Background
The Obstetrics Adverse Outcomes Index was designed to measure the quality of perinatal care and includes 10 adverse events that may occur at or around the time of delivery. We hypothesized that adverse outcomes in the labor and delivery suite, including hypoxic ischemic encephalopathy, could be decreased with a combination of interventions, even among high-risk pregnancies.
Objective
The objective of the study was to evaluate the impact of a labor and delivery care bundle on adverse obstetrics outcomes as measured by a modified Obstetrics Adverse Outcomes Index, Weighted Adverse Outcomes Index, and Severity Index.
Study Design
This is a retrospective cohort study including all women who delivered at our academic, tertiary care institution over a 3 year period of time, before and after the implementation of an intervention to decrease adverse outcomes. Outcome measures consisted of previously reported indices that were modified including the addition of hypoxic ischemic encephalopathy. The adverse outcomes index is a percentage of deliveries with 1 or more adverse events, the weighted adverse outcomes index is the sum of the points assigned to cases with adverse outcomes divided by the number of deliveries, and the severity index is the sum of the adverse outcome scores divided by the number of deliveries with an identified adverse outcome. A segmented regression analysis was utilized to evaluate the differences in the level and trend of each index before and after our intervention using calendar month as the unit of analysis.
Results
During the study period, 5826 deliveries met inclusion criteria. Comparing the pre- and postintervention periods, high-risk pregnancy was more common in the postintervention period (73.5% vs 79.4%, P < .001). Overall, there was a decrease in both the Modified Weighted Adverse Outcomes Index ( P = .0497) and the Modified Severity Index ( P = 0.01) comparing the pre- and postintervention periods; there was no difference in the Modified Adverse Outcomes Index ( P = .43). For low-risk pregnancies, there was no significant difference in the levels for any of the measured indices over the study period ( P = .61, P = .41, and P = .34 for the Modified Adverse Outcomes Index, Modified Weighted Adverse Outcomes Index, and Modified Severity Index, respectively). Among the high-risk pregnancies, the monthly Modified Weighted Adverse Outcomes Index decreased by 4.2 ± 1.8 ( P = .03). The monthly Modified Severity Index decreased by 53.9 ± 17.7 points from the pre- to the postintervention periods ( P = .01) and was < 50% of the predicted Modified Severity Index had the intervention not been implemented. The cesarean delivery rate was increasing prior to the intervention, but the rate was stable after the intervention, and the absolute rate did not differ between the pre- and the postintervention periods (28.4% vs 30.0%, P = .20).
Conclusion
Overall and for high-risk pregnancies, the implementation of the labor and delivery care bundle had a positive impact on the Modified Weighted Adverse Outcomes Index and Modified Severity Index but not the Modified Adverse Outcomes Index.
Recently, attention has been focused on establishing and reporting obstetrics quality indicators. In 2006, Mann et al published a set of 10 outcome measures and 3 quality improvement tools. The 3 quality improvement tools developed included the Adverse Outcomes Index (AOI), Weighted Adverse Outcomes Index (WAOI), and the Severity Index (SI). The AOI is the percentage of deliveries with 1 or more adverse events, the WAOI is the adverse score per delivery (the sum of the points assigned to cases with adverse outcomes divided by the number of deliveries), and the SI designates the severity of the outcomes (sum of the adverse outcome scores divided by the number of deliveries with an identified adverse outcome). Some but not all of the proposed measures have been endorsed by the National Quality Forum.
Previous studies have evaluated the AOI including the impact of specific interventions on the AOI. Pettker et al reported an unintended increase in cesarean delivery rates after implementing multiple interventions to improve patient safety at a university hospital. This study, however, did not include preintervention data and trends or account for the transition period following the intervention and thus was not able to account for secular trends. Additionally, the previously reported AOI lacked proven validity in a high-risk population and did not include hypoxic ischemic encephalopathy (HIE) as an outcome.
