- Email: [email protected]
Preterm Contractions in Community Settings: II. Predicting Preterm Birth in Women With Preterm Contractions
Preterm Contractions in Community Settings: II. Predicting Preterm Birth in Women With Preterm Contractions WILLIAM J. HUESTON, MD Objective: To examine risk factors for preterm delivery in women who present to nontertiary care hospitals with preterm contractions. Methods: Women who presented to a network of community hospitals in Wisconsin with preterm contractions were followed until delivery. The main outcomes were preterm delivery before completion of 36 weeks of gestation and delivery within 1 week of initial presentation. Results: Of the 266 women presenting with contractions over the 2-year study period, 90% (n 5 239) consented to participate. Multiple factors were associated with premature delivery, but when examined with a multivariate model, only four (ruptured membranes, multiple gestation, cervical effacement at least 80%, and nonwhite race) were associated with prematurity, whereas five (ruptured membranes, multiple gestation, cervical effacement at least 80%, dilation exceeding 1 cm, and being a nonsmoker) predicted delivery within 1 week. A decision model that combined the presence of ruptured membranes and effacement at least 80% could predict delivery at 34 weeks or less within 7 days of presentation with a sensitivity of 71% and a specificity of 98%. Conclusion: Only two clinical cues (ruptured membranes and effacement of the cervix of 80% or more) can predict premature delivery within 7 days. If these results are confirmed prospectively, women with either of these signs could be targeted for administration of steroids, antibiotics, or transfer to tertiary care facilities. (Obstet Gynecol 1998;92: 43– 6. © 1998 by The American College of Obstetricians and Gynecologists.)
The management of preterm labor is complicated by conflicting reports regarding the efficacy of some treatment strategies and confusion about when effective strategies should be used. Disappointing results regarding the effectiveness of tocolytic agents in wellFrom the Department of Family Medicine, University of WisconsinMadison Medical School, Madison, Wisconsin. This work was supported by a Robert Wood Johnson Generalist Faculty Scholar Award to Dr. Hueston.
VOL. 92, NO. 1, JULY 1998
controlled studies raise questions about whether any effective treatment strategies exist.1,2 Antenatal steroids have been shown to reduce the morbidity associated with prematurity3 and appear to be cost-beneficial4 but are under used.5 Finally, some evidence suggests that any treatment of preterm contractions after 33 weeks’ gestation6 or after premature rupture of membranes7 is not cost-effective and that women with these conditions should be allowed to deliver. One of the dilemmas facing clinicians who treat women with preterm contractions is recognizing which patients will and which patients will not deliver prematurely. Steroids, antibiotics, and the transfer of patients to level III nursery facilities all hinge on the ability to recognize which patients are likely to deliver soon. The purpose of this study was to attempt to define a small number of clinical factors that could identify which women with preterm contractions were likely to deliver prematurely and shortly after presentation.
Materials and Methods The study population was drawn from a network of 11 community hospitals located within a 60-mile radius of Eau Claire, Wisconsin. No facilities with level III neonatal nurseries are located in this area. Because referral of patients with continued contractions can be a source of selection bias and because one of our potential outcomes was the referral of a patient to a tertiary care facility, we limited the hospital sample to only level I and level II facilities. Women enrolled in the study consented to the review of their prenatal and delivery records and those of their newborn following delivery. Two hundred and thirty-nine (90%) of the 266 women invited to participate consented. Details about the hospitals participating in the study and data abstraction are reported elsewhere.8 Data were analyzed using bivariate comparisons of
0029-7844/98/$19.00 PII S0029-7844(98)00108-2
43
risk factors, institutional and clinician characteristics, and obstetric variables using x2 for categoric variables and the t test for continuous variables. Using those variables that were associated (at P , .10) with preterm, very preterm delivery, or delivery within 7 days of presentation, logistic regression models were constructed to adjust for potential confounding. Variables included in the regression models included all clinical variables with P , .05. Regression analyses were performed using True Epistat, version 2.0 (Epistat Services, Richardson, TX). On the basis of results of the regression analysis, prediction models then were constructed using all clinical variables that were associated independently with each outcome variable; sensitivity, specificity, and prediction accuracy were calculated.
