ADAM's APPLE: A Study to the Impact of Accuracy Problem Lists in Electronic Health Records on Correctness and Speed of Clinical Decision-making Performed by Dutch Healthcare Providers

Sponsor

Eva Klappe (Other)

Overall Status

Completed

CT.gov ID

NCT05657002

Collaborator

(none)

160

Enrollment

Location

Arms

Actual Duration (Days)

243.5

Patients Per Site Per Month

Study Details

Study Description

Brief Summary

The primary objective of this study is to determine whether patient records with complete, structured and up-to-date problem lists ('accurate problem lists'), result in better clinical decision-making, compared to patient records that convey the same information in a less structured way where the problem list has missing and/or duplicate diagnoses ('inaccurate problem lists'). The secondary objective is to determine whether the time required to make a correct decision is less for patient records with accurate problem lists compared to patient records with inaccurate problem lists.

Condition or Disease	Intervention/Treatment	Phase
Clinical Decision-Making Decision Making, Computer-assisted Medical Records, Problem-Oriented Quality of Health Care Evidence-Based Practice Data Accuracy Documentation / Standards Documentation / Statistics & Numerical Data Forms and Records Control / Standards Humans International Classification of Diseases / Standards	Other: patient A with accurate problem list Other: patient B with inaccurate problem list Other: patient A with inaccurate problem list Other: patient B with accurate problem list	N/A

Detailed Description

A problem list in Electronic Health Records (EHRs) is considered an essential feature in the collection of structured data. The problem list provides a centralized summary of each patient's medical problems and these problems or diagnoses are selected from the terminology underlying the problem list, such as SNOMED CT. If well maintained and structured, the problem list is a valuable tool for reviewing records of (unfamiliar) patients as it quickly shows the required information when needed. While studies have shown that the use of structured formats can serve as prompt for extra details, greater consistency of information and clinical decision-making there is little evidence whether a patient record with complete and structured problem lists results in more accurate and faster clinical decision making.

In the study (ADAM's APPLE: Adequate Data registration And Monitoring, subproject: Accurate Presentation of Problem List Elements), the investigators will perform a crossover randomized controlled trial in which a laboratory experiment will be performed among individual healthcare professionals to assess the impact of patient records with accurate and inaccurate problem lists on clinical decision-making. The participants will be presented with two records of two different patients in a training environment of the software system EPIC, one of them with an accurate problem list and the other that conveys the complete information in the patient record (in free text notes) but with an inaccurate problem list with missing diagnoses and duplicate information. The participants do not know which of the two records includes the accurate problem list and which record includes the inaccurate problem list. Participants are asked to decide whether or not to prescribe two medications for those two patients. One medication is not allowed per patient because the patient is allergic to that medication, which is documented on the allergy list. For the first patient record, the other medication is not allowed, because of a contraindicated diagnosis and for the second patient record the other medication is not allowed, because of a side effect that has occurred using that medication in the medical history. Based on the correctness of the motivation for correct answers and the time to the right answers, the research question if accurate problem lists in patient records lead to better and faster decision-making is answered.

Prior to this study, two healthcare professionals in the research team determined suitable use cases and questions for this study. These use cases were based on real-world unstructured versions of patient records. Two optimized accurate problem lists were also created for both patient records, which was defined according to the problem list policy at our institution (i.e. all current active problems and relevant medical history should be documented on the problem list).

Study Design

Study Type:

Interventional

Actual Enrollment :

160 participants

Allocation:

Randomized

Intervention Model:

Crossover Assignment

Intervention Model Description:

Participants will be presented with 2 patient records in EPIC, one with an accurate problem list and the other that conveys the complete information in the patient record (in free text notes) but with an inaccurate problem list with missing diagnoses and duplicates. Each participant answers a total of 4 questions for 2 separate patient cases (2 questions per case), one of those cases having the inaccurate and the other having the accurate problem list. Using 2 separate patient cases rather than one was decided to prevent a memory effect, i.e., participants would have had time to understand the patient's conditions described in the record which could impact the time required to answer questions in round 2. Additionally, the order in which patients cases are provided is not randomized, since in practice it also happens that a new patient is followed up by another new patient which requires professionals to systematically go through two separate records with little time in between.Participants will be presented with 2 patient records in EPIC, one with an accurate problem list and the other that conveys the complete information in the patient record (in free text notes) but with an inaccurate problem list with missing diagnoses and duplicates. Each participant answers a total of 4 questions for 2 separate patient cases (2 questions per case), one of those cases having the inaccurate and the other having the accurate problem list. Using 2 separate patient cases rather than one was decided to prevent a memory effect, i.e., participants would have had time to understand the patient's conditions described in the record which could impact the time required to answer questions in round 2. Additionally, the order in which patients cases are provided is not randomized, since in practice it also happens that a new patient is followed up by another new patient which requires professionals to systematically go through two separate records with little time in between.

