Comparisons using data from birth certificate records included race/ethnicity, maternal age, education, payment for delivery, participation in the Women, Infants, and Children program , parity, maternal birthplace, report of smoking during pregnancy, maternal body mass index , trimester when prenatal care began, and number of prenatal care visits.For multi-parous women, we examined the relationship between preterm birth and previous preterm birth, previous cesarean delivery, and interpregnancy interval. Interpregnancy interval was calculated from previous live birth as reported in linked records and estimated as months to conception of the index pregnancy. Given that the day of previous live birth was not available, the middle of the month was used for calculation purposes.Factors from hospital discharge ICD-9 diagnoses included: Preexisting hypertension without progression to preeclampsia, preexisting hypertension with progression to preeclampsia, gestational hypertension without progression to preeclampsia, gestational hypertension with progression to preeclampsia, preexisting diabetes, and gestational diabetes. We also compared preterm birth with respect to the frequency of coded infection, anemia, drug or alcohol dependence/abuse, and mental disorder . Multi-variable models of maternal risk and protective factors for preterm birth were built for each location of residence category using backwards-stepwise Poisson logistic regression wherein initial inclusion was determined by a threshold of p < .20 in crude analyses. Adjusted RRs and their 95% CIs were calculated for each residence stratum. In an effort to visualize overall risk of preterm birth by census tract,planting growing racks cumulative risk scores estimated the overall risk of preterm birth.
Scores were calculated for each woman by adding her risks and subtracting her protective factors – 1) remaining in the final multi-variable model. Risk scores were grouped into scores 0.0 or less, 0.1 to 0.9, 1.0 to 1.9, 2.0 to 2.9 and 3.0 or more. Drug dependence/abuse and mental illnesses were further classified based on ICD-9 diagnostic codes, although risks calculations were not computed due to small numbers. Drug dependence/abuse was defined by classification of drug: opioid, cocaine, cannabis, amphetamine, other drug dependence/abuse, and poly substance dependence/abuse. Mental illnesses were further classified as: schizophrenic disorders, bipolar disorder, major depression, depressive disorder, anxiety disorders, personality disorders, and more than one of the previously mentioned categories. Infection was further classified as asymptomatic bacteriuria, urinary tract infection, sexually transmitted infection, and viral infection . Additionally, rates of preterm birth by subgroup were examined. As previously described,pregnancies resulting in spontaneous preterm birth were considered to be those where birth certificate or hospital discharge records indicated premature rupture of membranes , premature labor, or those for whom tocolytic medications were administered. Pregnancies resulting in provider initiated preterm births were considered to be those without PROM, premature labor or tocolytic administration for which there was a code for “induction” or “artificial rupture of membranes”; or for which there was a cesarean delivery without any of the aforementioned codes. All analyses were performed using Statistical Analysis Software version 9.4 . Methods and protocols for the study were approved by the Committee for the Protection of Human Subjects within the Health and Human Services Agency of the State of California. Data used for the study were received by the California Preterm Birth Initiative at the University of California San Francisco by June 2016. The sample included 81,021 women: 29,052 with urban residence, 24,377 with suburban residence and 27,592 with rural residence. The majority of the women in the sample were Hispanic , between 18 and 34 years at delivery , WIC participants , and multi-parous .
The demographic makeup of the three residence locations differed. For example, 8.0% of the urban population, 6.7% of suburban mothers, and 2.2% of the rural population was Black race/ethnicity . Nine percent of women in urban residences, 8.0% of women in suburban residences, and 8.2% of women in rural residences delivered preterm . Of these, 1.4% of women living in urban residences delivered before 32 weeks, while 1.1% of women in suburban or rural residences delivered this early. More specifically, 1.7% of women in the Fresno East Central MSSA delivered before 32 weeks . Four individual census tracts within urban MSSAs had rates of birth at less than 32 weeks’ gestation of 2.0% or greater, with an n of 16 or more as the reporting threshold . In the final multi-variable logistic models, Black women were found to be at elevated risk of preterm birth across all residence strata . Similarly, women with inter pregnancy intervals less than six months were at elevated risk across residence strata . Women with comorbidities such as preexisting and gestational diabetes, preexisting hypertension, and infection were also at increased risk of having a preterm birth. Other factors, such as public insurance for delivery, less than 12 years of education, underweight BMI, and an inter pregnancy interval over 59 months, were only risk factors for women living in urban residences. Only Hispanic women in rural residences were at increased risk of preterm birth . Women living in urban and rural residences who participated in WIC were less likely to deliver preterm . For urban women, birth in Mexico and overweight BMI also showed a protective effect to preterm birth . Not only did the risk models differ by residence within Fresno County, but the percentage of women with the risk varied greatly for some factors. In urban residences, 12.2% of women with preterm births smoked, while 6.6% of women in rural residences with preterm birth smoked. Similarly, 8.9% of urban women with a preterm birth used drugs or alcohol and 4.4% women in rural residences with preterm birth did.