Results Seventy-eight (33%) of the 239 women in the study sample delivered prematurely; 52 (67%) of them delivered after 34 weeks’ gestation but before 37 weeks, and the remaining 26 (33%) delivered before 34 weeks’ gestation. To assess which factors could predict preterm birth, we performed bivariate analysis with a range of institutional, patient obstetric history, current pregnancy, and presenting clinical findings with the chance of preterm delivery (Tables 1 and 2). When we examined institutional and historic variables, we found several that were associated with preterm delivery (Table 1). Obstetricians tended to provide care for a greater proportion of women who delivered prematurely, especially those who delivered before 34 weeks’ gestation. This relationship remained significant even after stratifying for multiple-gestation pregnancies, which were more likely to be managed by obstetricians. Other factors associated with preterm delivery included nonwhite race, being employed outside the home, and nulliparity. Interestingly, being a nonsmoker increased the chance of delivery. The factors relating to the current pregnancy that were associated with preterm delivery included multiple gestation and rupture of membranes (Table 2). Additionally, cervical dilation greater than 1 cm and effacement of 80% or more were associated with preterm delivery. The frequency of contractions was not associated with prematurity. To adjust for confounding, we examined logistic regression models using factors that could be used by clinicians to help make a decision about rendering treatment with steroids, antibiotics, or by transferring a patient to a tertiary facility. Models evaluating two outcomes were constructed: one model used prematurity as an outcome and the second used delivery with 7 days as the outcome measure. Independent variables
44 Hueston
Predicting Preterm Birth
Table 1. Effect of Institutional, Clinician, and Past Obstetric History on Risk of Preterm Delivery Gestational age at delivery ,34 34 –36 weeks weeks Term (n 5 161) (n 5 52) (n 5 26) Institutional factors Deliveries/y #200 $201 Provider specialty Obstetrician FP or midwife Demographics and obstetric/ medical history Race White Nonwhite Insurance Private insurance Others Employed outside home Yes No Cigarette smoker Yes No Previous preterm delivery* Yes No Parity Zero 1 or more
P .08
41 (80) 120 (64)
7 (14) 45 (24)
3 (6) 23 (12)
40 (54) 121 (73)
19 (26) 33 (20)
15 (20) 11 (7)
156 (69) 5 (36)
47 (21) 5 (36)
22 (9) 4 (29)
100 (69) 61 (65)
33 (23) 19 (20)
12 (8) 14 (15)
104 (67) 57 (69)
29 (19) 23 (28)
23 (15) 3 (4)
40 (71) 121 (66)
16 (29) 36 (20)
0 (0) 26 (14)
26 (72) 76 (78)
10 (28) 14 (14)
0 (0) 8 (8)
16 (47) 145 (71)
7 (21) 45 (22)
11 (32) 15 (7)
.03
.02
.27
.02
.007
.06
,.001
FP 5 family practitioner. * Percentages based on all women with previous pregnancies.
