Economic evaluation of potential national childhood screening strategies for type 1 diabetes in Australia

Summary

Background

Type 1 diabetes (T1D) is an autoimmune condition affecting children. We aimed to investigate the costs and cost-effectiveness of potential national childhood screening strategies for T1D compared to no screening (usual care).

Methods

Screening costs were obtained from trial-based estimates. A Markov microsimulation model was developed to identify the most cost-effective childhood T1D screening strategy. The three screening strategies modelled were: Strategy 1) newborn genetic risk-stratification with bloodspot sampling, followed by autoantibody screening in at-risk children; Strategy 2) infant genetic risk-stratification using saliva sampling, followed by autoantibody screening in at-risk children; Strategy 3) population-wide autoantibody screening at two childhood ages. The model tracked 100,000 individuals from birth to 30 years of age. One-way and probabilistic sensitivity analyses were conducted.

Findings

Newborn bloodspot genetic risk-stratified screening (strategy 1) was the most cost-effective strategy. Incremental cost-effectiveness ratios (ICERs) were $50,682 per quality-adjusted life year (QALY) gained for strategy 1, $85,440 per QALY gained for strategy 2, and $133,285 per QALY gained for strategy 3. In the optimal strategy (strategy 1), the cost was $480,798 per screen-detected T1D and $12,183 per episode of diabetic ketoacidosis avoided. Results were sensitive to changes in time horizon, discount rates, and cost of the screening tests.

Interpretation

Of the three modelled T1D screening strategies, newborn bloodspot genetic risk-stratified screening was the most cost-effective. Varying cost inputs may change this hierarchy. Our economic evaluation will be useful for informing future T1D childhood screening policy in Australia and other high-income countries.

Funding

JDRF Australia.

Research in context.

Evidence before this study

Type 1 diabetes (T1D) is an autoimmune condition typically diagnosed in childhood. Early detection may help improve health outcomes. Pilot studies on population-wide T1D screening are underway in various countries, including Australia. Few studies to date have reported the cost-effectiveness of T1D screening and compared the cost-effectiveness of different screening strategies.

Added value of this study

We conducted a modelled economic evaluation of three potential national childhood screening strategies for T1D. Our modelled economic evaluation found that newborn bloodspot genetic risk-stratified screening is the most cost-effective strategy for T1D screening in Australia. The incremental cost-effectiveness ratio for newborn bloodspot screening compared to no screening was estimated at $50,682 per quality-adjusted life year (QALY) gained. The cost-effectiveness results were sensitive to changes in factors such as cost of screening tests. Infant saliva genetic risk-stratified screening and population-wide autoantibody screening were less cost-effective than newborn bloodspot screening.

Implications of all the available evidence

Our economic evaluation findings are important for informing design and implementation of national childhood T1D screening programs in Australia and other high-income countries.

Introduction

Type 1 diabetes (T1D) is an autoimmune condition with two-thirds of cases developing in young people under the age of 30 years.¹ In Australia, there are over 139,000 people living with T1D and T1D comprises approximately 10% of all cases of diabetes.^2,3 The economic burden of T1D to Australia, from a societal perspective, is estimated to be $AUD 2.9 billion per year.⁴

T1D is characterised by the progressive, immune-mediated destruction of beta cells in the pancreas. T1D is described in three stages, with stages 1 and 2 being presymptomatic, early-stage of the condition.⁵ Individuals are considered to have stage 1 T1D if they have two or more T1D-associated autoantibodies but remain normoglycaemic. The presence of autoantibodies indicates that the autoimmune condition has been initiated but sufficient functional beta-cell mass remains. In stage 2, individuals have T1D-associated autoantibodies and dysglycaemia due to progressive beta cell losses. In stage 3, individuals meet the clinical diagnostic thresholds for hyperglycaemia.⁶ Oral glucose tolerance test (OGTT) is considered gold standard for staging T1D in children. However, alternative approaches such as haemoglobin A1C (HbA1c) and random venous glucose are used where OGTT is not feasible or acceptable.⁷ The risk of progressing from stages 1 and 2 to stage 3 approaches 100% over the lifetime.^7,8

