The prevention of preterm birth (PTB) and its associated morbidity and mortality remains one of the most profound challenges in obstetrics. The inability to untangle the many causal threads that may lead to PTB and develop effective prevention strategies is reflected in the historical classification of cases of PTB not by causal factors, but by the final pathway: whether the delivery initiated spontaneously or occurred because of medical intervention.
Much effort has been directed towards understanding the causes of spontaneous PTB, and many theories have been proposed including but not limited to physical and social stress, maternal or fetal infections and inflammation, placental ischemia and vascular lesions, and uterine overdistension. Increasing attention has been directed towards the identification and management of causes of ‘iatrogenic PTB’, here referred to as ‘medical intervention PTB’. In the absence of standardised treatment protocols, some of these ‘medical intervention PTBs’ may be preventable. For many years, these medical intervention PTBs have been termed or assumed to be ‘indicated’ or clinically justified, in effect, becoming lesser targets of prevention strategies. However, especially in the wake of recent efforts to examine the preventability of early term births (i.e. those births occurring between 37 and <39 completed gestational weeks) characterised by medical intervention, attention has also turned to the prevention of PTB (i.e. those births occurring prior to 37 completed gestational weeks) associated with medical intervention.
In the US, two important trends have been observed recently regarding PTB. First, given that from 1990 to 2006, the national PTB rate rose by more than 20% from 10.6% to 12.8%, the proportion of PTBs occurring prior to 34 completed weeks appears to have remained relatively stable, rising only from 3.3% to 3.7% of all births during the same period. Increasing rates of overall PTB have been largely due to late PTBs occurring between ≥34 and <37 completed gestational weeks, which has increased from 7.3% of all births in 1990 to 9.2% in 2006. Overall PTB rates appear to have declined slightly in the subsequent 2 years to 12.3% in 2008, with most of this drop again relegated to the late PTB category, which in 2008 was 8.8% of all births.
The second trend is that the proportion of all births secondary to medical intervention has greatly increased, and most recently has been estimated to comprise between 35% and 40% of all PTBs.[5, 6] This is consistent with recent findings demonstrating rises in both late PTB and early term (37–39 gestational weeks) births. As investigators have become aware of these trends, they now estimate that 7–8% of medical intervention late PTBs may be preventable, that is, potentially delayed until term.[7, 8] The substantial savings in both costs and morbidity that might accrue with prevention strategies targeted to late PTB have been approximated.[9-11]
Estimates of the preventability of these medical intervention PTBs have rested on distinctions between ‘hard’ and ‘soft’ indications for delivery, the former being indications well known for causing severe maternal or fetal complications (e.g. severe preeclampsia, placenta previa), and the latter being indications that may include a range of maternal or fetal acuity that may impose a small or as yet unquantifiable risk to the mother or preterm fetus (e.g. prior maternal pelvic floor repair, fetal malpresentation or oligohydramnios), or that may be entirely for social benefit (e.g. upcoming military deployment). Cases with these soft indications (or no indications) are frequently referred to as ‘elective’ PTBs, a terminology that we will use here for clarity of our presentation.
The development of an accessible methodology for categorising and tracking elective PTBs should not only assist in quantifying the proportion of PTBs that could potentially be delayed, but also should provide a framework for exploring the rationales for medical intervention, and whether the timing of delivery could be improved to minimise both maternal and neonatal morbidity. Among PTBs caused by medical intervention, we propose a methodology for identifying and analysing ‘elective’ PTBs using administrative data. Furthermore, applying this methodology to California data, we identify the indications for elective PTB, to provide a framework for further exploration of the range of potential rationales for early delivery.
- Top of page
This study was approved by the appropriate Institutional Review Boards, and complies with all stipulated criteria for patient protection. Data were obtained from the linked birth cohort data sets from the California Office of Statewide Health Planning and Development for deliveries in calendar years 1999, 2002 and 2005. This data set combines hospital discharge data with vital statistics data, linking information for mothers and newborns, and has been demonstrated to link over 97% of California births. PTB was identified from the birth certificate gestational age of ≥24 and <37 completed weeks. Early PTB was defined as a gestational age <34 completed weeks, and late PTB was defined as a gestational age of ≥34 to <37 completed weeks. Pregnancies with multiple gestations and fetal demises were excluded.