were entered in a step-wise fashion in each model. Only variables that improved the fit of the model were retained in the final model. For predicting preterm delivery, only four variables were associated independently with prematurity (Table 3). Rupture of membranes was the strongest predictor of prematurity, with the other three factors (multiple gestations, nonwhite race, and effacement of 80% or more) being less powerful predictors. This model explained nearly 35% of all the variations in preterm delivery (R2 5 34.6%). Similarly, five factors were associated independently with delivery earlier than 34 weeks and within 7 days of the onset of preterm labor. Again, ruptured membranes was a powerful predictor of delivery within a week. Effacement of 80% of more was also a strong predictor, with multiple gestation, nonsmoker status, and dilatation greater than 1 cm less powerful predictors. This model explained almost 43% of all variations (R2 5 42.6%). Using the factors that were associated independently
Obstetrics & Gynecology
Table 2. Effect of Current Pregnancy and Presentation on Risk of Preterm Delivery Gestational age at delivery
Current pregnancy factors Multiple gestation Yes No Urinary infection during pregnancy Yes No Vaginitis or STD during pregnancy Yes No Trimester in which care began 1st 2nd Presentation factors Cervical dilatation #1 cm .1 cm Cervical effacement ,80% $80 Contraction frequency #5 min apart .5 min apart Rupture of membranes Ruptured Not ruptured
Term
34 –36 weeks
,34 weeks
6 (33) 155 (70)
7 (39) 45 (20)
5 (28) 21 (10)
Variable
P .004
.94 42 (66) 119 (67)
16 (25) 39 (22)
6 (9) 20 (11)
53 (60) 108 (72)
23 (26) 29 (19)
13 (15) 13 (9)
138 (78) 22 (71)
25 (14) 7 (23)
13 (7) 2 (6)
136 (72) 25 (50)
37 (20) 15 (30)
16 (8) 10 (20)
150 (72) 11 (37)
41 (20) 11 (37)
18 (9) 8 (27)
109 (66) 52 (69)
36 (22) 16 (21)
19 (12) 7 (9)
3 (16) 158 (72)
8 (42) 43 (20)
8 (42) 17 (8)
.89
.008
,.001
.86
,.001
with delivery within 1 week, a decision rule was constructed to assess the usefulness of these variables in predicting which patients would deliver within 7 days. The decision rule results are shown in Table 4. When considering all patients, reliance on rupture of membranes alone produces excellent specificity, but only captured a minority of the patients who would deliver. Table 3. Factors Associated With Preterm Labor Based on Logistic Regression Modeling Adjusted OR
95% CI
P
12.02 4.63 4.54 3.08
3.05, 47.37 1.47, 14.61 1.23, 16.67 1.20, 7.87
,.001 .009 .02 .02
66.79 10.57 8.44 4.17 3.31
12.43, 358.62 3.27, 34.09 1.99, 35.37 1.00, 16.67 1.15, 9.56
,.001 ,.001 .004 .05 .03
Preterm birth Ruptured membranes Multiple gestation Nonwhite race Effaced 80% or more Delivery within 7 d Ruptured membranes Effaced 80% or more Multiple gestation Nonsmoker Dilatation over 1 cm
OR 5 odds ratio; CI 5 confidence interval.
VOL. 92, NO. 1, JULY 1998
Presenting ,36 wk Ruptured membranes (A) A or effacement of $80% (B) A or B or multiple gestation (C) A or B or C or dilatation .1 cm Presenting ,36 wk Ruptured membranes (A) A or effacement of $80% (B) A or B or multiple gestation (C) A or B or C or dilatation .1 cm
Sensitivity
Specificity
47% 71% 74% 86%
99% 98% 86% 74%
42% 54% 77% 85%
98% 93% 86% 80%
.12
STD 5 sexually transmitted disease.
Variable
Table 4. Results of a Decision Rule for Predicting Delivery Within 7 Days of Contractions
The addition of effacement of 80% or more improved the sensitivity to 71% while retaining a specificity of 98%. The addition of multiple gestation or dilatation of more than 1 cm improves the sensitivity only marginally, while causing significant drops in specificity. On the basis of these models, it appears that the presence of either of two factors (ruptured membranes or advanced cervical effacement) is sufficient to identify almost three quarters of all women who will deliver within 1 week. For women presenting before 34 weeks in whom steroid administration might be considered, the prediction rule requires more variables to reach adequate sensitivity. Using ruptured membranes and effacement alone only identifies slightly more than half of all women who will deliver within a week. When multiple gestation and dilatation are added, 85% of all women who will deliver within a week can be identified with a 20% false-positive rate.
