Description of the condition

Preterm birth is a major public health problem worldwide. Preterm birth (before 37 completed weeks' gestation) occurs in 6% to 10% of births in high‐income countries and the proportion of pregnancies which end preterm, at 20 to 36 weeks, has not fallen in the 15 to 20 years for which Australian state and national data are available (Tracy 2007). This is true in virtually all high‐income countries (Ananth 2005; Buitendijk 2003; Joseph 2007; Tracy 2007). Although precise identification of preterm birth in lower‐income countries is poor because information on gestational age is rarely available, it is estimated to be at least as prevalent as in more prosperous countries (Kramer 1987).

Preterm birth is the major factor associated with perinatal mortality, admission to neonatal intensive care, severe morbidity in the first weeks of life, a prolonged hospital stay after birth, and re‐admission to hospital in the first year of life (Petrou 2003; Riley 2008). Surviving infants, especially those born before 32 weeks, have a substantially increased risk of chronic lung disease, and major and minor impairments (Anderson 2004; Costeloe 2006; Davis 2007; Doyle 2001a; Doyle 2001b; Doyle 2003; Doyle 2005; Doyle 2006; Ford 2000; Hack 2004; Larroque 2008; Wood 2005). There are considerable financial (Gilbert 2006; Petrou 2003; Petrou 2006) and emotional costs for the parents of preterm infants (Singer 1999).  

A number of strategies are used in an attempt to delay preterm birth. These include insertion of a ligature around the cervix to prevent it opening (cervical suture), drugs to dampen the contractions (tocolytics), antibiotics, and admission to hospital for bed rest. There is little evidence for the effectiveness of these interventions. Cochrane reviews have found no evidence in favour of bed rest in preventing preterm birth in singleton (Sosa 2015) and multiple (Crowther 2010) pregnancies; or any reduction in preterm or very preterm birth after treatment with magnesium sulphate (Crowther 2015), and there was an increase in perinatal mortality in the group treated with magnesium sulphate. Another Cochrane review found no reduction in preterm birth when prophylactic antibiotics were given to women in preterm labour with intact membranes (Flenady 2013). Social support for women with 'at‐risk' pregnancies did not reduce the proportion born preterm or low birthweight (Hodnett 2010). Reviews of progress on the primary and secondary prevention of preterm birth concluded that interventions tested in the past 15 years have not been found to be effective in well‐designed trials (Johnston 2001; Stevens‐Simon 1999).

In contrast to prevention, improved care of preterm infants before birth (transfer to tertiary care, antenatal steroids) (Crowther 2015), and afterwards (intensive care, surfactant) (Stevens 2007), has improved perinatal outcomes markedly over the last 15 years. These improvements mean that adverse outcomes of preterm birth are now concentrated in very preterm births (births at less than 32 weeks' gestation) (Buitendijk 2003; Slattery 2002).

A recent systematic review (Honest 2009) of the accuracy of screening tests to predict and reduce spontaneous preterm birth in symptomatic and asymptomatic women found the most cost‐effective method of reducing spontaneous preterm birth in asymptomatic women would be a universal intervention used in early pregnancy. Interventions suggested as appropriate but which require further investigation were periodontal care, fish oil, progesterone and antibiotics for asymptomatic bacteriuria. For symptomatic women in later pregnancy (those in threatened preterm labour), there is a need to define which among the currently available tests (e.g. cervical length, fibronectin, phIGBP‐1 and absence of fetal breathing movements) will provide the most efficacious outcomes when used alone or in combination. Other interventions such as calcium channel blockers and oxytocin antagonists require more evidence of efficacy and cost effectiveness (Honest 2009; Tsourapas 2009). Further, these authors recommend more methodological research into topics that could be considered in risk‐scoring systems.

Description of the intervention

Reviews of the epidemiology of preterm birth (Berkowitz 1993; Kramer 2001; Lumley 1993; Tucker 2004) have identified consistent associations with material and social adversity, multiple gestation, assisted conception, structural abnormalities of the uterus and cervix, serious medical, surgical, or gynaecological conditions in the mother, stressful life events, 'perceived' stress, poor psychological health, lack of family/social support, and tobacco and cocaine use. Many of the known risk factors are rare, and others are difficult to modify.

