How do you measure up?
August 1, 2007
Do you cringe when your boss asks how your safety and loss control efforts compare to the market? How truly effective is your return-to-work program? Do you pay more or less than other companies to treat injuries and supplement lost wages?
Sure, you have a mountain of historical loss information, but it reflects just one perspective â€” your situation. And that alone won’t give you the answers to these questions. Without solid comparative data, you’ll never know just how your programs measure up.
The American Heritage Dictionary defines “benchmarking” as the ability to measure a rival’s product according to specified standards in order to compare it with and improve one’s own product. In this case, the product is your loss prevention and control program.
Until recently, questions about the real effectiveness of safety and loss control programs could only be answered by the actuary â€” so secretive was “the data” that access was often kept even from company executives. (We are, after all, speaking about how those actuaries determine your risk; and hence, your insurance premiums.)
Today there are many affordable products companies can license to provide such analysis. When you integrate benchmarking tools into your loss control information system, you’ll be able to:
1. Obtain real-time, on-demand answers to your questions;
2. Identify trends; and
3. Trigger proactive management before problem cases become outliers.
How benchmarking works
The key to the solution requires that you start by thinking like an actuary.
First, categorize your injury/illness data using ICD9 codes. ICD9 codes are published by the World Health Organization and are used in the medical and insurance communities to numerically classify injuries and illnesses. These codes provide much greater detail than the text descriptions used in your injury reporting systems.
For example, ICD9 codes distinguish between open and closed fractures, simple sprains versus disc herniations, etc. This detail is very important when analyzing outcomes. You can license the ICD9 data for a nominal fee and you don’t need to be a clinician to make sense of the codes. But, chances are that your third-party administrator (TPA) or carrier already uses ICD9 codes, so simply request them with your monthly claim loss run reports.
Second, correlate the ICD9 code with the outcome data. The Official Disability Guidelines (ODG), published by the Work Loss Data Institute, compiles claims data from over 3.5 million cases. Using ODG, you’ll know the expected number of days lost and medical/indemnity costs for each claim in your spreadsheet. Drop in some formulas to compute averages and you’ll know just how you measure up with respect to national norms.
Your first milestone
Realize that the only thing you’ve accomplished so far is measuring the effectiveness of your program. Drawing actionable conclusions from this stage of the analysis may be more a result of luck than effective management. Even if you outsource all claims administration, there are internal, external and data factors that influence your outcome. For instance, your service company cannot perform to expectations if there is significant lag time in reporting or if the quality of care provided by practioneers is sub par, or if your data contains outliers.
The real key to successful benchmarking is to use this data to improve your program. You’ve laid the groundwork. Now to take action, you’ll need to analyze a host of other variables before determining root cause. This is where integrating external ODG guidelines into your loss control information system pays off.
Scrub the data
Whether you have 100 claims or 100,000, due diligence demands that you examine outliers. Outliers can simply be bad data or significant indicators. Let’s examine how to distinguish the difference.
One of the biggest sources of outliers involves the ICD9 code assigned to the claim. A single ICD9 may be adequate for the majority of your claims, but many injuries involve multiple body parts. When the guideline is properly integrated into your information system, it should be able to process multiple ICD9 codes, picking the condition that drives the greatest duration of disability and factors in the cost of all conditions.
The information system should also factor in surgeries that are not ICD9-based. For example, a lumbar sprain/strain (ICD9 = 847.2) has an expected return to work outcome of 17 days. But if that claim required lower back surgery, the return to work increases to 142 days. Other factors such as diabetes or a heart condition may warrant exclusion of the case altogether. Lastly, consider age-adjusting the outcome. Younger workers ages 18 to 24 typically recover in half the time of workers ages 55 to 64.
Reclassification of injuries, exclusion of outliers and age-adjustment alone could completely change your evaluation of the data. When the guidelines are integrated into your information system, these adjustments can be performed automatically.
Now that you’ve cleansed your data, you can begin to look for trends. The number of comparisons you create depends on the data in your system. A natural starting point is to compare the outcomes among your facilities. This sheds light on geographic differences and vendor performance. Next, compare department performance within and among facilities. If there is no significant difference among facilities, but like departments show differing outcomes, you may have identified a problem.
In each case, pay attention to the guideline â€” if the outlier department is at or near the norm, you could infer there is no problem as the other department is outperforming the guideline. Many users report being “shocked” when this analysis reveals a wide variation in outcomes among plants or even hospitals that are all part of the same organization.
If you are capturing provider information on the case level you can easily benchmark provider outcomes. Identify the top ten best and worst outcomes by ICD9, then filter by provider. Your results will pinpoint those providers who excel, meet the guideline, or perform poorly when treating specific injuries and illnesses.
Trigger proactive management
The holy grail of a solid benchmarking program is to use the outcome data to trigger proactive management. This is the best way to improve your program. Share your data with plant managers and supervisors, and demonstrate how your analysis grades certain facilities, departments, and work processes.
Ask providers for an expected return-to-work date from the onset of treatment. And if that falls outside the guideline, start asking questions. Set up triggers to flag cases containing problematic ICD9s so you can steer employees to those providers with good track records.
Finally, never forget that the best way to improve your loss control program is prevent injuries from occurring. If you’ve uncovered trends relative to work processes or work experience, share the results with loss prevention and implement the appropriate training programs.
The beauty with this type of benchmarking is that you are dealing with averages â€” something everyone understands. There are a tremendous number of dynamics that inhibit your efforts such as union rules, management support, etc. However, no one can dispute that your efforts are aligned with the best interests of the company and its employees. Once you demonstrate that a department, a work process, or provider consistently delivers poor outcomes with respect to what is considered normal, behaviors will change.
The key is to deliver your message with an emphasis on improving the results rather than disciplining offenders. This is the cornerstone to continuous improvement.
SIDEBAR: Benchmarking rules of thumb• No sample should contain fewer than 12 records. You cannot draw adequate conclusions about what you consider a problem area if you’re analyzing three injuries.
• Don’t over-rely on the published guidelines. They do not factor in outcomes relative to your industry or cost fluctuations in your geographic area.
• Look for consistent, repetitive patterns.
• Keep it simple, always comparing apples to apples.