SEARCH

SEARCH BY CITATION

Keywords:

  • body mass index;
  • cost and cost analysis;
  • economics;
  • healthcare costs;
  • managed care

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Research Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

Objective: To assess the relationship between body mass index (BMI) and future healthcare costs.

Research Methods and Procedures: We undertook a retrospective cohort study of the relationship between obesity and future healthcare costs at Kaiser Permanente Northwest Division, a large health maintenance organization in Portland, Oregon. Study subjects (n = 1286) consisted of persons who responded to a 1990 health survey that was mailed to a random sample of adult Kaiser Permanente Northwest Division members who were 35 to 64 years of age; had a BMI ≥ 20 kg/m2 (based on self-reported height and weight); did not smoke cigarettes; and did not have a history of coronary heart disease, stroke, human immunodeficiency virus, or cancer. Subjects were stratified according to their BMI in 1990 (20 to 24.9, 25 to 29.9, and ≥30 kg/m2; n = 545, 474, and 367, respectively). We then tallied their costs (in 1998 US dollars) for all inpatient care, outpatient services, and prescription drugs over a 9-year period (1990 through 1998).

Results: For persons with BMIs of 20 to 24.9 kg/m2, mean (±SE) annual costs of prescription drugs, outpatient services, inpatient care, and all medical care averaged $261 (±18), $848 (±59), $532 (±85), and $1631 (±120), respectively, over the study period. Cost ratios (95% confidence intervals) for persons with BMIs of 25 to 29.9 kg/m2 and ≥30 kg/m2, respectively, were 1.37 (1.12 to 1.66) and 2.05 (1.62 to 2.55) for prescription drugs, 0.96 (0.83 to 1.13) and 1.14 (0.97 to 1.37) for outpatient services, 1.20 (0.81 to 1.86) and 1.38 (0.91 to 2.14) for inpatient care, and 1.10 (0.91 to 1.35) and 1.36 (1.11 to 1.68) for all medical care.

Discussion: Future healthcare costs are higher for persons who are overweight, especially those with BMIs ≥ 30 kg/m2.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Research Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

Large-scale epidemiological cohort studies have established that excess body weight (as assessed by body mass index [BMI]) is an important risk factor for a range of chronic disease conditions, including coronary heart disease (CHD), type 2 diabetes, hypertension, selected cancers, and musculoskeletal disorders (1). The relationship between BMI and future healthcare costs, however, has received little attention. This issue is of increasing importance in the wake of recent reports that the average BMI of US adults has increased over the past two decades (2) and that more than one-half of US adults are now overweight or obese (3). In this study, we examine the relationship between BMI and future healthcare costs among the members of a large managed care plan.

Research Methods and Procedures

  1. Top of page
  2. Abstract
  3. Introduction
  4. Research Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

Study Design

Using a retrospective cohort study design, we examined the relationship between BMI and the cost of future healthcare services among people who were members of Kaiser Permanente Northwest Division (KPNW), a not-for-profit group-model health maintenance organization with a membership of ∼450,000 people in the Portland, Oregon metropolitan area. Study subjects were respondents to a 1990 health survey (which elicited information on height and weight) that was mailed to a random sample of KPNW adult members. We stratified survey respondents according to their BMI in 1990 (20 to 24.9, 25 to 29.9, and ≥30 kg/m2) and examined differences in their use and the cost of healthcare services over a 9-year period (1990 through 1998). Subjects in the two highest BMI strata (i.e., 25 to 29.9 kg/m2 and ≥30 kg/m2) were compared with those in the lowest stratum (i.e., 20 to 24.9 kg/m2).

Study Subjects

Study subjects were selected from 7021 adults (56.3%) who responded to a 1990 health survey that was mailed to a random sample (n = 12,474) of the KPNW membership. The survey was administered during the months of May through December. Survey respondents were included in our study if they provided information on their height and weight; were between the ages of 35 and 64 years at the time of survey; had a BMI ≥ 20 kg/m2 at the time of survey; reported that they did not smoke cigarettes; were enrolled in KPNW for at least 1 year before the date of survey and at least 1 month thereafter; if female, did not give birth between January 1, 1990 and June 1, 1991; had no medical encounters in the 3-year period before the survey in which malignancy (International Classification of Diseases [CD]-9-Clinical Modification [CM] 140–208, 230–234), human immunodeficiency virus (HIV), or acquired immunodeficiency syndrome (AIDS) (ICD-9-CM 042) was noted; had pharmacy benefits throughout their entire period of enrollment in the plan; and were not hospitalized for CHD (ICD-9-CM 410–414) or stroke (ICD-9-CM 431–434, 436) in the 3-year period before the survey.

