KP EOS Calculator: 2.0 update

While effective, why does the EOS calculator need updating?

• The original calculator models were derived from data from 1993-2007.
• United States Obstetric practice has since adopted universal antepartum screening for Group B Streptococcus (GBS).
• ACOG guidance has changed regarding which antibiotics constitute appropriate intrapartum antibiotic prophylaxis (IAP).

What changes were made to the classification and timing of antibiotics in the new EOS predictive model?

• Classification of maternal antibiotics – The original prediction model classified erythromycin, clindamycin, and vancomycin as GBS IAP per CDC 2002 recommendations.¹ With changes in GBS antibiotic susceptibility and updated information on transplacental antibiotic transfer, revisions were made to GBS IAP recommendations.^2-10 For the updated model, only the administration of ampicillin, penicillin, or cefazolin was considered adequate for GBS IAP.^2-10 The American College of Obstetricians and Gynecologists (ACOG) no longer recommends erythromycin for IAP due to high rates of GBS resistance to erythromycin. While ACOG and AAP recommend that clindamycin and vancomycin can be used in PCN allergic women, for the purposes EOS risk, we did not consider them adequate IAP because of current limited data on transplacental transfer and effectiveness. We classified GBS IAP in combination with an aminoglycoside, aztreonam, azithromycin, or metronidazole as broad-spectrum antibiotics. Also included as broad-spectrum antibiotics were second-, third-, or fourth-generation cephalosporins, extended-spectrum penicillins, fluoroquinolones, carbapenems, tetracyclines, and sulfamethoxazole/trimethoprim.
• Timing of antibiotics - Timing of antibiotics was classified in the original model as antibiotics given >4 hours prior to delivery, antibiotics given <4 hours prior to delivery, and no antibiotics given.¹¹ For clinical implementation, antibiotics given <2 hours prior to delivery were classified as "no antibiotics". GBS IAP administered >2 hours prior to delivery was considered adequate IAP as the effectiveness of IAP in preventing GBS-specific EOS if given >2 hours is >89%.^12-15 For the updated model, we classified antibiotic timing as >4 hours prior to delivery; 2-3.9 hours prior to delivery; and <2 hours prior to delivery or no antibiotics.

Why does the new calculator not include infants born at 34 weeks’ gestation?

• While the original model included infants > 34 weeks’ gestation, the new model was developed using a cohort of infants > 35 weeks gestation. The high prevalence of respiratory distress among 34-week infants diminishes the predictive value of their clinical status in EOS prediction. In most cases, decisions regarding EOS risk are not the primary reason these late preterm infants are admitted to intensive care units. Often these infants are routinely admitted to intensive care units based upon their gestational age alone. For these reasons, 34-week infants were not included in the cohort used to reestimate the EOS prediction model.

How did the methodology change in the new EOS predictive model?

• Cohort design: The original model used a nested case-control design because information was not available electronically on the entire cohort in this era. The new model uses data from 2010-2020 for which electronic data is available on all variables in the risk prediction model. Therefore, a cohort design could be used, allowing us to use data on all 412,543 infants. In the original model, the risk prediction model was developed on 350 infants with culture-confirmed sepsis in first 72 hours of age and 1063 randomly selected controls. Statistical methods were than used to adjust the model to the baseline population incidence of EOS.
• Likelihood ratios for clinical status: In the original calculator, the upper limit of the 95% CI of the LR for the clinical status was used to calculate the EOS posterior probabilities.^16,17 This was done intentionally to estimate the highest risk, as an added measure of safety. Now that numerous implementation studies demonstrate the safety of this approach,^18-22 in the updated model, we used the point estimates of the LRs to eliminate bias in the risk estimates. The LRs used were clinical illness, 14.5 (original LR 21.2), equivocal, 3.65 (original 5.0), well appearing, 0.36 (original 0.41).²³
• Intercept adjustment: In applying the model to populations outside of KPNC, it may be necessary to adjust the intercept for the prevalence of EOS in that population if the difference is due to factors other than differences in the prevalence of the predictors in the model. one can use standard formulas^24,25 To adjust the intercept (β0) to make the predicted EOS rate in the predicted population equal to the observed rate in the new population. Letting τ denote the outcome rate in the new population and ȳ denote the KPNC outcome rate of 0.0002763 (0.2763 cases per 1,000 births) then the corrected intercept (β1) is given by:

• The table below shows the adjusted model intercept for selected prevalence rates.

