Retrospective study finds heterogeneity in recorded postoperative symptoms across assessors

A five-year retrospective observational study of 121,755 routinely recorded postoperative patient-controlled analgesia follow-up assessments found substantial variation in symptom fields among assessors. The analysis included 45,398 surgical patients assessed between January 2020 and April 2025, and it found that adjusted models could not fully identify the sources of that variation. The results do not support ranking nurses or treating recorded rates as direct estimates of true symptom incidence.

Researchers removed 5,519 excess completely identical rows before analysis. They retained non-identical rows recorded for the same patient at the same time and disclosed this decision. American Society of Anesthesiologists physical status was linked only when both the patient and exact encounter or start time matched. The primary cohort comprised 120,680 assessments from 45,177 patients and 15 assessors. Logistic mixed models from M0 through M4 were used to account for repeated patients and to incorporate year, ward and available case-mix information.

Corrected assessor-specific rates ranged from 0.14% to 13.19% for a nausea or vomiting code of at least 1, from 0.10% to 16.43% for a dizziness code of at least 1, and from 0% to 7.78% for a documented sedation score of at least 1. Assessor-only variance partition coefficients were 28.1% for nausea or vomiting, 42.5% for dizziness and 67.6% for sedation. These figures describe variation in the corrected recorded rates, rather than causal effects of assessor behavior or differences in patients’ clinical risk.

The adjusted analysis was incompletely estimable. Of 14 primary-cohort models with numeric result records, nine met the prespecified numerical acceptance criteria and five were not estimable; three accepted fits were singular. For M3, two models produced median odds ratios of 1.00 and 1.02 but were singular, while the third was not estimable. Sedation M4 remained not estimable after the prespecified fallback attempts. A post hoc amendment approved by the corresponding author preceded the remaining restricted-cohort models, but those six models were not run; the restricted cohorts were instead summarized descriptively with Wilson 95% confidence intervals. The authors concluded that prospective validation is required before these routinely recorded fields are used for benchmarking.

Subject of Research: Variability in routinely recorded postoperative symptom fields among nursing assessors

Article Title: Between-assessor and temporal heterogeneity in routinely recorded postoperative symptom fields: a five-year retrospective observational study of 121,755 nursing assessments

Article References: Ge, J., Chen, S., Xie, F., Zhang, Q., Cao, N., Zhang, S., Wang, T., Ding, R., Li, Z., Cao, X., & Kong, Q. (2026). Between-assessor and temporal heterogeneity in routinely recorded postoperative symptom fields: a five-year retrospective observational study of 121,755 nursing assessments. BMC Nursing. https://doi.org/10.1186/s12912-026-05380-7

Image Credits: AI Generated

DOI: 10.1186/s12912-026-05380-7

Keywords: postoperative symptoms, patient-controlled analgesia, nursing assessments, assessor variability, nausea and vomiting, dizziness, sedation, retrospective cohort study, mixed-effects models, benchmarking, Between-assessor, temporal