Behavior
Conditions
Keywords
Medication safety, Prescribing practice, Prescribing behaviour
Brief summary
Background Medication errors are the leading cause of preventable harm in healthcare settings worldwide. An estimated 237 million medication errors occur in England alone every year, with 66 million considered clinically significant. There is an estimated cost to the NHS from definitely avoidable adverse drug reactions as a result of these errors of £98.5 million per year, consuming 181,626 bed-days and causing to 712 deaths. Medication related clinical decision support systems, often integrated with electronic prescribing systems, are rapidly increasing in number over the last few decades, ranging from drug-drug interaction alerts to allergy checks and formulary support. A recent systematic review summarised that these systems are still relatively immature, with limited use of patient-specific input or human factors research used to develop them. There is an opportunity to improve these systems significantly for the benefit of the user and for patient safety. The World Health Organization propose that interventions to reduce medication error should include the development of technologies that are well understood and designed for the systems and practice they are applied to. Human factors and usability engineering is an integral part of developing medical devices, such as clinical decision support (CDS) systems, to ensure that such devices are easy to use and can be used safely as intended. User testing / usability testing, which may incorporate several methods, should be conductive throughout the development process (at formative, summative assessment, and during post-market surveillance). These methods are now becoming more common place in healthcare technology research and should continue to support the development of new technologies. RxConnect RxConnect, a newly registered UKCA marked medical device, is an on-demand clinical decision support tool that receives medication and patient inputs and uses them to filter an underlying formulary, such as the BNF, and perform dosing calculations, as needed, to return patient-specific dosing recommendations. RxConnect does not have a user interface and relies on an integration with third-party systems, such as electronic prescribing systems, to deliver CDS services to clinical end users. For this study a prototype user interface for RxConnect that emulates a typical electronic prescribing system will be used. The study team hypothesise that use of RxConnect as a digital prescribing aid is quicker, easier, and as safe to use as currently available prescribing aids. This study aims to utilise user testing to prove or disprove the above hypothesis and to generate quantitative and qualitative outputs to support the continued development of RxConnect prior to clinical deployment.
Interventions
OTHERRxConnect
Participants use RxConnect, an on-demand clinical decision support tool that receives medication and patient inputs and uses them to filter an underlying formulary, such as the BNF, and perform dosing calculations, as needed, to return patient-specific dosing recommendations.
Sponsors
National Institute for Health Research, United Kingdom
Imperial College London
Study design
Allocation
RANDOMIZED
Intervention model
CROSSOVER
Primary purpose
BASIC_SCIENCE
Masking
SINGLE (Subject)
Eligibility
Sex/Gender
ALL
Healthy volunteers
Yes
Inclusion criteria
* Willingness to consent and participate * Medical doctor - Foundation year 1 and above OR registered non-medical prescriber (e.g. nurses or pharmacists) * Regular (at least weekly) experience in prescribing medications as part of working role
Exclusion criteria
* Infrequent prescribing practice (less than once a week) * Not willing to participate
Design outcomes
Primary
| Measure | Time frame | Description |
|---|---|---|
| Number of Prescribing Errors by Study Arm | 60 minutes | Sub analysis of errors by type available in full report |
Secondary
| Measure | Time frame | Description |
|---|---|---|
| Number of Medication Orders With a Large Magnitude Error (Greater Than 25% of the Recommended Dosing Range) | 60 minutes | Dosing errors with a deviation of more than 25% from the recommended range were categorised as large magnitude errors. |
| Time Taken to Prescribe Each Medication | 60 minutes | For the first scenario, TTP was calculated from the moment the participant began reading the scenario to task completion, while for subsequent scenarios, timing started from the completion of the previous scenario. The endpoint for each scenario was marked by the participant's submission of the medication order on the electronic prescribing (eP) system. |
| Measurement of the Prescribers Perceived Mental Load Per Prescribing Scenario | 60 minutes | Measurement of the Prescribers perceived mental load per prescribing scenario, Using NASA task load index (TLX). An overall workload score combining all 6 NASA TLX domains was calculated (minimum 0 lower workload - maximum 126 highest workload). |
Other
| Measure | Time frame | Description |
|---|---|---|
