Evaluation of the impact of fetal fibronectin test implementation on hospital admissions for preterm labour in Ontario: a multiple baseline time-series design

Authors

  • DB Fell,

    1. Better Outcomes Registry & Network (BORN) Ontario, Children's Hospital of Eastern Ontario Research Institute, Centre for Practice Changing Research, Ottawa, ON, Canada
    Search for more papers by this author
  • AE Sprague,

    Corresponding author
    1. Better Outcomes Registry & Network (BORN) Ontario, Children's Hospital of Eastern Ontario Research Institute, Centre for Practice Changing Research, Ottawa, ON, Canada
    • Correspondence: Dr AE Sprague, Better Outcomes Registry & Network (BORN) Ontario, Children's Hospital of Eastern Ontario Research Institute, 401 Smyth Road, Centre for Practice Changing Research, Ottawa, ON, Canada, K1H 8L1. Email asprague@bornontario.ca

    Search for more papers by this author
  • JM Grimshaw,

    1. Clinical Epidemiology Program, Ottawa Hospital Research Institute, Department of Medicine, University of Ottawa, Ottawa, ON, Canada
    Search for more papers by this author
  • AS Yasseen III,

    1. Clinical Research Unit, Children's Hospital of Eastern Ontario, Centre for Practice Changing Research, Ottawa, ON, Canada
    Search for more papers by this author
  • D Coyle,

    1. Department of Epidemiology and Community Medicine, University of Ottawa, Ottawa, ON, Canada
    Search for more papers by this author
  • SI Dunn,

    1. Better Outcomes Registry & Network (BORN) Ontario, Children's Hospital of Eastern Ontario Research Institute, Centre for Practice Changing Research, Ottawa, ON, Canada
    Search for more papers by this author
  • SL Perkins,

    1. Department of Pathology and Laboratory Medicine, University of Ottawa and The Ottawa Hospital, Ottawa, ON, Canada
    Search for more papers by this author
  • WE Peterson,

    1. School of Nursing, Faculty of Health Sciences, University of Ottawa, Ottawa, ON, Canada
    Search for more papers by this author
  • M Johnson,

    1. Better Outcomes Registry & Network (BORN) Ontario, Children's Hospital of Eastern Ontario Research Institute, Centre for Practice Changing Research, Ottawa, ON, Canada
    Search for more papers by this author
  • PS Bunting,

    1. Division of Biochemistry, Department of Pathology and Laboratory Medicine, University of Ottawa, Ottawa, ON, Canada
    Search for more papers by this author
  • MC Walker

    1. Better Outcomes Registry & Network (BORN) Ontario, Children's Hospital of Eastern Ontario Research Institute, Centre for Practice Changing Research, Ottawa, ON, Canada
    Search for more papers by this author

Abstract

Objective

To determine the impact of a health system-wide fetal fibronectin (fFN) testing programme on the rates of hospital admission for preterm labour (PTL).

Design

Multiple baseline time-series design.

Setting

Canadian province of Ontario.

Population

A retrospective population-based cohort of antepartum and delivered obstetrical admissions in all Ontario hospitals between 1 April 2002 and 31 March 2010.

Methods

International Classification of Diseases codes in a health system-wide hospital administrative database were used to identify the study population and define the outcome measure. An aggregate time series of monthly rates of hospital admissions for PTL was analysed using segmented regression models after aligning the fFN test implementation date for each institution.

Main outcome measure

Rate of obstetrical hospital admission for PTL.

Results

Estimated rates of hospital admission for PTL following fFN implementation were lower than predicted had pre-implementation trends prevailed. The reduction in the rate was modest, but statistically significant, when estimated at 12 months following fFN implementation (−0.96 hospital admissions for PTL per 100 preterm births; 95% confidence interval [CI], −1.02 to −0.90, P = 0.04). The statistically significant reduction was sustained at 24 and 36 months following implementation.

Conclusions

Using a robust quasi-experimental study design to overcome confounding as a result of underlying secular trends or concurrent interventions, we found evidence of a small but statistically significant reduction in the health system-level rate of hospital admissions for PTL following implementation of fFN testing in a large Canadian province.

