OJNI

Developing a Computer Interpretable Guideline with Nursing Knowledge: A Pilot Study of a Pressure Ulcer Risk Assessment and Prevention

By

Jeeyae Choi, DNSc, RN

and Hyeoneui Kim, PhD, RN

Citation

Choi, J. & Kim, H. (February 2013). Developing a Computer Interpretable Guideline with Nursing Knowledge:  A Pilot Study of a Pressure Ulcer Risk Assessment and Prevention. Online Journal of Nursing Informatics (OJNI), vol. 17 (1), Available at http://ojni.org/issues/?p=2393

Abstract

Nursing clinical practice guidelines have been developed to assist nurses’ care processes, but these clinical guidelines have not been used effectively in practice. In order to use clinical practice guidelines effectively at the point of care, they should be represented in a computer interpretable format. To facilitate the use of nursing guidelines, this pilot study explored the feasibility of developing a computer interpretable guideline (CIG) using a guideline modeling language with nursing guideline content. Pressure ulcer risk assessments and the related procedures and protocols available at a local hospital were encoded with Guideline Interchange Format (GLIF). A CIG was developed and was validated for its accuracy of the knowledge translated into a CIG using 30 patient scenarios. Findings of this study showed that it is feasible to encode a nursing guideline using a guideline modeling language and that a CIG can be developed for a computerized decision support system.

Keywords: computer interpretable guidelines, decision support systems, guideline modeling language, Guideline Interchange Format (GLIF), clinical practice guidelines, pressure ulcer risk assessment, pressure ulcer prevention, pressure ulcer guidelines

Introduction

Pressure ulcers are a common health problem and one of the most serious safety concerns among patients in health care settings. Pressure ulcers have an impact not only on morbidity and mortality, but also on human discomfort (Kottner & Dassen, 2008). Treatment of pressure ulcers is a financial burden on health care settings because it is complex and requires a longer period of hospitalization (Beeckman et al., 2010).

Prevention is considered the best nursing strategy to avoid painful treatment and unnecessary hospitalization related to pressure ulcers (Magnan & Maklebust, 2009).  Many standardized pressure ulcer risk assessment scales and prevention guidelines have been developed and are currently in use in many hospitals (de Laat et al., 2007; Meesterberends, Halfens, Lohrmann, & Wit, 2010), but they exist in a paper format as a reference for nursing care.  In addition, assessments and interventions are not described with sufficient specificity for different patient cases. These factors may prevent active adoption of pressure ulcer practice guidelines among nurses. Studies show that implementing guidelines as part of a computerized decision support system increases clinicians’ use of guidelines because they enable delivery of patient-specific information at the point of care (Choi, Currie, Wang, & Bakken, 2007; Peleg, Shachak, Wang, & Karnieli, 2009).

Accurately representing clinical practice guidelines in the form of a computer interpretable format is critical for developing and implementing a computerized decision support system.  Clearly defined nursing tasks and unambiguous recommendations from a clinical practice guideline could be translated and incorporated into a computerized decision support system by computer interpretable guidelines.  These computer interpretable guidelines are accurately represented and facilitate acceptance and utilization of a clinical practice guideline in nurses’ daily practice.  In addition, computer interpretable guidelines allow easy maintenance when updating and local adaptation of clinical practice guidelines (Clercq, Kaiser, & Hasman, 2008).

The overall purpose of this study was to explore the feasibility of encoding the nursing practice knowledge presented in a guideline into a computer interpretable format.  As a pilot study, we looked at pressure ulcer risk assessments and the related procedures and protocols available at Spaulding Rehabilitation Hospital (SRH).  This paper describes (1) encoding the nursing knowledge on pressure ulcer prevention into a computer interpretable format using a guideline modeling language, GuideLine Interchange Format (GLIF); (2) developing a computer interpretable guideline based on the GLIF-encoded guideline; and (3) evaluating the accuracy and completeness of the translated knowledge using a small set of patient scenarios.

