone files.
Results: Giving research staff direct access to collected data has reduced wait time for using data and has allowed for more exploratory "what if" analysis compared to traditional methods that involved programming staff.
Conclusions: Web-based access to SAS data has increased the speed and extent to which research data can be used by investigators. Building a front-end for staff use has allowed staff without SAS programming knowledge to work with the data. Retaining SAS as the data storage format allows this application to be a natural extension of the existing programming process, rather than requiring a separate set of analysis datasets.
The HIPAA AnswerBoard: Documenting and Sharing Organizational Learning
Christine C. Stewart, PhD, Group Health Cooperative-Center for Health Studies, 1730 Minor Avenue, Suite 1600, Seattle WA 98101. E-mail: firstname.lastname@example.org.
Background: The Health Insurance Portability and Accountability Act (HIPAA) raises many new compliance issues for researchers in healthcare organizations. While there is a great deal of HIPAA-related information on the Internet, the layering of federal, state, and organizational policy means that many resolutions are organization-specific.
Objective: To create retrievable documentation on HIPAA-related issues specific to research at our organization, starting early in the implementation of the HIPAA Privacy Rule in 2003.
Methods: The AnswerBoard is a web application that displays information in a categorized question-and-answer format on the Center for Health Studies (CHS) Intranet. The application was built using Microsoft ASP.NET and a SQL Server 2000 database, and was designed to be easily re-used for other topics. When new questions arise, they are brought to the CHS HIPAA Committee, and after the issue is resolved, it is added to the AnswerBoard's content.
Results: The AnswerBoard provides a first stop for staff to look for the answer to any HIPAA question. Each question can be listed under several categories, making it easy to find an area of interest. Answers may contain links to other web sites.
Conclusions: The HIPAA AnswerBoard has increased the transparency of emerging HIPAA policies in our department, and eliminated time spent "re-inventing the wheel" as the same issues recur over time in different projects. The AnswerBoard application is generic and appropriate for documenting any developing body of knowledge and has 2 significant advantages over the typical "frequently-asked questions" web page. First, the categorization of questions permits easy navigation through a large volume of items; second, the web interface for posting new information means that subject-area experts can update the information without relying on a web designer or developer.
The Human Infrastructure of Information Technology at the Center for Health Studies
Julia Hecht, PhD, MPH, Group Health Cooperative-Center for Health Studies, 1730 Minor Avenue, Suite 1600, Seattle, WA 98101. E-mail: email@example.com.
Background: Since the Center for Health Studies was established in 1983 there has been a net expansion in information technology (IT) staff from 1 to 31. CHS/IT is organized differently from IT at many of the other HMORN sites.
Objective: The primary objective is to explore the growth and the organization of IT staff and its relationship to the work of the Center. An additional aim of this descriptive presentation is to foster a discussion of different modes of organization in IT staff across the HMORN.
Methods: Growth is examined through study of tenure and years of hire. Patterns of organization and areas of concentration of IT staff are examined.
Results: Information technology staffing has expanded relatively steadily with marked growth in 1994-1998. Center for Health Studies IT staff has diverse backgrounds and experience. Programmers are assigned primarily to projects with overarching management of the programmers as a group and some development of subgroups. This project-centered model contrasts to a model of programmer-investigator dyads. Another organizational model is a programming pool where programmers are assigned to specific tasks rather than longer-term projects. Some form of overarching management may accompany each of these 3 models.
Conclusions: The Center for Health Studies IT group structure and organization has valuable advantages in parlaying programmer knowledge within subject areas while encouraging communication among programmers. There are disadvantages to this organizational form, one being that it may be difficult to meet unanticipated staffing needs under this structure. Other structures for IT management have different strengths and weaknesses. For example, a programming pool might be better able to respond to last-minute programming requests but exhibit less ability to manage long-term staffing needs. Dyads develop depth of subject-area knowledge and efficiencies based on knowing one's work partner very well. However, there may be less opportunity for exchange of ideas across groups without some overarching organization under any model, whether it be investigator-programmer dyad, project-centered, or programmer pool.