Without screening, individuals are typically diagnosed at stage 3 with symptoms of hyperglycaemia such as polyuria, polydipsia, fatigue and unintended weight loss. Diabetic ketoacidosis (DKA) is a serious and potentially life-threatening complication of T1D, and is currently present in approximately a third to a half of individuals with a new diagnosis of T1D in Australia.9, 10, 11, 12 Screening for T1D before the development of symptoms has the potential to significantly lower the risk of DKA in T1D,^13,14 with long-term health improvements.^11,15

Historically, screening for T1D has targeted relatives of individuals with T1D, given their 15-fold higher risk of the condition. However, approximately 90% of individuals with T1D do not have a first degree relative with T1D¹⁶ so a population-wide approach to screening is needed to capture the majority of future cases. Potential general population screening strategies include population-wide autoantibody screening to detect early stages of T1D, or initial polygenic risk screening followed by targeted autoantibody screening in those with an increased risk of developing T1D. In recent years, a number of general population screening programs for T1D have been studied in high-income countries.¹⁷

To date, no published study to date has examined the potential costs and cost-effectiveness of different options for national childhood screening programs for T1D in Australia. Concurrently, we, and others, are conducting an Australian pilot study to investigate the acceptability and feasibility of T1D screening strategies in the general population.¹⁸ The pilot started in 2022 and screened children across Australia using newborn bloodspot risk-stratified screening, infant saliva risk-stratified screening, or population-wide autoantibody screening. In the current paper we aim to describe the cost and cost-effectiveness of these three childhood screening strategies for T1D compared to no screening (usual care).

Methods

This study examined costs and conducted a modelled economic evaluation for three potential national childhood T1D screening strategies. A healthcare funder perspective was used, with an annual discount rate of 5% for both costs and outcomes.¹⁹ Costs are reported in 2024 Australian dollars (AUD), and Australian Institute of Health and Welfare health price deflators were used to convert costs to 2024 AUD where required.²⁰ The study followed the Consolidated Health Economic Evaluation Reporting Standards (CHEERS 2022) checklist to ensure transparent reporting (Supplemental Materials).

Strategies modelled

Three screening strategies were compared to no screening (incidental diagnosis)—see Fig. 1. The screening strategies mirror those included in the T1D screening feasibility and acceptability pilot underway in Australia:¹⁸

Fig. 1.

Overview and costs of no screening and screening strategies modelled. Abbreviation: T1D, type 1 diabetes.

Strategy 1—newborn bloodspot genetic risk-stratified screening: polygenic risk screening in newborns using heelprick dried bloodspots, followed by targeted autoantibody screening for increased-risk individuals at 1, 2, and 6 years of age.

Strategy 2—infant saliva risk-stratified screening: polygenic risk screening in infants (first year of life) using saliva swabs, followed by targeted autoantibody screening for increased-risk individuals at 1, 2, and 6 years of age.

Strategy 3—population-wide autoantibody screening: all children receive autoantibody screening using fingerprick dried bloodspots at 2 and 6 years of age (i.e. no polygenic risk stratification).

The rationale for strategy 1 is that T1D screening could be considered for inclusion within or alongside the existing Newborn Bloodspot Screening program in Australia, which has near-universal uptake.²¹ Strategy 2 is an alternative, non-invasive sampling method for infants that could be delivered by GPs or nurses during routine immunisation visits. It is envisaged that strategy 3 could be delivered during GP visits, or as part of in-school health checks.

Genetic risk stratification (strategies 1 and 2) in the pilot occurred via single nucleotide polymorphism (SNP) panel analysis of the bloodspot or saliva sample and calculation of the T1D Genetic Risk Score 2 (GRS2).²² Autoantibody analysis in the study tested for four islet autoantibodies—insulin (IAA), glutamic acid decarboxylase,²³ islet antigen 2 (IA-2A) and zinc transporter 8 (ZnT8A), using the antibody detection by agglutination-PCR (ADAP) assay.²⁴ Full details of the screening procedures are available in the pilot protocol paper.¹⁸

