## INTRODUCTION

In the United States, serious but treatable inherited metabolic diseases are detected through state-mandated newborn screening protocols.1 However, mitochondrial disorders (MitD) may not be captured. When a diagnosis is undetermined, newborns undergo a diagnostic odyssey involving multiple tests.2 During this period, quick diagnosis is critical for patient stabilization or palliative care. Diagnosis turnaround time and accuracy also have implications for inpatient health-care resource utilization due to resource-intensive medical management. Infants with severe MitD are often admitted to the neonatal intensive care unit (NICU), incurring significant daily costs as their illness remains undiagnosed.

MitD refers to a spectrum of genetic disorders that occurs in a variety of patients, ranging from immediate onset in newborns, to later onset in adults.3 Newborns with suspected MitD tend to have a worse prognosis and shorter life expectancy than those diagnosed later in life, often presenting with a combination of symptoms ranging from seizures to cardiomyopathies and persistent metabolic disturbances.4,5 However, the hallmark of MitD, left untreated, is steady progression and increased involvement of multiple organ systems.5 Due to clinical and genetic heterogeneity, MitD clinical presentation can overlap with other diseases, leading to longer NICU length-of-stay and complicating diagnosis.4,5 Rapid diagnosis in this context is critical for making informed treatment decisions.

## RESULTS

### Cost–utility

In the base case scenario, total, nondiscounted expected life-years per patient is 2.10 years for singleton and trio eES arms, 2.03 years for the TC arm under the singleton scenario, and 2.12 years under the trio ES TC arm, consistent with life expectancies seen in patients with severe MitD observed in real world settings.5,16 Total discounted costs for the singleton eES arm is $288,308 and$330,249 for the TC arm (Table 2). Discounted costs for trio eES and the trio TC arm is $282,281 and$334,170, respectively. Total discounted expected QALYs for each arm is approximately 1.1 (Table 2). These values imply that singleton and trio eES dominate current TC practices for diagnosing patients with severe MitD because they provide equivalent QALYs at a lower cost. iNMB of singleton eES compared to singleton TC at a WTP of $200,000 and$50,000 were also calculated to be $45,387 and$42,802, respectively (Table 2). We observed iNMBs of $49,826 and$48,432 at WTPs of $200,000 and$50,000 among trio eES relative to singleton TC.