Background

Pressure Ulcer Assessment and Prevention Guideline

Pressure ulcers are a persistent problem in health care settings because they cause tremendous suffering, frustration, and decreased quality of life to patients (de Laat, Scholte, & Achterberg, 2005). They burden not only the patients but also the care providers due to the high health care costs (Beeckman et al., 2010; Laat et al., 2007; Meesterberends et al., 2010).  The Center for Medicare and Medicaid (CMS) reported 257,412 cases of pressure ulcers as a secondary diagnosis and estimated an average treatment cost of $43,180 in 2007 (Armstrong, Ayello, Capitulo, Fowler, & Krasner, 2008). The most recent pressure ulcer prevalence survey (2006-2009) found that suspected deep tissue injury has increased three times while the overall prevalence has been slightly reduced to 12.3%. As of October 2008, the CMS ceased reimbursement of treatment costs for newly acquired pressure ulcers (VanGilder, MacFarlane, Harrison, Lachenbruch, & Meyer, 2010).

Preventing pressure ulcers is the best strategy to reduce its prevalence rate.  In order to facilitate pressure ulcer prevention practices, many guidelines have been developed and disseminated to guide the preventive care process in many health care settings (Meesterberends et al., 2010). However, these guidelines are not actively used in daily practice because (1) the paper format makes it difficult for nurses to effectively refer to the content during the busy patient care activities, and (2) the guidelines’ care recommendations are not specific enough to be directly applied to patients (Boxwala et al., 2004; Katz, Muehlenbruch, Brown, Fiore, & Baker, 2004).  In addition to guidelines, many hospitals mandate nurses to use the Braden Scale to assess the pressure ulcer risk level.  Sensitivity, specificity, and predictive validity of the Braden Scale have been extensively tested, and the scale is known to have a good balance between sensitivity and specificity (Kottner, J. & Dassen, 2010).

Developing a computerized decision support system is a resource-intensive task that involves the iterative processes of knowledge encoding, modification, and validation (Wang & Peleg, 2007).  However, computerized decision support systems have the potential to increase adherence of clinical guidelines because they are able to provide critical information on patient care at the point of care.  A popular way to operationalize a computerized decision support system is by modeling a clinical practice guideline into a computer interpretable format and incorporating it into a computerized decision support system (Hussain & Abidi, 2008).

The computerized decision support systems that are well integrated into an electronic health record system are especially useful because they facilitate effective decision making and improve the quality of the decision by providing necessary information without disrupting the clinical workflow (Anderson, Willson, Peterson, Murphy, & Kent, 2010; Lyerla, 2008).  Thus computerized practice guidelines that are successfully integrated into an electronic health record system will better assist nurses to make decisions regarding preventive care of pressure ulcers.  It is therefore important that the accurate representation of clinical practice guidelines in computer interpretable format should precede the development of a computerized decision support system.

Guideline Interchange Format (GLIF)

GLIF is an object-oriented model that was developed to represent sharable computer interpretable guidelines.  Its various classes and their attributes are used to describe and to illustrate complex clinical knowledge.  The current version of GLIF (i.e., GLIF3) encodes guideline knowledge at three levels:  (1) a conceptual flowchart, (2) a computable specification for validating logical consistency and completeness, and (3) an implementable specification that can be incorporated into an information system of an institution.  GLIF3 uses Health Level 7 standards to allow integration of computer interpretable guidelines into a clinical information system.  Its specification structure is based on the Resources Description Framework, which allows extending a computer interpretable guidelines’ specification (Boxwala et al., 2004).  GLIF3 has been used to encode different types of guidelines (Boxwala et al., 2004; Peleg et al., 2000; Peleg et al., 2004; Wang et al., 2004).

In this study, we explored the possibility of developing a computer interpretable guideline based on the Braden Scale and the paper-based nursing care protocol at Spaulding Rehabilitation Hospital with Guideline Interchange Format.

Method and Procedure

Spaulding Rehabilitation Hospital (SRH) and Pressure Ulcer Prevention Guideline

SRH is an academic acute rehabilitation hospital with 190-certified beds and a 160-average daily census’ (statistics of 2011).  SRH offers six different rehabilitation programs: five for adult patients and one for pediatric patients.  Adult programs cover various types of injuries including cardiac, musculoskeletal, stroke, brain injury, and spinal cord injuries. The pressure ulcer prevalence rate at SRH was 2.27% higher than rates of similar hospitals in Massachusetts, 0.86% in 2011(Patient CareLink, 2012).

