Missingness patterns in a comprehensive instrument identifying psychosocial vulnerability and substance use in antenatal care

Abstract

Background:

Psychosocial vulnerabilities (e.g., inadequate social support, financial insecurity, stress) and substance use elevate risks for adverse perinatal outcomes and maternal mental health morbidities. However, various barriers, including paucity of validated, simple and usable comprehensive instruments, impede execution of the recommendations to screen for such vulnerabilities in the first antenatal care visit. The current study presents findings from a newly implemented self-report tool created to overcome screening barriers in outpatient antenatal clinics.

Methods:

This was a retrospective chart-review of 904 women who completed the Profile for Maternal & Obstetric Treatment Effectiveness (PROMOTE) during their first antenatal visit between June and December 2019. The PROMOTE includes the 4-item NIDA Quick Screen and 15 additional items that each assess a different psychosocial vulnerability. Statistical analysis included evaluation of missing data, and exploration of missing data patterns using univariate correlations and hierarchical clustering.

Results:

Three quarters of women (70.0%) had no missing items. In the entire sample, all but four PROMOTE items (opioid use, planned pregnancy, educational level, and financial state) had < 5% missing values, suggesting good acceptability and feasibility. Several respondent-related characteristics such as lower education, less family support, and greater stress were associated with greater likelihood of missing items. Instrument-related characteristics associated with missing values were completing the PROMOTE in Spanish or question positioning at the end of the instrument.

Conclusions and Implications:

Conducting a comprehensive screening of theoretically and clinically meaningful vulnerabilities in an outpatient setting is feasible. Study findings will inform modifications of the PROMOTE and subsequent digitization.

Introduction

Leading health organizations recommend comprehensive psychosocial risk assessment, or Screening, Brief Intervention and Referral to Treatment (SBIRT), during the first antenatal visit (American College of Obstetrics and Gynecology, 2006; Patnode et al., 2020; Reddy, Davis, Ren, & Greene, 2017; Wright et al., 2016). SBIRT focuses on substance use and mental health problems. However, it does not regularly address other psychosocial factors which are known to affect women’s health. There is need for an evidence-based, feasible (e.g., short, self-report, acceptable), comprehensive psychosocial screening instrument to identify vulnerable women at risk of adverse perinatal (e.g., preterm birth, Neonatal Abstinence Syndrome), behavioural (e.g., inadequate care, substance use) and mental health outcomes (e.g., depression, anxiety).

Barriers to comprehensive screening

Stress, insufficient social support, interpersonal violence, residential instability, low income and education, and psychiatric co-morbidities are risk factors for poor maternal and infant outcomes due to their association with inadequate care, poor treatment adherence, and health behaviours during pregnancy (Lobel et al., 2008). Substances such as alcohol, tobacco, and opioids adversely affect outcomes due to their teratogenic or placental impact and their associated behavioural risks (Kozhimannil, Graves, Levy, & Patrick, 2017; Schempf & Strobino, 2009). Therefore, comprehensive screening should include the psychological, social, and behavioural risk factors that are directly or indirectly associated with adverse outcomes (Preis, Garry, Herrera, Garretto, & Lobel, 2020).

A dearth of validated simple clinical tools along with professional and patient driven barriers help account for the poor detection and assessment of psychosocial and substance use risk factors in antenatal care (Connell, Barnett, & Waters, 2018; Harrison & Sidebottom, 2008; Oni, Buultjens, Abdel-Latif, & Islam, 2019). Providers often lack time to conduct assessment, may be unaware that psychosocial factors affect treatment and health, may not be skilled in screening, and may be reluctant or unable to intervene if vulnerabilities are detected (Chang et al., 2008; Williams et al., 2015; Rahm et al., 2015; Wright et al., 2016). Patients underreport information they are not specifically asked about, may not understand the relevance of psychosocial questioning, and may allow perceived stigma to hinder disclosing psychosocial problems and substance use (Frazer, McConnell, & Jansson, 2019; Leis, Mendelson, Perry, & Tandon, 2011; Preis, Inman, & Lobel, 2020).