Nearly five percent of urban women delivering preterm had fewer than three prenatal care visits and 2.3% of women in suburban residences had this few number of visits. The percent of women with a preterm birth and with inter pregnancy intervals less than six months ranged from 7.7% to 11.2% . When examining these risk factors in more geographic detail, appropriate targets for preterm birth reduction are elucidated. For instance, in six census tracts 15% or more mothers of preterm infants smoked during their pregnancy – four in urban residences and two in suburban residences . Also, five census tracts in urban residences show that over 10% of mothers who delivered preterm used drugs or alcohol . Over 2,600 women delivering in Fresno County had a cumulative risk score for preterm birth ≥ 3.0: 2.2% of women living in urban residences, 4.1% in suburban, and 3.7% in rural residences had this high risk score . In this study of preterm births in Fresno County, we found that differences in the type and magnitude of risk and protective factors differed by the residence in which women reside. Black women and women with diabetes, hypertension, infection, fewer than three prenatal care visits, previous preterm birth or inter pregnancy interval less than six months were at increased risk of preterm birth, regardless of location of residence. Public insurance, maternal education less than 12 years,plant racks for vertical growing underweight BMI, and inter pregnancy interval of five years or more were identified as risk factors only for women in urban residences. Women living in urban locations who were born in Mexico and who were overweight by BMI were at lower risk for preterm birth; WIC participation was protective for women in both urban and rural locations. Taken together, these findings suggest targeted place-based interventions and policy recommendations can be pursued. The preterm birth risk factors identified in these analyses are not unique to Fresno County: previous work has also shown that women of color, lower education, lower socioeconomic status, women with co-morbidities such as hypertension and diabetes, smoking, and short inter pregnancy interval are at elevated risk of preterm birth.In Fresno County, however, we observed that these risks differ in magnitude. This is critical, as the percentage of women in each region with the risk factor can vary greatly. Hispanic women were at increased risk of preterm birth in rural residence. The degree of risk was mild – only a 1.1-fold increase in risk. However, 72% of the population giving birth in rural Fresno County is Hispanic, suggesting that focusing interventions reaching this population may provide the most impact. Similarly,Black women were at elevated risk of preterm birth regardless of location of residence. Since urban residences have the highest percentage of Black women and rural has the lowest , focusing prevention efforts for Black women in urban residences may be an effective approach. Others have found that with pre-pregnancy initiation of Medicaid , has been associated with earlier initiation of prenatal care,a factor that may reduce preterm birth rates.In addition, participation in the WIC program also has shown a moderate reduction of the risk of a small for gestational age infant and has been associated with reduced infant mortality in Black populations.Fresno women from both urban and rural residences who participated in the WIC program were less likely to deliver preterm, while those women living in urban locations who were publicly insured through Medi-Cal coverage for delivery were at increased risk for preterm birth.
Low income is a criterion for both public assistance programs, and over 32% of families in this region lives below the poverty line;it is apparent that social economic status is a complex risk factor for preterm birth. A key take away message from this study is that women who accessed prenatal care more frequently – three or more prenatal care visits – were less likely to deliver preterm. Fresno County may be able to improve preterm birth rates by addressing factors that encourage prenatal care access, which may include enrollment in Medi-Cal during the preconception period and increasing WIC participation. Identifying regions where a high percentage of women do not access three or more prenatal care visits may suggest locations for an intervention such as home visits or mobile clinic. Using a large administrative database allows for examination of rates and risks that would not be possible with other data sources. Despite these strengths, the study has some critical limitations. By design, the findings are very specific to one area of California and may not be as applicable to other areas of the state, country, or world. In fact, we recently conducted a similar study examining preterm birth risk factors by sub-type for all of California.Our findings in Fresno County identified both similar and different risk factors for preterm birth. Similar to the entire California population, we demonstrated increased risk of preterm birth for Fresno County women who were of Black race/ethnicity, who had diabetes or hypertension during pregnancy, or who had a previous preterm birth. However, Fresno County was different from the whole state in a few ways. Unlike the state of California as a whole, Hispanic women, women over 34 years at delivery, and underweight women in urban residences in Fresno County were at increased risk for preterm birth. Also, education over 12 years did not provide protection against preterm birth in any of the Fresno County residences, although higher education did provide protection when we looked at the whole state of California. These differences point to specific pathways occurring in Fresno County that may be distinct from the state as a whole, and demonstrate the value of place-based investigation of risk factors when examining a complex outcome such as preterm birth. Other residences may benefit from similar analyses to identify risk and protective factors that are important on a local level. An additional limitation, as with most administrative databases, is that accuracy and ascertainment of variables is not easily validated. Previous studies of California birth certificate data suggests that race/ethnicity is a valid measure of self-identified race/ethnicity for all but Native Americans, and best obstetric estimate of gestation may underestimate preterm delivery rates.Previously reported rates of preterm birth in Fresno County are around 9.5% and was 8.4% overall in our population after removing multiple gestation pregnancies and pregnancies with major birth defects. Additionally, United States estimates for drug dependence/use during pregnancy is 5.0% to 5.4% and was only 2.5% in our population. This under ascertainment may mean that we are capturing the most severe diagnoses, potentially overestimating our risk calculations. Alternatively, under ascertainment also implies that drug users were likely in our referent population, which would underestimate our risk calculations. This examination of Fresno County preterm birth may provide important opportunities for local intervention. Several populations were identified as at risk, regardless of location of maternal residence, that deserve targeted interventions.