Ten Modifiable Health Risk Factors and Employees’ Medical Costs

Abstract

Purpose:

To estimate the relationship between employees’ health risks and health-care costs to inform health promotion program design.

Design:

An observational study of person-level health-care claims and health risk assessment (HRA) data that used regression models to estimate the relationship between 10 modifiable risk factors and subsequent year 1 health-care costs.

Setting:

United States.

Participants:

The sample included active, full-time, adult employees continuously enrolled in employer-sponsored health insurance plans contributing to IBM MarketScan Research Databases who completed an HRA. Study criteria were met by 135 219 employees from 11 employers.

Measures:

Ten modifiable risk factors and individual sociodemographic and health characteristics were included in the models as independent variables. Five settings of health-care costs were outcomes in addition to total expenditures.

Analysis:

After building the analytic file, we estimated generalized linear models and conducted postestimation bootstrapping.

Results:

Health-care costs were significantly higher for employees at higher risk for blood glucose, obesity, stress, depression, and physical inactivity (all at P < .0001) than for those at lower risk. Similar cost differentials were found when specific health-care services were examined.

Conclusion:

Employers may achieve cost savings in the short run by implementing comprehensive health promotion programs that focus on decreasing multiple health risks.

Keywords

financial impact risk change and cost change worksite wellness health-care utilization workplace health promotion

Purpose

Half of small firms and 84% of large employers offered health promotion programs in 2019.¹ Despite these programs’ popularity, debate persists about whether they save money. A recent randomized control trial of one employer’s program found that workers at worksites exposed to wellness programs reported an 8.3% higher rate of regular exercise and a 13.6% higher rate of weight management behaviors.² However, the program did not produce cost savings for the employer over 18 months.

Although randomized trials may be the gold standard for testing specific medical treatments, they may not be appropriate for large, multicomponent employer interventions in which a variety of interventions are offered and a change in organizational culture may be needed to achieve population-wide health improvements. Also, results from randomized trials may not be as generalizable as those from observational studies that include more participants and multiple employers.

Observational studies examining single employer programs and those with multiple employers have produced mixed results as to whether these programs are cost neutral or save money.^3
-5 Best practice benchmarking studies have shown that results vary depending on how the particular program is designed and implemented, whether best practices are applied, leadership engagement, participation rates, strategic communication, and an overall culture of health framework.⁶

One particular aspect of program design that may influence the effectiveness of a health promotion program is which health risks are targeted. Employers wish to know which risks may achieve the most health-care cost savings, even in the short run.

Previous studies have found that certain health risks have a stronger relationship with health-care costs than others, namely, high blood glucose, high blood pressure, tobacco use, high stress, and physical inactivity.^7,8 These studies provided estimates of changes in expenditures by tracking individuals’ costs over an average of 3 years following risk measurement. However, those studies did not address the short-term impacts of having certain health risks—specifically, the relationship between health risks and health-care costs in just 1 year following risk measurement.

The current study offers this short-term view of the risk–cost relationship. It responds to employers’ interest in having more contemporaneous data that predict the following year’s health-care costs, which may prove useful in planning targeted workplace wellness programs. Going forward, examining data in 1-year intervals may allow employers and practitioners to recalibrate their interventions relative to the shifting health profile of employees and the effects such shifts may have on benefit program budgets.

Related Literature

Prior studies examined the relationship between modifiable health risks and employee health-care spending in an employed population. In 1998, Goetzel et al¹ collected self-reported behavioral risk and biometric data from 46 026 employees at 6 large companies and linked those data to the workers’ health insurance claims over a 3-year period. That analysis found that employees in 7 of 10 high-risk categories recorded significantly higher health-care costs compared with those at lower risk. Those risks, ordered from highest to lowest cost impact, were depression, stress, high blood glucose, obesity, tobacco use, high blood pressure, and physical inactivity.

A follow-up study by Goetzel et al in 2012 also² considered prescription drug costs when estimating total health-care spending to gain a more complete picture of expenditures borne by employers and employees. The sample included 92 486 workers from 7 companies. Workers again were followed for 3 years after completing a health risk assessment (HRA). Findings from the second and original studies were similar, with depression still the most predictive of future health-care spending, followed by high blood glucose, high blood pressure, obesity, tobacco use, physical inactivity, and high stress.

Other studies examining the risk–cost relationship have focused on a single employer. A study by Kowlessar et al⁹ that leveraged a large employer’s database and spanned 3 years connected health risk and cost data for 63 013 workers. It found that the following risk factors most influenced total health-care spending: high blood glucose, high blood pressure, inadequate exercise, overweight/obesity, and high triglycerides. Henke et al¹⁰ examined health risk information for 11 217 Pepsi Bottling Group employees and linked those records to health-care costs over 2 years. In that analysis, obesity, high blood pressure, high blood glucose, and high cholesterol had the greatest impact on total costs. Goetzel et al¹¹ studied health risks and health-care costs among 5875 Novartis employees over 2 years and found a significant and consistent association among 3 clusters of risk factors (high biometric laboratory values, cigarette and alcohol use, and poor emotional health) and health-care spending. Costs were 10% to 20% higher among employees with high-risk biometric laboratory values and those with emotional health problems than among those at lower risk.

