Source-linked AI summary
A dynamic network approach for the study of human phenotypes
Cesar A. Hidalgo, Nicholas Blumm, Albert-Laszlo Barabasi, Nicholas Christakis
TL;DR
Disease research lacked comprehensive population-based phenotypic relationships despite extensive clinical histories. The paper constructs a Phenotypic Disease Network from medical claims and finds that disease progression follows network proximity, varies across demographic groups, and is associated with connectivity and mortality.
Problem
Extensive population-based disease-association datasets were unavailable, although patient clinical histories provide phenotypic information relevant to disease relationships.
Method
The study constructs a Phenotypic Disease Network from Medicare hospitalization claims and publishes an extensive relational phenotypic data resource.
Results
Patients develop diseases close in the PDN; progression differs by gender and race, and highly connected or preceded diseases are associated with earlier death.
Takeaways & Limitations
Network methods can represent and study disease progression, while the published PDN provides a resource for investigating human disease relationships.
Takeaways & Limitations
The results should be interpreted within a population of elderly citizens in an industrialized country because the dataset consists of hospitalization claims from elderly United States residents.
Abstract
from arXiv · showhide
The use of networks to integrate different genetic, proteomic, and metabolic datasets has been proposed as a viable path toward elucidating the origins of specific diseases. Here we introduce a new phenotypic database summarizing correlations obtained from the disease history of more than 30 million patients in a Phenotypic Disease Network (PDN). We present evidence that the structure of the PDN is relevant to the understanding of illness progression by showing that (1) patients develop diseases close in the network to those they already have; (2) the progression of disease along the links of the network is different for patients of different genders and ethnicities; (3) patients diagnosed with diseases which are more highly connected in the PDN tend to die sooner than those affected by less connected diseases; and (4) diseases that tend to be preceded by others in the PDN tend to be more connected than diseases that precede other illnesses, and are associated with higher degrees of mortality. Our findings show that disease progression can be represented and studied using network methods, offering the potential to enhance our understanding of the origin and evolution of human diseases. The dataset introduced here, released concurrently with this publication, represents the largest relational phenotypic resource publicly available to the research community.
Author Summary
The study introduces the Phenotypic Disease Network (PDN), a map of disease connections, and uses it to examine disease progression and mortality across patient groups. It also provides an online database built from millions of patient records.
- The PDN summarizes phenotypic connections between diseases and represents disease progression as occurring along the links of this map.
- Disease progression along PDN links differs for patients with different genders and racial backgrounds.
- Patients with diseases connected to many other diseases in the PDN tend to die sooner than those with less connected diseases.
- The study created a queryable online database from 18 datasets generated from more than 31 million patients.
INTRODUCTION
Existing disease networks largely emphasize genetic, proteomic, and expression relationships, while extensive population-based comorbidity data remain unavailable. This study addresses that gap by constructing and publishing a PDN from medical claims, then examining disease associations, progression, demographic differences, and mortality.
- Existing network approaches: Prior network resources connect diseases through shared genes, proteins, metabolic reactions, or expression patterns, but do not provide comprehensive phenotypic relationships.
- Research gap: Extensive comorbidity datasets were unavailable partly because access to medical records is limited.
- Data resource: The study makes pairwise comorbidity correlations available for more than 10 thousand diseases reconstructed from over 30 million medical records.
- Data resource: The results are organized into 18 datasets and grouped by race, gender, and combined race-and-gender subsets.
- Study aims: The PDN captures diseases recorded through medical claims and is used to study illness progression, demographic differences, network connectivity, and mortality.
- Disease progression: More central diseases are more likely to occur after other diseases, while more peripheral diseases tend to precede other illnesses.
- Disease progression: Patients diagnosed with diseases preceded by other conditions tend to die sooner than patients diagnosed with conditions that precede other illnesses.
- Data limitations: The claims data cover hospitalizations of elderly United States citizens, limiting representation of diseases uncommon among elders and conditions treated solely as outpatient care.
RESULTS
The study constructs the Phenotypic Disease Network from disease comorbidity associations and uses it to examine disease progression, demographic variation, connectivity, and lethality. The results support network-based representations of illness development while highlighting limitations from incomplete disease histories and observation periods.
- Network construction: The PDN represents disease phenotypes as nodes and links phenotypes with statistically significant comorbidity associations.The analysis focuses on the strongest and most significant associations and provides RR- and φ-based visualizations.
- Network construction: RR and φ produce complementary network representations: RR emphasizes relatively infrequent diseases and ICD9-like modules, whereas φ emphasizes prevalent diseases with cross-category connections.The paper does not favor one representation; both capture significant associations at different prevalence scales.
- Progression along network links: 95.6% of patients in the φ-PDN and 81.5% in the RR-PDN showed inter-visit correlations above chance, by average factors of 10 and 1.5, respectively.The stronger effect in the φ-PDN supports studying illness progression as a spreading process along network links.
- Demographic variation: Comorbidity patterns differed across racial and gender groups, with several cardiovascular and pulmonary conditions more comorbid among white males and hypertension, diabetes, and renal disorders more characteristic of black males.The reported associations also reproduce known disease relationships and support examining demographic variation in comorbidity.
- Connectivity and lethality: Connectivity correlated with lethality in both PDNs, while prevalence showed only a weak correlation and did not explain the connectivity–lethality relationship.The connectivity association remained significant after controlling for age, gender, number of diagnoses, and number of visits, and was stronger for some disease groups such as neoplasms.
DISCUSSION
The PDN provides a large public resource for studying disease associations, progression, and population differences. Its structure suggests that network connectivity relates to disease lethality and may support future integration with genetic and proteomic data.
- DISCUSSION: The study introduces an extensive publicly available dataset quantifying comorbidity associations in a large population.The data are made available to the research community as a phenotypic resource.
- DISCUSSION: Phenotypic network information could complement genetic and proteomic data in studies of disease etiology and evolution.The authors also suggest that phenotypic maps could be used to study how diseases evolve in patients.
- DISCUSSION: Patients tend to develop diseases close to their existing diseases in the PDN, which represents disease progression through network proximity.The authors describe this as suggestive rather than conclusive evidence.
- DISCUSSION: More connected diseases are observed to be more lethal, possibly because highly connected diseases can be reached through multiple progression paths.The proposed explanation is that patients developing highly connected diseases may be at a more advanced disease stage.
- DISCUSSION: Comorbidity strengths differ among patients of different races and genders, making the PDN a starting point for studying population differences.The paper notes that these differences may relate to biological processes, environmental factors, or healthcare quality.