Clinical Decision Support Tool to Promote Adoption of New Neonatal Hyperbilirubinemia Guidelines

Abstract

Objective This study aimed to increase the adoption of revised newborn hyperbilirubinemia guidelines by building a clinical decision support (CDS) tool into templated notes.

Methods We created a rule-based CDS tool that correctly populates the phototherapy threshold from more than 2,700 possible values directly into the note and guides clinicians to an appropriate follow-up plan consistent with the new recommendations. We manually reviewed notes before and after CDS tool implementation to evaluate new guidelines adherence, and surveys were used to assess clinicians' perceptions.

Results Postintervention documentation showed a decrease in old risk stratification methods (48 to 0.4%, p < 0.01) and an increase in new phototherapy threshold usage (39 to 95%, p < 0.01) and inclusion of follow-up guidance (28 to 79%, p < 0.01). Survey responses on workflow efficiency and satisfaction did not significantly change after CDS tool implementation.

Conclusion Our study details an innovative CDS tool that contributed to increased adoption of newly revised guidelines after the addition of this tool to templated notes.

Keywords

hyperbilirubinemia - clinical decision support systems - pediatrics

Background and Significance

Hyperbilirubinemia affects almost all newborn infants due to normal physiologic differences in neonatal bilirubin production, metabolism, and excretion. In some cases, the bilirubin can reach pathologic levels and, if not managed appropriately, can lead to acute or chronic bilirubin encephalopathy causing profound neurologic impact.[1] This is considered a preventable adverse event, also known as a “never event.” As such, newborn providers routinely evaluate infants for hyperbilirubinemia and treat them with phototherapy if the level is high. In September 2022, the American Academy of Pediatrics revised their clinical practice guidelines for the management of newborn hyperbilirubinemia.[2]

The new guidelines recommend bilirubin screening for all newborns prior to nursery discharge. They also define more than 2,700 updated phototherapy thresholds based on gestational age (GA) in weeks, age in hours at the time of lab draw, and the presence of neurotoxicity risk factors. These neurotoxicity risk factors have been revised since the last guideline release in 2004 and include hypoalbuminemia, evidence of isoimmune hemolytic disease, sepsis, and any other clinical instability.[2] [3] For each newborn, clinicians refer to 1 of 10 nomograms, stratified by birth GA, and the presence (or absence) of neurotoxicity risk factors, to determine whether an infant requires phototherapy. For all newborns, providers need to select the appropriate graph and plot out the newborn's bilirubin against these nomograms to determine whether they require phototherapy treatment to lower their bilirubin to a safe level.

Additionally, the new guidelines delineate follow-up recommendations on if and when to check the next bilirubin based on how close the infant's bilirubin is to their nomogram-specific phototherapy threshold, a numerical value referred to as “delta.”[2] For example, a 30-hour-old infant born at 39 weeks gestation with no neurotoxicity risk factors has a bilirubin phototherapy threshold of 13.8 mg/dL. If the infant's bilirubin at 30 hours is 4 mg/dL, the delta is equal to the difference between the threshold level and 4, in this case 9.8. A recheck can occur in the nursery or after discharge, depending on the delta and other factors, such as discharge timing. This is a shift from prior guidelines that required referencing two separate nomograms: a less-specific risk stratification curve (known as the “Bhutani curve”) and another curve to assess the current need for phototherapy.[4] The original guidelines constituted standard practice for nearly 20 years, making it difficult for clinicians to abruptly adopt a new workflow in newborn nurseries.

Adoption of new guidelines and changing clinical practice is complex and often requires several interventions and multilayered approaches.[5] [6] There has been increasing literature to support the use of clinical decision support (CDS) tools as one of these layers to improve patient care and bridge the gap between evidence-based medicine and clinical practice. For example, CDS tools have been shown to reduce medication errors as well as increase compliance with vaccination and other guideline recommended preventative measures.[7] [8] [9] [10] [11] [12] [13] Given the complexity of hyperbilirubinemia evaluations, several popular tools currently exist for hyperbilirubinemia assessment including a web-based decision support tool, a report generated within the electronic health record (EHR), and other third-party applications. The web-based decision support tool experienced an increase in site visits over time since its creation, suggesting increased adoption of clinical guidelines.[14] [15] An automated report embedded in one major EHR has also been linked to improved physician satisfaction and reduced documentation errors.[16] An add-on third-party application helped reduce time spent on neonatal bilirubin management and demonstrated good usability.[17] All these studies supported the use of CDS tools in the evaluation of neonatal hyperbilirubinemia. Yet these existing tools all share certain limitations. Firstly, they require navigation to separate sections of the EHR or to an external web browser, which may affect efficiency and accuracy by necessitating toggling between windows and adding extra clicks. They also require manual entry of the patient's bilirubin level, age at the time of lab draw, phototherapy threshold, and delta into the clinician's note and oftentimes in the tool itself. Each of these steps presents an opportunity for error. Accurate documentation of these values is crucial for effective communication with providers, especially those at outside institutions, who make clinical decisions on when to perform and how to interpret follow-up bilirubin levels.