Discussion This study suggests that only two clinical cues can be used to assist physicians in identifying women with preterm contractions who will deliver within 1 week of their presentation. The clinical factors that were identified, ruptured membranes and advanced cervical effacement, can be readily determined at the time of presentation, are inexpensive and easy to assess, and can identify nearly three quarters of all women who will deliver within a week. Furthermore, if neither factor is present, women have a very low probability (2%) of delivering within the next week. For hospitals with poor capabilities of stabilizing potentially ill premature neonates, the presence of either of these factors might be cause for consultation or transfer of patients to tertiary centers. Conversely, when neither of these factors is present, there appears to be little need for transfer because of the low probability of delivery. When women present before 34 weeks’ gestation, including multiple gestations and cervical dilatation of
Hueston
Predicting Preterm Birth
45
2 cm or more improves the predictive value of the decision rule. In these cases, the presence of any of the four factors captures 85% of the women who will deliver within 1 week. Because evidence suggests that the duration of action of corticosteroids for acceleration of fetal lung maturity is at least 7 days,3 administration of steroids to women with either of these risk factors assures that the majority of women are treated, and subjects 20% of women to unnecessary therapy. Previous studies have suggested that steroids and other beneficial treatments in preterm labor often are under used.5,9 This is especially true in women who present to institutions that do not have level III nurseries or high-risk maternal units.9 In these hospitals, physicians often defer the decision to administer steroids or other treatment to referral centers, which may delay administration of appropriate agents and place the patient at excessive risk. In addition, clinicians do not have reliable tools to help decide which patients should be referred to tertiary centers, which can further delay appropriate care.10 The algorithm described in this study may help guide clinicians with these decisions. Although these findings suggest that only a few factors can be used to guide decisions about referral and treatment of women with preterm contractions, important limitations of this study should be considered. First, the study population was predominantly white and from only one area of the country. Previous studies have shown that black women have higher rates of preterm delivery even when correcting for other socioeconomic variables.11 Higher rates of premature delivery were observed in the nonwhite participants of this study, but insufficient numbers of women were included in the sample to allow for appropriate evaluation of the effect of race on the decision to treat women. Larger studies with a more diverse patient population would be useful in confirming and expanding the results presented in this study. Second, this study relied on pooled retrospective data to develop the decision model. Validation of the model prospectively will be necessary to see how the model performs under routine clinical conditions.
2. Higby K, Xenakis EMJ, Pauerstein CJ. Do tocolytic agents stop preterm labor? A critical and comprehensive review of efficacy and safety. Am J Obstet Gynecol 1993;168:1247–59. 3. Crowley RA. Antenatal corticosteroid therapy: A meta-analysis of the randomized trials, 1972–1994. Am J Obstet Gynecol 1995;173: 322–35. 4. Simpson KN, Lynch SR. Cost savings from the use of antenatal steroids to prevent respiratory distress syndrome and related conditions in premature infants. Am J Obstet Gynecol 1995;173: 316 –21. 5. Leviton LC, Baker S, Hassol A, Goldenberg RL. An exploration of opinion and practice patterns affecting low use of antenatal corticosteroids. Am J Obstet Gynecol 1995;173:312– 6. 6. Korenbrot CC, Aalto LH, Laros RK. The cost effectiveness of stopping preterm labor with beta-adrenergic treatment. N Engl J Med 1984;310:691– 6. 7. Weiner CP, Renk K, Klugman M. The therapeutic efficacy and cost-effectiveness of aggressive tocolysis for premature labor associated with premature rupture of membranes. Am J Obstet Gynecol 1999;159:216 –22. 8. Hueston WJ. Preterm contractions in community settings: 1. The treatment of preterm contractions at community hospitals. Obstet Gynecol 1998 (in press). 9. Hueston WJ. Variation between family physicians and obstetricians in the evaluation and treatment of preterm labor. J Fam Pract 1997;45:336 – 40. 10. Institute of Medicine Committee to Study the Prevention of Low Birth Weight. Preventing low birth weight. Washington, DC: National Academy Press, 1985:21–34. 11. Kempe A, Wise PH, Barkan SE, Sappenfield WM, Sachs B, Gortmaker SL, et al. Clinical determinants of the racial disparity in very low birth weight. N Engl J Med 1992;327:969 –3.