Masking:

Single (Participant)

Masking Description:

Participant will be blinded, they do not know in what group they belong (thus which record they receive with the accurate or inaccurate problem list) and the participants are not informed that impact of problem list accuracy is investigated. An independent researcher produced a randomization schema where per block of 10 participants a balanced random order of "patient A with accurate problem list + patient B with inaccurate problem list" or "patient A with inaccurate problem list + patient B with accurate problem list" is produced. The investigator (ESK) follows this list for the consecutive participants and gives access to the appropriate patient records as defined by the randomisation schema. The independent researcher checks afterwards whether the right order is followed based on time stamps and randomisation schema. We performed a power analysis before conducting any experiment, which resulted in a required number of 157 participants.

Primary Purpose:

Health Services Research

Official Title:

a Randomized Controlled Trial Study to Determine the Impact of Accuracy of Problem Lists in Electronic Health Records on Clinical Decision-making

Actual Study Start Date :

Dec 1, 2022

Actual Primary Completion Date :

Dec 21, 2022

Actual Study Completion Date :

Dec 21, 2022

Arms and Interventions

Arm	Intervention/Treatment
Active Comparator: accurate problem list, then inaccurate problem list in round 1, the participants will use the patient record of patient A, with an accurate problem list and answer the question: can the patient be prescribed Medication X and Y where medication X is a control question and medication Y is related to a contraindicated diagnosis (on problem list) In round 2, the participants will use the patient record of patient B, with an inaccurate problem list and answer the question: can the patient be prescribed Medication X and Y where medication X is a control question and medication Y is related to medical history (not on problem list)	Other: patient A with accurate problem list A problem list that contains a diagnosis code that is contraindicated with a type of medication (Y). Also, all other relevant diagnoses and medical history for the patient are up-to-date on the problem list, which was defined according to the problem list policy at our institution (i.e. all current active problems and relevant medical history should be documented on the problem list). Additionally, the problem list is included in eight out of thirteen notes using so-called smart phrases that can automatically import a (part of a) problem list. One note includes the problem list with the diagnosis relevant for the question asked. Other: patient B with inaccurate problem list A problem list that does not contain the diagnosis code and corresponding details explaining medical history of this diagnosis caused by a type of medication (Y). Additionally, the problem list is included in three out of thirteen notes using so-called smart phrases that can automatically import a (part of a) problem list. The relevant diagnosis is not documented on the problem list and hence is not included in the imported problem list in the notes. The expert panel provided and anonymized two real-world representative examples of hematology patient records that included inaccurate problem lists and that had many free-text notes. An 'inaccurate problem list' is defined as a problem list where diagnoses are missing resulting in missed trigger medication or order-alerts, where diagnoses are 'active' although they should be closed or removed and/or where the problem list contained duplicated diagnoses.
Active Comparator: inaccurate problem list, then accurate problem list in round 1, the participants will use the patient record of patient A, with an inaccurate problem list and answer the question: can the patient be prescribed Medication X and Y where medication X is a control question and medication Y is related to a contraindicated diagnosis (not on problem list) In round 2, the participants will use the patient record of patient B, with an accurate problem list and answer the question: can the patient be prescribed Medication X and Y where medication X is a control question and medication Y is related to medical history (on problem list)	Other: patient A with inaccurate problem list A problem list that does not contain the diagnosis code that is contraindicated with the type of medication (Y). Additionally, the problem list is included in eight out of thirteen notes using so-called smart phrases that can automatically import a (part of a) problem list. The relevant diagnosis is not documented on the problem list and hence is not included in the imported problem list in the notes. The expert panel provided and anonymized two real-world representative examples of hematology patient records that included inaccurate problem lists and that had many free-text notes. An 'inaccurate problem list' is defined as a problem list where diagnoses are missing resulting in missed trigger medication or order-alerts, where diagnoses are 'active' although they should be closed or removed and/or where the problem list contained duplicated diagnoses. Other: patient B with accurate problem list A problem list that contains the diagnosis code and corresponding details explaining medical history of this diagnosis caused by a type of medication (Y). Also, all other relevant diagnoses and medical history for the patient are up-to-date on the problem list, which was defined according to the problem list policy at our institution (i.e. all current active problems and relevant medical history should be documented on the problem list). Additionally, the problem list is included in three out of thirteen notes using so-called smart phrases that can automatically import a (part of a) problem list. One note includes the problem list with the diagnosis and details relevant for the question asked.