For the years studied, the California birth certificate relied on the last menstrual period (LMP) for determining the gestational age. This gestational age estimate is often inconsistent with the ‘obstetrical estimate’ of the gestational age, which requires that birth attendants base the estimate on all perinatal factors, including ultrasound, and specifically exclude the neonatal examination as a determinant of the estimate. Callaghan and Dietz have reported that birthweights for pregnancies estimated to be 28–36 weeks of gestational age by LMP, appear to be biased upward from ‘gold standard’ estimates, indicating that if the birthweights were accurate, these pregnancies would likely have an erroneously low estimate of the true gestational age.
Given this known bias, for the purposes of this investigation, we used two strategies to improve the case-definition for both early and late PTB. First, we excluded cases where the birthweight and gestational age appeared extremely unlikely. Using the birthweight charts derived using obstetrical clinical estimates (not LMP) as reported by Olsen et al., we excluded cases from the entire study population using the following rules: (1) exclude all cases with birthweight < 3rd percentile for 24 weeks (<464 g for females and < 497 g for males), (2) exclude all early PTB cases with birthweight > 97th percentile for 33 weeks (>2580 g for females and >2688 g for males), (3) exclude all late PTB cases with birthweight < 3rd percentile for 34 weeks (<1523 g for females and <1589 g for males), (4) exclude all late PTB cases with birthweight < 97th percentile for 36 weeks (>3667 g for females and > 3737 g for males), and (5) exclude all term deliveries with birthweight < 3rd percentile for 37 weeks (<1958 g for females and <2103 g for males). Second, because birth records linked to discharge data also had information regarding gestational age, we excluded deliveries with a birth certificate gestational age < 37 weeks if maternal codes for a normal term delivery (650) or post-dates (654.1645.2) existed simultaneously with the absence of a preterm code (765.22–765.28) for the neonate. We acknowledge that these exclusions would tend to decrease the estimated rate of overall PTBs because they differentially excluded more preterm than term deliveries, as preterm deliveries are more likely to be miscoded. However, these exclusions should improve the accuracy of our estimates within the cohorts of both early and late PTBs, which is the focus of our study.
Using codes from the International Classification of Diseases – Version 9 – Clinical Modification (ICD-9),we classified PTBs hierarchically into one of three groups: ‘Complicated’ pregnancies (with specified medical conditions defined below); uncomplicated pregnancies with medical intervention (for current purposes, termed ‘Elective’); and uncomplicated pregnancies with spontaneous onset of labour, including preterm premature rupture of membranes (PROM) (for current purposes, termed ‘Spontaneous’).
Relying largely on the classification scheme developed by Ananth and Vintzileos for medical intervention PTB using ICD-9 codes from administrative data, delivery records noting one or more of the following conditions were placed in the Complicated Group: hypertension (642, 401–405), vaginal bleeding, previa or accreta (641), intrauterine growth restriction (656.5), non-reassuring fetal heart rate monitoring (656.3, 659.7), isoimmunisation (656.1,2), and maternal cardiac and renal conditions (425, 648.5,6 and 646.2). Within this group, no attempt was made to distinguish between patients who did and did not undergo spontaneous labour.
Once the Complicated Group was created, the remaining ‘uncomplicated’ deliveries associated with an induction of labour or caesarean delivery were identified and categorised as Elective. Induction of labour was identified if it was noted on the birth certificate, or if ICD-9 procedure codes 73.1, 73.01 or 73.4, or ICD-9 codes for failed induction 659.0,1,2 were present. Pregnancies undergoing primary or repeat caesarean in the absence of labour were identified using a previously published algorithm. Women with codes for premature or prolonged rupture of membranes (658.1,2), amnionitis (658.4) or spontaneous labour (644.2) were excluded from this group, and they, and all remaining deliveries were classified in the Spontaneous Group. See Figure 1 for an illustration of this hierarchical classification.
Rates of PTB were calculated for each study year, and using hierarchical multivariable logistic and multinomial regression models, were adjusted for the following: maternal age, race, education, parity, prenatal care adequacy (Kotelchuck index 3–4 = Adequate vs. 1–2 = Inadequate), prior caesarean, hospital delivery volume, obstetrical teaching status and neonatal intensive care unit availability. Adjusted PTB rate differences were tested for the three period differences using pairwise t-tests (1999 vs. 2002, 1999 vs. 2005 and 2002 vs. 2005), and the overall period effect was tested with an F-test (overall effect of the three-level predictor for year), based on model estimates of effect sizes and standard errors. The following groupings were used in analyses with separate models fitted at each level of analysis: examination of all deliveries grouped by gestational age into term (>36 weeks), late PTB (34–36 weeks) and early PTB (<34 weeks), stratification of preterm deliveries by gestational age as late PTB or early PTB, and further stratification by Complicated, Elective or Spontaneous Groups. Generalised linear models with random intercepts for hospitals were fitted with a logit link function and Bernoulli distribution for the binary PTB outcome and with a generalised logit link function and multinomial distribution for the three-level PTB grouping. Analyses were performed using the generalised linear mixed model procedure GLIMMIX in SAS version 9.2 (Cary, NC, USA).