Introduction

Fetal fibronectin (fFN) is an extracellular matrix glycoprotein[1-4] produced in fetal tissues and mainly found in the amniotic fluid, placenta, the interface between the maternal and fetal components of the choriodecidual junction[5] and in cytotrophoblasts.[2, 6-9] Cervicovaginal secretions contain high levels of fFN early in gestation and again just before term delivery[2, 6]; however, fFN should not be detectable between approximately 24 and 34 weeks of gestation.[10] Detection levels of ≥50 ng/ml at ≥22 weeks of gestation have been associated with an increased risk of spontaneous preterm birth.[11]

The rapid fFN test exploits the gestational age-specific appearance of fFN in cervicovaginal secretions and is indicated for use in pregnant women presenting with signs and symptoms of preterm labour (PTL) between 24 and 34 weeks of gestation. The presence or absence of fFN detected by the test is one of the best predictors of preterm birth found to date, particularly for the determination of which women presenting with PTL are not at immediate risk for preterm birth.[11] Numerous systematic reviews/meta-analyses,[5, 11, 12] clinical trials[8, 13, 14] and cohort studies[15-18] reporting diagnostic test properties of the fFN test in different obstetrical populations have demonstrated high negative predictive values (between 75% and 100%), indicating that women who test negative for fFN have a high probability of not delivering within the subsequent 7–14 days. The utilisation of the fFN test for the evaluation of women presenting with PTL has achieved measurable reductions in hospital admissions, maternal transfer rates and length of stay for PTL in numerous settings[2, 9, 15, 19-25]; however, to our knowledge, an evaluation at a health system level has not been carried out previously.

In 2009, the government of Canada's most populous province (Ontario) provided funding, practice guidelines and educational materials to support the implementation of new fFN testing programmes, or to support existing programmes, based on recommendations from an Expert Panel.[26] We carried out an evaluation of fFN test implementation in Ontario to determine the effect on rates of hospital admission for PTL. Using a multiple baseline time-series design, our objective was to conduct a naturalistic evaluation of fFN test implementation within a complete healthcare system.

Methods

Study design

This population-based retrospective study used a multiple baseline time-series design.[27] This quasi-experimental approach is analogous to an interrupted time-series (ITS) design,[28, 29] except that the intervention is introduced in a staggered fashion at different sites, as compared with the singular intervention point across sites approach of the conventional ITS. Although ITS designs are useful for determining whether an intervention has had an effect on the outcome that is significantly different from underlying secular trends prior to implementation, the possibility of confounding by temporally concurrent interventions still poses an important threat to the validity of this approach.[27] As the implementation of fFN testing programmes in Ontario was staggered across several years and we knew there to be temporal trends in the rates of antepartum hospital admission in Canada,[30] we considered the multiple baseline time-series design to be most appropriate for accounting for the temporal trends, whilst simultaneously accommodating the natural experiment created by the staggered implementation dates of fFN test programmes across the province's hospitals.

Data sources

We obtained 8 years of obstetric admission records from the Canadian Institute for Health Information's Discharge Abstract Database (CIHI-DAD) for Ontario hospitals between 1 April 2002 and 31 March 2010. This hospitalisation database captures approximately 98% of all the births in the province, and contains demographic, clinical (i.e. medical diagnoses, interventions) and administrative information, such as hospital institution codes. Diagnoses and interventions in the CIHI-DAD were coded according to the Canadian implementation of the International Classification of Diseases, 10th Revision, Canada (ICD-10-CA) and the Canadian Classification of Health Interventions (CCI) coding systems, respectively. Record-level information on whether or not fFN testing was used prior to hospital admission was not available in the dataset. To confirm programme implementation dates and to obtain additional information on the characteristics of the fFN testing programme for each hospital site, we conducted an Internet-based hospital laboratory survey of all 111 Ontario hospital sites that had implemented an fFN testing programme and matched the information to the hospitalisation dataset using the hospital institution codes.