All these studies used similar methods to predict the correlation between having certain risk factors on health-care costs. The risk factors found to be most predictive of health-care costs overlapped in most studies. However, little research has used a multiemployer database to focus on the short-term associations between employee health risks on costs—an issue of importance to employers whose workforce exhibits high turnover rates and whose workers’ tenure with an organization is brief.

Study Purpose

In this study, we used data from 11 large employers included in a multiemployer database to measure the short-term relationship between workers’ health risks and their health-care costs for 10 common modifiable risk factors.

Methods

We leveraged a large multiemployer database, the IBM MarketScan Commercial Database, which contains deidentified person-level and linkable benefits eligibility, medical claims, and prescription drug information, and the MarketScan HRA Database. The MarketScan HRA Database includes self-reported biometric and behavioral risk data collected as part of corporate health and wellness programs. Employees completing HRAs may have been more motivated to improve their health than those who did not.

Health risk assessment responses were linked to administrative claims database, thereby reflecting real-world treatment patterns and costs for individual employees across all providers of care, maintaining health-care utilization and cost records at the patient level. Because these databases contain statistically deidentified data and are fully compliant with US privacy laws and regulations (ie, the Health Insurance Portability and Accountability Act), this study was exempt from institutional review board approval.

Compared with previous studies, our population included more workers from a greater number of organizations with complete health risk and claims data, thus making the findings more generalizable. To ensure improved data quality and integrity, we introduced additional filters to the raw files to guard against data entry errors and intentional reporting of misinformation. We linked employee benefit enrollment files, medical and prescription drug claims, and health risk data over a 1-year period beginning anytime in 2016, when an HRA would have been completed by an employee, and then for at least 12 months thereafter. Because question format and risk cutoff varied in HRA instruments administered by different vendors, we carefully aligned survey items from different instruments to make them comparable in content and scoring.

Sample

The study population included active, full-time employees, aged 18 to 64 years at the time of HRA completion. We excluded Medicare-eligible employees because we did not have access to Medicare claims. Employees had to be continuously enrolled in a self-insured medical plan during the year in which they completed the HRA and for 12 months thereafter. Employees were not required to have submitted claims during the study period to be included in the analysis. Employees who were pregnant at any time during the study period were excluded from the analysis.

Employers who contributed HRA and complete medical and prescription drug data to the MarketScan Commercial Database were included. Employers allowed employees to skip questions when completing the HRA. To ensure generalizability of HRA responses, we required that HRAs be completed by most workers at the organization (ie, no more than 50% of the responses in 9 of 10 health risk categories could be missing for a given risk factor to be included in the analysis). Demographic differences for those who did not take the HRA, or replied to less than 9 health risk categories, can be seen in Supplementary Table A3, enrollees included were more likely to be female and had a lower Charlson comorbidity index (CCI) then those who did not take the HRA.

Further, we excluded companies with unreasonably high (≥99%) or low (≤1%) health risk prevalence rates for any given risk factor. We also set clinical boundaries for all biometric values (ie, height, weight, blood pressure, blood glucose, cholesterol), excluding responses that were below or above a given range considered valid from a medical standpoint.

Health Risks

Respondents to HRA surveys were assigned “at high risk” or “not at high risk” on the basis of the following definitions mirroring national guidelines used in prior studies.¹² Individuals were classified as being at high blood pressure risk if the systolic value was ≥140 mm Hg or diastolic value was ≥90 mm Hg, as were employees with a total cholesterol ≥240 mg/dL. A body mass index (BMI) ≥30 was defined as high risk for body weight, and a value >115 mg/dL was considered high risk for blood glucose. Having a poor diet was defined as eating ≤4 fruits and vegetables on an average day. Employees who spent ≤2 days a week exercising for at least 20 minutes each day were considered physically inactive. Current smokers or current users of any form of tobacco were considered high risk for tobacco use. Male respondents who consumed >14 alcoholic drinks per week and females who consumed >7 per week were classified at high risk for alcohol use. Employees were considered high risk for stress if they responded to the question “How do you currently feel you are coping with life?” by answering either “I sometimes feel stressed and have trouble coping” or “I often feel stressed and have trouble coping.” Finally, depression was flagged when respondents answered that they “Quite often or always” felt depressed, sad, blue, down, or hopeless.

