Abstract
Purpose:
Perinatal epidemiology studies using healthcare utilization databases are often restricted to live births, largely due to the lack of established algorithms to identify non-live births. The study objective was to develop and validate claims-based algorithms for the ascertainment of non-live births.
Methods:
Using the Mass General Brigham Research Patient Data Registry 2000–2014, we assembled a cohort of women enrolled in Medicaid with a non-live birth. Based on ≥1 inpatient or ≥2 outpatient diagnosis/procedure codes, we identified and randomly sampled 100 potential stillbirth, spontaneous abortion, and termination cases each. For the secondary definitions, we excluded cases with codes for other pregnancy outcomes within ±5 days of the outcome of interest and relaxed the definitions for spontaneous abortion and termination by allowing cases with one outpatient diagnosis only. Cases were adjudicated based on medical chart review. We estimated the positive predictive value (PPV) for each outcome.
Results:
The PPV was 71.0% (95% CI, 61.1–79.6) for stillbirth; 79.0% (69.7–86.5) for spontaneous abortion, and 93.0% (86.1–97.1) for termination. When excluding cases with adjacent codes for other pregnancy outcomes and further relaxing the definition, the PPV increased to 80.6% (69.5–88.9) for stillbirth, 86.6% (80.5–91.3) for spontaneous abortion and 94.9% (91.1–97.4) for termination. The PPV for the composite outcome using the relaxed definition was 94.4% (92.3–96.1).
Conclusions:
Our findings suggest non-live birth outcomes can be identified in a valid manner in epidemiological studies based on healthcare utilization databases.
Keywords: claims-based algorithm, positive predictive value, spontaneous abortion, stillbirth, termination, validation study
1 |. INTRODUCTION
Healthcare utilization databases, such as the Medicaid Analytic eXtract and the Sentinel Distributed Database, are frequently utilized to evaluate the safety of exposure to medical products in pregnancy.1 The majority of studies using these databases are conducted in live birth pregnancy cohorts assembled via mother and infant linkage,2,3 enabling the assessment of the impact of maternal exposure to medical products on both maternal and neonatal outcomes (e.g., congenital malformations, small for gestational age, and neurodevelopmental disorders).4–10 However, of the 6 million pregnancies occurring each year in the US, approximately 2.2 million end with a non-live birth outcome: stillbirth (pregnancy losses at ≥20 weeks’ gestation), spontaneous abortion (pregnancy losses at <20 weeks’ gestation) or termination.11 Regulatory agencies are therefore particularly interested in being able to include non-live birth outcomes as a potential adverse outcome when evaluating the safety of medical products in pregnancy. Similar to maternal and neonatal outcomes among live births, healthcare utilization databases have significant potential to study the impact of medical products on non-live birth outcomes, if these outcomes can be identified with a high level of accuracy.
Close to 50% of all pregnancies in the US are publicly insured.12 Medicaid-eligible pregnant women represent a racially diverse and socioeconomically disadvantaged population that is traditionally understudied, but prone to non-live births.13,14 The objective of our study was to develop and validate diagnosis and procedure code-based algorithms to identify non-live birth outcomes (i.e., spontaneous abortion, termination, stillbirth) in this important, vulnerable population.
2 |. METHODS
2.1 |. Cohort selection
Within the Mass General Brigham (MGB) Research Patient Data Registry (RPDR),15 a clinical data warehouse of medical records, we identified Medicaid insured women 12–55 years of age who had a medical encounter related to a non-live birth at Massachusetts General Hospital or Brigham and Women’s Hospital in Boston between 2000 and 2014.
2.2 |. Algorithms to identify non-live birth outcomes
Among this source population, we first identified potential cases of stillbirth, spontaneous abortion, and termination, based on the presence of International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) diagnosis and procedure codes and Current Procedural Terminology (CPT) procedure codes recorded in the RPDR claims (Table 1). For stillbirths (≥20 weeks’ gestation), we required ≥1 inpatient code since delivery of the stillborn fetus typically occurs in an inpatient setting. For spontaneous abortion and termination, both inpatient and outpatient codes were considered since patients could receive outpatient care for these two non-live birth outcomes. The primary definition required ≥1 inpatient or ≥2 outpatient codes. We included both principal and secondary inpatient codes. These criteria were relaxed in a secondary definition where 1 outpatient code was also accepted.