Information Technology in Ambulatory Care: Use of the Automated Medical Record Over Time
Kinsman CJ, Huang J, Fireman B, Robertson N, Frankel R, Hsu J
Charles James Kinsman, Kaiser Permanente Division of Research, 2000 Broadway, Oakland, CA 94612. E-mail: firstname.lastname@example.org.
Background: There currently is little data on how often and extensively providers use information technology (IT) during individual visits.
Objective: To examine the frequency and level of use of an automated medical record (AMR) by adult primary care providers (PCP) during ambulatory care.
Methods: We conducted a longitudinal study on IT use by 9 PCPs in a large, prepaid integrated delivery system (IDS), after the introduction of an AMR into exam-rooms. We video-recorded and coded 159 study visits during 2 periods (84 in the first month after introduction [immediate], and 75 six months later [experienced]). We assessed 3 categories of AMR use: viewing, documentation, and advanced care management (eg, graphing laboratory results over time, reviewing online clinical guidelines.) We assessed the number and proportion of visits with any use, and with each category of use. Finally, we assessed whether use changed over time using a generalized estimating equation (GEE) model with adjustments for clustering by PCP and for patient and visit characteristics.
Results: The patient subjects had a mean age of 53 years, tended to be female (67%), reported incomes greater than $35 000 (63%) and have poor health (67% not excellent or very good). PCPs were trained in family practice (56%) or internal medicine (44%) with ≥3 years experience practicing in the IDS (80%) and tended to be female (56%). In addition, 78% of the visits were for urgent problems (very important or important), and 15% were initial visits with the PCP. Overall, PCPs used the AMR in 82% of visits (131/159): 83% (70/84) immediately, and 81% (61/75) with experience. PCPs used the AMR for viewing most frequently (81%), followed by documentation (62%), and care management (20%). In our multivariate model, there were no statistically significant differences in viewing, documentation, and care management between immediate and experienced use of the AMR.
Conclusion: Primary care providers appear to frequently use the AMR for basic functions such as viewing and documentation. In contrast, PCPs infrequently used care management features during the visit. Future research is needed to assess the perceived utility, usability, and predictors of AMR use in order to achieve potential quality benefits.
A Knowledge-Based Information Retrieval System (KBIRS)
Hazlehurst BL, Zade R, Sittig DF
Brian Hazlehurst, PhD, Kaiser Permanente Center for Health Research, 3800 N. Interstate Avenue, Portland, OR 97227. E-mail: email@example.com.
Background: The push for providing high quality healthcare based on the "evidence" has caused many clinical information systems developers to begin exploring ways to provide clinicians with access to clinical reference and patient education materials at the point of care. For a given care situation, the availability of vast amounts of patient-specific clinical data and generally applicable clinical knowledge often makes identification of the most relevant information for a particular patient difficult, if not impossible, for the patient or even the clinician. We have developed a prototype system capable of taking a copy of a patient's electronic medical record, which includes their current clinical problems, medications, and laboratory test results among other clinical data, and automatically generates a set of highly relevant, condition and patient-specific educational documents.
Objective: To design and implement a prototype knowledge-based information retrieval system (KBIRS) that tailors education materials to patient needs.
Methods: Briefly, KBIRS takes as input a copy of a patient's electronic medical record in an emerging HL-7 standard format called the Clinical Document Architecture (CDA). It then uses a condition-specific knowledge model created using the Protégé knowledge modeling software to identify key concepts and relationships from the patient's data. For example, it may identify that a patient with asthma has recently been started on a new corticosteroid. An algorithm, using additional clinical knowledge, then converts this set of facts and relationships into a query that can be run against a standard web-based patient education resource (eg, Healthwise) to identify and retrieve one or more patient-specific documents.