The Pressure Ulcer Risk Assessment and Prevention procedure and policy at Spaulding Rehabilitation Hospital, although in paper format, has been developed and regularly updated since 2004.  In this document, the Braden Scale is specified as an assessment tool.  It is mandated that nurses at Spaulding Rehabilitation Hospital assess the pressure ulcer risk of all hospitalized patients using the Braden Scale and document the Braden Score at least once a day.

Encoding the Local Pressure Ulcer Prevention Guideline with GLIF

Concepts that represented the nursing tasks related to pressure ulcer prevention were extracted from the procedure and policy document.  These concepts were then classified as action, decision, and patient state (Table 1).  For example, “Get last value of Braden Scale Score” was identified as an action and “Patient is in adult unit” as the patient status that acted as a triggering event associated with this action.  Each concept and its associated triggering event were tabulated.  Table 2 shows three examples of the mapping method between concepts and the GLIF steps.

Table 1 Glossy:  Definitions of steps and triggering events in a guideline and the GLIF notation (Peleg et al., 2004)

Table 1 Glossy:  Definitions of steps and triggering events in a guideline and the GLIF notation (Peleg et al., 2004)

Table 2: Mapping method between concepts and the GLIF steps

Table 2: Mapping method between concepts and the GLIF steps

The process of assessing pressure ulcer risks and recommending preventive interventions in the Pressure Ulcer Risk Assessment and Prevention Procedure and Policy of Spaulding Rehabilitation Hospital was encoded as follows.  When a patient is admitted in the adult unit, this triggers a guideline to assess the Braden Scale.  Then, a guideline recommends initiating preventive intervention for patients at risk based on results of the Braden Scale.  This process was designed as a main guideline.  Specific courses of actions related to assessing each Braden parameter were designed as six sub-guidelines.

Developing a Computer Interpretable Guideline

The concepts classified with GLIF3 steps were converted into a flowchart format using the GLIF authoring tool implemented as a plug-in application in Protégé 3.4.  Using the same tool, the main guideline and the six sub-guidelines were converted and visualized in a flowchart format.  A nurse specialized in wound care reviewed the flowchart and verified the logical consistency.  During this review, any unnecessary or illogical steps in the flowchart were identified and corrected.  For example, the procedure and policy document recommends scoring the Braden Scale more frequently than once a day when a patient’s wound status has changed.  However, this additional assessment was deemed unnecessary as the daily assessment of pressure ulcer risks can capture the risk level changes of the patients at Spaulding Rehabilitation Hospital as effectively.  This is related to patients at the rehabilitation stage being less likely to show abrupt aggravations in health conditions.  Therefore, the decision step suggesting “complete the Braden Scale as a PRN” was removed from the flowchart. The guideline flowchart was finalized after three iterations of review and revision.

Evaluation of a developed Computer Interpretable Guideline

In order to validate the accuracy of the knowledge translated into the computer interpretable format, the Automated Braden Scoring Tool was developed.  This prototype tool for automatically assessing pressure ulcer risks was programmed based on a GLIF-encoded guideline.  We used an open source tool, Rails (http://rubyonrails.org), to build the system for the testing.  Thirty patient scenarios were created with the help of the wound care nurse.  Every data element of each scenario was entered into the system.  Risk levels of 30 scenarios, including the value of each parameters of the Braden Scale from an expert and the system, were recorded and compared.

Results

Identified Guideline Interchange Format (GLIF) steps

A patient-state-step was identified as an entry point for both the main guideline and the six action steps that invoke six sub-guidelines. Only one branch step was needed to coordinate multiple and simultaneous paths between a main and sub-guideline. Two synchronization steps and three case steps were identified to pass clinical data appropriately and in a timely manner to the next step.

Figure 1 shows an example of identified action task encoded with GLIF with associated triggering events and specific tasks.  The subsequent task of each step was also specified at this level.