Using a screening questionnaire in antenatal care increases the likelihood of detecting substance use and psychosocial vulnerabilities (Carroll et al., 2005). Yet there are no validated, systematic, efficient, easily administered instruments to comprehensively yet simply assess psychosocial risk factors in pregnancy while being aligned with antenatal appointment time constraints and United States culture (Harrison & Sidebottom, 2008; Ko et al., 2020). Existing validated instruments that were developed outside the United States are sometimes limited in their focus (e.g., on mental health or substance use) or cultural context, are less feasible for clinical use, or do not quantify risk severity systematically and in a simple manner to enable development of sophisticated methods to predict risk (Johnson et al., 2012). For example, validates instruments such as ALPHA, KINDEX, Mind2Care, ANRQ-R, PDPI-R (Alves, Fonseca, Canavarro, & Pereira, 2019; Carroll et al., 2005; Priest, Austin, Barnett, & Buist, 2008; Quispel et al., 2014; Reilly et al., 2021; Spyridou, Schauer, & Ruf-Leuschner, 2015) are either not self-administered, making them less aligned with time constraints related to the many tasks providers have to accomplish during a busy first antenatal visit (ALPHA), are long and require calculations (KINDEX), do not assess important domains such as healthy behaviours or factors relevant to pregnant women in the United States such as home ownership (Mind2Care), or are focused on mental health risk (ANRQ-R, PDPI-R). We know of no well-validated screening tool in the United States that assesses substance use and other psychosocial risk factors that are critical for perinatal physical and mental health such as overall stress, major life events, health behaviours, partner support and family support, and economic indicators (Preis et al., 2020) in a single instrument. The current investigation is part of the process evaluation of a comprehensive instrument that assesses these factors.

Current study

In June 2019, we developed a novel substance use and psychosocial screening instrument, the Profile for Maternal & Obstetric Treatment Effectiveness (PROMOTE), for antenatal clinics affiliated with a large tertiary care public hospital. The objective of this study was to assess patients’ completion of the PROMOTE and evaluate whether the magnitude and pattern of missing items (i.e., item nonresponse) is associated with instrument structure or patient characteristics (Bulut, Xiao, Rodriguez, & Gorgun, 2020; Hughes et al., 2019). Items may be intentionally skipped or unintentionally missed, implicating problems with content (e.g., sensitive topic or irrelevant to respondent), ambiguous or incomprehensible wording, or formatting and placement issues. Examining item missingness helps ensure that instrument structure is sound, and items are comprehensible. Missingness can also diminish the fidelity of data, its generalizability, and its use to create sophisticated prediction models. In-depth investigation will identify any necessary modifications to the PROMOTE. Our main study questions were:

1. How fully do patients complete the screening instrument?

2. Are there distinguishable missing item patterns?

3. Which individual factors and structural factors are associated with missingness?

Materials and Methods

Study Design and Participants

We conducted retrospective chart review of all women who completed the PROMOTE during their first antenatal visit at an outpatient clinic associated with a large university hospital on Long Island, New York between June 6 and December 10, 2019.

Procedure

All patients are asked to arrive 15 minutes in advance of their antenatal care appointment. At check-in, they received a clipboard with the paper PROMOTE in English or Spanish together with additional forms in their intake packet to be completed before their appointment. The PROMOTE began with a prompt stating, “Below are questions that will help your doctors and nurses get to know you better and provide care that meets your needs. There are no right or wrong answers. Please answer as honestly as you can. If any question makes you feel uncomfortable, please skip it.” Patients completed the PROMOTE while sitting in the waiting room, sometimes with another person of their choice present with them. The completed form was then used by their provider to assess vulnerabilities following the SBIRT framework. Each completed PROMOTE was scanned into the Electronic Medical Record (EMR). The Institutional Review Board at Stony Brook University granted ethical approval for abstraction of records and entry into REDCap data collection software and approved a waiver of informed consent for this minimal risk study (IRB2019–00512).