Health-Care Costs

Total health-care costs were defined as allowed amounts reflecting insurer and patient contribution to payment including deductibles, copayments, and coinsurance but not insurance premiums. We annualized expenditures in 2016 by aggregating expenses from the date the employee completed the HRA through the end of the study period or by December 31, 2017, whichever occurred last. If the observation period was <12 months, the employee was removed from the study. If it was >12 months, the spending amount was multiplied by 365 and then divided by the number of days in which the employee was eligible for medical benefits. All dollar amounts reflected actual 2016 to 2017 values. Outlier cases, defined as employees whose annualized costs were equal to or greater than the 99th percentile of each of the 5 measures of health-care costs for the entire study sample, were included in the study and winsorized to the 99th percentile.¹³ Individuals with negative annual claims were set to $0 spending. Figure 1 shows the study timeline.

Figure 1.

Timeline.

The 5 measures of specific health-care services included inpatient (IP), outpatient (OP), emergency department, OP pharmaceutical, and preventive care (PC). Emergency department expenditures included both treat-and-release and treat-and-admit cases. Preventive care services were defined according to the Current Procedural Terminology code on specific claims on the basis of input from health-care experts. Remaining health-care expenditures were categorized as IP if there were individual facility or professional encounters and services associated with a hospital admission (ie, a room-and-board charges were included) or OP for services rendered at a doctor’s office, hospital OP facility, or other OP facility. Outpatient pharmaceutical claims included those delivered by mail order or retail prescription card services. Prescriptions received in an IP setting were included in the IP category.

Employee Characteristics

To estimate relationships between health risks and subsequent health-care costs, we controlled for employee and employment characteristics known to influence costs. Variables included employee age, sex, health plan enrollment type, geographic location of residence, employment category, and a dummy variable representing the specific employer contributing data. Because it is not always clear whether a given risk factor precedes an illness or co-occurs with it, we conducted several sensitivity analyses to control for illnesses present at baseline and disease comorbidities. The disease severity variables included the CCI¹⁴ and Psychiatric Diagnosis Group¹⁵ measures available in, or calculated from, the MarketScan Commercial Database.

Analysis

We specified a generalized linear model to estimate risk-spending relationships. The outcome for all models was total health-care costs and their subset categories post-HRA. The predictor variables included indicators for each health risk measured and the confounding variables listed above.

We compared log normal, log gamma, log Poisson, and log negative binomial with our final model, which used a log link with an underlying normal distribution that yielded residuals with no systematic association with predicted values. A sensitivity analysis compared model estimates that included and excluded outliers. To facilitate interpretation of the nonlinear scale of the model, we applied bootstrapping methods on the risk coefficients and calculated confidence intervals on adjusted costs along with differences among risk groups. Significance tests for the adjusted differences were set at the <.05 level. Detailed model analysis can be seen in Supplementary Tables A1 and A2.

Results

After applying all study inclusion and exclusion criteria, our sample included 135 219 employees from 11 employers. Table 1 displays the characteristics of the study sample. Most (54.8%) workers in the sample were female, and their average age was 43 years. The largest region represented was the South (42.6%). Almost all workers (83.8%) were enrolled in preferred provider organizations plans. They were employed at companies representing manufacturing (44.4%), retail (40.8%), and information and finance (14.7%).

Table 1.

Description of the Study Sample.^aVariable

	Value
Employees, n	135 219
Employers, n	11
Employee characteristics
Female, %	54.8
Age, years, %
18-25	3.6
26-35	26.2
36-45	26.9
46-64	43.3
Salaried, %	42.0
Charlson comorbidity index	0.284
Psychiatric diagnostic groupings	0.224
Region, %
Northeast	23.6
North central	20.9
South	42.6
West	12.9
Health plan type, %
Health maintenance organization	1.8
Preferred provider organization	83.8
High deductible	14.4
Industry type, %
Information and finance	14.7
Manufacturing	44.4
Retail	40.8

^a Source was the authors’ analysis of data for 2016 from the IBM MarketScan Commercial Database.

Table 2 shows the prevalence of the sample’s health risk factors along with their average annual health-care costs, overall and by category. The largest proportion of employees with health risks had poor diets (86.9%), were obese (30.8%) or physically inactive (23.6%), had high stress (18.4%), or were depressed (11.8%). Health risk for employees who were smokers (6.8%) or had high alcohol consumption (1.7%), high blood pressure (9.9%), high total cholesterol (5.9%), or high blood glucose (7.6%) was lower.

Table 2.

