veteran_comorbidity — Co-morbidity network of US military veterans (2001-2017)
- Description
A dense, directed and weighted network of co-morbidity among 95 physical and mental health conditions, derived from the longitudinal health records of approximately one million United States military veterans. The records come from the VA (Department of Veterans Affairs) National Corporate Data Warehouse, which holds the coded diagnoses of all VA inpatient and outpatient encounters, for veterans who served in the Operation Enduring Freedom, Operation Iraqi Freedom or Operation New Dawn (OEF/OIF/OND), enrolled in VA healthcare after 1 October 2001, and had at least one VA primary care visit before the end of the study period, on 1 October 2017. The conditions are diagnostic categories with a prevalence of at least 0.1%: categories of the AHRQ (Agency for Healthcare Research and Quality) Clinical Classifications Software, psychiatric diagnoses as categorized by the VA's Northeast Program Evaluation Center (NEPEC: alcohol use disorder, drug use disorder, schizophrenia, major depressive disorder, bipolar disorder, PTSD, other depression, unspecified and generalized anxiety), and diabetes, neoplasms, congenital anomalies and genital disorders. For each ordered pair of conditions, the association was estimated with a logistic regression of the odds of one condition given the other.
Nodes are conditions, and a directed edge from condition j to condition i carries the increase in the log odds of condition i given condition j (the logistic regression coefficient; negative values indicate a decreased odds), so that the out-edges of a node describe the conditions that it predicts. All 95 x 94 ordered pairs are present, and the network is complete. This edge direction follows the authors' sample code (which builds the network from the transpose of the matrix of log odds, whose rows are the predicted conditions). The upstream repository also includes matrices for the interaction effect of sex and its p-values, but in the available files these contain only zeros, and they are not included.
Node properties:
name, the variable name of the condition in the source (with prefix "AHRQ_" for the AHRQ categories and "NEPEC_" for the NEPEC psychiatric categories; e.g. "AHRQ_Asthma_DT", "NEPEC_PTSD_DT", "dmdxDT" for diabetes);short_name, the abbreviated label used in the authors' sample code;psychiatric, whether the condition is psychiatric, as marked in the authors' sample code. Edge properties:log_odds, the increase in log odds described above;pvalue, its p-value from the logistic regression (values below the floating point precision are reported as 0).
- Tags
- Biological Public Health Weighted Metadata
- Citation
-
- A. F. Alexander-Bloch, A. Raznahan, R. T. Shinohara, S. R. Mathias, H. Bathulapalli, I. P. Bhalla, J. L. Goulet, T. D. Satterthwaite, D. S. Bassett, D. C. Glahn, C. A. Brandt, "The architecture of co-morbidity networks of physical and mental health conditions in military veterans." Proceedings of the Royal Society A 476(2239), 20190790 (2020) 10.1098/rspa.2019.0790 Sci-Hub
- Upstream URL OK
- https://github.com/aaronab/comorbidity_networks
- Networks
-
Tip: hover over a column header for a description.
Column legend
- $\left<k\right>$
- Mean degree
- $\sigma_k$
- Standard deviation of the degrees
- $\lambda_h$
- Largest eigenvalue of the Hashimoto (non-backtracking) matrix
- $\tau$
- Random walk mixing time
- $r$
- Degree assortativity
- $c$
- Global clustering coefficient
- $\oslash$
- (Pseudo-) diameter
- $S$
- Relative size of the largest component
- NPs
- Node properties
- EPs
- Edge properties