Model overview

A Markov microsimulation model was developed in TreeAge Pro 2024, R1²⁵ to estimate the long-term costs and outcomes of three childhood screening strategies for T1D compared to no screening. See Supplementary Materials, Figure S1 for the simplified tree structure. The model was developed by WC with input from experts in T1D (KB, MC), health economics (KH, SN), and epidemiology (NN). The model followed a hypothetical cohort of 100,000 individuals from birth to 30 years of age, in annual cycles. A 30-year time horizon was selected to examine short to medium impacts of T1D screening on costs and outcomes, before the majority of long-term complications develop (e.g. renal, cardiovascular, retinopathy). Half-cycle corrections were applied to the Markov model.

The health states modelled included no disease, stages 1 and 2 T1D as a combined health state, stage 3 T1D, and death—which is modelled as the absorbing state. See Supplementary Materials, Figure S2 for the state transition diagram representing the model. Tracker variables in TreeAge were used to record specific events in the microsimulation, such as the years since diagnosis of stage 3 T1D, number of DKA events for an individual, and numbers of individuals in each screening arm with false positive screening results. Tracker variables were used to model a higher risk of DKA at initial year of diagnosis of stage 3 T1D compared to subsequent years, and an increased mortality associated with one or more episodes of DKA compared to those with stage 3 T1D and no DKA. The model was calibrated to result in a consistent lifetime T1D prevalence across all arms (i.e. a similar number of individuals with stage 3 T1D across no screening and screening arms, at the end of 30 years).

Parameters and model assumptions

Key model parameters are presented in Table 1. The background prevalence of T1D was estimated to be 415 out of 100,000 based on Australian Institute of Health and Welfare data of T1D prevalence in young people up to 20 years of age.¹ As per published studies, the 30-year probability of progression from presymptomatic stages of T1D (stages 1 or 2) to stage 3 is assumed to 99%.⁸ Australian Bureau of Statistics age-specific background mortality rates for the Australian population were used.³⁰ The additional cost of per episode of DKA ($1000) and annual probability of DKA (41% initial year of T1D diagnosis, 5% subsequent years) were estimated from a related study on T1D hospitalisations and costs in Australia.³¹

Table 1.

Key model parameters.

Parameter	Base case	One-way sensitivity range (min to max)	PSA (distribution)b	Reference
Time horizon	30 years	15 years–45 years	Not included	N/A
Discount rate	5%	0%–7%	Not included	19
Autoantibody screening—test properties	Sensitivity 0.8 Specificity 0.95	Sensitivity 0.5–1 Specificity 0.8–1	Beta	26, 27, 28, 29
Transition probabilities
Annual background probability of stage 3 T1D	0.00021 per year	Not includeda	Not includeda	1
Annual probability of stage 1&2 T1D, to stage 3	0.14 per year	+/−30%	Beta	8
Annual probability of death	Time-dependent, by age	Not includeda	Not includeda	30
Annual probability of DKA	Time-dependent, by years since T1D diagnosis: Initial year: 0.41 per year Subsequent years: 0.05 per year	+/−30%	PERT	9, 10, 11, 12,31
Costs
Stage 3 T1D annual healthcare costs	$6500 per year	+/−30%	Gamma	4
Stage 3 T1D hospitalisation at diagnosis (no screening)	$4500 once off	+/−30%	Gamma	31
Stage 1 & 2 T1D annual healthcare cost (with screening)	$975 per year (estimated 15% of T1D stage 3)	+/−30%	Gamma	4,32
DKA	$1000 per DKA episode	+/−30%	Gamma	31
Death	$20,000 once off	+/−30%	Gamma	33
Screening and follow-up costs	$41 newborn bloodspot genetic risk screen, per test $65 infant saliva genetic screen, per test $40 autoantibody screen, per test See Fig. 1 for detailed costs	+/−30%	Gamma	Pilot study18
Utilities
No stage 3 T1D	1	Not included	Not included	Assumed
Stage 3 T1D	0.9	0.85–0.95	Beta	34
DKA	−0.05 per DKA episode	−0.1 to 0 (no disutility)	Beta	35
Risk ratios
Probability of DKA with screening (vs no screening)	0.2	0.1–0.3	PERT	13,14
Probability of death with screening (vs no screening)	1.001	1 (no increase)–1.01	PERT	36
Probability of stage 3 T1D in genetically high risk individuals (vs low risk)	10.1	Not includeda	Not includeda	22

Abbreviations: DKA, diabetes ketoacidosis; PERT, Program Evaluation and Review Technique; PSA, probabilistic sensitivity analysis; T1D, type 1 diabetes.

