Anesthesia-related Deaths during Obstetric Delivery in the United States, 1979–1990

In 1987, the Centers for Disease Control and Prevention (CDC) established an ongoing National Pregnancy Mortality Surveillance System, which is a system to monitor maternal deaths at the national level and conduct epidemiologic studies of the deaths of pregnant women. [7]Health departments in all 50 states, the District of Columbia, and New York City provided the CDC with copies of death certificates (with patient and provider identification removed) and, when available, the linked pregnancy outcome records (birth certificates and fetal death records) for all identified pregnancy-related deaths from 1979 through 1990.

A woman's death was classified as pregnancy related if it occurred during pregnancy or within 1 yr after delivery and resulted from (1) complications of the pregnancy itself, (2) a chain of events initiated by the pregnancy, or (3) the aggravation of an unrelated condition by the physiologic or pharmacologic effects of pregnancy. [2]Clinical epidemiologists reviewed each pregnancy-related death certificate and, when available (89% of cases), the matched pregnancy outcome record for a possible relation to anesthesia. Three obstetric anesthesiologists independently reviewed the maternal death certificates and matched live birth or fetal death certificates for each of the cases to confirm that the death resulted from a complication of anesthesia. If all three anesthesiologists confirmed from the vital records information that the death resulted from an anesthesia-related complication, they then determined the cause of death, procedure for delivery, and type of anesthesia provided. Because vital records are often incomplete concerning the events surrounding the death, if all three anesthesiologists evaluating the records independently could not agree, the information was coded as unknown.

We classified deaths by the CDC system, which distinguishes among the immediate and underlying causes of death as stated on the death certificate, associated obstetrical conditions or complications, and the outcome of pregnancy. [2]Complications of anesthesia leading to death were further categorized as airway management problems, which included aspiration of gastric contents, problems with induction or intubation during general anesthesia and esophageal intubation, inadequate ventilation, and respiratory failure; high spinal or epidural block; toxicity of local anesthetics; drug reactions, anaphylaxis, or both; overdose of sedatives; or unknown cause. Because some of the death certificates only listed cause of death as “cardiac arrest,” these deaths were included in a category labeled “unspecified intraoperative cardiac arrest during anesthesia.”

All pregnancy-related deaths (n = 4,097) submitted to the CDC for 1979–1990 were reviewed, and it was determined that 155 deaths were caused by anesthesia-related complications. After an initial analysis of these cases, the 20 abortion-related deaths and the six deaths related to ectopic pregnancy were excluded so that 129 cases associated with an obstetric delivery (live births or stillbirths) were evaluated. We calculated pregnancy mortality rates per million live births using national data on live births from the 1979–1990 natality files of the National Center for Health Statistics.

To compare the risks of general versus regional anesthesia during cesarean section, we estimated case-fatality rates for two time periods: 1979–1984 and 1985–1990. We used these periods for three reasons:(1) The annual number of anesthesia-related deaths were too small to compute reliable annual rates, (2) the two time periods divided the study period into equal 6-yr blocks, and (3) changes in the clinical practice of obstetric regional anesthesia occurred after 1984 with the withdrawal of 0.75% bupivacaine for epidural injection from clinical practice. This latter change may have influenced rates of mortality from local anesthetic toxicity.

Case-fatality rates were computed by the following procedure. First, we determined the national number of live births for each year during the period 1979–1990. Second, we applied national cesarean section rates for each year, derived from National Hospital Discharge Survey data, to the number of live births, and then calculated the number of cesarean section deliveries for each of the two periods. [8]This estimate was then apportioned to administration of general and regional anesthesia according to estimated percentages from survey data from obstetric anesthesia personnel. For the years 1979–1984, the percentages of women who received regional or general anesthesia for cesarean section were estimated from a survey of obstetricians and anesthesia personnel conducted in 1981. [9]The average cesarean section rate during that time was 19%[8]; general anesthesia was used in approximately 41% of cases, and regional anesthesia was used in approximately 55%. [9](According to the authors, these values add up to less than 100% because some respondents failed to answer all questions.) For the period 1985- 1990, we obtained data on anesthesia administration from a similar survey of obstetricians and anesthesia personnel conducted in 1992. [10]Although 1992 falls outside the study period, no major changes in obstetric anesthesia practice occurred from 1990–1992. The average cesarean section rate during the later period was 24%, [8]the use of general anesthesia had declined to approximately 16% of cesarean sections, and regional anesthesia was used in approximately 84% of cases. [10] 