MitD patient NICU outcomes, predischarge, are better for patients in the eES singleton and trio arms. More patients are correctly diagnosed and more survive the NICU relative to singleton TC (Table 2). Total NICU cost per patient, cost per correct diagnosis, and cost per NICU survivor as observed from the health-care perspective are less costly for both eES arms relative to both TC arms. Trio eES has the lowest total NICU cost difference relative to singleton TC, lowest cost per correct diagnosis, and lowest cost per NICU survivor compared to each of the other arms at -$49,849, -$112,394, and -$95,158 respectively (Table 2). These analyses are primarily driven by more correct diagnoses, greater NICU survival, and lower costs for eES arms relative to TC arms. Earlier diagnosis by singleton and trio eES contributes to shorter NICU length-of-stay and thus fewer health care–associated expenses. ### One-way and probabilistic sensitivity analyses The parameters with largest effects on ICER and iNMB at the$200,000 and $50,000 WTP thresholds were the likelihoods of true positive diagnoses in both arms (TC and eES), length-of-stay in the NICU regardless of true/false positive/negative diagnostic rates, discharge or death likelihood following an initial unconfirmed diagnosis, and unconfirmed diagnosis likelihood following either eES or TC (Fig. 2). Key results were robust to univariate sensitivity analysis. Probabilistic sensitivity analysis confirms base case results at each WTP from US$0 to $1,000,000. At a$0 WTP, trio eES is the most cost-effective comparator in 58% of simulations. However, singleton eES becomes the best option as WTP threshold increases beyond $500,000. Figure 3 and Appendix Fig. 2 exhibit Monte Carlo simulations projected to a cost-effectiveness plane and the cost-effectiveness acceptability curve. Finally, EVPI was calculated according to our PSA results to determine the theoretical amount a decision maker would be willing to pay for perfect information on diagnostic outcome for three of our comparators relative to singleton TC. Singleton eES has an EVPI of$85,496 and $86,458 at a WTP of$50,000 and $200,000, trio eES has an EVPI of$69,207 and $63,107 respectively, and trio TC has an EVPI of$97,042 and $91,134, relative to singleton TC. ### Scenario analyses Relative to singleton TC, no ES was more costly, although with a positive iNMB of$6,778 and $1,135 at WTP thresholds of$200,000 and $50,000, relative to singleton TC. Total NICU cost, cost per correct diagnosis, and cost per NICU survival were also less than singleton TC (Appendix Table 5). No ES has a slightly lower probability of being a cost-effective alternative at a varying WTP threshold than singleton eES, and a higher probability of being cost-effective relative to singleton and trio TC (Fig. 3). Under the hypothetical curative orphan drug threshold analysis, no ES has the largest iNMB relative to singleton TC, under the base conditions that expected life expectancy due to therapy increases (Appendix Figure 3a). However, as total orphan treatment cost decreases (Appendix Figure 3b–f), eES yields the greatest iNMB. ## DISCUSSION ES is an expensive genetic diagnostic with clinical promise for diagnosing newborns with rare diseases.14 This analysis evaluated the economic value of eES for the diagnosis of neonates with severe MitD. By robustly analyzing this patient population and intervention from the societal perspective, eES is shown to have a positive iNMB relative to TC, while also dominating the current TC by reducing a major driver of ill newborn costs, i.e., NICU length-of-stay. Given the results presented here, one could assume that for other genetic diseases that are difficult to diagnose through current clinical means, eES represents a cost-effective and cost-minimizing way to quickly diagnose and optimally manage patients with severe disorders. Furthermore, the breadth of identifiable illnesses through ES suggests the analyses here are conservative since other illnesses may be diagnosed in these newborns at nearly zero marginal costs. This is particularly true if parents consent to testing of the 59 genes on the American College of Medical Genetics and Genomics–established list of incidental findings.30 These include BRCA1/2, other cancer genes, cardiomyopathy genes, and more. There are several limitations to this analysis that may be explored through threshold and scenario analyses. eES and length-of-stay costs have a large effect on this model. Establishing an appropriate neonatal cost-effectiveness threshold can help inform clinicians on the timeliness of incorporating eES into the care continuum. Other threshold analyses should also be completed regarding the likelihood of true positive and true negative diagnoses. As ES accuracy and testing precision improves, the value of eES should also improve. Although useful, PHIS has limitations. Since severe MitDs are rare and deadly, newborns in community hospitals may die prior to diagnosis. Since PHIS data are derived solely from children’s hospitals, similar patients from community hospitals are excluded, potentially underestimating MitD morbidity and mortality, which may bias the observed effect of implementing a novel diagnostic in the patient care continuum (Appendix Note 4). Like other decision-analytic models, this analysis had additional limitations that may bias or impact these results. MitD is rare and often underdiagnosed. Data pertaining to longer-term outcomes and health-care resource allocation are scarce. CP was used as a proxy for generating costs and health state utilities to populate this model. Although similar, CP is a different illness than MitD, and both have different hurdles pertaining to cost and health states. Given the lower life expectancy of severe MitD patients relative to those of CP, results here may be an overestimate with the inclusion of costs seen in infants with CP, but not in infants with MitD. Most available evidence regarding diagnostic likelihood and longer-term outcomes of patients with MitD is confined to single-institution studies at major academic hospitals. Patients seen at those centers likely have different outcomes and costs than those seen at smaller institutions, so the inclusion of such data may create bias. Finally, MitD is notoriously clinically and economically heterogeneous. Thus, the base case explored here may not represent the extreme clinical spectrum or diagnostic pathways for neonates with severe MitD. This paper highlights that interventions leading to earlier diagnoses for MitD neonates are cost-effective and cost-minimizing. Through earlier diagnosis, parents can be counseled earlier, reducing family-associated emotional burden; earlier treatment can potentially stabilize the patient; or palliative care for severe cases can be implemented earlier. Although this paper demonstrates the clinical and economic value of eES as a diagnostic for newborns with severe metabolic disorders, the findings also highlight research areas important for improving neonatal care of patients with metabolic disorders. Further, EVPIs calculated here indicate that decision makers stand to gain upwards of$90,000 per MitD patient, if the uncertain parameters are studied further.

Scenario analyses demonstrate potential limitations of ES as a MitD diagnostic. Due to eES cost and diagnostic accuracy, eES is best served as an early diagnostic. Used as a last resort, early diagnosis benefit, such as earlier patient discharge and shorter NICU length-of-stay, are not observed in the modeled population, suggesting in some cases, no exome sequencing is as beneficial as late exome sequencing (Fig. 3). This finding underscores the importance of including eES as a first-line diagnostic, especially as diagnostic efficiency improves.

Threshold analyses demonstrate an interesting dynamic in this model. As total post-NICU cost increases, no sequencing incurs the largest iNMB. However, that decreases as total post-NICU cost decreases, and eES starts to incur the largest iNMB. Although this analysis is limited with fixed parameters, these findings demonstrate the importance of later life costs in determining the value of early-life diagnostics. This highlights the need for future discussions to address cost of patient treatment after stabilization as well as rethinking some neonatal care value metrics.

When evaluating eES from the societal perspective for a genetic disorder, there are other scenarios worth considering. For example, the newborn diagnostic odyssey can lead to a substantial caregiver and parent quality-of-life burden.33 Future analyses might also consider the advent of curative cellular therapies. Capturing the impact of curative therapies will also underscore the potential benefit of exploring innovative diagnostics.

As an expensive yet innovative diagnostic, convincing payers and public agencies to reimburse eES has been difficult. By modeling the clinical, health status, and cost associated with a severe neonatal MitD, we have demonstrated the value of earlier inclusion of ES in the diagnostic process. In addition to reducing the emotional toll on families awaiting diagnosis, these findings can be informative for clinicians and decision makers while we continue to search for a curative therapy for this devastating group of diseases.