Portal de Periódicos da CAPES

Building a data warehouse for infection control

2008; Elsevier BV; Volume: 36; Issue: 3 Linguagem: Inglês

10.1016/j.ajic.2007.07.004

1527-3296

William E. Trick,

Clinical Laboratory Practices and Quality Control

Recently, the convergence of several themes in health care has stimulated interest in exploiting data collected during clinical encounters to monitor and improve the quality of health care. Infection control departments can be supported by assembling previously collected electronic data into a data warehouse to monitor infections, report communicable diseases, and detect clusters of organisms. I present the rationale for constructing a data warehouse and describe the steps and challenges of developing a data warehouse to support infection control activities. Recently, the convergence of several themes in health care has stimulated interest in exploiting data collected during clinical encounters to monitor and improve the quality of health care. Infection control departments can be supported by assembling previously collected electronic data into a data warehouse to monitor infections, report communicable diseases, and detect clusters of organisms. I present the rationale for constructing a data warehouse and describe the steps and challenges of developing a data warehouse to support infection control activities. For many years, computers have been used to improve the efficiency of health care delivery. For most institutions, progress has been realized primarily for recording and retrieving information for provider–patient encounters. In addition to improving care during an individual patient encounter, there is great potential for improving the completeness and efficiency of monitoring the quality of care through increased penetration of information technology.1Chaudhry B. Wang J. Wu S. et al.Systematic review: impact of health information technology on quality, efficiency, and costs of medical care.Ann Intern Med. 2006; 144: 742-752Crossref PubMed Scopus (2279) Google Scholar, 2Evans R.S. Pestotnik S.L. Classen D.C. Bass S.B. Burke J.P. Prevention of adverse drug events through computerized surveillance.Proc Annu Symp Comput Appl Med Care. 1992; : 437-441PubMed Google Scholar Recently, the convergence of several themes in health care has stimulated interest in exploiting data collected during the clinical encounter to monitor and improve the quality of health care. These themes include an increasing number of physicians who use electronic medical records to document health care encounters;3Centers for Disease Control and Prevention. National Center for Health Statistics. Electronic medical record use by office based physicians: United States, 2005. Available from: http://www.cdc.gov/nchs/products/pubs/pubd/hestats/electronic/electronic.htm. Accessed April 2007.Google Scholar prioritization of a national health information infrastructure,4Yasnoff W.A. Humphreys B.L. Overhage J.M. Detmer D.E. Brennan P.F. Morris R.W. et al.A consensus action agenda for achieving the national health information infrastructure.J Am Med Inform Assoc. 2004; 11: 332-338Crossref PubMed Scopus (101) Google Scholar and increased demands for public reporting of performance measures (eg, hospital-acquired infections).5Wong E.S. Rupp M.E. Mermel L. Perl T.M. Bradley S. Ramsey K.M. et al.Public disclosure of healthcare-associated infections: the role of the Society for Healthcare Epidemiology of America.Infect Contr Hosp Epidemiol. 2005; 26: 210-212Crossref PubMed Scopus (42) Google Scholar In addition to external demands that may stimulate change, many professionals within the infection control and public health fields recognize that use of electronic data can improve the efficiency and accuracy for monitoring health care–associated infections6Platt R. Yokoe D.S. Sands K.E. Automated methods for surveillance of surgical site infections.Emerg Infect Dis. 2001; 7: 212-216Crossref PubMed Scopus (66) Google Scholar, 7Evans R.S. Larsen R.A. Burke J.P. Gardner R.M. Meier F.A. Jacobson J.A. et al.Computer surveillance of hospital-acquired infections and antibiotic use.JAMA. 1986; 256: 1007-1011Crossref PubMed Scopus (195) Google Scholar and reporting communicable diseases to public health departments.8Panackal A.A. M'Ikanatha N.M. Tsui F.C. McMahon J. Wagner M.M. Dixon B.W. et al.Automatic electronic laboratory-based reporting of notifiable infectious diseases at a large health system.Emerg Infect Dis. 2002; 8: 685-691Crossref PubMed Scopus (75) Google Scholar, 9Overhage J.M. Suico J. McDonald C.J. Electronic laboratory reporting: barriers, solutions and findings.J Public Health Manag Pract. 2001; 7: 60-66Crossref PubMed Scopus (54) Google Scholar In addition to these core functions, a data warehouse also can be used for other purposes—for example, tracking the occupancy of rooms designated for patients on isolation precautions, monitoring the incidence of antimicrobial-resistant organisms, and calculating trends in antimicrobial use. As demands on infection control departments increase, constructing a data warehouse may increase the efficiency and productivity of the department by automating time-consuming repetitive tasks using computer systems.10Wisniewski M.F. Kieszkowski P. Zagorski B.M. Trick W.E. Sommers M. Weinstein R.A. Development of a data warehouse for hospital infection control.J Am Med Inform Assoc. 2003; 10: 454-462Crossref PubMed Scopus (119) Google Scholar In this article, I present the rationale for constructing a data warehouse to enhance infection control activities, discuss the steps and challenges of developing a data warehouse, and highlight a few advances that herald future uses of data warehouses to support infection control activities. Surveillance of health care–associated infections consumes many work hours for infection control personnel. Historically, identifying potential infections required manual reviews of paper-based microbiology reports with subsequent creation of lists of patients to evaluate. Often, after bedside evaluation of the patient in conjunction with medical record review, paper forms were completed, followed by manual data entry into local databases or a centralized reporting system. Over time, for some institutions, these tasks have been fully or partially automated depending on the sophistication of the institution's information system and support from informatics departments. The substantial amount of time spent identifying possible health care–associated infections or reportable diseases could be reallocated to efforts focused on improving processes of care that have been shown to improve health care quality and protect patients.11Berenholtz S.M. Pronovost P.J. Lipsett P.A. Hobson D. Earsing K. Farley J.E. et al.Eliminating catheter-related bloodstream infections in the intensive care unit.Crit Care Med. 2004; 32: 2014-2020Crossref PubMed Scopus (733) Google Scholar, 12Warren D.K. Zack J.E. Cox M.J. Cohen M.M. Fraser V.J. An educational intervention to prevent catheter-associated bloodstream infections in a nonteaching, community medical center.Crit Care Med. 2003; 31: 1959-1963Crossref PubMed Scopus (97) Google Scholar The Centers for Disease and Control and Prevention (CDC)'s National Nosocomial Infections Surveillance system, which has evolved into the National Healthcare Safety Network, established the definitions and surveillance methods to identify health care–associated infections. The CDC's system provides benchmarks for device-associated infection rates, compares unit-specific rates, and provides feedback to participants. Although widely recognized as the standard for health care–associated infection surveillance systems, as recognized by CDC experts, there are opportunities to improve the efficiency, and perhaps the reliability, of surveillance.13Tokars J.I. Richards C. Andrus M. Klevens M. Curtis A. Horan T. et al.The changing face of surveillance for health care-associated infections.Clin Infect Dis. 2004; 39: 1347-1352Crossref PubMed Scopus (106) Google Scholar In particular, there are concerns regarding the amount of time required for surveillance using manual methods, and inconsistent methods both for identifying potential infections and in the application of definitions. Excessively labor-intensive methods constrain surveillance activities to limited populations (eg, intensive care patients), which results in interventions focused on these same populations. Inconsistent application of definitions compromises the reliability of determinations, which complicates interhospital comparisons of infection rates. Recent legislative requirements to publicly disclose infection rates likely will result in deleterious economic consequences for institutions that report high infection rates, which may effect the application of infection definitions and further reduce interrater reliability. An example of a relatively simple form of automation is to create lists of patients that need further evaluation. For example, using the data warehouse at our institution, we monitor hospital-acquired bloodstream infections by creating a daily list of isolates from positive blood cultures obtained from the patient care units of interest. Despite the apparent simplicity of such a request, some hospital information systems do not allow for easy access to clinical datasets; therefore, information retrieval may depend on the system's proprietary report writer. These reports are not always easily transferred to spreadsheets or analytic programs, and may require manual review for interpretation (Fig 1). In other instances, infection control personnel may have the skills needed to extract data; however, informatics departments are focused on supporting the provider–patient clinical encounter. Reasons informatics departments may prevent access to data include the potential for degradation of system performance while data are being extracted from the system's transactional server and concerns about the security of data. Construction of a separate data warehouse provides the opportunity to extract data without compromising the performance of the clinical information system. Compared with generation of patient lists, a more sophisticated application of electronic data is to screen patients and identify those who have a high probability of infection. Automated detection of surrogates of infection becomes increasingly complicated as the number of requisite data tables increases. For