Address reprint requests to:
William J. Hueston, MD Department of Family Medicine Medical University of South Carolina 295 Calhoun Street Charleston, SC 29401 E-mail: [email protected]
Received November 6, 1997. Received in revised form January 15, 1998. Accepted March 5, 1998.
References 1. The Canadian Preterm Labor Investigators Group. Treatment of preterm labor with the beta-adrenergic agent ritodrine. N Engl J Med 1992;327:308 –12.
46 Hueston
Predicting Preterm Birth
Copyright © 1998 by The American College of Obstetricians and Gynecologists. Published by Elsevier Science Inc.
Obstetrics & Gynecology
The management of preterm labor is complicated by conflicting reports regarding the efficacy of some treatment strategies and confusion about when effective strategies should be used. Disappointing results regarding the effectiveness of tocolytic agents in wellFrom the Department of Family Medicine, University of WisconsinMadison Medical School, Madison, Wisconsin. This work was supported by a Robert Wood Johnson Generalist Faculty Scholar Award to Dr. Hueston.
VOL. 92, NO. 1, JULY 1998
controlled studies raise questions about whether any effective treatment strategies exist.1,2 Antenatal steroids have been shown to reduce the morbidity associated with prematurity3 and appear to be cost-beneficial4 but are under used.5 Finally, some evidence suggests that any treatment of preterm contractions after 33 weeks’ gestation6 or after premature rupture of membranes7 is not cost-effective and that women with these conditions should be allowed to deliver. One of the dilemmas facing clinicians who treat women with preterm contractions is recognizing which patients will and which patients will not deliver prematurely. Steroids, antibiotics, and the transfer of patients to level III nursery facilities all hinge on the ability to recognize which patients are likely to deliver soon. The purpose of this study was to attempt to define a small number of clinical factors that could identify which women with preterm contractions were likely to deliver prematurely and shortly after presentation.
Materials and Methods The study population was drawn from a network of 11 community hospitals located within a 60-mile radius of Eau Claire, Wisconsin. No facilities with level III neonatal nurseries are located in this area. Because referral of patients with continued contractions can be a source of selection bias and because one of our potential outcomes was the referral of a patient to a tertiary care facility, we limited the hospital sample to only level I and level II facilities. Women enrolled in the study consented to the review of their prenatal and delivery records and those of their newborn following delivery. Two hundred and thirty-nine (90%) of the 266 women invited to participate consented. Details about the hospitals participating in the study and data abstraction are reported elsewhere.8 Data were analyzed using bivariate comparisons of
0029-7844/98/$19.00 PII S0029-7844(98)00108-2
43
risk factors, institutional and clinician characteristics, and obstetric variables using x2 for categoric variables and the t test for continuous variables. Using those variables that were associated (at P , .10) with preterm, very preterm delivery, or delivery within 7 days of presentation, logistic regression models were constructed to adjust for potential confounding. Variables included in the regression models included all clinical variables with P , .05. Regression analyses were performed using True Epistat, version 2.0 (Epistat Services, Richardson, TX). On the basis of results of the regression analysis, prediction models then were constructed using all clinical variables that were associated independently with each outcome variable; sensitivity, specificity, and prediction accuracy were calculated.