TABLE 1.
Non-live birth outcome definitions
| Outcome | ICD-9-CM/CPT codes | Code Definition | N (%) of potential cases with the code | Algorithm Definition |
|---|---|---|---|---|
| Stillbirth | ICD-9 Dx: 656.4x | Intrauterine death affecting management of mother | 82 (82.0) | ≥ 1 inpatient codes |
| ICD-9 Dx: V27.1 | Single stillborn | 83 (83.0) | ||
| ICD-9 Dx: V27.4 | Twins, both stillborn | 3 (3.0) | ||
| ICD-9 Dx: V27.7 | Other multiple birth, all stillborn | 2 (2.0) | ||
| Spontaneous abortion | CPT4: 01965 | Incomplete or missed abortion | 19 (7.6) | Primary definition: ≥ 1 inpatient or ≥2 outpatient codes; Secondary definition: ≥ 1 inpatient or outpatient codes |
| CPT4: 59812 | Treatment of incomplete abortion, any trimester, completed surgically | 22 (8.8) | ||
| CPT4: 59820 | Treatment of missed abortion, completed surgically; first trimester | 57 (22.8) | ||
| CPT4: 59821 | Treatment of missed abortion, completed surgically; second trimester | 15 (6.0) | ||
| CPT4: 59830 | Treatment of septic abortion, completed surgically | 3 (1.2) | ||
| ICD-9 Dx: 632 | Missed abortion | 1 (0.4) | ||
| ICD-9 Dx: 634.xx | Spontaneous abortion | 171 (68.4) | ||
| ICD-9 Dx: 637.xx | Spontaneous abortion with other specified complications | 36 (14.4) | ||
| Termination | CPT4: 01966 | Anesthesia for induced abortion | 7 (2.5) | Primary definition: ≥ 1 inpatient or ≥2 outpatient codes; Secondary definition: ≥ 1 inpatient or outpatient codes |
| CPT4: 59840 | Induced abortion, by dilation and curettage | 31 (11.2) | ||
| CPT4: 59841 | Induced abortion, by dilation and evacuation | 192 (69.3) | ||
| CPT4: 59850 | Induced abortion, by one or more intra-amniotic injections | 1 (0.4) | ||
| CPT4: 59851 | Induced abortion, by one or more intra-amniotic injections, with dilation and curettage and/or evacuation | 1 (0.4) | ||
| CPT4: 59852 | Induced abortion, by one or more intra-amniotic injections, with hysterotomy | 0 (0) | ||
| CPT4: 59855 | Induced abortion, by one or more vaginal suppositories with or without cervical dilation | 2 (0.7) | ||
| CPT4: 59856 | Induced abortion, by one or more vaginal suppositories with or without cervical dilation; dilation and curettage and/or evacuation | 0 (0) | ||
| CPT4: 59857 | Induced abortion, by one or more vaginal suppositories with or without cervical dilation; with hysterotomy | 0 (0) | ||
| CPT4: 59866 | Multifetal reduction | 1 (0.4) | ||
| ICD-9 Dx: 635.xx | Legally induced abortion | 238 (85.9) | ||
| ICD-9 Dx: 636.xx | Illegal abortion | 0 (0) | ||
| ICD-9 Dx: 779.6 | Termination of pregnancy (fetus) | 0 (0) | ||
| ICD-9 Proc: 69.01 | Dilation and curettage for termination of pregnancy | 43 (15.5) | ||
| ICD-9 Proc: 69.51 | Aspiration curettage of uterus for termination of pregnancy | 7 (2.5) | ||
| ICD-9 Proc: 75.0 | Intra-amniotic injection for abortion | 0 (0) | ||
| ICD-9 Proc: 74.91 | Hysterotomy to terminate pregnancy | 0 (0) |
When there were non-live birth codes on more than 1 day for the same women, we identified independent episodes. After examining the distribution of gaps between consecutive visits with a code for the respective non-live birth outcomes in individual patients, we considered all consecutive visits with codes for the same outcome occurring less than 1 month apart as belonging to the same episode, which aligns with the duration of follow-up care for a non-live birth. The date of the first code of each episode was considered the event date for the non-live birth, and the start date of the episode. The date of the last code was considered the end date of the episode.
