Preferences for aspects of antenatal and newborn screening: a systematic review

In many countries there are population-wide screening programmes for unborn (antenatal) or newborn babies [1]. These screening programmes are used to detect whether the baby is at a higher risk of certain conditions and therefore lead to an earlier diagnosis and inform timely decisions about their care and/or treatment [2]. The extent and content of target conditions for antenatal screening varies between, and sometimes within, countries. For example screening for sickle cell disease, thalassaemia and blood tests for infectious diseases may or may not be included. The probability of Down’s syndrome, Edwards’ syndrome and Patau’s syndrome can also be determined through ultrasound scans combined with blood tests [3], with high risk women being offered an invasive diagnostic test (such as amniocentesis or chorionic villus sampling) which raises the risk of a miscarriage but provides a conclusive result [4, 5]. New non-invasive prenatal testing (NIPT) is a maternal blood test that relies on the detection of fetal cell free DNA that can be used for analysis [6], is also becoming available in some countries, although often only privately [7]. Newborn screening often includes a physical and hearing examination, and newborn bloodspot screening. Newborn bloodspot screening programmes (NBSPs) are recognised as being one of the most significant public health achievements in the developed world [8, 9].

There have been changes and proposed future changes to the configuration of both antenatal and newborn screening programmes in Europe and around the world. Many countries have expanded their programmes to include more conditions in recent years [8, 10]. The change has been driven by the development of new technologies such as mass spectrometry in NBSPs [11] and DNA sequencing in NIPT [12]. However, expanded screening and new technology for testing means the options available are more complex than ever and require difficult trade-offs for the parent(s) who must balance the benefits and risks of participation [13, 14]. There is also a need for parents to be provided with the relevant information about the methods, associated risks, and the available options if the test results are positive [15]. Furthermore, some tests can give an indicative but inconclusive result leaving the parents uncertain or facing further risky tests for a definitive result to confirm the presence of an abnormality [16].

Various studies have sought to understand people’s (mothers’, fathers’ and healthcare professionals’) views for newborn and antenatal screening using qualitative research methods [17‐20]. For example, one interview study of parents and professionals found they were concerned about information overload, insufficient information, how the information was provided and having autonomy in decision making [19]. Although able to illuminate factors of importance, qualitative studies are limited in their ability to determine the relative importance of different aspects of screening, the extent of heterogeneity within the population’s views or how changes to the screening programme may affect uptake.

A robust method for quantifying preferences, and the variation within, is a discrete choice experiment (DCE), and a recent 2002 [21] extension called best-worst scaling experiments (BWS), that can be embedded in a survey. In a DCE, respondents stated their preferred and/or least preferred option from a set of alternative goods or services described by the same attributes but differing in their amount (levels) [22]. From the choices made over a series of sets, it is possible to determine how the respondent balanced the different attributes and determine the relative impact each had on the probability of an option being selected. The data can also be analysed to understand how individuals balance the benefits and risks of participating in a healthcare programme and indirectly estimate the value they place on the attributes or the good or service as a whole.

Reviews of healthcare DCEs have found continued growth in studies using the method [23]. There has also been a rising interest in using the method to inform policy and regulatory decisions in the United States (US) and Europe [24‐27]. There have been systematic reviews of the DCE literature for cancer screening [28] and women’s preferences for birth place [29]. However, no reviews have summarised studies that have elicited preferences for newborn or antenatal screening programmes. This study aimed to identify published studies that have elicited preferences for antenatal and newborn screening programmes and provide an overview of key methods and findings.

A systematic review was conducted using a pre-defined protocol (available from the authors on request) based on standardised review methods [30]. For the purpose of this review, antenatal screening was defined as “population screening to identify people with a genetic risk, or a risk of having a child with a congenital or genetic disorder” through biochemical, genetic or ultrasound screening [31]. Newborn screening was defined as population screening of newborns to identify those at risk of congenital disorders to allow early intervention [32]. These definitions excluded childhood vaccination programmes.

A total of 19 studies relevant to antenatal or newborn screening programmes were identified and underwent detailed data extraction. The flow of studies through the review are shown in Fig. 1. The review process also identified a study eliciting women’s preferences for prenatal tests in Iceland [33], published in Icelandic and therefore not meeting our inclusion criteria. A qualitative study identifying women’s preferences for prenatal testing was also identified, although it did not contain an empirical DCE, the authors stated the identified themes will be used as attributes and levels in a future study [34]. Furthermore, another study elicited parents’ preferences for research on newborn dried bloodspots although this study used a rating rather than discrete choice exercise [35].

This review identified 19 studies that aimed to quantify preferences for aspects of antenatal and newborn screening. Accuracy of the test or screening programme was the most commonly included attribute. In contrast to advice in the risk communication literature, but in line with other healthcare DCEs [56], risk attributes (including risk of miscarriage) were most commonly presented as a percentage. In only three studies was ‘accuracy’ broken down into sensitivity or specificity. There is some evidence that individuals have different preferences for sensitivity and specificity and the balance of these aspects of antenatal and newborn screening require further investigation [57]. In addition, individuals find risk and percentages complex attributes to understand [58]. Studies comparing preferences frequently concluded that accuracy was relatively more important to healthcare professionals than women. However, it is unclear whether this is because of differences in each sample’s interpretation of this information. Future research could investigate if the heterogeneity is robust to different formats or more detailed explanations of the accuracy information.