Finally, to provide a framework for further exploration of the rationales for early delivery, the ICD-9 codes for other conditions (i.e. the ‘soft’ indications associated with the deliveries) in the Elective and Spontaneous Groups (for both early and late PTBs) from 2005 were examined and tabulated for descriptive purposes.
- Top of page
A total of 1 387 565 singleton liveborn deliveries meeting inclusion criteria occurred in the final study population, with 99 614 (7.2%) classified as preterm. Table 1 describes the derivation of the study population. Of the PTBs identified in the original data, 17.8% were excluded because of improbable birthweights; only 0.3% of the term births were excluded for this reason. Adjusted PTB rates (divided into early and late PTB rates) for each of the study years are shown in Figure 2. During the study period, the overall increase in the PTB rate was 9.0% (P < 0.0001), with a statistically insignificant increase in the early PTB rate of 8.1% (P = 0.2409) and a significant increase in the late PTB rate of 9.3% (P < 0.0001). Late PTBs accounted for 79.9% of all PTBs and given the similar rise in PTB rates in both early and late PTB groups, the proportion of late PTBs did not change substantially over the study period (F-test P = 0.4636).
Table 1. Derivation of study population
| ||Data description (California linked data for 1999, 2002, 2005)||n term cases excluded||n preterm cases excluded||Sample size||Preterm birth cases (%)|
|1||All singleton liveborn deliveries excluding missing or extreme gestational age (<168 or >350 days)|| || ||1 433 364||132 299 (9.23%)|
|2||Excluded missing newborn gender||9309||1123||1 422 932||131 176 (9.22%)|
|3||Excluded cases with low birthweight: females < 464 g and males < 497 g (third percentile of 24 weeks)||31||198||1 422 703||130 978 (9.21%)|
|4||Excluded early preterm deliveries with high birthweight: females > 2580 g and males > 2688 g (97th percentile of 33 weeks)||0||12 222||1 410 481||118 756 (8.42%)|
|5||Excluded late preterm deliveries with low birthweight: female < 1745 g and male < 1886 g (third percentile of 34 weeks)||0||961||1 409 520||118 756 (8.42%)|
|6||Excluded late preterm deliveries with high birthweight: females >3667 g and males > 3737 g (97th percentile of 36 weeks)||0||10 003||1 399 517||107 792 (7.70%)|
|7||Excluded term deliveries with low birthweight: females <1958 g and males < 2103 g (third percentile of 37 weeks)||3774||0||1 395 743||107 792 (7.72%)|
|8||Excluded preterm birth cases because of presence of ICD-9 codes for maternal full term or post-dates and no neonatal ICD-9 code for preterm birth||0||8178||1 387 565||99 614 (7.18%)|
Tables 2-4 describe the trends in PTB rates for all, early and late PTBs, stratified by Complicated, Elective and Spontaneous Groups. While Complicated PTBs increased 12.1% overall, with similar increases for both early and late PTBs, Elective PTBs demonstrated the largest increase (27.7%), rising from 8.7% to 11.1% of all PTBs, and was nearly entirely confined to the late PTB group.
Table 2. Adjusted trends in overall preterm delivery in all hospital deliveries in California in 1999, 2002 and 2005, by preterm birth causal groups
|Group||Yeara||Per cent change||Per cent of births that were late pretermc|
Table 3. Adjusted trends in preterm delivery in early preterm birth hospital deliveries in California in 1999, 2002 and 2005, by preterm birth causal groups
|Group||Yeara||Per cent change|
Table 4. Adjusted trends in preterm delivery in late preterm birth hospital deliveries in California in 1999, 2002 and 2005, by preterm birth causal groups
|Group||Yeara||Per cent change|
There was a wide distribution of conditions associated with deliveries in the Complicated Group. Many of the patients in this group had multiple conditions, and the overall proportion of patients with each of the following conditions was: hypertension (44.0%), bleeding/previa/accreta (22.0%), intrauterine growth restriction (10.5%), non-reassuring fetal heart rate/fetal distress (40.2%), isoimmunisation (4.4%), maternal renal condition (0.6%), maternal cardiac condition (2.2%) and maternal acute respiratory distress (0.2%).