Prevalence per 1,000 Births	β1	Prevalence Rate	1 - Ʈ Ʈ	ȳ 1 - ȳ
5.00	60.2	0.005	199	0.000276
4.00	60.0	0.004	249	0.000276
2.00	59.3	0.002	499	0.000276
1.00	58.6	0.001	999	0.000276
0.90	58.5	0.0009	1,110	0.000276
0.80	58.4	0.0008	1,249	0.000276
0.70	58.2	0.0007	1,428	0.000276
0.60	58.1	0.0006	1,666	0.000276
0.50	57.9	0.0005	1,999	0.000276
0.40	57.7	0.0004	2,499	0.000276
0.30	57.4	0.0003	3,332	0.000276
0.20	57.0	0.0002	4,999	0.000276
0.10	56.3	0.0001	9,999	0.000276
0.05	55.6	0.00005	19,999	0.000276

What are the coefficients in the new model?

• Updated EOS Model Coefficients

	Beta Coefficient	OR	p-value
Intercept	57.29929499		0.22120
Highest maternal temperature a	0.85194656	2.344	0.00000
Gestational age b	-7.72247124	0.000	0.00127
Gestational age squared	0.09842383	1.103	0.00141
Rupture of membranes c	0.86770862	2.381	0.00016
Abx1 d	-2.13142945	0.119	0.00000
Abx2 e	-2.33985917	0.096	0.00012
GBS f Positive	1.02265353	2.781	0.00102
GBS Status Unknown	1.13710111	3.118	0.00267

1. Temperature is captured to nearest 0.1°F
2. Gestational age is captured in exact days
3. Time in hours + 0.05 to the 1/5^th power
4. Abx1 = 1 if GBS IAP given > 2 hours prior to delivery or broad-spectrum antibiotics given > 2 hours but <4 hours prior to delivery
5. Abx2 = 1 if broad-spectrum antibiotics given > 4 hours prior to delivery
6. GBS = Group B Streptococcus

What likelihood ratios for clinical presentation are used in the new model?

• Likelihood ratio for clinical illness: 14.5
• Likelihood ratio for equivocal: 3.65
• Likelihood ratio for well appearing: 0.36

Is the new predictive model better than the old model?

• We compared the sensitivity of the models after applying the newborn’s clinical status during the first 24 hours. There was no statistically significant difference in sensitivity at 24 hours, 0.76 (95% CI 0.63-0.85) versus 0.80 (95% CI 0.68-0.89), p=0.15, for the revised calculator’s model. Comparing the original to the updated calculator, there was a 0.2% increase in recommended antibiotic use by 24 hours (3.5% vs. 3.7%).
• The updated calculator recommended empiric antibiotics (EOS risk >3 per 1,000 live births) or a blood culture (EOS risk 1-2.9) in the first 24 hours in 88% (58/66) of EOS cases.
• For each additional case identified in the first 24 hours with the updated model compared to original model, an additional 158 infants would be treated with empiric antibiotics in the cohort studied.

Which model should I use?