Measurement of hospital admissions for PTL

The algorithm for the identification of PTL hospital admissions in the hospitalisation database was developed using information from the literature,[30] and consultation with a nosologist. The definitions and ICD-10-CA codes for the measurement of hospital admission for PTL, and for distinguishing between obstetrical delivery records and antepartum, undelivered obstetrical records, are provided in Table 1. Hospital admissions for PTL were defined as undelivered, antepartum hospital admissions of pregnant women with a singleton fetus, an admission gestational age between 24 and 34 weeks, and an ICD-10-CA diagnostic code for PTL in any of the 25 diagnostic code fields. Records thus identified were not included in the PTL definition if they had a primary diagnosis of antepartum haemorrhage, incompetent cervix or premature rupture of membranes. To fully capture the burden of PTL admissions on the health system, all hospital admissions for PTL during a unique pregnancy were included in the analysis, regardless of whether the subsequent admission record corresponded to a separate episode of PTL during the same pregnancy or represented an inter-hospital transfer for the same episode of care. We expressed our outcome as a monthly rate, calculated as the number of hospital admissions for PTL per 100 preterm births over the same time interval. Although the clinical population of interest for our evaluation was all women presenting with PTL, we were only able to measure PTL once a woman was admitted to hospital (i.e. not those who were evaluated and sent home). Women with a preterm delivery were thus considered to be the most relevant proxy population for the denominator of our rate calculation, as preterm delivery is preceded by PTL a high proportion of the time.

Table 1. Algorithms and database codes used to define study variables
Study variableDatabase codes
ICD-10-CACCPCCI
  1. CCI, Canadian Classification of Health Interventions; CCP, Canadian Classification of Procedures; ICD-10-CA, International Classification of Diseases, 10th Revision, Canada.

  2. a

    Any pregnancy outcome (live birth or stillbirth).

  3. b

    Antepartum admissions between 24 and 34 weeks of gestation.

Delivered admissions a
(a) Last digit of ‘1’ or ‘2’ in any diagnostic code fieldO10–O99  
OR
(b) Code present in any diagnostic code field

Z37

O80–O84

860–864

868

869

871

872

5.MD.60

5.MD.50

Antepartum admissions
Last digit of ‘3’ or ‘4’ in any diagnostic code fieldO10–O99  
Multiple gestation pregnancies

O30

O31

O32.5

  
Preterm labour (PTL) definition b
(a) Inclusion criteria (code present in any diagnostic code field)

O47.003

O47.903

O60.003

O75.803

  
(b) Exclusion criteria (code present in primary diagnostic code field)

O34.303

O34.30

O46.003

O43.013

O46.093

O46.803

O46.903

O42

O42.013

O42.113

O46.0

P01.0

  

Statistical analysis

The rate of hospital admission for PTL was first calculated for each institution at monthly time points, producing a chronological time series of data points on either side of the institution's fFN test implementation date. The implementation date was denoted ‘time zero’ (T0), and ranged from Tn pre-fFN implementation to T+n post-fFN implementation, where n represents the number of months of data available for a given institution. Institutions that implemented their fFN programme early in the study period had a smaller number of pre-intervention monthly time points available for analysis, and a larger number of post-intervention time points, and the converse was true for those that implemented their fFN programme later in the study period. To ensure a continuous series of evenly spaced time points for each institution, we used regression imputation[31] to fill in time points with missing numerator or denominator values in a given month (regression imputation was required for seven hospitals in total, with a range of 1–25 imputed time points).

To provide a health system-level assessment of the effect of fFN testing, we aggregated the time series of data points from each hospital (regardless of the number of data points before and after fFN implementation) and aligned monthly observations for each institution with respect to the fFN testing programme implementation date (T0).[32] We then truncated this aggregated time series to isolate the most effective range of constant variance, to account for the limited number of institutions contributing data at time points at the extremes of the series. To this reduced series of data points, we applied a segmented (piecewise) generalised linear regression model with a negative binomial distribution[33] to fit a regression line for each segment of time (i.e. pre-fFN implementation and post-fFN implementation), where the intercept represented the baseline rate of hospital admission for PTL and the slope represented the trend in rates of hospital admission for PTL over the segment of time. We compared the estimated rate of PTL admissions post-fFN implementation (generated by the post-fFN intervention model) with the predicted rate of PTL admissions in the post-fFN time period had the pre-fFN trends in PTL admission rates prevailed (generated by extrapolating the pre-intervention model into the post-fFN period) by computing the difference between the two regression slopes at 12, 24 and 36 months post-fFN implementation (T12, T24 and T36, respectively). The magnitude of the change was expressed as a difference (with 95% confidence intervals [CI]). As the measure of effect is reported on the additive scale, a negative value indicates a reduction in the rate of hospital admissions for PTL following fFN implementation (and vice versa), and a 95% CI that does not include zero indicates that the change is statistically significant at the P = 0.05 level.