Assessments

The PROMOTE was created by a team of social scientists, obstetrics providers, and administrators. Each constituent item assesses a single psychosocial construct (e.g., perceived stress, support from partner, domestic violence) or health behaviour (e.g., tobacco use, opioid use, healthy antenatal behaviours). The first section contains 19 items that are the focus of this investigation. At the end of this section is a brief description of opioids and a question about use of opioids in the last year. The following items were included in the PROMOTE, listed by order of appearance. Tables 1 and 2 list response options.

Table 1.

Sociodemographic sample characteristics

	n (valid %)	Missing n (%)
Planned pregnancy		78 (8.6)
No	340 (41.2)
Yes	486 (58.8)
Employment		21 (2.3)
Full time	454 (50.2)
Part-time	129 (14.3)
Homemaker	114 (12.6)
Unemployed	105 (11.6)
Student	28 (3.1)
Other	53 (5.9)
Education level		60 (6.6)
Some high school (no diploma)	51 (5.6)
High school graduate	268 (29.6)
Associate’s degree	116 (12.8)
Bachelor’s degree or more	409 (45.2)
Financial state		73 (8.1)
Below average	137 (15.2)
Average	610 (67.5)
Above average	84 (9.3)
Residence		19 (2.1)
Homeowner	373 (40.5)
Renting	324 (34.7)
With family or friends	167 (18.5)
Group residence	6 (0.7)
Homeless/ shelter	4 (0.4)
Other	28 (3.1)
Current relationship		32 (3.5)
Married or cohabiting	588 (65.0)
Serious relationship	219 (24.2)
No relationship	38 (4.2)
Other	27 (3.0)

Table 2.

Behavioural health, substance use screening and distribution

	M±SD	Missing n (%)	Never n (valid %)	1–2 times n (valid %)	Monthly n (valid %)	Weekly n (valid %)	Daily n (valid %)
Alcohol (four drinks or more)	1.46±0.84	17 (1.9)	620 (68.6)	172 (19.0)	59 (6.5)	25 (2.8)	11 (1.2)
Tobacco	1.44±1.15	9 (1.0)	755 (83.5)	46 (5.1)	11 (1.2)	9 (1.0)	74 (8.2)
Prescription drugs	1.05±0.40	10 (1.1)	877 (97.0)	5 (0.6)	4 (0.4)	1 (0.1)	7 (0.8)
Illegal drugs	1.14±0.61	13 (1.4)	834 (92.3)	28 (3.1)	9 (1.0)	5 (0.6)	15 (1.7)
			Very little				Very much
			1	2	3	4	5
Family support	4.35±1.12	16 (1.8)	45 (5.1)	31 (3.5)	95 (10.7)	114 (12.8)	603 (67.9)
Partner support*	4.71±0.81	15 (1.7)	22 (2.5)	8 (0.9)	38 (4.1)	71(7.8)	752 (82.9)
Antenatal stress	4.06±1.18	18 (2.0)	179 (20.2)	215 (24.3)	293 (33.1)	133 (15.0)	66 (7.4)
Positive health behaviours	4.06±0.97	22 (2.4)	14 (1.5)	45 (5.0)	172 (19.0)	292 (32.3)	359 (39.7)
			No n (valid %)	Yes n (valid %)
Major life events		25 (2.8)	746 (84.9)	133 (15.1)
Abuse		25 (2.8)	851 (96.8)	28 (3.2)
Emotional difficulty		20 (2.2)	795 (89.9)	89 (10.1)
Psychiatric medication		21 (2.3)	833 (94.3)	50 (5.7)
Any kind of opioids		90 (10.0)	778 (95.6)	36 (4.4)