| Erroneous Medication Orders by Hierarchial Task Analysis (Identifying Vulnerable Steps in the Prescribing Workflow). | 60 minutes | Erroneous orders identified as a primary outcome of the study will then be analysed using hierarchical task analysis (HTA). The HTA is a qualitative outcome, different from the primary outcome (error yes/no) by instead identifying 'where' within the prescribing process an error occurred. Workflow steps, representing tasks or actions in both the control and intervention arms, were developed based on established and anticipated prescribing workflows and refined as new, unanticipated steps emerged during study observations. These workflows were then employed for hierarchical task analysis, as detailed in the data analysis section. Hierarchical task analysis was conducted by reviewing recordings of all erroneous medication orders, breaking down the prescribing process into discrete steps. This structured approach allowed for identification of potential risks or inefficiencies in the workflow, helping trace each error's likely origin within the process. |
| Number of Participants That Gave Qualitative Feedback | 60 minutes | Audio of interviews will be transcribed verbatim and thematically analyses to provide insights from participants that can be utilised for recommendations for practice and research. |
Countries
United Kingdom
Participant flow
Participants by arm
| Arm | Count |
|---|---|
| All Study Participants All study participants exposed to both control and intervention, therefore reported together | 24 |
| Total | 24 |
Baseline characteristics
| Characteristic | All Study Participants | — |
|---|---|---|
| Age, Categorical <=18 years | 0 Participants | — |
| Age, Categorical >=65 years | 0 Participants | — |
| Age, Categorical Between 18 and 65 years | 24 Participants | — |
| Participant grade (self reported titles used) Clinical Fellow | 3 Participants | — |
| Participant grade (self reported titles used) Consultant | 1 Participants | — |
| Participant grade (self reported titles used) CT/ST Years 1-5 | 5 Participants | — |
| Participant grade (self reported titles used) Foundation Year 1 | 1 Participants | — |
| Participant grade (self reported titles used) Pharmacist Pay Band 8A | 1 Participants | — |
| Participant grade (self reported titles used) Pharmacist Pay Band 8B | 3 Participants | — |
| Participant grade (self reported titles used) Registrar | 1 Participants | — |
| Participant grade (self reported titles used) Senior House Officer | 1 Participants | — |
| Participant grade (self reported titles used) ST Years 6-8 | 6 Participants | — |
| Participant grade (self reported titles used) Trust Grade | 2 Participants | — |
| Profession Doctor | 20 Participants | — |
| Profession Pharmacist | 4 Participants | — |
| Race and Ethnicity Not Collected | — | — Participants |
| Sex/Gender, Customized Gender Female | 15 Participants | — |
| Sex/Gender, Customized Gender Male | 8 Participants | — |
| Sex/Gender, Customized Gender Not stated | 1 Participants | — |
| Speciality Adults | 10 Participants | — |
| Speciality Paediatric Emergency | 1 Participants | — |
| Speciality Paediatric Intensive Care Unit | 4 Participants | — |
| Speciality Paediatrics | 9 Participants | — |
| Years using Cerner 1-2 years | 2 Participants | — |
| Years using Cerner <1 year | 6 Participants | — |
| Years using Cerner 2-3 years | 6 Participants | — |
| Years using Cerner 3-4 years | 2 Participants | — |
| Years using Cerner 4-5 years | 2 Participants | — |
| Years using Cerner 5+ years | 6 Participants | — |
Adverse events
| Event type | EG000 affected / at risk | EG001 affected / at risk |
|---|---|---|
| deaths Total, all-cause mortality | 0 / 24 | 0 / 24 |
| other Total, other adverse events | 0 / 24 | 0 / 24 |
| serious Total, serious adverse events | 0 / 24 | 0 / 24 |
Outcome results
Primary
Number of Prescribing Errors by Study Arm
Sub analysis of errors by type available in full report
Time frame: 60 minutes
| Arm | Measure | Group | Value (COUNT_OF_UNITS) |
|---|---|---|---|
| RxConnect (Experiment Arm) | Number of Prescribing Errors by Study Arm | Erroneous medication order | 8 Medication orders |
| RxConnect (Experiment Arm) | Number of Prescribing Errors by Study Arm | Dose error | 8 Medication orders |
| RxConnect (Experiment Arm) | Number of Prescribing Errors by Study Arm | Route error | 3 Medication orders |
| RxConnect (Experiment Arm) | Number of Prescribing Errors by Study Arm | Patient error | 0 Medication orders |
| RxConnect (Experiment Arm) | Number of Prescribing Errors by Study Arm | Frequency error | 0 Medication orders |
| RxConnect (Experiment Arm) | Number of Prescribing Errors by Study Arm | Formulation error | 3 Medication orders |
| Current Practice (Control) Arm | Number of Prescribing Errors by Study Arm | Frequency error | 3 Medication orders |
| Current Practice (Control) Arm | Number of Prescribing Errors by Study Arm | Erroneous medication order | 34 Medication orders |
| Current Practice (Control) Arm | Number of Prescribing Errors by Study Arm | Patient error | 1 Medication orders |
| Current Practice (Control) Arm | Number of Prescribing Errors by Study Arm | Dose error | 31 Medication orders |
| Current Practice (Control) Arm | Number of Prescribing Errors by Study Arm | Formulation error | 6 Medication orders |
| Current Practice (Control) Arm | Number of Prescribing Errors by Study Arm | Route error | 4 Medication orders |
Secondary
Measurement of the Prescribers Perceived Mental Load Per Prescribing Scenario
Measurement of the Prescribers perceived mental load per prescribing scenario, Using NASA task load index (TLX). An overall workload score combining all 6 NASA TLX domains was calculated (minimum 0 lower workload - maximum 126 highest workload).