Prevalence and Medical Expenditures Associated With Health Risk Factors.^a

Variable	Prevalence, % at High Risk	Annual Health-Care Spending
Risk factor (n = 135 219)
High blood pressure	9.9
High cholesterol	5.9
High blood glucose	7.6
High BMI	30.8
Poor diet	86.9
Tobacco use	6.8
High alcohol consumption	1.7
High stress	18.4
Depression	11.8
Lack of adequate exercise	23.6
Total spending
Mean, $		4182
SD, $		8040
% with $0 spending		4.85
Inpatient spending
Mean, $		854
SD, $		5665
% with $0 spending		96.2
Outpatient spending
Mean, $		1995
SD, $		3046
% with $0 OP spending		7.8
Emergency department spending
Mean, $		353
SD, $		1251
% with $0 spending		86.8
Preventive care spending
Mean, $		170
SD, $		264
% with $0 spending		28.3
Outpatient pharmaceutical spending
Mean, $		810
SD, $		1566
% with $0 spending		16.4

Abbreviations: BMI, body mass index; SD, standard deviation; OP, outpatient.

^a Source was the authors’ analysis of data for 2016 from the IBM MarketScan Commercial Database. Health-care expenditures are in 2016 dollars.

Relationship Between Risks and Costs

Table 3 shows the unadjusted mean dollar amounts for employees at high versus low risk, overall and by medical service category. Positive values represent higher spending for those at high versus lower risk. With the exception of poor diet and high alcohol use, total health-care costs were greater for those at high risk than for those at lower risk. Also, for 8 of 10 categories preventive service costs were greater for high- than for lower-risk employees.

Table 3.

Unadjusted Medical Spending Differentials Between High- and Low-Risk Employees, Overall and by Medical Service.^a,b

Medical Service Category	Health Risk	Spending by Risk Factor, $
Medical Service Category	Health Risk	High Blood Pressure	High Cholesterol	High Glucose	High BMI	Poor Diet	Tobacco Use	High Alcohol Use	High Stress	Depression	Low Exercise
Total	Lower	4150	4169	4021	3738	4269	4180	4184	4027	4010	4046
	High	4475	4395	6044	5178	4169	4208	4107	4873	5357	4969
	Dif	325	226	2023	1440	−100	28	−76	846	1347	923
Inpatient	Lower	833	837	807	704	675	852	854	812	817	833
	High	1040	1129	1392	1189	881	885	831	1041	1080	1247
	Dif	207	292	585	485	206	33	−23	229	264	414
Outpatient	Lower	1995	1996	1957	1850	2246	2004	1995	1937	1919	1923
	High	1994	1981	2411	2321	1957	1868	2001	2252	2539	2183
	Dif	−1	−15	455	471	−288	−136	6	315	620	259
Emergency department	Lower	348	353	344	300	303	345	354	327	334	340
	High	403	359	478	473	361	471	315	471	508	437
	Dif	55	6	134	172	58	126	−39	144	174	97
Retail prescriptions	Lower	802	813	744	712	849	807	810	779	770	778
	High	883	757	1595	1029	804	842	781	945	1057	936
	Dif	81	−56	851	318	−45	35	−29	166	287	157
Preventive	Lower	172	170	170	172	197	172	170	172	169	171
	High	155	169	168	166	166	142	179	165	172	167
	Dif	−17	−1	−2	−6	−31	−31	9	−7	3	−4

Abbreviations: BMI, body mass index; Dif, difference.

^a Source was the authors’ analysis of data for 2016 from the IBM MarketScan Commercial Database.

^bSpending categories (except Total) are mutually exclusive.

Table 4 displays the total and service category values of health-care costs for employees with and without a specified risk factor, adjusted for confounders. Employees at high risk for high blood glucose, obesity, stress, depression, and physical inactivity were significantly more costly than those at lower risk. Similar cost differentials were found when examining service types.

Table 4.

Adjusted Medical Spending Differentials Between High- and Low-Risk Employees.^a,b

Medical Service Category	Health Risk	Blood Pressure	Cholesterol	Glucose	BMI	Diet	Tobacco Use	Alcohol	Stress	Depression	Exercise
All health	Lower	4215	4206	4058	3873	4371	4211	4205	4078	4101	4176
	High	4079	4112	5752	4873	4175	4056	3956	4720	4717	4597
	Dif	−136	−94	1694^c	1000^c	−195	−155	−248	642^c	616^c	421^c
Inpatient	Lower	853	844	823	750	853	863	858	836	844	877
	High	908	1076	1213	1072	859	803	885	950	889	1115
	Dif	56	232^c	390^c	322^c	6	−60	27	114^c	45	238^c
Outpatient	Lower	2018	2012	1969	1899	2135	2015	2007	1954	1959	1973
	High	1885	1884	2374	2229	1983	1862	1853	2224	2292	2059
	Dif	−133^c	−128^c	405^c	329^c	−151^c	−153^c	−154^c	270^c	333^c	86^c
Emergency department	Lower	349	352	346	314	357	346	353	336	340	357
	High	374	348	428	426	351	424	286	411	433	380
	Dif	24^d	−4	82^c	112^c	−6	78^c	−67^d	75^c	93^c	23^d
Retail prescriptions	Lower	816	815	741	734	821	807	808	784	784	796
	High	742	688	1578	953	805	806	753	910	916	860
	Dif	−74^c	−128^c	837^c	219^c	−16	−1	−55	125^c	133^c	63^c
Preventive	Lower	173	172	171	173	174	173	171	170	170	171
	High	153	157	171	166	170	141	172	175	170	166
	Dif	−20^c	−14^c	1	−8^c	−3	−32^c	1	5^d	0	−5^d

Abbreviations: BMI, body mass index; Dif, difference.

