There is a built-in tendency to over-count the size of productivity and performance impairment in employee populations.  This is because the count is based on the impairment measured in or reported by individual employees, attributed to individual impairment factors, then totaled over all such factors.  Since the vast majority of employees have more than one impairment factors, adding the overall impairment each has for each of the impairment factors each has counts that impairment as many times as each has such factors.

For example, when HealthMedia counted the overall impairment of roughly 175,000 employees working in a number of businesses, it found that impairment linked to eight factors averaged:

•     Unhealthy weight – 3.59% impairment — 64.9% affected
•     Smoking – 5.39% impairment – 13.0% affected
•     Stress – 9.12% impairment – 27.06% affected
•     Depressed Feelings – 9.36% impairment – 28.06% affected
•     Unhealthy Sleep – 3.51% impairment – 79.46% affected
•     Chronic Risk/Disease – 8.82% impairment – 43.09% affected
•     Physical Inactivity – 3.75% impairment – 51.46% affected
•     Poor Nutrition – 1.93% impairment – 42.8% affected

[The raw data come from “HRAs Evolving to Drive Program Participation, Measure Productivity” Disease Management News 11:20 Oct 25, 2006.  The calculations of percent impairment were made by the author.]

If the average impairment per person affected is multiplied by the percentage of employees affected, then the results for each of the eight factors summed, the total impairment would come to 17.55%.  And this is a conservative figure, since HealthMedia only counted impairment above what it estimated as the “normal” impairment of 6.31% unrelated to health factors.  But if the total percentages of employees affected were summed, they would amount to 356.66% of the workforce that are affected by such factors.

This clearly represents significant over-counting of the amount of impairment in the total workforce, as well as the amount due to any one factor, since each impairment reflects how much each individual is impaired, but attributes that total impairment to one factor, when each individual had, on average, more than four.  There are at least three different ways this over-counting can either be avoided or used to adjust the total impairment, though not to identify the impairment actually attributable to each factor, individually.

Dividing the Total by the Degree of Over-Counting

The first method is the simplest.  Since the total impairment sum was 17.55%, and the total prevalence added up to 356.66%, dividing the total impairment by the measured degree of over-counting can yield a good estimate of the actual total impairment.  In this case, dividing 17.55% by 3.5666 = 4.92%, a more likely total, and a more conservative basis for predicting the potential gains from health management (HM) interventions.

Actual potential gains from individual interventions will probably be less than the amount attributed to each, since interventions tend to address only one factor, while individuals affected have an average of more than five. On the other hand, most interventions, while focusing on a single factor, will tend to improve participants’ overall health, so the gains will also tend to be greater than one-fifth the impairment attributed to each separately.

Classifying impairment by the Numbers of Factors Each Individual Has

HealthMedia adjusted for over-counting by separate calculations of the impairment of individual employees based on the numbers of factors each individual reported.  Since each would appear only once in this calculation, this avoids double counting.  It calculated that the combined effects of the seven impairment factors amounted to 2.67% for risk factors, and that of the risk/disease conditions amounted to 2.24% for the population as a whole.  This adds up to 4.92% total impairment, essentially the same figure as the simple adjustment yielded, accounting for rounding effects.

Identifying “Primary” Factors

When the Dupont Chemical Co. counted the overall impairment of its employees, it measured overall losses in absence and presenteeism, together with healthcare costs, as amounting to $7219 per employee.  Figured against a workforce with an average annual compensation of $50,000, this would amount to 11.98%.  [This would be comparable to HealthMedia’s figures if the adjusted total of 4.92% were combined with its “normal” impairment level of 6.31% = 11.23%.

[Calculations were based on: J. Collins, et al. “The Assessment of Chronic Health Conditions on Work Performance, Absence and Total Economic Impact for Employers” JOEM (Journal of Occupational and Environmental Medicine June 2005 547-557; and S. Nicholson, et al.  “How to Present The Business Case for Health Care Quality to Employers” Applied Health Economics and Health Policy  4:4 2005 209-218 – though were made by the author]

Dupont avoided double-counting individual impairment reported by individuals by asking each employee to identify one of the ten impairment conditions identified as the primary cause of each’s impairment.  This would tend to over-count the amount of impairment actually due to any single factor, since most had many, but would be compensated by the fact that other employees were not attributing some of their total impairment to any but one cause.  In any case, the total impairment of each individual would be counted only once, so the total should be accurate.  This should then represent the total potential for an overall HM strategy, even though the potential for individual interventions would likely be over-, or under-stated.

Double-counting is usually less likely in counting healthcare costs.  First of all, the total costs will be known, based on employers’ own self-insured expenses, or its insurer’s experience-rating figures.  And second, medical costs are far more easily attributed to specific causes based on the diagnosis involved, though counting only costs attributable to a single disease can under-estimate the savings potential.  For example, diabetes management typically includes efforts to reduce blood pressure and cholesterol, and promote exercise, which can have wide impact in reducing medical expenses beyond that attributable solely to diabetes.

We need many years of experience dealing with the productivity and performance impairment of many different factors, from chronic diseases to risk behaviors to risk conditions and behaviors that only have impairment effects with no healthcare cost impacts, to identify the true potential of factor-specific HM interventions.  Until then, using one or more of the above adjustments or counting methods can at least ensure that neither HM providers nor their clients exaggerate the overall potential for such interventions.