Physician Referrals During Prenatal Care

Abstract

Objective:

Referrals are an important component of patient care, and have been increasing over time. During pregnancy, people have intensive contact with the healthcare system, but little is known about the involvement of different physicians for pregnant patients during this period. This study examines referral patterns during prenatal care visits.

Methods:

Using the 2006–2015 National Ambulatory Medical Care Survey and national birth certificate data, we estimate the number of referrals per pregnancy from prenatal care visits with OB/GYN and family medicine physicians. We use multivariable regression analysis to compare the probability of receiving a referral during a prenatal visit for visits with family medicine and OB/GYN physicians, controlling for visit, patient, and physician characteristics. Analyses are weighted to make results nationally representative.

Results:

224,335,436 prenatal visits over 19,893,015 pregnancies were included; 60% of these visits were covered by private insurance. On average, 0.3 referrals are made per pregnancy (95% confidence interval [CI]: 0.22,0.38). A prenatal visit with an OB was 5.5 percentage points less likely to result in a referral than a visit with a family medicine physician, controlling for other characteristics.

Conclusions:

Referrals are relatively common in prenatal care, and are more commonly initiated by family medicine physicians than by OB/GYNs. Understanding the contribution of multiple clinicians to a pregnant person’s health during the prenatal period and how coordination among clinicians impacts care receipt is an important next step. As healthcare becomes more specialized, better understanding care teams of individuals during the perinatal period is important for improving prenatal care.

Introduction:

Nearly 4 million people give birth annually in the United States (Martin et al., 2018a; Martin et al., 2018b) and during this time birthing people have frequent contact with the healthcare system: 94% of pregnant people have at least some prenatal care (Martin et al., 2019) and the average pregnant person attends more than 10 prenatal care visits (United States Department of Health and Human Services et al., 2019). Many studies have assessed factors associated with the receipt of adequate prenatal care (Blakeney et al., 2019; Gadson et al., 2017; Kitsantas et al., 2012). However, as evidence has mounted that having the recommended number and sequence of prenatal visits is not sufficient for ensuring good maternal and infant health outcomes (Alexander & Kotelchuck, 2001; Atrash et al., 2008; Noonan et al., 2013; Siddique et al., 2012), attention has turned to examining the content and quality of perinatal care (Attanasio et al., 2021; Coley et al., 2018; Dyer et al., 2018; Geissler et al., 2020; Sword et al., 2012). Referrals are an important element of patient care (Agha et al., 2019; Barnett et al., 2012; Mehrotra et al., 2011) and may contribute to prenatal care quality, but little is known about referrals in this context.

In healthcare generally, referrals have been increasing over time (Barnett et al., 2012; Geissler, 2020). Although the optimal referral rate for most clinical situations is not known, the frequency and nature of referrals varies based on clinical condition (Geissler & Zeber, 2020); geographic location (Geissler, 2020); and insurance, physician, and organizational characteristics (Agha et al., 2019; Barnett et al., 2012; Geissler et al., 2021; Geissler et al., 2020; Mehrotra et al., 2011). Care for individuals during pregnancy may be distinctive in ways that matter for referrals; for example, some specialists may be hesitant to treat pregnant people (Jones et al., 2014; Noonan et al., 2018; Wright et al., 2016). As the prevalence of chronic conditions has increased in the U.S. and as the average age at first birth has risen, more individuals are entering pregnancy with complex healthcare needs and developing complications during pregnancy (Ananth et al., 2019; Correa et al., 2015; Fisher et al., 2013). In some cases these conditions may be managed solely by the prenatal care clinician, but in other cases they may require consultation or co-management with specialists (Barnett et al., 2012). However, previous research has not demonstrated whether prenatal care has followed the rest of healthcare in becoming more specialized, and what factors may be associated with different referral patterns. Better understanding the frequency of referrals and their correlates is important to identify points in the healthcare system where coordination of care and attention to these transitions among physicians is required.