A variety of conditions were identified among deliveries in both the Elective and Spontaneous Groups (Table 5) including malpresentation, prior caesarean (scar type not distinguishable by ICD-9 code), mental health conditions, fetal anomalies, human immunodeficiency virus infection, cervical incompetence and prior pelvic floor repair. Among all births in the Elective Group, 50.3% had at least one, 14.5% had at least two and 3.0% had at least three of the conditions listed in Table 5. At least one of these conditions was present in 66.9% of early PTBs and 50.1% of late PTBs (data not shown). Even if all patients in the Elective late PTB Group with ‘soft’ indications were reclassified to the Complicated Group, 6.0% of patients with a late PTB would have had a medical intervention with no hard or soft indication for delivery. This proportion rose across the study period, from 4.8% in 1999, to 5.8% in 2002, to 7.1% in 2005 (P < 0.0001).
Table 5. Conditions identified in both early and late preterm birth patients in the Elective and Spontaneous Groups from 1999, 2002 and 2005
n (%) identified in Elective Group
n (%) identified in Spontaneous Group
|Malpresentation|| || ||1455 (14.9)||5144 (9.3)|
|652.21||Breech||1086 (11.1)||3220 (5.8)|
|652.31||Transverse lie||119 (1.2)||488 (0.9)|
|652.81||Compound presentation||243 (2.5)||1356 (2.4)|
|660.01||Malpresentation with obstruction||22 (0.2)||1310 (2.4)|
|Fluid/placenta|| || ||717 (7.3)||1702 (3.1)|
|658.01||Oligohydramnios||607 (6.2)||1133 (2.0)|
|657.01||Polyhydramnios||87 (0.9)||326 (0.6)|
|656.71||Placental infarct||28 (0.3)||255 (0.5)|
|Fetal compromise|| || ||1372 (14.1)||9997 (18.0)|
|663.11, 663.21, 663.31||Cord compression||1305 (13.4)||9873 (17.8)|
|655.71||Decreased fetal movement||89 (0.9)||160 (0.3)|
|Suspected fetal macrosomia||656.61||Fetal macrosomia||96 (1.0)||89 (0.2)|
|Fetal anomaly||655.01||Central nervous system malformation||32 (0.3)||68 (0.1)|
|Fetal exposure|| || ||221 (2.3)||1462 (2.6)|
|655.81||Suspected fetal damage: toxin||103 (1.1)||408 (0.7)|
|655.51||Suspected fetal damage: drug||2 (<0.1)||45 (0.1)|
|305.70||Amphetamine use||44 (0.5)||422 (0.8)|
|305.50||Opioid use||9 (0.1)||49 (0.1)|
|305.60||Cocaine use||16 (0.2)||195 (0.4)|
|305.20||Cannibis use||27 (0.3)||271 (0.5)|
|648.31||Drug dependence||32 (0.3)||273 (0.5)|
|Asthma||493.90||Asthma||155 (1.6)||754 (1.4)|
|Diabetes|| || ||917 (9.4)||3697 (6.7)|
|648.01||Diabetes mellitus in pregnancy||135 (1.4)||479 (0.9)|
|648.81||Gestational diabetes mellitus||779 (8.0)||3218 (5.8)|
|250.00, 250.01||Diabetes mellitus||103 (1.1)||395 (0.7)|
|V58.67||Insulin use (0 cases in 1999, 2002)||51 (0.5)||145 (0.3)|
|Thyroid|| || ||117 (1.2)||600 (1.1)|
|648.11||Thyroid||111 (1.1)||589 (1.1)|
|244.9||Hypothyroid||82 (0.8)||394 (0.7)|
|Mental/social condition|| || ||540 (5.5)||2601 (4.7)|
|648.41||Mental health condition||263 (2.7)||1864 (3.4)|
|311||Depressive disorder||39 (0.4)||284 (0.5)|
|659.81||<16 years old||273 (2.8)||711 (1.3)|
|Maternal soft tissue|| || ||567 (5.8)||2188 (3.9)|
|654.01, 752.3||Bicornuate uterus||76 (0.8)||355 (0.6)|
|654.11, 218.9||Uterine fibroids||188 (1.9)||546 (1.0)|
|654.51||Cervical incompetence||73 (0.7)||901 (1.6)|
|654.61||Cervix soft tissue condition||24 (0.2)||140 (0.3)|
|654.91||Other uterine scar||62 (0.6)||64 (0.1)|
|654.41||Pelvic floor repair||180 (1.8)||284 (0.5)|
|Maternal infection|| || ||376 (3.9)||2108 (3.8)|
|646.61||Genito-urinary tract infection||175 (1.8)||1202 (2.2)|
|599.0||Urinary tract infection||56 (0.6)||571 (1.0)|
|647.61||Viral infection||162 (1.7)||593 (1.1)|
|647.81||Other infection/parasite||30 (0.3)||304 (0.5)|
|054.10||Herpes||43 (0.4)||150 (0.3)|
|042||Human immunodeficiency virus||11 (0.1)||16 (<0.1)|
|Other condition|| || ||50 (0.5)||145 (0.3)|
|643.21||Hyperemesis||9 (0.1)||30 (0.1)|
|780.39||Convulsions||41 (0.4)||115 (0.2)|
Nearly half of the patients in the Spontaneous Group (42.8%) also had at least one of the conditions from Table 5 (data not shown). For example, among patients with late PTB, oligohydramnios was present in 6.1% of the Elective Group and 1.9% of the Spontaneous Group.