We then randomly selected 100 potential cases for stillbirth, spontaneous abortion and termination each based on the primary definitions. For the secondary definitions, we randomly selected an additional 150 spontaneous abortion and 177 termination cases based on one outpatient code only, in order to preserve the distribution of cases (i.e., number of cases identified based on ≥1 inpatient or ≥2 outpatient code versus number of cases identified based on 1 outpatient code) as observed in the source population.
Potential cases with diagnosis or procedure codes for other pregnancy outcomes (e.g., other non-live birth outcomes, or live births, Table S1) within the time window from 5 days before the start of the non-live birth episodes until 5 days after the end of the episodes were flagged for exclusion in sensitivity analyses. The analysis was repeated using a time window of 28 days.
2.3 |. Medical chart review
Information contained in the medical chart was considered the reference standard. Patient medical records were identified based on a unique patient identifier, date of birth and the non-live birth event date for each of the potential cases. The case adjudication was performed by physicians trained in obstetrics and gynecology (KJG, RMR, BM-G, EY) who were blinded to the claims-based classification of the birth outcome. In case of uncertainty, a second clinician reviewed the medical record before a final determination was made based on consensus (Table S1 for details).
2.4 |. Statistical analysis
We estimated the positive predictive value (PPV) and its 95% confidence interval (CI) for the primary definitions of stillbirth, spontaneous abortion, and termination, using the medical chart-based classification as the reference standard. In addition to the individual non-live birth outcomes, we also evaluated a composite outcome of any non-live birth, since this may be the outcome of interest in perinatal epidemiologic studies.
We also estimated the PPV for the more sensitive secondary definitions of spontaneous abortion and termination requiring ≥1 inpatient or ≥1 outpatient codes. For both the primary and the secondary definitions, we further evaluated the performance of each algorithm after excluding cases with ICD-9-CM or CPT codes indicative of other pregnancy outcomes within ±5 days of their occurrence.
Since perinatal epidemiologic studies heavily draw upon data from a period when ICD-9 codes were in use (i.e., any time before October 2015 in the US), we developed the algorithms based on ICD-9, and subsequently mapped the ICD-9 definitions to ICD-10 to enable implementation of the algorithms in more recent years of data, using the forward-backward mapping method created by the Centers for Medicare & Medicaid Services and the Centers for Disease Control and Prevention.16,17 We reviewed the identified ICD-10-CM codes, and explored the ICD-10-CM data dictionary to identify any additional codes of interest that had not been selected through the mapping.
2.5 |. Ethics approval
The research was approved by the Institutional Review Board at Mass General Brigham, which waived the requirement for informed consent.
3 |. RESULTS
Using RPDR 2010–2014, we identified a total of 314 stillbirth, 4321 spontaneous abortion (1729 based on the primary definition), and 5527 termination (2005 based on the primary definition) cases in Medicaid insured individuals. The characteristics were very similar between the non-live birth samples and the cases in RPDR (Table S2). In the random sample of 100 algorithm-identified cases based on the primary definition, the PPV was 71.0% (95% CI, 61.1–79.6) for stillbirth, 79.0% (69.7–86.5) for spontaneous abortion, and 93.0% (86.1–97.1) for termination (Table 2). Out of the 297 randomly sampled potential non-live births (3 potential cases had codes for ≥2 non-live birth outcomes), 274 were true non-live births based on the medical chart review, resulting in a PPV of 92.3% (88.6–95.0).
TABLE 2.