Understanding if, and how, preferences for screening are affected by information and the communication of probabilities may have implications beyond completing a valuation study such as a DCE; for example, the preparation of invitations and screening leaflets or tailoring these to target subgroups. Several studies have explored how to facilitate informed decision making in antenatal and newborn screening [59, 60]; however, there appears to be no “one size fits all” approach [61]. It has been suggested that risk information should be tailored but there is a great challenge in adapting to the social, religious and cultural background of the user(s) or their attitude or knowledge [15]. In this review, Beulen et al. [46] and Barrett et al. [42] both suggested personalised counselling for women when providing information about or results from prenatal testing.

Other commonly occurring attributes were when the screening or testing occurred and time to results. Studies which included these attributes often found that they were not statistically significant or relatively unimportant in women’s decision-making. This could be because the advantages of early diagnosis and treatment may have been unclear, uncertain or unknown. Carroll et al. [52] found that in one of their preference classes waiting time was the most important attribute, however, this group accounted for less than 12% of the sample. For prenatal screening, the importance of early testing or rapid results for decisions about pregnancy termination may be irrelevant to some groups of women; future studies should consider heterogeneity in preferences around these aspects of screening.

Commissioners of screening programmes need to determine which screening programmes should be provided and by whom, when the tests should be offered, and how information should be given and consent obtained. No studies included in this review predicted demand or uptake for services or screening programmes. The validity of the included studies was also not explicitly discussed, despite uptake rates of screening being published in many countries [62, 63] and a desire to understand if stated preferences match those ‘revealed’ by the individual in the real-world [64]. Uptake rates can also be used for model parameterisation [65] particularly for new technologies where demand may be unknown. Similarly, only six studies included a cost attribute to calculate willingness-to-pay. Monetary valuation of aspects of screening or the programmes as a whole can be useful for comparing across subgroups but also in cost-benefit analysis to understand, for example, the net-benefit of expanded programmes [66].

In this review, only two studies specifically considered partners’ views [51, 52], although some studies with a broader sampling strategy of the public may have picked up these opinions indirectly [36‐38]. Partner preferences may be an important consideration when understanding choices about screening, and there is evidence to suggest partners may have different decision-making processes with different reactions to test information [67]. Future studies may wish to consider the preferences of partners and investigate if, and how, they are related to the mother’s choices.

Researchers seeking to contribute to the development of evidence reporting preferences for antenatal and newborn screening services should make use of advancements in experimental designs, for example using Bayesian or D-efficient approaches [68], to ensure attribute or level interactions of interest can be measured. Although the experimental design often included the option of ‘no screening’, the coefficient on the constant term was rarely reported in the results tables. This provides important information about women’s preferences in general for/again screening and includes information about the role of other attributes not included in the choice set. The studies included in the review also had large samples when compared with standard healthcare DCEs so there was potential to conduct sophisticated analyses and investigate preference heterogeneity either through subgroup analysis or other model specifications. Although not always considered in the studies included in this review, understanding the degree of heterogeneity in individuals’ preferences for healthcare could help decision-makers configure screening services to improve uptake and the utility of those who have participated.

The largest methodological issues related to 1) scale heterogeneity and 2) calculation of marginal rates of substitution. The pooling of data from multiple sources and assumption that the error term is homoscedastic, meant that the coefficients of the pooled models were uninterpretable. Methods which allow researchers to disentangle issues of scale include the heteroskedastic conditional logit or [69] generalised multinomial [70] models or scale-adjusted latent class analysis [71]. Alternatively, researchers can compare ratios of coefficients in marginal rates of substitution such as willingness-to-pay or willingness-to-wait. In this review, some studies reported point estimates of these ratios, presumably assuming the ratio of two significant coefficients could be interpreted as significant too. The Delta or Krinsky Robb methods [72] can be used to estimate the confidence intervals for a ratio, revealing to readers the degree of uncertainty in the trade-off calculations [73].

This review has shown that DCEs and BWS are currently being under used to understand preferences for antenatal and newborn screening programmes. Multiple studies concluded that accuracy was the most important aspect of testing to healthcare professionals whereas women placed more importance on the risks of participation. Further research is required to understand if these valuations are robust to different approaches to framing information. Limited reporting of the methodological component in some studies made interpretation of the findings challenging and in future studies, more sophisticated approaches to experimental design and/or the discrete choice modelling may improve confidence in the results. Researchers wanting to use DCEs for future applications in this area may want to compare estimated demand to actual participation rates as a test for study validity and generalisability. Furthermore, these estimates could be used by decision-makers to configure screening services to maximise uptake and, as a consequence, the health benefit to the population.