^aThe ICER outputs of the model are highly sensitive to variations in number of cases of T1D across no screening and screening arms. Consequently, several parameters were calibrated in the base case or kept constant throughout the sensitivity analysis to ensure stability in model results.

^bIn the PSA, beta distributions were used for probabilities and utilities, gamma distributions for costs, and PERT distributions for other parameters such as risk ratios, and time-dependant probabilities.

In strategies 1 and 2, the top 10% of children in the general population are deemed to be at increased genetic risk, whereas the remaining 90% of children are at low risk.²² For the risk-stratified individuals, the increased-risk children (top 10% of population) are assumed to have a 10 times greater annual probability for developing T1D compared to the general population.²² As the Australian pilot is not evaluating screening effectiveness, intervention effects were estimated from existing literature.^13,14 Across all screening strategies, we assumed a 0.2 risk ratio (RR) for annual probability of DKA for children with stage 3 T1D, vs those who did not receive screening.

For the base case, test performance of the bloodspot autoantibody screening was assumed to have 80% sensitivity and 95% specificity for a correct diagnosis of T1D (2 or more autoantibodies present).²⁶ The ADAP test used for autoantibody screening in the Australian pilot has >95% sensitivity and specificity compared to gold standard assays for detecting 2 or more autoantibodies.³⁷ However, reported test performance of autoantibody screening for the presence of T1D varies widely across studies, complicated by differences in the combination of autoantibodies studied, cut-off levels for presence of autoantibodies, definition of T1D used, age at which screening was performed, and difficulties in following sufficient number of individuals over time to confirm the development of T1D.26, 27, 28, 29 For example, Knip et al. reports a 61% sensitivity (95% CI 36–83)²⁹ whilst Ghalwash et al. report a 82% sensitivity (95% CI 79–86).²⁶ A wide range of sensitivities and specificities were modelled in the sensitivity analysis to investigate the impact of test performance uncertainty.

Costs

Screening costs were estimated from pilot study costs and clinical estimates (Fig. 1). These costs included the cost of the screening tests and postage, confirmatory pathology tests, and specialist consultations to discuss positive results. The initial cost of genetic risk screening was $41 in strategy 1 (newborn bloodspot risk-stratified screening) and $73 in strategy 2 (infant saliva risk-stratified screening), which includes both the test kits and pathology testing of the screening sample. Subsequently, those at high genetic risk were offered autoantibody screening at a cost of $40 per dried bloodspot pathology test. The same autoantibody screening test was employed for population-wide autoantibody screening (without genetic risk stratification, strategy 3) at the same cost of $40 for each dried bloodspot pathology test. For all screening strategies, if autoantibody screening is positive, a follow-up confirmatory autoantibody test ($100 per venous pathology test) and an initial specialist consult (MBS item 110, $175 per consult) cost are incurred in the model.

Only ongoing running costs of the screening strategies were considered—i.e. set-up costs such as those associated with infrastructure, training, staffing or community awareness campaigns for a potential national screening program were not included in the cost-effectiveness analysis. We estimated routine annual healthcare costs for stage 3 T1D ($6500 per person, per year) from an Australian report commissioned by JDRF Australia, which took a societal perspective.⁴ As described in the JDRF report, we assumed that 25% of the total annual cost was attributable to direct healthcare costs. As per New South Wales (NSW) data, an additional once-off cost of $4500 is assumed in the no screening group at time of diagnosis of stage 3 T1D to account for the hospitalisation costs at diagnosis for this group.³¹ In the screening group, we applied an annual healthcare cost of $975 for stage 1 and 2 T1D (15% of the annual healthcare cost for stage 3 T1D)³² to account for a model of care that provides education and monitoring for progression to stage 3 T1D in an outpatient setting. Consequently, we assumed no routine hospitalisations at initial diagnosis of stage 3 T1D, given this prior education. A lower annual healthcare cost is applied in stage 1 and 2 compared to stage 3 T1D as no pharmacological management or intensive glycaemic monitoring is required at this stage. For those with stage 3 T1D in any group, in addition to the ongoing healthcare costs, a cost for DKA is incurred in the model at each DKA event ($1000 estimate based on NSW data).³¹