Figure 1 An encoded Action Step, “Get last value of Braden Scale” in protégé authoring tool

Figure 1 An encoded Action Step, “Get last value of Braden Scale” in Protégé authoring tool

A Computer Interpretable Guideline in flowchart format

Figure 2 shows a partial view of a pressure ulcer risk assessment and prevention guideline presented in GLIF flowchart format.  In the flowchart, “Patient in adult unit” describes the patient state and serves as a starting point of a guideline.  It triggers the next step, “Get last value of Braden Scale,” and this action step retrieves the value of the Braden Scale. The decision step, “What is the last value of Braden Scale?” decides which next step should be triggered.  When the retrieved value is null, then the action step, “Complete Braden Scale” will be triggered.  But if the value is not null, the decision step, “Is time to score Braden Scale again?” will be triggered.

Figure 2 Partial view of a main guideline

Figure 2 Partial view of a main guideline

Figure 3 shows one of the nested sub-guidelines for assessing and setting sensory perception value.  In the sub-guideline flowchart, “Start sensory perception parameter calculation” represents a starting point of a sub-guideline rather than a patient state.  The guideline triggers the “Assess level of consciousness” action step, and will get the value of Glasgow Coma Scale-Best Eye Opening which will trigger the decision step that asks, “What is score of GCS-eye opening?” If the value is 3, which is “Opens eyes in response to voice”, then it triggers the “What is score of GCS-verbal response?” decision step.

Figure 3 Partial view of sub-guideline that assesses sensory perception

Figure 3 Partial view of sub-guideline that assesses sensory perception

Evaluation of a Developed Computer Interpretable Guideline

Table 3 shows the percentage of exact matches and near matches between an expert and a computer interpretable guideline.  Exact match rate was 82%, and the near match rate was 100%.  Correlation coefficient of risk levels (total score of Braden Scale) assessed by an expert and a computer interpretable guideline was 0.96, meaning that the computer interpretable guideline assessed risk level was close to that of an expert.

Table 3: Percentage of exact and near matches between an expert and a CIG


*Near match: When (expert’s score) – (CIG’s score) = ±1, it was considered as a matched case.

Table 3: Percentage of exact and near matches between an expert and a CIG

Discussion

We demonstrated the encoding of a clinical practice guideline using GLIF and developed a computer interpretable guideline.  Through this study we found that the high level workflow information on pressure ulcer risk assessments, as described in the procedure and protocol document, was suitable for GLIF encoding because the workflow was described as sequential tasks. Various encoding steps defined in GLIF also supported representing the tasks involved in assessing pressure ulcer risks.

While converting a guideline to a GLIF flowchart format, we realized that the “Complete Braden Scale” action step was overly simplified when it was converted into a computer interpretable format. To be executed in a software system, this action tasks need to be defined at a more specific level. Therefore, each parameter of Braden Scale was encoded with GLIF and formed a sub-guideline, which contained enough details for a computer execution.

Although the evaluation was done using a small number of patient scenarios, results showed the high level of consistency between the assessment produced by the computer interpretable guideline and the expert nurse.  This result implies that the knowledge presented in the procedure and protocol and the Braden Scale was accurately translated into a computer interpretable guideline.

The use of one expert assessing the risk for pressure ulcers of patient scenarios might weaken evaluation credibility of a developed Computer Interpretable Guideline. Although this was not the best method, it was necessary and practical for evaluation. This expert is the only wound care nurse at the study site and has been assessing patients’ pressure ulcer risk for 35 years in rehabilitation hospital settings. In addition, she has educated nurses in pressure ulcer risk assessment and prevention at the study site.

Our original goal was to encode the intervention recommendations as well.  However, encoding the interventions described in the procedure and protocol document added little value as the interventions were described at the general pressure ulcer prevention level rather than the specific risk level of a patient.  In other words, the interventions described in the document applied equally to the patients who have a Braden Score of less than 19.  This might be another reason that this procedure and protocol document has not been actively utilized by nurses in practice.