Results Seventy-eight (33%) of the 239 women in the study sample delivered prematurely; 52 (67%) of them delivered after 34 weeks’ gestation but before 37 weeks, and the remaining 26 (33%) delivered before 34 weeks’ gestation. To assess which factors could predict preterm birth, we performed bivariate analysis with a range of institutional, patient obstetric history, current pregnancy, and presenting clinical findings with the chance of preterm delivery (Tables 1 and 2). When we examined institutional and historic variables, we found several that were associated with preterm delivery (Table 1). Obstetricians tended to provide care for a greater proportion of women who delivered prematurely, especially those who delivered before 34 weeks’ gestation. This relationship remained significant even after stratifying for multiple-gestation pregnancies, which were more likely to be managed by obstetricians. Other factors associated with preterm delivery included nonwhite race, being employed outside the home, and nulliparity. Interestingly, being a nonsmoker increased the chance of delivery. The factors relating to the current pregnancy that were associated with preterm delivery included multiple gestation and rupture of membranes (Table 2). Additionally, cervical dilation greater than 1 cm and effacement of 80% or more were associated with preterm delivery. The frequency of contractions was not associated with prematurity. To adjust for confounding, we examined logistic regression models using factors that could be used by clinicians to help make a decision about rendering treatment with steroids, antibiotics, or by transferring a patient to a tertiary facility. Models evaluating two outcomes were constructed: one model used prematurity as an outcome and the second used delivery with 7 days as the outcome measure. Independent variables
44 Hueston
Predicting Preterm Birth
Table 1. Effect of Institutional, Clinician, and Past Obstetric History on Risk of Preterm Delivery Gestational age at delivery ,34 34 –36 weeks weeks Term (n 5 161) (n 5 52) (n 5 26) Institutional factors Deliveries/y #200 $201 Provider specialty Obstetrician FP or midwife Demographics and obstetric/ medical history Race White Nonwhite Insurance Private insurance Others Employed outside home Yes No Cigarette smoker Yes No Previous preterm delivery* Yes No Parity Zero 1 or more
P .08
41 (80) 120 (64)
7 (14) 45 (24)
3 (6) 23 (12)
40 (54) 121 (73)
19 (26) 33 (20)
15 (20) 11 (7)
156 (69) 5 (36)
47 (21) 5 (36)
22 (9) 4 (29)
100 (69) 61 (65)
33 (23) 19 (20)
12 (8) 14 (15)
104 (67) 57 (69)
29 (19) 23 (28)
23 (15) 3 (4)
40 (71) 121 (66)
16 (29) 36 (20)
0 (0) 26 (14)
26 (72) 76 (78)
10 (28) 14 (14)
0 (0) 8 (8)
16 (47) 145 (71)
7 (21) 45 (22)
11 (32) 15 (7)
.03
.02
.27
.02
.007
.06
,.001
FP 5 family practitioner. * Percentages based on all women with previous pregnancies.
were entered in a step-wise fashion in each model. Only variables that improved the fit of the model were retained in the final model. For predicting preterm delivery, only four variables were associated independently with prematurity (Table 3). Rupture of membranes was the strongest predictor of prematurity, with the other three factors (multiple gestations, nonwhite race, and effacement of 80% or more) being less powerful predictors. This model explained nearly 35% of all the variations in preterm delivery (R2 5 34.6%). Similarly, five factors were associated independently with delivery earlier than 34 weeks and within 7 days of the onset of preterm labor. Again, ruptured membranes was a powerful predictor of delivery within a week. Effacement of 80% of more was also a strong predictor, with multiple gestation, nonsmoker status, and dilatation greater than 1 cm less powerful predictors. This model explained almost 43% of all variations (R2 5 42.6%). Using the factors that were associated independently
