UHR Volume 12 (2007)_OCR

Introduction and Editor's Note On behalf of this year's Editorial Board, I am pleased to present to you the twelfth volume of Utah's Health: An Annual Review. To keep up with technology, great effort was put forth this year to enhance the online version of our journal. I invite the public to access it via our updated website, www.uhreview.com. Here you can not only access the information from Volume XII, but previous volumes as well. The website is now searchable, allowing you the opportunity to view individual articles based on a search for key words. As with previous editions, Volume XII includes Original Research Articles and a Data Review section. The Original Research presented here represents the work of university faculty, students, physicians, statisticians, and other professionals from the community. The Data Review section follows the journal's tradition of providing a variety of health statistical trends in Utah and the United States. It is important to note that some pages were not updated since the last publication and were therefore not included in Volume XII. While previous issues have included a section on Special Topics, we are also pleased to announce the first special supplement of Utah's Health. This stand-alone supplement focuses on the status of women's health in Utah, and is the result of collaboration between Utah's Health and the National Center of Excellence in Women's Health at the University of Utah. The full version of the supplement will be accessible on our website. For additional information, visit http://uuhsc.utah.edu/coe/womenshealth/. The successful publication of a journal involves hard work and dedication from all of those involved. I wish to thank all of the authors and contributors to this year's research articles, data pages, and special supplement. The journal would not exist without your commitment to research and other scholarly activities that focus on health-related issues in Utah. I would like to acknowledge the contributions of my fellow students who served on the Editorial Board, and extend a sincere thank you for their efforts. I would also like to thank everyone who served on this year's Advisory Board for their willingness to work along side us to ensure a quality publication. I wish to thank Dr. Richard Sperry, our faculty advisor, for his guidance and support throughout the publication process. Lastly, I extend a special thank you to our readers for their continued support. Julia Franklin Summerhays, Editor-in-Chief, Volume XII, 2007 2007 UTAH'S HEALTH: AN ANNUAL REVIEW Cesarean Deliveries and Newborn Injuries: Evidence from Linked Utah Birth Certificate and Inpatient Discharge Data Gulzar Shah, MStat, MSS, PhD, Pamela A. Clarkson Freeman, PhD, Keely Cofrin, PhD, Wu Xu, PhD Key Words: Cesarean Deliveries, Newborn Injury, Probabilistic Linkage, Inpatient Discharge, Birth Data CORRESPONDENCE: Gulzar H . Shah, PhD Director o f Research National Association o f Health Data Organizations 448 East 400 South Suite 301 Salt Lake City, Utah 84111 Phone: [phone] [e-mail]; [e-mail] Abstract Since 1999, the rate of cesarean deliveries has increased considerably in Utah. W e examined the relationship between cesarean deliveries and newborn injury among 171,114 women across the state of Utah during the years 2001-2004, using linked birth certificate and hospital discharge data. During this four-year period, there were 958 newborn injuries, yielding a rate of 6 per 1,000 live births. We found that cesarean deliveries, with or without complications, do not increase the risk of neonatal injury in Utah, particularly after the effect of other factors is statistically controlled. Risk of neonatal injuries was greater for Non-Ob/Gyn birth attendants, births with Emergency Department as a source of admission, urban hospitals, and teaching hospitals, but lower for mothers who delivered at a hospital outside their county of residence. Important policy and research implications of these findings are discussed. Introduction Neonatal injuries are an important public health concern because of their far-reaching consequences, (Siva, 2006; Visco et al., 2006; Kruszewski & Ferriero, 2005; Ferriero, 2004). The injuries could include a local infection, broken collar bone, or a head injury. Most birth injuries are minor and heal without complications. In rare cases, however, serious injuries can cause death or permanent harm, leading to lifelong suffering and cost of care to the injured child. A recent report of injuries to newborns in Utah showed large variation in birth injuries across the hospitals in the state (Office of Healthcare Statistics, Utah Department of Health [UDOH], 2006), which prompted the current study. 10 Cesarean Deliveries © 2007 The University of Utah. All Rights Reserved UTAH'S HEALTH: AN ANNUAL REVIEW The primary research question for this study is: "Are cesarean deliveries associated with elevated risk of newborn injuries?" A cesarean delivery is a birth that happens through an incision in the abdominal wall and uterus rather than through the vagina (American Pregnancy Association, 2007). While some cesarean deliveries are planned or elective, roughly one in four are unplanned and may have to be performed immediately because of certain complications that are preventing a baby from being delivered vaginally. Such procedures are often called emergency cesareans. Emergency cesarean sections vary in urgency and can be categorized into three main groups -- ‘Crash' sections, ‘Emergency' sections, and ‘Urgent' sections -- based on time frame within which a decision is made (AIMS, 2004). A large proportion of emergency cesarean are performed within 30 minute of the decision to perform the procedure. For instance, a study by Bloom and colleagues (2006) found that 23.5% of all primary cesarean deliveries were performed for an emergency indication, of which 65% began within 30 minutes of the decision to perform the procedure. Cesarean deliveries, particularly those performed on emergency basis, are a major risk factor for newborn injury (Janer et al., 2007; Alexander et al., 2006; Aburezq, Chakrabarty, & Zuker, 2005). Rate of cesarean deliveries in the United States is on the rise again, after undergoing a steady decrease from 1991 to 1996. Additionally, the primary cesarean rate has increased dramatically since 1996. Currently, over 25% of deliveries in the United States involve surgical birth via cesarean section (Declercq, Menacker, & MacDorman, 2006). The rate of "no indicated-risk" deliveries has also risen recently (Declercq, Menacker, & MacDorman, 2005). Our own research indicates that Utah's rates of cesarean deliveries are following this national trend. Cesarean deliveries are absolutely necessary in some cases to prevent complications and even to save lives. However, there is overwhelming evidence that many c-sections, performed due to non-medical factors, including attending physicians' convenience or at the mother's request, and are medically unnecessary (Bettes et. al., 2007; Baicker, Buckles, & Chandra, 2006; Gossman, Joesch, & Tanfer, 2006; Visco et al, 2006;). For instance, Kabir et al. (2005) classified 11% of all primary and 65% of all repeat cesareans in 2001 in the United States as potentially unnecessary. Indications for cesarean deliveries are well documented in literature (Lydon-Rochelle et al., 2006; Shipp et al, 2000; Gregory et al., 1998). Factors contributing to the risk of emergency cesarean deliveries include the inability of the baby to fit through the pelvic canal (cephalopelvic disproportion), malpresentation, fetal stress, elective induction of labor, uterus rupture, failed progress in labor associated with an epidural (particularly when administered early in labor), and severe labor pain (Klien, 2005; Vrouenraets et al., 2005; Maslow & Sweeny, 2000; Hess et al., 2000; Albright & Forster, 1997; Leiberman et al., 1996; Neuhoff, Burke, & Porreco, 1989). Other factors associated with emergency cesarean section are increased maternal age, lower maternal height, male fetus, no previous vaginal birth, and greater gestational age (Smith et al., 2005). © 2007 The University of Utah. All Rights Reserved Cesarean Deliveries 11 2007 UTAH'S HEALTH: AN ANNUAL REVIEW The existing body of research literature supports the notion that cesarean deliveries are associated with unfavorable outcomes for the mother and newborn including elevated risk of mortality (Silver et al., 2006; MacDorman et al., 2006; Harper et al., 2003; Lal, Callender, & Radley, 2003). Cesarean deliveries particularly increase the risk of injury to the newborn, including iatrogenic fetal injury and such injuries are often underreported (Aburezq, Chakrabarty, & Zuker, 2005). Risk of lung injury to the newborn is also associated with cesarean deliveries (Janer et al., 2007). Cesarean sections are associated with the risk of cuts to baby during delivery (Wiener & Westwood, 2002). At a hospital in Newport, it was found that 1.5 % of the babies delivered via cesarean sections had skin lacerations (Van-Ham, van Dongen, & Mulder, 1997). Although a skin cut may not be something of great consequence, these cuts can cause infections and other complications. The risk of neonatal injuries is even greater for emergency cesarean deliveries (Aburezq, Chakrabarty, & Zuker, 2005; Dessole, 2004; Okaro & Anya, 2004; Wiener & Westwood, 2002; Rydhstrom et al., 1998). A recent study examining a large sample of 37,110 cesarean deliveries concluded that although cesarean deliveries posed a risk of injury to 1.1% of newborns, the risk was far greater for cesarean deliveries that were conducted on emergency basis, particularly when primary cesarean deliveries had a failed forceps or vacuum attempt and when the type of uterine incision was "T " or "J", as opposed to a vertical incision. The most common injury was a skin laceration. Other injuries included cephalohematoma, clavicular fracture, brachial plexus, skull fracture, and facial nerve palsy (Alexander et al., 2006). The rate of intracranial hemorrhage (bleeding within skull) to newborns has also been found to be higher in cases of emergency cesarean sections as opposed to a vaginal or elective cesarean delivery (Towner et al., 1998). When delivered by emergency cesarean section, a baby is more likely to have breathing and respiratory difficulties (Annibale et al., 1995). In addition to injury, babies born by emergency cesarean section are 50% more likely to have lower APGAR scores than those born vaginally (Burt, Vaughan & Daling, 1988). While it is possible that breathing difficulties and low APGAR scores may have been a function of complications leading to the emergency cesarean, researchers believe that cesarean emergency cesarean deliveries are certainly a contributory factor (see, e.g., Burt, Vaughan & Daling, 1988). Increased awareness regarding the use of cardiotocography (CTG) and Doppler ultrasonography (US) for diagnosis of complications such as brain death syndrome may prevent unnecessary emergency cesarean section (Chen et al., 2006). Data and Methods A unique feature of the data used in this research is that record-level probabilistic linkage of two unique databases was performed using the software AutoMatch (MatchWare Technologies Inc.). For this, we obtained 2001 through 2004 Utah Birth Certificate Data and the Utah Inpatient Hospital Discharge Data from the Utah Department of Health. The former database is maintained by the Office of Vital Records and Statistics, and the 12 Cesarean Deliveries © 2007 The University of Utah. All Rights Reserved UTAH'S HEALTH: AN ANNUAL REVIEW latter is maintained by the Office of Health Care Statistics. The Inpatient Discharge Data includes information related to diagnosis codes, procedure codes, Diagnosis Related Group (DRG), and provider identification (i.e. hospital, county). Information related to birth parents, birth methods, complications, place of birth, type of attendant, and antepartum procedures is available in the Birth Certificate Data. In order to make the comparisons across various categories of deliveries, we removed records involving multiple births. After this exclusion, a total of 171,114 women, whose hospital discharge records were matched with their baby's birth records, were included in this analysis. The outcome of interest, or dependent variable, for our research was ‘Birth Trauma - Injury to Neonate' or injuries that happen to the newborn baby during birth. The operational definition of the birth injury was adopted from a standard measure -- Patient Safety Indicator 17 -- "Birth Trauma: Injury to Neonate, PSI 17"-- developed by federal Agency for Healthcare Research and Quality (AHRQ, 2006). A discharge record was flagged as having an injury if it contained a diagnosis field with any one of the following International Classification of Diseases, Version 9, Clinical Modification (ICD-9-CM) codes: 767.0 - Subdural and cerebral hemorrhage 767.11 - Epicranial subaponeurotic hemorrhage (massive) 767.3 - Injuries to skeleton (excludes clavicle) 767.4 - Injury to spine and spinal cord 767.7 - Other cranial and peripheral nerve injuries 767.8 - Other specified birth trauma 767.9 - Birth trauma, unspecified Linked records for all live births were used as the denominator for computing rates. While the use of a patient safety indicator implies that the injury could have been prevented, some of the birth injuries flagged using PSI-17 may be non-preventable. For instance, some instances of eye damage are complications of a long or difficult labor. Injuries included in this category can also be unspecified. If the practitioner describes an injury that does not have a code, or does not specify what the injury was, it will also be included in this category. For our research, a newborn was identified as any neonate with either: (a) an ICD-9-CM diagnosis code for an in-hospital liveborn birth, or (b) an admission type of newborn, age in days at admission equal to zero, and not having an ICD-9-CM diagnosis code for an out-of-hospital birth. The primary independent variable for our research was the interaction of two dummy variables: "Whether the delivery involved complications-based on the DRG", and "Whether the delivery method was cesarean". This variable has the following attributes: (a) Cesarean delivery with complication, (b) Cesarean delivery without complication, (c) Vaginal birth. In order to control for other factors associated with our dependent variable "neonatal injuries," we included several control © 2007 The University of Utah. All Rights Reserved Cesarean Deliveries 13 2007 UTAH'S HEALTH: AN ANNUAL REVIEW variables such as hospital location and teaching status, patient county of residence, severity level, mortality risk index, and other relevant variables from the birth certificate. The patient severity level and mortality risk index were derived from the Inpatient Discharge Data using the 3M All Patient Refined (APR) DRG software. To examine bivariate differences, we used chi-square tests of significance. Multivariable analysis were performed using logistic regression with 'Injury to neonate (PSI7)' as the dependent variable and 'cesarean delivery' as independent variable with a variety of control variables in the model. Results Our analysis included 171,114 records where we were successful in linking the discharge record for a newborn with that of his/her mother. The average age of the mother at the time when discharged was 26.7 years and the average birth weight was 3,321 grams or 7.3 pounds. Figure 1: Trend in Cesarean Delivery Rate, Utah Hospital Discharge Data, 1996-2005 Note: these are overall cesarean rates, and may differ slightly from those reported in Table 1, which involve exclusions of multiple births from analysis. In recent years, rates of cesarean sections in Utah have increased (Figure 1), as was the case for National rates, highlighted in the review of literature. These rates hovered around 16 per 100 deliveries from 1997 to 1999, after which a sharp and steady increase occurred from 16.6 per 100 deliveries in the year 2000, to 20.7 in 2004, an increase of over 4 per 100 deliveries during a short period of 4 years. The rate of birth trauma was 6 per 1,000 live births, which amounts to 958 births with trauma during the time period 2001-2004. Nearly 82% of all births were vaginal. Table 1 shows the frequency and percentage distribution of discharges by their various characteristics. The overall rate of cesarean deliveries was 18.1% (for singleton births). Over 19% (5,894) of all c-sections involved complications while 81% or 25,238 cesarean deliveries had no complications. As in the case of overall cesarean deliveries (Figure 1), there is a substantial increase in cesarean 14 Cesarean Deliveries © 2007 The University of Utah. All Rights Reserved UTAH'S HEALTH: AN ANNUAL REVIEW deliveries involving singleton births as well (Table 1). Only a small proportion (1.3%) of births had Emergency Department (ED) as the source of admission. Demographic characteristics of the patients as well as their geographic location are likely to have a bearing on the primary outcomes of interest to this research, namely, the neonatal injury at birth and the cesarean deliveries. Approximately 18% of mothers migrated out of their county for childbirth. About 82% of the births occurred in urban hospitals and about 33% of all deliveries occurred in teaching hospitals. A large proportion (89%) of deliveries had a Medical Doctor (MD) as the birth attendant. The attending physician's specialty was Obstetrics/Gynecology (Ob/Gyn) for 55% of all deliveries, while the remaining 45% were attended by family physicians or others. Risk Factors for Neonatal Injury - A Bivariate Analysis Table 2 shows that the risk of neonatal injuries was slightly but statistically non-significantly greater for cesarean deliveries with complications. The greatest risk of neonatal injury was associated with the mother's severity level and risk of mortality. Mothers with a risk of mortality at level 3 had injury rates of 18 per 1,000 deliveries. Statistically, migration out of county for hospital care was the strongest predictor of injury, with those migrating having a lower risk of neonatal injury. Urban location of hospitals was the next most important risk factor for the injuries, with 6 injuries per 1,000 births at urban hospitals, compared with 2 per 1,000 births in rural hospitals. Similar disparity existed by attending physician's specialty. The rate of neonatal injury was 3 per 1,000 births for the Ob/Gyn, whereas for other specialties it was 9 neonatal injuries per 1,000 live births. The risk of injuries was higher for births in teaching hospitals -- 9 per 1,000 live births for the teaching hospitals as opposed to 4 per 1,000 live births for the non-teaching hospitals. This difference was statistically significant. There was no difference in risk of injury by whether the birth attendant was an MD or other practitioner. Deliveries with the source of admission as ED had also higher rate of injury. The proportion of births with injuries was 16 per 1,000 deliveries for deliveries with ED as source of admission. This was more than double the rate for births with non-ED as source of admission. However, this difference was not statistically significant because of small cell size for the deliveries with ED as a source of admission. Risk factors for Neonatal Injury - Results of Logistic Regression Analysis Table 3 shows the results of logistic regression analysis. Consistent with the bivariate analysis, results of the logistic regression analyses show that the risk of neonatal injury is not significantly greater in cesarean deliveries when compared with vaginal deliveries, particularly after other factors are statistically controlled. Among the covariates, the most important determinant of neonatal injury, after controlling for other factors, was whether the specialty of the attending physician was Obstetrics/Gynecology. Births attended by a Non-Ob/Gyn © 2007 The University of Utah. All Rights Reserved Cesarean Deliveries 15 2007 UTAH'S HEALTH: AN ANNUAL REVIEW were 2.6 times more likely to involve neonatal injuries compared to those attended by an Ob/Gyn. Urban status of a hospital was the next most important factor in predicting neonatal injury, with the odds of injury twice as high in urban hospitals as compared to rural hospitals. Conversely, urban or rural status of mothers' county of residence was not associated with risk of neonatal injury. Migration of mother out of her county of residence was, on the other hand, significantly associated with reduced risk of neonatal injury; non-migrants were 1.3 times more likely to have neonatal injury. Teaching hospitals were associated with approximately twice the risk of injury as in non-teaching hospitals, even after adjusting for other risk factors such as patient severity and risk of mortality. Births attended by MDs and those with ED as a source of admission were associated with a greater risk of injury (1.47 and 1.46, respectively). This difference could be attributable partially to the difference in mix of severity and complications, though some of this difference was statistically controlled. After controlling for all other variable in the model, both ‘risk of mortality' and ‘level of severity' were significantly associated with the risk of neonatal injury. Discussion and Conclusions The primary concern of this research was to determine if cesarean deliveries are associated with an elevated risk of birth injury in Utah. Cesarean deliveries, particularly those with complications, were thought be an important contributor to such injuries. In an effort to provide evidence-based recommendations for quality improvement in maternal care, we examined the relationship between cesarean deliveries and newborn injury among 171,114 women across the state of Utah from 2001 to 2004. Linked birth certificate and hospital discharge data pertaining to women with single births were analyzed, with the relationship between cesarean deliveries and newborn injury examined using logistic regression analysis. Our analyses revealed that since 1999, rates of cesarean sections have undergone a noticeable increase in Utah, rising from 16.0 per 100 deliveries in 2000 to 20.8 in 2005. As was the case in overall cesarean deliveries, there is a substantial increase in cesarean deliveries involving singleton births as well. Healthcare policy can benefit from future studies in Utah, examining the patterns of trial of labor in repeat cesareans, elective cesareans including those upon maternal request, and other reason for the increase in cesarean deliveries. Such studies may require the use of more sophisticated data sources than administrative Hospital Inpatient Abstracts, including chart reviews, key informant interviews, and in-depth surveys. From 2001 to 2004 there were 958 newborn injuries. This amounts to a rate of birth trauma of 6 per 1,000 live births. This is substantially lower than the injury rates reported in other large studies (Alexander et al., 2006). Contrary to many previous studies, we found that cesarean deliveries, with or without complications, were not associated with neonatal injury in Utah, particularly after the effect of other factors is statistically controlled. This is an important finding given the sharp rise in cesarean deliveries in Utah as well as nationally. Among the other factors, risk of injury was 2.6 times greater for Non-Ob/Gyn birth attendants than Ob/Gyn. Births with ED as a source of admission were also more often associated with injury. The fact that MDs, urban hospitals and teaching 16 Cesarean Deliveries © 2007 The University of Utah. All Rights Reserved UTAH'S HEALTH: AN ANNUAL REVIEW hospitals had a higher risk may still be partially attributable to "sicker patients," because the patient severity level used in this study was a simplified measure of the APR-DRG software. Coding variations among urban or rural, teaching or non-teaching hospitals might also affect the analytical results. Migration of mother out of her county of residence was associated with reduced risk of neonatal injury. A detailed analysis of migration status revealed that the migration mostly occurred to large urban counties of hospital location, with migration to hospitals in Salt Lake County being highest (41.6%), followed by Weber County (22.3%). Most of the migrants were either from small rural counties or adjacent urban counties. The primary reason for the migration for nearly all cases was physician referral (97.7%). Only a small proportion of mothers, i.e. 1.3%, with source of admission as Emergency Department migrated from their county of residence to a hospital with a different county. Although a mother may travel out of her county because of severity of illness, a complicated pregnancy, or the lack of a hospital in her county, the positive outcome may imply that such mothers are better informed consumers, or in some cases, mothers crossing the county boundary may have less risky pregnancy, giving them more time to travel across county lines. The migration may not have occurred primarily because of shortage of ObGyn in rural areas because the difference between rural and urban counties in percentage of ObGyn as birth attendants was minimal, 53% and 56% respectively. The variables in the model resulted in a combined R-Squared (Nagelkerke R-Squared) of 0.056, indicating that these variables explained a small percent of variation in the dependent variable, "Injuries to newborns." Since we selected from the most significant of the explanatory variables available in the data, the low R-Squared implies that the important determinants of the neonatal injury are not available from the hospital discharge data or the birth certificate data. We recommend further research with other sources of data to improve understanding of factors contributory to neonatal injury in Utah. Although limited in scope by the nature of the data, our research has some important policy implications. First, record-level linkage of administrative data such as hospital discharge data, with other data on the same individuals (e. g., birth certificate data), offers the opportunity to answer research questions not possible from a single data source. Secondly, although they do not appear to significantly increase risk of neonatal injuries, the recent rise in cesarean deliveries is a matter of concern because of its other potential unintended effects on maternal health outcomes, including complication of surgery such as post-surgical wound infections. Further research is imperative to investigate reason for such increase. Acknowledgments This research was funded by Agency for Healthcare Research and Quality (AHRQ) grant # R24 HS11844-05 "Intermountain BRIC Consortium." The linkage of data was performed under the CDC Assessment Initiative Cooperative agreement. We wish to acknowledge Denise Love, Executive Director NAHDO for her valuable guidance and contribution during design and formative stage of this study. © 2007 The University of Utah. All Rights Reserved Cesarean Deliveries 17 2007 UTAH'S HEALTH: AN ANNUAL REVIEW Table 1. Characteristics of discharges related to delivery and birth, Utah, 2001-2004 (N=171,114) Characteristic Number Percent Newborn Birth Trauma? No 170156 99.4 Yes 958 0.6 Type of Delivery Cesarean Section C-Section with Complications 5894 3.4 C-Section without complications 25238 14.7 Vaginal Deliveries 139982 81.8 Trend in Cesarean Deliveries1 2001 6,719 16.2 2002 7,646 18.0 2003 8,082 18.6 2004 8,685 19.8 County of Hospital's Location Rural 30934 18.9 Urban 140180 81.9 Mother's County of Residence Rural 36871 22.0 Urban 130720 78.0 Type of Birth Attendant Midwife or Other 19066 11.1 MD 152048 88.9 Attending Physicians Specialty Family Practitioner or Other 76681 44.8 ObGyn 94433 55.2 Was the Hospital a Teaching Hospital? No 115381 67.4 Yes 55733 32.6 Source of Admission Other 168832 98.7 Emergency Room 2282 1.3 Did the mother migrate across county for care? No 140295 82.0 Yes 30725 18.0 1 The rates are only for singleton births; the overall cesarean rates for all deliveries are presented in Figure 1. 18 Cesarean Deliveries © 2007 The University of Utah. All Rights Reserved UTAH'S HEALTH: AN ANNUAL REVIEW Table 2.Prevalence of Birthing Characteristics Across Women who Experienced Neonatal Injuries (PSI 17) During Delivery, Utah, 2001-2004 (N=171,114) Birthing Characteristic Neonatal Injury occurred? X 2 No Yes (n=170,156) (n=958) County of Hospital's Location Urban 99.4 0.6 69.73** Rural 99.8 0.2 Mother's County of Residence 57.63** Urban 99.4 0.6 Rural 99.7 0.3 Is the type of hospital a teaching hospital? No 99.6 0.4 216.79** Yes 99.1 0.9 Attending Physicians Specialty Family Practitioner or Other 99.1 0.9 210.50** ObGyn 99.7 0.3 Type of Birth Attendant - MD vs. Other 0.56 Midwife or others 99.4 0.6 MD 99.4 0.6 Migrated Across County? No 99.4 0.6 1.14* Yes 99.5 0.5 Source of Admission 43.03** Emergency room Other 98.4 99.5 1.6 0.5 Method of Delivery C-Section with complications C-Section without complications Vaginal Delivery 99.3 99.4 99.4 0.7 0.6 0.6 1.16 Level of Severity 1 99.5 0.5 142.00** 2 98.9 1.1 3 99.0 1.0 4 98.8 1.2 Risk of Mortality 1 99.5 0.5 131.01** 2 98.3 1.7 3 98.2 1.8 4 99.2 0.8 **p<0.01; *p<0.05 © 2007 The University of Utah. All Rights Reserved Cesarean Deliveries 19 2007 UTAH'S HEALTH: AN ANNUAL REVIEW Table 3. Summary of Logistic Regression Analysis Predicting Neonatal Trauma, 2001-2004 Table 3. Summary of Logistic Regression Analysis Predicting Neonatal Trauma, 2001-2004 Variable B SE Odds Ratio Wald statistic Hospital County 0.73 0.21 2.08 11.74* Mother's County of Residence 0.001 0.18 1.00 0.0 Migration Status -0.27 0.11 0.76 6.17* Teaching 0.69 0.07 1.98 92.46* Delivery by OB/GYN -0.95 0.07 0.39 171.05* MD was Birth Attendant 0.39 0.11 1.47 13.68* Mother's Age -0.04 0.007 0.96 41.46* Severity Level 1 2 3 4 -0.16 0.50 -0.88 0.369 0.366 0.350 0.85 1.66 0.92 61.45* 0.19 1.90 0.63 Risk for Mortality 1 2 3 4 -0.577 0.380 0.691 0.763 0.761 0.753 0.56 1.46 2.00 44.84* 0.572 0.249 0.841 Birth Weight in Grams 0.00 0.00 1.00 41.86* Source of Admission 0.38 0.18 1.46 4.53** Type of Delivery C-section with complication C-section without complications Vaginal deliveries -0.074 -0.14 0.170 0.094 0.905 0.928 0.986 0.201 0.191 0.022 * =p<0.01; ** =p<0.05; --- = reference category 20 Cesarean Deliveries © 2007 The University of Utah. All Rights Reserved UTAH'S HEALTH: AN ANNUAL REVIEW APPENDIX A: Definition of Variables Birth Trauma -Injury to Neonate (PSI 17) Definition: Cases of birth trauma, injury to neonate, per 1,000 live born births. Numerator: Discharges with ICD-9-CM codes for birth trauma in any diagnosis field. ICD-9-CM Birth Trauma diagnosis codes: 7670 SUBDURAL AND CEREBRAL HEMORRHAGE (DUE TO TRAUMA OR TO INTRAPARTUM ANOXIA OR HYPOXIA) 767.11 EPICRANIAL SUBAPONEUROTIC HEMORRHAGE (MASSIVE) (OCT 03) 7673 INJURIES TO SKELETON (EXCLUDES CLAVICLE) 7674 INJURY TO SPINE AND SPINAL CORD 7677 OTHER CRANIAL AND PERIPHERAL NERVE INJURIES 7678 OTHER SPECIFIED BIRTH TRAUMA 7679 BIRTH TRAUMA, UNSPECIFIED Note: Because 767.1 was not previously included in the numerator specification, the addition of 767.11 may cause an increase in the rate. Exclude: 1. Infants with a subdural or cerebral hemorrhage (subgroup of birth trauma coding) and any diagnosis code of pre-term infant (denoting birth weight of less than 2,500 grams and less than 37 weeks gestation or 34 weeks gestation or less). 