Conclusion

Many nursing clinical practice guidelines have been developed and implemented in hospitals to improve patient care and outcomes.  However, these clinical practice guidelines are not used effectively in practice because they are disseminated in the paper format and their recommendations are not specific enough to apply in practice.  As a way to facilitate use of nursing guidelines, this study explored the feasibility of developing a computer interpretable guideline with nursing guideline content.  Findings of this study showed that it is feasible to encode a nursing guideline with a guideline modeling language and that an implementable computer interpretable guideline can be developed for a clinical decision support system.  Well-constructed computer interpretable guidelines can facilitate building and implementation of a quality clinical decision support system.  This will assist nurses’ decision making at the point of care, which will lead to improved quality of patient care and outcomes.

Limitation

There are several limitations of this study.  Only one wound care nurse evaluated 30 scenarios.  If more than one wound care nurse had evaluated the scenarios, the outcome might have been different.  Although clinically possible scenarios were created, actual patient data should be used to test the accuracy of a computer interpretable guideline.  If a different guideline modeling language other than GLIF had been used to encode a paper-based guideline, the outcome might have been different as well.

Lessons Learned

From this study, we learned that nursing paper-based guidelines can be translated into a computer interpretable guideline to facilitate use of clinical practice guidelines at the point of care for nurses.  Also, we established that a quality computer interpretable guideline can be developed that includes workflow of the local health care setting while encoding a local procedure and protocol.

Acknowledgments

The authors thank Dr. Jinho Lee for his valuable input on building and programming a software system for this study.

References

Anderson, J. A., Willson, P., Peterson, N. J., Murphy, C., & Kent, T. A. (2010). Prototype to practice developing and testing a clinical decision support system for secondary stroke prevention in a veterans healthcare facility. Computers, Informatics, Nursing, 28, 353-363.

Armstrong, D. G., Ayello, E. A., Capitulo, K. L., Fowler, D. L., & Krasner, J. M. (2008). New opportunities to improve pressure ulcer preventionand treatment: Implications of the CMS inpatient hospital care Present on Admission (POA) indicators/hospital-acquired conditions (HAC) policy.  Journal of Wound, Ostomy, and Continence Nursing, 35, 485-492.

Beeckman, D., Vanderwee, K., Demarre, L., Paquay, L., Van Hecke, A., & Defloor, T. (2010). Pressure ulcer prevention: Development and psychometric validation of a knowledge assessment instrument. International Journal of Nursing Studies, 47, 399-410.

Boxwala, A. A., Peleg, M., Tu, S., Ogunyemi, O., Zeng, Q. T., Wang, D., . . . Shortliffe, E. H.  (2004). GLIF3: A representation format for sharable computer-interpretable clinical practice guidelines. Journal of Biomedical Informatics, 37, 147-161.

Choi, J., Currie, L. M., Wang, D., & Bakken, S. (2007). Encoding a clinical practice guideline using guideline interchange format: A case study of a depression screening and management guideline. International Journal of Medical Informatics, 76S, S302-S307.

Clercq, P. D., Kaiser, K., & Hasman, A. (2008). Computer-interpretable Guideline formalisms. Studies in Health Technology and Informatics, 139, 22-43.

de Laat, E. H., Pickkers, P., Schoonhoven, L., Verbeek, A. L., Feuth, T., & van Achterberg, T. (2007). Guideline implementation results in a decrease of pressure ulcer incidence in critically ill patients. Critical Care Medicine, 35, 815-820.

de Laat, E. H., Scholte op Reimer, W. J., & Achterberg, T. (2005). Pressure ulcers: diagnostics and interventions aimed at wound-related complaints: A review of the literature. Journal of Clinical Nursing, 14, 464-472.

Hussain, S., & Abidi, S. (2008). An ontology-based framework for authoring and executing clinical practice guidelines for clinical decision support systems. The Journal on Information Technology in Healthcare, 6, 8-22.

Katz, D. A., Muehlenbruch, D. R., Brown, R. L., Fiore, M. C., Baker, T. B., & AHRQ Smoking Cessation Guideline Study Group. (2004). Effectiveness of implementing the Agency for Healthcare Research and Quality smoking cessation clinical practice guideline: A  randomized controlled trial. Journal of the National Cancer Institute, 96, 594-603.

