< Back to previous page
Publication
Improving patient safety through health information technology
Book - Dissertation
Subtitle:focus on drug-drug interactions and drug allergies
Primum non nocere, or above all, do no harm! This axiom is central in medical education and a potent reminder that each medical and pharmacological decision carries a certain potential for harm. Nevertheless, in 2013, patient harm as a result of unsafe medical care was estimated to be the 14th leading cause of global disease burden, comparable to tuberculosis, malaria and multiple sclerosis. Harm and costs due to medication errors and adverse drug events are substantial and a global issue. Many patient safety reports have advocated for the use of health information technology (IT) as (one of) the most promising strategy(ies) to improve patient safety, clinical outcomes, and efficiency with most focus on electronic health records with computerized physician order entry and clinical decision support systems (CDSS). CDSS are health IT applications that combine clinical knowledge with patient information to support clinicians at appropriate times in decision making, often by presenting safety alerts. These applications have predominantly achieved positive results, but there are important areas of improvement.
In this thesis, we focused on CDSS for drug-drug interaction screening and on allergy documentation in electronic health records as an essential building block for CDSS for drug allergy screening.
We demonstrated that implementation of multiple CDSS optimizations at once resulted in a higher alert acceptance and that the inclusion of a single context factor for potassium-increasing drug-drug interactions resulted in a significant decrease in alert burden and an increase in alert acceptance without a significant difference in occurrence of hyperkalemia, thus preserving patient safety. Pharmacist interventions for drug-drug interactions through the CDSS follow-up system were consistently more often accepted than CDSS alerts for drug-drug interactions. This indicates, next to complementarity of both approaches, more impact of such pharmacist interventions. For pharmacist interventions for QT drug-drug interactions, patients had on average five additional risk factors. Prediction models were developed to incorporate such additional risk factors in the CDSS rules for QT drug-drug interactions in order to increase the clinical relevance of QT alerts. The linear prediction model outperformed the current QT alert risk stratification based on severity level alone, but performance should still be further improved before implementation in clinical practice. We also found that machine learning techniques have the potential to outperform classical statistical methods in the prediction of QTc prolongation.
In a first step towards the development of a CDSS for drug allergy screening in our institution, we analyzed the technical possibilities to develop such a CDSS, but we observed that the allergy documentation did not fulfill technical requirements. Through a survey study, we learned that also
231from a clinical point of view, revision of the allergy documentation module was required. Based on results of this survey study, additional literature review, and discussions with end users and software experts, two new versions of an allergy documentation module were developed with focus on user experience. In a usability study, we evaluated and compared the user satisfaction, task time, and accuracy of the old allergy documentation module and these two new versions. The key difference with the old documentation module was that in the new versions more information was requested which was also encoded and thus usable for real-time CDSS. For both new versions, satisfaction and accuracy were significantly improved compared to the old version. Task time did not significantly differ between the three versions. However, satisfaction scores were not yet excellent. Based on qualitative feedback, a third new version was designed with an extra quick symptom selection feature, revised diagnosis options, and more information and guidance on the new inactivation feature. Next to the graphical user interface of the allergy documentation itself, much effort went into the integration of this allergy module throughout the electronic health record in the notification area, clinical and hospitalization notes, and both the medical and nursing record. The revised allergy documentation module was implemented in May 2022 in the electronic health record of UZ Brussel.
To advance the research into patient-specific CDSS rules, it is extremely important to invest in a properly coded problem list and automatic data collection through the creation of data pipelines. Such a coded problem list would not only benefit CDSS optimization, but also facilitate other types of research, most notably clinical real-world data studies. A coded problem list will also improve efficiency in clinical practice by decreasing administrative burden and by facilitating clinical patient dashboards. The realization of the new coded allergen list can be considered as a successful proof of concept that could be translated to a coded problem list. Evaluation studies on the allergy documentation module should investigate both quantitative and qualitative parameters. Another future perspective is the development and implementation of an advanced CDSS for drug allergy screening with stratification of alerts according to contextual factors.