Results: We have recently completed end-to-end testing of a prototype that includes clinical knowledge to interpret an asthma patient's clinical record and recommend relevant documents. The system has been developed entirely using freely available open-source software tools and techniques. Preliminary results suggest that the system performs as well as patients with intermediate-level Internet searching capabilities.
Conclusions: We believe that such a knowledge-based information retrieval system illustrates the future of emerging trends in clinical knowledge-management and distributed clinical computing. Bringing context-dependent clinical knowledge to bear at the right time and place to help clinicians and patients make the right decision is one of the key factors in facilitating patient-focused clinical decision making.
Transferring Confidential Research Data: Using a Secure Web-Based File Transfer Protocol
Hornbrook MC, Ansell GC, Bachman DJ, Eddy C, Thornton M
Mark C. Hornbrook, PhD, Kaiser Permanente Center for Health Research, 3800 North Interstate Avenue, Portland, OR 97227-1110. E-mail: firstname.lastname@example.org.
Background: The Health Insurance Portability and Accountability Act (HIPAA) significantly increased the required level of protection of personal health information. In many studies, it is necessary to move research data from one site to another.
Objective: To describe the development of a secure Web-based file transfer application to increase the security of transmitting data from one physical location to another.
Methods: This application uses e-commerce encryption and verification technology and standards to provide secure and rapid file transfers. This application transfers data for approved healthcare research (including sensitive data) from an approved user to an approved receiver using a secured Web site. Confidentiality and security of the file transmissions are ensured using a combination of certificate authentication when clients connect to the host server, custom programming to secure individual access to forms and other documents, and Secure Sockets Layer (SSL) encryption to ensure that confidential data cannot be revealed even if packets were intercepted in transmission. Access is restricted to 128-bit key versions on Netscape and Internet Explorer, which provide a higher level of encryption than other browsers. Person-level user IDs and passwords and client certificates are required to log on.
Results: During the past 26 months, the CHR has established about 200 users with client certificates, encompassing 40 projects. Once access is approved, the user uploads a file to the Web site and the receiver logs onto the approved folder and file and downloads it. Notifying the receiver that the file will be posted at a specified time can minimize security risk exposure. The system automatically logs out a user after 15 minutes of nonuse.
Conclusions: Users have successfully and easily transferred 100MB files and larger. Most Institutional Review Boards asked to approve use of the secure file transfer system has approved it. This successful application is being used by many multisite studies. This method reduces the risk of inadvertent delivery to unintended location as well as the risk of interception by unauthorized users compared to other transfer modes.
Using Claims Data to Identify Cases of Asthma
Williams LK, Kolk D, Krajenta R, et al.
L. Keoki Williams, MD, MPH, Henry Ford Health System, Center for Health Services Research, 1 Ford Place, Suite 3A, Detroit, MI 48202. E-mail: email@example.com.
Background: Although claims data are a relatively inexpensive and abundant source of patient-level data for epidemiologic research, there are few established algorithms for identifying cases of asthma. The Health Plan Employer Data and Information Set (HEDIS) include criteria for persistent asthma; however, these may be too restrictive for epidemiologic purposes.
Objective: To compare the number of asthmatics identified by HEDIS to those identified by a combination of diagnosis and medication use (DMU).
Methods: Patients were members of a large health maintenance organization in southeast Michigan. We included individuals aged 5-56 years who had been enrolled in the health plan between January 1, 1999 and December 31, 2000. Individuals met the HEDIS criteria for persistent asthma if they had the following in the year 2000: ≤4 fills of an asthma medication (AM); ≤1 emergency department visit for asthma ≤1 hospitalization for asthma; or ≤4 outpatient visits for asthma and ≤ 2 fills of an AM. We also identified individuals who had one or all of the following: ≤1 visit for asthma; ≤1 fill of a controller medication (CM); and ≤1 fill of a beta-agonist medication (BAM). An individual was also considered to have persistent asthma if they had all three (DMU).