Kottner, J., & Dassen, T. (2008). An interrater reliability study of the Braden Scale in two nursing homes. International Journal of Nursing Studies, 45, 1501-1511.

Kottner, J., & Dassen, T. (2010). Pressure ulcer risk assessment in critical care: Interrater reliability and validity studies of the Braden and Waterlow scales and subjective ratings in two intensive care units. International Journal of Nursing Studies, 47, 671-677.

Lyerla, F. (2008). Design and implementation of a nursing clinical decision support system to promote guideline adherence. Computers, Informatics, Nursing, 26, 227-233.

Magnan, M. A., & Maklebust, J. (2009). The nursing process and pressure ulcer prevention: Making the connection. Advances in Skin and Wound Care, 22, 83-92.

Meesterberends, E., Halfens, R., Lohrmann, C., & Wit, R. (2010). Pressure ulcer guideline development and dissemination in Europe. Journal of Clinical Nursing, 19, 1495-1503.

Patient CareLink. (2012). Hospital Data. Retrieved from http://www.patientcarelink.org/nqf/NQF_Template.cfm?ID=20078&Name=Spaulding

Peleg, M., Boxwala, A., Ogunyemi, O., Zeng, Q., Tu, S., & Lacson, R. (2000). GLIF3: the evolution of a guideline representation format. Paper presented at the AMIA Annual Symposium, Los Angeles, CA.

Peleg, M., Boxwala, A., Tu, S., Wang, D., Ogunyemi, O., & Zeng, Q. (2004). Guideline Interchange Format 3.5 technical specification: InterMed Collaboratory. Retrieved from http://mis.hevra.haifa.ac.il/~morpeleg/Intermed/

Peleg, M., Shachak, A., Wang, D., & Karnieli, E. (2009). Using multi-perspective methodologies to study users’ interactions with the prototype front end of a guideline-based decision support system for diabetic foot care. International Journal of Medical Informatics, 78, 482-493.

VanGilder, C., MacFarlane, G. D., Harrison, P., Lachenbruch, C., & Meyer, S. (2010). The demographics of suspected deep tissue injury in the United States: An analysis of the International Pressure Ulcer Prevalence Survey 2006-2009. Advances in Skin and Wound Care, 23, 254-261.

Wang, D., & Peleg, M. (2007). Using GLIF and GLEE to facilitate knowledge management in development of clinical decision support systems. Medinfo, 12, 2268-2270.

Wang, D., Peleg, M., Tu, S. W., Boxwala, A. A., Ogunyemi, O., Zeng, Q., . . . Shortliffe, E. H.  (2004). Design and implementation of the GLIF3 guideline execution engine. Journal of Biomedical Informatics, 37, 305-318.

Author Bios

Jeeyae Choi, DNSc, RN

Jeeyae Choi received her Doctorate (DNSc) in Nursing Informatics from Columbia University. She is an Assistant Professor at the College of Nursing at University of Wisconsin-Milwaukee. Dr. Choi has two educational backgrounds, one in Nursing and the other in Computer Systems Engineering. Her research interest is in developing and evaluating decision support systems utilizing clinical guideline modeling language, Guideline Interchange Format, in an attempt to translate a text-based clinical practice guideline (CPG) into a computer-interpretable guideline. Dr. Choi, as a primary investigator, completed the project that would validate and enhance a decision support system on pressure ulcer risk assessment for generalizations in 2011.

Hyeoneui Kim, PhD, RN

Hyeoneui Kim earned PhD in Health Informatics from University of Minnesota, Twin Cities.
Her research areas include standardized concept representation (standardized terminologies, ontologies, and information modeling) and consumer health informatics. She also has experienced in developing various clinical applications such as Electronic Medical Records, clinical research data warehouse and a nursing decision support tool on pressure ulcer risk assessment. She is currently an assistant professor of Division of Biomedical Informatics at University of California, San Diego.

 

Back to Issue Index

Back to Issue Index

Be Sociable, Share!

Comments are closed.