Outcomes

The primary outcome of interest was the incremental cost-effectiveness ratio (ICER), expressed as the cost per quality-adjusted life year (QALY) gained. The ICER was calculated by dividing the difference in total cost between the screening strategy and usual care group, by differences in QALY outcomes between the two groups. A utility weight of 0.9 was used for people living with stage 3 T1D, in line with Xie et al., a recently published 2024 meta-analysis of T1D utilities.³⁴ A disutility weighting of 0.05 was applied per DKA episode, as per a previous T1D screening modelling study by McQueen et al.³⁵ Secondary outcomes calculated included cost per additional case of T1D detected through screening, and cost per episode of DKA avoided.

Sensitivity analysis

One-way sensitivity analysis and probabilistic sensitivity analysis were conducted around most key parameters to investigate the effect of uncertainty on the incremental cost-effectiveness ratio (ICER; Table 1). For the PSA, 10,000 simulations were conducted—gamma distributions used for costs, beta distributions for proportions and utilities, and PERT (Program Evaluation and Review Technique) distributions for other parameters such as RR, and time-dependant probabilities.^38,39 PERT distributions are an alternative to triangular distributions, and incorporate expert estimates for the minimum, most likely, and maximum values of a parameter. Additional scenario analyses were conducted to explore the effect of changes to the modelled time horizon, and alternative screening intervals.

Ethics approval

The T1D pilot study had ethics approval through Sydney Children's Hospitals Network Human Research Ethics Committee (2022/ETH00537). Individual patient consent was obtained for this study.

Role of the funding source

The funder of the study had no role in study design, data collection, data analysis, data interpretation, writing of the report, or the decision to submit for publication.

Results

Number of individuals with type 1 diabetes and DKA

A microsimulation was conducted for 100,000 individuals from birth to 30 years. By the end of 30 years, the prevalence of stage 3 T1D within the modelled strategy was approximately 600 per 100,000 across all groups (no screening and screening groups; Table S1). The number of screen-identified stage 1 and 2 T1D cases was slightly lower in the risk-stratified strategies (strategies 1 and 2, n = 45 each) compared to the population-wide autoantibody group (strategy 3, n = 51). In the simulation, all screen-identified stage 1 and 2 T1D cases progressed to stage 3 T1D by the end of the 30-year time horizon. The number of DKA episodes was highest in the no screening group (n = 467) and substantially lower in the risk-stratified groups (n = 84 for strategy 1 and 2) and the population-wide autoantibody group (n = 91).

For the 100,000 individuals modelled in each screening strategy, the risk-stratified strategies had a relatively low number of total autoantibody screens (strategies 1 and 2, n = 30,297) compared to the population-wide autoantibody screening strategy, where all children were screened twice during childhood (strategy 3, n = 199,577). Therefore, assuming the same sensitivity and specificity for autoantibody screening across all strategies, the total number of autoantibody screening false positives was much higher in strategy 3 (n = 9969) compared to strategy 1 or 2 (n = 1511 in each). The number of false negatives were similar across the three strategies (n = 11 in strategies 1 and 2, n = 9 in strategy 3).

Costs and outcomes

The costs, outcomes, and ICER results for the three screening strategies compared against no screening are shown in Table 2 and Fig. 2. At a whole of population level, including those with and without T1D, average costs over 30 years were lowest in the no screening group ($238 per person), followed by strategy 1 ($286 per person), strategy 2 ($318 per person), and strategy 3 ($322 per person). Compared to no screening, newborn bloodspot risk-stratified screening was the optimal strategy with the lowest ICER (strategy 1, $51,782 per QALY gained), followed by infant saliva risk-stratified screening (strategy 2, $86,540 per QALY gained) and population-wide autoantibody screening (strategy 3, $111,776 per QALY gained).