We additionally carried out sensitivity analyses to determine whether there was consistency in our province-level findings across each of the individual hospital-level time series.[27] Only those hospital sites with a minimum of 12 months of data both pre- and post-implementation of fFN testing were included in individual hospital-level models to ensure an adequate number of time points for the establishment of a stable trend in each time segment.[28, 29]

All data processing steps and descriptive statistics were produced using SAS version 9.2 on a Windows platform (SAS Institute Inc., Cary, NC, USA). All other analyses were carried out using R version 2.13.1.[34]

Results

The Internet-based hospital laboratory survey was distributed to all 111 hospital sites identified by the Ontario Ministry of Health and Long-Term Care as having implemented an fFN testing programme, and complete survey information on the timing and characteristics of the fFN testing programme was received from 98 hospital sites (88%). The remaining 13 hospital sites either had not yet implemented their fFN programme or did not respond to the survey. Of the 98 responding hospitals with complete information, 58 had already implemented their fFN testing programmes prior to the government funding becoming available (i.e. before March 2009). Overall, fFN testing programmes in these 98 Ontario hospital sites were implemented across a 9-year time period, ranging from January 2002 to December 2010.

Between 1 April 2002 and 31 March 2010, ICD-10-CA diagnostic codes identified 1 196 145 obstetrical admissions to Ontario hospitals, of which 1 195 810 were admissions within the 98 hospitals included in our analysis. There were 1 084 485 deliveries in the 98 study hospitals and 13 707 undelivered hospital admissions for PTL, corresponding to 11 897 unique pregnancies (the latter reflecting the reduced number of records after accounting for multiple admissions within an episode of care involving an inter-hospital transfer, as well as for separate episodes of PTL during the same pregnancy). The maternal, pregnancy and hospital characteristics of the full study population are provided in Table 2.

Table 2. Distribution of maternal, pregnancy and hospital characteristics of obstetrical admissions in 98 study hospitals from 1 April 2002 to 31 March 2010
 All delivered obstetrical admissions (n = 1 084 485)Preterm labour admissionsa (n = 13 707)
n (%)n (%)
  1. a

    Preterm labour admissions were defined as undelivered antepartum admissions with a single fetus and an admission gestational age between 24 and 34 weeks with a diagnostic code for preterm labour in any diagnostic code field, and without a primary diagnosis of antepartum haemorrhage, incompetent cervix or premature rupture of membranes.

Maternal age (years)
<35864 892 (79.8)11 793 (86.0)
≥35219 593 (20.3)1914 (14.0)
Gestational age at admission (weeks)
<243984 (0.4)
24–274393 (0.4)2307 (16.8)
28–317739 (0.7)5022 (36.6)
32–3419 018 (1.8)6378 (46.5)
35–3647 734 (4.4)
≥371 000 589 (92.4)
Missing1028 (0.1)
Hospital level of care
No designation716 (0.1)21 (0.2)
1117 448 (10.8)964 (7.0)
2460 875 (42.5)3722 (27.2)
2+260 659 (24.0)3679 (26.8)
3172 261 (15.9)4238 (30.9)
M372 526 (6.7)1083 (7.9)
Region of hospital
Central West161 782 (14.9)2089 (15.3)
East South East139 680 (12.9)2288 (16.7)
Greater Toronto Area563 280 (52.0)6116 (44.6)
North88 114 (8.1)1243 (9.1)
South West130 596 (12.1)1971 (14.4)
Area size of hospital location
Rural22 294 (2.1)171 (1.3)
Urban1 062 191 (97.9)13 536 (98.8)

After aligning the fFN implementation date (i.e. T0) for each of the 98 hospitals included in the health system-level multiple baseline analysis, the most effective range of constant variance extended from 72 months pre-fFN implementation (i.e. T−72) to 38 months post-fFN implementation (i.e. T+38), and this was the range of data points used in the segmented regression models. Figure 1 displays the time series plot of rates of hospital admission for PTL at each time point (graphed using the primary y axis), representing a combined rate across the total number of hospitals contributing to each given time point (shown using the secondary y axis), which varied over time owing to the multiple baseline (i.e. the staggered fFN programme implementation dates across provincial hospitals). Figure 1 also presents the fitted regression lines corresponding to the pre-and post-fFN implementation time periods.

