Delayed EHR availability of vital signs is associated with nursing shift structure Authors: Patrick G Lyons, Brenna Park-Egan, Tanuj Devara, Jeffrey A Gold Published: 2026 ABSTRACT AbstractDelayed EHR availability of vital signs may jeopardize patient safety and data integrity. In 11.4 million measurements from 26,479 encounters, median measurement-to-validation lag ranged 12–25 minutes across locations; one-third of abnormal values were delayed over an hour. Lag peaked before nursing shift changes, including in settings with device-EHR integration. These findings implicate workforce-driven workflow, rather than technical failure, as a primary barrier to real-time surveillance and health system responsiveness.Subject terms: Engineering, Health care, Medical research Introduction In acute care settings, recording patient vital sign measurements in the electronic health record (EHR) enables timely medical decision-making, automated clinical decision support, and quality benchmarking1,2. Although vital signs are intended to represent a patient’s physiology at the time of measurement, the timeliness of their availability in the EHR depends on manual data entry or device transmission followed by nurse validation. As a result, the time at which a vital sign becomes available in the EHR may lag behind the time at which it was measured3. Such delays may hinder escalation of care, compromise the responsiveness of surveillance systems, and introduce temporal ambiguity into retrospective analyses – consequences that may be especially pronounced when vital signs are abnormal. A recent study demonstrated the existence of these delays3, but neither their determinants nor the populations most affected have been systematically characterized. We aimed to quantify the lag time between vital sign measurement and EHR validation across the acute care spectrum, hypothesizing that lag times would differ based on vital sign abnormality and nursing shift structure. Measurement-to-validation lag across acute care We retrospectively evaluated 26,479 consecutive adult patient encounters (20,071 patients; Table 1) in the emergency department (ED), inpatient wards, or intensive care units (ICUs) at a 500-bed urban, academic, quaternary care hospital between January 1, 2023, and June 30, 2024. From these encounters, we evaluated 11,376,701 unique vital sign measurements taken at 3,215,702 different moments (Table 2). Vital signs were validated in the EHR at 2,408,220 distinct time points, including 362,937 instances in which several measurements of a single vital sign, taken at different times, were “batch” validated simultaneously. Table 1Patient characteristicsCharacteristicUnique patients, n = 20,071Age at first encounter, years, median (IQR)62 (46–73)Female, n (%)9153 (46%)Race, n (%) American Indian or Alaska Native541 (2.7%) Asian562 (2.8%) Black or African American663 (3.3%) Native Hawaiian or Other Pacific Islander198 (1.0%) White16,913 (84%) Other/Unknown1194 (5.9%)Ethnicity, n (%) Hispanic or Latino1745 (8.7%) Not Hispanic or Latino17,186 (86%) Other/Unknown1140 (5.7%) Patient characteristics Table 2Measurement times, validation times, and values of routine vital signs from the electronic health recordHR, n = 2,875,781SBP, n = 1,882,139Respirations, n = 2,511,042SpO2, n = 2,693,860Temperature, n = 1,413,879Normal range60–120 beats/min90–180 mmHg10–20 breaths/min>= 90%96.0–100.3 FAbnormal ValuesNormal ValuesAbnormal ValuesNormal ValuesAbnormal ValuesNormal ValuesAbnormal ValuesNormal ValuesAbnormal ValuesNormal ValuesValue, median (IQR)59 (55–128)85 (74–97)86 (81–89)123 (109–138)24 (22–27)16 (15–18)87 (84-89)97 (95–99)100.8 (100.4–101.3)98.2 (97.7–98.7)Unique measurements, n (%)296,655 (10%)2,579,126 (90%)116,231 (6.2%)1,765,908 (94%)633,581 (25%)1,877,461 (75%)65,548 (2.4%)2,628,312 (98%)83,413 (5.9%)1,330,466 (94%)Unique validation times, n (%)*191,553 (10%)1,720,254 (90%)91,747 (5.8%)1,498,002 (94%)372,069 (21%)1,362,119 (79%)45,401 (2.5%)1,753,530 (97%)52,072 (4.4%)1,144,221 (96%)Batched measurements, n (%)*304,381 (16%)178,633 (11%)226,245 (13%)264,150 (15%)82,678 (6.9%)Measurements per batch, med (IQR)2 (2–3)2 (2–3)2 (2–3)2 (2–3)2 (2–3)Lag minutes, med (IQR)**21 (5–76)20 (4–72)11 (3–36)14 (3–47)28 (7–93)20 (5–72)22 (5–85)21 (5–76)20 (5–62)14 (3–52)Lag time > 30 minutes, n (%)126,664 (43%)1,074,578 (42%)33,089 (28%)596,650 (34%)305,724 (48%)783,554 (42%)28,880 (44%)1,122,585 (43%)33,875 (41%)467,735 (35%)Lag time > 60 minutes, n (%)86,428 (29%)727,888 (28%)18,331 (16%)354,595 (20%)211,517 (33%)531,871 (28%)20,391 (31%)768,786 (29%)21,359 (26%)296,326 (22%)Lag time > 90 minutes, n (%)65,023 (22%)540,370 (21%)11,325 (9.7%)229,904 (13%)160,980 (25%)394,861 (21%)15,658 (24%)575,607 (22%)15,356 (18%)206,997 (16%)HR heart rate, SBP systolic blood pressure, IQR interquartile range.*Denominators for validation time: HR, 1,911,807; SBP, 1,589,749; Respirations 1,734,188; SpO2 1,798,931; Temperature 1,196,293. Normal/abnormal numerators do not sum to these denominators because some validation events contain both normal and abnormal values for a vital sign.