Referral patterns may differ between obstetrician-gynecology (OB/GYN) physicians and family medicine physicians, who provide prenatal care for a small but substantial proportion of pregnant people, particularly in rural areas (Cohen & Coco, 2009; Tong et al., 2013; Young, 2017). In the vast majority of prenatal visits with family physicians, the family physician is also serving as a primary care provider, while the clinician serving as the primary care provider is the case in less than a third of prenatal visits with OB/GYNs (Coco, 2009). As primary care providers, family physicians may be more likely than OB/GYN physicians to refer patients to specialty care in general (Barnett et al., 2012; Vintzileos et al., 2001), but it is also possible that they may make fewer referrals overall because they can care for primary care issues during pregnancy themselves (Coco, 2009).

Using the National Ambulatory Medical Care Survey (NAMCS) and natality records, we examine the frequency of referrals during prenatal care and contributors to these referrals, including the type of physician providing prenatal care and pregnancy complexity. This study is a first step in quantifying and understanding referrals made during prenatal care and how different physician specialties may work together to care for pregnant people during the perinatal period. We hypothesize that referrals will be more frequent for prenatal visits with family medicine physicians than OB/GYNs, and will be more likely for visits for high-risk pregnancies.

Methods:

Data and Sample:

Data are from the 2006–2015 NAMCS, a survey of office-based physicians conducted by the National Center for Health Statistics (Centers for Disease Control and Prevention National Center for Health Statistics). The survey includes detailed information about 50 patient visits in a one-week period. Visit weights are included to account for physician selection and non-response, making the weighted sample representative of all office-based physician visits (Centers for Disease Control and Prevention National Center for Health Statistics, 2018). We do not include 2016 data to avoid the mandated change to International Classification of Disease, Tenth Revision (ICD-10) in October 2015, which may impact the identification and classification of prenatal visits. We use additional data from the National Center for Health Statistics Natality files (United States Department of Health and Human Services et al., 2019) to determine the average number of prenatal visits per pregnancy by patient characteristics.

The primary analytic sample includes prenatal care visits by females age 18 and older to office-based physicians. Following prior literature, we identified visits as prenatal care visits if there was a primary or secondary diagnosis related to normal pregnancy, supervision of high-risk pregnancy, complications of pregnancy, or pregnancy-associated diagnoses (limiting to prenatal diagnoses using the fifth digit of the ICD-9 code as appropriate), or if the visit had a NAMCS reason for visit code indicating prenatal care (see Appendix Table 1 for full list) (Coco, 2009; Siddique et al., 2009; Siddique et al., 2012). We additionally limit to visits with family medicine and OB/GYN physicians, including physicians employed by practices owned by hospitals in 2014–2015. The sample excludes visits to physicians practicing in community health centers, federally employed physicians, and physicians practicing primarily in hospital outpatient departments (Centers for Disease Control and Prevention National Center for Health Statistics, 2018). Certified nurse-midwives practicing independently were not included in the NAMCS, and thus are excluded from the analysis. Visits were excluded from the analytic sample if they are missing data on the outcome or control variables.

Measures:

The primary outcome measure is whether the visit resulted in a referral to another physician, as determined by the visit disposition. The primary independent variable of interest is whether the visit was with a family medicine or OB/GYN physician.

In most analyses, we control for visit, patient, and physician characteristics. Visit characteristics include the supervision of high risk pregnancy (see Appendix Table 2) (Siddique et al., 2012), major reason for visit (e.g., visit for a new problem), number of diagnoses coded for the visit, and year of visit. Patient characteristics include race/ethnicity, age, and payment type (i.e., private insurance, Medicaid, and other). We include indicators of whether the patient had five specific comorbid conditions: asthma, depression, diabetes, hypertension, and obesity. Physician and organizational characteristics include physician specialty (OB/GYN or family medicine), census region of the physician office location, private practice office setting, solo practice physician, physician or physician group practice ownership, and physician location in a metropolitan statistical area.