• The updated KP EOS calculator exhibited similar performance in identifying EOS cases in the first 24 hours after birth as the original version. Our results suggest that either version can be used safely for EOS risk stratification. While there was no statistical difference in the sensitivity of the two calculators, the updated model demonstrated a slight advantage in case identification, balanced by a higher recommended use of empiric antibiotics in the first 24 hours after birth (3.7% versus 3.5%). Among 264,347 newborns, the updated calculator identified one additional EOS case in the first 24 hours at the expense of empiric antibiotics in an additional 158 uninfected infants.
• To better reflect changes in ACOG appropriate IAP and a modern birth cohort with universal GBS screening, the new model would be preferred.
• If obstetrical practice in your location does not include universal GBS screening, we recommend that you continue to use the original model. The original calculator was developed on a population with a 48.6% GBS unknown rate compared to the updated calculator population which has GBS unknown rate of 4.5% . The difference in GBS screening rates reflects ACOG recommendations for universal GBS screening. As a result, the coefficient and corresponding odds ratio for GBS unknown differs substantially between the two models. In the original model the odds ratio for GBS unknown was 1.01 and in the updated model the odds ratio was 3.11. This likely reflects that many untested (GBS unknown) in the original population would probably have been GBS negative if tested. For countries/areas that don’t do universal screening, using the updated model will result in a ~3X increase in the EOS risk in pregnancies where GBS is unknown. We therefore recommend using the original model if universal GBS screening is not performed.