2. Infants with injury to skeleton (767.3, 767.4) and any diagnosis code of osteogenesis imperfecta (756.51). ICD-9-CM Preterm Infant diagnosis codes: 76501 EXTREME IMMATURITY, LESS THAN 500 GRAMS 76502 EXTREME IMMATURITY, 500 - 749 GRAMS 76503 EXTREME IMMATURITY, 750 - 999 GRAMS 76504 EXTREME IMMATURITY, 1000 - 1249 GRAMS 76505 EXTREME IMMATURITY, 1250 - 1499 GRAMS 76506 EXTREME IMMATURITY, 1500 - 1749 GRAMS 76507 EXTREME IMMATURITY, 1750 - 1999 GRAMS 76508 EXTREME IMMATURITY, 2000 - 2499 GRAMS 76511 O THER PRETERM INFANTS, LESS THAN 500 GRAMS 76512 O THER PRETERM INFANTS, 500 - 749 GRAMS 76513 O THER PRETERM INFANTS, 750 - 999 GRAMS 76514 OTHER PRETERM INFANTS, 1000 - 1249 GRAMS 76515 O THER PRETERM INFANTS, 1250 - 1499 GRAMS © 2007 The University of Utah. All Rights Reserved Cesarean Deliveries 21 2007 UTAH'S HEALTH: AN ANNUAL REVIEW 76516 O THER PRETERM INFANTS, 1500 - 1749 GRAMS 76517 OTHER PRETERM INFANTS, 1750 - 1999 GRAMS 76518 O THER PRETERM INFANTS, 2000 - 2499 GRAMS 76521 LESS THAN 24 COMPLETED WEEKS OF GESTATION 76522 24 COMPLETED WEEKS OF GESTATION 76523 25-26 COMPLETED WEEKS OF GESTATION 76524 27-28 COMPLETED WEEKS OF GESTATION 76525 29-30 COMPLETED WEEKS OF GESTATION 76526 31-32 COMPLETED WEEKS OF GESTATION 76527 33-34 COMPLETED WEEKS OF GESTATION Denominator: All liveborn births. Liveborn DRGs: 385 NEONATES, DIED OR TRANSFERRED TO ANOTHER ACUTE CARE FACILITY 386 EXTREME IMMATURITY OR RESPIRATORY DISTRESS SYNDROME OF NEONATE 387 PREMATURITY W / MAJOR PROBLEMS 388 PREMATURITY W/O MAJOR PROBLEMS 389 FULL TERM NEONATE W / MAJOR PROBLEMS 390 NEONATE W / OTHER SIGNIFICANT PROBLEMS 391 NORMAL NEWBORN 22 Cesarean Deliveries © 2007 The University of Utah. All Rights Reserved UTAH'S HEALTH: AN ANNUAL REVIEW References Aburezq, H., Chakrabarty, K. H ., & Zuker, R. M. (2005). Iatrogenic fetal injury. Obstetetrics and Gynecology, 106(5 P t 2), 1172-1174. The Agency for Healthcare Research and Quality (2006). Quality Indicators (QIs): Patient Safety Indicators Overview. Retrieved Sept 2006 from http://www.qualityindicators.ahrq.gov/psi_overview.htm. AIMS. (2002). Emergency cesareans- How long does it take. AIMS Journal, 4(1), Retrieved Dec 2006 from http://www.aims.org.uk/Journal/Vol14No1/resspring2002.htm. Albright, G. A., & Forster, R. M. (1997). Does combined spinal-epidural analgesia with subarachnoid sufentanil increase the incidence of emergency cesarean delivery? Regional Anesthesia, 22, 400-405. Alexander, J. M., Leveno, K. J., Hauth, J., Landon, M. B., et al. (2006). Fetal injury associated with cesarean delivery. Obstetrics and Gynecology, 108(4), 885-890. American Pregnancy Association. (2007). Reasons fo r a Cesarean Birth, 2006. Retrieved Jan 02, 2007 from http://www.americanpregnancy.org/labornbirth/reasonsforacesarean.html. Annibale, D. J., Hulsey, T. C. , Wagner, C. L., & Southgate, W . M. (1995). Comparative neonatal morbidity of abdominal and vaginal deliveries after uncomplicated pregnancies. Archives o f Pediatric and Adolescent Medicine, 149(8), 862-867. Baicker, K., Buckles, K. S., & Chandra A. (2006). Geographic variation in the appropriate use of cesarean delivery. Health Affairs, 25(5), w355 - w367. Bettes, B. A., Coleman, V. A., Zinberg, S., Spong, C. Y. et al. (2007). Cesarean Delivery on Maternal Request: Obstetrician- Gynecologists' Knowledge, Perception, and Practice Patterns. Obstetrics and Gynecology, 109, 57-66. Bloom, S, L., Leveno, K. J., Spong, C. Y., Gilbert, S., et al.(2006). Decision-to-incision times and maternal and infant outcomes. Obstetrics and Gynecology, 108(1), 6-11. Burt, R. D., Vaughan, T. L., & Daling, J. R. (1988). Evaluating the risks of cesarean section: low Apgar score in repeat C-section and vaginal deliveries. American Journal o f Public Health, 78(10), 1312-1314. Chen, Y. T., Hsu, S. T., Tseng. J. J., Chen, W . C., Ho, E. S., & Chou, M . M. (2006). Cardiotocographic and Doppler ultrasonographic findings in a fetus with brain death syndrome. Taiwan Journal of Obstetrics and Gynecology, 45(3), 279-282. Declercq, E., Menacker, F., & MacDorman, M. (2006). Maternal risk profiles and the primary cesarean rate in the United States, 1991- 2002. American Journal o f Public Health, 96, 867-872. Declercq, E., Menacker, F., & MacDorman, M. (2005). Rise in "no indicated risk" primary cesareans in the United States, 1991-2001. British MedicalJournal, 330, 71-72. Dessole, S., Cosmi, E., Balata, A., Uras, L., Caserta, D., Capobianco, G., & Ambrosini, G. (2004). Accidental fetal lacerations during cesarean delivery: experience in an Italian level III university hospital. American Journal o f Obstetrics and Gynecology, 191(5), 1673- 1677. Ferriero, D. M. (2004). Neonatal brain injury. The N ew England Journal o f Medicine, 351, 1985-1995. Gossman, G. L., Joesch, J. M ., & Tanfer, K. (2006). Trends in Maternal Request Cesarean Delivery From 1991 to 2004. Obstetrics and Gynecology, 108, 1506-1516. Gregory, K. D., Curtin, S. C., Taffel, S. M., & Notzon, F. C. (1998). Changes in indications for cesarean delivery: United States, 1985 and 1994. American Journal o f Public Health, 88, 1384-1387. Harper, M.A., Byington, R. P., Espeland, M .A., Naughton, M ., Meyer, R., & Laneb, K. (2003).Pregnancy-Related Death and Health Care Services. Obstetrics and Gynecology, 102, 273-278. Hess, P. E., Pratt, S. D., Soni, A. K., Sarna, M. C., & Oriol, N .E. (2000). An association between severe labor pain and cesarean delivery. Anesthesia and Analgesia, 90(4), 881-886. © 2007 The University of Utah. All Rights Reserved Cesarean Deliveries 23 2007 UTAH'S HEALTH: AN ANNUAL REVIEW Janer, J., Andersson, S., Haglund, C., & Lassus, P. (2007). Pulmonary endostatin perinatally and in lung injury of the newborn infant. Pediatrics, 119(1), e241-e246. Kabir, A. A., Pridjian, G., Steinmann, W . C., Herrera, E. A., & Khan, M . M. (2005). Racial differences in cesareans: an analysis of U.S. 2001 National Inpatient Sample Data. Obstetrics and Gynecology, 105, 710-718. Lal, M., Mann, C., Callender, R., & Radley, S. (2003). Does cesarean delivery prevent anal incontinence? Obstetrics and Gynecology, 101(2), 305-312. Lieberman, E., Lang, J. M., Cohen, A., et al. (1996). Association of epidural analgesia with cesarean delivery in nulliparas. Obstetrics and Gynecology, 88, 993-1000. MacDorman, M. F., Declercq, E., Menacker, F., & Malloy, M. H. (2006). Infant and neonatal mortality for primary Cesarean and vaginal births to women with "No indicated risk," United States, 1998-2001 birth cohorts. Birth: Issue in Perinatal Care, 33(3), 175-182. Klein, M. C. (2005). Early epidurals increase cesarean rate, meta-analysis shows. British Medical Journal, 330, 790-790. Kruszewski, S. P., & Ferriero, D. M. (2005). Neonatal Brain Injury. New EnglandJournal o f Medicine, 352, 839-839. Lydon-Rochelle, M. T., Gardella, C., Cardenas, V., & Easterling, T. R. (2006). Repeat cesarean delivery: what indications are recorded in the medical chart? Birth, 33(1), 4-11. Maslow, A. S., & Sweeny, A. L. (2000). Elective induction of labor as a risk factor for cesarean delivery among low-risk women at term. Obstetrics and Gynecology, 95, 917. Neuhoff, D., Burke, M . S., & Porreco, R. P. (1989). Cesarean birth for failed progress in labor. Obstetrics and Gynecology, 73, 915-920. Okaro, J. M ., & Anya, S. E. (2004). Accidental incision o f the fetus at caesarian section. Nigerian Journal o f Medicine, 13(1), 56-58. Rydhstrom, H ., Ryding, L. E., Wijma, B., & Wijma K. (1998). Fear of childbirth during pregnancy may increase the risk of emergency cesarean section", Acta obstetricia etgynecologica Scandinavica, 77, 542-547. Shipp, T. D., Zelop, C. M., Repke, J. T., Cohen, A., Caughey, A. B., & Lieberman, E., (2000). Labor after previous cesarean: influence of prior indication and parity. Obstetrics and Gynecology, 95(6), 913-916. Silver, R. M., Landon, M. B., Rouse, D. J., Leveno, K. J., et al. (2006). Maternal morbidity associated with multiple repeat cesarean deliveries. Obstetrics and Gynecology, 107(6), 1226-1232. Towner, D., Castro, M ., Eby-Wilkens, E., & Gilbert W. M. (1999). Effect o f mode of delivery in nulliparous women on neonatal intracranial injury. The New England Journal of Medicine, 341, 1710-1714. Utah Department of Health. (2006). 2006 Hospital Comparison Report on Maternity and Newborns. Office of Healthcare Statistics, Utah Department of Health (UDOH). Retrieved Jan 07, 2007 from http://health.utah.gov/myhealthcare/reports/maternity2006/. Silva, A. M., Smith, R. N., Lehmann, C. U., Johnson, E. A. Holcroft, C. J., & Graham, E. M.(2006). Neonatal nucleated red blood cells and the prediction of cerebral white matter injury in preterm infants. Obstetrics and Gynecology, 107(3), 550 - 556. Van Ham, M. A., van Dongen, P. W ., & Mulder, J. (1997). Maternal consequences o f cesarean section. A retrospective study of intraoperative and postoperative maternal complications o f cesarean during a 10-year period. European Journal o f Obstetrics, Gynecology and Reproductive Biology, 74 (1), 1-6. Visco, A. G., Viswanathan, M ., Lohr, K. N., Wechter, M . E., Gartlehner, G., W u, J. M., Palmieri, R., Funk, M . J., Lux, L., Swinson, T ., & Hartmann, K. (2006). Cesarean delivery on maternal request: maternal and neonatal outcomes. Obstetrics and Gynecology, 108, 1517-1529. Vrouenraets, F. P. J. M., Roumen, F. J. M. E., Dehing, C. J. G. , van den Akker, E. S. A., Aarts, M . J. B., & Scheve, E. J. T. (2005). Bishop Score and risk of cesarean delivery after induction of labor in nulliparous women. Obstetrics and Gynecology, 105(4), 690- 697. Wiener, J., & Westwood, J. (2002). Fetal lacerations at cesarean section. Journal o f Obstetrics and Gynecology, 22, 23-24. 24 Cesarean Deliveries © 2007 The University of Utah. All Rights Reserved UTAH'S HEALTH: AN ANNUAL REVIEW Characterization of Women with Hysterectomy Ray M. Merrill, PhD, MPH, Whitney A. Johnson, MPH Department of Health Science, College of Health and Human Performance, Brigham Young University CORRESPONDENCE: Ray M. Merrill, PhD, Professor Department o f Health Science, Brigham Young University 229-A Richards Building Provo, Utah 84064 Phone: [phone] Fax: [phone] [e-mail] Abstract This study describes the association between hysterectomy and demographic characteristics, reproductive history and hormonal contraception use in Utah. Data were obtained from a cross-sectional telephone survey developed and conducted during March and April, 2002. Analyses involve 926 women. Twenty-seven percent of women surveyed reported having had a hysterectomy. Age was directly associated with hysterectomy. After adjusting for age, hysterectomy was not associated with race, tobacco and alcohol use, yearly income, and education. However, hysterectomy prevalence significantly decreased with older age at first pregnancy, older age at menarche, older age at starting oral contraceptive use, and with increased duration of breastfeeding. In conclusion, there is a significant relationship between the prevalence of hysterectomy and variables associated with lower circulating estrogen levels, such as breastfeeding, and delayed commencement of menarche, pregnancy, and oral contraceptive use. Introduction Hysterectomy, a surgical procedure to remove the uterus, is the second most common surgery performed each year in the United States, behind cesarean section delivery (NCHS, 2004). During 1994-1998, one in nine women between the ages of 35-45 years had a hysterectomy in the United States. Women between the ages of 40-44 have the highest rate of hysterectomy compared with other age groups (Keshavarz, Hillis, & Kieke, 2002). Reasons why women undergo a hysterectomy include uterine fibroids, endometriosis, abnormal uterine bleeding, and cancer. Uterine fibroids are the primary reason for hysterectomies, accounting for approximately 31% of all hysterectomies (Keshavarz, Hillis, & Kieke, 2002; Wilcox et al., 1994; Weaver et al., 2001). Abnormal uterine bleeding is the underlying cause of 14% of hysterectomies and endometriosis accounts for 11% (Weaver et al., 2001). Cancers affecting the reproductive organs (endometrial cancer, cervical cancer, and cancer of the ovaries or © 2007 The University of Utah. All Rights Reserved Women With Hysterectomy 25 2007 UTAH'S HEALTH: AN ANNUAL REVIEW fallopian tubes) explain about 10% of all hysterectomies (Rowe et al., 1999). Hysterectomies are also performed because of chronic pelvic pain, uterine prolapse, and chronic pelvic inflammatory disease. Uterine fibroids are benign tumors that grow on or within the smooth muscle tissue of the uterus. Although the cause of uterine fibroids is unclear, fibroids enlarge with increasing levels of circulating estrogen and progesterone, which may result from oral contraceptive use or pregnancy (Nowak, 1999; Barbieri, 1997; Beral, Bull, & Reeves, 2005). Heavy menstrual bleeding, clotting, and pelvic pain often result from uterine fibroids and lead a woman to seek treatment. African American women have a higher prevalence of uterine fibroids compared with Caucasian American women (Kjerulff et al., 1993; Baird et al., 2003). This likely explains why African American women have a higher prevalence of hysterectomy (Meilahn et al., 1989). Women who use oral contraceptives over an extended period of time are at increased risk of cervical cancer (Lacey et al., 1999). On the other hand, oral contraceptives may protect against endometrial cancer, ovarian cancer, and pelvic inflammatory disease (There's good news about birth control pills, 1992). A combination of estrogen and progestin, as found in oral contraceptives, is sometimes used to treat endometriosis and heavy bleeding during or between menstrual periods (Van Gorp & Neven, 2002). A younger age at first full term pregnancy has been associated with an increased risk of cervical cancer (La Vecchia et al., 1993; Kvale, Heuch, & Nilssen, 1988; Mogren, Stenlund, Hogberg, 2001; Guo et al., 1994; Bjorge & Kravdal, 1996) and endometrial cancer (Lochen & Lund, 1997; Parslov et al., 2000; Hinkula et al., 2002; Parazzini et al., 1991; Lambe et al., 1999). Uterine prolapse and chronic pelvic pain are commonly associated with damage to the pelvic floor during vaginal deliveries, instrumental deliveries, and vaginal delivery of large babies (Swift et al., 2005; Coleman, 2005; Davis & Kumar, 2003). Advancing age and menopause can weaken the pelvic floor (Swift et al., 2005; Davis & Kumar, 2003; What to do about pelvic organ prolapse, 2005). The effects from age and menopause may result from decreased circulating estrogen levels. Other contributing factors may involve heavy manual labor, heavy lifting, use of a tight abdominal girdle, chronic coughing, and straining during bowel movements (Davis & Kumar, 2003; What to do about pelvic organ prolapse, 2005). This paper describes demographic characteristics, reproductive history and hormonal contraception use associated with hysterectomy in Utah. Possible indirect effects of these factors on hysterectomy will be discussed. Methods Survey Instrument Prevalence among Utah women of reproductive factors, hysterectomy, and selected demographic variables was obtained through a statewide, random-digit-dialed survey. Telephone numbers were computer generated in a 26 Women With Hysterectomy © 2007 The University of Utah. All Rights Reserved UTAH'S HEALTH: AN ANNUAL REVIEW random fashion, providing numbers that were representative of households throughout the state of Utah. These numbers included listed, unlisted, and non-published numbers. Every potential telephone