Scholarlink

Background <p>Bronchopulmonary dysplasia (BPD) remains a major complication of very preterm birth and very low birth weight. We aimed to develop and internally validate a pragmatic prediction model for BPD using routinely available clinical variables obtained before outcome ascertainment.</p> Methods <p>In this multicenter retrospective study, infants with gestational age &lt; 32&#xa0;weeks and/or birth weight &lt; 1500 g admitted to three tertiary neonatal intensive care units in China between January 2023 and December 2024 were included. BPD was defined as the need for respiratory support at 36 + 0/7&#xa0;weeks’ postmenstrual age. Candidate clinical variables included perinatal characteristics, first-24-h laboratory and blood gas variables, and major morbidities diagnosed before 36 weeks’ postmenstrual age. Variables showing between-group differences in baseline comparisons were considered for least absolute shrinkage and selection operator regression after excluding variables overlapping with the outcome definition or representing downstream clinical course. Predictors selected by least absolute shrinkage and selection operator regression were entered into a multivariable logistic regression model to construct a nomogram. Model performance was assessed by discrimination, calibration, overall accuracy, and decision-analytic performance, with bootstrap internal validation using 1000 resamples.</p> Results <p>A total of 796 infants were included, of whom 345 (43.34%) developed BPD. The final model retained 10 predictors: birth weight, gestational age, 5-min Apgar score, duration of premature rupture of membranes, Ureaplasma urealyticum infection, neonatal respiratory distress syndrome, early-onset sepsis, late-onset sepsis, purulent meningitis, and assisted reproductive technology. The model showed good discrimination [C-index 0.848, 95% confidence interval (CI) 0.821–0.874], a Brier score of 0.158, and acceptable apparent calibration (slope 1.0). Decision curve analysis suggested potential net benefit across a range of threshold probabilities. In a sensitivity analysis excluding late-onset sepsis and purulent meningitis, the supplementary model retained acceptable discrimination (AUC 0.8283, 95% CI 0.7999–0.8567).</p> Conclusions <p>We developed an internally validated hospitalization-based nomogram for estimating BPD risk in very preterm and/or very low birth weight infants. The model demonstrated good discrimination, acceptable calibration, and potential decision-analytic value. External validation and prospective impact studies are required before routine clinical implementation.</p> <p><i>Clinical Trial</i> Retrospective Observational Study.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Predicting bronchopulmonary dysplasia before 36 weeks’ PMA using routine NICU data: a multicenter model and nomogram

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月5日

查看全文
  • Xiaowei Sun,
  • Rui Jing,
  • Jialin Wen,
  • Wenying Meng,
  • Qianqian Jiang,
  • Qingqing Li,
  • Yang Li

摘要

Background

Bronchopulmonary dysplasia (BPD) remains a major complication of very preterm birth and very low birth weight. We aimed to develop and internally validate a pragmatic prediction model for BPD using routinely available clinical variables obtained before outcome ascertainment.

Methods

In this multicenter retrospective study, infants with gestational age < 32 weeks and/or birth weight < 1500 g admitted to three tertiary neonatal intensive care units in China between January 2023 and December 2024 were included. BPD was defined as the need for respiratory support at 36 + 0/7 weeks’ postmenstrual age. Candidate clinical variables included perinatal characteristics, first-24-h laboratory and blood gas variables, and major morbidities diagnosed before 36 weeks’ postmenstrual age. Variables showing between-group differences in baseline comparisons were considered for least absolute shrinkage and selection operator regression after excluding variables overlapping with the outcome definition or representing downstream clinical course. Predictors selected by least absolute shrinkage and selection operator regression were entered into a multivariable logistic regression model to construct a nomogram. Model performance was assessed by discrimination, calibration, overall accuracy, and decision-analytic performance, with bootstrap internal validation using 1000 resamples.

Results

A total of 796 infants were included, of whom 345 (43.34%) developed BPD. The final model retained 10 predictors: birth weight, gestational age, 5-min Apgar score, duration of premature rupture of membranes, Ureaplasma urealyticum infection, neonatal respiratory distress syndrome, early-onset sepsis, late-onset sepsis, purulent meningitis, and assisted reproductive technology. The model showed good discrimination [C-index 0.848, 95% confidence interval (CI) 0.821–0.874], a Brier score of 0.158, and acceptable apparent calibration (slope 1.0). Decision curve analysis suggested potential net benefit across a range of threshold probabilities. In a sensitivity analysis excluding late-onset sepsis and purulent meningitis, the supplementary model retained acceptable discrimination (AUC 0.8283, 95% CI 0.7999–0.8567).

Conclusions

We developed an internally validated hospitalization-based nomogram for estimating BPD risk in very preterm and/or very low birth weight infants. The model demonstrated good discrimination, acceptable calibration, and potential decision-analytic value. External validation and prospective impact studies are required before routine clinical implementation.

Clinical Trial Retrospective Observational Study.