References

1. Schrag S, Gorwitz R, Fultz-Butts K, Schuchat A. Prevention of perinatal group B streptococcal disease. Revised guidelines from CDC. MMWR Recomm Rep. 2002;51(RR-11):1-22. doi:papers3://publication/uuid/661AF9EC-77CA-4A7E-8C1A-6DB761CE6E77.
2. Prevention of Group B Streptococcal Early-Onset Disease in Newborns: ACOG Committee Opinion, Number 797. Obstet Gynecol. Feb 2020;135(2):e51-e72. doi:10.1097/AOG.0000000000003668.
3. Nanduri SA, Petit S, Smelser C, et al. Epidemiology of Invasive Early-Onset and Late-Onset Group B Streptococcal Disease in the United States, 2006 to 2015: Multistate Laboratory and Population-Based Surveillance. JAMA Pediatr. Mar 1 2019;173(3):224-233. doi:10.1001/jamapediatrics.2018.4826.
4. Castor ML, Whitney CG, Como-Sabetti K, et al. Antibiotic resistance patterns in invasive group B streptococcal isolates. Infect Dis Obstet Gynecol. 2008;2008:727505. doi:10.1155/2008/727505.
5. Teatero S, Ferrieri P, Martin I, Demczuk W, McGeer A, Fittipaldi N. Serotype Distribution, Population Structure, and Antimicrobial Resistance of Group B Streptococcus Strains Recovered from Colonized Pregnant Women. J Clin Microbiol. Feb 2017;55(2):412-422. doi:10.1128/JCM.01615-16.
6. Bulska M, Szczesniak P, Pieta-Dolinska A, Oszukowski P, Orszulak-Michalak D. The placental transfer of erythromycin in human pregnancies with group B streptococcal infection. Ginekol Pol. Jan 2015;86(1):33-9. doi:10.17772/gp/1896.
7. Creti R, Imperi M, Berardi A, et al. Neonatal Group B Streptococcus Infections: Prevention Strategies, Clinical and Microbiologic Characteristics in 7 Years of Surveillance. Pediatr Infect Dis J. Mar 2017;36(3):256-262. doi:10.1097/INF.0000000000001414.
8. Nanovskaya T, Patrikeeva S, Zhan Y, Fokina V, Hankins GD, Ahmed MS. Transplacental transfer of vancomycin and telavancin. Am J Obstet Gynecol. Oct 2012;207(4):331 e1-6. doi:10.1016/j.ajog.2012.06.064.
9. Hnat MD, Gainer J, Bawdon RE, Wendel GD, Jr. Transplacental passage of vancomycin in the ex vivo human perfusion model. Infect Dis Obstet Gynecol. Jun 2004;12(2):57-61. doi:10.1080/10647440400009821.
10. Puopolo KM, Lynfield R, Cummings JJ, Committee On F, Newborn, Committee On Infectious D. Management of Infants at Risk for Group B Streptococcal Disease. Pediatrics. Aug 2019;144(2) doi:10.1542/peds.2019-1881.
11. Puopolo KM, Draper D, Wi S, et al. Estimating the probability of neonatal early-onset infection on the basis of maternal risk factors. Pediatrics. Nov 2011;128(5):e1155-63. doi:10.1542/peds.2010-3464.
12. Colombo DF, Lew JL, Pedersen CA, Johnson JR, Fan-Havard P. Optimal timing of ampicillin administration to pregnant women for establishing bactericidal levels in the prophylaxis of Group B Streptococcus. Am J Obstet Gynecol. Feb 2006;194(2):466-70. doi:10.1016/j.ajog.2005.07.098.
13. de Cueto M, Sanchez MJ, Sampedro A, Miranda JA, Herruzo AJ, Rosa-Fraile M. Timing of intrapartum ampicillin and prevention of vertical transmission of group B streptococcus. Obstet Gynecol. 1998;91(1):112-4.
14. Lin FY, Brenner RA, Johnson YR, et al. The effectiveness of risk-based intrapartum chemoprophylaxis for the prevention of early-onset neonatal group B streptococcal disease. Am J Obstet Gynecol. May 2001;184(6):1204-10. doi:10.1067/mob.2001.113875.
15. Illuzzi JL, Bracken MB. Duration of intrapartum prophylaxis for neonatal group B streptococcal disease: a systematic review. Obstet Gynecol. Nov 2006;108(5):1254-65. doi:10.1097/01.AOG.0000241539.86451.11.
16. Kuzniewicz MW, Puopolo KM, Fischer A, et al. A Quantitative, Risk-Based Approach to the Management of Neonatal Early-Onset Sepsis. JAMA Pediatr. Apr 1 2017;171(4):365-371. doi:10.1001/jamapediatrics.2016.4678.
17. Kuzniewicz MW, Walsh EM, Li S, Fischer A, Escobar GJ. Development and Implementation of an Early-Onset Sepsis Calculator to Guide Antibiotic Management in Late Preterm and Term Neonates. Jt Comm J Qual Patient Saf. May 2016;42(5):232-9. doi:10.1016/s1553-7250(16)42030-1.
18. Achten NB, Dorigo-Zetsma JW, van der Linden PD, van Brakel M, Plotz FB. Sepsis calculator implementation reduces empiric antibiotics for suspected early-onset sepsis. European journal of pediatrics. May 2018;177(5):741-746. doi:10.1007/s00431-018-3113-2.
19. Achten NB, Klingenberg C, Benitz WE, et al. Association of Use of the Neonatal Early-Onset Sepsis Calculator With Reduction in Antibiotic Therapy and Safety: A Systematic Review and Meta-analysis. JAMA Pediatr. Sep 3 2019;doi:10.1001/jamapediatrics.2019.2825.
20. Deshmukh M, Mehta S, Patole S. Sepsis calculator for neonatal early onset sepsis - a systematic review and meta-analysis(). The journal of maternal-fetal & neonatal medicine : the official journal of the European Association of Perinatal Medicine, the Federation of Asia and Oceania Perinatal Societies, the International Society of Perinatal Obstet. Aug 11 2019:1-9. doi:10.1080/14767058.2019.1649650.
21. Dhudasia MB, Mukhopadhyay S, Puopolo KM. Implementation of the Sepsis Risk Calculator at an Academic Birth Hospital. Hospital pediatrics. May 2018;8(5):243-250. doi:10.1542/hpeds.2017-0180.
22. Strunk T, Buchiboyina A, Sharp M, Nathan E, Doherty D, Patole S. Implementation of the Neonatal Sepsis Calculator in an Australian Tertiary Perinatal Centre. Neonatology. 2018;113(4):379-382. doi:10.1159/000487298.
23. Escobar GJ, Puopolo KM, Wi S, et al. Stratification of risk of early-onset sepsis in newborns >/= 34 weeks' gestation. Pediatrics. Jan 2014;133(1):30-6. doi:10.1542/peds.2013-1689.
24. Hosmer D, Lemeshow S. Applied logistic regression, 2nd edn. John Wiley and Sons; 2000.
25. King S, Zeng L. Logistic regression in rare events data. Polit Anal 2001;9:137–63.
26. Puopolo KM, Lynfield R, Cummings JJ; Committee on Fetus and Newborn; Committee on Infectious Diseases. Management of Infants at Risk for Group B Streptococcal Disease. Pediatrics. 2019 Aug;144(2):e20191881.