Objectives

The primary objective of the study was to determine if a novel CDS tool integrated into a templated nursery discharge summary would increase the adoption of the new guidelines by inpatient clinicians. Our secondary objective was to increase clinician efficiency and satisfaction in interpreting newborn bilirubin levels. To our knowledge, no hyperbilirubinemia CDS tool exists that integrates into a patient's note, nor are there any current publications evaluating such a tool.

Methods

We conducted a retrospective study evaluating newborns born at ≥35 weeks gestation who were admitted to and discharged from either of the two newborn nurseries within our health system. We excluded any infants who were <35 weeks as the new guidelines do not apply to this population. We also excluded infants admitted or transferred to the neonatal intensive care unit or pediatric ward during their admission because these units have service-specific discharge summary templates that were not part of the intervention. Newborns without a bilirubin drawn were included and classified as nonadherent to the new guidelines.

We created a unique CDS tool that translates several tables containing over 2,700 phototherapy threshold values specific to the hours of life a bilirubin lab was drawn, the GA of the newborn, and the presence (or absence) of neurotoxicity risk factors. We translated these values into text by creating 1,018 configurable logic rules that could then be built into our EHR, Epic Systems, using native build tools (e.g., “SmartLinks,” “SmartTexts,” and “SmartLists”). Providers were prompted via a drop-down list to select the presence or absence of neurotoxicity risk factors. Based on their selection, the CDS tool used the patient-level data from various parts of the chart to autopopulate the bilirubin level, age at the time of lab draw, and phototherapy threshold within the note including guidance for whether the level is above or below the threshold. A drop-down list is then allowed to select a follow-up plan based on the delta and age at discharge. [Fig. 1] shows the provider's view of the CDS tool, where blue- and gray-shaded areas represent data automatically pulled into the note and pink-shaded areas require provider selection from a drop-down list. This tool was integrated into existing newborn discharge summary note templates, and these updated templates were released to all nursery providers in August 2023 along with an educational email with stepwise instructions on the tool.

Fig. 1 Example of a CDS tool within a templated note. Blue- and gray-shaded areas represent automatically populated portions in the note. Pink-shaded areas are SmartLists requiring provider selection. ©2024 Epic Systems Corporation. CDS, clinical decision support; HOL, hours of life; TcB, transcutaneous bilirubin; TsB, total serum bilirubin.

Newborn discharge summaries were reviewed for new guidelines adherence as measured by documentation of old risk stratification based on the Bhutani curve[4] versus new phototherapy thresholds and follow-up plans based on updated guidelines. New guidelines were released in September 2022. Subsequently, data on newborns discharged from the nursery were collected over a 1-month period in December 2022, referred to as “pre-CDS tool cohort.” We implemented the CDS tool in August 2023, and data for the “post-CDS tool cohort” were collected over a 1-month period in November 2023. Study outcomes were measured for each cohort 3 months after the revised guidelines release and 3 months after CDS tool implementation to allow time for uptake of the guidelines and tool, respectively. We compared outcomes for pre- and post-CDS tool cohorts using Fisher's exact test.

We evaluated provider efficiency and satisfaction with a voluntary, anonymized survey emailed to all providers caring for newborns in the nurseries, which included neonatology fellows, neonatologists, neonatal nurse practitioners, hospitalists, and hospitalist fellows. The survey asked providers to rate their efficiency and satisfaction using a 5-point Likert scale. The proportion of positive survey results was compared between pre- and post-CDS tool cohorts using chi-square analysis.

The study was determined to meet the criteria for exemption by the Institutional Review Board.