Results: A total of 18 734 individuals had either ≤1 visit for asthma (8459)≤1 fill of a CM (7121); or ≤1 fill of a BAM (16 170); 3821 individuals had all three. Of the 5342 individuals who fulfilled the HEDIS criteria, a majority (2832 (53.0%) also fit the criteria for DMU. However, 989 (25.9%) of 3821 individuals with likely persistent asthma as defined by DMU did not meet HEDIS criteria.
Conclusion: The HEDIS criteria, which have been used to identify cases of persistent asthma, may be too restrictive as they missed a number of individuals whose profiles were also consistent with persistent disease. Both HEDIS and DMU may be required to identify persons with persistent asthma.
Using Discussion Mailing Lists for Communication and Information Storage on Multi-Site Research Studies
Robert Hartl, MS, Group Health Cooperative-Center for Health Studies1730 Minor Avenue, Suite 1600, Seattle, WA 98101. E-mail: firstname.lastname@example.org.
Background: The Center for Health Studies (CHS) participates in a number of collaborative research studies with other health research organizations. One of the challenges of multisite collaborative research is the communication and discussion of study issues among study team members across all sites. Related to this is the need for an information and document store, of research study materials, that is readily accessible by all study team members.
Objective: To report on the usefulness of web-based discussion mailing lists for facilitating and promoting multisite collaborative research communication at CHS.
Methods: For each of several multi-site research studies of which the CHS is a participant, a discussion mailing list has been created using simple and inexpensive listserv software running on a CHS web server. A listserv-hosted discussion mailing list is a combination e-mail and web solution. A mailing list member contributes a message to the mailing list and the listserv distributes the message to all mailing list members. The listserv also archives all mailing list messages, providing a readable and searchable message database via an easy-to-use web—based user interface.
Results: A listserv-hosted discussion mailing list effectively facilitates the discussion of research issues among study team members. When research study documents are distributed to study team members as mailing list message attachments, the listserv provides not only an effective means of message/document distribution, but also a readily accessible store of those documents. Use of a listserv is particularly effective on a long-term study and with changes to study team members (ie, the listserv manages the email distribution list and the message/document store).
Conclusions: The use of discussion mailing lists is not only an effective solution, but also can be implemented using relatively inexpensive standard listserv software. Once implemented, additional mailing lists can be easily added and maintained.
Web-based Dx/Procedure Frequency Plot Generator
Pardee R, Rubanowice D, Hart G
Roy E. Pardee, JD, MA, Group Health Cooperative-Center for Health Studies, 1730 Minor Avenue, Suite 1600, Seattle, WA 98101. E-mail: email@example.com.
Background: In order to probe for practice and/or data differences resulting from the gradual clinic-by-clinic rollout of a new clinical information system (EPIC) we built a browser-based application that generates time series plots of individual diagnosis and procedure code frequencies by clinic over time. The application allows investigators and programmers to check for discernible differences in the frequency with which diagnoses and procedures are coded at clinics pre- and post-EPIC rollout, and compare these to clinics that have not yet implemented EPIC.
Objective: To explore the feasibility and use of a web-based tool for investigation of effects of a new Clinical Information System on automated data.
Methods: SAS outpatient encounter data were summarized by code, month, and clinic. These summarized data were fed into a server database. We then built a browser-based interface by which users can select codes clinics, and dates of interest, and generate lineplots of selected code/clinic frequencies. The web server application uses asp.net, the lineplots are displayed using SAS Institute's SAS/Graph ActiveX control.
Results: The lineplots generated by the tool are very revealing of the variability and trend of code frequencies over time at GHC clinics. There is little evidence thus far of precipitous differences in code frequencies over time, although the EPIC data are only now starting to become available.
Conclusions: Investigators and programmers will be able to use this tool to explore the extent to which the switch to EPIC has affected the data, and to visualize time trends in outpatient encounter coding generally.