number within the sampled telephone area had a known and equal probability of selection. The sample telephone numbers were randomly assigned to interviewers by the Computer Assisted Telephone Interview (CATI) system. All women age 18 years and older in the state of Utah were eligible for participation in the survey. Upon contacting a household with eligible women, interviewers requested to speak to the oldest available woman in the household. If no woman was available, interviewers continued to call back until one became available or until 15 attempts were made. All selected phone numbers were called 15 times or until resident eligibility and willingness to participate could be determined. The 37-item survey instrument for this study was designed to assess reproductive behaviors among women in Utah. The Utah Health Status Survey (1996) served as a model for question design. In order to establish content and face validity, the instrument was reviewed by a number of experts whose training represents epidemiology, biostatistics, and women's health, and two other individuals with extensive experience in survey sampling. A pilot version of the survey was tested on 27 women, selected from the Utah population, to assess the instrument for clarity of questions and ease of administration. Pegus Research, a Salt Lake City-based firm specializing in survey research, administered the questionnaire. Demographic and Reproductive Variables The questionnaire was originally designed as a breast cancer risk factors questionnaire. A study based on the results of this instrument was published in 2004 (Daniels et al., 2004). Several of the variables in the questionnaire, however, allow us to explore relationships between selected demographic and reproductive variables and hysterectomy. Demographic variables collected and used in this study include current age, race, tobacco use, alcohol use, income, and education. Because over 95% of the study population was white, the race variable was dichotomized into whites and non-whites. Tobacco use was classified as smoker, former smoker, and never smoker. Alcohol use was classified as "yes" if the participant indicated that they consumed alcoholic beverages at lease once per month in the prior year and "no" otherwise. Categories of age, average household income, and education appear in Table 1. Reproductive variables considered in this study include whether the woman was ever pregnant (including live births, miscarriages, or other pregnancy outcomes), number of births (also called parity), multiple births, still births, miscarriages, lifetime duration of breastfeeding (years), use of oral contraceptives, use of oral contraceptives before first full term pregnancy, age started oral contraceptives, and age at menarche. Age at first birth is defined as the age at the end of a woman's first pregnancy of at least 20 weeks gestation. Pregnancies terminating prior to 20 weeks were classified as either spontaneous or induced abortions. Pregnancies © 2007 The University of Utah. All Rights Reserved Women With Hysterectomy 27 2007 UTAH'S HEALTH: AN ANNUAL REVIEW of at least 20 weeks were classified as full term pregnancies. Age at menarche is defined as a woman's age at her first menstrual period. Number of children (parity) is defined as the total number of live births, including stillbirths, in a woman's pregnancy history. All births after 20 weeks gestation were classified as either live or stillbirths. Use of oral contraceptive (OC) involved only women who reported using birth control pills. Birth control methods Depo-Provera and Norplant were not considered in this study because they did not receive Food and Drug Administration (FDA) approval until the mid 1990s. Depo-Provera is a brand name for the synthetic hormone, progesterone, which is injected. Norplant, which also uses progesterone, is encased in small silicone rubber tubes inserted in the arm. The phone survey produced 1,316 women aged 18 years or older determined to be eligible for the study. O f these women eligible for participation in the study, the interviewers were not allowed to speak directly with 78, 224 directly refused to participate, and language difficulties prevented 87 respondents from identifying their household status and eligibility. Interviewers conducted the survey in English. An additional subject terminated the telephone call before we were able to complete half the survey. We completed interviews with 926 women; the proportion of completed interviews among eligible women contacted was 70.4%, based on the formula defined by the American Association for Public Opinion Research (1998). Statistical Methods Hysterectomy status by selected demographic and reproductive variables was described using cross-tabulations. Relative risks and confidence intervals were computed and adjusted for age. Statistical significance was evaluated using the Pearson 2 test, the t test, and confidence intervals. Tests of significance were evaluated against the null hypothesis of no association, using the 0.05 level. Analyses were performed with standard packages of the Statistical Analysis System, Release 9.1 (SAS Institute Inc., Cary, NC, USA, 2003). Results Respondents ranged in age from 18 to 90 (M = 47.0, SD = 16.5). There were 26.6% (n = 246) who indicated that they had previously had a hysterectomy. The percentage having had a hysterectomy sharply increased from 1.2% in women aged 18-29 years to 51.3% for women aged 60-69 years, and then decreased slightly to 50.5% in women aged 70 years and older (Figure 1). 28 Women With Hysterectomy © 2007 The University of Utah. All Rights Reserved UTAH'S HEALTH: AN ANNUAL REVIEW Figure 1. Prevalence of hysterectomy in Utah according to age Age group A summary of selected demographic variables by hysterectomy status is presented in Table 1. The risk of hysterectomy significantly increased with age. For example women aged 60-69 were 6.17 times more likely to have had a hysterectomy than women aged 18-29 years. After adjusting for current age, the risk of hysterectomy was not significantly associated with any of the other demographic variables shown in the table. A summary of selected reproductive variables by hysterectomy status is presented in Table 2. After adjusting for age, there is a significant association between hysterectomy and age at first full term pregnancy, years breast-fed among parous women, age started birth control, and age first started having menstrual periods. The risk of having had a hysterectomy decreased with older age at first full term pregnancy, more years breast-fed, older age started birth control, and older age at menarche. Mean age at first full term pregnancy for women having undergone a hysterectomy compared with women not having had a hysterectomy is shown across current age in Figure 2a. Women having had a hysterectomy had a consistently younger age at first full term pregnancy across the age groups. The largest difference occurred in women 18-29 and the smallest difference is in women 60-69. Overall, the mean age at first full term pregnancy, adjusted for current age, is 21.5 years for women having had a hysterectomy compared with 23.7 years for women without a hysterectomy (t = -5.76, p < .0001). © 2007 The University of Utah. All Rights Reserved Women With Hysterectomy 29 2007 UTAH'S HEALTH: AN ANNUAL REVIEW Table 1. Summary of selected demographic variables by hysterectomy status Hysterectomy No Hysterectomy No. % No. % x 2 Risk Ratio* 95% CI Age 18-29 2 0.81 170 25.11 183.47 --- --- 30-39 14 5.69 157 23.19 <.0001 .98 .76, 1.28 40-49 40 16.26 144 21.27 1.13 0.88, 1.46 50-59 82 33.33 102 15.07 2.32 1.66, 3.26 60-69 59 23.98 56 8.27 6.17 3.60, 10.58 70+ 49 19.92 48 7.09 43.44 10.80, 174.74 Race Non-White 7 2.85 36 5.29 2.45 --- --- White 239 97.15 644 94.71 .1178 1.01 .53, 1.90 Hispanic origin Yes 4 1.63 18 2.65 .82 --- --- No 242 98.37 660 97.35 .3648 1.15 .53, 2.51 Tobacco smoking Current 22 8.94 67 9.85 6.60 --- --- Former 45 18.29 80 11.76 .0369 .92 .61, 1.40 Never 179 72.76 533 78.38 1.22 .86, 1.74 Alcohol drinking Yes 45 18.29 150 22.09 1.57 --- --- No 201 81.71 529 77.91 .2108 .92 .71, 1.19 Yearly income Less than $15,000 35 15.35 59 9.31 11.34 --- --- $15,000 to less than $30,000 41 17.98 119 18.77 .0230 1.23 .89, 1.70 $30,000 to less than $45,000 37 16.23 153 24.13 1.31 .92, 1.86 $45,000 to less than $60,000 54 23.68 129 20.35 .99 .70, 1.39 Greater than $60,000 61 26.75 174 27.44 1.07 .76, 1.51 Education Less than High School 11 4.49 20 2.95 13.69 --- --- Some college/technical school 73 29.80 138 20.32 .0177 1.10 .72, 1.69 Technical or associates degree 83 33.88 236 34.76 1.10 .72, 1.67 Bachelor's degree 24 9.80 83 12.22 1.69 .95, 3.01 Post graduate work or degree 30 12.24 124 18.26 1.25 .78, 2.00 *Adjusted for age. Mean years of breastfeeding among parous women with a hysterectomy compared with women without a hysterectomy is shown across current age in Figure 2b. Women with a hysterectomy had fewer years of breast feeding, except among women in the youngest and oldest age groups. Overall, mean years breastfeeding, adjusted for age, is 1.2 years for parous women with a hysterectomy and 1.7 years for parous women without a hysterectomy (t = -2.91, p = .0038). Mean age of first menstrual period for women with a hysterectomy compared with women without a hysterectomy is shown by age in Figure 2c. Women having undergone a hysterectomy had, on average, a younger age at menarche, with the exception of women aged 60-69. Overall, mean age at menarche, adjusted for age, is 12.6 years for women with a hysterectomy and 13.2 years for women without a hysterectomy (t = -4.02, p < .0001). 30 Women With Hysterectomy © 2007 The University of Utah. All Rights Reserved UTAH'S HEALTH: AN ANNUAL REVIEW Table 2. Summary of selected reproductive variables by hysterectomy status Hysterectomy No Hysterectomy No. % No. % x2 Risk Ratio 95% CI Pregnant Yes 227 92.28 576 84.71 8.99 --- --- No 19 7.72 104 15.29 .0027 1.16 .81, 1.64 Parity 1 12 5.29 79 13.72 20.82 --- --- 2 34 14.98 119 20.66 .0009 1.32 .77, 2.26 3 51 22.47 125 21.70 1.12 .68, 2.00 4 37 16.30 88 15.28 1.07 .63, 1.83 5 34 14.98 60 10.42 1.24 .73, 2.12 6+ 59 25.99 105 18.23 1.10 .64, 1.90 Age at first full term pregnancy 18-17 19 8.48 25 4.50 31.91 --- --- 18-20 78 34.82 116 20.86 <.0001 .90 .65, 1.26 21-23 72 32.14 174 31.29 .64 .46, .88 24-26 26 11.61 116 20.86 .47 .31, .73 27+ 29 12.95 125 22.48 .43 .29, .64 Multiple births in a single pregnancy Yes 16 7.05 24 4.17 2.86 --- --- No 211 92.95 552 95.83 .0910 .75 .52, 1.08 Stillbirths Yes 13 5.73 28 4.86 0.25 --- --- No 214 94.27 548 95.14 .6157 .90 .56, 1.44 Miscarriage Yes 91 40.09 186 32.29 4.37 --- --- No 136 59.91 390 67.71 .0365 .91 .74, 1.11 Years breast-fed among parous women 0 82 37.27 128 23.40 15.36 --- --- Some but less than 2 year 78 35.45 228 41.68 .0005 .89 .71, 1.12 2 or more years 60 27.27 191 34.92 .75 .58, .98 Ever used birth control pills Yes 160 65.31 501 73.68 6.18 --- --- No 85 34.69 179 26.32 .0129 .92 .74, 1.14 Birth control pills before first pregnancy Yes 73 32.16 269 46.70 14.07 --- --- No 154 67.84 307 53.30 .0002 .92 .72, 1.17 Age started birth control pills 7-15 8 5.10 19 3.83 1.52 --- --- 16-18 31 19.75 110 22.18 .8222 .80 .46, 1.38 19-21 57 36.31 180 36.29 .67 .40, 1.15 22-24 25 15.92 88 17.74 .48 .26, .86 25+ 36 22.93 99 19.96 .37 .21, .63 Age at menarche 6-10 18 7.41 28 4.17 14.20 --- --- 11-12 96 39.51 201 29.96 .0026 .91 .64, 1.30 13-14 98 40.33 317 47.24 .72 .51, 1.02 15+ 31 12.76 125 18.63 .53 .33, .86 *Adjusted for age. Mean age of first use of birth control for women having had a hysterectomy compared with women without a hysterectomy is shown across current age in Figure 2d. Women having had a hysterectomy had consistently younger age at which they started taking birth control pills, except for in women aged 70 years and older. Overall, mean age started taking birth control pills, adjusted for current age, is 19.8 years for women having had a hysterectomy and 22.7 years for women without a hysterectomy (t = -5.32, p < .0001). © 2007 The University of Utah. All Rights Reserved Women With Hysterectomy 31 2007 UTAH'S HEALTH: AN ANNUAL REVIEW Figures 2a through 2d Figure 2a. Mean age at first full term pregnancy in Utah Current Age Figure 2b. Mean years breast-fed in Utah Current Age Hysterectomy No Hysterectomy Hysterectomy No Hysterectomy Figure 2c. Mean age at menarche in Utah 14 12 10 S e 8 S 6 4 2 0 18-29 30-39 40-49 50-59 Curre nt Age 60-69 70+ Figure 2d. Mean age started birth control in Utah Current Age Hysterectomy----------No H ysterectomy Hysterectomy ----------No Hysterectomy Discussion In this study of Utah women, the prevalence of hysterectomy significantly increased with age. This is consistent with previous reports. After adjusting for age, no significant association was found between hysterectomy prevalence and race/ethnicity, tobacco use, alcohol use, yearly income, and education. Hysterectomy prevalence decreased with older age at first full term pregnancy, more years of breastfeeding, older age at menarche, and older age at starting oral contraception. All of the factors associated with decreased hysterectomy prevalence share the common link of being associated with decreased lifetime exposure to circulating estrogen. Although this study does not directly associate exposure to estradiol and risk of hysterectomy, the literature indicates that the body produces estradiol, a natural form of estrogen, during every menstruation, thus an older age at menarche means less lifetime exposure to estradiol (Hynes, 2005; Heitz et al., 1999). Similarly, a later age at first full term pregnancy means later exposure to increased levels of estradiol, another form of natural estrogen 32 Women With Hysterectomy © 2007 The University of Utah. All Rights Reserved UTAH'S HEALTH: AN ANNUAL REVIEW that increases in a woman's body during pregnancy (Fujimoto et al., 2005). During breastfeeding, a woman's body produces minimum estradiol, thus more years of breastfeeding translates to less lifetime exposure to estradiol (Hynes, 2005). Furthermore, standard birth control pills contain ethinyl estradiol, a synthetic form of estrogen. Longer duration of birth control use would be associated with increased exposure to ethinyl estradiol. While the current study did not consider duration of oral contraceptive use, there was a direct inverse relationship between age they started taking oral contraceptives and risk of hysterectomy. The increasing number of hysterectomies with age may also be related to levels of estrogen. Between the ages of 50 and 60 years, the number of hysterectomies increases, but at a much slower rate than before age 50. After 60 years of age the number of hysterectomies begins to decline. This corresponds with the median age for the onset of menopause, which is between 50 and 52 years, with only 2% of women still pre-menopausal by age 54 (Brambillam & McKinlay, 1989; McKinlay, Bifano, & Mckinlay, 1985; Mishra & Kuh, 2006). Menopause is marked by a decline in estrogen. As discussed above, the major factors associated with hysterectomies in the United States are uterine fibroids, endometriosis, and cancer. Increased exposure to estrogen is positively associated with increased rates of endometriosis, uterine prolapse, and cervical, endometrial, and ovarian cancers. Though estrogen has not been established as a cause of uterine fibroids, it is associated with their growth and contributes to risk factors for fibroids (Flake, Anderson, & Dixon, 2003). Progesterone is made by the ovaries and is essential for conception and the survival of the fertilized egg and the fetus throughout