example, algorithms that include data stored in a single table (eg, billing codes or microbiology data) only require accessing and understanding data from a single table. There have been several investigations of automating infection detection—a few examples appear in Table 1. In general, automated methods of infection detection improve the sensitivity of event detection. For example, surgical site infections often are missed due to incomplete monitoring of patients during their hospital stay or after hospital discharge. By applying algorithms to electronic data, postoperative patients who have an increased likelihood of surgical site infection can be identified for further record review.6Platt R. Yokoe D.S. Sands K.E. Automated methods for surveillance of surgical site infections.Emerg Infect Dis. 2001; 7: 212-216Crossref PubMed Scopus (66) Google Scholar Using standardized screening criteria should improve reliability by standardizing the method for identifying potential infections. In addition to identifying surgical site infections, algorithms to fully or partially automate detection of other sites have been tested and have demonstrated performance that equals or exceeds manual methods (Table 1).Table 1Studies that have evaluated full or partial automation of infection detectionAuthorsEventData sourcesSummaryEvans et al7Evans R.S. Larsen R.A. Burke J.P. Gardner R.M. Meier F.A. Jacobson J.A. et al.Computer surveillance of hospital-acquired infections and antibiotic use.JAMA. 1986; 256: 1007-1011Crossref PubMed Scopus (195) Google ScholarBSIADT (All sites)Reference standard—interpretation by infectious disease physicians.LRTIMicrobiology (all sites)SSIChemistry (UTI)Sensitivity of automated detection exceeded that of infection control professional review (78% vs 68%)UTISurgery (SSI)Radiology (LRTI)Respiratory therapy (LRTI)Haas et al21Haas J.P. Mendonca E.A. Ross B. Friedman C. Larson E. Use of computerized surveillance to detect nosocomial pneumonia in neonatal intensive care unit patients.Am J Infect Control. 2005; 33: 439-443Abstract Full Text Full Text PDF PubMed Scopus (35) Google ScholarPneumoniaRadiology (text interpretation)Performance of infection detection: sensitivity = 71%, specificity = 99.8%Spolaore et al23Spolaore P. Pellizzer G. Fedeli U. Schievano E. Mantoan P. Timillero L. et al.Linkage of microbiology reports and hospital discharge diagnoses for surveillance of surgical site infections.J Hosp Infect. 2005; 60: 317-320Abstract Full Text Full Text PDF PubMed Scopus (22) Google ScholarSSIICD9 codesPositive predictive value = 72%Microbiology (wound cultures)Trick et al24Trick W.E. Zagorski B. Tokars J.I. Vernon M.O. Welbel S.F. Wisniewski M.F. et al.Computer algorithms to detect bloodstream infections.Emerg Infect Dis. 2004; 10: 1612-1620Crossref PubMed Scopus (133) Google ScholarBSIADT–bed informationComputer algorithm with manual central venous catheter determination better correlated (К) with reference standard compared with separate manual review (0.72 vs 0.48)MicrobiologyPharmacy (antibiotic exposure)Wright et al.16Wright M.-O. Perencevich E.N. Novak C. Hebden J.N. Standiford H.C. Harris A.D. Preliminary assessment of an automated surveillance system for infection control.Infect Contr Hosp Epidemiol. 2004; 25: 325-332Crossref PubMed Scopus (40) Google ScholarOutbreaksMicrobiologyPositive predictive value of automated alerts (62%), sensitivity of routine surveillance (45%)Yokoe et al25Yokoe D.S. Noskin G.A. Cunnigham S.M. Zuccotti G. Plaskett T. Fraser V.J. et al.Enhanced identification of postoperative infections among inpatients.Emerg Infect Dis. 2004; 10: 1924-1930Crossref PubMed Scopus (97) Google Scholar∗Algorithms were tested at multiple centers; not all centers could run the algorithms on electronic records.SSIICD-9 codesInfection prevalence higher using a system to augment infection detection (К = 0.66 for separate manual reviews)Pharmacy (antibiotic exposure)Abbreviations: BSI, bloodstream infection; LRTI, lower respiratory tract infection; SSI, surgical site infection; UTI, urinary tract infection.∗ Algorithms were tested at multiple centers; not all centers could run the algorithms on electronic records. Open table in a new tab Abbreviations: BSI, bloodstream infection; LRTI, lower respiratory tract infection; SSI, surgical site infection; UTI, urinary tract infection. Using manual methods, reports of communicable diseases (eg, hepatitis A, tuberculosis, and bacterial meningitis) are delayed and often incomplete.9Overhage J.M. Suico J. McDonald C.J. Electronic laboratory reporting: barriers, solutions and findings.J Public Health Manag Pract. 2001; 7: 60-66Crossref PubMed Scopus (54) Google Scholar, 14Birkhead G. Chorba T.L. Root S. Klaucke D.N. Gibbs N.J. Timeliness of national reporting of communicable diseases: the experience of the National Electronic Telecommunications System for Surveillance.Am J Public Health. 