The elements that make up the various risk‐scoring systems will be likely to affect their accuracy. Indeed, a review of the area found widely divergent levels of accuracy (Honest 2004). For example, events that took place in a previous pregnancy may be most informative, making the prediction better for multigravid women. In evaluating risk‐scoring systems related to preterm birth, it is important to take account of several of their characteristics: their ease of use; their accuracy amongst different groups, e.g. multiparous and primiparous women; using them at earlier versus later gestation; whether they differentiate between spontaneous labours and births that are induced or by elective caesarean section; the factors they measure; the cut‐off points used to define preterm and very preterm birth; how well and appropriately the predictor variables are defined; the accuracy with which variables are measured; weightings given to elements of the score; whether 'dose' is considered, e.g. amount of smoking, duration of abnormal presentation; and their utility in varying healthcare environments (e.g. when no neonatal intensive care unit (NICU) is available).

There are a number of ethical issues involved in the decision to implement risk scoring: e.g. interventions may be used more frequently, with no evidence of better outcomes, and the possibility of harm to women and/or their babies (e.g. adverse effects of medications; deep venous thrombosis with rest in bed; time away from usual activities); resources may be inappropriately re‐allocated from areas of greater need; the woman may prefer not to disclose some matters included in the instrument, e.g. socio‐economic details and domestic violence; sensitive information needs to be recorded in the medical record; labelling may in fact be an additional risk factor; and women who refuse the interventions that flow from a high‐risk label can be regarded by their caregivers as misguided (Alexander 1989), with the risk that this may affect the (real or perceived) quality of their care. Recent research indicates that labelling women to be 'at risk' in pregnancy may have an adverse effect on their psychosocial well‐being, and may not accurately predict the outcome of interest (Stahl 2003).

Many of the factors included in the various risk‐scoring systems are probably markers for unknown causes of preterm birth, rather than being causal themselves. As a result, 'treating' the marker is unlikely to alter the outcome. Many of the markers are not amenable to change even if they were aetiological, e.g. maternal weight at the beginning of the pregnancy, age, past reproductive performance. The labelling that results from scoring may thus cause extra anxiety, with no possible counterbalancing potential for effective intervention.

How the intervention might work

The main purpose of screening for an increased risk of preterm birth is to enable high‐level antenatal care, aimed at prevention or delay of preterm birth in those identified as being at increased risk, or transfer in utero to a hospital with neonatal intensive care available. The focus of this review is on the capacity of screening to influence care and outcomes, rather than on the predictive power of the instruments per se. Accurate prediction of preterm birth is not in itself useful unless it enables interventions to be implemented that in some way reduce the mortality and morbidity associated with preterm birth (e.g. by delaying the birth; ensuring that birth occurs in a facility with a NICU). The primary outcomes of interest in this review are preterm birth and very preterm birth. The other neonatal outcomes (like mortality and morbidity) are included because delaying preterm birth may not be advantageous in some circumstances; i.e. preterm birth may remove some babies from a hostile environment (e.g. infection, poor placental function) and so reduce their risk of mortality or morbidity.

It is useful to know what proportion of women who go on to experience preterm labour are identified by risk scoring as being at high risk (sensitivity of the test), and what proportion of those who do not go on to experience preterm labour are identified by the test as being at low risk (specificity). Unless these measures of the screening system are quite high, the system will mis‐classify women, with the resulting lack of intervention for those wrongly classified as low risk and unnecessary interventions for those wrongly classified as high risk. Another measure of the value of a system like this is the proportion of those identified as being at high risk who will go on to experience the outcome in question (preterm birth in this case). This is known as the positive predictive power, and is useful information for the clinical management of individual women.

Why it is important to do this review

The results of this review have the potential to guide caregivers and women alike in making decisions about the management of pregnancies at higher than average risk of preterm birth.