- Top of page
Using California administrative data from 1999 to 2005 for liveborn singletons ≥24 and <37 completed gestational weeks, we hierarchically classified deliveries into three groups (Complicated, Elective and Spontaneous), and quantified the increases in the Complicated and Elective Groups over the study period. While Complicated PTBs increased 12.1% overall, with similar increases for both early and late PTBs, Elective PTBs demonstrated the largest increase (27.7%), rising from 8.7% to 11.1% of all PTBs, with the increase nearly entirely confined to the late PTB group. Although other studies have used a variety of methods and data sources, our results appear to be consistent with those reported by other investigators using data from administrative records, and from both paper and electronic medical record (EMR) reviews.
Davidoff et al. examined birth certificates for 1992–2002 from the US National Center for Health Statistics and used a hierarchical scheme to classify patients into those with PROM, then medical intervention, and finally spontaneous onset of labour. For 2002, Davidoff et al. identified a 41% intervention rate (similar to the rate of our Complicated and Elective Groups combined of 44% in 2002), a 57% spontaneous rate and a 2% PROM rate (similar to the rate of our Spontaneous Group of 56% in 2002). Our methodology differed from theirs in three aspects: (1) we did not consider PROM as distinct from spontaneous onset of labour, (2) instead of a single medical intervention group that assumed that all medical interventions were ‘indicated’, we first created a ‘complicated’ group of patients that had standard indications for preterm delivery (regardless of medical intervention), and (3) we subsequently created an additional group, the Elective Group, in which medical intervention was used in the absence of standard indications for preterm delivery.
As the obstetrical community became aware of growing numbers of cases of ‘late’ and ‘near term’ PTB in which medical intervention might have been prevented or delayed, investigators began to develop methods to quantify ‘elective’ late PTBs. Laughon et al. used EMR data from the Safe Labor Consortium of 31 clinical centres and hospitals over calendar years 2002–08 to estimate the proportion of late PTBs that may have been preventable. They identified 62% of the cases of late PTB as attributable to spontaneous labour or PROM, 31% as having an ‘indicated’ medical intervention (induction or elective caesarean delivery), and 7% as having an unknown or ‘soft’ precursor to iatrogenic PTB. These soft precursors included suspected macrosomia without diabetes, history of a fetal, maternal or obstetric complication in a previous pregnancy, and lack of documentation of any pregnancy complication. Although our case definitions differ slightly, our results are similar for 2005, demonstrating that 32% of late PTB deliveries have standard, or hard indications for medical intervention, and 13% have unknown or soft indications.
Approximately half of the patients in the Elective and Spontaneous Groups had at least one of the soft indications for delivery listed in Table 5. Using the condition of oligohydramnios as an example of an indication that may have a wide range of acuity, Elective late PTB patients had a rate of approximately 6%, compared with 2% in the Spontaneous Group. Such results suggest that in some cases, oligohydramnios is considered an indication for delivery while in others it is not. There is insufficient detail in administrative data to characterise the severity of oligohydramnios to permit distinction between those cases that would and would not justify intervention. In addition, there is no evidence-based or consensus-based standard to suggest the degree of oligohydramnios that requires intervention. What is most important is that clearer standards might have allowed some of these early deliveries to be delayed.