^a Overall and by medical service spending by risk factor, $.

^b Source was the authors’ analysis of data for 2016 from the IBM MarketScan Commercial Database.

^c Statistical significance of the adjusted difference at P < .001.

^d Statistical significance of the adjusted difference at P < .05.

Table 5 summarizes these analyses and presents differences in costs as percentage values. As shown, statistically significant differences in costs were found for employees with high blood glucose levels—high-risk employees were 41.8% more expensive than were those at lower risk. Additionally, obese employees were 25.8% more expensive than nonobese employees; costs for those at high risk for stress and depression were 15.7% and 15.0% higher, respectively; and expenses for those not exercising enough were 10.1% higher.

Table 5.

Average Adjusted and Unadjusted Medical Expenditures by Risk Level.^a

Risk Measure and Level	Unadjusted Mean, $	Adjusted Mean, $	Unadjusted Difference, %	Adjusted Difference, %
Blood glucose
High	6044	5752	50.3	41.8^b
Lower	4021	4058	-^c	-^c
BMI
High	5178	4873	38.5	25.8^b
Lower	3738	3873	-^c	-^c
Stress
High	4873	4720	21.0	15.7^b
Lower	4027	4078	-^c	-^c
Depression
High	5357	4717	33.6	15.0^b
Lower	4010	4101	-^c	-^c
Poor exercise
High	4969	4597	22.8	10.1^b
Lower	4046	4176	-^c	-^c
Total cholesterol
High	4395	4112	5.4	−2.2
Lower	4169	4206	-^c	-^c
Blood pressure
High	4475	4079	7.8	−3.2
Lower	4150	4215	-^c	-^c
Tobacco use
High	4208	4056	0.7	−3.7
Lower	4180	4211	-^c	-^c
Poor diet
High	4169	4175	−2.3	−4.5
Lower	4269	4371	-^c	-^c
Alcohol consumption
High	4107	3956	−1.8	−5.9
Lower	4184	4205	-^c	-^c

Abbreviation: BMI, body mass index.

^a Source was the authors’ analysis of data for 2016 from the IBM MarketScan Commercial Database.

^b Statistical significance of the adjusted difference at P < .001 level.

^c Reference category.

Table 6 provides a population-level cost impact analysis when considering both positive and negative cost drivers regardless of statistical significance. The summary considers the incremental cost of each risk factor multiplied by the number of individuals with the given risk factor in the sample, thus summarizing the overall and per capita cost for each risk relative to overall expenditures.

Table 6.

Estimated Effect of Health Risks on Annual per Capita Medical Expenditures.^a

Risk Factor	Estimated Annual Effect per High-Risk Person, $	Prevalence: People With High Risk, n	High-Risk Group Annual Effect
Risk Factor	Estimated Annual Effect per High-Risk Person, $	Prevalence: People With High Risk, n	Cost, $	Total Annual Expenditures, %	Cost per Capita, $
BMI	1000	41 683	41 700 340	7.4	308.4
Glucose	1694	10 257	17 380 137	3.1	128.5
Stress	642	24 838	15 940 519	2.8	117.9
Exercise	421	26 287	11 074 477	2.0	81.9
Depression	616	15 606	9 612 803	1.7	71.1
Alcohol consumption	−248	2297	−570 388	−0.1	−4.2
Cholesterol	−94	7442	−699 940	−0.1	−5.2
Tobacco use	−155	9247	−1 431 378	−0.3	−10.6
Blood pressure	−136	13 321	−1 812 421	−0.3	−13.4
Diet	−195	117 538	−22 975 546	−4.1	−169.9

Abbreviation: BMI, body mass index.

^a Source was the authors’ analysis of data for 2016 from the IBM MarketScan Commercial Database.

Because of the large number of obese employees, the largest cost driver was high BMI, which contributed to an increase of $308 in annual per capita costs (across the entire population). When considering only other statistically significant findings, the following risk factors markedly increased annual per capita costs: high blood glucose ($129 increase), stress ($118), depression ($71), and physical inactivity ($82). Three risk factors resulted in lower per capita spending, although none significantly, poor diet had the largest negative impact at $170. On an aggregate basis and considering only statistically significant increases or decreases in total health-care spending along with the prevalence of the risk factors examined, the net additional costs for the 10 risk factors studied was $708 or 16.9% of total spending. This percentage amount is likely an underestimate given the many other risk factors affecting costs, which were not measured.