gestation. Progesterone increases during ovulation and, if fertilization does not occur, falls dramatically. During pregnancy progesterone influences an increase in the size of alveoli and lobes. Progesterone in oral contraceptives is a hormone that makes the body "believe" it is pregnant, thereby stopping the ovary from releasing an egg, thickening the mucus between the uterus and the vagina, and making the lining of the uterus less likely to let a released egg attach to the wall of the uterus. Thus, age at menarche, first birth, and first use of oral contraceptives influence progesterone. The role of progesterone in the incidence of hysterectomies is unclear. Progesterone appears to be preventive in endometriosis, with one study even showing a dose response relationship between progesterone levels and risk of endometriosis (Beral, Bull, & Reeves, 2005). The role of progesterone in fibroids is unclear, with different studies showing different results. Progesterone appears to temper the effects of estrogen in some studies, while other studies show that it also contributes to the growth of fibroids (Flake, Anderson, & Dixon, 2003; Rein, Barbieri, & Friedman, 1995). The relationship between progesterone and reproductive conditions is complex and it is difficult to differentiate between effects of progesterone and estrogen and the combined effect of both together. More research is needed to determine the effects of progesterone on reproductive conditions and hysterectomies. © 2007 The University of Utah. All Rights Reserved Women With Hysterectomy 33 2007 UTAH'S HEALTH: AN ANNUAL REVIEW Conclusion This study shows a significant relationship between the prevalence of hysterectomy and activities that lower estrogen levels in a woman's body, such as breastfeeding, and delayed commencement of menarche, pregnancy, and oral contraceptive use. A possible explanation for this relationship is that increased exposure to circulating estrogen produces health conditions, such as uterine fibroids, endometriosis, and cancers that explain the majority of hysterectomies in Utah. Further, it appears that activities that result in higher levels of estrogen increase the risk of hysterectomy. Thus, women may be able to decrease their risk of hysterectomy by avoiding excess estrogen exposure through delaying the use of oral contraceptives and increasing the duration of breastfeeding. References Baird, D. D., Dunson, D. B., Hill, M. C., Cousins, D., Schectman, J. M. (2003). High cumulative incidence o f uterine leiomyoma in black and white women: Ultrasound evidence. Am J Obstet Gynecol, 188(1), 100-7. Barbieri, R. L. (1997). Reduction in the size of a uterine leiomyoma following discontinuation of an estrogen-progestin contraceptive. Gynecol Obstet Invest, 43(4), 276-7. Beral, V., Bull, D., Reeves, G. (2005). Endometrial cancer and hormone-replacement therapy in the million women study. Lancet, 365(9470), 1543-51. Bj0rge, T., Kravdal, O. (1996). Reproductive variables and risk of uterine cervical cancer in Norwegian registry data. Cancer Causes Control, 7(3), 351-7. Brambillam, D. J., McKinlay, S. M. (1989). A prospective study of factors affecting age at menopause. J Clin Epidemiol, 42, 1031-9. Coleman, C. (2005). Elective primary cesarean birth: Issues for educators. Int J Childbirth Educ, 20(1), 34-9. Daniels, M., Merrill, R. M., Lyon, J. L., Stanford, J. B., White, G. L. J. (2004). Associations between breast cancer risk factors and religious practices in Utah. Prev Med, 38(1), 28-38. Davis, K., Kumar, D. (2003). Pelvic floor dysfunction: A conceptual framework for collaborative patient-centred care. J Adv Nurs, 43(6), 555-68. Flake, G. P., Anderson, J., Dixon, D. (2003). Etiology and pathogenesis of uterine leiomyomas: A review. Environ Health Perspect, 111(8), 1037. Fujimoto, J., Nakagawa, Y., Toyoki, H., Sakaguchi, H., Sato, E., Tamaya, T. (2005). Estrogen-related receptor expression in placenta throughout gestation. J Steroid Biochem Mol Biol, 94(1-3), 67-9. Guo, W. D., Hsing, A. W., Li, J. Y., Chen, J. S., Chow, W. H., Blot, W. J. (1994). Correlation of cervical cancer mortality with reproductive and dietary factors, and serum markers in China. Int J Epidemiol, 23(6), 1127-32. Heitz, N. A., Eisenman, P. A., Beck, C. L., Walker, J. A. (1999). Hormonal changes throughout the menstrual cycle and increased anterior cruciate ligament laxity in females. J A thl Train, 34(2), 144-9. Hinkula, M., Pukkala, E., Kyyronen, P., Kauppila, A. (2002). Grand multiparity and incidence of endometrial cancer: A population-based study in Finland. Int J Cancer, 98(6), 912-5. Hynes, A. (2005). The breast health handbook. Natural Health, 35(9), 56-65. Keshavarz, H., Hillis, S. D., Kieke, B. A. (2002). Hysterectomy surveillance - United States, 1994-1999. Morb Mortal Wkly Rep, 51(SS- 5), 1-8. Kjerulff, K. H., Guzinski, G. M., Langenberg, P. W ., Stolley, P. D., Moye, N. E., Kazandjian, V. A. (1993). Hysterectomy and race. Obstet Gynecol, 82(5), 757-64. 34 Women With Hysterectomy © 2007 The University of Utah. All Rights Reserved UTAH'S HEALTH: AN ANNUAL REVIEW Kvale, G., Heuch, I., Nilssen, S. (1988). Reproductive factors and risk of cervical cancer by cell type. A prospective study. Br J Cancer, 58(6), 820-4. La Vecchia, C., Negri, E., Franceschi, S, Parazzini, F. (1993). Long-term impact of reproductive factors on cancer risk. Int Journal Cancer, 53(2), 215-9. Lacey, J. V. J, Brinton, L. A., Abbas, F. M ., Barnes, W. A., Gravitt, P. E., Greenberg, M. D., et al. (1999). Oral contraceptives as risk factors for cervical adenocarcinomas and squamous cell carcinomas. Cancer Epidemiol Biomarkers Prev, 8(12), 1079-85. Lambe, M., Wuu, J., Weiderpass, E., Hsieh, C. C. (1999). Childbearing at older age and endometrial cancer risk (Sweden). Cancer Causes Control, 10(1), 43-9. L0chen, M. L., Lund, E. (1997). Childbearing and mortality from cancer of the corpus uteri. Acta Obstet Gynecol Scand, 76(4), 373-7. Meilahn, E. N., Matthews, K. A., Egeland, G., Kelsey, S. F. (1989). Characteristics of women with hysterectomy. Maturitas, 11(4), 319- 30. McKinlay, S. M ., Bifano, N. L., Mckinlay, J. B. (1985). Smoking and age at menopause in women. Ann Intern Med, 103, 350-6. Mishra, G., Kuh, D. (2006). Perceived change in quality of life during the menopause. Soc Sci Med, 62(1), 93-102. Mogren, I., Stenlund, H ., Hogberg, U. (2001). Long-term impact of reproductive factors on the risk of cervical, endometrial, ovarian and breast cancer. Acta Oncol, 40(7), 849-54. National Center for Health Statistics. (2004). Health, United States, 2004 with chartbook on trends in the health o f Americans, Table 97. Hyattsville, Maryland. Nowak, R. A. (1999). Fibroids: Pathophysiology and current medical treatment. Baillieres Best Pract Res Clin Obstet Gynaecol, 13(2), 223-38. Parazzini, F., La Vecchia, C., Negri, E., Fedele, L., Balotta, F. (1991). Reproductive factors and risk of endometrial cancer. Am J Obstet Gynecol, 164(2), 522-7. Parslov, M., Lidegaard, O., Klintorp, S., Pedersen, B., Jonsson, L., Eriksen, P. S., et al. (2000). Risk factors among young women with endometrial cancer: A Danish case-control study. Am J Obstet Gynecol, 182, 23-9. Rein, M. S., Barbieri, R. L., Friedman, A. J. (1995). Progesterone: a critical role in the pathogenesis o f uterine myomas. Am J Obstet Gynecol, 172(1 Pt 1), 14-8. Rowe, M. K., Kanouse, D. E., Mittman, B. S., Bernstein, S. J. (1999). Quality o f life among women undergoing hysterectomies. Obstet Gynecol, 93(6), 915-21. Swift, S., Woodman, P., O'Boyle, A., Kahn, M., Valley, M., Bland, D., et al. (2005). Pelvic organ support study (POSST): The distribution, clinical definition, and epidemiologic condition of pelvic organ support defects. Am J Obstets Gynecol, 192(3), 795- 806. The American Association for Public Opinion Research. (1998). Standard definitions: Final dispositions o f case codes and outcomes rates for RDD telephone surveys and in-person household surveys. Ann Arbor, Michigan. There's good news about birth control pills. (1992). Contracept Technol Update, 13(10 Suppl), 1-2. Utah Health Status Survey Codeboook, 2nd ed. (1996). Bureau of Surveillance and Analysis, Office of Public Health Data, Utah Department o f Health. Van Gorp, T., Neven, P. (2002). Endometrial safety of hormone replacement therapy: Review o f literature. Maturitas, 42(2), 93-104. Weaver, F., Hynes, D., Goldberg, J. M, Khuri, S., Daley, J., Henderson, W. (2001). Hysterectomy in Veterans Affairs Medical Centers. Obstet Gynecol, 97(6), 88-4. What to do about pelvic organ prolapse. (2005). Harv Women's Health Watch, 12(10), 4-6. Wilcox, L. S, Koonin, L. M., Pokras, R., Strauss, L. T , Xia, Z. S., Peterson, H. B. (1994). Hysterectomy in the United States, 1988-1990. Obstet Gynecol, 83(4), 549-55. © 2007 The University of Utah. All Rights Reserved Women With Hysterectomy 35 2007 UTAH'S HEALTH: AN ANNUAL REVIEW Excess Risk of Gestational Diabetes among Native American Mothers in Utah Brenda Ralls, PhD, William F. Stinner, PhD, Richard Bullough, PhD, Michael F. Friedrichs, MS, Jeffrey Duncan, MS, Jenny Billy Key Words: Gestational Diabetes, Native American, Diabetes during Pregnancy, Pre-pregnancy Weight, Maternal Age Abstract The prevalence of gestational diabetes mellitus (GDM) and its underlying risk factors within minority populations has become a matter of increasing concern. National data indicate a particularly high risk for Native American mothers. This study focused on the patterns of GDM and two risk factors (maternal age and pre-pregnancy weight status) between Native American and non-Native American mothers in Utah. Data were obtained from a pooling of five years of Utah birth records. Descriptive and multivariate logistic regression techniques were used. The effect of being Native American on GDM was more than double that observed for their non-Native American counterparts. When age and pre-pregnancy weight status were simultaneously adjusted, the higher prevalence of diabetes among Native American mothers persisted. Introduction Gestational diabetes mellitus (GDM) increases the risk for adverse pregnancy outcomes, such as stillbirths, congenital malformations, macrosomia, and cesarean sections. There is also emerging evidence that infants born to mothers with diabetes may be at increased risk of developing diabetes themselves (Dabelea, Hanson, Pettitt, et al., 2000). While it usually disappears after delivery, GDM can also lead to an increased risk of permanent diabetes for women who develop it. About 135,000 women are diagnosed with GDM each year in the U.S., accounting for about four percent of all pregnancies (American Diabetes Association, 2007). In Utah, birth records indicate that about two and one-half percent of all pregnancies, or approximately 1,100 women a year, are impacted by GDM. However, there is considerable variation by race and ethnicity, with higher prevalence observed among African American, Hispanic/Latina, Asian American mothers and most notably, Native American mothers (Berkowitz, Lapinski, Wein, & Lee, 1992). This study focuses on Native American mothers. Two factors often cited as risk factors for GDM are maternal age and pre-pregnancy weight status. These two risk factors may interact with and mediate the influence of being a Native American mother on developing GDM. Two ways in which this might operate are: (1) the two risk factors might exert a stronger effect among 36 Gestational Diabetes © 2007 The University of Utah. All Rights Reserved UTAH'S HEALTH: AN ANNUAL REVIEW Native American than among non-Native American mothers; or (2) Native American mothers may have higher or lower prevalence of the two risk factors. For example, older mothers and mothers who were overweight or obese prior to pregnancy tend to have a higher propensity for GDM than younger mothers and mothers who were not overweight or obese. Thus, to the extent that Native American mothers differ from non-Native American mothers regarding these risk factors, one would expect age and pre-pregnancy weight status to account for any initially observed difference between the two populations. This study, therefore, has two objectives: 1) To examine the extent to which Native American mothers have a higher prevalence of GDM and its risk factors than non-Native American mothers, and the degree to which the two risk factors account for any observed differences in GDM between the two populations, and 2) To examine the degree to which the two risk factors might exert a stronger effect among Native American than among non-Native American mothers. Methods Data for this study were obtained from a pooling of five years (2000-2004) of singleton Utah births (N = 228,680), including 2,567 births to Native American mothers and 226,113 births to non-Native American mothers. Native American mothers account for about one percent (1.1%) of all births to Utah mothers, while Native American women of child-bearing age (15-44) make up about two percent (2.1%) of the total number of women in Utah of child-bearing age (Center for Health Data, 2006). Cases for which mothers had less than two prenatal care visits and where any of the relevant information was missing were excluded from the analysis. Gestational diabetes (GDM) was measured as a two-category variable (present/not present). In the descriptive phase of the study, two risk factors (maternal age and pre-pregnancy weight status) were measured as follows: (1) maternal age - a four-category variable based on mother's age at time of delivery (less than 25, 25-29, 30-34 and 35 and older); (2) maternal pre-pregnancy weight status - a three-category variable based on the mother's body mass index (BMI) (weight in kilograms divided by height in meters squared). The three weight status categories were: not overweight (BMI <=24.9 kg/m2), overweight (25<=BMI<=29.9), and obese (BMI>=30). Mean age and mean BMI were also computed. Logistic regression is commonly used as a statistical method in epidemiological studies, and is especially useful where the number of cases in the comparison groups (e.g., exposed vs. unexposed populations) may be quite different (Friis & Sellers, 2004). Logistic regression was used to examine the effect of being Native American on the odds of having GDM. In this segment of the analysis, age category "less than 25" and the weight status category "not overweight" served as the reference categories. In the descriptive phase of the investigation, differences in GDM prevalence as well as variations across the categories of maternal age and pre-pregnancy weight status between Native American and non-Native American © 2007 The University of Utah. All Rights Reserved Gestational Diabetes 37 2007 UTAH'S HEALTH: AN ANNUAL REVIEW mothers were examined. In the logistic phase, separate analyses for Native American and non-Native American mothers were conducted to examine the patterning and relative strength of the odds ratios across the age and weight status categories, considered simultaneously. The main effects of being Native American on the odds of being diagnosed with GDM both without and with controls on age and weight status were assessed. Findings Figure 1. Percentage of Native American and Non-Native American Mothers with Gestational Diabetes Mellitus (GDM) Utah Office of Vital Records and Statistics 2000-2004 The prevalence of GDM for Native American and non-Native American mothers in Utah is shown in Figure 1. As can be seen, prevalence of GDM among Native American mothers was more than double that found for non- Native American mothers in Utah (5.6% vs. 2.3 %; p<.001). As may be seen in Table 1, Native American mothers were actually younger than non-Native American mothers. More than half (51.4%) of Native American mothers with live births were under age 25, compared to about two of five (39.0%) non-Native American mothers. However, the percentages of mothers aged 35 and over were similar for Native American and non-Native American mothers, 8.5% and 8.7%, respectively. The mean age for Native American mothers was lower than that for non-Native American mothers, 25.2 vs. 26.6 years, respectively. The percentage of Native American mothers who were overweight or obese was greater than that for non-Native American mothers. Over one of four (27.7%) Native American mothers were overweight, compared to about one of five (20.5%) non-Native American mothers. Fully 26.6 percent of Native American mothers were obese, nearly double that observed for non-Native American mothers (13.9%). The average BMI for Native American mothers was higher than that for non-Native American mothers ( 26.8 vs. 24.4 kg/m2, respectively). 