^*Partner support data does not include n = 16 (1.8%) women who mentioned that this was not applicable to them

Planned pregnancy was assessed by asking women ”was this pregnancy planned”. Sociodemographic data included employment, education, financial state, residential stability, and relationship status. Social Support was assessed with two items: extent of help or support from family or friends, and from partner or spouse. Antenatal stress was assessed by asking, “How much stress do you currently have in your life?” Healthy antenatal behaviours were assessed with, “To what extent are you involved in healthy activities (eat well, take vitamins, come to medical appointments, etc.)?” Substance use was assessed using the 4-item NIDA quick screen (National Institute on Drug Abuse, 2012): frequency of past year risky alcohol use (four or more drinks), tobacco, prescription drugs for non-medical use, and illicit drugs. Women indicated whether they had experienced major life events (e.g., “break-up, moving, death of someone close, etc.”) since becoming pregnant. Intimate Partner violence was assessed by asking if women had “ever been emotionally or physically abused by a partner or someone important to you”. Emotional difficulty was assessed by asking about current “emotional or psychiatric problems” and if yes, which. Respondents were asked if they are regularly taking psychiatric medication and if so, which. Opioid use was the last question in the first section of the PROMOTE and began with an explanation and examples of opioids such as heroin, Fentanyl, painkillers (Oxycodone, Percocet) or medication assisted treatment (Subutex, Methadone, Suboxone) followed by a question regarding any past year opioid use. The second section of the PROMOTE is comprised of in-depth questioning on substance use completed only by women who endorse past year opioid use. These data are not analysed in the current investigation.

Age, gestational week of pregnancy, race/ethnicity, and language in which the PROMOTE was completed (English or Spanish) were also abstracted.

Analysis

Analyses were performed using SPSS 27.0 and Python (NumPy, SciPy, Seaborn, Scikit-learn libraries). We examined overall completion rates of the PROMOTE and missing values patterns, including missing pairs. Distribution patterns were compared across items with differences evaluated by χ² tests. Hierarchical clustering of respondents based on missingness was used to identify and visualize missingness patterns. Y_N×K represented a matrix of responses of N respondents to K questions; the missingness indicator was defined by the matrix M, such that m_ij = 1 if y_ij is missing, and m_ij = 0 if y_ij is not missing (Little & Rubin, 2019). The matrix M, has 268 rows (number of respondents with missing item) and 19 columns (number of items). Next, we clustered respondents and items into similar groups in terms of missingness patterns. We defined M′ as the clustered version of M, where its rows are clusters of respondents and columns are clusters of items. We applied hierarchical clustering with a bottom-up approach, where each observation starts in its own cluster, and pairs of clusters merge based on similarity of their missingness pattern.

Results

Study sample

The sample included 904 women, on average 30.7±5.9 years old (range 15–47), average gestational age 12.7±8.2 weeks. The majority completed the PROMOTE in English (n=871, 96.5%). The sample was representative of the suburban population where the study was conducted with half of the participants working full-time (n=454, 50.2%), a fifth identifying as Hispanic/ Latinx (n=193, 21.8%), over a third that were homeowners (n=373, 40.5%), and nearly two-thirds (n=588, 65.0%) who were married or cohabiting. Additional sample characteristics and item distributions are in Tables 1 and 2.

Missing value patterns

95.7% of women had three or fewer missing items from all 19 items. 636 women (70.0%) had no missing items, 166 (18.4%) had one, 46 (5.0%) had two, 17 (1.8%) had three, and 39 women (4.3%) had four or more missing items. Five women completed fewer than half of the items. Little’s test to estimate Missing Completely at Random (R. J. A. Little, 1988) indicated that missingness was not random (χ²(110)=196.5, p<0.001). Therefore, we conducted further investigation of missing item patterns.