Discussion

Our study confirms previous research findings that employees with modifiable health risks (ie, high blood glucose, obesity, stress, depression, and physical inactivity) are significantly more costly than those at lower risk. These results suggest that employers, who pay a share of employee health-care costs, have a financial interest in improving the health of their workers. In addition, employers who measure employees’ risk profile on a periodic basis may use these data to identify population segments most likely to incur higher medical costs in the near term, thus providing an opportunity for targeted interventions.

Employers designing new programs or refining existing programs may want to consider focusing on the 5 health risks with the highest short-term cost increases to maximize the potential for medical savings. Having a poor diet, which is a focus for many health promotion programs, was not linked to increased health-care spending. However, poor diet is a known risk factor for many diseases including cancer, type 2 diabetes, heart disease, and related disorders that result in premature mortality.¹⁶ Despite ongoing debate, most experts agree that a plant-based diet that is high in fruits and vegetables, fiber, and unprocessed foods and low in sugar and sodium is ideal. But measurement of healthy diets is complex and often unreliable.

Our findings are directionally similar to those of prior studies, although the cost impacts were somewhat muted compared with the earlier analyses. It may be that short-term impacts are less pronounced, and a longer time horizon is needed to determine the cumulative cost effects of being at high risk for poor outcomes. Indeed, the health habits and biometric measures observed typically require several years to take hold, but once they are part of a person’s behavioral or biometric footprint, they may be difficult and expensive to change. Other studies that have followed workers’ risk and cost profile over several years have shown that costs are higher for individuals who move from high to low risk than for those who remain at high or low risk.^17,18 We reason that the accumulation of poor lifestyle habits takes years, and additional years are required to actualize the ameliorative effects of improved health.

Separating the risk–cost relationship findings from other observations from the data, we found other results to be sobering. For example, the analysis of health risk data for our very large sample of 135 219 workers revealed that almost all employees (86.9%) reported having poor diets, although the risk was defined narrowly as consuming fewer than 5 fruits and vegetables a day. Not considered in that definition was the amount of food consumed; whether the diet is vegetarian or meat-based, processed, or organic; the time of day when, or pace at which, food is consumed; whether the food is homemade or restaurant purchased; and whether the food contains large amounts of sugar or sodium. Although most agree that adequate fruit and vegetable intake is an important marker of a healthy diet, it is by no means the only barometer of healthy eating. Numerous studies have shown that self-reports of eating habits are unreliable and often reflect aspirational rather than actual eating habits.¹⁹ Previous studies have reported similar counterintuitive findings, further underscoring the need for better validation of self-reported diet as reported in HRAs.

It was also notable how many employees reported having other modifiable health risks. Nearly a third of workers (30.8%) were obese, 7.6% had high blood glucose levels, and almost a quarter (23.6%) were physically inactive. Also noteworthy were the psychosocial predictors of higher costs, including high levels of stress (18.4% of workers) and depression (11.8%). Taken together, the 5 risk factors with significant correlation with higher health-care spending represented 16.9% of total expenditures.

Recent studies of workplace health promotion programs at BJ’s Wholesale Club²⁰ and the University of Illinois²¹ have called into question the value of workplace programs in reducing the incidence of chronic diseases and containing health-care costs. Although these studies used randomized trials whereby individual workers or worksites were randomly assigned into treatment or control conditions, they were subject to significant limitations common in real-world evaluations of programs. Both studies followed workers over a relatively short period—approximately 12 to 18 months. The short-term effects of health risk factors on disease and cost outcomes that we found were less pronounced than those found in previous work with longer program exposures.

Cutler et al²² showed that Medicare expenditures for seniors had slowed dramatically since 2005, and as much as half of that reduction may be attributed to reduced spending on cardiovascular disease largely due to patients following preventive medicine advice and taking medications to prevent or control heart disease. As the authors discussed in the article, preventive care may help people live longer and healthier lives while exerting a dampening on health-care cost trends.