Validation of non-live birth outcome algorithms
| Analysis | No. of cases reviewed | % retained | No. of true cases | % retained | PPV (95% CI), % |
|---|---|---|---|---|---|
| Stillbirth | |||||
| Primary definition: ≥ 1 inpatient codes | 100 | 71 | 71.0 (61.1–79.6) | ||
| Exclude cases with codes for other pregnancy outcomes within ±5 days | 72 | 72.0% | 58 | 81.7% | 80.6 (69.5–88.9) |
| Spontaneous abortion | |||||
| Primary definition: ≥ 1 inpatient or ≥2 outpatient codes | 100 | 79 | 79.0 (69.7–86.5) | ||
| Exclude cases with codes for other pregnancy outcomes within ±5 days | 59 | 59.0% | 51 | 64.6% | 86.4 (75.0–94.0) |
| Secondary definition: ≥ 1 inpatient or outpatient codes | 250 | 193 | 77.2 (71.5–82.3) | ||
| Exclude cases with codes for other pregnancy outcomes within ±5 days | 171 | 68.4% | 148 | 76.7% | 86.6 (80.5–91.3) |
| Termination | |||||
| Primary definition: ≥ 1 inpatient or ≥2 outpatient codes | 100 | 93 | 93.0 (86.1–97.1) | ||
| Exclude cases with codes for other pregnancy outcomes within ±5 days | 85 | 85.0% | 83 | 89.2% | 97.7 (91.8–99.7) |
| Secondary definition: ≥ 1 inpatient or outpatient codes | 277 | 234 | 84.5 (79.7–88.5) | ||
| Exclude cases with codes for other pregnancy outcomes within ±5 days | 217 | 78.3% | 206 | 88.0% | 94.9 (91.1–97.4) |
| Composite non-live birth outcomes | |||||
| Primary definition | 297a | 274 | 92.3 (88.6–95.0) | ||
| Secondary definition | 623b | 588 | 94.4 (92.3–96.1) |
Out of 300 potential non-live birth cases, 3 duplicates were removed as they were selected for more than one non-live birth sample for validation. This duplication occurred because each non-livebirth cohort was created independently, and codes indicative of different non-live birth outcomes could have been recorded close in time. For the composite non-live birth outcome, a case was considered to be a true positive as long as the non-live birth outcome was confirmed, regardless of the specific type of non-live birth.
Out of 627 potential non-live birth cases, 4 duplicates were removed as they were selected for more than one non-live birth sample for validation.
The algorithm performance for all three outcomes improved substantially after removing the cases with codes for other pregnancy outcomes within ±5 days: PPV of 80.6% (95% CI, 69.5–88.9) for stillbirth, 86.4% (75.0–94.0) for spontaneous abortion, and 97.7% (91.8–99.7) for termination (Table 2). Out of the potential cases with adjacent codes for other pregnancy outcomes, about half were true stillbirths, spontaneous abortions, and terminations. Approximately 90% of the false positives were true cases of other pregnancy outcomes and had a corresponding code. Notably, false positives of stillbirth were most likely to be true spontaneous abortions or terminations, while false positives of spontaneous abortion were most likely to be true terminations and vice versa (Table S3). Extending the window to ±28 days did not meaningfully affect the PPV (Table S4).
Expanding the algorithm definition to also include potential cases based on 1 outpatient code resulted in a slightly lower PPV: 77.2% (95% CI, 71.5–82.3) for spontaneous abortion, and 84.5% (95% CI, 79.7–88.5) for termination (Table 2). The PPV of the composite non-live birth outcome was not affected much: 94.4% (95% CI, 92.3–96.1) versus 92.3%. However, after further removing the cases with codes for other pregnancy outcomes within ±5 days, the performance of the algorithms based on the secondary definition was comparable to that based on the primary definition (spontaneous abortion: 86.6% vs. 86.4%; termination: 94.9% vs. 97.7%), while the number of identified cases was considerably higher.
4 |. DISCUSSION
Using information in the medical charts as the gold standard, we developed and validated claims-based algorithms for non-live birth outcomes, including stillbirth, spontaneous abortion, and termination. Based on both the primary and second definitions, the estimated PPV for all three outcomes was high – ranging from 80.6% to 97.7% – after restricting to algorithm-derived cases without adjacent codes for other pregnancy outcomes within ±5 days. The estimated PPV for the composite non-live birth outcome was equally high.
Further exploration of the potential non-live birth cases with codes for other pregnancy outcomes recorded within ±5 days of the occurrence of the case revealed coding errors and recording of codes for medical history of the outcome, leading to potential misclassification of non-live births using claims. Of note, based on medical chart review, these cases with codes for different non-live birth outcomes were often confirmed as a false-positive for one type but a true case for another. This also explains the high PPV (>92%) we observed for the composite non-live birth outcome, which avoids the misclassification between individual non-livebirth outcomes. Considering that imperfect specificity of outcome definitions may have more impact on relative risk estimates than imperfect sensitivity,18 restricting to cases without adjacent codes for other pregnancy outcomes will be the preferred approach for identifying individual non-live birth outcomes.