**p-values for normal/abnormal comparisons of the primary outcome are <0.001 by Wilcoxon rank-sum tests for all vital sign measures. Measurement times, validation times, and values of routine vital signs from the electronic health record HR heart rate, SBP systolic blood pressure, IQR interquartile range. *Denominators for validation time: HR, 1,911,807; SBP, 1,589,749; Respirations 1,734,188; SpO2 1,798,931; Temperature 1,196,293. Normal/abnormal numerators do not sum to these denominators because some validation events contain both normal and abnormal values for a vital sign. **p-values for normal/abnormal comparisons of the primary outcome are <0.001 by Wilcoxon rank-sum tests for all vital sign measures. The study’s primary outcome was measurement-to-validation lag, defined as the difference (in minutes) between the EHR-documented measurement time and validation time for each vital sign. Because unvalidated values are not permanently retained in our EHR, validation time marks the first retained EHR timestamp at which a value became available in the record; we therefore interpret measurement-to-validation lag as an EHR-availability lag. Measurement-to-validation lag varied by vital sign, with respiratory rate having the longest lag times (22 minutes; interquartile range [IQR] 5–77) followed by SpO2 (21 minutes [5–76]) and heart rate (20 minutes [5–72]); lag was shorter for temperature (14 minutes [3–53]) and blood pressure (13 minutes [3–46]; p < 0.001). Measurement-to-validation lag also varied by clinical location, with longer lag times in the ICU (median 25 minutes [7–71]; n = 5,528,405) than in the ED (16 minutes [3–73]; n = 1,202,994) or ward settings (12 minutes [3–57]; n = 4,645,302; p < 0.001). Although lag was shortest on the wards, 24% (n = 1,115,303) of observations had delays greater than 60 minutes. Lag and vital sign abnormality Hypothesizing that vital sign perturbations would spur timelier documentation and validation, we compared lag times between normal and abnormal values. A total of 1,195,428 vital signs (11%) had abnormal values. The relationship between abnormal values and lag varied by vital sign (Fig. 1). Abnormal systolic blood pressures were validated more quickly than normal values (median 11 minutes [3–36] vs 14 minutes [3–47]), while abnormal respiratory rates and temperatures had longer lag times than their normal counterparts (28 minutes [7–93] vs 20 minutes [5–72]; 20 minutes [5–62] vs 14 minutes [3–52]). Regardless of vital sign type, 44% (n = 528,232) of abnormal values still had delays greater than 30 minutes and 30% (n = 358,026) had delays exceeding 60 minutes. Cumulative distribution of measurement-to-validation lag by vital sign type and abnormal status. Curves represent the proportion of measurements not yet available in the EHR at each time point following measurement. Darker lines indicate abnormal values; lighter lines indicate values within normal range. Abnormal systolic blood pressures (SBP) were validated faster than normal values, while abnormal respiratory rates (RR) and temperatures were validated more slowly. Regardless of vital sign type, substantial proportions of abnormal measurements remained unavailable in the EHR through 120 minutes. Lag and proximity to nursing shift change We also evaluated the relationship between time of day and measurement-to-validation lag. We anticipated longer lag times near shift changes, reasoning that end-of-shift workflow patterns—such as retrospective charting or validation of earlier measurements—might delay EHR availability. At our hospital, ICU and ward shift changes occur at 7 AM and 7 PM, whereas ED shifts are variable. In the ICU and on the wards, lag time varied bimodally over the course of the day, peaking within the 2 hours prior to shift change (Fig. 2). After adjusting for location, shift period, vital sign type, and abnormal values, mean lag increased by 5.2 minutes (95% confidence interval 5.13-5.30) for every hour closer to shift change. Relationship between time of day and measurement-to-validation lag. Mean measurement-to-validation lag (y-axis) varied by the hour of validation (x-axis) for all clinical vital signs. Lag peaked just prior to nursing shift change (7:00 AM/PM) on the inpatient wards and in the intensive care unit (ICU). In the emergency department (ED), where nursing shifts vary, this pattern did not hold. Interpretation and implications In summary, we found substantial lag between vital sign measurement and EHR availability across the spectrum of acute care, including long delays for abnormal measurements. These findings have implications for both patient safety and health system performance; measurement-to-validation lags introduce systematic temporal noise — discrepancies between when physiology changes and when the EHR reflects it — into the EHR, which serves as the foundational data layer for hospital operational analytics and AI-driven predictive models. First, delayed vital sign availability could impede timely recognition of physiological instability by clinicians while also directly impacting the timeliness of EHR-based early warning systems. Second, accurate vital sign timestamps are essential for the validity of operational analytics and other downstream processes (e.g., sepsis bundle compliance)4,5. If the interval between measurement and validation is substantial or varies systematically, the resulting data could misrepresent physiological