Statistical Analysis:

We present descriptive statistics to compare prenatal care visits that result in a referral to visits that do not, weighted to the pregnancy level as detailed below. We use t-tests for continuous variables and chi-square tests for binary and categorical variables to identify statistically significant differences.

We then combine NAMCS visit-level data with data on the number of prenatal visits during a pregnancy from the natality files based on patient characteristics (age, race/ethnicity, and year of pregnancy) to estimate the average number of referrals that are made during a pregnancy. We calculate the number of referrals during a pregnancy overall, for patients receiving care from an OB/GYN versus family medicine physician, and for high risk versus non-high risk pregnancies. We calculate both unadjusted estimates and adjusted estimates controlling for the visit, patient, and physician characteristics described above.

We use a process based on the method developed in Siddique and colleagues (2009) to use visit-level estimates to calculate the number of referrals an individual obtains on average over the course of prenatal care during a pregnancy (Siddique et al., 2009; Siddique et al., 2012). We use a logistic regression model and marginal effects to estimate the average probability that visits overall, and visits of each type described above, will result in a referral. These probabilities are then multiplied by the average number of visits per pregnancy. The visit weights included in the NAMCS can be used to make results representative of office-based outpatient physician visits nationally. We additionally adjust these weights using the number of prenatal care visits by age-race/ethnicity-year categories to create weights representative of pregnancies (Siddique et al., 2009; Siddique et al., 2012). We calculate standard errors for these estimates using bootstrapping with 1000 replicates that accounts for the complex survey design and the additional pregnancy-level down-weighting (Siddique et al., 2009).

To determine whether there are differences in the probability of a prenatal care visit resulting in a referral by physician specialty, we estimate a logit model with the outcome of whether or not the visit resulted in a referral with the primary independent variable of the visit being with an OB/GYN vs. family physician. We then control for visit, patient, and physician characteristics as described above to isolate the association between prenatal visits and OB/GYN, using average marginal effects (AME) to interpret the results. Results indicate the change in the predicted probability of a visit resulting in a referral associated with a one unit change in the independent variable of interest.

As prenatal visits for specific concerns may differ from visits that are classified as being for preventive care, we conduct a sensitivity analysis limited to prenatal care visits for which the major reason for visit is “preventive care.” We conduct this sensitivity analysis to determine differences in the probability of referral by physician type. To determine the sensitivity of our results to the inclusion of obstetric specialists, we conduct an additional sensitivity analysis excluding prenatal visits to maternal-fetal medicine physicians and reproductive endocrinologists.

Analyses are conducted in Stata-MP version 16.0. The University of Massachusetts-Amherst Institutional Review Board determined this study was not human subjects research. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline (von Elm et al., 2007).

Results:

The primary analytic sample includes 224,335,436 prenatal care visits (n=7,323 unweighted visit observations) to office-based family medicine or OB/GYN physicians for 19,893,015 pregnancies over the 10-year period with complete information (Table 1). The sample began with 8,818 unweighted visit observations to office-based family medicine or OB/GYN physicians for females aged 18–44; 627 visits were excluded from years in which other settings (e.g., community health centers) were reported for consistency across time; 868 visits were excluded due to missing data on the outcome or covariates; the final analytic sample consists of 7,323 unweighted visit observations.

Table 1:

Descriptive statistics for prenatal visits (2006–2015)