1991; 81: 1313-1315Crossref PubMed Scopus (22) Google Scholar, 15Curtis A.B. McCray E. McKenna M. Onorato I.M. Completeness and timeliness of tuberculosis case reporting. A multistate study.Am J Prev Med. 2001; 20: 108-112Abstract Full Text Full Text PDF PubMed Scopus (55) Google Scholar Some public health departments have used automated electronic laboratory reporting to improve the timeliness and completeness of reporting.8Panackal A.A. M'Ikanatha N.M. Tsui F.C. McMahon J. Wagner M.M. Dixon B.W. et al.Automatic electronic laboratory-based reporting of notifiable infectious diseases at a large health system.Emerg Infect Dis. 2002; 8: 685-691Crossref PubMed Scopus (75) Google Scholar Because direct connection to laboratory information systems is not always possible, and because stand-alone laboratory information systems do not contain all data relevant to reporting communicable diseases, creation of a data warehouse can facilitate the transfer of information. Even without full automation, case identification can be improved by generating a list of organisms or laboratory results that indicate a communicable disease. If a local health department has the capacity to electronically accept results and all of the relevant data elements are contained in the data warehouse, results can be messaged directly to the health department. Alternatively, the form required by the health department can be electronically created, prepopulated with data from the data warehouse, and faxed to the health department. Our infectious disease division's data warehouse prepopulates the acute hepatitis infection form with serologic results and demographic data. Then, the form is manually reviewed for accuracy and completeness and sent to the Chicago Department of Public Health. In addition to infection surveillance, another core function of infection control departments is to detect and respond to clusters of infections or organisms. Rather than relying on astute laboratory personnel or clinicians to identify an increased incidence of infection by specific organisms, the sensitivity of detection of clusters of organisms can be enhanced using analytic methods applied to electronically stored data.16Wright M.-O. Perencevich E.N. Novak C. Hebden J.N. Standiford H.C. Harris A.D. Preliminary assessment of an automated surveillance system for infection control.Infect Contr Hosp Epidemiol. 2004; 25: 325-332Crossref PubMed Scopus (40) Google Scholar, 17Samore M. Lichtenberg D. Saubermann L. Kawachi C. Carmeli Y. A clinical data repository enhances hospital infection control.Proc AMIA Annu Fall Symp. 1997; : 56-60PubMed Google Scholar Although these are intriguing and innovative approaches to outbreak detection, increased detection of clusters of organisms has the cost of increasing the number of alerts that require investigation, regardless of whether intervention is required or would be beneficial. Additional research is needed to quantify the work required to respond to alerts and evaluate the circumstances for which reduced transmission can be achieved through timely intervention. At our hospital, infection control professionals use the data warehouse to assist with implementation of contact isolation precautions. For this function, the bed location of patients on isolation precautions is reported and the movement of patients through the hospital is tracked. Also, patients who had an antibiotic-resistant organism recovered during a previous hospitalization are flagged by the system and contact isolation precautions are initiated. For infection control departments that have limited experience working with their institution's informaticists, the prospect of creating a data warehouse is daunting. However, the complexity of developing a data warehouse can be considered as a continuum; the process can begin with extracting, cleaning, and understanding a single relevant table (eg, microbiology table). Over time, the data warehouse can evolve to incorporate data elements contained in other tables of interest as determined by the goals of the infection control department. Examples of candidate tables of interest to infection control departments are displayed (Fig 2, Table 2). Although it is worthwhile to contemplate the entire scope of the project, if the ultimate goal is not feasible, the project should be scaled back to meet less lofty goals.Table 2Examples of tables and fields relevant to infection control functions∗The tables are organized by department or function.Tables of interestFieldsExamples of dataUtilityCommentMicrobiologyCulture sourceWound, blood, sputum, urine, stoolCreation of line list for evaluating potential infectionsAdditional, more descriptive field likely available (eg, left foot)Organism nameStaphylococus aureusCalculate organism-specific incidenceMay be a free text field that requires cleaningAntibiotic susceptibilityResult by antibiotic and cut-point (eg, S, R)Identification of patients who may need isolation precautionsLikely a free text field that requires cleaningAdmission, discharge, and transferAdmission and discharge datesCalculation of difference between event and hospital admissionRegistration and admission dates may differ due to admission delaysIdentifier, visit levelVisit number; financial numberProvides key to link data between tables and identify a visitSome facilities assign a new visit number for intrafacility transfersIdentifier, patient levelMedical record numberAllows for tracking patients on repeat visits to the health care systemBed informationSpecific bed locationShould be populated dailyAssignment of events to a unit; calculation of unit-specific censusPatient service may be assigned once in a separate tableBilling codesCPTScreening for potential infectionsInterinstitutional variability; may be coding delayICD-9-CMPharmacyAntibiotic transactionsCalculation of antimicrobial use; used in screening postoperative patients for infectionFree text entries common; calculation complicated by returned dosesRouteIntravenous, oral, intramuscular, topicalTopical use excluded from calculationsCPT, current procedural terminology.