Missingness was more common in some items (see Tables 1 and 2) and some item pairs. As can be seen in Figure 1, four items had more than 5% missing values: Opioid Use (10.0%), Pregnancy Planning (8.6%), Financial State (8.1%), and Education (6.6%). Missing values for other items ranged from 1.1% (past year tobacco use) to 3.5% (current relationship).While not exhibiting high missingness, the five sets of items most frequently missing in pairs included Education and Financial State (n=27, 2.9% missing both), Education and Relationship Status (n=20, 2.2% missing both), Psychiatric Medication and Opioid Use (n=20, 2.2% missing both), Emotional Problems and Opioid Use (n=19, 2.1% missing both), and Emotional Problems and Psychiatric Medication (n=19, 2.1% missing both). Generally, missingness in pairs was more common for items within the same domain, such as sociodemographic data (e.g., Education and Financial State; Education and Relationship), or for items located closely (e.g., Opioid Use, Emotional Difficulty, Psychiatric Medication).

Figure 1:

Count and percentage of missing values per item

Correlates of missing patterns

A missing value on the Opioid Use item was more common among women with a high school education than those with an Associates or college degree (χ²(2)=31.59, p<0.001), among women with below average rather than average or above average Financial State (χ²(2)=6.13, p=0.047), and among women who completed the instrument in Spanish rather than English (χ²(2)=12.19, p<0.001). Women who missed the Opioid Use item were younger (t(897)=2.21, p=0.04) and reported less Family Support than those who responded (t(886)=2.17, p=0.03).

A missing Pregnancy Planning response was more common among women who did not use any tobacco (χ²(2)=6.59, p=0.01) or alcohol (χ²(2)=8.91, p=0.003) compared to those who did, and among multiparous women versus primiparous women (χ²(2)=4.28, p=0.04).

A missing value for Financial State was more common among those with a high school education than an Associates or college degree (χ²(2)=40.78, p<0.001), among women who did not use psychiatric medication versus those who did (χ²(2)=4.35, p=0.04), among those who were in a serious relationship or not in a relationship compared with women who were married or cohabiting (χ²(2)=21.78, p<0.001), among those who did not use alcohol (χ²(2)=7.22, p=0.007) or illicit drugs (χ²(2)=5.03, p=0.025) compared to those who did, and among women who completed the instrument in Spanish rather than English (χ²(2)=17.93, p<0.001).Women who missed the Financial State item reported more Stress than those who did respond to this item (t(884)=2.30, p=0.02).

Finally, a missing Education Level response was more common among women who reported having below average rather than average or above average Financial State (χ²(2)=25.56, p<0.001), among those who did not use illicit drugs versus those who did (χ²(2)=4.01, p=0.045), and among women who completed the instrument in Spanish versus English (χ²(2)=249.90, p<0.001). Women who missed the Education Level item reported less Family Support than those who did respond (t(886)=3.35, p<0.001), and the former reported more Antenatal Stress than those who responded (t(884)=2.56, p=0.01).

Clustering patients by missingness patterns

Figure 2 shows the heatmap visualization of M′ among 268 women who had any missing items. The red colour intensity illustrates the percentage of respondents who missed the question corresponding to that column (labelled at the bottom of the map). Values to the left of the map are the number of respondents in the corresponding cluster. For example, the first row has four respondents who exhibited the highest missing percentage, the bottom row has 13 respondents who missed the last few PROMOTE questions (items 16 to 19), and the row second to the bottom had 67 respondents who missed only the Opioid Use question. Figure 2 also illustrates various ways to cluster respondents based on patterns of missingness. For example, to produce two clusters, respondents in the first row would comprise cluster 1, and the remaining respondents would comprise cluster 2. To produce three clusters, cluster 2 is split into clusters of respondents in the bottom row versus a cluster of remaining respondents. The heatmap also illustrates that items placed proximally, or which are thematically similar, were more often missing together. For example, in the second row from the top, the cluster is missing items from the psychosocial background section (Family Support, Partner Support, Stress, Healthy Behaviours).

