Each year in the United States, approximately 6 million women become pregnant. Although traditionally, the effect of pregnancy on a woman’s health has been assessed in terms of maternal mortality, maternal morbidity has become an increasingly important indicator of a woman’s health and her health care during pregnancy.1 Nevertheless, estimating the prevalence of maternal morbidity presents many challenges. For example, many databases previously used to track morbidity, such as national- and state-based hospital discharge databases, provide information only on morbidities that result in hospitalization or morbidities occurring during delivery.2–5 Moreover, changes in medical care and management of pregnancy-related complications have led to increasing reliance on outpatient treatment for some conditions previously treated in the hospital; conditions treated in outpatient settings are not included in hospital discharge databases. Thus, hospital-based surveillance leads to underestimates of the prevalence of maternal morbidity.
Recent advances in the medical informatics of large, vertically integrated health care delivery systems include the adoption of computerized inpatient and outpatient clinical, laboratory, pathology, radiology, and other records. In a previous report, we described an algorithm that used the electronic data systems of a large, nonprofit, group-model health maintenance organization (HMO) to identify all pregnancies and pregnancy outcomes occurring in a defined time period.6 For this report, we employed similar methodology and identified maternal morbidities during the antepartum, intrapartum, and postpartum periods in a defined population. We present rates of the most common morbidities by pregnancy outcome and selected covariates.
MATERIALS AND METHODS
This study was conducted by using electronic data from Kaiser Permanente Northwest, a prepaid, nonprofit group practice HMO with approximately 467,000 members. Members include individuals with employer-sponsored coverage (commercial), Washington Basic Health Plan (state-sponsored coverage for uninsured), Medicare, or Medicaid. We developed and validated a computerized algorithm that identifies indicators of pregnancy. The algorithm obtains data from Kaiser Permanente Northwest individual-level data systems, including enrollment, hospital discharge, outpatient care, emergency department, outside claims and referrals, radiology and imaging, laboratory, and pharmacy. Using a hierarchy of decision rules, the algorithm linked indicators and dates of pregnancies and pregnancy outcomes to create pregnancy “episodes.” A detailed description of the methods is published elsewhere.6
A pregnancy episode was defined as the interval between the estimated date of the last menstrual period and 8 weeks after delivery or pregnancy termination. Pregnancy outcomes included were live birth, stillbirth, spontaneous abortion, therapeutic abortion, gestational trophoblastic disease, and ectopic gestation. The study population was female members of Kaiser Permanente Northwest, aged 12–55 years, whose pregnancy episode occurred completely within the study period (1998–2001) and who were still members at the date of the pregnancy outcome. We defined maternal morbidity as any condition occurring during a pregnancy episode that adversely affects the physical or psychological health of the woman. We included conditions that are unique to pregnancy (eg, preeclampsia) and conditions that occur outside of pregnancy but are exacerbated by pregnancy (eg, diabetes). Ectopic pregnancy and gestational trophoblastic neoplasia were considered complications in and of themselves; therefore, we did not attempt to identify other morbidities in pregnancy episodes with these outcomes.
To determine the morbidities assessed in this study, we first conducted a comprehensive review of all codes, including those outside the pregnancy chapter, of the International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) system. We grouped codes that met our definition of maternal morbidity into 38 clinically relevant categories. For example, we grouped transient hypertension of pregnancy (642.3), chronic hypertension (401–405, 642.0, 642.1, 642.2, 642.9), chronic hypertension with superimposed preeclampsia or eclampsia (401–405, 642.0, 642.1, 642.2, 642.4, 642.5, 642.6), preeclampsia (642.4, 642.5, 642.6), and hypertension during pregnancy, unspecified (642.9, 760.0) into a category called hypertensive disorders of pregnancy.
The majority of data on race and ethnicity and parity came from Oregon and Washington birth certificates, which were linked to the Kaiser Permanente Northwest data. The remaining variables (age, insurance type, and pregnancy outcome) were obtained from Kaiser Permanente Northwest administrative or clinical data bases. Race and ethnicity was categorized as white non-Hispanic, black non-Hispanic, Hispanic, Asian, Native American, or other; the latter two categories had insufficient numbers for meaningful comparisons. Insurance coverage status was categorized as commercial, Washington Basic Health Plan, or Medicaid; there were too few women covered by Medicare during pregnancy to permit meaningful comparisons.
