Presentation
Simulation-Based Multi-Site Evaluation of EHR Safeguards for Pediatric Weight-Based Dosing
SessionDH8: Connected Care Systems: Integrating Digital Health, Clinical Workflows, and Patient Safety
DescriptionBackground
Weight-based dosing (WBD), the cornerstone of accurate and safe pediatric medication ordering and administration, is associated with substantial medication errors that can result in disproportionately harmful effects for children. Despite their intended role in supporting safe prescribing, electronic health records (EHRs) often present usability challenges that contribute to WBD errors, including mismatches with clinical workflows, alert fatigue, and insufficient support with calculating doses. As a result, providers are frequently left to rely on manual double-checks and workarounds, which increase cognitive burden and risk of error. To better understand these vulnerabilities, our team developed a human factors-informed checklist and associated test cases to systematically evaluate EHR safeguards against WBD errors. By testing this checklist across multiple institutions in EHR test environments, we aimed to capture a cross-site view of the EHR safeguards to promote safe pediatric prescribing and to provide actionable insights for improving usability and safety in safety-critical clinical contexts.
Methods
A multidisciplinary team of pediatric clinicians, pharmacists, human factors engineers, and health IT experts created a 20-item checklist to assess EHR safeguards against pediatric WBD errors in the stages of pre-ordering (e.g., weight documentation and plausibility checks; 3 checklist items), ordering (e.g., dosing calculations, order sets, alerts; 16 checklist items), and administration (e.g., verification and device checks; 1 checklist items). Four standardized test cases were designed to assess each item within EHR test environments.
Usability testing was conducted across 13 U.S. healthcare institutions spanning the Northeast (n=3), Mid-Atlantic (n=2), South (n=3), Midwest (n=2), and West (n=2) in two EHR vendor test environments between February to August of 2025. 13 nurses (female = 12; average age = 35 years) and 13 physicians (female = 8; average age = 31 years) participated in the sessions, which were conducted virtually using secure screen sharing. Nurses tested one test case (4 checklist items) on pre-ordering and administration items related to weight documentation, and a typical session lasted for about 30 minutes. Physicians tested three test cases (16 checklist items) involving ordering and participated for approximately one hour. All sessions were facilitated by human factors researchers (SK, GZ, GF). Researchers read aloud the specific task to be performed on the EHR, then observed participants’ actions while encouraging them to describe their workflow and decision-making. Each checklist item was rated as either passed or failed depending on whether the EHR safeguard functioned as intended. Results were aggregated across institutions to capture cross-site patterns in safeguard performance.
Results
Across 13 institutions, 5 of 20 (25%) safeguards functioned at all sites, while 15 of 20 (75%) failed at one or more sites, indicating broad variability between EHR design and safeguards implemented to prevent WBD errors. Results are organized by phase.
Pre-Ordering Stage
Core safeguards around weight documentation were uniform across health systems: the EHR prevented weight documentation in non-metric units (e.g., pounds) or auto-converted non-kilogram entries (n=13; 100%) and alerted when a new weight exceeded a recent documented value (n=13; 100%). The primary vulnerability was age-appropriate plausibility: only (n=9; 69%) flagged weights that were too high for age (absent in n=4; 31%). Several implementations emphasized percent-change warnings or showed expected ranges but did not clearly cite a growth-chart source.
Ordering Stage
This phase showed the most inconsistency with 14 of 16 ordering safeguards failing in at least one site. The EHR system in most sites enabled capturing a medication indication (e.g., empiric meningitis; n=12; 92%). However, the EHR system in none of the tested sites (0%) provided feedback to the ordering provider if the pharmacy could review the medication indication. At the point of order entry: only two sites (15%) displayed the weight, how it was obtained, and when it was obtained at order entry: 10 sites (77%) displayed the total dose; 11 sites (85%) displayed the formula and mathematics to calculate the specific dose (e.g., 15 kg × 10 mg/kg → 150 mg) hidden behind an icon; and no site (0%) identified the scientific source of suggested doses within the ordering window, but required clicking a link to an external source. Eleven sites (85%) supported ordering multi-component medications (e.g., amoxicillin-clavulanic acid) based on a single component. Sites differed in the ability to detect and interrupt critical overdoses: seven sites (54%) alerted when the single dose of a medication exceeded the safe maximum dose for patients; six sites (46%) alerted when the daily maximum dose of a medication was exceeded; 10 sites (77%) when a dose exceeded ten times the recommended safe dose. Only nine sites (69%) required weight documentation before ordering WBD medications.
Administration Stage
All sites (n = 13, 100%) enabled nurses to document medication administration without barcode scanning and after a time lapse between administration and documentation to support medication administration in emergencies.
Discussion
A total of 20 checklist items were tested across 13 healthcare institutions, yielding 260 assessments. Overall, 175 (67.3%) items passed, 83 (31.9%) failed, and 2 (0.8%) were not tested. The EHR design largely supported the WBD phases of pre-ordering (weight documentation) and medication administration. However, there were several safety-critical gaps in the ordering stage when the EHR system permitted placing an order for WBD medication without a weight documentation, and either did not catch several types of potential overdose inputs (e.g., exceeding single dose or daily dose maximum, or exceeding 10 times the safe dose), or permitted the ordering provider to override the alert and place the order. In these cases, ensuring safe and accurate data entry and guarding against EHR errors was incumbent on providers’ vigilance, which is widely recognized as a brittle safeguard to interrupt errors in busy clinical environments. When present, alerts were inconsistent in phrasing, timing, severity, and ability to be overridden, risking alert fatigue and inability to differentiate between alerts based on their clinical risk.
