Using Artificial Neural Networks to Predict Claim Duration in a Work Injury Compensation Environment

Anthony Almudevar

stat.APFeb 17, 2026

Using Artificial Neural Networks to Predict Claim Duration in a Work Injury Compensation Environment

Authors: Anthony Almudevar

Organizations: Department of Biostatistics and Computational Biology, University of Rochester, Rochester, New York

Abstract

Currently, work injury compensation boards in Canada track injury information using a standard system of codes (under the National Work Injury Statistics Program (NWISP)). These codes capture the medical nature and original cause of the injury in some detail, hence they potentially contain information which may be used to predict the severity of an injury and the resulting time loss from work. Claim duration easurements and forecasts are central to the operation of a work injury compensation program. However, due to the complexity of the codes traditional statistical modelling techniques are of limited value. We will describe an artificial neural network implementation of Cox proportional hazards regression due to Ripley (1998 thesis) which is used as the basis for a model for the prediction of claim duration within a work injury compensation environment. The model accepts as input the injury codes, as well as basic demographic and workplace information. The output consists of a claim duration prediction in the form of a distribution. The input represents information available when a claim is first filed, and may therefore be used in a claims management setting. We will describe the model selection procedure, as well as a procedure for accepting inputs with missing covariates.

Figures & tables

Short	Long	Comments
Variable Name
NOI	Nature of Injury	NWISP injury code
POB	Part of Body	NWISP injury code
SOI	Source of Injury	NWISP injury code
TOA	Type of accident	NWISP injury code
AGE	Age of claimant	Categorized by quartile

TABLE I: Predictor Variables.

Nature of Injury - Medical condition underlying injury
03400	Cuts, lacerations
05302	Second-degree heat burns, scalds
14530	Silicosis
Part of Body - Location of injury
22100	Esophagus
34000	Finger(s), fingernail(s), unspecified

TABLE II: Coding examples

Code	No. Distinct
	Original	Consolidated
NOI	154	61
POB	119	58
SOI	655	72
TOA	169	28
SIC	373	80

TABLE III: Coding consolidation

Fig. 1: Long term trend in claim duration. 1) and 2) Mean and median trends using piecewise linear Cox regression (solid) and quarterly summaries (dashed). 3) Quarterly summaries of censoring rates.

Fig. 2: Log cumulative hazard functions partitioned by NOI, POB, OCC, PAY, AGE and SEX (by column, then by row).

Fig. 3: Multiple boxplots of log duration given for each predictor.

NOI	longer durations	P value
	Sex with
Dislocations	Female	.0025
Sprains, strains, tears	Female	$<$ .0001
Rheumatism, except the back, unsure	Female	$<$ .0001
Avulsions	Male	$<$ .0001
Burns, n.e.c.	Male	.0081

TABLE IV: Interactions between Nature of Injury and Sex (out of 25 codes)

POB	longer durations	P value
	Sex with
Chest, except int. loc. of diseases …	Female	.0373
Wrist(s)	Female	.0003
Knee(s)	Female	.0106
Groin	Male	.0283
Finger(s), fingernail(s)	Male	$<$ .0001

TABLE V: Interactions between Part of Body and Sex (out of 38 codes)

Fig. 4: Generalized

R^{2}

by decay parameter, for reduced and full models. Separate lines corresponding to number of hidden nodes are superimposed.

Prediction	Actual Duration Quintile
Quintile	1	2	3	4	5
1	0.38	0.27	0.20	0.11	0.04
2	0.25	0.22	0.23	0.19	0.12
3	0.19	0.21	0.20	0.23	0.17
4	0.16	0.16	0.19	0.22	0.27
5	0.09	0.12	0.13	0.26	0.40

TABLE VI: Proportion of actual closed duration quintiles by predicted quintiles.

Fig. 5: Actual durations (for closed claims) partitioned into deciles on the basis of prediction terms calculated from selected model. Plots are given in original and logarithm scale, for clarity. The location of 1, 4 and 10 weeks is indicated.

