Mediaplex_tag

Willis and Towers Watson have merged. Visit willistowerswatson.com

Analytics in an ergonomics world

Eric Kennedy and Tom Hilgen | 20th April 2015


In a recent meeting, a client said, “We do not have $50k to invest in ergonomics.” Yet they were averaging $8M in workers’ compensation claims per year. Where is that money coming from?

Many companies are paying a lot of money in claims that could be avoided with the right ergonomics program—but to determine the right program you need to understand the kind of injuries and losses your company is actually experiencing.

Risk management, human resources, claims and safety professionals charged with reducing an organization’s total cost of risk (TCOR) focus a lot of their attention on employee musculoskeletal disorders (MSDs)—those maladies affecting muscles, tendons, ligaments, nerves and blood vessels. With an increasingly aging and/or out-of-shape workforce, preventing workplace MSDs will become even more challenging.

Investments in ergonomics such as improving work flow and workstation design, plus improved tools and equipment, can significantly reduce the constant onslaught of annually recurring claims—and result in improvements in workplace efficiency, production, quality and scheduling.

But to determine which ergonomic investments will have the greatest impact on your company’s losses, you must first understand your losses.

First, understand the loss

Regardless of enterprise sophistication, we rarely see that a company’s loss experience is truly reflected in the time, talent and treasure they allocate to control those losses. For example, looking at one company’s chart, below, we see that musculoskeletal disorders account for 50% of incurred costs and 38% of claim frequency over a five-year period. Thus, it would be reasonable to allocate roughly 40% to 50% of enterprise loss control resources on the prevention and control of musculoskeletal disorders (ergonomics). However, a review of loss control activities, training, audits, checklists and other process related items indicates that less than 5% of resources are being allocated to control the dominant loss driver. This type of scenario is all too common.

A common scenario: Though musculoskeletal disorders for this client accounted for 50% of incurred costs, those losses received less than 5% of cost-control resources.

What is driving your loss cost?


Loss dataset size

To answer this question you must view a representative sample of loss data in a way that corresponds well to loss control programs or processes. We typically ask for 3 to 5 years of data (either calendar or policy years) in order to get a sample size that is adequate to provide high confidence in your conclusions.

The higher your frequency of claims the shorter your period can be, but you will not want to choose a sample size so large that older data is not representative of current conditions. For example, perhaps you began using a new process that substantially altered exposures in an area that has typically generated many losses. Including data from periods before the change will skew your results towards loss exposures that no longer exist.

Valuation dates

Now that you have an idea about how much data you want to view, it is critical to think through how you will value it, as this may dramatically affect your financial conclusions. There is a general upward trend in liability and workers’ compensation claim totals after the initial reporting period, called "loss development."

Trend data

If you select five years of data with a single valuation date at the end of the five-year period then the incurred values for less mature years will most always look lower in more recent years due to a lack of loss development. This kind of trend data might have period and valuation criteria as follows:


These lower incurred values in more recent history might be interpreted by audiences unfamiliar with loss development as improvement in the financial trend, while that may or may not be true.

There are two methods that can be used to deal with this problem and they both have strengths and limitations. The two methods are Point-in-Time Data and Growth-Development Factors.

Point-in-time data

In this method, each year of data is valued at the end of the year. For example, a five loss run using this method would require five different queries of your risk management information system (RMIS). The queries might look something like the following:

The advantage to point-in-time data analysis is that each period is equally mature and thus provides a more realistic view as to whether there is real improvement in the financials. The disadvantage is that point-in-time data has no development to any of the periods and thus understates the true significance of the financial impact on the enterprise.

Loss development factors

A common method of adjusting losses for the development/growth in claims and incurred but not reported (IBNR) losses is to apply loss development factors (LDFs). LDFs are used to arrive at the ultimate value that can be expected for a claim. For example, an LDF of 1.50 means that for every $1 of current claims, the ultimate pay out will be $1.50. A total of $50,000 in current claims would result in an ultimate pay out of $75,000. Loss development factors are typically calculated through actuarial analysis, and these may or may not be readily available to you.

The table below highlights some of the key issues to consider in determining which method to use.

Method Pros Cons
Trend data Easiest to pull from RMIS, only requiring one query. Incurred trend typically looks better in less mature years. This can be misinterpreted as improvement in the financial impact of losses.
Point-in-time data comparison Provides an honest comparison of periods with the same maturity for each period. Most difficult to pull from RMIS requiring a separate query for each period. Understates, significantly what the ultimate incurred values will be.
Loss development factors Provides a realistic prediction of ultimate incurred costs based upon previous actuarial analysis.Used with trend data, thus easy to pull the needed data with a single query. Loss development factors can be applied to the data as a whole but cannot be applied with much accuracy to subsets of data. For example, not all cause groups develop at the same rate. MSDs typically develop more and over a longer time than do cuts.

