Laboratory for Optimization and Computation in Orthopaedic Surgery (LOCOS)

Laboratory for Optimization and Computation in Orthopaedic Surgery (LOCOS)

Mission

The Laboratory for Optimization and Computation in Orthopaedic Surgery (LOCOS) is the academic laboratory of Dr. Hughes. The mission of LOCOS is to apply computational methods to improve musculoskeletal health.

Current research areas

Quality improvement in joint replacement surgery

Dr. Hughes co-founded the Michigan Arthroplasty Registry Collaborative Quality Initiative (MARCQI), which is a state-wide network of hospitals and surgeons dedicated to improving the quality of care for total hip and knee replacement (“arthroplasty”) patients. An example of what MARCQI has accomplished is the reduction in the risk of transfusion over time in Michigan (Figure 1). Each transfusion increases the risk of infection, death, etc.; therefore, reducing transfusions improves the quality of care. For more information on MARCQI, see the paper, Hughes et al. (2015), that described the early history of MARCQI and its structure. A more in-depth description can be found in MARCQI’s first annual report.

Figure 1. Risk of transfusion during hip and knee replacement surgery in Michigan.

Causal inference in health care quality improvement

The gold standard for finding causal relationships in medicine is the randomized control trial (RCT), and observational data are considered suspect for finding causal relationships. Yet registries such as MARCQI that are dedicated to quality improvement rely on the analysis of observational data because RCTs are expensive, time consuming, and sometimes unethical to conduct. Therefore, a knowledge gap exists in the health care quality improvement realm, i.e. how to strengthen causal inference when analyzing registry data. An example from our previous work was understanding the relationship between tranexamic acid (TXA) and blood loss during surgery. Understanding the safety and effectiveness of TXA was an important stem in achieving the success shown in Figure 1. LOCOS investigators are involved in this research. The methods involve analysis of directed acyclic graph models of causality as well as optimal matching. An example of a study conducted at LOCOS on causal inference using MARCQI data is the paper by a Biomedical Engineering student, Camden Cheek:  Cheek et al. (2018)

Translating arthroplasty registry data into practice

MARCQI produces and posts annual reports that provide revision risk data by implant. The most recent one is Hughes, R.E., Zheng, H., and Hallstrom, B.R. 2019 Michigan Arthroplasty Registry Collaborative Quality Initiative (MARCQI) Annual Report. University of Michigan, Ann Arbor. Please open this document and look at it. You will see it is 243 pages long, with extensive information about each implant. However, a practicing orthopaedic surgeon is unlikely to wade through an entire report like this to find the information necessary to make an informed choice of what implant to use. Thus, the challenge LOCOS faces is creating novel visualization methods for presenting data from this report that will be maximally impactful for improving the selection of implants in Michigan and around the world. In addition, LOCOS students are developing online tools for accessing these data to facilitate adoption of the best orthopaedic implants by Michigan surgeons.

Adapting occupational biomechanics models for use in litigation

Biomechanical analyses have often been used in support of civil litigation. In area of personal injury law, forensic experts are called upon to “opine” on the causes of injuries as well as whether some accepted design standard was or was not met by the defendant. The standard of proof used in civil litigation is “more probable than not.” However, most biomechanical models are deterministic while this standard is stochastic. Thus, there is a gap between what the research literature has generally produced and what experts need. LOCOS is working to bridge this gap using hybrid Bayesian network modeling tools. I have published three papers related to this topic. Using a Bayesian Network to Predict L5/S1 Spinal Compression Force from Posture, Hand Load, Anthropometry, and Disc Injury Status implemented a static biomechanical model of lifting that incorporated disc injury status, which is something that is often available in civil litigation cases involving back injury. This paper was the foundation for a paper that proposes a method for developing hybrid Bayesian network models in occupational biomechanics that can be used to address the question of negligence (Hybrid Bayesian Network models of spinal injury and slip/fall events). Finally, Hybrid Bayesian Network Modeling of Slips and Falls for Forensic Analysis in Civil Litigation proposed a Bayesian network modeling framework that meets the need of an expert opining of two competing theories of causation of a slip and fall event.

Working with LOCOS during COVID-19

All LOCOS work is being conducted remotely.