The use of electronic health record systems (EHRs) has increased in tertiary hospitals in Korea since 2000. The Korean government has an electronic health record initiative to implement EHRs throughout public sector 2010.1 The Korean government strongly recommended the use of standardized medical terminology for this initiative so as to facilitate clinical data exchange and the secondary use of data. The nursing working group associated with this initiative translated, extended, and recommended the International Classification for Nursing Practice (ICNP) as a standardized nursing terminology, and the Korean government adopted this extended Korean ICNP as part of the Korean unified healthcare terminology.2
The authors participated in the development of an ICNP-based electronic nursing record (ENR) system at Seoul National University Bundang Hospital in 2004. The next year, Bundang Hospital confronted many demands from users regarding the construction of a centralized clinical data repository (CDR), which allows end-users to access and retrieve patient data directly. The hospital initiated a project to develop a CDR.3 The CDR of Bundang Hospital was designed to feed data from the management information system, billing system, physicians order entry system, and electronic medical record (EMR) system, including nursing records. The CDR has the capacity to provide great opportunities to allow real-time profiling, ad hoc data search, and data mining according to end-user purposes. The nursing department at Bundang Hospital participated in this project to support various nursing needs. The nursing practice data collected by the ENR system are encoded with a standardized nursing data dictionary (NDD) populated by combining concepts from the extended Korean ICNP.4,5
As EMR and EHR are being spread widely in Korea, interest in the use of data stored in the CDR is increasing. However, it is hard to find use cases showing how a CDR could be used to contribute to the improvement of nursing. Hall and Thornton6 investigated the average number of times nurses documented nursing activities for each patient from the structured labor and delivery data at Intermountain Healthcare. They identified facility-specific nursing patterns associated with a particular patient's profile by comparing the practice and outcomes for similar patient cases at multiple hospitals. They used enterprise data warehouses from the various operational systems, and the data were typically loaded from these systems at regular intervals.
In medicine, many studies including the work of Owen et al7 have investigated the use of data from EMR systems for quality improvement. They have emphasized that replacing reviews of written medical records with performance measurement and quality improvement projects can potentially improve efficiency. Owen et al7 showed the feasibility and validity of using data electronically extracted from the Veterans Health Administration computer database. Pesut also highlighted the detailed, flexible, and rapid retrospective view of both clinical and financial patient data within the hospital information system.8
A CDR is a set of databases that form a centralized data core to which may be linked departmental data collected for clinical purposes. A CDR generally consists of three major components: the data repository, the custom Web-based interface, and a collection of programs for loading data and performing data validation. Access to a CDR has been used mainly for departmental statistics, physician order history, disease statistical analysis, and hospital revenue analysis. The CDRs have been used for retrospective research and clinical trials in studying practice patterns (eg, numbers and types of patients seen, diagnosis/procedure mix, and financial information), quality improvement, teaching students/residents, and reviewing practices of healthcare professionals.7,9
Ad hoc queries about specific groups of patients can be answered in seconds with the use of a CDR. Because the different databases are integrated together, researchers can easily link information from one database to another. Currently, any data item entered in a structured way into the EMR can be found in a search. For example, to search for hypertension patients who smoke, users can select the department and the medical diagnosis, and then filter the patients by age, level of hypertension, smoking, date, and other factors. The complexity of questions that can be asked will continue to grow as more data are added to existing databases and new databases are integrated into the repository.
The CDR at Bundang Hospital is expected to be used for nursing research, practice, education, and management in real time. There are several predefined simple query functions for monitoring nursing indicators, such as pressure ulcers, medication recording errors, and breast-feeding. The system also supports customized user-defined query functions with Boolean conditions on a daily basis. For example, users can search admission assessments, vital signs and measurements, operating-room summaries, and discharge plans. Also, tracking nursing interventions is possible since all nursing records are encoded with ICNP-based NDD mapped to extended Korean ICNP concepts. For complex or sophisticated questions, data mining techniques and statistical software are needed. Several nurse researchers have already used the CDR in research on quality assurance and pressure ulcers.8,10
This article introduces two examples of use of the CDR in nursing. The first example is to study the gaps between the required nursing care time based on patient classification and the actual nursing care time based on current staffing level for nursing administrative purposes. The first purpose of this study was to get information and insight on long-term demands for nurses by nursing unit based on trend analysis. The second is to explore the practice variations of nurses by comparing documented nursing interventions to prevent and treat iatrogenic pressure ulcers. With these examples, we expected to show how to compare nursing interventions and to identify the areas that need more improvement. Although these studies are preliminary trials, they demonstrate the usefulness of a CDR.