This study, and those preceding it, is not enough to support a conclusive business case for workplace health promotion in any particular company. To do so, companies must demonstrate that programs, policies, and environmental supports that make up their workplace culture are sufficiently potent to change health behaviors and achieve population-wide risk reduction, which eventually may result in cost savings. In the past 4 decades, a growing body of evidence has emerged suggesting that specific individual and organizational health promotion practices that fall under the broad banner of health and well-being programs and supported by a culture of health may achieve population health improvements, cost savings, and productivity gains.²³ To strengthen the business case for workplace health promotion, more research like the current study is needed to identify the most costly health risks, the most impactful programs, and the most cost-effective programs, individually or in combination.²⁴

Future research is needed to further explore the relationship between organizational health and employees’ health risk profile, medical expenditures, and other relevant business outcomes such as its stock price.^25,26 Good measures are becoming available to evaluate the organizational health of a company, which can capture such elements as its culture, leadership support, programming elements, and feedback loops.^27
-29 Among the hypotheses to be tested is whether after controlling for confounders, healthy company cultures improve employees’ health risk profiles and reduce medical spending. Early results suggest that improving a company’s culture of health predicts reductions in employees’ health risks and health-care utilization.³⁰

This study has several strengths and limitations. Unlike previous analyses, we accessed data from multiple health promotion vendors used by employers. This presented a challenge of harmonizing responses to the variety of HRA tools administered and operationally defining high risk across instruments. However, expanding the number of HRA vendors supplying data for our analysis allowed us to increase our sample size.

As is the case in most studies that leverage previously collected HRA data, we used self-reported behavior and biometric values to define health risks, which may be unreliable, although some studies have found evidence that self-reporting can be an appropriate substitute for administrative data.³¹ While we required HRA responses to be collected from at least 50% of the population, our sample may be biased because only respondents who completed an HRA were included.

Our analysis was cross-sectional meaning that no cause-effect relationship could be established. Studies such as those by Nyce et al³² and Carls et al³³ that examine changes in risk as these correspond to changes in costs are instructive in terms of establishing closer links between behavior change and cost savings.

Finally, the employers from whom the sample was drawn are large and likely to offer generous medical benefits. Because small and medium-sized employers are not well represented in the MarketScan Commercial Database, our results may not be applicable to smaller businesses that represent the majority of employers in the United States.

Employers wishing to maximize their potential for health-care cost savings may view our findings as instructive on how to prioritize their investments in worker health and well-being. Future work should continue to evaluate the most important ingredients of wellness programs to further distill the right recipe for success.

SO WHAT?

What is already known about this topic?

Prior research into the relationship between employee health risk factors and future health-care spending found significant associations between certain modifiable health risks and medical costs.

What does this article add?

We revisited this research by examining more contemporaneous data collected in 2016 and 2017 and focused on the short-term (1 year) relationships between risks and medical costs.

What are the implications for health promotion practices or research?

Employers wishing to maximize their potential for health-care cost savings may choose to focus first on reducing health risks with the largest association with medical expenditures: high blood glucose, obesity, depression, stress, and physical inactivity.

Supplemental Material

Supplemental Material, sj-docx-1-ahp-10.1177_0890117120917850 - Ten Modifiable Health Risk Factors and Employees’ Medical Costs—An Update

Supplemental Material, sj-docx-1-ahp-10.1177_0890117120917850 for Ten Modifiable Health Risk Factors and Employees’ Medical Costs—An Update by Ron Z. Goetzel, Rachel Mosher Henke, Michael A. Head, Richele Benevent and Kyu Rhee in American Journal of Health Promotion

Footnotes

Acknowledgments

The authors would like to acknowledge Frank Yoon for his statistical expertise and Mary Beth Schaefer and Paige Jackson for their editorial assistance.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iDs

Ron Z. Goetzel, PhD

Michael A. Head, MS

Supplemental Material

Supplemental material for this article is available online.

References

Kaiser Family Foundation. 2019 Employer health benefits survey; 2019. https://www.kff.org/health-costs/report/2019-employer-health-benefits-survey/. Accessed December 10, 2019.

Song

Baicker

. Effect of a workplace wellness program on employee health and economic outcomes: a randomized clinical trial. JAMA. 2019;321(15):1491–1501.

Nyman

Abraham

Jeffery

Barleen

. The effectiveness of a health promotion program after 3 years: evidence from the University of Minnesota. Med Care. 2012;50(9):772–778.

Caloyeras

Liu

Exum

Broderick

Mattke

. Managing manifest diseases, but not health risks, saved PepsiCo money over seven years. Health Aff (Millwood). 2014;33(1):124–131.

Baicker

Cutler

Song

. Workplace wellness programs can generate savings. Health Aff (Millwood). 2010;29(2):304–311.

Goetzel

Pronk

. Worksite health promotion: how much do we really know about what works? Am J Prev Med. 2010;38(2):S223–S225.

Goetzel

Anderson

Whitmer

, et al. The relationship between modifiable health risks and health care expenditures. J Occup Environ Med. 1998;40(10):843–854.

Goetzel

Pei

Tabrizi

, et al. Ten modifiable health risks factors are linked to more than one-fifth of employer-employee health care spending. Health Aff (Millwood). 2012;31(11):2074–2484.