The current CDSS for drug-drug interaction screening as well as the future CDSS for drug allergy screening, just as some other functionalities in the electronic health record, need certification under the Medical Device Regulation. This new law with its stringent safety and performance requirements for medical software will require an enormous culture change and will lead to a new interdisciplinary field of expertise with many opportunities to improve patient safety.
In this thesis, we focused on CDSS for drug-drug interaction screening and on allergy documentation in electronic health records as an essential building block for CDSS for drug allergy screening.
We demonstrated that implementation of multiple CDSS optimizations at once resulted in a higher alert acceptance and that the inclusion of a single context factor for potassium-increasing drug-drug interactions resulted in a significant decrease in alert burden and an increase in alert acceptance without a significant difference in occurrence of hyperkalemia, thus preserving patient safety. Pharmacist interventions for drug-drug interactions through the CDSS follow-up system were consistently more often accepted than CDSS alerts for drug-drug interactions. This indicates, next to complementarity of both approaches, more impact of such pharmacist interventions. For pharmacist interventions for QT drug-drug interactions, patients had on average five additional risk factors. Prediction models were developed to incorporate such additional risk factors in the CDSS rules for QT drug-drug interactions in order to increase the clinical relevance of QT alerts. The linear prediction model outperformed the current QT alert risk stratification based on severity level alone, but performance should still be further improved before implementation in clinical practice. We also found that machine learning techniques have the potential to outperform classical statistical methods in the prediction of QTc prolongation.
In a first step towards the development of a CDSS for drug allergy screening in our institution, we analyzed the technical possibilities to develop such a CDSS, but we observed that the allergy documentation did not fulfill technical requirements. Through a survey study, we learned that also
231from a clinical point of view, revision of the allergy documentation module was required. Based on results of this survey study, additional literature review, and discussions with end users and software experts, two new versions of an allergy documentation module were developed with focus on user experience. In a usability study, we evaluated and compared the user satisfaction, task time, and accuracy of the old allergy documentation module and these two new versions. The key difference with the old documentation module was that in the new versions more information was requested which was also encoded and thus usable for real-time CDSS. For both new versions, satisfaction and accuracy were significantly improved compared to the old version. Task time did not significantly differ between the three versions. However, satisfaction scores were not yet excellent. Based on qualitative feedback, a third new version was designed with an extra quick symptom selection feature, revised diagnosis options, and more information and guidance on the new inactivation feature. Next to the graphical user interface of the allergy documentation itself, much effort went into the integration of this allergy module throughout the electronic health record in the notification area, clinical and hospitalization notes, and both the medical and nursing record. The revised allergy documentation module was implemented in May 2022 in the electronic health record of UZ Brussel.
To advance the research into patient-specific CDSS rules, it is extremely important to invest in a properly coded problem list and automatic data collection through the creation of data pipelines. Such a coded problem list would not only benefit CDSS optimization, but also facilitate other types of research, most notably clinical real-world data studies. A coded problem list will also improve efficiency in clinical practice by decreasing administrative burden and by facilitating clinical patient dashboards. The realization of the new coded allergen list can be considered as a successful proof of concept that could be translated to a coded problem list. Evaluation studies on the allergy documentation module should investigate both quantitative and qualitative parameters. Another future perspective is the development and implementation of an advanced CDSS for drug allergy screening with stratification of alerts according to contextual factors.
The current CDSS for drug-drug interaction screening as well as the future CDSS for drug allergy screening, just as some other functionalities in the electronic health record, need certification under the Medical Device Regulation. This new law with its stringent safety and performance requirements for medical software will require an enormous culture change and will lead to a new interdisciplinary field of expertise with many opportunities to improve patient safety.
Number of pages: 244
Publication year:2022
Keywords:drug allergies, patient safety, Health information, Drug-drug interactions
Accessibility:Embargoed