Table 2.

Incremental cost-effectiveness ratio results, all screening strategies vs no screening.

	No screening	Strategy 1—newborn bloodspot risk-stratified screen	Strategy 2—newborn infant risk-stratified screen	Strategy 3—population-wide autoantibody screen
Costs
Average cost per person	$238	$286	$318	$322
Incremental (vs no screening)	–	$48	$80	$84
Outcome—QALYs
Average QALYs per person	15.0803	15.0813	15.0813	15.0811
Incremental (vs no screening)	–	0.0009	0.0009	0.0008
ICER ($ per QALY gained)	–	51,782	86,540	111,776
Outcome—cases of screen identified T1D
Cases of screen identified T1D	0.0000	0.0005	0.0005	0.0005
Incremental (vs no screening)	–	0.0005	0.0005	0.0005
ICER ($ per additional screen identified T1D)	–	634,700	715,365	622,774
Outcome—DKA episodes
DKA episodes	0.0047	0.0008	0.0008	0.0009
Incremental (vs no screening)	–	0.0038	0.0038	0.0038
ICER ($ per DKA episode avoided)	–	12,447	21,925	21,190

Results are per person, over a 30-year time horizon. Rounded to 4 decimal places for outcomes and whole dollars for costs and ICERs.

Abbreviations: DKA, diabetes ketoacidosis; ICER, incremental cost-effectiveness ratio; QALYs, quality adjusted life years; T1D, type 1 diabetes.

Fig. 2.

Incremental cost-effectiveness results—QALYs as outcome. Abbreviations: QALYs, quality adjusted life years; WTP, willingness-to-pay (threshold).

In the optimal screening strategy (strategy 1), the cost per additional case of screen-detected T1D was $634,700, and the cost per DKA episode avoided was $12,447. Note that due to T1D being relatively uncommon on a population-level, the microsimulation ICER results in terms of QALYs were subject to wide changes due to random variability in the modelled number of individuals with T1D in each group and their disease trajectories over the time horizon. The variations in ICER results in terms of QALYs persisted, even when a larger number of simulations were run (e.g. 1,000,000 individuals for microsimulation). The ICER results in terms of screen-detected T1D and DKA episodes avoided was reasonably stable across simulations.

Sensitivity analysis results

One-way sensitivity analyses were conducted for the optimal screening scenario (i.e. strategy 1). Time horizon, discount rate, and parameters such as the cost of the risk-stratified risk screening test were influential on the ICER (Figure S3). Due to the gradual development of T1D and DKA over the time horizon, reducing the discount rate or increasing the time horizon improves the cost-effectiveness of strategy 1 in terms of QALY outcomes. For example, a longer time horizon of 45 years yields an ICER of $50,092 per QALY gained compared to $51,782 per QALY gained with the base case of 30 years. Conversely, a shorter time horizon and an increased discount rate improves the cost-effectiveness of strategy 1 in terms of cost per screen-detected case of T1D (Table S2).

In terms of test performance, improved specificity of the autoantibody test in strategy 1 had varying effects on cost-effectiveness with an 80% specificity (lower limit) producing an increased ICER of $64,390 per QALY gained for strategy 1, and a 100% specificity (higher limit) producing an increased ICER of $62,310 per QALY gained for strategy 1. Changes to test sensitivity of the autoantibody test between 50% sensitivity (lower limit) and 100% sensitivity (higher limit) had minimal impact on ICER results for strategy 1.

Sensitivity analysis showed that both risk-stratified screening options for T1D (strategy 1 and 2) would be considered cost-effective at a $50,000 per QALY gained threshold, if the initial cost of newborn bloodspot risk-stratified screening was <$39 per person screened (base case strategy 1 $41, base case strategy 2 $65) or the subsequent autoantibody screening for increased-risk individuals was <$34 per test (base case $40 in both strategy 1 and 2). The population-wide autoantibody screening strategy would be considered cost-effective at the same $50,000 per QALY gained threshold if the autoantibody screening cost was <$12 per test (base case $40 in strategy 3). See Figures S4–S6 for results of the threshold analysis.