In another retrospective study of precursors to late PTB using a clinical database, Holland et al. estimated that up to 17% of these deliveries may have been avoidable. This 17% represented not only cases with standard indications that might have been temporised or delayed until term, but also those with unknown or soft indications. A key to avoidability was the ability to determine the acuity of the condition and stability of the mother and fetus, for example whether a medical condition was worsening, or Doppler studies were reassuring. Such data are clearly limited in administrative data sets. Yet, in spite of the different methodologies used, results from these various studies have been remarkably consistent and suggest the need for clearer guidance regarding when to intervene for ‘soft’ indications. The Society for Maternal Fetal Medicine has recently suggested some general guidelines for ‘hard’ indications and called for additional research for other conditions where the maternal/fetal risks and benefits of early vs. delayed delivery are less clear.
A large majority of late PTBs do not appear to be avoidable; however, this leaves a substantial number of patients experiencing PTB (7–17% of all PTBs according to these studies, and 12% from our study) for whom temporisation to reach term may be possible. Even after accounting for all potential indications, up to 6% of late PTBs remained classified as undergoing medical intervention without any documented indication, suggesting that there remains ample room to employ preventive strategies within targeted segments of this population. Because many of the conditions identified as soft indications for delivery may present within a wide range of acuity, such a categorisation may best serve for case finding, with further exploration of the rationale for delivery provided by medical chart review.
Similarly, medical chart data regarding patients in the Complicated Group may deserve further exploration to determine opportunities to delay delivery. The severity of the condition and the stability of both mother and fetus cannot be assessed with administrative data, and it is likely that the early delivery of some patients in this category might be preventable. Of note, some cases in the Complicated Group were delivered spontaneously. Such complicated deliveries may have a wide range of acuity, and likely require complex decision making regarding delivery, irrespective of whether delivery is or is not initiated spontaneously. For example, severe preeclampsia is traditionally considered an absolute indication for delivery regardless of gestational age, yet more recently, there is a growing precedent for ‘expectant’ inpatient management for patients near viability or early preterm. Future studies that identify and categorise such patients should be able to more thoroughly explore the variety of conditions justifying medical intervention PTB and the optimal timing of delivery.
This study has attempted to create an accessible methodology for the identification and tracking of both complicated and ‘elective’ preterm deliveries for the purpose of laying the foundation for determining their preventability. To our knowledge, no other current methodology is using administrative data to achieve this set-up. Our results are consistent with similar efforts that have relied on EMR data or clinical data sets and are less resource-intensive. They do require the use of linked administrative data sets for mother and baby (or potentially, use of the EMR or birth certificate) to obtain gestational age, birthweight and gender, as these items are not included in discharge data. Increasingly, linked maternal and neonatal records are becoming more of a necessity for the assessment of the quality and safety of obstetrical care.
The strengths of this study include the ability to estimate elective PTB rates from administrative data, adjusting for demographic shifts and minimising inaccuracies because of estimation of gestational age from LMP, and separating out ‘hard’ from ‘soft’ indications for delivery.
As mentioned above, we opted to exclude 18% of reported PTBs because of mismatched gestational age and birthweight, which largely occurs because of the use of LMP for calculating gestational age instead of using a clinical or obstetrical estimate of gestational age on the birth certificate. Such exclusions were likely to have decreased the PTB rate reported in this study by as much as 1.6% from the original study sample, although they should have greatly improved the specificity of the categorisation of PTB as early and late, and within these strata, the categorisation of PTB as Complicated, Elective or Spontaneous. Current national efforts to shift to a clinical or obstetrical estimate of gestational age for the birth certificate is an important effort and should improve our identification of PTB. Further decreasing the overall PTB rate in our report was our exclusion of multiple gestations (which are more likely to be preterm) and gestations under 24 weeks. A further concern is that the sensitivity of hospital discharge data for many relevant maternal and fetal conditions is often low, and may account for why some deliveries are classified as Elective.
In summary, administrative data appear useful for the tracking of increasing rates of complicated and elective preterm deliveries. The use of such a methodology may enable further exploration of practice opportunities to temporise or prevent these early deliveries among at-risk women – particularly among late preterm gestations with ‘soft’ or ‘no’ indications for delivery.