Figure 1.

Health system-level time series plot of rates of hospital admission for preterm labour (PTL) per 100 preterm deliveries before and after implementation of fetal fibronectin (fFN) testing programmes (n = 98 hospital sites). Time point zero on the x axis (T0) corresponds to fFN test implementation. Negative and positive time points represent months prior to and post-fFN test implementation, respectively. Observations plotted at each time point represent a combined rate across the number of hospitals contributing data to each time point (number of hospitals shown on the secondary y axis). The absolute change in rates is estimated at time points 12, 24 and 36 months post-fFN implementation (T12, T24 and T36) (see Table 3 for actual values).

The estimated change in rates at 12, 24 and 36 months post-fFN implementation (T12, T24 and T36, respectively) for the health system-level aggregate multiple baseline analysis are provided in Table 3. The values indicate a statistically significant reduction of approximately one antepartum hospital admission for PTL per 100 preterm deliveries at each of the post-fFN evaluation time points (e.g. at 12 months post-intervention, the rate changed by −0.96; 95% CI, −1.02 to −0.90). In the absence of fFN test implementation, the expected rates of antepartum hospital admission for PTL at 12, 24 and 36 months post-fFN implementation would have been 12.87, 12.63 and 12.38 per 100 preterm deliveries, respectively, had the pre-implementation temporal trends prevailed. However, the predicted rates of PTL admissions at 12, 24 and 36 months post-fFN implementation were 11.92, 11.58 and 11.24 per 100 preterm deliveries, respectively.

Table 3. Health system-level results from multiple baseline analysis of rates of hospital admission for preterm labour (PTL) per 100 preterm deliveries before and after implementation of fetal fibronectin (fFN) testing programmes (n = 98 hospital sites)
Months after fFN test implementation Rates of hospital admission for preterm labour per 100 preterm births Difference between models after fFN test implementationc (95% CI) P
Predicteda (i.e. ‘without’ fFN test implementation)Estimatedb (i.e. ‘with’ fFN test implementation)
  1. a

    Estimated by extrapolating the pre-fFN regression line shown in Figure 1 into the post-fFN time period.

  2. b

    Estimated from the post-fFN regression line shown in Figure 1.

  3. c

    Difference between predicted and estimated rates at each post-fFN time point indicated, where a negative value indicates a reduction in the rate and a 95% confidence interval (CI) that does not include zero indicates P < 0.05.

12 months (T12)12.8711.92−0.96 (−1.02 to −0.90)0.04
24 months (T24)12.6311.58−1.06 (−1.18 to −0.94)0.003
36 months (T36)12.3811.24−1.15 (−1.32 to −0.98)<0.0001

Of the 98 hospital sites included in the health system-level multiple baseline analysis, 55 were eligible for the sensitivity analysis of individual sites by virtue of having a minimum of 12 time points pre- and post-fFN implementation. A further eight sites were excluded as a result of model convergence issues, leaving 47 individual hospitals for analysis. At 12 months following fFN implementation, 26 of the 47 individual hospitals (55%) demonstrated a statistically significant reduction in the rate of hospital admissions for PTL, whereas 10 of the 47 (21%) showed a significant increase, and results for the remaining 11 sites were inconclusive. At 24 months post-fFN implementation, 22 hospitals (of 36 with sufficient observations to be evaluated at this time point) had a statistically significant reduction and, at 36 months, 10 hospitals (of 18 with sufficient post-fFN implementation data points) showed a statistically significant reduction in the rate of hospital admissions for PTL. At each of the three post-fFN implementation evaluation time points, we observed a significant increase in the rate of hospital admissions for PTL in a small number of hospital sites. For instance, of the 10 hospitals with a significant increase in their rate of antepartum hospital admission for PTL at 12 months, three had a low birth volume (i.e. fewer than 500 births annually), resulting in very imprecise confidence intervals around the point estimate. The remaining seven, however, had sufficient birth volume and availability of time points for the analysis. The magnitude of the increase in the rate of PTL admissions at 12 months post-fFN implementation for these seven sites ranged from 1.1 per 100 preterm births to 8.5 per 100 preterm births (data not shown but available on request).