HIE is a subcategory of neonatal encephalopathy and is defined as the clinical syndrome of disturbed neurological function in the early days of life in an infant born at ≥ 35 weeks’ gestation. Although uncommon, occurring in 2–6 per 1000 deliveries, HIE can be devastating for families including prolonged intensive care hospitalization and expense. HIE is the most common and costly birth injury claim among medical malpractice lawsuits.
The objective of this study was to compare the modified indices including HIE before and after the implementation of our labor and delivery patient safety bundle using robust methodology and to evaluate the impact on cesarean delivery rates. We modified the AOI elements proposed by Mann et al including the addition of HIE, thus creating the Modified Adverse Outcomes Index (M-AOI), Modified Weighted Adverse Outcomes Index (M-WAOI), and the Modified Severity Index (M-SI).
Materials and Methods
The protocol for this cohort study was approved by the Mayo Clinic Institutional Review Board. All women who delivered at the Mayo Clinic Rochester from Jan. 1, 2011, through Dec. 31, 2013, were considered. Deliveries were not included in the analysis if the research authorization status was declined for either the mother or infant in accordance with Minnesota law. Delivery of multiple gestations was counted as a single delivery. Pregnancies were categorized as high-risk or low-risk based on the presence or absence of specific risk factors derived from the Eunice Kennedy Shriver National Institute of Child Health and Human Development document on high-risk pregnancy ( Appendix 1 ).
At our tertiary care academic center, we perform approximately 2500 deliveries annually and serve as a referral center for an additional 6000 annual deliveries within the Mayo Clinic Health Systems. All providers are salaried Mayo Clinic employees who deliver patients at a single Mayo Clinic-owned hospital with no private or community patients or providers. Prior to 2012, the labor and delivery unit was staffed by a certified nurse midwife and resident staff and supervised by an attending obstetrician who was not required to be on the labor and delivery unit at all times.
In the spring of 2012, an event review was conducted following a case of HIE. The results of the root cause analysis were consistent with the analyses performed for previous HIE event reviews, and it was concerning to note the frequency of the diagnosis was increasing. In the majority of the HIE cases, there was a demonstrated lack of oversight by the attending obstetrician and ineffective communication among team members.
Another commonality was the variation in the interpretation of fetal heart rate tracings and the frequent loss of situational awareness in cases of prolonged labor. There was little standardization of practice, leading to a wide variation in approaches based on the team members who were staffing a delivery.
As a result of these findings, our institution implemented a labor and delivery care bundle with the primary aim of reducing cases of HIE. The care bundle included the following initiatives: (1) dedicated obstetric attending presence on the unit 24 hours a day (the majority of shifts covered by a laborist physician); (2) communication training for all providers including physicians, certified nurse midwives, and nurses; (3) mandatory use of a labor partogram; (4) mandatory fetal heart rate tracing interpretation training (K2 Medical Systems); and (5) regular feedback on individual provider AOI. The unit is staffed 24 hours a day, 7 days a week with obstetricians, certified nurse midwives, anesthesiologists, and consultative maternal-fetal medicine specialists. All women are Mayo Clinic patients cared for under the supervision of attending obstetricians including a minority of patients primarily managed by certified nurse midwives or family medicine attending physicians.
Maternal demographics and maternal and neonatal outcomes data were collected from a combination of manual extraction from the electronic medical record and diagnosis codes. Manually extracted data included gravidity, parity, gestational age at delivery, mode of delivery, and neonatal outcomes including disposition, birthweight, and Apgar scores. Other variables were extracted from the electronic medical record including maternal age, race, and baseline body mass index (BMI). Diagnosis codes were utilized to identify high-risk factors ( Appendix 1 ), HIE ( International Classification of Codes , ninth revision, codes 768.70, 768.71, 768.72, and 768.73), and other adverse events.
The collected data were utilized to derive the M-AOI, M-WAOS, and M-SI outcome measures including the incorporation of HIE ( Appendix 2 ). HIE was assigned 350 points, 50 points less than neonatal death because of the long-term care impact of HIE. Maternal death was increased to 1100 such that the total elements did not exceed maternal death.