We searched pregnancy episodes for ICD-9-CM codes for the 38 predetermined morbidity groups. Using these results, we estimated the prevalence of morbidity during the study period. Variations in morbidity were examined by pregnancy outcome, age, race and ethnicity, parity, and insurance payer; we used the χ2 statistic to assess statistically significant variations among groups with sufficient numbers to make comparisons. In this paper, we present first the most common complications experienced by women with all pregnancy outcomes. Thereafter, we focus on women whose pregnancy ended in a live birth. The study was approved by institutional review boards at the Centers for Disease Control and Prevention and Kaiser Permanente Northwest.
Between January 1, 1998, and December 31, 2001, there were 251,251 female Kaiser Permanente Northwest members between 12 and 55 years of age. We identified 24,481 pregnancies among 21,001 women. Most women were white, between the ages of 20 and 29, and insured by the Kaiser Permanente Northwest commercial plan. More than two thirds of pregnancies ended in a live birth (Table 1).
Table 2 displays the ICD-9-CM codes and overall prevalence for each of the 38 morbidity groups identified during the pregnancy episodes. At least one morbidity was reported in 50.4% of pregnancy episodes. Overall, the five most common complications were anemia, hypertensive disorders of pregnancy, urinary tract infections, pelvic and perineal trauma occurring at delivery, and mental health conditions.
The overall prevalence of pregnancy-related morbidity varied by the outcome: 60.3% among live births, 69.2% among stillbirths, 31.0% among spontaneous abortions, and 20.3% among therapeutic abortions (Table 3). Specific complications also differed by pregnancy outcome. Ten complications had a prevalence of 5% or greater in at least one pregnancy outcome group. Urinary tract infections were common (5–13%) in women with all outcomes, as was asthma (2–7%). Rates of anemia, hypertensive disorders of pregnancy, pelvic and perineal trauma, excessive vomiting, and postpartum hemorrhage each occurred more frequently in women who had a live birth or stillbirth. Antepartum hemorrhage and obstetric infections were most common among women with a stillbirth (11% and 22%, respectively). Rates of mental health conditions were common in women with all outcomes (7–23%); the rate was particularly elevated (23%) among women who had a stillbirth.
Limiting our analyses to pregnancies ending in a live birth, we examined the five most common complications by the woman’s race and ethnicity, age, parity, and insurance coverage (Figs. 1–4). With the exception of pelvic and perineal trauma, which did not vary by age (P=.24), and hypertensive disorders of pregnancy, which did not vary by insurance coverage (P=.11), statistically significant differences were found in the prevalence of these five complications by each of the covariates (P<.001).
The most striking differences observed were the following: Asian women were diagnosed with substantially more pelvic and perineal trauma and substantially fewer mental health conditions than women of other racial and ethnic groups; urinary tract infections declined considerably with increasing age; hypertensive disorders were more common among nulliparous women than parous women and urinary tract infections and pelvic and perineal trauma decreased as parity increased; and women who received Medicaid were diagnosed more frequently with anemia and mental health conditions than women with other insurance coverage.
Using a validated computerized algorithm, we estimated rates of maternal morbidity among women enrolled in a large HMO. A distinct feature of our study was the full spectrum of morbidity throughout pregnancy and postpartum, including morbidity identified in all treatment settings and for all pregnancy outcomes. Although prevalence and type of morbidity varied by outcome, overall, 50% of women had at least one complication during the pregnancy episode. Notably, between 7% and 23% of women were diagnosed with mental health conditions during the pregnancy episode.
Direct comparison of our results with those from studies that used only hospital discharge data3,7,8 is difficult due to marked differences in methodology. These studies reported on hospitalizations for complications, which do not provide a comprehensive picture of maternal morbidity. Hospital discharge data do not capture multiple hospitalizations for an individual woman, complications treated in an outpatient setting, or complications that occurred postpartum. A noteworthy strength of our study was the availability of all electronic, individual-level databases, which allowed us to identify morbidity not found in many other studies. The most common complications we detected—anemia, hypertensive disorders of pregnancy, urinary tract infections, and mental health conditions—usually do not require hospitalization; they would be underestimated or missed in studies that use only hospitalization data. For example, our results indicated a prevalence of mental health conditions of 7% to 23%. Others have found much lower rates, most likely because they did not have access to data sources other than hospitalizations. Proxies used for mental health conditions were “hospitalizations for mental disorders” (1.6%) and “hospitalizations for psychiatric problems” (0.7%).3,7. Because we had data on every encounter with the Kaiser Permanente Northwest health care system, we were able to estimate morbidity rates per woman.