Finally, an alternative scenario was modelled where the final autoantibody screen occurred at 10 years of age rather than 6 years of age for individuals at increased genetic risk in the optimal screening scenario (strategy 1). Increasing the age for the final autoantibody screen resulted in an increased ICER of $78,364 per QALY gained.

PSA conducted for the optimal screening scenario (strategy 1) compared to no screening showed wide variations in ICER results between the 1000 simulations (Figure S7). Figure S8 shows the PSA acceptability curve for all screening strategies compared to no screening. At a willingness-to-pay (WTP) threshold of $50,000 per QALY gained, the likelihood of favouring no screening is 72%, followed by strategy 1 (20%), strategy 3 (8%), and strategy 2 (0%). The PSA results indicates that small variations in costs or outcomes across simulations can influence which strategy is deemed most cost-effective.

Discussion

We present the first study of the potential costs and cost-effectiveness of T1D national screening strategies in Australia. Internationally, this economic evaluation is also the only study to date to consider the cost-effectiveness of risk-stratified strategy for a childhood T1D national screening program. Our results demonstrate that of the three modelled screening strategies, newborn bloodspot risk-stratified screening was the most cost-effective screening strategy (strategy 1 vs no screening, ICER of $51,782 per QALY gained). Australia does not have an explicit WTP threshold but a WTP threshold of <$50,000 per QALY gained is often quoted in the literature.⁴⁰ In the modelled analysis, all screening strategies were above this threshold. However, either risk-stratified strategies (strategy 1 and 2) may be considered cost-effective at this threshold if lower test costs could be negotiated in the context of a national screening program (initial genetic risk test costs <$39 per test, or <$34 per test for follow-up autoantibody screening tests). Reduced costs may also be realised if the initial polygenic risk test costs for T1D risk stratification are covered by the potential future use of genomics in newborn bloodspot screening.⁴¹ Current universal newborn screening programs such as the universal newborn bloodspot screening, and newborn hearing screening programs are generally considered to be cost-effective.^42,43 However, the decision to add or remove conditions from the existing newborn bloodspot screening program remains complex. The complexity of these ethical, equity, and economic considerations are amplified in the era of genomics—and remain an area of active research in Australia.^41,44

In the optimal screening strategy (strategy 1), the cost per screen-detected T1D was $634,700 and cost per DKA episode avoided was $12,447. The high cost per screen-detected T1D reflects the relatively small number of additional screen-detected T1D cases compared to usual care (strategy 1, n = 45 for 100,000 simulated individuals). The population-wide autoantibody screening strategy (strategy 3) captured marginally more cases of T1D (n = 6 additional cases) and avoided several additional cases of DKA compared to risk-stratified strategies (n = 7 additional cases) but had the disadvantage of requiring a larger number of individuals to be screened for autoantibodies at two timepoints, and a high number of false positive results (n = 9969 for 100,000 simulated individuals).

A number of studies on T1D childhood screening in the general population have commenced in recent years,¹⁷ but few economic evaluation studies of these screening strategies are available for comparison. All economic evaluations to date have described population-wide autoantibody screening only, and do not have genetic risk-stratified screening strategies for comparison. Our cost per autoantibody screening test was $40, comparable to an approximate range of $20AUD⁴⁵ to $200AUD³⁵ in previously published studies. In terms of cost-effectiveness, McQueen et al. describes the T1D Autoimmunity Screening for Kids (ASK) program to be cost saving if the percentage reduction in DKA events is 40% or more (screening vs no screening).³⁵ One reason for better cost-effectiveness in ASK compared to the screening strategies in our study is that their Markov cohort model incorporates future diabetes-related complications (e.g. cardiovascular, renal) over a lifetime horizon. Our model focuses on short to medium-term outcomes up to 30 years of age due to limited primary data on longer-term T1D morbidity and mortality outcomes. Karl et al. describes the cost-effectiveness of the Fr1da T1D screening program in Germany. Comparisons between our study and Fr1da further demonstrates that cost-effectiveness is strongly influenced by whether short or long-term costs and outcomes are considered. Cost-effectiveness of screening in Fr1da was approximately AUD $11,600 per child diagnosed (our study, AUD $634,700 per screen-detected T1D).⁴⁵ A key difference between the two studies is that we included the substantial yearly ongoing healthcare cost of T1D in our modelling, whereas costs in Fr1da are only described up to the point of diagnosis.