The two manpower surveys used for these estimates were conducted in a similar manner. [9,10]Using the American Hospital Association registry, hospitals were differentiated by number of births occurring in that year and by U.S. Census region, and a stratified random sample of hospitals was selected. Two copies of the survey were sent to the administrator of each hospital, one for the chief of obstetrics and one for the chief of anesthesiology. In 1981, 1,193 surveys were returned (48% from obstetricians and 52% from anesthesiologists) and responding hospitals represented about 30% of the total births in the United States that year. In 1992, 902 surveys were returned (51% from obstetricians and 49% from anesthesiologists) and births at responding hospitals represented 56% of the total U.S. births occurring that year.

Finally, we calculated case-fatality rates by dividing the number of deaths associated with general or regional anesthesia during cesarean section in a given period by the estimated number of administrations of that type of obstetric anesthesia during cesarean section delivery in the same period. We used regional anesthesia as the referent for comparison with the risk associated with general anesthesia. For the resulting case-fatality rates and risk ratios, we based 95% confidence intervals (CI) on the logit case-fatality rate and used the Taylor series method for variances. [11] 

To compare U.S. data for 1979–1990 with data from England and Wales, we calculated pregnancy mortality rates by triennium. Because we based the denominator for U.S rates on the number of live births, whereas the rates from England and Wales are based on all “maternities”(tile count of mothers who delivered live births, stillbirths, pregnancy terminations, ectopic pregnancies, and abortions), these rates are not perfectly analogous but should allow for some degree of comparison. [5] 

Characteristics of the 129 women who died of complications of anesthesia during delivery in the study period 1979–1990 are shown in Table 1. Seventy-nine percent of these women were 20 to 34 yr old, and 52% were black. They tended to have 12 or more years of education and to have begun prenatal care in the first trimester of pregnancy (Table 1). Most women who died from complications of anesthesia in this study were undergoing a cesarean delivery (82%). Approximately 5% of the deaths were associated with a vaginal delivery, and for 13% the delivery procedure could not be determined.

The causes of anesthesia-related death varied by the type of anesthesia administered (Table 2). The 67 women who died of complications of general anesthesia (52% of the deaths) primarily died of airway management problems, which included aspiration, induction or intubation problems, inadequate ventilation, and respiratory failure. About one fourth (n = 33) of the 129 deaths from anesthesia-related complications were associated with problems that occurred during the administration of regional anesthesia. Most deaths from regional anesthesia (70%) occurred among women who had epidural anesthesia; the remaining 30% were associated with spinal anesthesia. These deaths usually resulted from local anesthetic toxicity or an inadvertent high spinal or epidural block. Four deaths (3%) were related to complications of parenteral opioids or sedatives. In three or four cases, airway management problems were the cause of these deaths. The type of anesthesia was not available in the vital records for 25 (19%) of the women, but the causes of death appeared to be similar to those for women who died of complications of general anesthesia (Table 2). Cardiac arrest during anesthesia was listed as the only cause of death in 30 cases (23% of the total), but information was inadequate to determine the underlying cause. Analysis by type of anesthesia showed that 50% of these events occurred during general anesthesia and 7% during regional anesthesia; in 43% of these cardiac arrest cases we could not determine the type of anesthesia used.

The type of anesthesia associated with maternal deaths changed during the 12-yr study period (Figure 1). The number of deaths from complications of anesthesia during delivery decreased with time. In the first half of the study (trienniums 1979–1981 and 1982–1984), deaths from general anesthesia accounted for about 41% of the cases and deaths from regional anesthesia accounted for 28–31%. Most deaths related to general anesthesia during this period resulted from aspiration and intubation problems, and most deaths related to regional anesthetic resulted from local anesthetic toxicity (Table 2). In contrast, after 1984 there was an abrupt decrease in deaths due to regional anesthesia. Fewer deaths due to local anesthetic toxicity were reported after 1984. The number of anesthesia-related deaths in which cause of death and characteristics of the woman were unknown also decreased with time.