Figure 2: Hierarchical clustering of 268 patients with missing items.

Note: Rows represent patients, and the number at the left represents the number of patients in that row. Columns represent items and the colour represents missingness (i.e., darker is more likely to be missing). The figure represents clustered respondents and items based on their missingness patterns.

Discussion

In this study, we evaluated missingness patterns in the PROMOTE, a new comprehensive instrument to conduct psychosocial and substance use screening in antenatal care. Although overall, few items were missing, missingness patterns implicate instrument characteristics (e.g., positioning, language) and patient characteristics (e.g., age, education) in item nonresponse and indicate that more vulnerable patients (e.g., less social support, greater stress) tend to complete less of the PROMOTE. The high completion rate (95.7% of women had three or fewer missing items) establishes the feasibility of the PROMOTE as a tool self-administered by patients and results of this investigation will inform modification and use of the PROMOTE.

Substance use screening in the PROMOTE was conducted with the NIDA quick screen (National Institute on Drug Abuse, 2012) and a separate question regarding use of any opioid in the last year. The missingness of items in the NIDA quick screen, which was located in the middle section of the PROMOTE, was minimal (<2%), whereas the opioid question, located at the end of the form, had relatively high missingness (10.0%). Pregnant women often do not report substance use due to shame and concerns about potential consequences of disclosure (Palamar, 2019). Compared with other substance use screening instruments that specify a time frame within the antenatal period (e.g., since becoming pregnant, in the past month), the one-year time frame used in the PROMOTE item encompasses the preconception period. Thus, it does not spotlight substance use during pregnancy, but rather a risk for use during pregnancy. Indeed, over a third of participants reported harmful substance use in the year prior to their first antenatal visit, putting them at greater risk during pregnancy, since preconceptional substance use is a strong predictor of antenatal use (Harrison & Sidebottom, 2009; Skagerstróm, Chang, & Nilsen, 2011).

While focusing on past year (preconceptional) use might be associated with “false positives”, this type of screening method is likely to reduce more consequential “false negatives” (i.e., not identifying antenatally substance using women), by substance-using women reluctant to disclose substance use in pregnancy. For women, this approach could provide an opening to discuss substance use during pregnancy, and for providers, it could indicate a reason for additional substance use screening (e.g., direct questioning about substance use or urinalysis) in an empathic encounter. Additional research is needed to assess the predictive validity of this questioning for antenatal substance use. Nevertheless, a tenth of participants did not respond to the past year opioid use item. Since this was the last item on the PROMOTE, and was on the second page together with in-depth questioning about substance use, it might have been mistakenly overlooked. It is unlikely that missingness of the opioid use item was related to insufficient time to complete the full measure, because the item preceding opioid use had high completion rates (psychiatric medication; 2.3% missingness).

Modifications of the PROMOTE will include grouping this item with the other substance use items. Finally, the NIDA Quick Screen does not explicitly list marijuana. This might contribute to under-identification of women who use marijuana if they do not consider marijuana an illicit drug, a possibility that we confirmed in qualitative, in-depth interviews of some respondents (Authors, in preparation). Only 7.7% of respondents reported illicit substance use, which is lower than national rates of marijuana use which are 34.3% past year use among women 18–25 and 12.1% among women 26 and older (Center for Behavioral Health Statistics and Quality, 2020). Therefore, we will modify the PROMOTE to add a separate question regarding marijuana use in the year before pregnancy.