Obstetrics & Gynecology
Table 2. Effect of Current Pregnancy and Presentation on Risk of Preterm Delivery Gestational age at delivery
Current pregnancy factors Multiple gestation Yes No Urinary infection during pregnancy Yes No Vaginitis or STD during pregnancy Yes No Trimester in which care began 1st 2nd Presentation factors Cervical dilatation #1 cm .1 cm Cervical effacement ,80% $80 Contraction frequency #5 min apart .5 min apart Rupture of membranes Ruptured Not ruptured
Term
34 –36 weeks
,34 weeks
6 (33) 155 (70)
7 (39) 45 (20)
5 (28) 21 (10)
Variable
P .004
.94 42 (66) 119 (67)
16 (25) 39 (22)
6 (9) 20 (11)
53 (60) 108 (72)
23 (26) 29 (19)
13 (15) 13 (9)
138 (78) 22 (71)
25 (14) 7 (23)
13 (7) 2 (6)
136 (72) 25 (50)
37 (20) 15 (30)
16 (8) 10 (20)
150 (72) 11 (37)
41 (20) 11 (37)
18 (9) 8 (27)
109 (66) 52 (69)
36 (22) 16 (21)
19 (12) 7 (9)
3 (16) 158 (72)
8 (42) 43 (20)
8 (42) 17 (8)
.89
.008
,.001
.86
,.001
with delivery within 1 week, a decision rule was constructed to assess the usefulness of these variables in predicting which patients would deliver within 7 days. The decision rule results are shown in Table 4. When considering all patients, reliance on rupture of membranes alone produces excellent specificity, but only captured a minority of the patients who would deliver. Table 3. Factors Associated With Preterm Labor Based on Logistic Regression Modeling Adjusted OR
95% CI
P
12.02 4.63 4.54 3.08
3.05, 47.37 1.47, 14.61 1.23, 16.67 1.20, 7.87
,.001 .009 .02 .02
66.79 10.57 8.44 4.17 3.31
12.43, 358.62 3.27, 34.09 1.99, 35.37 1.00, 16.67 1.15, 9.56
,.001 ,.001 .004 .05 .03
Preterm birth Ruptured membranes Multiple gestation Nonwhite race Effaced 80% or more Delivery within 7 d Ruptured membranes Effaced 80% or more Multiple gestation Nonsmoker Dilatation over 1 cm
OR 5 odds ratio; CI 5 confidence interval.
VOL. 92, NO. 1, JULY 1998
Presenting ,36 wk Ruptured membranes (A) A or effacement of $80% (B) A or B or multiple gestation (C) A or B or C or dilatation .1 cm Presenting ,36 wk Ruptured membranes (A) A or effacement of $80% (B) A or B or multiple gestation (C) A or B or C or dilatation .1 cm
Sensitivity
Specificity
47% 71% 74% 86%
99% 98% 86% 74%
42% 54% 77% 85%
98% 93% 86% 80%
.12
STD 5 sexually transmitted disease.
Variable
Table 4. Results of a Decision Rule for Predicting Delivery Within 7 Days of Contractions
The addition of effacement of 80% or more improved the sensitivity to 71% while retaining a specificity of 98%. The addition of multiple gestation or dilatation of more than 1 cm improves the sensitivity only marginally, while causing significant drops in specificity. On the basis of these models, it appears that the presence of either of two factors (ruptured membranes or advanced cervical effacement) is sufficient to identify almost three quarters of all women who will deliver within 1 week. For women presenting before 34 weeks in whom steroid administration might be considered, the prediction rule requires more variables to reach adequate sensitivity. Using ruptured membranes and effacement alone only identifies slightly more than half of all women who will deliver within a week. When multiple gestation and dilatation are added, 85% of all women who will deliver within a week can be identified with a 20% false-positive rate.