We calculated case-fatality rates for women who had a cesarean section delivery with general or regional anesthesia for the two periods 1979–1984 and 1985–1990 (Table 3). The estimated rate of death from complications of general anesthesia during cesarean section increased from 20 deaths per million general anesthetics administered in the earlier time period to 32.3 deaths per million during the latter. In contrast, the rate of death during cesarean section delivery using regional anesthesia decreased from 8.6 per million regional anesthetics administered to 1.9 per million during the last half of the study period. From 1979–1984, the risk of death from complications of general anesthesia during cesarean delivery was 2.3 (95% CI, 1.9–2.9) times higher than that for regional anesthesia. From 1985–1990, the rate of death related to regional anesthesia decreased, and concomitantly the use of regional anesthesia for cesarean section delivery increased. Therefore the risk of death from complications of general anesthesia increased to 16.7 (95% CI, 12.9–21.8) times that for regional anesthesia.

We have used the format of the Report on Confidential Enquiries into Maternal Deaths in the United Kingdom to show anesthesia-related maternal mortality rates (deaths per million live births or maternities) in the United States and England and Wales by triennium (Table 4). In the United States, this maternal mortality rate has decreased in each subsequent 3-yr period from 4.3 per million in 1979–1981 to 1.7 per million in 1988–1990. The rates in England and Wales have also decreased.

Maternal mortality rates have decreased during the 12-yr period of this study. This decrease is reflected by data from both the United States and England and Wales. [12]The 1991 Confidential Enquiries report noted, “The actual reduction in the direct mortality rate due to anaesthesia is probably greater than that suggested by the decrease in the number of deaths, since the number of anaesthetics for operative procedures continues to rise.”[5]This position is also supported by data from the United States. [8]Cesarean delivery rates in the United States increased from 16% to 24% during the years of our study. Anesthesia-related deaths primarily occur during cesarean section, with at least 82% in this study (Table 1). Other studies have found that the risk for maternal death is significantly greater for women undergoing cesarean section than for those who have a vaginal delivery. [13]The increased risk of cesarean delivery may be related not only to the labor or delivery complications necessitating the cesarean section but also to the specific risks of anesthesia for this method of delivery. Management of anesthesia for cesarean section is more complex than for vaginal delivery and is more likely to involve general anesthesia and airway management. In this country, the American Society of Anesthesiologists Closed Claims Study has also shown that maternal death claims are predominantly related to use of general anesthesia. [14]Furthermore, 100% of parturients are exposed to the risk of anesthesia when they have a cesarean delivery. In contrast, a 1981 survey of obstetricians and anesthesiologists showed that only 16% of parturients received an anesthetic for labor (other than parenteral medications). [9]A similar survey in 1992 showed that 37% received regional analgesia. [10] 

This study represents the first assessment of national data on maternal deaths related to anesthesia in the United States. Because this report is based on a national surveillance system and vital records, it lacks the detailed case data used in the reports from the United Kingdom. Nevertheless, it provides the best possible view of trends in anesthesia-related death in the United States. Previous studies of maternal death due to complications of anesthesia have been conducted only on a state level. [15–17]However, the number of anesthesia-related deaths for each state is so small it precludes definitive analysis. In addition, concerns about confidentiality when such small numbers are involved have curtailed the activities of many state committees on maternal death. [6,18]When studies use maternal mortality data from a national data bank, with larger numbers, confidentiality is better preserved and results have wider applicability.

Several important limitations of this study should be considered. First, not all maternal deaths are identified as such on death certificates. For example, if a woman dies from complications of pulmonary aspiration after an extended 6-week stay in an intensive care unit, the person completing the death certificate may not remember to indicate that the woman was pregnant. It has been estimated that as many as 37% of maternal deaths are missed due to underreporting on vital statistics records. [19,20]Second, because death certificates are not designed to collect medical data and are completed by persons with varying backgrounds and training, these certificates often do not contain enough detail to determine the circumstances and pathophysiologic conditions leading to the maternal death. For example, the cause of death may be listed simply as “cardiac arrest under anesthesia.” In this instance, not only was the true cause of death unknown, but the type of anesthesia was also unknown. Nevertheless, it is obvious that the cause of death was related to anesthesia. Finally, reporting bias can lead to missed cases. In the review by Kaunitz et al. [6]of maternal mortality in the United States, the investigators reported that some deaths reportedly due to amniotic fluid embolism may actually have been due to complications of anesthesia. Syverson et al. [21]also noted that deaths classified as “cardiac arrest” on vital records often mentioned complications of anesthesia when autopsy reports or medical records were reviewed.

In this study, all cases mentioning anesthesia as a cause of death were included. If all three anesthesiologists evaluating the records independently could not agree on the rest of the information, it was coded as unknown. This approach was in part responsible for much of the of data classified as “unknown.” However, it reduced the extent of speculation as to what happened.