Arm

Intervention/Treatment

Active Comparator: accurate problem list, then inaccurate problem list

in round 1, the participants will use the patient record of patient A, with an accurate problem list and answer the question: can the patient be prescribed Medication X and Y where medication X is a control question and medication Y is related to a contraindicated diagnosis (on problem list) In round 2, the participants will use the patient record of patient B, with an inaccurate problem list and answer the question: can the patient be prescribed Medication X and Y where medication X is a control question and medication Y is related to medical history (not on problem list)

Other: patient A with accurate problem list

A problem list that contains a diagnosis code that is contraindicated with a type of medication (Y). Also, all other relevant diagnoses and medical history for the patient are up-to-date on the problem list, which was defined according to the problem list policy at our institution (i.e. all current active problems and relevant medical history should be documented on the problem list). Additionally, the problem list is included in eight out of thirteen notes using so-called smart phrases that can automatically import a (part of a) problem list. One note includes the problem list with the diagnosis relevant for the question asked.

Other: patient B with inaccurate problem list

A problem list that does not contain the diagnosis code and corresponding details explaining medical history of this diagnosis caused by a type of medication (Y). Additionally, the problem list is included in three out of thirteen notes using so-called smart phrases that can automatically import a (part of a) problem list. The relevant diagnosis is not documented on the problem list and hence is not included in the imported problem list in the notes. The expert panel provided and anonymized two real-world representative examples of hematology patient records that included inaccurate problem lists and that had many free-text notes. An 'inaccurate problem list' is defined as a problem list where diagnoses are missing resulting in missed trigger medication or order-alerts, where diagnoses are 'active' although they should be closed or removed and/or where the problem list contained duplicated diagnoses.

Active Comparator: inaccurate problem list, then accurate problem list

in round 1, the participants will use the patient record of patient A, with an inaccurate problem list and answer the question: can the patient be prescribed Medication X and Y where medication X is a control question and medication Y is related to a contraindicated diagnosis (not on problem list) In round 2, the participants will use the patient record of patient B, with an accurate problem list and answer the question: can the patient be prescribed Medication X and Y where medication X is a control question and medication Y is related to medical history (on problem list)

Other: patient A with inaccurate problem list

A problem list that does not contain the diagnosis code that is contraindicated with the type of medication (Y). Additionally, the problem list is included in eight out of thirteen notes using so-called smart phrases that can automatically import a (part of a) problem list. The relevant diagnosis is not documented on the problem list and hence is not included in the imported problem list in the notes. The expert panel provided and anonymized two real-world representative examples of hematology patient records that included inaccurate problem lists and that had many free-text notes. An 'inaccurate problem list' is defined as a problem list where diagnoses are missing resulting in missed trigger medication or order-alerts, where diagnoses are 'active' although they should be closed or removed and/or where the problem list contained duplicated diagnoses.

Other: patient B with accurate problem list

A problem list that contains the diagnosis code and corresponding details explaining medical history of this diagnosis caused by a type of medication (Y). Also, all other relevant diagnoses and medical history for the patient are up-to-date on the problem list, which was defined according to the problem list policy at our institution (i.e. all current active problems and relevant medical history should be documented on the problem list). Additionally, the problem list is included in three out of thirteen notes using so-called smart phrases that can automatically import a (part of a) problem list. One note includes the problem list with the diagnosis and details relevant for the question asked.