		Visit Resulted in a Referral
	Overall	Yes	No
	(N pregnancies = 19,893,015)	(N pregnancies = 527,282)	(N pregnancies = 19,365,733)
% or Mean (standard deviation)	(N visits = 224,335,436)	(N visits = 5,937,299)	(N visits = 218,398,138)
Referral to other physician	2.7	100	0	*
Prenatal visit for high-risk pregnancy	34.5	59.7	33.8	*
Number of diagnoses for visit	1.4 (0.08)	1.6 (0.10)	1.4 (0.08)	*
Physician characteristics
Physician specialty category
OB/GYN (vs. family medicine)	93.0	75.5	93.5	*
Physician is full or part owner of practice	71.9	50.3	72.5	*
Office setting is a private solo or group practice	91.1	87.9	91.2
Physician is in a Solo Practice	32.7	28.8	32.8
Practice is owned by physician or physician group	82.9	73.1	83.2
Practice uses any electronic medical records	59.5	55.8	59.6
Office location is in a metropolitan statistical area	90.2	92.7	90.2
Patient characteristics
Patient age in years	28.2 (0.2)	28.6 (0.7)	28.2 (0.2)
Patient payment type
Private	60.0	59.1	60.0
Medicaid	33.9	38.1	33.8
Other - worker’s comp, self-pay, no charge, and other	6.1	2.8	6.2
Patient race/ethnicity
Non-Hispanic white	57.3	57.4	57.3
Non-Hispanic Black	13.8	19.8	13.6
Hispanic	21.4	14.5	21.6
Non-Hispanic Other	7.5	8.3	7.5
Patient comorbidities
Asthma	4.3	4.1	4.3
Diabetes	4.1	6.1	4.0
Depression	4.2	4.1	4.2
Hypertension	3.8	7.5	3.7	*
Obesity	5.2	2.9	5.3

Note: Summary statistics are calculated using per-pregnancy weights as described in the text, correcting for differences in the number of prenatal visits by different age-race/ethnicity categories.

^*indicates difference statistically significant at p<0.05 based on t-test for continuous variables and chi-squared test for binary and categorical variables. Standard errors that account for the sample design are in parentheses for continuous variables,. Overall unweighted sample size is 7,323 observations.

Of the included prenatal visits in the analytic sample, 34.5% were for high risk pregnancies. About 3 percent of visits resulted in a referral to another physician, with referrals significantly less common for visits with OB/GYNs versus family medicine physicians. Referrals were statistically significantly more likely for visits for high risk pregnancies, visits with a higher number of diagnoses, and for patients with hypertension. Additionally, referrals were significantly more common for visits where the physician was not in a practice owned by a physician or physician group and whether the physician was in a metropolitan statistical area.

On average, there were 0.30 referrals per pregnancy (95% confidence interval [CI]: 0.21,0.39) for the full sample of prenatal care during pregnancy (Figure 1; Supplemental Table 3, Panel A). The estimated number of referrals for patients receiving OB/GYN care adjusted for visit, patient, and physician characteristics of 0.25 referrals per pregnancy (95% CI: 0.16,0.33) is significantly lower than patients receiving care from a family medicine physician where the average number of referrals is 0.86 (95% CI: 0.39,1.32), adjusting for visit, patient, and physician characteristics (Figure 1; Supplemental Table 3, Panel B). High risk pregnancies have more than twice the number of referrals per pregnancy as non-high-risk pregnancies (0.46 adjusted referrals vs. 0.20 adjusted referrals) (Figure 1; Supplemental Table 3, Panel C).

Figure 1:

Estimated number of referrals per pregnancy

Note: 95% confidence interval shown with vertical lines. Confidence intervals for unadjusted estimates are calculated adjusted for the pregnancy-level downweighting as described in the text. Confidence intervals for regression adjusted estimates are estimate using bootstrapping with 1000 replicates as described in the text. Regression adjusted estimates include controls for (as appropriate based on comparison shown) OB/GYN versus family medicine physician, high risk pregnancy, patient age, patient payment type, patient race/ethnicity, number of diagnoses coded for visit, major reason for visit, year of visit, physician (Census) region, office setting is private practice, solo practice, physician ownership of practice, physician location in metropolitan statistical area, and patient comorbid conditions of asthma, depression, diabetes, hypertension, and obesity. Overall unweighted sample size is 7,323 observations.

Visits for prenatal care with OB/GYNs are associated with a significantly lower probability of referral (unadjusted AME=−0.072; p<0.001) than visits for prenatal care with family medicine physicians (Table 2; full results in Supplemental Table 4). After controlling for visit, patient, and physician characteristics, prenatal care visits with OB/GYNs have a 5.5 percentage point (pp) lower probability of referral, compared to prenatal care with family medicine physicians (adjusted AME=−0.055; p<0.001).