Results

A total of 464 patients met the inclusion criteria, with 223 patients in the pre-CDS tool cohort and 241 patients in the post-CDS tool cohort. In the pre-CDS tool cohort, 48% of notes continued to document old risk stratification in discharge documentation, 39% reported the correct revised phototherapy threshold, and 28% documented the correct follow-up recommendations based on the new guidelines. As seen in [Table 1], after CDS tool integration into discharge summaries, old risk stratification significantly reduced to 0.4% (p < 0.01), correct documentation of revised phototherapy thresholds significantly increased to 95% (p < 0.01), and correct documentation of new follow-up recommendations increased to 79% (p < 0.01). The incorrect documentation of revised phototherapy thresholds in the post-CDS tool cohort in 5% (n = 13) of patients was due to either manual removal of the tool from notes (n = 8), not collecting bilirubin levels prior to discharge (n = 1), or incorrect phototherapy threshold selection based on neurotoxicity risk factors (n = 4). An incorrect delta was calculated in 12% (n = 28) of charts and 9% (n = 23) had no delta documented.

Table 1 Discharge summary documentation before and after clinical decision support tool implementation

Guideline adherence	Before CDS tool (n = 223)	After CDS tool (n = 241)	p-Value
Risk stratification (old guidelines)	106 (48%)	1 (0.4%)	<0.01
Phototherapy threshold (new guidelines)	87 (39%)	228 (95%)	<0.01
Follow-up plan (new guidelines)	62 (28%)	190 (79%)	<0.01

Abbreviation: CDS, clinical decision support.

Notably, during the study period, all newborns who were above the phototherapy threshold received appropriate phototherapy treatment consistent with the new guidelines (n = 2, both in the post-CDS tool cohort). Overtreatment with subthreshold phototherapy occurred for four patients in the pre-CDS tool cohort and one patient in the post-CDS tool cohort. Given the relative infrequency of phototherapy at baseline, we could not draw any conclusions from these results.

A total of 65 providers caring for infants in the newborn nurseries were surveyed. The response rate was 77% (n = 50) in the pre-CDS tool cohort and 51% (n = 33) in the post-CDS tool cohort. In the pre-CDS tool cohort, 49% of survey respondents found their workflow for finding and interpreting bilirubin values to be “efficient” or “very efficient,” which improved to 61% (p = 0.30) in the post-CDS tool cohort. In the pre-CDS tool cohort, 49% of survey respondents were “somewhat satisfied” or “extremely satisfied” with their workflow, which improved to 64% (p = 0.19) in the post-CDS tool cohort.

Discussion

As shown in this study, automated CDS within templated notes can influence provider adherence to best practices and clinical decision-making. Additionally, our tool improved workflow by minimizing manual calculations and note entry and eliminating navigation outside of a note to complete this necessary newborn assessment. Although the build was complex, it resulted in clinical documentation that was consistent with new guidelines and seamlessly integrated within existing workflows. Notably, our providers' survey responses did not show a statistically significant change in efficiency or satisfaction after CDS tool implementation, though this may be confounded by the lower response rate in the post-CDS tool cohort.

There were no significant challenges with provider adoption of the CDS tool as it was integrated into existing standardized discharge summaries at our institution. These discharge summaries automatically populate when searching for nursery discharge summaries, so system-wide templated notes are preferentially used over personalized note templates. One potential trend seen in our study involved the correct documentation of follow-up recommendations. This practice significantly improved with our CDS tool; however, the degree of improvement fell short of that seen in the other metrics. This may have been due to errors in manual calculation required to determine the correct delta.

Prior studies detailing efforts to institute evidence-based best practices indicate that education alone is often inadequate and that multifaceted interventions may be required to realize sustained change.[5] [6] CDS tools have helped align clinical and best practices with varied success.[18] [19] Our tool has several features previously shown to improve adoption, such as the gathering of patient-specific data while demanding minimal effort from physicians.[20] [21] Additionally, it provides recommendations that are actionable by the provider if necessary, including the need for phototherapy and when the newborn should be seen for a follow-up visit. It is also easily integrated into existing provider workflows by being automatically included in note templates and is easy to use, thus requiring minimal education and training.[22]

The main limitation of this study is reproducibility. Creating the CDS tool was labor-intensive, necessitating manual input of 10 tables of bilirubin levels, each with 119 to 359 discrete data entries.[2] The stratified structure of the new hyperbilirubinemia guidelines was also amenable to a rule-based logic approach, which may not be the case for other clinical practice guidelines. Also, because there was a lapse between the new guidelines release and the inception of our CDS tool, adherence to new guidelines may be partially attributable to a secular trend; however, the significant improvement shortly after the implementation of the CDS tool suggests that it likely played an important role. Lastly, although guidelines change infrequently (in this case, 18 years), any future updates will require potentially significant manual editing of the logic rules.