Data Sources

For each study subject, information on medical history as of the date of survey and on the use of healthcare services from 1990 through 1998 was obtained from the electronic records of the plan. These included records of all prescription drugs and supplies (e.g., syringes and testing strips) dispensed by KPNW outpatient pharmacies, ambulatory services provided within and outside the plan, and all hospital admissions. KPNW enrollment records were used to ascertain dates of enrollment and disenrollment (if applicable) from the plan as well as deaths.

Cost estimates were obtained from a variety of sources. Pharmacy costs for prescription drugs and supplies were estimated based on local retail prices at the time of purchase. Algorithms developed by KPNW's Center for Health Research were used to assign costs to all outpatient and inpatient services provided by the health plan (4). Outpatient visit costs include direct and overhead expenses as well as laboratory services. Hospital inpatient costs reflect costs related to the provision of care in four service departments (critical care, routine care, surgery, and surgical recovery) plus overhead expenses. In instances in which study subjects received services from out-of-plan providers, amounts paid by KPNW were used to estimate costs. All costs were adjusted to 1998 price levels using the Medical-Care Component of the Consumer Price Index for All Urban Consumers.

Study Measures

For the period 1990 through 1998, we examined the use and cost of all pharmacy services, including dispenses of prescription drugs and supplies; outpatient services, including visits to physician offices, clinics, and hospital outpatient facilities; and hospital inpatient services. In analyses of healthcare costs, we also examined the total across these three categories.

We stratified pharmacy services by therapeutic category as follows: diabetes medications (i.e., insulin or oral antihyperglycemics); cardiovascular medications (i.e., antihypertensives, antihyperlipidemics, antiarrhythmics, and antianginal agents); and all other medications. We similarly stratified outpatient services by provider specialty (i.e., primary versus specialty care).

Data Analyses

We examined the demographic and clinical characteristics of study subjects in 1990, including their age, sex, race, and prestudy use of selected prescription drugs (diabetes medications, antihypertensives, and lipid-lowering agents). The statistical significance of observed differences was assessed using an F test for continuous measures and a χ2 test for categorical measures.

To compare use and costs across BMI strata, we calculated annualized means for each measure of interest, weighting the observation for each subject by his or her number of months of continuous enrollment during the 9-year period of follow-up. We then computed rate ratios for subjects with BMIs of 25 to 29.9 kg/m2 and ≥30 kg/m2 relative to those with BMIs of 20 to 24.9 kg/m2 based on their respective weighted means. We estimated 95% confidence intervals (95% CIs) for these ratios using methods developed by Fieller (5) and Willan and O'Brien (6). We also performed regression analyses to compare annual levels of use and costs across BMI strata controlling for differences in gender and age.

Finally, using techniques described by Lin et al. (7) for analyzing healthcare costs when data are censored, we compared cumulative healthcare costs across BMI strata over the period 1990 through 1998. Using this methodology, Kaplan–Meier estimates of the probability of survival to each month of follow-up were calculated for subjects in each BMI group (8). The expected cost in each BMI stratum was calculated by summing the products of these monthly survival probabilities and the corresponding mean monthly costs of care. We also estimated 95% CIs for the cumulative cost measures (7).

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Research Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

Sample Characteristics

A total of 1286 respondents to the 1990 survey met all criteria for inclusion in the study (Table 1). More than one-half of study subjects (n = 741) were overweight or obese (BMI ≥ 25 kg/m2; Table 2). Subjects with BMIs of 25 to 29.9 kg/m2 were slightly older and more likely to be men than those with lower or higher BMIs. Baseline use of diabetes medications and antihypertensive agents was higher for those with higher BMIs. The mean duration of follow-up was 88.5 months; it did not differ significantly according to BMI (p = 0.176).

