Predicting Severe Maternal Morbidity and Mortality - An Informatics Opportunity

Citation:

Carroll, W. (November, 2018). Predicting Severe Maternal Morbidity and Mortality - An Informatics Opportunity. Online Journal of Nursing Informatics (OJNI), 22(3).  Available at http://www.himss.org/ojni

A disturbing trend

In the last year, there has been a lot of press bringing to light the alarming maternal morbidity and mortality crisis in the United States (National Public Radio, 2018; USA Today, 2018). These investigative reports have highlighted new information from the Agency for Healthcare Research and Quality’s (AHRQ) 2018 statistical brief, “Trends and disparities in delivery hospitalizations involving Severe Maternal Morbidity (SMM), 2006-2015.” The reports highlighted that between 2006 and 2015, SMM increased at a rate of 45% for U.S. mothers, from 101.3 to 146.6 per 10,000 delivery hospitalizations (AHRQ, 2018). Both healthy women and women with comorbidities are dying in childbirth at far higher rates than women in other countries (USA Today, 2018).

The Centers for Disease Control and Prevention (CDC) does not fully understand why SMM is increasing (2018). The hypothesis based on the statistics links this rise to an increase in maternal age, pre-pregnancy obesity, pre-existing chronic medical conditions, and cesarean delivery (CDC, 2018). The agency asserts that “tracking and understanding patterns of SMM, along with developing and carrying out interventions to improve the quality of maternal care are essential” (CDC, 2018, pgh. 2). An opportunity exists for nurse informaticists to further the use of emerging and disruptive technologies to identify relationships in data and paint a clearer picture of what is contributing to SMM.

The need for predictive data

Much descriptive and diagnostic data exist to show the trends of SMM, with visualizations depicting the staggering increase in incidence. These data used in hindsight are not enough to improve outcomes for mothers. It is time to manipulate these data to answer the questions of why SMM occurs and, more urgently, to identify these events before they happen. Within the Big Data in today’s healthcare ecosystem, assembly of large, unbiased, clean and validated data sets can assist researchers to proactively determine what is happening to mothers before giving birth that may cause SMM or post-partum mortality.

With a strong focus on racial and ethnic disparities, social determinants of health and health equity, datasets must represent populations at the highest risk for SMM, since rates of SMM were found to be greatest among mothers who are poor, over age 40, uninsured, on Medicaid, and residing in large urban areas (AHRQ, 2018). SMM occurred more often among black, Hispanic and Asian/Pacific Islander women than among white women in 2015. Also, black women were three times more likely to die in childbirth than white women (AHRQ, 2018), making a case for more inclusive data sets that accurately represent all maternal races and ethnicities. 

How artificial intelligence can help

By adopting artificial intelligence (AI), specifically predictive analytics and machine learning (ML), the health informatics community can work with data scientists to identify mothers at risk of becoming sick during the intrapartum and post-partum periods. Predictive analytics is beginning to gain traction for its use in determining care variations, population health risks, and disease onset and progression. We can begin to use fresh data sets from agencies such as the CDC, medical associations and women’s health initiatives, to gather the patient and care delivery-specific variables needed to predict at-risk mothers.

Using ML, we can begin to unearth the unknown patterns in maternal data from multiple sources including claims, electronic health records (EHR), medication prescriptions, and patient-reported data that attribute to poor outcomes. A study conducted in Bogotá, Columbia, where SMM is also a public health issue, used ML techniques to develop a probabilistic classifier based on mothers’ risk factors during pregnancy to predict adverse events during childbirth and the puerperium (Rodríguez, Estrada, Torres, & Santos, 2016). This project is ongoing with the intent to further build improved ML techniques and data sets that are more inclusive of population demographics, with enhanced diagnostic features, using socioeconomic variables for risk prediction. Studies such as this serve as an example of using predictive analytics to find out what attributes to poor outcomes for mothers in childbirth. 

An informatics opportunity

A newly formed initiative, the Alliance for Innovation on Maternal Health (AIM), works “through state teams and health systems to align national, state, and hospital level quality improvement efforts to improve overall maternal health outcomes” (n.d., pgh. 2). One of the goals of this initiative is to build datasets to collect features and variables that can contribute to maternal early warning systems (MEWS) and track criteria for predicting maternal care early, for better outcomes (AIM, 2018).

Along with programs like AIM, nurse informaticists have an immense opportunity to partner with health information and technology vendors, health system and medical experts in SMM and maternal mortality to further develop comprehensive data sets. These data can lead to novel technology solutions that decrease errors, adverse events and near misses and create better maternal outcomes. Nursing informaticists can use their expertise to identify, prepare and use predictive data that forecasts potential SMM.

The insurgence of information about SSM speaks to the need for clinical practice standards for the care of mothers during and after childbirth. Rich datasets for machine-driven approaches to improve population outcomes and incite policy development are needed now. With the understanding, use and adoption of emerging and disruptive technologies such as AI, nurse informaticists can be at the forefront to help mothers in crisis.

References

Agency for Healthcare Research and Quality (AHRQ). (2018, September). Statistical Brief #243. Trends and disparities in delivery hospitalizations involving severe maternal morbidity, 2006-2015. Retrieved from https://www.hcup-us.ahrq.gov/reports/statbriefs/sb243-Severe-Maternal-Morbidity-Delivery-Trends-Disparities.jsp

Alliance for Innovation on Maternal Health. (2018). About AIM. Retrieved from https://safehealthcareforeverywoman.org/aim-program-3/about-aim/

Centers for Disease Control and Prevention (CDC). (2018). Severe Maternal Morbidity in the United States. Retrieved from https://www.cdc.gov/reproductivehealth/maternalinfanthealth/severematernalmorbidity.html

National Public Radio. (2018). Lost mothers: Maternal mortality in the U.S. Retrieved from https://www.npr.org/series/543928389/lost-mothers

Rodríguez, E. A., Estrada, F. E., Torres, W. C., & Santos, J. C. (2016). Early prediction of severe maternal morbidity using machine learning techniques. Lecture Notes in Computer Science Advances in Artificial Intelligence - IBERAMIA 2016, 259-270. https://doi.org/10.1007/978-3-319-47955-2_22

USA Today. (July 2018). Hospitals know how to protect mothers, they just aren’t doing it. Retrieved from https://www.usatoday.com/in-depth/news/investigations/deadly-deliveries/2018/07/26/maternal-mortality-rates-preeclampsia-postpartum-hemorrhage-safety/546889002/