The Pediatrics Department at UCLA Health conducts ongoing tracking of infants readmitted for phototherapy treatment. We continue to monitor these charts to ensure compliance with new phototherapy thresholds. We plan to analyze additional metrics, such as an overall decline in phototherapy use and readmissions, as part of a follow-up study.

Conclusion

CDS within EHRs is widely used in attempts to change clinical practices to align with evidence-based best practices and to improve provider efficiency. Our study contributes to this growing body of literature by demonstrating the utility of a novel CDS tool implemented soon after the release of new guidelines. We significantly improved documentation to be more consistent with new guidelines. The next steps include evaluating the clinical impact, specifically the appropriate provision of phototherapy treatment when indicated, as well as the critical step in effective CDS tool implementation—consistent evaluation and monitoring to ensure accuracy and user compliance.[22] [23]

Clinical Relevance Statement

There exists a persistent gap between best practice guidelines and clinical practice. CDS systems have had mixed success in changing provider practices. Our study illustrates a tool that allowed our institution to change clinical documentation to increase the adoption of new guidelines.

Multiple-Choice Questions

1. Which feature should be included for the successful adoption of a CDS tool?

   • Assessment feature

   • References for recommendations

   • Automated provision

   • Delayed recommendations

   Correct Answer: The correct answer is option c. CDS tools are used throughout different health systems to help promote best practices with few studies looking into successful implementation. One systematic review noted automated provision as the feature most correlated with the successful use of CDS tools.[17] An understanding of provider workflow and providing a tool that seamlessly integrates is associated with increased tool use. Although an assessment feature is often included in CDS tools, actionable recommendations are preferred. References or data representing recommendations have not been shown to change CDS tool effectiveness. Ideally, a CDS tool will quickly provide information at the setting and time it is needed.

2. Which of the following statements regarding the CDS tool in this study is incorrect?

   • There was a decrease in the use of old phototherapy guidelines for patient risk stratification.

   • There was an increase in correct documentation of phototherapy thresholds.

   • Physician efficiency was qualitatively increased.

   • Documentation of follow-up recommendations increased.

   Correct Answer: The correct answer is option c. The CDS tool was associated with a decrease in the use of old phototherapy guidelines, an increase in the accurate documentation of phototherapy thresholds, and an increase in the documentation of follow-up recommendations. Although a greater percentage of physicians reported that bilirubin workflows were “efficient” or “very efficient” after the implementation of the tool, these results were not statistically significant.

Conflict of Interest

None declared.

Protection of Human and Animal Subjects

The study did not involve human and/or animal subjects. The study was determined to meet the criteria for exemption by the Institutional Review Board.

Authors' Contributions

P.J.L. and D.K. contributed equally and are considered co-principal investigators of this work.