Table 1.  Attrition of KPNW plan members in sample selection process
Inclusion criteriaSample size
1. Responded to 1990 member health survey7021
2. (1) And had nonmissing self-reported height and weight6111
3. (2) And were 35 to 64 years of age at survey3413
4. (3) And had BMI ≥ 20 kg/m23234
5. (4) And were enrolled in plan for ≥1 year before and ≥1 month after study initiation2202
6. (5) And, if female, did not give birth between January 1, 1990 and June 1, 19912195
7. (6) And did not have a history of malignancy (ICD-9-CM 140–208, 230–234) or HIV, or AIDS (ICD-9-CM 042) in the 3-year period before survey2080
8. (7) And had pharmacy benefits throughout period of enrollment in plan1375
9. (8) And were not hospitalized for CHD (ICD-9-CM 410–414) or stroke (ICD-9-CM 431–434,436) in the 3-year period before survey1286
Table 2.  Characteristics of study subjects by BMI
 BMI (kg/m2) 
Characteristic20 to 24.925 to 29.9≥30p
Number of subjects545474267 
Gender (% female)65.0%44.3%58.4%<0.001
Race (% white)95.0%95.5%96.6%0.586
Mean age46.848.948.1<0.001
Age group (%)    
35 to 44 years48.4%37.6%41.2%0.012
45 to 54 years28.6%34.0%35.6% 
55 to 64 years22.9%28.5%26.2% 
Use of selected medications (%)    
Diabetes medications1.1%1.5%6.0%<0.001
Antihypertensives2.9%5.7%13.5%<0.001
Lipid-lowering agents2.0%3.2%4.1%0.217

BMI and Annual Healthcare Use and Costs

Use of healthcare services generally increased with BMI. Relative to persons with BMIs of 20 to 24.9 kg/m2, those with BMIs of 25 to 29.9 kg/m2 and ≥30 kg/m2, respectively, had 1.20 (95% CI: 1.04 to 1.38) and 1.84 (95% CI: 1.56 to 2.16) times the annual number of pharmacy dispenses, including 1.45 (95% CI: 0.54 to 6.77) and 6.08 (95% CI: 2.88 to 27.78) times the number of dispenses of diabetes medications and 2.04 (95% CI: 1.54 to 2.72) and 3.44 (95% CI: 2.57 to 4.65) times the number of dispenses of cardiovascular medications (Table 3). Relative to subjects in the lowest BMI stratum, the number of outpatient visits to primary-care providers was 1.12 (95% CI: 1.02 to 1.23) and 1.38 (95% CI: 1.23 to 1.54) times higher among those with BMIs of 25 to 29.9 kg/m2 and ≥30 kg/m2, respectively; however, a similar pattern was not observed in visits to specialists. The annual number of inpatient days increased with BMI, but the differences relative to persons in the lowest BMI stratum were not statistically significant.

Table 3.  Annual use of medical care services in relation to BMI by type of service, 1990–1998
  Rate ratio (95% CI) by BMI (kg/m2)
MeasureWeighted* mean (SE), BMI = 20.0 to 24.9 kg/m225.0 to 29.9≥30
  • *

    All measures weighted by number of months of follow-up; annualized utilization (SE) presented for subjects with a BMI of 20.0 to 24.9 kg/m2.

  • Rate ratios based on weighted mean annualized utilization relative to subjects with a BMI of 20 to 24.9 kg/m2.

  • See text for definitions.

Number of pharmacy dispenses7.66 (0.38)1.20 (1.04–1.38)1.84 (1.56–2.16)
Diabetes medications0.13 (0.05)1.45 (0.54–6.77)6.08 (2.88–27.78)
Cardiovascular medications0.84 (0.09)2.04 (1.54–2.72)3.44 (2.57–4.65)
All other medications6.69 (0.34)1.09 (0.95–1.26)1.56 (1.30–1.84)
Number of outpatient visits5.70 (0.19)0.98 (0.90–1.08)1.16 (1.03–1.30)
Primary care1.75 (0.06)1.12 (1.02–1.23)1.38 (1.23–1.54)
Specialty care3.95 (0.15)0.93 (0.83–1.03)1.07 (0.93–1.22)
Number of inpatient days0.16 (0.03)1.19 (0.77–2.03)1.49 (0.88–2.63)

Although differences in costs generally mirrored those in healthcare use, BMI-related increases in pharmacy costs actually exceeded those of pharmacy dispenses, particularly for diabetes medications (Table 4). Total annual costs were 1.10 (95% CI: 0.91 to 1.35) and 1.36 (95% CI: 1.11 to 1.68) times higher among subjects with BMIs of 25 to 29.9 kg/m2 and ≥30 kg/m2, respectively, relative to those with BMIs of 20 to 24.9 kg/m2.