Weight-based dosing (WBD), the cornerstone of accurate and safe pediatric medication ordering and administration, is associated with substantial medication errors that can result in disproportionately harmful effects for children. Despite their intended role in supporting safe prescribing, electronic health records (EHRs) often present usability challenges that contribute to WBD errors, including mismatches with clinical workflows, alert fatigue, and insufficient support with calculating doses. As a result, providers are frequently left to rely on manual double-checks and workarounds, which increase cognitive burden and risk of error. To better understand these vulnerabilities, our team developed a human factors-informed checklist and associated test cases to systematically evaluate EHR safeguards against WBD errors. By testing this checklist across multiple institutions in EHR test environments, we aimed to capture a cross-site view of the EHR safeguards to promote safe pediatric prescribing and to provide actionable insights for improving usability and safety in safety-critical clinical contexts.
Methods
A multidisciplinary team of pediatric clinicians, pharmacists, human factors engineers, and health IT experts created a 20-item checklist to assess EHR safeguards against pediatric WBD errors in the stages of pre-ordering (e.g., weight documentation and plausibility checks; 3 checklist items), ordering (e.g., dosing calculations, order sets, alerts; 16 checklist items), and administration (e.g., verification and device checks; 1 checklist items). Four standardized test cases were designed to assess each item within EHR test environments.
Usability testing was conducted across 13 U.S. healthcare institutions spanning the Northeast (n=3), Mid-Atlantic (n=2), South (n=3), Midwest (n=2), and West (n=2) in two EHR vendor test environments between February to August of 2025. 13 nurses (female = 12; average age = 35 years) and 13 physicians (female = 8; average age = 31 years) participated in the sessions, which were conducted virtually using secure screen sharing. Nurses tested one test case (4 checklist items) on pre-ordering and administration items related to weight documentation, and a typical session lasted for about 30 minutes. Physicians tested three test cases (16 checklist items) involving ordering and participated for approximately one hour. All sessions were facilitated by human factors researchers (SK, GZ, GF). Researchers read aloud the specific task to be performed on the EHR, then observed participants’ actions while encouraging them to describe their workflow and decision-making. Each checklist item was rated as either passed or failed depending on whether the EHR safeguard functioned as intended. Results were aggregated across institutions to capture cross-site patterns in safeguard performance.
Results
Across 13 institutions, 5 of 20 (25%) safeguards functioned at all sites, while 15 of 20 (75%) failed at one or more sites, indicating broad variability between EHR design and safeguards implemented to prevent WBD errors. Results are organized by phase.
Pre-Ordering Stage
Core safeguards around weight documentation were uniform across health systems: the EHR prevented weight documentation in non-metric units (e.g., pounds) or auto-converted non-kilogram entries (n=13; 100%) and alerted when a new weight exceeded a recent documented value (n=13; 100%). The primary vulnerability was age-appropriate plausibility: only (n=9; 69%) flagged weights that were too high for age (absent in n=4; 31%). Several implementations emphasized percent-change warnings or showed expected ranges but did not clearly cite a growth-chart source.
Ordering Stage
This phase showed the most inconsistency with 14 of 16 ordering safeguards failing in at least one site. The EHR system in most sites enabled capturing a medication indication (e.g., empiric meningitis; n=12; 92%). However, the EHR system in none of the tested sites (0%) provided feedback to the ordering provider if the pharmacy could review the medication indication. At the point of order entry: only two sites (15%) displayed the weight, how it was obtained, and when it was obtained at order entry: 10 sites (77%) displayed the total dose; 11 sites (85%) displayed the formula and mathematics to calculate the specific dose (e.g., 15 kg × 10 mg/kg → 150 mg) hidden behind an icon; and no site (0%) identified the scientific source of suggested doses within the ordering window, but required clicking a link to an external source. Eleven sites (85%) supported ordering multi-component medications (e.g., amoxicillin-clavulanic acid) based on a single component. Sites differed in the ability to detect and interrupt critical overdoses: seven sites (54%) alerted when the single dose of a medication exceeded the safe maximum dose for patients; six sites (46%) alerted when the daily maximum dose of a medication was exceeded; 10 sites (77%) when a dose exceeded ten times the recommended safe dose. Only nine sites (69%) required weight documentation before ordering WBD medications.
Administration Stage
All sites (n = 13, 100%) enabled nurses to document medication administration without barcode scanning and after a time lapse between administration and documentation to support medication administration in emergencies.
Discussion
A total of 20 checklist items were tested across 13 healthcare institutions, yielding 260 assessments. Overall, 175 (67.3%) items passed, 83 (31.9%) failed, and 2 (0.8%) were not tested. The EHR design largely supported the WBD phases of pre-ordering (weight documentation) and medication administration. However, there were several safety-critical gaps in the ordering stage when the EHR system permitted placing an order for WBD medication without a weight documentation, and either did not catch several types of potential overdose inputs (e.g., exceeding single dose or daily dose maximum, or exceeding 10 times the safe dose), or permitted the ordering provider to override the alert and place the order. In these cases, ensuring safe and accurate data entry and guarding against EHR errors was incumbent on providers’ vigilance, which is widely recognized as a brittle safeguard to interrupt errors in busy clinical environments. When present, alerts were inconsistent in phrasing, timing, severity, and ability to be overridden, risking alert fatigue and inability to differentiate between alerts based on their clinical risk.
Event Type
Oral Presentations
TimeWednesday, March 259:15am - 9:37am EDT
LocationNassau
Digital Health