Retrospective Validation Of a Patient-Initiated Preconception Screener Against Obstetric Comorbidity Indices To Assess Pregnancy Complications

This retrospective study validates a patient-initiated electronic health record preconception screener (PreMA) by demonstrating that its scores are robustly associated with severe maternal morbidity outcomes, performing comparably to established clinician-facing comorbidity indices across diverse populations.

The Gist

Imagine you are about to go on a long, adventurous road trip. Before you hit the highway, you'd probably want to check your car's engine, tires, and fuel gauge to make sure you won't break down halfway through, right?

In the world of pregnancy, that "pre-trip check" is called preconception care. But for a long time, getting that check-up was like trying to read a complex mechanic's manual written in a foreign language. It was hard for regular people to understand their own health risks before they even started trying to have a baby.

This paper is about a new tool called PreMA (Preconception Medical Assessment) that changes the game. Here is the story of what the researchers did, explained simply:

1. The Problem: The "Secret Manual"

Doctors have long had secret "scorecards" (like the Bateman and Leonard indices) to predict if a pregnancy might get complicated. These scorecards are based on medical records and are great, but they are like mechanic's manuals: only doctors can read them, and they usually only get filled out after you are already pregnant or in the hospital.

Meanwhile, many serious pregnancy problems (like high blood pressure or heart issues) start before the baby is even conceived. We needed a way for regular people to check their own "engine" early, in plain English.

2. The Solution: The "DIY Dashboard"

The researchers created PreMA. Think of this as a simple, user-friendly dashboard that anyone can fill out at home. It asks simple questions about your health history (like "Do you have heart issues?" or "Do you have diabetes?") and turns those answers into a risk score.

The big question was: Does this simple dashboard actually work? Does it predict trouble as well as the complex doctor's manuals?

3. The Experiment: The "Big Data Test Drive"

To find out, the researchers didn't just ask people to fill out the form; they went on a massive digital treasure hunt.

• The Data: They looked at the electronic health records (the digital medical files) of nearly 60,000 women across the University of California system.
• The Method: They took the answers these women would have given on the PreMA form (by looking at their past medical diagnoses) and calculated a "PreMA Score."
• The Comparison: They compared this new PreMA Score against the old, complex doctor's scores (Bateman and Leonard) to see which one was better at predicting Severe Maternal Morbidity (SMM).
  • SMM is like a "flat tire" or "engine failure" during the trip—it means the mother had a life-threatening complication during or right after birth.

4. The Results: The Dashboard Wins!

The results were exciting. The simple PreMA score worked just as well as the complex doctor's manuals.

• The Prediction: If a woman had a higher PreMA score (meaning more health risks), she was significantly more likely to have a severe complication during pregnancy.
• The Fairness: This worked for everyone, regardless of their race, ethnicity, or how much money they made. It was like finding a universal key that unlocked the door to understanding risk for all types of drivers.
• The Details: The study found that heart and blood pressure issues were the biggest "red flags" on the dashboard, followed closely by diabetes.

5. Why This Matters: Empowering the Driver

The most important takeaway is empowerment.

• Before: You had to wait until you saw a doctor, who would then look at your complex file and say, "You might be at risk."
• Now: With PreMA, a person can look at their own health history, fill out a simple form, and get a clear signal: "Hey, your engine needs a tune-up before you start the trip."

This allows people to fix health problems (like managing diabetes or blood pressure) before they get pregnant, potentially preventing the "flat tires" and "engine failures" that lead to severe complications.

The Bottom Line

This study proves that a simple, patient-led checklist is just as powerful as the complex tools doctors use. It's like upgrading from a mechanic's manual to a GPS navigation system that anyone can use. It helps people take control of their health journey before they even leave the driveway, making the road to motherhood safer for everyone.

Technical Summary

1. Problem Statement

Severe Maternal Morbidity (SMM) and mortality remain critical public health challenges in the United States, with significant disparities affecting marginalized communities. While California has successfully reduced maternal mortality through hospital-based interventions, the preconception period remains an underutilized frontier for risk mitigation. Many severe complications (e.g., preeclampsia, hemorrhage) are linked to preexisting conditions that often go unrecognized before conception.

Existing risk assessment tools, such as the Bateman and Leonard indices, are clinician-facing and rely on administrative diagnosis codes recorded during clinical care. There is a lack of validated, patient-initiated screening tools that can empower individuals to identify risks prior to pregnancy and bridge the gap between patient self-awareness and clinical decision-making. The Preconception Medical Assessment (PreMA) was developed as a brief, self-administered tool to address this, but it requires rigorous validation against objective clinical data to ensure its predictive validity.

