01:30 PM - 02:00 PM
Risk Adjustment of Stroke Outcomes for Comparing Hospital Performance: A Modelling Perspective
Stroke is one of the three leading causes of death and a leading cause of long-term disability worldwide. Measuring outcomes after stroke has both major policy and operational implications. Recently the American Heart and Stroke Associations have published a joint statement on this subject. The statement provides an overview of statistical considerations for the evaluation of hospital-level outcomes after stroke and promotes the use of hierarchical or multilevel random effect regression models that use random effect terms to describe hospital-specific effects. It is argued in the joint statement that the proposed approach helps in overcoming the failure to account for variation in case numbers across hospitals and for intra-hospital clustering effects. Using mortality outcome as an example, the random effect modelling approach would generate both the population-expected hospital mortality levels and the hospital-specific predicted mortality levels by applying the regression coefficients generated from all the patients in the sample, but using different values of random intercept. The predicted-over-expected ratio is then multiplied by the unadjusted mortality rate to yield a risk-standardized mortality rate. Generating the predicted values of the hospital-specific intercepts is a non-trivial task and these are computed as a part of Bayesian framework. It is important to note that this strategy is strongly cautioned against in the random-effect regression modelling literature due to the specific technical problems arising with estimation of the cluster-specific random effects.
In this talk we provide a modelling perspective on the use of random-effect regression modelling for risk-adjusted comparison of hospital performance, discuss its advantages and limitations, and compare the random-effect approach to other potential alternative approaches using Australian stroke hospital mortality data as an illustrative context.
02:00 PM - 02:30 PM
Histopathology Laboratory Operations Analysis and Improvement
Histopathology laboratories aim to deliver high quality diagnoses based on patient tissue samples. Indicators for quality are the accuracy of the diagnoses and the diagnostic turnaround times. However, challenges exist regarding employee workload and turnaround times, which both impact the diagnostic quality. This paper proposes a decomposed planning and scheduling method for the histopathology laboratory using (mixed) integer linear programming to improve the spread of workload and reduce the diagnostic turnaround times. First, the batching problem is considered, in which batch completion times are equally divided over the day to spread the workload. This reduces the peaks of physical work available in the laboratory. Thereafter, the remaining processes can be scheduled to minimize the tardiness of orders. Results show that using this decomposed method, the peaks in histopathology workload in UMC Utrecht, a large university medical center in the Netherlands, are reduced with up to 50% by better spreading the workload over the day. Furthermore, turnaround times are reduced with up to 20% compared to current practices. This approach is currently being implemented in the histopathology laboratory of UMC Utrecht.
02:30 PM - 03:00 PM
Knowledge-based Quality Assessment of Radiotherapy Treatment Plans
The aim of radiotherapy treatment (RT) for cancer is to irradiate the tumour while avoiding damage to surrounding healthy organs at risk (OARs). RT planning involves assessment of several objectives related either to the tumour or OARs, leading to multi-objective optimisation problems. The major commercial treatment planning systems use weighted sums of objectives in deriving treatment plans. RT planners iteratively modify plans by adjusting the weights of objectives until clinical planning criteria are met. This iterative trial-and-error nature of RT planning results in an inefficient planning process. At the completion of the planning phase, a plan is produced for the oncologist to review. If a plan is not deemed acceptable it takes more time to produce another plan, without knowing in advance whether the new plan will be considered preferable to previous plans. Assessing RT plan quality is a topic that is receiving attention by researchers and health practitioners.
We propose to measure RT plan quality using a linear-programming-based technique called data envelopment analysis (DEA) that compares each new plan being generated to a library of previously generated clinically acceptable plans. This is not straight-forward as every patient is different, the tumour and OARs will have different shapes in each case, and the relative closeness to OARs will vary. DEA assesses quality based on plan parameters such as dose delivered to the tumour and dose received by OARs while taking into account patient specific variations.
We demonstrate that DEA-based plan assessment works well for prostate and head and neck cancer which can have several OARs and patient dependent volume of OAR overlap with the tumour. We thus obtain a knowledge-based plan quality assessment methodology that builds on a library of plans and captures several plan characteristics such as radiation dose received as well as the patient-specific geometry.