Table 1 Summary of reported methods

Barrett et al. [42] (Singapore, 2017)

Four: accuracy, time of results, risk of miscarriage, and amount of information provided.

Pregnant women (n = 69) and healthcare professionals (n = 301).

Ten unlabelled choices of two alternatives generated using an unreported design with level balance, minimal overlap and orthogonality and incorporating a test for internal validity.^a

Conditional logit model

Beulen et al. [46] (the Netherlands, 2015)

Seven: minimal gestational age, time to test results, level of information, detection rate, false positive rate, miscarriage risk, and cost.

Pregnant women (n = 596) and healthcare professionals (n = 297).

Seventeen unlabelled choices of two alternatives and a dual non-response option generated using an unreported design with level balance, minimal overlap and orthogonality and incorporating a test for internal validity.

Conditional logit model

Bishop et al. [48] (UK, 2004)

Three: time of test, detection rate, and risk of miscarriage of a baby unaffected by Down’s Syndrome.

Pregnant women (n = 291) and healthcare professionals (n = 98).

Four unlabelled choices of two alternatives and a dual non-response option generated using an unreported design with unspecified methods and incorporating a test for internal validity.

Random effects probit model.

Boormans et al. [53] (the Netherlands, 2010)

Five: detection capacity, anxiety, waiting time, failure rate, and consequences of detected chromosomal abnormalities.

Pregnant women (n = 103).

Thirty-two labelled choices of three alternatives generated using a D-efficient design.

Conditional logit model.

Carroll et al. [52] (UK, 2013)

Four: detection rate, gestation, time to wait for results, and cost.

Women and partners (n = 103).

Sixteen unlabelled choices of two alternatives generated using a main effects design with maintaining orthogonality

Conditional logit and latent class models.

Chan et al. [43] (China, Hong Kong, 2009)

Three: level of the test information, waiting time for result availability, and cost of test.

Pregnant women (n = 300).

Eight labelled choices of three alternatives and an optout generated using a main effects design with level balance, minimal overlap and orthogonality and tested for non-traders always choosing a certain alternative or ‘no test.’

Conditional logit model with subgroup analysis.

Hill et al. [44] (Canada, Denmark, Iceland, Israel, Italy, the Netherlands, Portugal, Singapore, and UK, 2015)

Four: accuracy, time of test, risk of miscarriage, and provision of information about Down syndrome only, or Down syndrome and other conditions.

Pregnant women (n = 2666) and healthcare professionals (n = 1245).

Nine unlabelled choices of two alternatives and an optout generated using a main effects design with level balance, minimal overlap and orthogonality and incorporating a test for internal validity.

Conditional logit model.

Hill et al. [47] (UK, 2012)

Four: accuracy, time of test, risk of miscarriage, and provision of information about Down syndrome only, or Down syndrome and other conditions.

Pregnant women (n = 355) and healthcare professionals (n = 181).

Ten unlabelled choices of two alternatives generated using a main effects design with level balance, minimal overlap and orthogonality and incorporating a test for internal validity.

Conditional logit model.

Hill et al. [50] (UK, 2017)

Three: accuracy, time in pregnancy when the test result is received and risk of miscarriage.

Service users (carriers/affected with sickle-cell) (n = 67) and healthcare professionals (n = 62).

Eight unlabelled choices of two alternatives and an optout generated using a main effects design with level balance, minimal overlap and orthogonality and incorporating a test for internal validity. ^b

Conditional logit model with subgroup analysis.

Hill et al. [49] (UK, 2014)

Three: risk of miscarriage, accuracy, and time in pregnancy when the test result is received.

Adult cystic fibrosis patients (n = 92), carriers (n = 50) and healthcare professionals (n = 70).

Eight unlabelled choices of two alternatives and an optout generated using a main effects design with level balance, minimal overlap and orthogonality and incorporating a test for internal validity.

Conditional logit model with subgroup analysis.

Lewis et al. [41] (Australia, 2006)

Three: risk of miscarriage, accuracy and time in pregnancy when the test result is received.

Pregnant women (n = 322), midwives (n = 266) and obstetricians (n = 34).

None unlabelled choices of two alternatives and a dual non-response option generated using an unreported design with maintaining orthogonality and incorporating a test for internal validity.

Random effects probit model.

Lewis et al. [45] (UK, Australia 2006)

Three: risk of miscarriage, accuracy and time in pregnancy when the test result is received.

Midwives (n = 146 in Australia, 53 in UK) And obstetricians (n = 29 from Australia, 41 in UK) .

Twelve unlabelled choices of two alternatives generated using an unreported design with unspecified methods and incorporating a test for internal validity.

Random effects probit model.

Lund et al. [51] (Denmark, 2018)

Four: accuracy, time of test, risk of miscarriage, and provision of information about Down syndrome only, or Down syndrome and other rare conditions.

Women (n = 315) and their partners (n = 102) in addition to foetal medicine experts and sonographers (n = 57) and midwives not involved in screening (n = 48).

Ten unlabelled choices of two alternatives generated using a main effects design with level balance, minimal overlap and orthogonality and incorporating a test for internal validity. ^c

Conditional logit model with subgroup analysis.

Lynn et al. [54] (UK, 2015)

Four: health-care professional conducting the scan, detection rate for abnormal foetal growth, provision of non-medical information, and cost.

Pregnant women (n = 146)

Sixteen unlabelled choices of three alternatives and an optout generated using a main effects design with maintaining orthogonality and incorporating a test for internal validity.

Mixed logit model.

Miller et al. [36] (Canada, 2015)

Five: clinical benefits of improved health, earlier time to diagnosis, reproductive risk information, false-positive (FP) results, and overdiagnosed infants.

Members of the public (n = 1213).

Eight unlabelled choices of three alternatives and an optout generated using a D-efficient design.

Mixed logit and generalised multinomial logit models.

Ryan et al. [55] (UK, 2005)

Three: level of information, number of days’ wait for results, and cost to you.

Pregnant women (n = 40).

Eight labelled choices of two alternatives and an optout generated using an unreported design with unspecified methods.

Conditional logit model.

Tarini et al. [38] (USA, 2018)

Ten: number of babies diagnosed, chance of false positive, cost, likelihood of developing symptoms, seriousness of symptoms without treatments, age of symptoms and life expectancy without treatment, time to start of treatment, success of treatment, side effects of treatment, impact of diagnosis.

Members of the public (n = 502).

Four choice sets to select the ‘most important’ and ‘least important’ characteristic generated using an efficient experimental design.

Generalized estimating equation logit model.

Wright et al. [37] (UK, 2017)

Four: how information is provided, when information is provided, parents’ ability to make a decision, cost to the parents.

Current and future patients aged 18–45 (n = 702).

Ten unlabelled choices of three alternatives and an optout generated using a D-efficient design with Ngene and incorporating a test for internal validity.

Heteroskedastic conditional logit model.

Wright et al. [39] (UK, 2018)

Four: how information is provided, when information is provided, parents’ ability to make a decision, cost to the parents.

Midwives (n = 134).

Ten unlabelled choices of three alternatives and an optout generated using a D-efficient design with Ngene and incorporating a test for internal validity.

Conditional logit model.