Kowlessar

Goetzel

Carls

Tabrizi

Guindon

. The relationship between 11 health risks and medical and productivity costs for a large employer. J Occup Environ Med. 2011;53(5):468–477.

10.

Henke

Carls

Short

, et al. The relationship between health risks and health and productivity costs among employees at Pepsi Bottling Group. J Occup Environ Med. 2010;52(5):519–527.

11.

Goetzel

Carls

Wang

, et al. The relationship between modifiable health risk factors and medical expenditures, absenteeism, short-term disability and presenteeism among employees at Novartis. J Occup Environ Med. 2009;51(4):487–499.

12.

Johns Hopkins Medicine. Updated cardiovascular disease prevention guidelines announced. 2019. https://www.hopkinsmedicine.org/news/newsroom/news-releases/2019-updated-cardiovascular-disease-prevention-guidelines-announced. Accessed February 17, 2020.

13.

Ghosh

Vogt

Outliers: an evaluation of methodologies. Paper presented at: Joint Statistical Meetings 2012 ; August 1, 2012; San Diego, CA.

14.

Charlson

Pompei

Ales

MacKenzie

. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis. 1987;40(5):373–383.

15.

Ashcraft

Fries

Nerenz

, et al. A psychiatric patient classification system. An alternative to diagnosis-related groups. Med Care. 1989;27(5):543–557.

16.

Wang

Chiuve

Willett

. Improvements in US diet helped reduce disease burden and lower premature deaths, 1999–2012; overall diet remains poor. Health Aff (Millwood). 2015;34(11):1916–1922.

17.

Carls

Goetzel

Henke

Bruno

Isaac

McHugh

. The impact of weight gain or loss on health care costs for employees at the Johnson & Johnson Family of Companies. J Occup Environ Med. 2011;53(1):8–16.

18.

Rolando

Byrne

McGown

Goetzel

Elasy

Yarbrough

. Health risk factor modification predicts incidence of diabetes in an employee population results of an 8-year longitudinal cohort study. J Occup Environ Med. 2013;55(4):410–415.

19.

Archer

Hand

Blair

. Validity of U.S. nutritional surveillance: National Health and Nutrition Examination Survey caloric energy intake data, 1971-2010. PLoS One. 2013;8(10):e76632.

20.

Song

Baicker

. Effect of a workplace wellness program on employee health and economic outcomes: a randomized clinical trial. JAMA. 2019;321(15):1491–1501.

21.

Jones

Molitor

Reif

. What do workplace wellness programs do? Evidence from the Illinois workplace wellness study. Q J Econ. 2019;134(4):1747–1791.

22.

Cutler

Karaca

Henke

Head

Wong

. The effects of Medicare Accountable Organizations on inpatient mortality rates. Inquiry. 2018;55:46958018800092.

23.

Soler

Leeks

Razi

, et al. A systematic review of selected interventions for worksite health promotion the assessment of health risks with feedback. Am J Prev Med. 2010;38(2 suppl):S237–S262.

24.

Goetzel

. Perspective: do prevention or treatment services save money? The wrong debate. Health Aff (Millwood). 2009;28(1):37–41.

25.

Henke

Head

Kent

Goetzel

Roemer

McCleary

. Improvements in an organization’s culture of health reduces workers’ health risk profile and health care utilization. J Occup Environ Med. 2019;61(2):96–101.

26.

Goetzel

Fabius

Roemer

, et al. The stock performance of American companies investing in a culture of health. Am J Health Promot. 2019:33(3):439–447.

27.

Roemer

Kent

Mummert

, et al. Validity and Reliability of the updated CDC Worksite Health ScoreCard. J Occup Environ Med. 2019. 61(9):767–777.

28.

HERO. All resources. https://hero-health.org/resources/allresources/. Accessed October 16, 2019.

29.

Workplace Health Achievement Index. American Heart Association Website. 2019. https://www.heart.org/en/professional/workplace-health/workplace-health-achievement-index. Accessed November 19, 2019.

30.

Henke

Head

Kent

Goetzel

Roemer

McCleary

. Improvements in an organization’s culture of health reduces workers’ health risk profile and health care utilization. J Occup Environ Med. 2019;61(2):96–101.

31.

Short

Goetzel

Pei

, et al. How accurate are self-reports? Analysis of self-reported health care utilization and absence when compared with administrative data. J Occup Environ Med. 2009;51(7):786–796.

32.

Nyce

Grossmeier

Anderson

, et al. Association between changes in health risk status and changes in future health care costs: A multiemployer study. J Occup Environ Med. 2012; 54(11):1364–1373.

33.

Carls

Goetzel

Henke

Bruno

Isaac

McHugh

. The impact of weight gain or loss on health care costs for employees at the Johnson & Johnson family of companies. J Occup Environ Med. 2011;53(1):8–16.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.08 MB