Discussion

Main findings

In our health system-level multiple baseline analysis, we estimated a statistically significant reduction in the rate of hospital admission for PTL following the introduction of fFN testing. The ability to distinguish women in true PTL from those who will ultimately deliver at term gestation is highly relevant to obstetrical care providers making time-sensitive clinical management decisions. In most cases, women with a negative fFN test can be monitored as outpatients, avoiding unnecessary interventions, hospitalisations or maternal transport to a hospital with a higher level of neonatal care. The latter point is particularly important in a jurisdiction such as Ontario, which has a vast geography of more than 1 million square kilometres. Women residing in sparsely populated and remote areas of the province require air or land transport to larger centres for obstetrical emergencies, such as preterm birth.

Although the magnitude of the reduction in PTL admissions estimated was small, a reduction of one per 100 preterm births would correspond to approximately 293 avoided admissions in the study time period following the introduction of fFN (i.e. based on 29 377 total preterm births that accrued in the study hospitals following fFN implementation). An in-depth examination of the economic impact of the fFN programme is beyond the scope of this article; however, it was a component of our broader programme evaluation. Even without factoring in avoided maternal transports (a comprehensive source of data on transports was not available), we estimated a cost savings associated with the implementation of the fFN programme in the province.[35] We speculate that any cost savings would only increase with the inclusion of transport costs into the economic models.

Interpretation

Despite different study designs, our results concur with other studies reporting a significant reduction in hospital admissions for PTL.[2, 15, 20, 22, 23] Most previous studies assessed the impact of fFN testing on healthcare utilisation using cohort study designs comparing pre- and post-intervention periods[2, 15, 22, 23, 36]; however, such comparisons are vulnerable to confounding from temporal trends and concurrent interventions.[28] We implemented a multiple baseline design, making use of the non-experimental variation in the timing of fFN programme implementation to estimate the effect of the intervention.[27, 28] Although we cannot definitively rule out all possible threats to the internal validity of our results (such as temporally concurrent policies or interventions, health system changes or changes in the characteristics of the obstetrical population that were associated with the study outcome), the staggered introduction of the fFN programme, as compared with introduction at a single point in time, provides additional protection against confounding from these factors.[27, 28]

Our study cannot discern whether healthcare providers managing women with PTL are already using the fFN test optimally (i.e. systematically using the test in all eligible women; using the test results to inform clinical management decisions regarding which women to admit, treat or transfer to a hospital with a higher level of neonatal care), or whether further health system benefits related to fFN testing could be achieved. The incorporation of new evidence into clinical practice is often slow[37] and influenced by individual-level factors (e.g. care provider awareness about a practice issue[38]; perceived credibility of new evidence[39-41]) and organisation- and health system-level factors (e.g. efficacy of an intervention; extent to which an intervention is appropriately provided and utilised by all those who could benefit from it[42]). In the qualitative component of our broader evaluation project, obstetrical care providers noted problems with the use of the fFN test outside of obstetrical settings, such as Emergency Departments, where women with symptoms of PTL may be triaged by non-obstetrical staff who are unaware of the need to collect the fFN swab before any other pelvic examination. Similar concerns were raised about current policies requiring a pelvic examination prior to air transport to a hospital with obstetrical services. As some remote settings are not equipped to screen for fFN prior to performing the pelvic examination, women arrive at hospital having had a recent pelvic examination, rendering them ineligible for fFN screening.[35]

Although a provincial implementation strategy was provided together with fFN programme funding in 2009, about one-half of the province's hospitals had already initiated their fFN programme prior to 2009 and were thus unable to make use of the strategy. The modest results from our naturalistic evaluation of the fFN programme may partly reflect the variable implementation strategies, particularly prior to 2009. A recent Cochrane review suggests that tailored implementation strategies are more effective at realising practice change than no intervention or simple dissemination of practice guidelines.[43] Additional strategies targeting hospital and health system-level barriers to fFN implementation may be needed to realise further health system benefits in Ontario. Ideally, future programmes should prospectively identify site-specific and system-wide barriers to change prior to programme implementation to permit the development of tailored implementation approaches.[43]