Fig. 6: Estimated relationship between actual and predicted duration means and quartiles by individual subject.

Fig. 7: Predicted mean and median duration against actual mean and median duration partitioned by POB, for each sex. Predictions are calculated using Method A (averaging over prediction terms).

Fig. 8: Predicted mean and median duration against actual mean and median duration partitioned by POB, for each sex. Predictions are calculated using Method B (averaging over survival curves).

Fig. 9: Sex difference in median duration partitioned by POB, predicted versus actual. Durations are transformed using a centered logistic function.

Event Time Prediction

Event time prediction estimates when future events will occur from event histories, time series, and contextual information.

See all 44 papers

SSH-Net: A Deep Neural Network for Predicting Failure Time Distribution Functions under Competing Risks with Application to GPU Data
Jun 18, 2026Jie Min, Yueyao Wang, Mengkun ChenSurvival Analysis Modular Neural Networks
Department of Mathematics & Statistics, University of South Florida, Tampa, FL · School of Statistics and Data Science, Zhejiang Gongshang University, Hangzhou, China · Department of Statistics,AbstractVirginia Tech, Blacksburg, VA
KAPLAN: Kolmogorov-Arnold Prognostic Learnable Activation Networks for Survival Analysis
May 21, 2026Stelios Boulitsakis Logothetis, Angela Wood, Pietro LiòSurvival Analysis Kolmogorov-Arnold Networks
University of Cambridge
Joint Distributional Copula Regression for Insurance Claim Frequency and Severity via Boosting
Oct 27, 2026Banghee So, Na ZhangDistributional Regression Copula Models
Department of Mathematics, Towson University, 7800 York Rd RM335, Towson, 21204, MD, USA. · Department of Mathematics, Towson University, 7800 York Rd RM371, Towson, 21204, MD, USA.

Jun 18, 2026J/K move · Enter open · S save

SSH-Net: A Deep Neural Network for Predicting Failure Time Distribution Functions under Competing Risks with Application to GPU Data

Jie Min, Yueyao Wang, Mengkun Chen

Department of Mathematics & Statistics, University of South Florida, Tampa, FL · School of Statistics and Data Science, Zhejiang Gongshang University, Hangzhou, China · Department of Statistics,AbstractVirginia Tech, Blacksburg, VA

Survival Analysis Modular Neural Networks

Open paper AskPDF arXiv

Competing risks are commonly observed in engineering fields and can bring challenges to time-to-event data modeling when the application scenarios are complicated. Recently, deep neural networks have received great attention for prediction with competing risks, due to their flexibility and high learning capability. However, the complexity of neural network structure brings extra difficulty in hyperparameter tuning based on different data inputs. Additionally, when an engineered system has complex physical structures with multiple hierarchical levels, treating all structural levels as a single group of inputs may fail to capture critical information. To address the issues, we propose a Structured Segmented Hazard Deep Neural Network (SSH-Net) for failure time prediction under cause-specific competing risks framework. Our approach associates neural network structure with data structures, and allows different covariate groups to impact the failure prediction through separate sub-networks. The neural network is constructed based on a cause-specific competing risks model. The SSH-Net outputs cause-specific hazard functions, and utilizes the penalized log-likelihood as the loss function. The prediction accuracy of SSH-Net is validated through simulation studies by evaluating the Brier score, the area under receiver operating characteristic curves (AUC), and the root mean square error (RMSE) of the predicted cause-specific cumulative incident function. We further demonstrate the model's ability to predict failure time distribution functions using the Titan GPU failure time data.

Abstract

Event Time Prediction

Explore similar work

SSH-Net: A Deep Neural Network for Predicting Failure Time Distribution Functions under Competing Risks with Application to GPU Data

KAPLAN: Kolmogorov-Arnold Prognostic Learnable Activation Networks for Survival Analysis

Joint Distributional Copula Regression for Insurance Claim Frequency and Severity via Boosting