Our study has several limitations. We had to rely on ICD-9-CM codes for diagnoses; analyses using these data are subject to an unknown extent of coding errors. However, Lydon-Rochelle et al10,11 reported reasonable sensitivities for many maternal morbidities identified by ICD-9-CM codes in hospital discharge data. Moreover, Kaiser Permanente Northwest has strong controls on diagnostic coding for inpatient and ambulatory care and conducts internal assessments of their coding practices. Regular audits are performed and have improved capture rates and accuracy of coding. External audits of coding data are conducted by the Centers for Medicare and Medicaid Services.
Information on some pregnancy episodes, particularly therapeutic abortions and deliveries at nonplan hospitals, was limited to claims data. Some covariates, such as parity, had many missing values. Data on length of gestation came from several sources. For nearly all hospital births, gestational age was obtained from medical records. When it was not documented, the algorithm searched for other information to estimate gestational age. As a last resort, outcome-specific estimates of mean gestational age based on published literature were assigned.6 To address our concern about potential errors in the algorithm’s assignment of gestational length, we conducted a validation analysis and found 94% agreement on gestational age within 4 weeks between algorithm data and medical record data.6
Numerous applications of an algorithm similar to the one we developed are possible. It could be used to institute an ongoing surveillance system of maternal morbidity in a defined population. A database of all pregnancies occurring in an HMO could be developed, permitting creation of an electronic pregnancy history file, a valuable resource for monitoring pregnancy health as well as long-term health of women. As more years of data are added, the algorithm could be used to identify less common complications. Moreover, the longitudinal nature of an ongoing system renders it possible to monitor and study the development of chronic medical conditions, such as type 2 diabetes mellitus among women who had a pregnancy with gestational diabetes or chronic hypertension among women who had a pregnancy with preeclampsia.
Better sources of maternal health data and enhanced methods for obtaining this information provide opportunities for a clearer understanding of the impact of morbidity and factors associated with it. In our study, the availability of all individual-level databases allowed us to identify pregnancy-related complications not found in many other studies. Ongoing collection, analysis, and dissemination of such data have the potential to improve services, influence policies and practice, and reduce complications of pregnancy.
1. Berg CJ, Bruce FC, Callaghan WM. From mortality to morbidity: the challenge of the twenty-first century. J Am Med Womens Assoc 2002;57:173–4.
2. Franks AL, Kendrick JS, Olson DR, Atrash HK, Saftlas AF, Moien M. Hospitalization for pregnancy complications, United States, 1986 and 1987. Am J Obstet Gynecol 1992;166:1339–44.
3. Bacak SJ, Callaghan WM, Dietz PM, Crouse C. Pregnancy-associated hospitalizations in the United States, 1999–2000. Am J Obstet Gynecol 2005;192:592–7.
4. Adams MM, Harlass FE, Sarno AP, Read JA, Rawlings JS. Antenatal hospitalization among enlisted servicewomen, 1987–1990. Obstet Gynecol 1994;84:35–9.
5. Scott CL, Chavez GF, Atrash HK, Taylor DJ, Shah RS, Rowley D. Hospitalizations for severe complications of pregnancy, 1987–1992. Obstet Gynecol 1997;90:225–9.
6. Hornbrook MC, Whitlock EP, Berg CJ, Callaghan WM, Bachman DJ, Gold R, et al. Development of an algorithm to identify pregnancy episodes in an integrated health care delivery system. Health Serv Res 2007;42:908–27.
7. Gazmararian JA, Petersen R, Jamieson DJ, Schild L, Adams MM, Deshpande AD, et al. Hospitalizations during pregnancy among managed care enrollees. Obstet Gynecol 2002;100:94–100.
8. Danel I, Berg C, Johnson CH, Atrash H. Magnitude of maternal morbidity during labor and delivery: United States, 1993–1997. Am J Public Health 2003;93:631–4.
9. Lydon-Rochelle MT, Holt VL, Martin DP, Easterling TR. Association between method of delivery and maternal rehospitalization. JAMA 2000;283:2411–6.
10. Lydon-Rochelle MT, Holt VL, Nelson JC, Cardenas V, Gardella C, Easterling TR, et al. Accuracy of reporting maternal in-hospital diagnoses and intrapartum procedures in Washington State linked birth records. Paediatr Perinat Epidemiol 2005;19:460–71.
11. Lydon-Rochelle M, Holt VL, Cardenas V, Nelson JC, Easterling TR, Gardella C, et al. The reporting of pre-existing maternal medical conditions and complications of pregnancy on birth certificates and in hospital discharge data. Am J Obstet Gynecol 2005;193:125–34.
© 2008 The American College of Obstetricians and Gynecologists
12. Andrade SE, Gurwitz JH, Davis RL, Chan KA, Finkelstein JA, Fortman K, et al. Prescription drug use in pregnancy. Am J Obstet Gynecol 2004;191:398–407.