Strengths and limitations

The strengths of this study include the use of a robust quasi-experimental design and the evaluation of the impact of fFN on hospital admissions for PTL across a health system in a large jurisdiction. Several study limitations require discussion. Approximately one-half of all hospital sites in our study could not be included in individual hospital-level sensitivity analyses because of an insufficient number of time points either before or after the fFN test implementation date. Generally, there was a lack of post-fFN test implementation time points, which not only reduced the number of institutions in the individual-level analyses, but also introduced variability in estimates at the health system level. A future analysis could overcome this limitation by using the additional data that have accrued since this study was carried out. Maternal transfers recorded in the provincial hospitalisation database were examined as part of our broader programme evaluation; however, as only inter-hospital transfers were captured, there was no information on transfers occurring prior to hospital admission (e.g. those processed directly from Emergency Departments). We were unable to incorporate other data sources with information on ground and air transfers, and therefore could not comprehensively evaluate this indicator of health services utilisation related to PTL. We did not have record-level information on the uptake or appropriate use of the fFN test by healthcare providers, necessitating the use of a quasi-experimental approach that utilised the variation in the implementation date of each hospital's fFN programme. Although we believe that the implementation date is a good proxy for the administrative rule which ‘assigns’ women to the pre- or post-fFN time period, any bias of our estimates resulting from healthcare provider non-compliance with the fFN test protocol would be in the direction of the null value. Finally, although the study database was not specifically designed for perinatal research, the validity of the diagnostic code information related to obstetrical hospital admissions has been shown to be high.[44] We expect some error in our method of classifying antepartum admissions for PTL; however, as any misclassification would be expected to be non-differential according to the timing of fFN implementation (i.e. exposure), any bias would be towards the null value.

Conclusions

Using a robust, quasi-experimental design, we estimated a small but statistically significant absolute decrease in antepartum hospital admissions for PTL following implementation of fFN testing in Ontario hospitals. Further research is required to determine whether additional health system benefits of the fFN programme could be realised through the use of targeted implementation strategies.

Disclosure of interests

The authors have no conflicts of interest to declare.

Contribution to authorship

DBF designed the study, supervised the statistical analysis and drafted the article. AES supervised and designed the study, provided critical input on the analysis and revised the article for intellectual content. JMG designed the study, provided critical input on the analysis and revised the article for intellectual content. ASY executed the statistical analysis and revised the article for intellectual content. DC contributed to the design of the study, provided critical input on the analysis and revised the article for intellectual content. SID contributed to the interpretation of the study results and revised the article for intellectual content. SLP contributed to the design and analysis of the study and revised the article for intellectual content. WEP contributed to the design of the study and the interpretation of the study results and revised the article for intellectual content. MJ contributed to the design of the study and revised the article for intellectual content. PSB contributed to the design of the study and revised the article for intellectual content. MCW designed the study, provided critical input on the analysis and revised the article for intellectual content.

Details of ethics approval

This study was approved by The Ottawa Hospital Research Ethics Board (#2010-013).

Funding

This study received funding from ECHO: Improving Women's Health in Ontario, an agency of the Ontario Ministry of Health and Long-Term Care.

Acknowledgements

The authors would like to acknowledge the Ontario Ministry of Health and Long-Term Care and ECHO: Improving Women's Health in Ontario for providing the opportunity to conduct this evaluation, as well as the Canadian Institute for Health Information for providing us with data. We are grateful to the laboratory staff in Ontario hospitals who responded to our laboratory survey, to project coordinators Bdour Dandies and Carly Lang, and to Yanfang Guo who assisted us with the preparation of the manuscript. Dr Mark Walker is supported by a University of Ottawa Tier 1 Chair in Perinatal Epidemiology.

Disclaimer

The opinions expressed in the manuscript do not necessarily reflect the views of ECHO: Improving Women's Health in Ontario or the Ontario Ministry of Health and Long-Term Care. Parts of this study are based on data and information compiled and provided by the Canadian Institute for Health Information; however, the analyses, conclusions, opinions and statements expressed herein are those of the authors, and are not necessarily those of the Canadian Institute for Health Information.

Ancillary