The primary outcomes of the study were the 3 modified outcomes indices. The secondary outcome was the cesarean delivery rate.
Statistical analyses were performed using the SAS version 9.3 software package (SAS Institute, Inc, Cary, NC). Baseline maternal characteristics and delivery outcomes were compared between the pre- and postintervention periods using the 2-sample Student t test for maternal age, the Wilcoxon rank-sum test for gravidity and parity, and the χ test for categorical variables. Segmented regression analysis, appropriate for an interrupted time series design, was utilized to evaluate differences in the level and trend of each outcome index before and after the intervention, using calendar month as the unit of analysis.
The outcomes were analyzed for all pregnancies combined, and separately by high- and low-risk pregnancy status. There were a total of 36 monthly intervals: 14 during the preintervention period (January 2011 through February 2012), 4 during the transition period (March 2012 through June 2012), and 18 during the postintervention period (July 2012 through December 2013).
The outcomes in the transition period were not included in the analysis. For each outcome, the following regression model was fit: E(Y) = β 0 + β 1 time + β 2 intervention + β 3 time after intervention + error, where time denotes the number of months sequentially from the first month in the baseline period, intervention is an indicator variable denoting the 2 time periods (before and after the implementation of the intervention), time after intervention denotes the number of months after the intervention, β 0 represents a constant term for the level of the outcome at the start of the preintervention period, β 1 represents the change in outcome per month (ie, linear trend) during the preintervention period, β 2 represents the change in the level of the outcome after the implementation of the intervention, and β 3 represents the change in the linear trend after the implementation of the intervention, compared with the trend during the preintervention period.
The regression models were fit using SAS Proc Autoreg. The Durbin-Watson (D-W) statistic was evaluated for evidence of serial autocorrelation. Because the Durbin-Watson statistic indicated a lack of autocorrelation, the models were not corrected for autocorrelation.
Results
During the study period, 7137 deliveries occurred at our institution; 5826 deliveries met inclusion criteria. There were 2349 deliveries in the preintervention period, 621 in the intervention period, and 2856 in the postintervention period. Comparing the pre- and postintervention periods, there was no difference in the mean maternal age in years (29.7 [SD 5.2] vs 29.9 [SD, 5.0], P = .17), obesity (BMI ≥ 30 kg/m 2 ) (23.6% vs 21.8%, P = .20), or median parity (2 [1, 3] vs 2 [1, 3], P = .62) ( Table 1 ). High-risk pregnancy was more common in the postintervention period (73.5% vs 79.4%, P < .001).
| Characteristic | Before intervention (n = 2349) | After intervention (n = 2856) | P value a |
|---|---|---|---|
| Age, y, mean (SD) | 29.7 (5.2) | 29.9 (5.0) | .17 |
| Race, n, % | .73 | ||
| American Indian/Alaska Native | 5 (0.2) | 5 (0.2) | |
| Asian | 109 (4.6) | 145 (5.1) | |
| Black | 135 (5.7) | 168 (5.9) | |
| White | 1913 (81.4) | 2340 (81.9) | |
| Native Hawaiian/Pacific Islander | 2 (0.1) | 1 (0.04) | |
| Other | 138 (5.9) | 154 (5.4) | |
| Unknown/chose not to disclose | 47 (2.0) | 43 (1.5) | |
| BMI, kg/m 2 b | .20 | ||
| < 30.0 | 1114/1459 (76.4) | 1757/2248 (78.2) | |
| ≥ 30.0 | 345/1459 (23.6) | 491/2248 (21.8) | |
| Gravidity, median (IQR) c | 2 (1, 3) | 2 (1, 3) | .88 |
| Parity, median (IQR) c | 2 (1, 3) | 2 (1, 3) | .62 |
| Previous cesarean delivery, n, % | 413 (17.6) | 482 (16.9) | .50 |
| VBAC, n, % | 91 (3.9) | 123 (4.3) | .43 |
| High-risk pregnancy, n, % | 1726 (73.5) | 2269 (79.4) | < .001 |
a BMI was calculated from available height and weight measurements closest to gestational age of 0 (possible weight measurements from 9 months prior to gestational age of 0-12 weeks after the delivery date; height from 1 year prior to gestational age of 0-1 year after the delivery date). BMI was unavailable for 890 (37.9%) in the preintervention group and 608 (21.3%) in the postintervention group
b Gravidity and parity reports include the current pregnancy
c χ 2 P value is presented for categorical variables, Wilcoxon rank-sum P value is presented for gravidity and parity, and t test P value is presented for age.