Missingness of Pregnancy Planning, Financial State, and Education were somewhat higher than other items, despite their location at the beginning of the PROMOTE. Women may be reluctant to answer these items because of social desirability, shame or inner conflict, or not understanding their relevance to antenatal care. Having an unplanned pregnancy could be embarrassing or women might feel judged that they were not ‘careful enough’. In addition, some women may not endorse a pregnancy as unplanned if they interpret this to mean that the pregnancy was unwanted (Aiken, Borrero, Callegari, & Dehlendorf, 2016). Yet this is an important factor to assess early in pregnancy since unplanned pregnancies are associated with greater preconception substance use, poorer mental health, and worse perinatal outcomes (Lundsberg, Peglow, Qasba, Yonkers, & Gariepy, 2018; Meiksin et al., 2010). Therefore, in clinical practice, missingness of the Pregnancy Planning item will warrant further inquiry by provider. As for the assessment of financial status, respondents often consider this a private matter. Reluctance to disclose one’s income is a known problem in data collection (Yan, Curtin, & Jans, 2010). Moreover, nonresponse to financial state was more common among women who were more stressed and had less family support, suggesting that they may have had financial strains that they did not want to disclose, potentially out of embarrassment. Since the aim of this question is to assess financial instability, we will modify it to directly assess difficulty obtaining or paying for necessities.

Finally, the higher rate of missing items among patients who completed the PROMOTE in Spanish requires further investigation of possible contributors. These could be related to wording, cultural misfit, or to unique strains faced by Spanish speaking patients.

Limitations and Strengths

One strength of the PROMOTE is that it includes single items to assess each risk construct, making it brief and usable. Although single-item assessment has been shown to be acceptable in previous antenatal contexts (Sagrestano et al., 2002), further research is needed to confirm the validity of this approach when assessing the specific vulnerabilities and risk factors included in the PROMOTE. Additionally, although self-report instruments are recommended to screen for substance use in pregnancy, reluctance to disclose could lead to under-identification of substance using pregnant women (Elertson & Schmitt, 2019; Reddy et al., 2017). There is a need for further investigation of ways to indirectly assess risk for substance use and to confirm the validity of such assessment based on prospective research that includes biological sampling (González-Colmenero et al., 2021). Finally, our assessment of the PROMOTE relied on outpatients of a single university-affiliated medical centre and further study with the modified PROMOTE should assess its feasibility in other antenatal settings (e.g., diverse population, varying appointment times).

Conclusion

The PROMOTE exhibited acceptability and overall, produced little missing data. Nevertheless, the current analysis laid the groundwork for some modifications of item placement and wording to improve completion of the instrument. Together with results of qualitative investigations of providers, administrators, and patients examining the ease of distribution and use of the instrument, a modified version of the PROMOTE will be created and digitized. The modified and digitized PROMOTE is expected to be as acceptable and accessible to patients (Kingston et al., 2017), and less likely to result in missing data (Highet, Gamble, & Creedy, 2019; Spark et al., 2015). Moreover, digitizing the PROMOTE will enable directly linking it to the EMR, and facilitate identifying women at risk. Administering the PROMOTE electronically will also make it easy to use in subsequent antenatal visits (Kiely, Gantz, El-Khorazaty, & El-Mohandes, 2013), but we believe that its utility in identifying risk and conducting SBIRT is best achieved as early in pregnancy as possible.

With proper implementation (Reilly, Brake, Kalra, & Austin, 2020), use of the PROMOTE will allow providers to deliver evidence-based care for their pregnant patients based on the patients’ perspectives, experiences, and the providers’ clinical judgment (Ben-David, Jonson-Reid, & Tompkins, 2017). As demonstrated by some of our findings and other recent research (Ray, Gitz, Hu, Davis, & Miller, 2020), missingness of social needs items is more likely among vulnerable groups. Further research and techniques are needed to indirectly and directly identify individuals with psychosocial risks, especially when missing items are related to disclosure reluctance. Currently, our team is working on developing machine learning algorithms to flag patients who are more vulnerable to social, behavioural, or medical risks including perinatal outcomes (e.g., inadequate antenatal care, preterm birth, positive urine toxicology, antenatal and postnatal depression assessed using the Edinburgh Depression Scale).The successful use of the PROMOTE to identify women at risk and foster appropriate interventions for them is a critical means to protect the health and well-being of pregnant women and their offspring.