Balancing Model Performance and Simplicity to Predict Postoperative Primary Care Blood Pressure Elevation.

Area of Science:

• Cardiology
• Anesthesiology
• Public Health

Background:

• Anesthesiologists face uncertainty in managing perioperative blood pressure and lack consensus on postoperative hypertension follow-up.
• Traditional views limit the anesthesiologist's role in long-term patient care.
• A predictive model could guide referrals and integrate surgical care with public health for hypertension management.

Purpose of the Study:

• To determine the prevalence of poorly controlled blood pressure after surgery.
• To develop and assess predictive models for identifying patients needing postoperative blood pressure follow-up.
• To evaluate the performance of simple blood pressure thresholds for referral decisions.

Main Methods:

• Analysis of national data from US veterans undergoing surgical care.
• Assessment of four logistic regression models with increasing complexity to predict postoperative uncontrolled blood pressure (≥140/90 mm Hg).
• Internal validation, C-statistics, Net Reclassification Index, and calculation of referral threshold performance characteristics.

Main Results:

• 25.7% of 215,621 surgical patients had poorly controlled postoperative blood pressure.
• A model using only 2 preoperative blood pressure readings performed nearly as well as more complex models (C-statistic 0.721 vs 0.736).
• A preoperative threshold of ≥150/95 mm Hg identified patients likely to have hypertensive readings post-surgery, with a negative predictive value of 79.6%.

Conclusions:

• Poorly controlled postoperative blood pressure is common in surgical patients.
• Predictive models using preoperative blood pressure measurements offer acceptable performance for guiding referral decisions.
• Further research is needed on the feasibility and effectiveness of postoperative referrals for long-term cardiovascular health.