Table 2:

Associations between referrals and physician type

Outcome: Visit resulted in referral to other physician	(1)	(2)
OB/GYN (vs. family medicine)	−0.072***	−0.055***
	(0.021)	(0.020)
Controls?	No	Yes
Number of pregnancies (weighted)	19,893,015	19,893,015
Number of visits (weighted)	224,335,436	224,335,436

Note:

^***indicates statistically significant at the p<0.001 level. Associations shown are average marginal effects calculated after logit regression with standard errors in parentheses calculated to account for sampling design. Overall unweighted sample size is 7,323 observations. Weights are used as described in the text to adjust for differences in the number of prenatal visits by different age-race-ethnicity categories. For this table the unit of analysis remains the visit rather than the pregnancy. However, the weighting corrects for the over-influence of pregnant people with more pre-natal visits. We report both the weighted number of pregnancies and the weighted number of office visits. Controls included in column 2 for high risk pregnancy, patient age, patient payment type, patient race/ethnicity, number of diagnoses coded for visit, major reason for visit, year of visit, physician (Census) region, office setting is private practice, solo practice, physician ownership of practice, physician location in metropolitan statistical area, and patient comorbid conditions of asthma, depression, diabetes, hypertension, and obesity.

In the sensitivity analysis limited to prenatal visits classified as preventive care visits, visits for prenatal care with OB/GYNs are associated with a significantly lower probability of referral than visits with family medicine physician (Supplemental Table 5; unadjusted AME = −0.041; p<0.01). A preventive prenatal care visit with an OB/GYN is associated with a 2.8 pp lower probability of referral (adjusted AME=−0.028; p<0.001) than a visit with a family medicine physician, controlling for visit, patient, and physician characteristics. When we restrict our sample by excluding prenatal visits to reproductive endocrinologists and maternal fetal medicine physicians, we find very similar results to our primary analyses (Supplemental Table 6).

Discussion:

Although visit-level estimates of referrals during prenatal care are low, over the course of a pregnancy, many pregnant people receive a referral. Referrals are more common from visits with family medicine physicians than visits with OB/GYNs, but occur in both types of care. Our results are consistent with previous findings that specialists have lower referral rates than primary care physicians (Barnett et al., 2012), but extend this finding to the context of prenatal care. Understanding the use of multiple physician types during pregnancy, particularly care teams initiated by referral, is important for optimizing care birthing people receive during and after pregnancy.

Family medicine physicians continue to play an important role in maternity care services, particularly in rural areas where they provide the majority of obstetric care (Eden & Peterson, 2018; Young, 2017). However, the proportion of family physicians that provide obstetric care has declined from 43% in 1986 (Cohen & Coco, 2009) to 18% in 2018 (American Academy of Family Physicians, 2020). The higher rate of referrals among family physicians may be due to family physicians referring to OB/GYNs for pregnancy-related issues that are outside their scope of care. Past research has highlighted barriers to family physicians practicing advanced maternity care including surgical obstetrics, potentially leading to the need for more referrals among family physicians that do provide obstetric services (Eden & Peterson, 2018). However, family medicine physicians are more likely to address other issues while delivering prenatal care (Coco, 2009), and thus some referrals may be for non-pregnancy related problems that family physicians identify at higher rates than OB/GYNs.

The relationship between referrals and quality is complex. An appropriate referral rate is not clear, and in prenatal care the optimal referral rate may vary based on the clinician’s scope of care, as we found. Physician and patient characteristics, including comorbidities, differ between visits with and without referrals although most of these differences do not persist in regression adjusted analyses. Furthermore, while making referrals is an important step in connecting patients to specialty care, the NAMCS data do not allow us to measure whether patients who are referred to specialty care actually receive specialty care. Pregnant individuals from marginalized social groups may be more likely to have health needs requiring specialist referral (Ananth et al., 2019; Breathett et al., 2014; Grobman et al., 2018; Lisonkova & Joseph, 2013; Tucker et al., 2007), but face barriers to actually receiving specialist services following a referral (Cook et al., 2007; Healy et al., 2006; Henderson et al., 2013). Our study examines referral rates and correlates, but does not capture whether referrals result in the receipt of additional care.