38 Gestational Diabetes © 2007 The University of Utah. All Rights Reserved UTAH'S HEALTH: AN ANNUAL REVIEW Table 1. Maternal Age and Pre-Pregnancy Weight Status for Native American and Non-Native American Mothers 2000-2004 Age and Pre-Pregnancy Weight Status Native American Mothers (N=2,567) Non-Native American Mothers (N=226,113) Age Group Less than 25 51.4 39.0 25-29 26.1 33.3 30-34 13.9 19.0 35 and over 8.5 8.7 Total 100.0% 100.0% Mean age (years) 25.2 26.6 BMI Not overweight 45.7 65.6 Overweight 27.7 20.5 Obese 26.6 13.9 Total 100.0% 100.0% Mean BMI (kg/m2) 26.8 24.4 Source: Utah Birth Records 2000-2004 Totals may not sum to 100.0% due to rounding Cases where any of the relevant information was missing were excluded Figures 2 and 3 depict the patterning of GDM differences across maternal age and weight status subgroups for Native American and non-Native American mothers. The older the maternal age subgroup, the greater the GDM for both Native American and non-Native American mothers, with the difference between Native American and non-Native American mothers expanding the older the age subgroup (Figure 2). A relatively small percentage of both Native American and non-Native American mothers less than 25 years of age had GDM, with only a slight difference between the two (2.4% vs. 1.2%, p<.001). However, the GDM difference between Native American and non-Native American mothers grew wider the older the age subgroup. In the oldest age subgroup (35 and older), GDM prevalence among Native American mothers was 12.3 percent-more than five times that observed for Native American mothers less than 25 years of age, and more than twice that of non-Native American mothers in the oldest age subgroup (12.3% vs. 5.4%, p<.001). Equally pronounced was the response difference between Native American and non-Native American mothers in GDM rates across categories of maternal pre-pregnancy weight status (Figure 3). GDM rates among both Native American and non Native American mothers who were not overweight were low and differed only slightly in magnitude (2.3% vs. 1.3%, p<.001). The GDM rates, however, were higher among overweight and obese mothers, albeit more so among Native American than non Native American mothers. © 2007 The University of Utah. All Rights Reserved Gestational Diabetes 39 2007 UTAH'S HEALTH: AN ANNUAL REVIEW Figure 2. Percentage of Native American and Non-Native American Mothers with Gestational Diabetes Mellitus (GDM) by Maternal Age Utah Office of Vital Records and Statistics 2000-2004 Figure 3. Percentage of Native American and Non-Native American Mothers with Gestational Diabetes Mellitus (GDM) by Maternal Pre-Pregnancy Weight Status Utah Office of Vital Records and Statistics 2000-2004 40 Gestational Diabetes © 2007 The University of Utah. All Rights Reserved UTAH'S HEALTH: AN ANNUAL REVIEW Native American mothers who were overweight, but not obese, had a rate over one and one-half times that for non-Native American mothers (5.1% vs. 2.9%, p<.001). GDM prevalence among obese Native American mothers was 11.8 percent, nearly twice that for non Native American mothers who were obese (11.8% vs. 6.0%, p<.001). In sum, the oldest and obese mothers exhibited the highest GDM rates for both Native American and non Native American mothers, but the magnitude was greater for Native American mothers. Table 2. Odds Ratios and Confidence Intervals for Logistic Regression Analysis of Impact of Maternal Age and Pre-Pregnancy Weight Status on GDM for Native American and Non-Native American Mothers 2000-2004 Maternal Age and Pre- Pregnancy Weight Status Native American Mothers Non-Native American Mothers Age Group Odds (Confidence Interval) Odds (Confidence Interval) Less than 25 REF REF 25-29 3.2 (2.0 - 5.0) (NS) 1.8 (1.7 -- 2.0)*** 30-34 4.5 (2.8 - 7.4) ** 2.8 (2.6 -- 3.0)*** 35 and over 5.7 (3.3 - 9.7) *** 4.7 (4.3 - 5.1)*** BMI Not overweight REF REF Overweight 2.3 (1.4 -- 3.8) (NS) 2.3 (2.1 -- 2.4) (NS) Obese 5.7 (3.7 -- 8.9) *** 4.8 (4.5 -- 5.1) *** Source: Utah Birth Records, 2000-2004 REF= reference category * p<.05; ** p<.01; ***p<.001 NS=Not Statistically Significant The results (odds ratios) for the logistic regression analysis of the impact of maternal age and pre-pregnancy weight status on GDM for Native American and non-Native American mothers are shown in Table 2. Generally, the older the mother and the greater the degree of being overweight, the greater were the odds of having GDM among both groups of mothers. Native American mothers aged 25 - 29, however, were not significantly different from their counterparts less than 25 years of age in their odds of having been diagnosed with GDM. For both Native American and non-Native American mothers, those aged 35 and over exhibited the highest odds of having GDM compared to those less than 25 years of age, with the odds ratio being higher among Native American mothers (O.R. 5.7 vs. 4.7). © 2007 The University of Utah. All Rights Reserved Gestational Diabetes 41 2007 UTAH'S HEALTH: AN ANNUAL REVIEW On the other hand, merely being overweight, but not obese, increased the odds of having GDM compared to mothers who were not overweight, but the increase was not statistically significant. This pattern was observed regardless of whether those mothers were Native American or not. Obesity, however, significantly increased the odds of having GDM among both groups of mothers compared to those mothers who were not overweight. The impact of obesity on risk of GDM was higher for Native American mothers than for their non-Native American counterparts (O.R. 5.7 and 4.8, respectively). Figure 4. Odds Ratios for Effect of Being a Native American Mother on Risk of Gestational Diabetes Mellitus (GDM): Bivariate, Age-Adjusted, Weight-Adjusted, and Age/Weight-Adjusted 4 to T3 2 o 2.5 2.8 2.0 2.1 Bivariate Age-Adjusted Weight-Adjusted Age-and-Weight Adjusted Controls 3 1 0 Utah Office of Vital Records and Statistics 2000-2004 While the response effect was higher for Native American mothers, there did not appear to be a major difference in the pattern of the influence of the two risk factors between the two populations. In order to examine the effects of being a Native American mother on GDM, a logistic regression analysis was also conducted with being a non- Native American mother as the reference category. Odds ratios are presented for four assessments of being a Native American mother on the likelihood of having GDM (Figure 4).The four assessments include a bivariate analysis and three multivariate analyses including age adjustment, weight adjustment, and age-and-weight adjustment. In the bivarate analysis, being a Native American mother increased the risk of developing GDM by 150 percent. Because Native American mothers were younger, on average, age adjustment actually increased the risk to 180 percent (p<.001). When weight status was adjusted, the odds decreased; nonetheless, Native American mothers remained two times more likely to develop GDM than non-Native American mothers (p<.001). With both age 42 Gestational Diabetes © 2007 The University of Utah. All Rights Reserved UTAH'S HEALTH: AN ANNUAL REVIEW and weight adjustment, the higher prevalence of GDM persisted, with Native American mothers more than twice as likely to develop GDM as non-Native American mothers (p<.001). In other words, the effect of being a Native American mother increased the odds of GDM, even with maternal age and pre-pregnancy weight status controlled. Conclusions The basic aims of this study were to evaluate the manner in which being a Native American mother interacted with and is mediated by two risk factors, namely, maternal age and pre-pregnancy weight status, in affecting the odds of developing GDM. Descriptive analysis suggested that the pattern of the impact of maternal age and pre-pregnancy weight status were similar for Native American and non-Native American mothers, but the magnitude of the impact was greater for Native American mothers. However, when the joint effects of the two risk factors, namely maternal age and weight status, were considered separately for Native American and non-Native American mothers, their operation was roughly similar for the two groups of women. In other words, there did not appear to be support for the notion that maternal age and weight status impact the risk of GDM differently for Native American and non-Native American mothers. The above finding, nonetheless, does not necessarily mean that being a Native American mother has no effect on GDM, independent of the two risk factors; only that the two risk factors, when considered simultaneously, do not affect GDM differently for Native American and non-Native American mothers. When the main effect of being a Native American mother on GDM was examined, the excess risk of GDM became clearer. While some of this excess could be traced to the greater prevalence of overweight and obesity in this population, it only partially accounted for the higher prevalence of GDM among Native American mothers. Moreover, were it not for the fact that Native American mothers were younger, their rates of GDM would be even higher. In fact, even accounting for the "protective" effect of younger age, Native American mothers still had a rate of GDM over twice that of their non-Native American counterparts. Because the excess risk was particularly pronounced among Native American mothers who were obese, efforts aimed at promoting weight reduction prior to pregnancy may help to mitigate the risk of GDM in this population. If interventions are to be effective, however, they must be suitable and reflect cultural and lifestyle practices prominent in the Native American population. Involvement of members of the Native American communities is important for developing successful interventions. Limitations Native American status was obtained from birth records, and the accuracy of the findings in this study is limited to the accuracy of reporting on birth certificates. Presence of GDM during pregnancy may not always be listed on © 2007 The University of Utah. All Rights Reserved Gestational Diabetes 43 2007 UTAH'S HEALTH: AN ANNUAL REVIEW birth certificates; however, because a check box method of reporting is used, there is likely to be less risk of underreporting than there would be if it had to be written in. Pre-pregnancy weight status was self-reported by the mother and subject to reporting error. Use of body mass index, while a reasonable proxy measure for body fat, may not capture perfectly a person's weight status. Finally, the records did not have a unique identifier, and therefore, a woman may have been included more than once in the dataset. References American Diabetes Association. What is Gestational Diabetes? In Gestational Diabetes. Retrieved January 4, 2007, from http://www.diabetes.org/gestational-diabetesjsp Berkowitz, G. S., Lapinski, R. H., Wein, R., & Lee, D. (1992). Race/Ethnicity and Other Risk Factors for Gestational Diabetes. American Journal of Epidemiology, 135, 965-73. Center for Health Data. Population Query Module. In Utah's Indicator-Based Information System for Public Health (IBIS-PH). Retrieved February 22, 2007, from http://ibis.health.utah.gov Dabelea, D., Hanson, R. L., Pettitt, D. J., Imperatore, G., Gabir, M. M., Roumain, J., et al. (2000, December). Intrauterine Exposure to Diabetes Conveys Risks for Type 2 Diabetes and Obesity: A Study of Discordant Sibships. Diabetes, 49(12), 2208-2211 Friis, R.H., & Sellers, T.A. (2004). Measures of Effect. In Epidemiology for Public Health Practice (3rd ed., pp. 327-344). Sudbury, MA: Jones and Bartlett Smalley, K. J., Knerr, A. N., Kendrick, Z. V., Colliver, J. A., & Owen, O. E. (1990). Reassessment of Body Mass Indices. American Journal o f Clinical Nutrition, 52, 405-8. 44 Gestational Diabetes © 2007 The University of Utah. All Rights Reserved UTAH'S HEALTH: AN ANNUAL REVIEW Home Birth Trends in Utah, 1992-2005 Shaheen Hossain, PhD; Brenda Ralls, PhD; Rob Satterfield, MStat; Nan Streeter, MS, RN; Angeni Marque, BS; Jeffrey D. Duncan, MS Key Words: Home Birth, Birth Outcomes, Midwife, Complications of Labor, Birth Certificate Data CORRESPONDENCE: Shaheen Hossain, PhD Program Manager Data Resources Program, M CH Bureau PO Box 142001 Salt Lake City, Utah, 84114-2001 [phone] [e-mail] Abstract Home births remain a topic of considerable debate. Studies comparing birth outcomes associated with home deliveries and hospital deliveries have been contradictory. Some studies have documented home births as relatively safe with positive outcomes while others have observed an elevated risk in home births. In 2000, the proportion of home births in Utah was higher than the nation (1.4% vs. 0.7%), but too little is known about the birth outcomes among this population. This study was conducted to profile home birth and assess trends and outcomes in Utah from 1992 to 2005 using birth records. Introduction For centuries giving birth at home was the norm throughout the world. Since the beginning of the 20th century, with advances in scientific medicine, emphasis has been increasingly placed on the hospital as the safest birth environment for both mother and newborn. Childbirth was increasingly viewed as a surgical procedure performed on an anesthetized woman in a sterile environment. This medical view of childbirth largely persists to the present day. A nationwide 2002 survey of women's childbearing experiences reported that technology-intensive labors are currently the norm. One or more of the following medical interventions were used in a majority of births: electronic fetal monitoring (93%), intravenous drip (86%), epidural analgesia (63%), artificially ruptured membranes (55%), and artificial oxytocin to strengthen contractions (53%). In spite of the heavy prevalence of medical intervention in childbirth, 45% of the women polled still reported feeling that "giving birth is a natural process that should not be interfered with unless absolutely medically necessary" (Declercq, Sakala, Corry, Applebaum, & Risher, 2002). Dissatisfaction with the medicalization of childbirth and a growing desire for a more ‘natural' childbirth experience led to a renewed interest in home birth (O'Conner, 1993). In the 1970s the homebirth movement began, and increasing numbers of women chose to give birth at home instead of in the © 2007 The University of Utah. All Rights Reserved Home Birth Trends 45 2007 UTAH'S HEALTH: AN ANNUAL REVIEW hospital. However, much controversy has arisen regarding the relative safety of home birth. Numerous studies have been conducted to compare birth outcomes associated with home birth and hospital birth. The results of these studies have often been contradictory. Some studies have documented home birth as a relatively safe