∗ The tables are organized by department or function. Open table in a new tab CPT, current procedural terminology. As previously described,10Wisniewski M.F. Kieszkowski P. Zagorski B.M. Trick W.E. Sommers M. Weinstein R.A. Development of a data warehouse for hospital infection control.J Am Med Inform Assoc. 2003; 10: 454-462Crossref PubMed Scopus (119) Google Scholar one path to creating a data warehouse is as follows:1.Establish a business understanding of the project. This involves determining and prioritizing the core functions for which you intend to use a data warehouse. At this stage, it is helpful to engage colleagues who access a data warehouse at their own institution and formulate goals within your infection control department.2.After an initial business plan is developed, the data tables required to realize your plan need to be identified and understood. There is considerable variability in database structures, so it may help to consider the tables as entities within the larger structure of separate departments or hospital functions (Fig 2, Table 2). Understanding the structure and content of the data tables includes learning the department-specific processes that generate data. For example, microbiology results are dynamic, progressing from Gram stain results to preliminary and final organism identification that includes the results of susceptibility tests.3.Discuss the options for receiving and housing data with your institution's informaticists. Options include the following: the informatics department can establish a data mart within the institution's servers with permission granted for infection control department access, transfer of data to a server maintained by the infection control department through batch downloads, or transfer of data to the server through real-time messages (most messages use HL7 standards for health care information exchange).18Health level seven I. Available from: http://www.hl7.org/. Accessed July 3, 2007.Google Scholar This final option requires relatively advanced technical skills to receive and parse messages into a usable format.4.Develop a plan to access and extract data from your data warehouse into analytic programs or reports that were specified in your business plan. For example, reports may be a list of patients who have a reportable communicable diseases, or antibiotic-resistant organisms, or a report can be a template formatted for data transfer to the local health department.5.Obtain approval for your proposed data warehouse with your institution's administrators. Important issues to be considered include describing intended uses of the data, compliance with HIPAA regulations, and assurances of data security—including identifying the individuals who can access the data. Also, inform local department heads (eg, laboratory or pharmacy) about the project and be prepared to justify the benefits of using "their" data for infection control purposes.6.After data acquisition, validate completeness and accuracy of the data transfer and further understand the data.7.Maintain the data warehouse through the evolution of your facility's information system. This includes establishing a plan to monitor the completeness of ongoing data transmission. Most challenges should fall into of the following three often intertwined domains: financial, technical, and political. Because infection control departments often have limited discretionary funds, financing the project requires convincing decision-makers that an investment in information technology would increase your efficiency and lead to better quality patient care. It may be possible to enlist the support of administrators from other departments who also desire improved access to data. One critical selling point is that the number of work hours saved from automation of manual processes can be substantial.7Evans R.S. Larsen R.A. Burke J.P. Gardner R.M. Meier F.A. Jacobson J.A. et al.Computer surveillance of hospital-acquired infections and antibiotic use.JAMA. 1986; 256: 1007-1011Crossref PubMed Scopus (195) Google Scholar, 10Wisniewski M.F. Kieszkowski P. Zagorski B.M. Trick W.E. Sommers M. Weinstein R.A. Development of a data warehouse for hospital infection control.J Am Med Inform Assoc. 2003; 10: 454-462Crossref PubMed Scopus (119) Google Scholar, 19Bouam S. Girou E. Brun-Buisson C. Karadimas H. Lepage E. An intranet-based automated system for the surveillance of nosocomial infections: prospective validation compared with physicians' self-reports.Infect Control Hosp Epidemiol. 