Care coordination services in pregnancy have been proposed as a potentially effective approach to improving outcomes, particularly for pregnant people at high risk of poor birth outcomes (Hillemeier et al., 2018; Kroll-Desrosiers et al., 2016). Maternity care coordination services typically include elements such as communication among members of the care team, case management, patient education, and social services referrals (Kroll-Desrosiers et al., 2016). One goal is to ensure that patients receiving these services, who may be more medically complex, receive appropriate medical care, including specialists as needed (Hillemeier et al., 2018; Kroll-Desrosiers et al., 2016).

Several limitations to the analysis exist. First, the NAMCS only captures visits to office-based physicians (and in later years, hospital employed physicians) and does not include visits to hospital outpatient departments or community health centers. NAMCS also does not include certified nurse-midwives practicing independently, but does include visits to advanced practice clinicians who practice within physician practices. In 2016, certified nurse-midwives attended approximately 9% of births (Martin et al., 2019). In 2011, the last year for which both the NAMCS and the National Hospital Ambulatory Care Survey (capturing hospital outpatient department visits) were both available, the NAMCS captured 89% of adult primary care physician visits (Geissler, 2020). However, prenatal care may be disproportionately delivered in hospital outpatient departments; prior to 2011 when both surveys were available, multiple studies estimated approximately 14% of prenatal visits occurred in hospital outpatient departments (Coco et al., 2009; Yamamoto et al., 2015) and prenatal care provided by community health centers were a small percentage of total prenatal care (Coco et al., 2009). Referral rates tend to be higher in hospital outpatient departments (Barnett et al., 2012), and thus our estimated referral rates are likely to be an underestimate. Additionally, the 34% of prenatal visits we see covered by Medicaid is slightly lower than the national average of 38% of people who gave birth having Medicaid coverage for prenatal care (MACPAC, 2018), although these estimates are not directly comparable. These factors may impact the generalizability of our findings, although the data we use remains representative of the majority of prenatal care. A second limitation is that we do not know the target or intended timing of the referral. However, the NAMCS are the best available data for understanding referral patterns across physicians and organizations – other types of data are either within-organization (e.g. electronic medical records data) or only show completed referrals (e.g. health insurance claims data). The NAMCS referral indicator has been shown to be highly specific, but may underestimate referrals (Gilchrist et al., 2004), and not all referrals are made during office visits (Forrest et al., 2002). Given that we cannot see completed referrals, the referral rates we observe are likely overestimates of completed visits resulting from a referral. Future research should examine whether referrals are more likely to be completed in the perinatal period than in other health care encounters, and whether referrals are intended to be completed during or after pregnancy. A third limitation is that although we control for maternal medical conditions, ICD-9 codes for supervision of high-risk pregnancy, and physician office location in a metropolitan statistical area (a measure of rurality), there may be unobserved differences between pregnant people receiving care from family medicine physicians versus OB/GYNs.

Better understanding the scope of care for obstetric clinicians and the inclusion of multiple clinicians in prenatal care through referrals is important for improving the quality and coordination of care patients receive during pregnancy. As healthcare becomes more specialized, increased knowledge of care teams during the perinatal period, particularly for birthing people with comorbid conditions, is important for optimizing prenatal care.

Significance.

What is already known on this subject?

Referrals are an important component of patient care, and have been increasing over time. During pregnancy, pregnant people have intensive contact with the healthcare system, but referral patterns during pregnancy are unknown.

What does this study add?

In this cross-sectional study of a national sample of office-based physician visits, an average of 0.3 referrals was made over the course of a pregnancy. Referrals were more frequently initiated by family medicine physicians providing prenatal care than by OB/GYNs. Findings suggest that referrals are relatively common during pregnancy, particularly for birthing people treated by family medicine physicians.