There are several limitations to note. Firstly, our study reports costs and cost-effectiveness in an Australian context, which may not be generalisable to other healthcare settings. However, we have transparently presented our modelling methods and assumptions to allow adaptation to other healthcare contexts, using local parameters (e.g. epidemiological data, cost inputs). Furthermore, costs of a national screening program may vary from the costs described here, depending on the actual model of care implemented for screening and follow-up. In addition, our analysis did not include set-up or broader infrastructure costs associated with establishing a national screening program. Secondly, the feasibility and acceptability pilot study on which this economic evaluation is based was not designed to investigate differences in effectiveness or long-term outcomes of screening between screened and non-screened populations.¹⁸ Although there are some existing studies outlining the potential effectiveness of screening in reducing DKAs,^13,14 more robust evidence on effectiveness of screening is needed to inform our economic evaluation modelling. Thirdly, we have considered a simplified model in which uptake does not differ between the three strategies due to a lack of data on differences in screening uptake and follow-up adherence between strategies. Fourthly, the potential harms of screening, such as disutility related to psychological distress from receiving positive autoantibody results, were not incorporated into this model.^14,46 Separate studies addressing this important issue are underway. Conversely, the reverse may also apply where screening with early diagnosis can result in reduced psychological impacts from a smoother transition to T1D care.⁷ There is limited research available on the qualitative impacts of screening and more work is needed to understand what is appropriate to incorporate into future models in terms of potential impacts of screening on young people, their parents and caregivers.

We considered a 30-year time horizon in the modelled economic evaluation and did not include longer-term effectiveness outcomes of T1D screening. There are several challenges in incorporating longer-term effectiveness outcomes. Although Australian researchers have developed comprehensive epidemiological models for T1D, these models are based on data from international cohorts rather than from individuals with T1D in Australia.^47,48 There have also been major changes in the management of T1D in recent years, in particular the use of medical technologies, such as continuous glucose monitors and insulin pumps, which are coupled with reductions in complications.⁴⁹ Consequently, previous long-term predictions may no longer be accurate for individuals in a future T1D screening program. When suitable data are available in the future, our T1D economic evaluation model could be expanded to incorporate longer term T1D-related complications through adapting T1D complication models for an Australian population.

In conclusion, we conducted an economic evaluation of three potential T1D national screening strategies using a Markov microsimulation model. The cost-effectiveness of three potential T1D screening strategies in Australia ranged from $51,782 to $111,776 per QALY gained. Among the strategies evaluated, newborn bloodspot risk-stratified screening was most cost-effective in terms of cost per QALY gained, cost per screen-detected T1D, and cost per DKA episode avoided. Risk-stratified strategies offer the benefit of screening fewer individuals and yield fewer false positive results. This economic evaluation is useful for informing future policy decisions regarding national childhood screening programs for T1D in Australia and other high-income countries.

Contributors

KJB, KH, NN conceived the study and all authors contributed to the study design. WC developed the economic model, conducted the analysis, and wrote the initial draft. All authors contributed to model structure and inputs, interpretation of findings, editing the manuscript, and approved the final version of the manuscript. WC and KJB are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Data sharing statement

All data supporting the findings of this study are available within the paper and its Supplementary Information.

Declaration of interests

KJB and MEC have received consulting fees from Sanofi. KH is a member of the Pharmaceutical Benefits Advisory Committee (PBAC) and SN is a member of the Medical Services Advisory Committee (MSAC). SN and KH receive sitting fees from the Australian Commonwealth Department of Health for their roles on PBAC and MSAC. All other authors have no conflict of interest to disclose.