2003; 24: 51-55Crossref PubMed Scopus (44) Google Scholar Depending on the scope of the project, the technical expertise within your department, and the availability of assistance from your informatics department, you may require assistance from an external technical consultant. If assistance is contracted, the work product should be transparent (eg, the programming code and decisions made for data extraction should be provided to you for future modifications). Regarding political challenges, expect to confront issues of data ownership. For example, some individuals may be threatened by the prospect of someone analyzing "their" data, whereas others may consider themselves responsible for data security. A leverage point is to offer to provide reports that can help the local departments despite these reports not being central to meeting your objectives. Some of the lessons learned:•Political barriers may be more challenging than technical barriers. To overcome these barriers, establish protocols to safeguard data and receive assurances from risk management.•Free text data entry will be allowable for some fields (eg, "Late growth of… YEAST"), mapping these entries to usable terms requires an initial investment of time.•After data transfer, manually review a sample of records to validate electronically transferred data. At our institution, data are stored using the Microsoft SQL Server (Microsoft Inc. Redmond, WA) platform. There are separate tables for the following domains: demographic, radiographic, laboratory, microbiology, pharmacy, daily bed information, coded diagnoses, and operating room procedures (Fig 2). Data are stored in a relational database, allowing for joining data from separate domains either through a visit- or patient-level identifier. These data are accessible to preapproved users either by direct query using structured query language or open database connectivity connection to analytic programs (eg, SAS [SAS Institute, Cary, NC]) or through proprietary report writing software (Crystal Reports; Business Objects Inc., San Jose, CA). In addition to these primary tables, additional tables of aggregate data are created (eg, patient census) to increase the efficiency of report generation. In the future, purchasers of information systems likely will recognize the value of facilitating local access to data; such as improvements to patient care. Eventually, vendors should respond to requests for improved access to data. In particular, data access should be enabled through use of common database languages and nonproprietary tools. In addition to improved access to data already present in most systems, information systems will evolve to represent the increasing amounts of data captured during the patient care encounter in a usable format as coded data elements. For example, to calculate infection rates, it would be useful to capture patient-level invasive devices use (eg, central lines and urinary catheters). Capturing additional data would allow for refinement of infection detection algorithms, automated denominator determination, and improved risk adjustment. Another important advance will be the extension of federally adopted standard vocabularies20Office of the national coordinator for health information technology. Available from: http://www.hhs.gov/healthit/chiinitiative.html. Accessed June 26, 2007.Google Scholar to information systems vendors. Use of standard vocabularies (eg, SNOMED-CT for laboratory result contents) not only can directly benefit patients—by facilitating transfer of clinical data between health care facilities and providers—but also would facilitate local application of algorithms for infection detection, and transmission of data to public health departments. Advances in the representation of captured date have been realized in a few select institutions, and over time should become more readily available. One such advance is the use of natural language processing systems to translate textual information into coded fields. For example, interpretation of chest radiographs is critical to identification of ventilator-associated pneumonia. Because these reports are stored as textual information, most data warehouses would not be able to use the data; a natural language processing system has been used to identify patients diagnosed with pneumonia.21Haas J.P. Mendonca E.A. Ross B. Friedman C. Larson E. Use of computerized surveillance to detect nosocomial pneumonia in neonatal intensive care unit patients.Am J Infect Control. 2005; 33: 439-443Abstract Full Text Full Text PDF PubMed Scopus (35) Google Scholar, 22Fiszman M. Chapman W.W. Aronsky D. Evans R.S. Haug P.J. Automatic detection of acute bacterial pneumonia from chest X-ray reports.J Am Med Inform Assoc. 2000; 7: 593-604Crossref PubMed Scopus (163) Google Scholar Improved access to previously collected data through a data warehouse provides value through automation of time-consuming manual tasks required of most infection control professionals and standardization of methods for infection detection. Over time, data warehouses will become more valuable as health care providers increase their documentation of care in electronic medical records, systems evolve to better represent the data captured by interpreting textual reports, and vendors allow non-proprietary tools to access data and adopt recommended standard vocabularies.