Discussion This study suggests that only two clinical cues can be used to assist physicians in identifying women with preterm contractions who will deliver within 1 week of their presentation. The clinical factors that were identified, ruptured membranes and advanced cervical effacement, can be readily determined at the time of presentation, are inexpensive and easy to assess, and can identify nearly three quarters of all women who will deliver within a week. Furthermore, if neither factor is present, women have a very low probability (2%) of delivering within the next week. For hospitals with poor capabilities of stabilizing potentially ill premature neonates, the presence of either of these factors might be cause for consultation or transfer of patients to tertiary centers. Conversely, when neither of these factors is present, there appears to be little need for transfer because of the low probability of delivery. When women present before 34 weeks’ gestation, including multiple gestations and cervical dilatation of
Hueston
Predicting Preterm Birth
45
2 cm or more improves the predictive value of the decision rule. In these cases, the presence of any of the four factors captures 85% of the women who will deliver within 1 week. Because evidence suggests that the duration of action of corticosteroids for acceleration of fetal lung maturity is at least 7 days,3 administration of steroids to women with either of these risk factors assures that the majority of women are treated, and subjects 20% of women to unnecessary therapy. Previous studies have suggested that steroids and other beneficial treatments in preterm labor often are under used.5,9 This is especially true in women who present to institutions that do not have level III nurseries or high-risk maternal units.9 In these hospitals, physicians often defer the decision to administer steroids or other treatment to referral centers, which may delay administration of appropriate agents and place the patient at excessive risk. In addition, clinicians do not have reliable tools to help decide which patients should be referred to tertiary centers, which can further delay appropriate care.10 The algorithm described in this study may help guide clinicians with these decisions. Although these findings suggest that only a few factors can be used to guide decisions about referral and treatment of women with preterm contractions, important limitations of this study should be considered. First, the study population was predominantly white and from only one area of the country. Previous studies have shown that black women have higher rates of preterm delivery even when correcting for other socioeconomic variables.11 Higher rates of premature delivery were observed in the nonwhite participants of this study, but insufficient numbers of women were included in the sample to allow for appropriate evaluation of the effect of race on the decision to treat women. Larger studies with a more diverse patient population would be useful in confirming and expanding the results presented in this study. Second, this study relied on pooled retrospective data to develop the decision model. Validation of the model prospectively will be necessary to see how the model performs under routine clinical conditions.
2. Higby K, Xenakis EMJ, Pauerstein CJ. Do tocolytic agents stop preterm labor? A critical and comprehensive review of efficacy and safety. Am J Obstet Gynecol 1993;168:1247–59. 3. Crowley RA. Antenatal corticosteroid therapy: A meta-analysis of the randomized trials, 1972–1994. Am J Obstet Gynecol 1995;173: 322–35. 4. Simpson KN, Lynch SR. Cost savings from the use of antenatal steroids to prevent respiratory distress syndrome and related conditions in premature infants. Am J Obstet Gynecol 1995;173: 316 –21. 5. Leviton LC, Baker S, Hassol A, Goldenberg RL. An exploration of opinion and practice patterns affecting low use of antenatal corticosteroids. Am J Obstet Gynecol 1995;173:312– 6. 6. Korenbrot CC, Aalto LH, Laros RK. The cost effectiveness of stopping preterm labor with beta-adrenergic treatment. N Engl J Med 1984;310:691– 6. 7. Weiner CP, Renk K, Klugman M. The therapeutic efficacy and cost-effectiveness of aggressive tocolysis for premature labor associated with premature rupture of membranes. Am J Obstet Gynecol 1999;159:216 –22. 8. Hueston WJ. Preterm contractions in community settings: 1. The treatment of preterm contractions at community hospitals. Obstet Gynecol 1998 (in press). 9. Hueston WJ. Variation between family physicians and obstetricians in the evaluation and treatment of preterm labor. J Fam Pract 1997;45:336 – 40. 10. Institute of Medicine Committee to Study the Prevention of Low Birth Weight. Preventing low birth weight. Washington, DC: National Academy Press, 1985:21–34. 11. Kempe A, Wise PH, Barkan SE, Sappenfield WM, Sachs B, Gortmaker SL, et al. Clinical determinants of the racial disparity in very low birth weight. N Engl J Med 1992;327:969 –3.
Address reprint requests to:
William J. Hueston, MD Department of Family Medicine Medical University of South Carolina 295 Calhoun Street Charleston, SC 29401 E-mail: [email protected]
Received November 6, 1997. Received in revised form January 15, 1998. Accepted March 5, 1998.
References 1. The Canadian Preterm Labor Investigators Group. Treatment of preterm labor with the beta-adrenergic agent ritodrine. N Engl J Med 1992;327:308 –12.
46 Hueston
Predicting Preterm Birth
Copyright © 1998 by The American College of Obstetricians and Gynecologists. Published by Elsevier Science Inc.
Obstetrics & Gynecology