The management of spontaneous abortions and terminations can occur exclusively in an outpatient setting. In this study, a single outpatient code identified over 60% of potential cases. The algorithms requiring a single inpatient or outpatient code with restriction to cases without adjacent codes for other pregnancy outcomes increased the number of potential cases and maintained a high PPV, and therefore is considered the preferred approach.
Only few previous studies have developed and validated ICD-9-CM claims-based algorithms for non-live birth outcomes.19 Two studies reported high accuracy (PPV >90%) to identify stillbirth (or intrauterine death) using ICD-9-CM code-based algorithms in small samples of potential cases (≤14), leading to concerns over the reproducibility of the estimated algorithm performance.20,21 One study – using the Vaccine Safety Datalink – also validated algorithms using ICD-9-CM and CPT codes for spontaneous abortion and termination in a random sample of 105 pregnancies each.20 While the PPV for spontaneous abortion was similar to estimates in the present study (85.7%), the PPV for termination was lower (81.0%), possibly due to use of different codes. Details on the algorithms, which were a modification of existing algorithms with complex implementation steps,22 were not provided, preventing replication of the findings.20
One recent study, using the Sentinel Distributed Database, has developed and validated an ICD-10-CM code-based algorithm for stillbirth, utilizing the same stillbirth codes identified in our mapping, plus additional codes for mixed births (i.e., multiple births with some stillborn and others liveborn).23 They reported a PPV of 82.5% (95% CI: 70.9–91.0) based on 63 potential cases identified with their best performing algorithm. The PPV was improved to 94% when restricting to inpatient codes (as we did in the present study). Another study, conducted among commercially insured pregnancies, validated an ICD-10-CM code-based hierarchical algorithm to identify pregnancy outcomes. In that study, 70.8% (95% CI: 50.2–85.5) of the 24 stillbirth cases (based on inpatient or outpatient codes) and 100% of the 75 spontaneous abortion cases were adjudicated as true cases. The ICD-10-CM codes for stillbirth and spontaneous abortion were identical to those identified in our mapping (Table S1).24
Strengths of our study include relatively large validation samples, focus on a publicly insured vulnerable population, and transparent, easy to implement algorithms. Nevertheless, our study also has some limitations. The Medicaid-insured samples in this study were selected from patients who received services at two major hospitals in Boston known for their high quality of care. It is possible that the coding practices at these hospitals do not generalize to other healthcare settings. Second, we do not have data on the number of pregnancies ending in non-live birth that were not captured by our claims-based algorithms and therefore cannot report on the algorithms’ sensitivity. If sensitivity is low, using our non-live birth algorithms in the context of medical product safety studies in pregnancy would result in an underestimation of absolute risks and risk differences. We do not expect this to be the case, however. Delivery of a stillborn fetus typically occurs in an inpatient setting and for spontaneous abortions and terminations any inpatient or outpatient code was considered evidence of the outcome occurrence. Early spontaneous abortions that occur before clinical recognition of the pregnancy will not be captured regardless of which data source is used, be it healthcare utilization databases, electronic health records, or prospective pregnancy registries. Finally, the algorithms were developed using ICD-9-CM codes, and converted to ICD-10-CM codes using a mapping process. While the code mapping is quite straightforward, ideally the accuracy of these mapped ICD-10-CM should be confirmed in future studies.
5 |. CONCLUSIONS
In summary, algorithms for non-live birth outcomes among publicly insured pregnancies using ICD and CPT codes had high PPVs, providing confidence in studying these outcomes in healthcare utilization databases, such as the nationwide Medicaid data. The best performing algorithms excluded cases with adjacent codes for other pregnancy outcomes. Inclusion of spontaneous abortion and termination cases with one outpatient code only increased the number of identifiable cases without impairing the PPVs of the algorithms.
Supplementary Material
Key Points.
Healthcare utilization databases are frequently utilized to evaluate the safety of exposure to medical products in pregnancy. However, there is a lack of established algorithms to identify non-live births.
The study developed and validated claims-based algorithms with high positive predictive values for the ascertainment of stillbirth, spontaneous abortion, and termination.
The study findings suggest non-live birth outcomes can be identified in a valid manner in epidemiological studies based on healthcare utilization databases.