Our results indicate that the number of deaths due to complications of general anesthesia, although small, is not decreasing over time-a discouraging finding given the marked decrease in the use of general anesthesia in the last decade. The decrease in anesthesia-related deaths seems to have been due to a decrease in complications of regional anesthesia, but because of large categories of unknowns, this interpretation is offered with caution. During the study period there was a marked increase in the use of regional anesthesia. In 1981, regional anesthesia was used in 55% of cesarean deliveries, whereas in 1992 it was used in 84%. [9,10]In contrast, use of general anesthesia for cesarean delivery has decreased sharply, from 41% to 16%, during the same period. The increased relative risk of general anesthesia may result from the fact that regional anesthesia was being used for lower-risk elective procedures in the later time period (1985–1990), such as patients requesting labor analgesia or elective cesarean sections. General anesthesia may have been used primarily in emergency situations, in patients for whom regional anesthesia was difficult to administer (a morbidly obese patient), or when regional anesthesia was contraindicated (a patient with a coagulation disorder due to hemorrhage or preeclampsia). [22]This hypothesis is supported by the findings of Endler et al., [17]who noted three risk factors for anesthesia-related maternal death: obesity (present in 80% of cases), emergency surgery (occurring in 80% of cases), and hypertension (present in 53% of cases), which are all potential reasons why general anesthesia rather than regional anesthesia might be necessary. Morbid obesity has been recognized in other analyses as a risk factor for airway problems and death from complications of anesthesia. [23]Although patient weight was not available for most cases we analyzed, when additional records were available, obesity was frequently noted in “can't intubate-can't ventilate” situations. Morgan et al. [24]recommend that a “prophylactic” epidural catheter be placed in patients in whom intubation is judged to be potentially difficult so that regional anesthesia can be used if emergent delivery is needed.

In 30 cases, the cause of death was listed as cardiac arrest, and a more direct cause could not be determined from the data available (Table 2). In the cases in which anesthesia type was known, 15 of 17 involved general anesthesia. In addition, data pooled from cases in the Confidential Enquiries revealed that 66% of cardiac arrest deaths involved airway management problems (esophageal intubation, difficult tracheal intubation, postoperative hypoventilation, and equipment failure such as a ventilator disconnect or kinked endotracheal tube). [22]Thus it is unlikely that a significant number of these deaths were related to regional anesthesia as described in the Closed Claims analyses. [25]However, without further information about each case, we cannot know definitively.

We were intrigued that the number deaths due to local anesthetic toxicity decreased abruptly after 1984 (Figure 1). The decrease in deaths due to regional anesthesia seems temporally related to the national debate about 0.75% bupivacaine, its relation to large intravenous doses having special cardiotoxic effects, and its discontinuance from obstetric anesthesia practice in 1984. [26,27]* At that time, anesthesiologists developed an increased awareness of the toxicity of local anesthetics, and protocols for test doses and fractionation of large doses of local anesthetics became standard practice. Our results reveal a reduction in deaths temporally related to these changes in practice, although we did not have information on the use of a test dose or the total doses of local anesthetic used.

Case-fatality rates computed in this analysis (Table 3) must be interpreted with caution because they were computed from other estimations (see Materials and Methods). Nevertheless, they indicate a notable difference in the risk for death from general rather than regional anesthesia complications; the risk ratio has increased eight times from 1979–1984 to 1985–1990. Although the case-fatality rates from regional anesthesia have decreased markedly, the great difference in risk during delivery is still cause for concern. If we assume that the patient with a more acute or complex condition (e.g., severe preeclampsia or morbid obesity) is more likely to receive general anesthesia in some settings, the apparent higher risk from complications of general anesthesia may be, in part, a result of the higher acuity level of these patients.

To ensure continuing awareness of the anesthetic causes of maternal death and to improve preventive measures, more states should develop active maternal mortality committees and involve anesthesiologists in case reviews. Efforts should be made to improve availability and reporting of information regarding maternal deaths. More complete information about each maternal death associated with obstetric anesthesia and continued comprehensive studies of maternal deaths due to complications of anesthesia will allow researchers and clinicians to develop strategies to prevent this type of death.

The authors thank Paul M. Gargiullo, Ph.D., for statistical assistance.

*Abbott Laboratories: Letter to doctors: Urgent new recommendations about bupivacaine. Westboro, MA, Astra Pharmaceutical Products, Breon Laboratories, 1984.