How one provider org used SDOH, analytics and texting to help reduce pre-term births

The Parkland Center for Clinical Innovation increased prenatal visit attendance by 24%, reduced early preterm delivery by 27%, and reduced first-year baby costs by 54%.
By Bill Siwicki
11:01 AM
How one provider org used SDOH, analytics and texting to help reduce pre-term births

Parkland Hospital in Dallas, Texas.

The Parkland Center for Clinical Innovation aims to improve health for underserved, vulnerable populations. Its staff says health begins where people live, learn, work and play, so they endeavor to bring together data science and social determinants of health to create connected communities of care.

THE PROBLEM

“One of the vulnerable populations we’ve been focusing on are women at risk of pre-term delivery since their health and pregnancy is heavily influenced not only by their medical history, but their social and economic situation and environment,” said Dr. Steve Miff, president and CEO of the Parkland Center for Clinical Innovation. “Our data scientists and clinicians took a fresh look at this challenge and explored how data science and SDOH can drive better segmentation, engagement, and ultimately better outcomes and lower costs for those who need help the most.”

As top healthcare problems go, pre-term birth is a big one because one in 10 children in the United States is born prematurely. This challenge is even more significant among inner city African-American and Hispanic women due to SDOH challenges.

“Preterm births are associated with a high societal cost and burden in the short- and long-term, and the costs and burden increase exponentially as prematurity is severe,” Miff explained. “For instance, delaying a preterm delivery by only 4 weeks may decrease short-term costs per patient by as much as $80,000 in the first year of life.”

"More than 800 at-risk patients have received text messaging interventions and more than 75% of patients report satisfaction with the program."

Dr. Steve Miff, Parkland Center for Clinical Innovation

Maternal stress, tobacco, alcohol and substance use, infections, and a history of preterm delivery are known to be associated with increased risk for preterm delivery, he added.

“Since preterm births disproportionately affect African Americans, Latinos and socio-economically disadvantaged populations, addressing social needs, providing social support and enhancing prenatal care attendance can help curb the risk for preterm birth among these disadvantaged populations,” Miff stated. “Risk-driven patient education and clinical decision support targeted and tailored to patient risk profiles can drive patient engagement, prenatal visit attendance and timely evidence-based interventions.”

PROPOSAL

There was a pioneering study and paper published in the American College of Obstetricians and Gynecologists by Drs. Anderson, Bloom and colleagues that concluded that pre-term birth significantly can be decreased in minority women delivered at an inner-city hospital through increased pre-natal care via targeted public healthcare programs, Miff recalled.

“The challenge we embarked on was to use data science and SDOH to risk stratify pregnancies and drive increases in pre-natal care via direct patient engagement using digital technology and coordinated case management outreach,” he explained.

The Parkland Center for Clinical Innovation Preterm Birth Prevention Program, led at PCCI by Senior Medical Director Yolande Pengetnze, is a comprehensive program combining: 1) Accurate risk prediction, 2) Provider notification, 3) Risk-driven, tailored patient education via digital technology, and 4) Workflow redesign to improve birth outcomes and reduce the rate of preterm birth.

The Parkland Center for Clinical Innovation Preterm Birth Risk Prediction Model leverages machine learning and multiple data sources, including claims, eligibility, EHR and community data. The model is unique in its ability to incorporate demographic, clinical and socioeconomic data from multiple sources to predict the risk for preterm delivery among pregnant women at any point during the pregnancy.

The model is three to four times more accurate than clinical standards for identifying pregnant women at risk for preterm birth, including first-time mothers, Miff stated.

“Using our predictive risk model to drive provider notification, point-of-care workflow redesign, case management outreach, and patient education, leads to improved patient engagement, positive health behavior change, increased prenatal care attendance and timely implementation of evidence-based care at the point-of-care, and thus better birth outcomes,” he said.

MEETING THE CHALLENGE

Parkland Center for Clinical Innovation staff worked with Dr. Barry Lachman’s team at the Parkland Community Health Plan, a provider-owned Texas Medicaid HMO owned by the Parkland Health and Hospital Systems.

“We used Parkland Community Health Plan claims and eligibility data along with block-level community data for SDOH metrics to develop the Parkland Center for Clinical Innovation Preterm Birth Risk Prediction Model,” Miff explained. “A key finding was that in order to meaningfully use and integrate SDOH data, it needs to be modeled at the block level and not just the Zip code level.”

There is too much variation within Zip codes for the data to be meaningful at that level, but when analyzed and modeled at the block level, SDOH features are in the top three overall predictive variables, he added.

“We used the risk-stratified scores to build insightful patient- and provider-level reports, including patient-specific risk scores and risk factors and provider-level outcomes reports,” he said. “The reports are presented as a Case Management list with practical and actionable information to drive seamless intervention. The reports are updated monthly and sent to clinical providers in the Parkland Community Health Plan Providers Network.”

Additionally, staff designed an educational text messaging program tailored in content and intensity to patients’ risk profiles, such that lower risk patients receive less frequent text messages whereas higher risk patients receive more frequent and content-specific interventions.

RESULTS

In the first year of intervention, more than 21,000 unique pregnancies have been prospectively risk-stratified, with about 7,000 pregnancies risk-stratified every month. More than 800 at-risk patients have received text messaging interventions and more than 75% of patients report satisfaction with the program.

“When compared with matched controls, patients receiving the text messaging have 24% increase in prenatal visit attendance, 27% drop in early preterm delivery, and 54% drop in baby costs in the first year of life,” Miff reported. “The program has resulted in substantial savings to Parkland Community Health Plan with a return on investment about 2.5 times in the first year of full program implementation.”

As the Parkland Center for Clinical Innovation moves past initial program set-up costs, it estimates an increase in ROI reaching four to five times per year in subsequent implementation years, he added.

ADVICE FOR OTHERS

“Patient awareness and engagement is a key element not only for pre-term birth, but for other population health programs,” Miff advised. “Risk stratification using clinical history, clinical profile, chronic conditions/comorbidities and SDOH are critical to tailor patient outreach, education and engagement based on individual risks.”

Another key recommendation is not to let technology be a barrier to adoption, he said.

“Digital technology used for patient engagement needs to be easily accessible and convenient for the population you’re trying to engage,” he advised. “While we tested a mobile app and in-home voice-activated personal assisted devices, for the inner-city Medicaid population, texting is currently the most convening and most accessible modality.”

Tailoring messages and frequency to patient risk profiles is important to decrease the risk for digital fatigue and enhance satisfaction with the program, he added. Claims-data based models are very powerful and meaningful, but in some cases, they need to be enhanced with other data sources, he said.

“For our model, we’re enhancing the claims model with EHR data to address the 90-day lag in claims data processing and adding mental-behavioral history into the predictive risk models to further improve the predictive power of the model and connect patients to appropriate provider-driven services and workflows,” Miff concluded. “We are starting to collaborate with Drs. Kenneth Rogers and Kimberly Roaten to study and leverage the more than 2 million suicide screenings done at Parkland since 2014 to gain further insights into the mental-behavioral co-morbidities of pre- and post-natal patients and other at-risk populations.”

These insights will enable staff to further understand patient complexities beyond their primary medical diagnoses and align and connect services across medical providers and across the community, he said.

Twitter: @SiwickiHealthIT
Email the writer: bill.siwicki@himssmedia.com
Healthcare IT News is a HIMSS Media publication.