The provider discovery problem

A parent whose newborn was identified by screening with a metabolic disorder typically has access to the regional metabolic genetics center within days of the screen. The infrastructure for that referral pathway works because the universe of metabolic genetics centers is small and the public health structure that runs newborn screening has direct relationships with each one.

The same parent, two years later, looking for a metabolic dietitian who has experience with their child's specific condition, may have no equivalent infrastructure. A few academic centers maintain provider lists. Some advocacy organizations maintain partial directories. Facebook groups and conference hallway conversations fill in. The directory of providers with expertise in a given rare disease, accessible by the affected community, with quality signals based on outcomes, does not exist for most conditions.

The same gap multiplies as the affected child grows up. Adult metabolic care for inborn errors of metabolism is a small subspecialty that most adult primary care providers have no training in. Adult EDS care is concentrated in a handful of specialty centers and not consistently available elsewhere. Adult care for survivors of childhood-onset cardiomyopathies, lysosomal storage disorders, and genetic epilepsies is similarly fragmented. The transition from pediatric to adult care is recognized as a structural problem in the rare disease literature. The provider discovery infrastructure is the unsolved part.

Why directories built by other people fail

Three structural reasons make conventional provider directories incomplete for rare disease.

The first is that the directories are built by parties who do not see the outcomes. A medical society directory lists members. A health system directory lists employed providers. A consumer search platform lists providers who have paid for placement. None of the three has the data to identify which providers achieve consistently better outcomes for which conditions. The rankings, when present, reflect satisfaction surveys, peer reputation, or institutional affiliation rather than outcome data.

The second is that the directories are condition-specific where the patient's question is patient-specific. A parent looking for a metabolic dietitian for a child with PKU plus emerging cognitive features wants a dietitian who handles both. A patient with hEDS plus POTS plus MCAS wants a clinician who has experience with the cluster, not three separate specialists each handling one feature. The directories that exist are organized by condition or specialty, which forces the patient to do the integration work the directory should do.

The third is that the directories age. Provider expertise changes as people retire, change institutions, develop new clinical interests, or stop seeing rare disease patients because of practice economics. A directory updated annually misses substantial flux. A directory updated by self-attestation reflects what providers say about themselves, which is not the same as what the outcomes data would say.

What the data would show

If the data trust captures structured information about which providers patients see, what treatments they receive, and what outcomes result, the provider directory emerges from the data organically.

A provider who treats a substantial number of patients with a specific condition becomes visible in the dataset. A provider whose patients consistently achieve better metabolic control, fewer hospitalizations, or better quality-of-life scores than the population mean becomes identifiable in the aggregate. The signal is the outcomes of their patients, not the provider's self-report.

The construction of the signal requires care. The provider's case mix matters. A provider who takes the more complex referrals will have outcomes that look worse than a provider who takes the easier cases, even if the more complex provider is the better clinician for the population the patient population needs. Risk adjustment for case mix, severity, and comorbidities is a methodological problem that the cardiac surgery quality reporting infrastructure has been working on for thirty years and has produced workable solutions for. The rare disease equivalent is achievable with the same methodological approach.

The signal supports patient navigation that does not currently exist. A parent looking for a metabolic dietitian sees the dietitians who handle their child's condition profile, ranked by outcomes within risk-adjusted comparisons. A patient looking for an EDS specialist sees the providers with experience in their specific phenotype constellation, ranked by outcomes for similar patients. The decision still belongs to the patient and family. The information that supports the decision is the information that has been missing.

What the provider gets

The provider whose outcomes are visible has both an opportunity and an incentive that the current system does not provide.

The opportunity is that providers with consistently good outcomes become discoverable to patients who need them. The current system produces provider visibility through institutional reputation, conference presentations, and word of mouth among the affected community. The data-driven system produces visibility through the actual quality of care provided. The latter rewards the clinician who has done the work.

The incentive is that providers who are aware their outcomes are visible have an external reason to invest in the rare disease subspecialty knowledge that the patient population needs. The current system has limited feedback mechanisms. A primary care physician who has been mismanaging a PKU patient for years receives no signal that the management is suboptimal. A data-rich ecosystem provides that signal in a form that supports the provider's improvement and, if necessary, the patient's referral elsewhere.

What the trust adds

The data trust governance protects the patient interest in two ways.

The provider quality signals are aggregated and risk-adjusted in a way that does not allow individual patients to be identified from the data that supports the signal. A specific patient's outcomes contribute to the aggregate. The aggregate reveals provider quality. The individual patient's contribution remains private.

The data trust supports patient choice rather than insurer or institutional choice. The provider quality data is available to the patient looking for a provider, not primarily to the insurer looking for a network or the health system looking for hiring decisions. The architectural choice matters because the question the data answers depends on whose tool the data is.

Building the dataset that supports provider discovery is the same project as building the dataset that supports the natural history work, the comparative effectiveness analyses, and the coverage decisions. The single longitudinal cohort dataset answers all of these questions if it is structured for them. The construction of the dataset is the project. The provider discovery use case is one of the byproducts.