Fields to include

Obviously, the more fields you have the deeper your data mining exercise can go and the more insights you can generate to bolster your loss-control approach.

Below, we detail minimal mandatory and optional but helpful fields that we typically request. While your actual RMIS field names may vary from those supplied you can typically find some equivalent field to those supplied below.

Mandatory minimum

  • Loss Date (for trends)
  • Incurred Total (for financial impact)
  • Case #/Claim# (for error checking and follow-up questions)
  • Cause Description (for checking cause codes)
  • Cause Detail/Claim Detail Description (for creating cause drivers)

Optional helpful

  • Body Part Description
  • Nature Description
  • Source/Agent Description
  • Business Unit/DIVISION/Branch/Department (any and all organizational hierarchy)
  • Tenure/Hire Date
  • Age/Birth Date
  • Paid Total
  • Reserve Total
  • Recoveries Total
  • Report Date
  • Receipt Date
  • Status
  • Loss Type
  • State
  • Claimant Name

Super-sizing causes

It is absolutely critical to group causes, as best as possible, into categories that will map well to loss-control approaches. For example, all causes that would typically be controlled through ergonomics should be grouped together; all losses that could be controlled through machine guarding and lock-out, tag-out should be grouped together, etc. Below is an example which causes we might map to the MSD group.

  • Bending
  • Continuous Trauma
  • Lifting
  • Overexertion
  • Pushing or Pulling
  • Reaching
  • Repetitive Motion
  • Twisting

Once you have created the map for all causes you will need to add a look-up column to your data set so that a cause group is available for every claim.

Some RMIS systems will supply a Cause Group or Cause Category field for you. We have not typically been satisfied with the way these groups are set up. You will want to review their existing map to make sure that it works for your purposes

Drilling down

If all has gone well then you will now know what cause groups have been driving frequency and severity over the past three to five years. This alone can go a long way to helping you set priorities. We usually choose to drill down on the top few cause drivers to understand how best to shape any targeted loss mitigation plans.

For example, you will want to filter your results down to only one cause group at a time and ask questions like:

  • What it the trend for this loss driver?
  • What is the percent of total trend for this cause group?
  • What detailed causes are most prevalent within this cause group?
  • What body parts are most frequently affected?
  • Which business units, locations, departments, occupations, etc. are most affected?

Seeing the problem


Data visualization

The way people present data—also known as data visualization—is critical to seeing it clearly. Data visualization is the graphical presentation of multidimensional data so that viewers can understand the underlying structure and relationships hidden in the data. It is what moves data to information, information to knowledge and knowledge to wisdom. We could all use much more wisdom when it comes to loss control. Our employees, managers, executives and shareholders are counting on us to deliver.

Each chart/exhibit has a relationship to the other charts/exhibits that you create from the same dataset but is a different way of looking at it. However, do not discount the importance of viewing the charts/exhibits together to build understanding within your audience about the interaction of the various charts/exhibits.

Volumes have been written on the best ways to visualize data, and the ready-to-use chart templates within Excel are very limiting. Thus, making powerful visualizations will require someone with some knowledge and skill within these topics. To help you along the way, we will share a few visualizations that we have found very impactful below.

Do not discount the importance of viewing the charts together to build understanding within your audience about the interaction of them.

Dashboard reporting

Annual data mining exercises go a long way toward steering your loss-control activities from a strategic perspective. Keeping all of your loss-control stakeholders informed such that the entire enterprise is setting priorities within their own operations appropriately generally requires more frequent and concise communications. To facilitate this type of tactical communication, a dashboard report is often effective.

A dashboard should be easy to read and preferably a single page. It should show graphics of the current status and historical trends of your organization’s key loss-performance indicators. Its goal is to enable quick and informed decisions to be made with only a brief scan.

In particular, it should assist your loss-control stakeholders to decide how they will spend the time, talent and treasure that they will allocate in the most meaningful way to the enterprise.

We generally recommend that dashboards be updated quarterly as it pertains to loss-control efforts. Depending upon the size of your enterprise, several levels of reporting may be needed. For example, one level of reporting may be needed for corporate, another for regions and yet another for individual locations.

A dashboard should be a single, easy-to-read page, graphically showing the current status and historical trends of your organization’s key loss-performance indicators.

Critical steps towards reducing your TCOR start with developing performance metrics based on your loss leaders, establishing goals, and tracking and communicating performance throughout the year.

When the appropriate loss dataset and valuation methods are used, ergonomics and the prevention of MSDs can play a large role in reducing an organization’s TCOR.



About the authors
Eric Kennedy
Eric Kennedy is Senior Risk Control Consultant at Willis Risk & Analytics
Tom Hilgen
Thomas Hilgen is Senior Risk Control Consultant at Willis Risk & Analytics
  • Copyright © 2019 Willis Towers Watson