option, while others have observed elevated risk. A large prospective study conducted in North America in 2005 found that planned home births were associated with lower rates of episiotomy, forceps use, vacuum extraction use, and cesarean section compared to hospital births (Johnson & Daviss, 2005). A study using national birth certificate data reported that the neonatal outcomes associated with home birth compared favorably with the national rates (Declercq, Paine, & Winter, 1995). On the other hand, Pang et al. (2002) observed an increased risk of neonatal death and maternal postpartum bleeding associated with planned home birth. The rate of home birth in Utah has fluctuated around 1.2% over the past decade. However, Utah's rate has been much higher than the nation's. In 2000, the proportion of home births in Utah was double the national rate (1.4% vs. 0.6%, Figure 1). Too little is known about the birth outcomes among the home birth population. This study was conducted to develop a profile of birth outcomes among Utah women planning and delivering infants at home. Another focus of the study was to understand and compare Utah home birth trends with those of the nation. Figure 1: Percentage of live births planned and delivered at home, Utah and U.S., 1992-2005 Methods Utah birth certificate data from 1992 to 2005 were used for this study. A home birth was defined as a birth both intended for delivery at home and occurring at home. Therefore, this study excluded unplanned home births. The 46 Home Birth Trends © 2007 The University of Utah. All Rights Reserved UTAH'S HEALTH: AN ANNUAL REVIEW outcomes of home births were compared with all births in Utah. Statistical analyses were performed using SAS version 9.1 and included chi square tests, t-tests, and bivariate logistic regression. The study methodology paralleled that of a national study that profiled home birth in the United States using birth certificate data (Declercq, Paine, & Winter, 1995). Results There were 624,897 births during 1992-2005, of which 7,605 (1.2%) were planned home births. Table 1 compares demographic profiles of home birth mothers with all birth mothers. Compared to all births, women who planned home births were more likely to be older, married, and multiparous. They were less likely to be Hispanic. Home birth mothers were much less likely to have used tobacco during pregnancy than overall birth mothers (1.3% vs. 8.1%, p<.05). The majority of home births (79.8%) were attended by midwives. Table 1: Demographic profiles of planned home birth mothers and all birth mothers, Utah, 1992-2005 Percent Maternal Characteristic Planned Home Births All Births (n=7,605) (n=624,897) Age Under 25 years 33.1% 40.4% 25-34 years 50.3% 50.7% Over 34 years 16.6% 8.9% Education Less than high school graduate 17.3% 16.9% High school graduate 40.9% 37.9% Post-high school education 41.8% 45.2% Hispanic 3.5% 11.3% Married 90.9% 83.3% Smoked during pregnancy 1.3% 8.1% Prenatal care in 1st trimester 70.8% 81.8% Parity Nulliparous (without previous live birth) 17.0% 35.6% Multiparous (with previous live birth) 83.0% 64.4% Complications of labor and delivery, as collected in birth certificate data, were generally less common in home births than all births, with the exception of measures of length of labor (prolonged labor and precipitous labor, Figure 2). Home births had lower incidence of bleeding such as abruptio placenta. However, home births were associated with increased risk in the other excessive bleeding category compared to all births (OR = 2.7, 95% CI 2.3 - 3.1). © 2007 The University of Utah. All Rights Reserved Home Birth Trends 47 2007 UTAH'S HEALTH: AN ANNUAL REVIEW Figure 2: Comparison of obstetric complications, home births and all births, Utah, 1992-2005 6% □ All Births □ Planned Home Births 4.7% <2 4% m <>u o a> a Sc0) p <u 2% Q. 2.6% 2.2% 2.2% 0% 1.6% 0.9% 0.8% 4.1% 2.0% 1.3% 1.1% 0.2% 2.7% 1.1% Premature Rupture of Membrane Other Excessive Bleeding Prolonged Labor Fetal Distress Complications Abruptio Precipitus Labor Dysfunctional Labor Lower rates of adverse birth outcomes were observed among home birth newborns compared to all newborns. Home birth newborns were much less likely to be found in either the low birth weight (< 2,500 grams) or preterm (< 37 weeks gestation) categories (Figure 3). The average Apgar scores both at 1-minute and 5-minute were higher among newborns born at home than among all newborns. Conclusions This study provides a profile of home birth trends and birth outcomes in Utah. Women with a low risk of obstetric complications are generally eligible to deliver at home. This study found that home birth outcomes were similar compared to all births. However, there were some noteworthy differences. Home birth mothers had higher rates of precipitous labor, prolonged labor, and other excessive bleeding. This higher incidence of precipitous labor may be at least partially attributable to the higher proportions of home birth mothers who were multiparous. Prolonged labor among home births may be a function of avoiding drug induced labor and artificial stimulation of labor. It is difficult to explain the higher risk of excessive bleeding among home birth mothers. It is possible that episiotomy rates between home birth mothers and other mothers may differ and may relate to the higher risk of excessive bleeding observed among home birth mothers. 48 Home Birth Trends © 2007 The University of Utah. All Rights Reserved UTAH'S HEALTH: AN ANNUAL REVIEW Figure 3: Percentage of births that were preterm or low birth weight, home births and all births, Utah, 1992-2005 The findings of this study should be considered with several limitations in mind. Birth certificate data does not identify births that were intended for home delivery but were transferred to the hospital for delivery. Therefore, the outcomes of such births are absent from the home birth group and are included among the all birth group. This omission of a portion of outcomes from the intended home birth group may distort the profile of outcomes. Additionally, in this study there may have been unmeasured differences (e.g. psychological, financial, cultural) between women who planned home birth and the general birth population. Such measures are not available in the birth certificate records and may have influenced estimates. Another limitation of this study was that the all births group included both high and low-risk pregnancies. It would be prudent to compare the outcomes of low-risk women giving birth at home with those of low-risk women giving birth in the hospital. It will be important to further examine the causes of the excessive bleeding observed in this study among home birth women. These unanswered questions, along with the contradictory findings of previous studies, all point toward the imperative for additional study regarding the safety of home births. References Declercq, E. R., Paine, L. L., & Winter, M. R. (1995). Home birth in the United States, 1989-1992: A longitudinal descriptive report of national birth certificate data. Journal of Nurse-Midwifery, 40(6), 474-482. Declercq, E. R., Sakala, C., Corry, M. P., Applebaum, S., & Risher, P. (2002). Listening to mothers: Report of the first national U.S. survey of women's childbearing experiences. New York: Maternity Center Association. © 2007 The University of Utah. All Rights Reserved Home Birth Trends 49 2007 UTAH'S HEALTH: AN ANNUAL REVIEW Johnson, K. C., & Daviss, B. A. (2005). Outcomes o f planned home births with certified professional midwives: Large prospective study in N orth America. British Medical Journal, 330, 1416. O'Conner, B. B. (1993). The home birth movement in the United States. The Journal of Medicine and Philosophy, 18, 147-174. Pang, J. W . Y., Heffelfinger, J. D., Huang, G. J., Benedetti, T. J., & Weiss, N. S. (2002). Outcomes of planned home births in Washington state: 1989-1996. The American College of Obstetrics and Gynecologist, 100(2), 253-259. Acknowledgements The authors would like to acknowledge the contributions o f Pete Barnard, CNM with this project. 50 Home Birth Trends © 2007 The University of Utah. All Rights Reserved UTAH'S HEALTH: AN ANNUAL REVIEW Patterns of Medicaid Expenditures in Utah and Nevada for Prescription Drugs by Urban Rural Frontier Location Gulzar H. Shah, MStat, MSS, PhD Key Words: Pharmacy Expenditures, Urban/Rural, Nervous System Drugs, Medicaid Enrollees CORRESPONDENCE: Gulzar H . Shah, PhD Director o f Research, National Association o f Health Data Organizations 448 East, 400 South, Ste 301 Salt Lake City, Utah 84111 Phone: [phone] [e-mail]; [e-mail] Abstract Purpose: This research examines rural vs. urban differences in cost of prescription drugs for nervous system and other therapeutic categories among Medicaid enrollees in Utah and Nevada. Data and Methods: Using enrollee For the measure of urban-rural residence, county of residence was classified into three categories: urban, rural, and frontier, based on a county's population density. To provide context to the study findings, discussions were held with several experts. Results: Expenditures per Medicaid enrollee were substantially higher in Utah than in Nevada. Expenditures were higher in urban areas, lower in rural areas, and lowest in frontier areas in Utah, but the opposite pattern was observed in Nevada. In both states, a large proportion of the Medicaid expenditures are for drugs directed at the nervous system--46% in Utah and 42% in Nevada. The study reveals sharp geographic differences in Medicaid spending for prescription drugs, particularly those related to nervous system. Introduction and Background A long-standing health policy issue is whether urban residents are different than those residing in rural areas in healthcare access, cost and consumption. Examining variation in cost of prescription drugs by geographic location of Medicaid enrollees is imperative now than ever before, given the recent dramatic increase in healthcare cost and expenditure in the United States, including Medicaid expenditures (Borger et al., 2006; Smith et al., 2006). Growth in Medicaid spending has been faster for prescription drugs than any other component of care (Holahan & Ghost, 2005). States have been intensifying their efforts to control rising costs of various components of Medicaid Programs, including prescription drugs (Kitchener, Miller, & Harrington, 2005; Mello, Studdert, & Brennan, 2004). Historically, disparities in healthcare have been examined using a variety of approaches, including a focus on specific populations such as merchants and seamen, and study of specific outbreaks of disease. In the © 2007 The University of Utah. All Rights Reserved Patterns Of Medical Expenditures 51 2007 UTAH'S HEALTH: AN ANNUAL REVIEW contemporary era, a greater emphasis has been on examining geographic disparities (see, e.g., Cayce et al., 2005; Fortney, 1999; Hall, Kaufman, & Ricketts, 2006; Larson & Fleishman, 2003; Phillips & McLeroy, 2004; Rost et al., 1998; Sheikh & Bullock, 2001). Discussions of this issue inevitably involve differences in the structure and processes of health care in rural and urban areas. Rural-urban disparities are often thought to stem from differences in demographic compositions characterizing these populations as well as contextual factors such as limited education, lower household income, poor access to and availability of medical care and awareness of such availability, lower insurance coverage, and lower economic opportunity (Blumenth et al., 2002; Probst et al., 2004). Rural residents often have relatively poorer socioeconomic status, face greater insurance coverage problems, and have more health problems and co-morbidities (Eachus et al, 1999; Gundy & Glaser, 2000; Ketsche, 2005; Koster et al., 2004). When coverage is available through Medicaid, it may result in greater utilization of healthcare resources by the rural area residents, (Dis, 2004; Merwin, Snyder, & Katz, 2006; Saag et al., 1998). Studies examining patterns of utilization and cost of prescription drugs by rurality status of residence are sporadic at best. Although rare, such studies often show no conclusive evidence of urban rural differences in drug price or use patterns (Carrie, Grymonpre, & Blandford, 2006; Casey, Klingner, & Moscovice, 2002). A few studies of the urban-rural difference in healthcare access specifically among Medicaid Enrollees do exist and they may provide context to the findings of this study (e.g., Kemper, Cohn, & Dombkowski, 2004; Kuhlthau et al., 2001; McAuley et al., 2004). Most of these studies show a higher use of services/resources by rural Medicaid enrollees. For instance, among Medicaid enrollee children 18 or younger in Michigan, those residing in rural counties had higher odds of receiving eye care or lens services compared with children in urban counties (Kemper, Cohn, & Dombkowski, 2004). A plausible explanation for existence of differences can be attributed to urban-rural differences in provider acceptance of Medicaid coverage. For instance, providers in urban markets with higher densities of eye care professionals were less likely to care for Medicaid-enrolled children (Kemper, Cohn, & Dombkowski 2004). Another study found that the use of formal home healthcare by Medicaid enrollees was significantly higher among the rural residents (McAuley et al., 2004). High costs of prescription drugs result in access problems even among Medicaid enrollees, in spite of their coverage through Medicaid, because of co-payments required in many states (Cunnigham, 2002). Existing research literature offers some important explanations to the differences in expenditure on pharmacy drugs between urban and rural enrollees. There is evidence of price differences by geographic areas even after pharmacy differences are statistically controlled for. The geographic difference in price is lower for repeatedly purchased prescriptions because expected benefits of searching for better prices and willingness to purchase from other locations is greater in case of repeated drugs (Sorenson, 2000). However, due to competing forces on pharmacy 52 Patterns Of Medical Expenditures © 2007 The University of Utah. All Rights Reserved UTAH'S HEALTH: AN ANNUAL REVIEW prices, direction of impact of rural residency on cost is unclear. For instance, urban-rural differences in costs can be attributed to differences in rent, salaries, general cost of living, insurance, and other factors, making the cost of drugs likely to be lower in rural areas. On the other hand, rural areas may have a fewer pharmacy vendors monopolizing the market. This may give these vendors leverage to charge higher prices. Urban areas, in contrast, may have higher vendor competition, leading to relatively lower prices. In addition, changes in drug product payment policy may have different consequences for providers and pharmacies across urban-rural geographic locations (Sorenson, 2000). Prescription drug cost may vary across urban and rural areas because of additional factors, including urban-rural differences in state dispensing fees. For instance in Utah, dispensing fee is slightly lower for the urban pharmacies with $3.90 for urban and $4.40 for rural pharmacies. Finally, rural pharmacies tend have lower annual prescription volume compared with urban pharmacies, which may result in higher unit costs (United States Congress, 2004). Various interest groups in a state may also have impact on the cost of Medicaid prescription drugs, resulting in policies that may lead to disparities (Pracht & Moore, 2003). In general the out of pocket expenses and cost of prescription drugs is higher in rural areas than in urban areas (Brangan & Caplan, 2004). Because of relatively lower incomes, the impact of high cost of prescriptions drugs is more severe for people in rural areas (Averill, 2003). Studies indicate that there is a particular need to examine the patterns of use and cost of antipsychotic and other nervous system drugs because their use is widespread, yet cost is alarmingly high (Gilmer et al., 2004; Rothbard, Metraux, & Blank, 2003). The rising cost of prescription drugs has caused two interrelated problems -- greater expense for "haves" and the limited access to medications for "have-nots" (Steinbrook, 2002). For this study, the initial concern was with examining if use of pharmaceutical prescription drugs was concentrated into few categories of enrollees, and to uncover potential high users, by eligibility category, including age, sex, and disability status. After conducting review of literature, certain gaps in the existing body of knowledge were found, suggestive of implications for this study. Given the gaps in literature, the main thrust of analyses was shifted on rural-urban differences in cost of prescription drug, especially on drugs associated with mental health. Data and Methods Two Medicaid files for 2002 were obtained for both Utah and Nevada: (a) Pharmacy file, wherein each record represented a prescription for a given enrollee on a given day for a given drug; and (b) Enrollment file, with variables such as ID number, eligibility category, county of residence, and months of enrollment in 2002. The two files were merged using ID number. Enrollees in managed care at mid-year were dropped from the analysis. © 2007 The University of Utah. All Rights Reserved Patterns Of Medical Expenditures 53 2007 UTAH'S HEALTH: AN ANNUAL REVIEW For the measure of urban-rural residence, county of residence was classified into three categories, urban, rural, and frontier. A county was designated as urban if its density was at least 100 persons per square mile. When a county had density of between six and 99 persons, it was designated as rural. If county density was five persons or less per square mile, it was designated as frontier. Isolated rural or "frontier" areas are sometimes defined as having population density of six persons per squares mile, although other commonly used definitions involve distance to services (Rural Assistance Center, 2006). Various population densities are used to distinguish urban and rural areas; a density of 100 persons per square mile is one such density (U.S. Bureau of Census, 2006). Within this study, locational analyses pertain to the residence of the enrollee, not the pharmacy or prescriber. In a similar study of Medicaid enrollees population, Kemper and colleagues (2004) defined urban-rural status of the enrollees based on whether their county of residence contained a metropolitan area as defined by the U.S. Census Bureau. In addition to the operational definitions discussed in this paper, many other are used by social scientists and demographers. A detailed review of approaches to classify geographic areas has been presented by a recent study, concluding that dichotomous definitions may fail to capture variability in rural areas (Hall, Kaufman, & Ricketts, 2006). In view of available data, and based on review of literature and gaps therein, the scope of this study was kept to exploratory, with its focus on nervous system drugs. Almost all payers in the U.S. identify drugs using the National Drug Code (NDC) system of classification. Because of the vast number of NDC codes, any analysis of therapeutic groups requires a categorization scheme. In this regard, Utah uses First Data Bank, whereas Nevada uses First Health Services. The pharmacy claims data from both Utah and Nevada Medicaid programs contained a therapeutic variable. To create comparable therapeutic groups, early in this project, Jeffrey Geppert, JD of Stanford University developed a crosswalk between these two schemes. Utah's nervous system category was deemed analogous to the combination of eight of Nevada's categories. Differences in the spending per member per month (PMPM) could simply be the result of differences in the distributions of eligibility categories. For instance, disabled beneficiaries have more drug spending than other eligibility categories. If rural areas have a higher proportion of disabled beneficiaries than urban areas, this could explain the higher spending in rural areas. To ensure that the expenditure figures do not reflect the impact of differing proportions within eligibility categories across areas, indirect standardization by eligibility categories was applied. 54 Patterns Of Medical Expenditures © 2007 The University of Utah. All Rights Reserved UTAH'S HEALTH: AN ANNUAL REVIEW To understand the factors that might lie behind the figures produced by this project, discussions were held with several experts, including: • Mr. Joseph A. Greenway, Center for Health Information Analysis, University of Nevada at Las Vegas • Dr. Mark Babbitt, Department of Family and Preventive Medicine, School of Medicine, University of Utah • Dr. William Custer, PhD, Georgia State University, Department of Risk Management, School of Business • Dr. Barbara Rudolph, PhD, MSSW, University of Wisconsin-Madison, Center for Health Systems Research and Analysis Their insights are incorporated into the results section. Results Spending for all Therapeutic Categories Patterns of difference in Medicaid expenditure on prescription drugs in Utah and comparative figures for Nevada are reported in Table 1. Across all therapeutic categories, Utah's Medicaid program spent about $145 million on drugs in 2002, where Nevada's spent about $85 million. In each of those states, the percentage of expenditures on behalf of enrollees living in urban areas was in the 70-73% range. Analogous figures for rural and frontier areas were 17-20% and 9-10%. These expenditures reflect, in part, the distribution of enrollees, the bulk of whom live in urban areas but their substantial proportions live in rural and frontier areas. Table 1. Urban-Rural-Frontier Differences in Expenditures for prescriptions in all Therapeutic Categories , Utah, and Nevada,2002 Expenditure Measures Utah Nevada All Urban Rural Frontier All Urban Rural Frontier Total expenditures 145,566,248 106,087,541 25,416,750 14,061,957 $84,966,527 $60,040,758 $16,971,986 $7,953,783 % Distribution 100.0% 72.9% 17.5% 9.7% 100% 70.7% 20.0% 9.4% Total Member months 1,905,871 1,333,628 366,168 206,075 1,821,296 1,366,821 312,628 141,847 % Distribution 100.0% 70.0% 19.2% 10.8% 100% 75.0% 17.2% 7.8% © 2007 The University of Utah. All Rights Reserved Patterns Of Medical Expenditures 55 2007 UTAH'S HEALTH: AN ANNUAL REVIEW Differences in average cost of prescription drugs "per member per month" (PMPM) for all drug categories and for the nervous system drugs are presented in Table 2 for both Utah and Nevada. The two states' Medicaid programs differ, in two ways, in terms of expenditures per enrollee, or more precisely, PMPM. First, for drugs in all therapeutic categories, expenditures are substantially higher in Utah than in Nevada, $76 vs. $47. This difference could reflect differences in prices paid for similar drugs, such as policy on generics or on discounts for brand-name drugs. Table 2. Urban-Rural-Frontier Differences in Expenditures for Utah, and Nevada, by Type of Drugs, 2002 Expenditure Measures Utah Nevada All * Urban Rural Frontier All* Urban Rural Frontier All Therapeutic Categories $$ PMPM $76.38 $79.55 $69.41 $68.24 $46.65 $43.93 $54.29 $56.07 Ratio of mean 1.000 1.042 0.909 0.893 1.000 0.942 1.164 1.202 Nervous System Drugs $$ PMPM $35.18 $37.90 $30.80 $25.39 $19.52 $18.67 $22.20 $21.86 Ratio of mean 1.000 1.077 0.875 0.722 1.000 0.957 1.137 1.120 All Therapeutic Categories other than Nervous System Drugs $$ PMPM $41.20 $41.65 $38.61 $42.85 $27.13 $25.26 $32.09 $34.22 Ratio of Mean 1.000 1.011 0.937 1.040 1.000 0.931 1.183 1.261 * The "all" column excludes the few members who are out of state. Second, expenditures per Medicaid enrollee fell with decreasing population density in Utah; that is, it was higher in urban areas ($80), lower in rural areas ($69), and lowest in frontier areas ($68). The opposite pattern was observed in Nevada - highest in frontier area ($56), lower in rural areas ($54) and lowest in urban areas ($44). In contrast to some of the figures (presented in Table 3), these figures are not adjusted for eligibility category. The average expenditure PMPM for all therapeutic categories except for those related to nervous system followed the same pattern in Nevada as for all therapeutic categories including nervous system -- highest in frontier area and lowest in urban areas (Table 2). However, for Utah, the corresponding figures showed a different pattern of expenditure by urban-rural-frontier residency status compared with the one for drugs in all therapeutic categories. 56 Patterns Of Medical Expenditures © 2007 The University of Utah. All Rights Reserved UTAH'S HEALTH: AN ANNUAL REVIEW Spending on Nervous System Drugs In both states, a large minority of the Medicaid expenditures are for drugs directed at the nervous system, 46% in Utah and 42% in Nevada. In Utah, spending PMPM in frontier areas is 72% of the statewide mean for nervous system drugs (computed from Table 3), but 104% of drugs in other therapeutic categories (computed from Table 2) - a sharp difference across urban vs. rural geographic areas by therapeutic category. In Nevada, in frontier areas, expenditures for drugs related to nervous system as well as for other therapeutic categories are higher than statewide mean, 112% of the statewide mean for the former and 126% higher for the latter. Table 3. Urban-Rural-Frontier Differences in Expenditures for Nervous System Prescriptions, Adjusted by Eligibility Category, Utah, and Nevada, 2002 Expenditure Measures Utah Nevada All Urban Rural Frontier All Urban Rural Frontier Total Expenditures ($$) 67,053,484 50,542,506 11,279,653 5,231,324 $35,557,672 $25,517,109 $6,940,110 $3,100,453 $$ PMPM -- Observed $35.18 $37.90 $30.80 $25.39 $19.52 $18.67 $22.20 $21.86 $$ PMPM -- Expected $35.18 $36.36 $31.31 $34.48 $19.54 $19.02 $21.69 $19.76 Ratio of Observed/ Expected $$ ** PMPM 1.000 1.042 0.984 0.736 0.999 0.981 1.023 1.106 $ PMPM as a ratio of urban areas -- Unadjusted NA 1.000 0.813 0.670 NA 1.000 1.189 1.171 $ PMPM as a ratio of urban areas -- Adjusted NA 1.000 0.944 0.706 NA 1.000 1.043 1.127 NOTE1: Nervous system (except autonomic) expenditures constitute 46% of all expenditures in Utah and 42% of all expenditures in Nevada; NOTE2: Autonomic nervous system expenditures constitute only 4%. * The "all" column excludes the few members who are out of state ** To apply adjustment for the eligibility category When these figures for nervous systems drugs are adjusted for eligibility categories, they are largely unchanged. In Utah, nervous system drugs are 74% of the mean (72% unadjusted); in Nevada, those drugs are 111% of the mean. These figures can be presented using urban areas as the denominator: adjusted for eligibility, spending in frontier areas for the nervous system drugs is 71% of urban areas in Utah. The corresponding figure for Nevada is 113%, indicating that even after adjustment in eligibility category, expenditures per Medicaid enrollee fell with decreasing population density in Utah, but opposite was true in Nevada. © 2007 The University of Utah. All Rights Reserved Patterns Of Medical Expenditures 57 2007 UTAH'S HEALTH: AN ANNUAL REVIEW Expert Panel's Discussion of Results: In discussions with the expert panel, a number of explanations regarding the geographic differences in prescription drugs, observed in the data were mentioned. It is useful to start by mentioning explanations that can be rejected: • Different eligibility patterns by location do not explain the results indicating geographic variation in prescription drugs expenditures, because they are adjusted for eligibility. • Differences in prices are unlikely to explain the results, because the prices vary little within a state. Dispensing fees vary slightly, but these constitute a small proportion of the cost of a prescription. • Other Medicaid policies (e.g., formulary) do not explain the results, because they apply equally within a state. Other explanations are less easily dismissed: Beneficiaries may differ by location in ways other than eligibility category. For instance, in Nevada the urban population is more heavily Hispanic than the rural population, but the same is not the case in Utah. In Utah, on the other hand, the urban population is more religiously heterogeneous than the rural population. In contrast, experts in Nevada did not think that level of religious homogeneity varied by urban-rural residence status in Nevada. Finally, in Nevada the urban population is more transient than the rural population, but that is not the case in Utah. Perhaps these cultural groups are more or less likely to take antidepressants than the population at large. Factors that could differ among cultural groups include social support network size and quality and stigma associated with help-seeking behavior and use of mental health medications. Another set of explanations pertain to providers. Rural pharmacies tend to be smaller and hence have fewer brands within a therapeutic category, due to which the cost per supply day (i.e., cost for a day's supply) differs by urban-rural-frontier location. Because the data were analyzed on the basis of enrollee residence, and not pharmacy or prescriber location, it is difficult to ascertain whether rural enrollees' usual source of care was located in a rural area or whether care and medications were from an urban area. Specialty mix of providers varies by location, with rural and frontier areas having a greater proportion of general practitioners. It is hypothesized that general practitioners prescribe less "talk therapy" and more psychotropic drugs. If so, one would expect more nervous system medication spending in rural areas than in urban ones. However, it is likely that general practitioners' training varies by medical college attended and this could clearly affect prescribing patterns. 58 Patterns Of Medical Expenditures © 2007 The University of Utah. All Rights Reserved UTAH'S HEALTH: AN ANNUAL REVIEW Summary, Conclusions, and Suggestions for Further Research This study represents an initial foray into differences across states and urban-rural Medicaid use of pharmaceuticals addressing the nervous system. The results are striking-with clear differences in usage of medications across states and across urban-rural-frontier locations. Across all therapeutic categories, Utah's Medicaid program spent about $145 million on drugs in 2002, where Nevada's spent about $85 million. Expenditures per enrollee were substantially higher in Utah than in Nevada. Expenditures per Medicaid enrollee were higher in urban areas, lower in rural areas, and lowest in frontier areas in Utah, but the opposite pattern was observed in Nevada. In both states, a large proportion of the Medicaid prescription drug expenditures are for drugs directed at the nervous system--46% in Utah and $42% in Nevada. In Utah, spending per member per month (PMPM) in frontier areas is 72% of the statewide mean for nervous system drugs. The corresponding figure for other therapeutic categories is 104% of the statewide mean - suggesting sharp geographic differences in Medicaid spending per enrollee by therapeutic category. In Nevada, in frontier areas, average cost per enrollee on both nervous system drugs and other drugs are above the statewide mean. After adjusting for eligibility, spending