Outcome Measures

Primary Outcome Measures

the correctness of the answer of medication B including the right motivation [during the experiment/questionnaire]
Measured using a questionnaire showing the two question(s) per patient record on a separate tablet during the experiment. Two Yes/No questions per patient record (so a total of four questions per participant) are answered. Answers are stored using an automatically assigned unique anonymized identifier. The question for both patient records is: "can the patient be prescribed medication A and/or B?". Medications A and B can both not be prescribed, but medication B is related to the problem list diagnoses and medication A is related to the allergy list which is the same for both versions of the patient records. A motivation is required per Yes/No answer to determine the correctness of the answer and prevent from a chance of gambling. An independent researcher from the research team who did not perform the experiments will categorize the motivation of the answers for medication B.

Secondary Outcome Measures

the total time to answer the two questions correctly, where the answer of medication B also includes the right motivation [during the experiment/questionnaire]
The time to enter the motivation was not measured. However, the participants might give the right answer but the wrong motivation. For this secondary outcome measure, only the answers that have the right motivation are considered as 'right answer'.
the total time to answer the two questions correctly including the right motivations for medication A and B [during the experiment/questionnaire]
A time stamp is registered when the participant opens the question, and a time stamp is registered when the participant confirms the answers. A time stamp is therefore registered for the total of two Yes/No answers per patient record.
the correctness of both the answer for medication A and B including the right motivations for A and B [during the experiment/questionnaire]
Measured using a questionnaire showing the two question(s) per patient record on a separate tablet during the experiment. Two Yes/No questions per patient record (so a total of four questions per participant) are answered. Answers are stored using an automatically assigned unique anonymized identifier. The question for both patient records is: "can the patient be prescribed medication A and/or B?". Medications A and B can both not be prescribed, but medication B is related to the problem list diagnoses and medication A is related to the allergy list which is the same for both versions of the patient records. A motivation is required per Yes/No answer to determine the correctness of the answer and prevent from a chance of gambling. An independent researcher from the research team who did not perform the experiments will categorize the motivation of the answers for medication A and B.
the correctness of the answer for medication A including the right explanation [during the experiment/questionnaire]
Measured using a questionnaire showing the two question(s) per patient record on a separate tablet during the experiment. Two Yes/No questions per patient record (so a total of four questions per participant) are answered. Answers are stored using an automatically assigned unique anonymized identifier. The question for both patient records is: "can the patient be prescribed medication A and/or B?". Medications A and B can both not be prescribed, but medication B is related to the problem list diagnoses and medication A is related to the allergy list which is the same for both versions of the patient records. A motivation is required per Yes/No answer to determine the correctness of the answer and prevent from a chance of gambling. An independent researcher from the research team who did not perform the experiments will categorize the motivation of the answers for medication A.

Eligibility Criteria

Criteria

Ages Eligible for Study:

N/A and Older

Sexes Eligible for Study:

All

Accepts Healthy Volunteers:

Yes

Inclusion Criteria:

Healthcare professionals who are allowed to prescribe medication, thus hold a position as: medical specialist, medical resident, nurse specialist or physician assistant, research-specialists
Healthcare professionals must have followed at least the 'basic EHR Epic course'. This electronic health record course lasts for three days and includes how to send letters, register diagnoses in a record, request testing, all in the software system EPIC, which concludes with an exam on the theory.

Exclusion Criteria:

Non-Dutch speaking employees as the patient cases and the exercises are described in Dutch

Contacts and Locations

Locations

	Site	City	State	Country	Postal Code
1	Amsterdam UMC, Location AMC	Amsterdam	Noord-Holland	Netherlands	1105AZ

Sponsors and Collaborators

Eva Klappe

Investigators

Principal Investigator: Eva Klappe, MSc, Academisch Medisch Centrum - Universiteit van Amsterdam (AMC-UvA)

Study Documents (Full-Text)

None provided.

More Information

Publications

None provided.

Responsible Party:

Eva Klappe, PhD Candidate, Academisch Medisch Centrum - Universiteit van Amsterdam (AMC-UvA)

ClinicalTrials.gov Identifier:

NCT05657002

Other Study ID Numbers:

Amsterdam UMC 2019-AMC-JK-7
2019-AMC-JK-7

First Posted:

Dec 20, 2022

Last Update Posted:

Dec 23, 2022

Last Verified:

Dec 1, 2022

Individual Participant Data (IPD) Sharing Statement:

Plan to Share IPD:

Studies a U.S. FDA-regulated Drug Product:

Studies a U.S. FDA-regulated Device Product:

Study Results

No Results Posted as of Dec 23, 2022