Time frame: 60 minutes
| Arm | Measure | Value (MEAN) |
|---|---|---|
| RxConnect (Experiment Arm) | Measurement of the Prescribers Perceived Mental Load Per Prescribing Scenario | 41.45 score on a scale |
| Current Practice (Control) Arm | Measurement of the Prescribers Perceived Mental Load Per Prescribing Scenario | 57.21 score on a scale |
Secondary
Number of Medication Orders With a Large Magnitude Error (Greater Than 25% of the Recommended Dosing Range)
Dosing errors with a deviation of more than 25% from the recommended range were categorised as large magnitude errors.
Time frame: 60 minutes
| Arm | Measure | Value (COUNT_OF_UNITS) |
|---|---|---|
| RxConnect (Experiment Arm) | Number of Medication Orders With a Large Magnitude Error (Greater Than 25% of the Recommended Dosing Range) | 6 Medication orders |
| Current Practice (Control) Arm | Number of Medication Orders With a Large Magnitude Error (Greater Than 25% of the Recommended Dosing Range) | 22 Medication orders |
Secondary
Time Taken to Prescribe Each Medication
For the first scenario, TTP was calculated from the moment the participant began reading the scenario to task completion, while for subsequent scenarios, timing started from the completion of the previous scenario. The endpoint for each scenario was marked by the participant's submission of the medication order on the electronic prescribing (eP) system.
Time frame: 60 minutes
| Arm | Measure | Value (MEAN) |
|---|---|---|
| RxConnect (Experiment Arm) | Time Taken to Prescribe Each Medication | 179.7 seconds |
| Current Practice (Control) Arm | Time Taken to Prescribe Each Medication | 224.8 seconds |
Other Pre-specified
Erroneous Medication Orders by Hierarchial Task Analysis (Identifying Vulnerable Steps in the Prescribing Workflow).
Erroneous orders identified as a primary outcome of the study will then be analysed using hierarchical task analysis (HTA). The HTA is a qualitative outcome, different from the primary outcome (error yes/no) by instead identifying 'where' within the prescribing process an error occurred. Workflow steps, representing tasks or actions in both the control and intervention arms, were developed based on established and anticipated prescribing workflows and refined as new, unanticipated steps emerged during study observations. These workflows were then employed for hierarchical task analysis, as detailed in the data analysis section. Hierarchical task analysis was conducted by reviewing recordings of all erroneous medication orders, breaking down the prescribing process into discrete steps. This structured approach allowed for identification of potential risks or inefficiencies in the workflow, helping trace each error's likely origin within the process.
Time frame: 60 minutes
| Arm | Measure | Group | Value (NUMBER) |
|---|---|---|---|
| RxConnect (Experiment Arm) | Erroneous Medication Orders by Hierarchial Task Analysis (Identifying Vulnerable Steps in the Prescribing Workflow). | Overall erroneous medication orders | 8 Medication orders |
| RxConnect (Experiment Arm) | Erroneous Medication Orders by Hierarchial Task Analysis (Identifying Vulnerable Steps in the Prescribing Workflow). | Review written patient scenario | 5 Medication orders |
| RxConnect (Experiment Arm) | Erroneous Medication Orders by Hierarchial Task Analysis (Identifying Vulnerable Steps in the Prescribing Workflow). | Access prefered/necesary dosing resource | 0 Medication orders |
| RxConnect (Experiment Arm) | Erroneous Medication Orders by Hierarchial Task Analysis (Identifying Vulnerable Steps in the Prescribing Workflow). | Identify medication, appropriate indication and dose recommendation | 0 Medication orders |
| RxConnect (Experiment Arm) | Erroneous Medication Orders by Hierarchial Task Analysis (Identifying Vulnerable Steps in the Prescribing Workflow). | Identify any relevant dose considerations | 0 Medication orders |
| RxConnect (Experiment Arm) | Erroneous Medication Orders by Hierarchial Task Analysis (Identifying Vulnerable Steps in the Prescribing Workflow). | Calculate dose as per resource directions | 0 Medication orders |
| RxConnect (Experiment Arm) | Erroneous Medication Orders by Hierarchial Task Analysis (Identifying Vulnerable Steps in the Prescribing Workflow). | Condsider if any min/max dose constraints need taking into account | 0 Medication orders |