The CDR at Bundang Hospital is an almost real-time data warehouse managed by the informatics department of the hospital that contains information about patients seen at the hospital (Figure 1). As shown in Figure 1, A and B show the data sources and the ETL process that the predefined data in the operating system are loaded into a database through a series of customized programs every 24 hours. The ETL process permits special data transformation and encryption of specific fields in the raw data to ensure data integrity and to protect confidentiality when merging raw data from different subsystems. C shows CDR's Web-based interface for users. This interface offers a number of tools to assist them in finding information in the repository, including a custom Query Builder, an interactive SQL interface, and various analysis functions predefined. Also, the CDR supports the data download for further analyses for research and administrative purposes. It contains more than 4 years' worth of data, about 6 704 900 days for inpatients and outpatients, including emergency-room visits. The Bundang hospital has 902 beds for inpatients and about 4000 to 4500 encounters a day for outpatients. Since the CDR is used not only for ad hoc inquiries but also for reviewing patient records, the data are not deidentified. We enforced strict security measures to protect patient confidentiality and privacy for this study. The data for this study which we downloaded from the CDR did not have any identifiable information. Identifiable information was replaced by a coded serial number, which serves to link the data of an unidentifiable individual from different tables. This study was approved by the institutional review board of the hospital. One of the coauthors of this study is a nursing manager who is in charge of the nursing information system in the hospital.
Study 1: Monitoring the Gap Between Patient Classification-Based Nursing Care Time and Actual Nursing Care Time for Nurse Staffing Forecasting
For study 1, we compared two nursing care times: ideally required nursing case time based on patient classification, and actual nursing care time based on nurse staffing level. Patient classification-based nursing care time is the average nursing care time estimated to care for each patient derived from patients' severity index by the patient classification system. Patients are classified into four groups based on severity scores collected at the point of care using the ENR. It is the nursing care time required for optimal nursing care. Bundang Hospital uses the Patient Severity Classification, which was developed by Park and Song (1990)11 and validated and maintained by the Korean Hospital Nurses Association. This classification has been used in hospital accreditation by the Korean Hospital Association since 2005. According to this system, the patients are classified into mildly ill (class I), moderately ill (class II), acutely ill (class III), and critically ill (class IV) based on level of self-care ability. Each class has different nursing care times for direct and indirect nursing care driven from time-motion studies. Nursing care time is estimated by multiplying the nursing care time for each class by the number of patients of that class and summed for all classes.
The actual nursing care time based on staffing level is the average nursing time used for patient care. Staff-mix information comes from the management information system and admission, and discharge and transfer information of patients comes from the billing system. Without the CDR system, these two types of nursing care time can be computed separately, one from the management information system and the other from an EMR system, However, this is possible only for currently hospitalized patients.
Bundang Hospital has experienced understaffing problems because of their limited budget. The nursing department decided to identify the gaps between the patient classification-based nursing care time and the actual staffing-based nursing care time and to use that information for nurse staffing and recruitment plans for coming years.
To compare the two nursing care times, we selected 124 416 patients who were discharged from January 1, 2005, and December 31, 2007, in seven different nursing units: pediatric, women's health, orthopedic and rehabilitation, medical, surgical, geriatric, and psychiatric units. We examined the overall trends of nursing care time gaps between required nursing care time and actual nursing care time by nursing unit over 3 years with analysis-of-variance statistics. Data for this study, including the nurse staff mix, patient's demographics, patient classification, and admission information were extracted from the CDR. The query builder in the CDR was used to select the period, nursing units, and patient classification data and to retrieve the patients' data. These data were sorted by time and preprocessed using a spreadsheet and graphic software. The SAS v.9.1 program (SAS Institute, Cary, NC) was used for further statistical analysis.
Study 2: Exploring Practice Variations in Pressure Ulcer Care
Nurse managers have been monitoring the incidence of hospital-acquired pressure ulcers daily since 2006. The incidence of pressure ulcers has been considered one of the quality indicators of the clinical outcome for nursing services. The purpose of this study is to show how we can retrieve retrospective nursing intervention data from the CDR to see variability in nursing intervention to prevent and treat pressure ulcer patients.
To investigate nursing interventions provided to pressure ulcer patients, we selected 41 891 patients who were discharged from January 1 to December 31, 2007. The nursing intervention data were obtained from narrative nursing notes documented at the point of care. These narrative notes are not structured documents, but rather composed of a series of NDDs. In the NDD master, each statement is classified into nursing assessment, problem, and intervention categories. The NDD is also used to describe nursing outcomes. To identify nursing interventions provided for pressure ulcer patients, we needed to limit nursing problems first and then link the interventions to the patient problems.