2. Methodology

The study employed a retrospective observational design using two large, de-identified Electronic Health Record (EHR) datasets conforming to the Observational Medical Outcomes Partnership (OMOP) Common Data Model (CDM):

• UCSF Cohort: Single-center data (UCSF Health + San Francisco Dept. of Public Health), $n \approx 21,427$ deliveries.
• UC-wide Cohort: Multi-center data (6 UC medical centers), $n \approx 37,122$ deliveries (excluding UCSF to avoid overlap).

Data Processing & Variables:

• Inclusion Criteria: Live-birth deliveries for individuals aged 12–55 with at least 10 recorded clinical events (conditions, observations, procedures) in the 280 days prior to conception.
• Risk Scores:
  • PreMA: A patient-initiated screener mapped to ICD-10/CPT codes (0–8 range).
  • Bateman & Leonard: Established clinician-facing comorbidity indices (0–51 and 0–281 ranges, respectively).
  • All scores were standardized to z-scores (mean=0, SD=1) for comparability.
• Outcomes:
  • Severe Maternal Morbidity (SMM): CDC indicators occurring within 7 days of delivery.
  • Non-Transfusion SMM (NT-SMM): SMM excluding cases defined solely by blood transfusion.
• Covariates: Maternal age (modeled via splines), race, ethnicity, and Area Deprivation Index (ADI) for socioeconomic stratification.

Statistical Analysis:

• Models: Logistic regression (for Odds Ratios) and Poisson regression with robust standard errors (for Relative Risks).
• Stratification: Analyses were stratified by race, ethnicity, and ADI quintiles.
• Item-Level Analysis: Adjusted Poisson regression was used to evaluate the association of individual PreMA domains (e.g., cardiovascular, diabetes) with outcomes.
• Correction: Benjamini-Hochberg procedure applied for multiple testing ($q=0.05$).

3. Key Contributions

• Validation of Patient-Initiated Screening: This is the first study to retrospectively validate the PreMA score against established clinician-facing indices using real-world EHR data, demonstrating that patient-reported risk domains align closely with objective clinical comorbidity burdens.
• Generalizability: The study confirms the robustness of the PreMA score across diverse datasets (single-center vs. multi-state), racial/ethnic groups, and socioeconomic strata (ADI quintiles).
• Granular Domain Analysis: The research moves beyond aggregate scores to identify specific PreMA domains (e.g., cardiovascular conditions) that drive the strongest associations with adverse outcomes, offering insights for targeted clinical interventions.
• Equity Focus: The study explicitly evaluates performance in marginalized populations, showing that PreMA remains a valid predictor even in high-deprivation neighborhoods where structural barriers to care exist.

4. Key Results

• Predictive Performance: Higher standardized PreMA scores were consistently associated with increased risk of SMM and NT-SMM.
  • Effect Sizes: Per standard deviation (SD) increase in PreMA score yielded Relative Risks (RR) of 1.23 (95% CI 1.19–1.28) for SMM in the UCSF cohort.
  • Comparison: The PreMA score demonstrated effect sizes comparable to the Bateman (RR 1.26) and Leonard (RR 1.26) indices, despite PreMA being patient-initiated and the others being clinician-derived.
• Stratified Findings:
  • Race/Ethnicity: Associations persisted across all groups. Effect estimates were notably higher for American Indian/Alaska Native (RR ~2.03), Black/African American (RR ~1.33), and Native Hawaiian/Pacific Islander patients compared to White patients.
  • Socioeconomic Status: Risk amplification was observed in higher-deprivation neighborhoods. In the highest ADI quintile, the PreMA score was associated with an RR of 1.31 for SMM, compared to 1.11 in the lowest quintile.
• Domain-Specific Drivers:
  • Cardiovascular conditions showed the strongest association with NT-SMM (RR 3.08).
  • Diabetes also showed strong associations (RR 1.83).
  • Other domains like lung disease and GI surgery showed more modest or non-significant associations after adjustment.

5. Significance and Implications

• Scalable Public Health Tool: The findings support PreMA as a valid, scalable, and patient-centered tool for early risk identification. Because it does not require clinician administration, it can be deployed widely to engage individuals before they enter the healthcare system.
• Clinical Workflow Integration: An EHR-derived PreMA score could be integrated into clinical workflows to flag high-risk patients for enhanced prenatal monitoring, referral to preconception care, or targeted counseling.
• Addressing Disparities: The tool's ability to predict risk across diverse socioeconomic and racial groups suggests it could help mitigate disparities by identifying high-risk individuals in marginalized communities who might otherwise be missed by traditional, clinic-dependent screening.
• Future Directions: While the study validates the association of PreMA with outcomes, future work is needed to assess its predictive performance (discrimination and calibration) in prospective implementation studies and to explore weighted versions of the tool that prioritize high-yield domains like cardiovascular health.

Limitations: The study relies on EHR proxies for patient-reported data, which may underrepresent conditions in individuals with limited healthcare access. Additionally, the data is primarily from California, limiting generalizability to non-urban or non-California populations.