Plain Language Summary.
Healthcare data, such as Medicaid data, are often used to study whether medication is safe to be used in pregnancy. However, it is unclear whether we could accurately identify which pregnancies end in non-live births. This study developed approaches for identifying stillbirth, miscarriage, and abortion using a healthcare database, and studied how accurate those approaches are by comparing against information in medical records. Out of all the potential pregnancies ending in stillbirth identified using this new approach, 80.6% of them were true stillbirth cases based on the medical records. Out of all the potential miscarriage cases, 86.6% of them were true miscarriage cases. Out of all termination cases, 94.9% of them were true abortion cases. We found that using the information in the healthcare databases we could accurately identify pregnancies ending in non-live births, which will allow us to study whether certain medication use in pregnancy may be linked to non-live births using these approaches in future studies.
Funding information
The U.S. Food and Drug Administration (FDA) through the Department of Health and Human Services, Grant/Award Number: HHSF223201400043I
CONFLICT OF INTEREST
This article reflects the views of the authors and should not be construed to represent FDA’s views or policies. Yanmin Zhu is an investigator on a research grant to her institution from Takeda for an unrelated study. Krista F. Huybrechts is an investigator on a research grant to her institution from Takeda and UCB for unrelated studies. Sonia Hernandez-Diaz is an investigator on grants to her institution from Takeda for unrelated studies; personal fees from UCB and Roche outside the submitted work; and having served as an epidemiologist with the North America AED pregnancy registry, which is funded by multiple companies. Brian T. Bateman is an investigator on research grants to his institution from Pacira and Takeda for unrelated studies. He is a consultant to Aetion Inc. and the Alosa Foundation. Kathryn J. Gray has served as a consultant to Illumina Inc., Aetion, Roche, and BillionToOne outside the scope of the submitted work. Loreen Straub, Rebecca M. Reimers, Beryl Manning-Geist, Elizabeth Yoselevsky, Lockwood G. Taylor, Rita Ouellet-Hellstrom, Yong Ma, Yandong Qiang, and Wei Hua have nothing to declare.
Footnotes
ETHICS STATEMENT
The study was approved by the Mass General Brigham Institutional Review Board.
SUPPORTING INFORMATION
Additional supporting information can be found online in the Supporting Information section at the end of this article.
This study was presented at the 35th International Conference on Pharmacoepidemiology & Therapeutic Risk Management, August 2019, Philadelphia, USA.
DATA AVAILABILITY STATEMENT
The data are not available for replication directly from the authors because of IRB restrictions. More information on access to the data can be found at https://rpdrssl.partners.org/.
REFERENCES
- 1.Huybrechts KF, Bateman BT, Hernandez-Diaz S. Use of real-world evidence from healthcare utilization data to evaluate drug safety during pregnancy. Pharmacoepidemiol Drug Saf. 2019;28(7):906–922. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Palmsten K, Huybrechts KF, Mogun H, et al. Harnessing the medicaid analytic eXtract (MAX) to evaluate medications in pregnancy: design considerations. PloS One. 2013;8(6):e67405. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Baldwin E, Johnson K, Berthoud H, Dublin S. Linking mothers and infants within electronic health records: a comparison of deterministic and probabilistic algorithms. Pharmacoepidemiol Drug Saf. 2015;24(1): 45–51. [DOI] [PubMed] [Google Scholar]
- 4.Molgaard-Nielsen D, Pasternak B, Hviid A. Use of oral fluconazole during pregnancy and the risk of birth defects. N Engl J Med. 2013; 369(9):830–839. [DOI] [PubMed] [Google Scholar]
- 5.Kieler H, Artama M, Engeland A, et al. Selective serotonin reuptake inhibitors during pregnancy and risk of persistent pulmonary hypertension in the newborn: population based cohort study from the five Nordic countries. BMJ. 2012;344:d8012. [DOI] [PubMed] [Google Scholar]
- 6.Furu K, Kieler H, Haglund B, et al. Selective serotonin reuptake inhibitors and venlafaxine in early pregnancy and risk of birth defects: population based cohort study and sibling design. BMJ. 2015;350:h1798. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Nordeng H, Lupattelli A, Romøren M, Koren G. Neonatal outcomes after gestational exposure to nitrofurantoin. Obstet Gynecol. 2013; 121(2 Pt 1):306–313. [DOI] [PubMed] [Google Scholar]
- 8.Huybrechts KF, Palmsten K, Avorn J, et al. Antidepressant use in pregnancy and the risk of cardiac defects. N Engl J Med. 2014; 370(25):2397–2407. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Hernandez-Diaz S, Huybrechts KF, Desai RJ, et al. Topiramate use early in pregnancy and the risk of oral clefts: a pregnancy cohort study. Neurology. 2018;90(4):e342–e351. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Huybrechts KF, Kulldorff M, Hernandez-Diaz S, et al. Active surveillance of the safety of medications used during pregnancy. Am J Epidemiol. 2021;190(6):1159–1168. [DOI] [PubMed] [Google Scholar]
- 11.Prevention CfDCa. Picture of Americia – Reproductive Outcomes Fact Sheet. Accessed September 15, 2021. https://www.cdc.gov/pictureofamerica/pdfs/picture_of_america_reproductive_outcomes.pdf.