We selected five nursing problems related to the risk of pressure ulcers and actual pressure ulcers: risk of impaired tissue integrity, impaired physical mobility, imbalanced nutrition, high risk for infection related to exposures of the ulcer base, and high risk for ineffective therapeutic regimen management. Then seven areas of nursing intervention related to the previously mentioned five nursing problems were searched: sensory/mobility condition, nutritional condition, ulcer wound care, exercise, position change, skin condition, and use of devices. We retrieved nursing interventions using the NDD in the query builder module in the CDR. For example, the skin condition category in the NDD has 40 standardized action statements, such as "dry up the skin," "observe redness on skin," "keep the bed as flat as possible," "observe skin dryness," and "assess skin color."
The data retrieval processes using the query builder of the CDR are described in detail in another work.11 We examined the incidence rates of pressure ulcers and frequencies of intervention documented in nursing notes by nursing unit.
Monitoring the Gap Between Required Nursing Care Time Based on Patient Classification and Actual Nursing Care Time for the Purpose of Staffing Forecasting
Just over half of the 124 416 patients (53.1%) discharged during 2005 to 2007 were male. The average age of patients was 51.4 (SD, 12.3) years, and 8.4% (10 500) were children admitted to the pediatric nursing unit, while 5.8% were admitted to the women's health, 25.3% to the surgical, 11.1% to the orthopedic and rehabilitation, 38.9% to the medical, 0.9% to the psychiatric, 4.5% to the geriatric, and 5.1% to the ICUs.
Figure 2 presents the overall trends in nursing care time for seven nursing units over 3 years. The geriatric, pediatric, and psychiatric units showed high required nursing care times, ranging from 5.1 to 5.7 hours. The required nursing care time for the psychiatric unit has been increasing continuously. The medical and surgical units were positioned in the moderate level of 4.3 to 4.9, while the orthopedic and rehabilitation unit and women's health units showed lower levels of required nursing care time. Figure 3 displays the pattern of time difference between required nursing care time based on patient classification and actual nursing care time based on current nurse staffing level, which reveals noticeable changes by nursing unit. For the pediatric unit, the time gap increased continuously over time from 2.7 to 3.2. The geriatric unit showed a little fluctuation from 3.0 to 3.3, a difference of no significance. However, the gap remained high over the years, while the women's health and psychiatric units showed the gap decreasing significantly. The gaps of the surgical, orthopedic and rehabilitation, and medical units remained about 2 hours, with significant fluctuations ranging from 1.7 to 2.4 over time.
Exploring Practice Variations in Pressure Ulcer Care
The average age of these patients was 52.0 (SD, 12.3) years, and 58.5% of them (22 336) were male. Table 1 presents the incidence rates of pressure ulcers and frequencies of preventive and therapeutic nursing interventions documented in nursing notes by each unit. We did not include the psychiatric or neonatal ICUs in the analysis because these two units had extremely low rates of pressure ulcers. The overall incidence rate was 1.9%, 776 of the 41 891 patients discharged during 2007. The ICU showed the highest incidence rate of 8.6. The geriatric and the orthopedic and rehabilitation units followed next. The order of incidence was similar to that of the average number of interventions provided to a patient. Table 2 shows the frequencies of each nursing intervention. The most frequent intervention overall was "position change," which was followed by "ulcer wound care" and "use of devices." "Patient condition observation" was documented with low frequencies. Figure 4 shows the distribution of nursing interventions by nursing unit.
The nursing record contains information on the care planned and/or given to individual patients by qualified nurses or other caregivers under the direction of a qualified nurse.12 The usefulness of nursing records has been improved by the introduction of ENR systems, which are more than a series of documents in electronic form-they will be the cornerstone of a new way of managing nursing information. The use of ENR allows data collected at the point of care to be used for nursing practice, research, education, and management.
This article has introduced two examples of the clinical uses of nursing data stored in a CDR. We examined the pattern of required nursing care time based on patient classification and gaps between the required nursing care time and the actual nursing care time. We also explored the variations of nursing practice in preventing and caring for patients with hospital-acquired pressure ulcers. Our investigation of nursing data stored in the CDR revealed the picture of each nursing unit as well as the overall picture of the institution. This was possible because the ENR was built using standardized nursing terminology, the ICNP.
With the nursing care time study, we found that there are gaps between required nursing care time computed based on patient classification and actual nursing care time based on current nurse staffing level. Each nursing unit had more than 2 hours of nursing care time differences as we anticipated because of understaffing problems encountered at Bundang Hospital. The geriatric unit and the pediatric unit had the highest gaps, with more than 3 hours. This implies that these two units urgently need more nurses. The pediatric unit displayed a continuously increasing trend, while the geriatric unit was slightly improved in 2007. This nursing care time gap information can be very useful when nurse managers set priorities to allocate personnel resources. Without a CDR, it could take a couple of days to integrate, convert, clean, and analyze data to get this kind of information. However, it took us only a minute to retrieve and aggregate the information we wanted.