Trends in the monthly values of the modified indices are shown in Figure 1 . Seven of the 5826 deliveries were complicated by HIE; all 7 of these had concurrent events. Thus, the incidence of HIE during the study was 1.2 per 1000 deliveries. Overall, there was a decrease in both the M-WAOI ( P = .0497) and the M-SI ( P = .01) comparing the pre- and postintervention periods; there was no difference in the M-AOI ( P = .43). For low-risk pregnancies, there was no significant difference in the levels for any of the measured indices over the study period ( P = .61, P = .41, and P = .34 for the M-AOI, M-WAOI, and M-SI, respectively) ( Figure 2 ).
Based on the high-risk pregnancies, at the start of the preintervention period, the average monthly M-AOI was 9.3 ( Figure 2 ). Over the course of the 14 month preintervention period, the estimated mean M-AOI decreased nonsignificantly at a rate of –0.07 ± 0.16 per month ( P = .66 for testing whether the slope is different from zero). After the transition period, the estimated mean monthly M-AOI decreased by 2.0 ± 1.7 ( P = .24).
Over the course of the 18 month postintervention period, the estimated mean M-AOI increased at a rate of 0.15 per month ( P = .26 for testing the difference in slopes between the 2 time periods). Based on the regression model, the estimated mean monthly M-AOI for the final month of the study period was 9.0 for high-risk pregnancies. However, had the intervention not been implemented and if the trend during the preintervention period were to have continued, the predicted M-AOI for the final month of the study period would have been lower (7.0) for high-risk pregnancies.
For the M-WAOI among the high-risk pregnancies, the slope or trend in the monthly index was stable across the preintervention period ( P = .59), and this slope was not significantly different before and after the intervention ( P = .97) ( Figure 2 ). However, at the start of the preintervention period, the average monthly M-WAOI was 4.6, and after the transition period, the estimated mean monthly M-WAOI decreased by 4.2 ± 1.8 ( P = .03), indicating an immediate intervention impact.
Based on the high-risk pregnancies, at the start of the preintervention period, the average monthly M-SI was 39.6 ( Figure 2 ). However, over the course of the 14 month preintervention period, the estimated mean M-SI increased at a rate of 3.1 ± 1.7 per month ( P = .07). After the transition period, the estimated mean monthly M-SI decreased significantly by 53.9 ± 17.7 ( P = .01).
Over the course of the 18 month postintervention period, the estimated mean M-SI gradually increased but only at a rate of 0.4 per month. Based on the regression model, the estimated mean monthly M-SI for the final month of the study period was 35.7 for high-risk pregnancies. However, had the intervention not been implemented and if the trend during the preintervention period were to have continued, the predicted M-SI for the final month of the study period would have been considerably higher (exceeded 100) for high-risk pregnancies. Thus, the M-SI was < 50% of the predicted M-SI had the intervention not been implemented.
Although the difference did not reach statistical significance, there was a decrease in deliveries of < 37 weeks’ gestation during the postintervention period (11.5% compared with 13.2%, P = .07). There was no difference in mode of delivery ( P = .30), multiple gestations ( P = .11), or neonatal birthweight < 2500 g ( P = .43) ( Table 2 ). The cesarean delivery rate was increasing prior to the intervention, but the rate was stable after the intervention, and the absolute rate did not differ between the pre- to the postintervention periods (28.4% vs 30.0%, P = .20) ( Figure 3 ).
