Image Credit: OpenAI DALL-E.
Biostatistical analysis is a cornerstone of healthcare research. Applications span a broad spectrum, including designing clinical trials and healthcare quality metrics. The sensitive nature of healthcare data presents a challenge to biostatisticians, however, as de-identifying then aggregating Real World Data (RWD) from their partners for analysis is not always realistic. Rhino Health’s Federated Biostatistics - allowing biostatisticians to run code on data distributed behind multiple custodians’ firewalls - is the answer. The Federated Biostatistics capability is empowering members of the Rhino Network to easily run their preferred biostatistical analyses across large, persistent networks of data partners. This article describes both the capability, and one illustrative collaboration between an Israeli HMO and a global pharmaceutical company using the Rhino Federated Computing Platform (FCP).
There are many applications of biostatistics in healthcare and life sciences. Examples include the design and analysis of clinical trials to assess treatment efficacy, epidemiological studies investigating disease distribution and determinants, genetic association studies identifying links between genetic variants and disease risk, survival analysis techniques examining time-to-event data, longitudinal data analysis to study disease progression, meta-analysis for synthesizing evidence from multiple studies, diagnostic test evaluation to assess test accuracy, and public health surveillance for monitoring disease outbreaks and health indicators. These analyses inform our understanding of disease dynamics, the efficacy of interventions, and the utility of health policy decisions.
Despite the value of biostatistics, they can be hard to implement in the real world. Biostatisticians often cannot find all of the data they would like to use. When they do, it is often locked in proprietary silos such as EHRs, claims databases, and registries. Even if the silos are connected, the data are often in different data models, presenting a data engineering challenge. And then if the data are harmonized, real time access to up to date data is a challenge. Underpinning all of these is the concern of patient privacy. To date, biostatisticians must do their best to overcome these hurdles.
The FCP’s Federated Biostatistics capability solves all of these issues, by providing a safe way for biostatisticians to build persistent pipelines to a broad set of high quality, harmonized data. The FCP’s Federated Biostatistics capability includes the biostatistician’s full toolkit, including:
Descriptive statistics
Tools for testing both parametric and non-parametric data (Chi Square, t tests, ANOVA, Wilcoxon Test, etc.)
Statistical correlations (e.g. Pearson’s, Intraclass, Spearman’s)
Risk adjustment methods, including regression analysis (linear, logistic, Poisson), Kaplan Meier, and Cox Proportional Hazards
The best part: all of the above tools are run using the same privacy preserving features as the full FCP including k anonymization and differential privacy AND with the same extensible architecture. Biostatisticians will now be able to deploy their code on data behind partner firewalls, taking advantage of the core capabilities of the FCP including data ingestion, pre-processing, annotation, and harmonization - all with a persistent network allowing real time access to partner data, and all with no data transfer.
Rhino Health has validated the efficacy of Federated Biostatistics with a leading HMO in Israel as part of the HMO’s ongoing collaboration with a global pharmaceutical company. Our HMO partner compared the results of the federated analyses to those of an analysis on centralized data and found equivalent results. This validation exercise has given the HMO partner the confidence they needed to deepen their collaboration with the pharmaco, creating value for both. Together, are now able to responsibly execute these critical analyses without risk of violating patient privacy, nor of giving away sensitive patient population data.
Rhino’s Federated Biostatistics package is available for use today. Contact us to learn more!
Yuval Baror
Co-founder and CTO
Lili Lau
Director of Product Marketing