trends and skew conclusions about patient status or the performance of monitoring infrastructure. We observed that for all vital signs, abnormal values were associated with shorter measurement-to-validation times than values within the normal range. This finding may be unsurprising in light of previously described documentation patterns associated with patients’ clinical conditions (e.g., more frequent measurements often reflect nurses’ clinical concern for impending deterioration)6. However, nearly one in three abnormal measurements had a lag greater than one hour, underscoring that clinical concern alone is likely insufficient for driving timely vital sign validation. Moreover, lag times were significantly associated with chronological proximity to nursing shift change even after adjustment for important confounders, including the clinical setting and factors intrinsic to the vital signs themselves. This finding suggests that workforce structure and clinical workflows—rather than technical failures alone—may be the primary drivers of these delays, which is in keeping with prior descriptions of other documentation “workarounds” in service of workflow efficiency7,8. Relatedly, differential lag patterns between day and night shifts may be related to task load, competing priorities, nurse characteristics, or some combination of these. Importantly, increasing lag times near nursing shift change in the ICU and on the wards suggests that direct device integration (predominant in the ICU) is insufficient for solving this problem9,10, because integrated values still require manual validation before they become available. Addressing these delays may therefore require organizational interventions—such as automatic validation of data from EHR-integrated devices—that reduce reliance on end-of-shift charting practices. An important limitation is that the available EHR timestamps do not allow us to decompose measurement-to-validation lag into its constituent workflow steps. Depending on clinical location and device integration, this interval may reflect delays in manual entry, device transmission, nurse review, validation, or retrospective release of previously measured values. Accordingly, our findings should be interpreted as delays in EHR availability rather than as a direct measure of documentation delay alone. Second, our study involved a single academic hospital; however, the ubiquity of EHRs and their well-documented effects on clinical workflows and other sociotechnical systems make it likely that our findings would generalize to other hospitals11,12. Third, delayed EHR availability may sometimes paradoxically reflect appropriate clinical prioritization (i.e., focusing on direct patient care when time-sensitive bedside demands compete with charting responsibilities)12. Even in these circumstances, though, delayed availability could negatively impact data integrity for purposes other than immediate clinical care. Notwithstanding these limitations, our findings indicate the need to study and improve the timeliness and integrity of vital sign data in the EHR. More broadly, these results illustrate how the reliability of the EHR and its associated technologies—including the predictive algorithms, clinical decision support, and integrated AI tools that depend on timely data —rests not only on technical infrastructure but also on the organizational and workforce contexts in which they operate. Whereas prior theory-based work has delineated determinants of documentation completeness13, our findings characterize a complementary dimension of data quality: the interval between documented measurement time and the point at which vital sign values become validated and available in the EHR. Methods Study design, setting, and patients We retrospectively evaluated consecutive adult patients presenting to the emergency department (ED), inpatient wards, or intensive care units (ICUs) at Oregon Health & Science University Hospital (OHSU, an urban, academic, quaternary care hospital with approximately 500 adult acute care beds) between January 1, 2023, and June 30, 2024. We included all encounters with at least one measured systolic blood pressure, heart rate, respiratory rate, and temperature. We excluded observations in the operating rooms, inpatient dialysis and procedural units, and labor, delivery, and postpartum units because we expected clinical and documentation workflows to differ substantially from acute care areas. Data From these encounters, we evaluated all vital signs measured between hospital presentation and discharge (i.e., excluding measurements charted as “pre-admission” or “post-discharge”). We excluded physiologically implausible recordings (validation times prior to measurement times, heart rates > 300 beats/min, respiratory rates > 60 breaths/min, temperatures > 111 F, and pulse oximetry values > 100), as these were likely to be charting errors. Because systolic and diastolic blood pressures are almost always recorded concurrently, we considered only systolic values in the analysis. The primary outcome was measurement-to-validation lag, defined as the difference, in minutes, between the EHR-documented measurement time (i.e., the time at which the vital sign was taken from a patient) and validation time (i.e., the time at which the vital sign was made available in the EHR) for each clinical vital sign. Vital sign recording workflows differ by clinical location at OHSU. On the wards, vital signs are recorded and validated manually. In the ICU and ED, vital signs are recorded by direct device-EHR integration followed by manual validation by the nurse. Unvalidated values are not retained in our EHR data; we therefore interpret measurement-to-validation lag as an EHR-availability lag. For observations validated simultaneously, measurement-to-validation lag was calculated separately for each vital sign measurement using its own EHR-documented measurement time and the shared validation time. Thus, multiple measurements validated in a single batch could have different lag values if their documented measurement times differed. Analysis A priori, we analyzed two exposures. First, because vital sign perturbations might spur prompt documentation and validation, we compared lag times between normal and abnormal vital sign values using Wilcoxon-Mann-Whitney tests. We considered measurements to be abnormal if they fell outside prespecified standard ranges: heart rate < 60 or > 120, systolic blood pressure < 90 or > 180, respiratory rate < 10 or > 20, SpO2 < 90, and temperature < 96.0 or > 100.314. Second, we evaluated the relationship between time of day and measurement-to-validation lag, hypothesizing that validation times closer to nursing shift change would have longer lag times (i.e., because nurses might retroactively document and/or validate previously-measured vital signs at the end of a shift). ICU and ward shift changes occur at 7 AM and 7 PM at OHSU, whereas ED shifts are variable. To estimate the relationship between time relative to shift change and the primary outcome – a right-skewed positive number – we fit a generalized estimating equation with a gamma distribution, a log-link, and clustering by hospital encounter to account for correlation among repeated vital sign measurements within the same hospitalization15,16. We specified an exchangeable correlation structure rather than an autoregressive approach due to computational constraints with the large number of repeated measurements per encounter. This model adjusted for location (categorical: ED, wards, ICU), day vs night shift (dichotomous), the vital sign (categorical: heart rate, systolic blood pressure, respiratory rate, SpO2, or temperature), and whether the measurement was abnormal (dichotomous, based on the values described above). We set two-sided statistical significance at p = 0.05. We used R version 4.4 (R Project for Statistical Computing) and the arrow, collapse, tidytable, tidyverse, and gtsummary packages for all analyses. OHSU’s institutional review board approved this study (#00025188), waiving the requirement for informed consent because the use of routinely collected clinical data posed minimal risk and obtaining individual consent was impracticable. All data were handled in accordance with institutional governance policies, HIPAA regulations, and the principles of the Declaration of Helsinki. We followed the STROBE guidelines for reporting observational research (Supplementary Information)17. Supplementary information STROBE-checklist-v4-cohort. Acknowledgements This work was supported by the National Cancer Institute of the National Institutes of Health under award number K08CA270383 (P.G.L.). The funder had no role in the design or conduct of the study; the collection, management, analysis, or interpretation of the data; the preparation, review, or approval of the manuscript; or the decision to submit the manuscript for publication. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. Author contributions Concept and design: P.G.L., J.A.G. Data acquisition, analysis, or interpretation: P.G.L., T.D., B.P.E. Drafting of the manuscript: P.G.L. Critical revision of the manuscript for important intellectual content: P.G.L., T.D., B.P.E., J.A.G. Supervision: J.A.G. Data availability The data underlying this study contain protected health information from electronic health records and cannot be shared publicly due to patient privacy regulations and institutional policy. De-identified, aggregate summary statistics may be available from the corresponding author upon reasonable request and with appropriate institutional review board approval. Code availability The code used for the analyses in this study is publicly available at https://github.com/p-lyons/vital-sign-lags. Declarations Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Supplementary information The online version contains supplementary material available at https://doi.org/10.1038/s44401-026-00151-7. References Associated Data Supplementary Materials STROBE-checklist-v4-cohort. Data Availability Statement The data underlying this study contain protected health information from electronic health records and cannot be shared publicly due to patient privacy regulations and institutional policy. De-identified, aggregate summary statistics may be available from the corresponding author upon reasonable request and with appropriate institutional review board approval. The code used for the analyses in this study is publicly available at https://github.com/p-lyons/vital-sign-lags.