Meta-modeling and validation of a risk prediction model for intradialytic hypotension in maintenance hemodialysis patients
Objective Meta modelling was employed to develop a risk prediction model for intradialytic hypotension(IDH)and validate the model.Methods Literature on risk factors for IDH published up to March 31,2023 was retrieved from 8 databases,including Cochrane Library,PubMed,Web of Science,EBSCO,Scopus,CINAHL,CNKI and Wanfang Database.Random effects model was used to combine ORs,and factors with P＜0.05 were selected to establish the model based on their regression coefficients.286 maintenance hemodialysis patients were selected as a validation cohort to evaluate the model's discrimination,calibration and clinical utility.Results 39 studies were included,involving 25 546 patients.14 factors were identified to establish the risk prediction model.The risk score for IDH occurrence was calculated as-0.301 × male+0.015 × age+0.004 × dialysis vintage+0.988 × diabetes+0.730 × cardiovascular disease-0.042 × predialysis systolic blood pressure+0.666 × dialysis mode of hemodialysis filter+0.076 × temperature+0.159 × ultrafiltration rate+0.476 × ultrafiltration volume+1.024 × weight gain between dialysis+0.053 × serum phosphorus+0.023 × blood urea nitrogen+0.040 ×β2-microglobulin.Definition in the Kidney Disease Outcomes Quality Initiative guideline,nadir intradialytic systolic blood pressure＜90 mmHg(1 mmHg=0.133 kPa),falling intradialytic systolic blood pressure ≥20 mmHg,and definition in the United Kingdom Kidney Association guideline were selected as 4 outcomes.The areas under the curve for the prediction model with respect to these 4 outcomes were 0.830,0.648,0.647,and 0.763,respectively.Calibration curves showed that the model predictive outcomes were consistent with actual outcomes for the first 2 outcomes(x2=14.824,P=0.064;x2=12.016,P=0.149).Decision curve analysis indicated that the model had better net benefit compared to either intervention/no intervention for all definitions.Conclusion The IDH risk prediction model developed by meta-modeling in this study has good predictive performance and certain application value.
HemodialysisHypotensionPrediction ModelMeta-AnalysisNursing Care
万方数据
维普
