Data and code for: Veterinary Expert System for Outcome (VESOP) Prediction
Data files
Aug 30, 2023 version files 154.23 KB
-
README.md
-
VESOPdata_1yr.csv
-
VESOPdata_2yr.csv
Abstract
Timely detection and understanding of causes for population decline are essential for effective wildlife management and conservation. Assessing trends in population size has been the standard approach but we propose that monitoring population health could prove more effective. We collated data from seven bottlenose dolphin (Tursiops truncatus) populations in the southeastern U.S. to develop the Veterinary Expert System for Outcome Prediction (VESOP), which estimates survival probability using a suite of health measures identified by experts as indices for inflammatory, metabolic, pulmonary, and neuroendocrine systems. VESOP was implemented using logistic regression within a Bayesian analysis framework, and parameters were fit using records from five of the sites that had robust stranding network and frequent photographic identification (photo-ID) surveys to document definitive survival outcomes. We also conducted capture-mark-recapture (CMR) analyses of photo-ID data to obtain separate estimates of population survival rates for comparison with VESOP survival estimates. VESOP analyses found multiple measures of health, particularly markers of inflammation, were predictive of 1- and 2-year individual survival. The highest mortality risk one year following health assessment related to low alkaline phosphatase, with an odds ratio of 10.2 (95% CI 3.41–26.8), while 2-year mortality was most influenced by elevated globulin (9.60; 95% CI 3.88–22.4); both are markers of inflammation. The VESOP model predicted population-level survival rates that correlated with estimated survival rates from CMR analyses for the same populations (1-year Pearson’s r=0.99; p=1.52e-05, 2-year r=0.94; p=0.001). While our proposed approach will not detect acute mortality threats that are largely independent of animal health, such as harmful algal blooms, it is applicable for detecting chronic health conditions that increase mortality risk. Random sampling of the population is important and advancement in remote sampling methods could facilitate more random selection of subjects, obtainment of larger sample sizes, and extension of the approach to other wildlife species.
README: Veterinary Expert System for Outcome Prediction (VESOP) data and code
We collated health data from seven bottlenose dolphin (Tursiops truncatus) populations in the southeastern U.S. to develop the Veterinary Expert System for Outcome Prediction (VESOP), which estimates survival probability using a suite of health measures. VESOP was implemented using logistic regression within a Bayesian analysis framework. Data include morphometrics, hematologic and serum biochemical results, pulmonary ultrasound scores, hormone measurements, and one- and two-year survival outcomes.
Description of the Data and file structure
The following files were used to implement the VESOP model for the manuscript and associated supplemental information:
Schwacke, LH, L Thomas, RS Wells, TK Rowles, G Bossart, F Townsend, M Mazzoil, JB Allen, BC Balmer, AA Barleycorn, A Barratclough, ML Burt, S De Guise, D Fauquier, FM Gomez, NM Kellar, JH Schwacke, TR Speakman, E Stolen, BM Quigley, ES Zolman, and CR Smith. An expert-based system to predict population survival rate from health data.
Data input files:
VESOPdata_1yr.csv, VESOPdata_2yr.csv- data from the 7 dolphin sampling locations used to fit the VESOP model and generate graphs for 1- and 2-year survival, respectively; each row is a separate dolphin health record and includes:
- Site- site where dolphin was sampled
- Year- year dolphin was sampled
- panelclass- age classification for the dolphin
- sex- sex of the dolphin
- Neutrophils (thousand/µL), Albumin (g/dL), Globulin (g/dL), Alkaline.Phosphatase (µ/L), Hemoglobin (g/dL), weight (kg), maximum.girth (cm), length (cm), Cholesterol (mg/dL), TotalT3 (ng/dL), Glucose (mg/dL), Lung (1 indicating moderate-severe; 0 indicating normal-mild), Cortisol (µg/dL), health measurements made on the dolphin
- refSiteLung, refSiteCort- indicates if site was an oiled site (1) or unoiled site (2)
- Survival- binary variable indicating if the dolphin survived 1 (or 2) years
- S.beta_par1, S.beta_par2- beta distribution parameters if CMR survival probability was calculated
- CMRSurvival- estimated CMR survival probability
phi_est_1Yr_bysite_CMR.Rds, phi_est_2Yr_bysite_CMR.Rds data in R format (can be read in with readRDS) giving 1000 replicate predictions of survival per site. Sites are the rows and replicate predictions are the columns. Rownames give the site names.
R code files:
- applyPanelsVESOP.R code that reads in VESOPdata file and adds health index flags, calling functions in calcFlags.R to estimate appropriate threshold values. Creates a temp.csv file that is input for run VESOP model.R
- calcFlags.R - functions that estimate threshold values using either non-linear quantile regression or calling R quantile function with stratified data
- run VESOP model.R - reads in data file (temp.csv), sets up model health factors and outcome fields, runs the rjags model (VESOPModel.R), and processes output to create graphs.
- VESOPModel.txt - rjags model code
- printSurvs.R - various functions that use VESOP model outputs to calculate population survival statistics and plot results.
Sharing/access Information
Links to other publicly accessible locations of the data:
Was data derived from another source? No
If yes, list source(s):
Methods
We collated data from seven bottlenose dolphin (Tursiops truncatus) health assessment studies in the southeastern U.S. to develop the Veterinary Expert System for Outcome Prediction (VESOP), which estimates survival probability using a suite of health measures. VESOP was implemented using logistic regression within a Bayesian analysis framework. Data include morphometrics, hematologic and serum biochemical results, pulmonary ultrasound scores, hormone measurements, and one- and two-year survival outcomes.
Usage notes
All code was implement in R and Rjags.