- 12.Garcia G Maternal and Child Health (MCH) Update: States Increase Eligibility for Children’s Health in 2007. National Governors Association; 2008. http://www.nga.org/files/live/sites/NGA/files/pdf/0811MCHUPDATE.PDF;jsessionid=7B47A647247DD4E5CB9B709C8F9797AE [Google Scholar]
- 13.Dehlendorf C, Harris LH, Weitz TA. Disparities in abortion rates: a public health approach. Am J Public Health. 2013;103(10):1772–1779. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Pruitt SM, Hoyert DL, Anderson KN, et al. Racial and ethnic disparities in fetal deaths – United States, 2015–2017. MMWR Morb Mortal Wkly Rep. 2020;69(37):1277–1282. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Colugnati FA, Staras SA, Dollard SC, Cannon MJ. Incidence of cytomegalovirus infection among the general population and pregnant women in the United States. BMC Infect Dis. 2007;7:71. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.CMS. ICD-10. Centers for Medicare and Medicaid Services; 2015. 2015. http://www.cms.gov/Medicare/Coding/ICD10/. [PubMed] [Google Scholar]
- 17.Fung KW, Richesson R, Smerek M, et al. Preparing for the ICD-10-CM transition: automated methods for translating ICD codes in clinical phenotype definitions. EGEMS (Wash DC). 2016;4(1):1211. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Rothman K, Greenland S, Lash T. Modern Epidemiology. 3rd ed. LIPPINCOTT WILLIAMS & WILKINS; 2008. [Google Scholar]
- 19.Likis FE, Sathe NA, Carnahan R, McPheeters ML. A systematic review of validated methods to capture stillbirth and spontaneous abortion using administrative or claims data. Vaccine. 2013;31(Suppl 10): K74–K82. [DOI] [PubMed] [Google Scholar]
- 20.Naleway AL, Gold R, Kurosky S, et al. Identifying pregnancy episodes, outcomes, and mother-infant pairs in the vaccine safety datalink. Vaccine. 2013;31(27):2898–2903. [DOI] [PubMed] [Google Scholar]
- 21.Yasmeen S, Romano PS, Schembri ME, Keyzer JM, Gilbert WM. Accuracy of obstetric diagnoses and procedures in hospital discharge data. Am J Obstet Gynecol. 2006;194(4):992–1001. [DOI] [PubMed] [Google Scholar]
- 22.Hornbrook MC, Whitlock EP, Berg CJ, et al. Development of an algorithm to identify pregnancy episodes in an integrated health care delivery system. Health Serv Res. 2007;42(2):908–927. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Andrade SE, Shinde M, Moore Simas TA, et al. Validation of an ICD-10-based algorithm to identify stillbirth in the sentinel system. Pharmacoepidemiol Drug Saf. 2021;30(9):1175–1183. [DOI] [PubMed] [Google Scholar]
- 24.Moll K, Wong HL, Fingar K, et al. Validating claims-based algorithms determining pregnancy outcomes and gestational age using a linked claims-electronic medical record database. Drug Saf. 2021;44(11): 1151–1164. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The data are not available for replication directly from the authors because of IRB restrictions. More information on access to the data can be found at https://rpdrssl.partners.org/.