Generally, data warehouses store historical data with a different data structure from the operational system to support administrative and regular analytic purposes, while a CDR serves as a centralized data repository on a daily basis to support ad hoc queries. However, both of them contain nonvolatile data, and more detailed analysis is possible with sophisticated statistics tools or data mining technologies. Raw data and their analysis are managed in centralized stores that can be accessed by any authorized person. It is a transactional environment in the sense that the data are retrievable and analyzable with all of the inputs and outputs actively managed by the software rather than procedurally by the users.13 Thus, a large volume of data can be summarized quickly in an ad hoc fashion. Although the CDR at Bundang Hospital is still under development and has room for improvement, the current data in the database can provide valuable information and insight about current practice and outcomes to clinical nurses, researchers, and administrators. Currently, ad hoc queries on any items entered in a structured way into the EMRS are allowed.
With the practice variation study, we had to deal with more data and went through more analysis. We found differences in the incidence of ulcers and nursing interventions from unit to unit. This result might have happened because of differences in patient demographics, disease, and acuity, even though we would hope to see no variation under an ideal nursing care environment. It was found that position change was the most frequent intervention, covering more than 60% to 90% of nursing interventions in all but the pediatrics and women's health units. The interventions of sensory and mobility condition, nutritional condition, exercise, and skin condition were much lower than other intervention categories. For the ICU, most of the patient's ability to move is assumed to be limited, which is the reason for lower nursing interventions relevant to exercise compared with other units. However, it is not clear if the variation among different units is desirable or not in other units. Although the ICU and the geriatric nursing unit showed the highest incidence of ulcers, the medical unit presented the highest frequency and proportion of interventions. This result implies that we need further studies for identifying the reason for practice variations.
The ENR system at Bundang Hospital is built around a standardized nursing terminology, the ICNP, and a library of nursing statements, the NDD, populated by combining concepts from the ICNP and coding. These statements are used to document patient problems, interventions, and outcomes. The ENR system provides ways to capture nursing practice data according to the nursing process, which makes it possible to find patterns of practice and discover nursing knowledge that is embedded in the system.
The complexity of questions that nurses can ask will continue to grow as more data are added to existing databases and nurse users become more informed about what the CDR can offer them. Patient outcomes such as length of stay, number of complications, and medical expenditures can also be investigated, depending on variations in nursing activities within a certain group of patients with similar nursing problems. Any nurse-related factors that affect the variations in nursing activity can also be analyzed. Such analyses make it possible to monitor what is really happening in nursing practice, thereby facilitating improvements to this practice.
However, this study has two limitations. First, the two study designs would not be sufficient in answering all questions that could be driven from each topic. Because the purpose of this article was to demonstrate the use of a CDR in nursing, further research should follow with more sophisticated data selection and analysis methods. Second, the study was performed in a hospital with an enterprise EMR system and a CDR using a standardized nursing terminology. Therefore, the results of this study would have limitations in generality.
Our follow-up research could focus on variations in nursing activities based on patient severity, medical diagnosis, and patient characteristics, as well as the analysis of patient outcomes such as length of stay, number of complications, and cost resulting from variations in nursing activities within a homogeneous group of patients with similar characteristics. We also would like to analyze the variations in activities by nurse-related factors.
The two parts of the study reported in this article exemplified how nursing practice data generated with a standardized nursing terminology-based information system can be used to create useful information for nursing practice. We examined the patterns of gaps between the required nursing care time based on patient classification and the actual nursing care time, and variations of nursing intervention provided for pressure ulcer care.
With the nursing care time gap analysis, we found that the pediatric and geriatric units had relatively high staffing needs, with understaffing for 3 years. This information is very useful to nurse managers in forecasting nurse staffing and recruitment planning. With the practice variation study, we were able to tract nursing interventions for pressure ulcer care. We found that there are nursing practice variations by nursing unit. If we expand our research to the outcome variations of nursing interventions, we will be able to identify the most effective care. This study showed that the current data in the database are very useful for identifying nursing care time gap and practice variations.
The increasing interest of the healthcare industry in electronic patient records increases the importance of finding ways to use data collected at the point of care. To maximize the usefulness of the obtained information, it is essential that all data entered are coded and retrievable using a standardized vocabulary. Furthermore, the shifting of the healthcare industry to an outcome and evidence-based practice makes it necessary for nursing practice to be visualized and improved through the continuous feedback from practice data.
The authors thank Mr Kyeong Seop Kim, senior chief at ezCareTech, for providing invaluable information regarding the configuration and operation of the CDR and for his help in extracting data for the study.
© 2011 Lippincott Williams & Wilkins, Inc.