• References

• 1 Maisels MJ, Newman TB. Kernicterus in otherwise healthy, breast-fed term newborns. Pediatrics 1995; 96 (4 pt 1): 730-733
• 2 Kemper AR, Newman TB, Slaughter JL. et al. Clinical Practice Guideline Revision: Management of hyperbilirubinemia in the newborn infant 35 or more weeks of gestation. Pediatrics 2022; 150 (03) e2022058859
• 3 American Academy of Pediatrics Subcommittee on Hyperbilirubinemia. Management of hyperbilirubinemia in the newborn infant 35 or more weeks of gestation. Pediatrics 2004; 114 (01) 297-316
• 4 Bhutani VK, Johnson L, Sivieri EM. Predictive ability of a predischarge hour-specific serum bilirubin for subsequent significant hyperbilirubinemia in healthy term and near-term newborns. Pediatrics 1999; 103 (01) 6-14
• 5 Bero LA, Grilli R, Grimshaw JM, Harvey E, Oxman AD, Thomson MA. The Cochrane Effective Practice and Organization of Care Review Group. Closing the gap between research and practice: an overview of systematic reviews of interventions to promote the implementation of research findings. BMJ 1998; 317 (7156) 465-468
• 6 Lannon C, Stark AR. Closing the gap between guidelines and practice: ensuring safe and healthy beginnings. Pediatrics 2004; 114 (02) 494-496
• 7 Kwan JL, Lo L, Ferguson J. et al. Computerised clinical decision support systems and absolute improvements in care: meta-analysis of controlled clinical trials. BMJ 2020; 370: m3216
• 8 Lukac PJ, Bell D, Sreedharan P, Gornbein JA, Lerner C. The application of dental fluoride varnish in children: a low cost, high-value implementation aided by passive clinical decision support. Appl Clin Inform 2023; 14 (02) 245-253
• 9 Dexter PR, Perkins SM, Maharry KS, Jones K, McDonald CJ. Inpatient computer-based standing orders vs physician reminders to increase influenza and pneumococcal vaccination rates: a randomized trial. JAMA 2004; 292 (19) 2366-2371
• 10 York JB, Cardoso MZ, Azuma DS, Beam KS, Binney Jr GG, Weingart SN. Computerized physician order entry in the neonatal intensive care unit: a narrative review. Appl Clin Inform 2019; 10 (03) 487-494
• 11 Utidjian LH, Hogan A, Michel J. et al. Clinical decision support and palivizumab: a means to protect from respiratory syncytial virus. Appl Clin Inform 2015; 6 (04) 769-784
• 12 Hum RS, Cato K, Sheehan B. et al. Developing clinical decision support within a commercial electronic health record system to improve antimicrobial prescribing in the neonatal ICU. Appl Clin Inform 2014; 5 (02) 368-387
• 13 Alrifai MW, Mulherin DP, Weinberg ST, Wang L, Lehmann CU. Parenteral protein decision support system improves protein delivery in preterm infants: a randomized clinical trial. JPEN J Parenter Enteral Nutr 2018; 42 (01) 219-224
• 14 Longhurst C, Turner S, Burgos AE. Development of a web-based decision support tool to increase use of neonatal hyperbilirubinemia guidelines. Jt Comm J Qual Patient Saf 2009; 35 (05) 256-262
• 15 Palma JP, Arain YH. Development of a web-based decision support tool to operationalize and optimize management of hyperbilirubinemia in preterm infants. Clin Perinatol 2016; 43 (02) 375-383
• 16 Petersen JD, Lozovatsky M, Markovic D. et al. Clinical decision support for hyperbilirubinemia risk assessment in the electronic health record. Acad Pediatr 2020; 20 (06) 857-862
• 17 Kawamoto K, Kukhareva P, Shakib JH. et al. Association of an electronic health record add-on app for neonatal bilirubin management with physician efficiency and care quality. JAMA Netw Open 2019; 2 (11) e1915343
• 18 Hunt DL, Haynes RB, Hanna SE, Smith K. Effects of computer-based clinical decision support systems on physician performance and patient outcomes: a systematic review. JAMA 1998; 280 (15) 1339-1346
• 19 Garg AX, Adhikari NKJ, McDonald H. et al. Effects of computerized clinical decision support systems on practitioner performance and patient outcomes: a systematic review. JAMA 2005; 293 (10) 1223-1238
• 20 Kilsdonk E, Peute LW, Jaspers MWM. Factors influencing implementation success of guideline-based clinical decision support systems: a systematic review and gaps analysis. Int J Med Inform 2017; 98: 56-64
• 21 Kawamoto K, Houlihan CA, Balas EA, Lobach DF. Improving clinical practice using clinical decision support systems: a systematic review of trials to identify features critical to success. BMJ 2005; 330 (7494) 765
• 22 Van de Velde S, Kunnamo I, Roshanov P. et al; GUIDES expert panel. The GUIDES checklist: development of a tool to improve the successful use of guideline-based computerised clinical decision support. Implement Sci 2018; 13 (01) 86
• 23 Bates DW, Kuperman GJ, Wang S. et al. Ten commandments for effective clinical decision support: making the practice of evidence-based medicine a reality. J Am Med Inform Assoc 2003; 10 (06) 523-530

Publication History

Received: 15 April 2024

Accepted: 18 June 2024

Accepted Manuscript online:
19 June 2024

Article published online:
11 September 2024