| RxConnect (Experiment Arm) | Erroneous Medication Orders by Hierarchial Task Analysis (Identifying Vulnerable Steps in the Prescribing Workflow). | Select/search for or confirm route from unfiltered list | 0 Medication orders |
| RxConnect (Experiment Arm) | Erroneous Medication Orders by Hierarchial Task Analysis (Identifying Vulnerable Steps in the Prescribing Workflow). | Select/search for form from unfiltered list | 0 Medication orders |
| RxConnect (Experiment Arm) | Erroneous Medication Orders by Hierarchial Task Analysis (Identifying Vulnerable Steps in the Prescribing Workflow). | Launch required patient in Cerner | 1 Medication orders |
| RxConnect (Experiment Arm) | Erroneous Medication Orders by Hierarchial Task Analysis (Identifying Vulnerable Steps in the Prescribing Workflow). | Select one or more dose recommendation(s) | 1 Medication orders |
| RxConnect (Experiment Arm) | Erroneous Medication Orders by Hierarchial Task Analysis (Identifying Vulnerable Steps in the Prescribing Workflow). | Enter/confirm dose | 1 Medication orders |
| Current Practice (Control) Arm | Erroneous Medication Orders by Hierarchial Task Analysis (Identifying Vulnerable Steps in the Prescribing Workflow). | Select one or more dose recommendation(s) | 0 Medication orders |
| Current Practice (Control) Arm | Erroneous Medication Orders by Hierarchial Task Analysis (Identifying Vulnerable Steps in the Prescribing Workflow). | Overall erroneous medication orders | 34 Medication orders |
| Current Practice (Control) Arm | Erroneous Medication Orders by Hierarchial Task Analysis (Identifying Vulnerable Steps in the Prescribing Workflow). | Condsider if any min/max dose constraints need taking into account | 1 Medication orders |
| Current Practice (Control) Arm | Erroneous Medication Orders by Hierarchial Task Analysis (Identifying Vulnerable Steps in the Prescribing Workflow). | Review written patient scenario | 8 Medication orders |
| Current Practice (Control) Arm | Erroneous Medication Orders by Hierarchial Task Analysis (Identifying Vulnerable Steps in the Prescribing Workflow). | Launch required patient in Cerner | 0 Medication orders |
| Current Practice (Control) Arm | Erroneous Medication Orders by Hierarchial Task Analysis (Identifying Vulnerable Steps in the Prescribing Workflow). | Access prefered/necesary dosing resource | 1 Medication orders |
| Current Practice (Control) Arm | Erroneous Medication Orders by Hierarchial Task Analysis (Identifying Vulnerable Steps in the Prescribing Workflow). | Select/search for or confirm route from unfiltered list | 1 Medication orders |
| Current Practice (Control) Arm | Erroneous Medication Orders by Hierarchial Task Analysis (Identifying Vulnerable Steps in the Prescribing Workflow). | Identify medication, appropriate indication and dose recommendation | 7 Medication orders |
| Current Practice (Control) Arm | Erroneous Medication Orders by Hierarchial Task Analysis (Identifying Vulnerable Steps in the Prescribing Workflow). | Enter/confirm dose | 0 Medication orders |
| Current Practice (Control) Arm | Erroneous Medication Orders by Hierarchial Task Analysis (Identifying Vulnerable Steps in the Prescribing Workflow). | Identify any relevant dose considerations | 12 Medication orders |
| Current Practice (Control) Arm | Erroneous Medication Orders by Hierarchial Task Analysis (Identifying Vulnerable Steps in the Prescribing Workflow). | Select/search for form from unfiltered list | 1 Medication orders |
| Current Practice (Control) Arm | Erroneous Medication Orders by Hierarchial Task Analysis (Identifying Vulnerable Steps in the Prescribing Workflow). | Calculate dose as per resource directions | 3 Medication orders |
Other Pre-specified
Number of Participants That Gave Qualitative Feedback
Audio of interviews will be transcribed verbatim and thematically analyses to provide insights from participants that can be utilised for recommendations for practice and research.
Time frame: 60 minutes
Population: There is only one arm reported for the qualitative feedback as all participants provided feedback.
| Arm | Measure | Value (NUMBER) |
|---|---|---|
| RxConnect (Experiment Arm) | Number of Participants That Gave Qualitative Feedback | 24 participants |