Table 4.  Annual cost of medical care services in relation to BMI by type of service, 1990–1998 (1998 US dollars)
  Rate ratio (95% CI) by BMI (kg/m2)
MeasureWeighted* mean (SE), BMI = 20.0 to 24.9 kg/m225.0 to 29.9≥30
  • *

    All measures weighted by number of months of follow-up; annualized costs (SE) presented for subjects with a BMI of 20.0 to 24.9 kg/m2.

  • Rate ratios based on weighted mean annualized costs relative to subjects with a BMI of 20 to 24.9 kg/m2.

  • See text for definitions.

Pharmacy costs$261 (18)1.37 (1.12–1.66)2.05 (1.62–2.55)
Diabetes medications$3 (1)3.38 (1.24–12.80)13.25 (6.40–48.68)
Cardiovascular medications$36 (5)2.38 (1.70–3.39)3.42 (2.40–4.94)
All other medications$222 (17)1.17 (0.94–1.46)1.67 (1.25–2.16)
Outpatient visit costs$848 (59)0.96 (0.83–1.13)1.14 (0.97–1.37)
Primary care$260 (9)1.13 (1.03–1.24)1.39 (1.24–1.56)
Specialty care$587 (56)0.89 (0.73–1.11)1.03 (0.83–1.32)
Inpatient costs$532 (85)1.20 (0.81–1.86)1.38 (0.91–2.14)
Total costs$1641 (120)1.10 (0.91–1.35)1.36 (1.11–1.68)

The pattern of increase in healthcare costs in relation to BMI differed between men and women. Men with BMIs of 25 to 29.9 kg/m2 and ≥30 kg/m2, respectively, had total annual healthcare costs that were 1.02 (95% CI: 0.73 to 1.56) and 1.53 (95% CI: 1.03 to 2.40) times higher than their peers in the lowest BMI stratum; among women, these cost ratios were 1.22 (95% CI: 0.96 to 1.55) and 1.27 (95% CI: 1.01 to 1.59), respectively.

The pattern of increase in healthcare costs in relation to BMI also varied by age. Among subjects aged 35 to 44, 45 to 54, and 55 to 64 years, ratios for total annual healthcare costs were 1.14 (95% CI: 0.87 to 1.46), 0.86 (95% CI: 0.57 to 1.42), and 1.19 (95% CI: 0.94 to 1.53), respectively, among those with BMIs of 25 to 29.9 kg/m2; these same ratios were 1.36 (95% CI: 1.04 to 1.73), 1.10 (95% CI: 0.73 to 1.82), and 1.50 (95% CI: 1.09 to 2.02), respectively, among those with BMIs ≥ 30 kg/m2.

BMI and Cumulative Healthcare Costs

Cumulative total healthcare costs over the period 1990 through 1998 increased with BMI, averaging $15,583 (95% CI: $13,478 to $17,688), $18,484 (95% CI: $16,327 to $20,641), and $21,711 (95% CI: $18,825 to $24,596) for subjects with BMIs of 20 to 24.9, 25 to 29.9, and ≥30 kg/m2, respectively. Cost differences relative to those in the lowest BMI stratum emerged during the second year of follow-up for subjects with BMIs ≥ 30 kg/m2, and in the fifth year of follow-up for those with BMIs of 25 to 29.9 kg/m2 (Figure 1).

image

Figure 1. Cumulative total healthcare costs shown by year of follow-up and BMI.

Download figure to PowerPoint

Cumulative costs of pharmacy services averaged $2450 (95% CI: $2148 to $2751), $3310 (95% CI: $2905 to $3714), and $5000 (95% CI: $4201 to $5798) among those with BMIs of 20 to 24.9, 25 to 29.9, and ≥30 kg/m2, respectively (Figure 2). Cumulative costs of outpatient services were similar in magnitude among those with BMIs of 20 to 24.9 kg/m2 and 25 to 29.9 kg/m2, averaging $7673 (95% CI: $6712 to $8634) and $7391 (95% CI: $6871 to $7911), respectively, but were $8826 per patient (95% CI: $7927 to $9725) among those with BMIs ≥ 30 kg/m2 (Figure 3). The cumulative cost of inpatient care averaged $5460 (95% CI: $4107 to $6814) for those with BMIs of 20 to 24.9 kg/m2, $7783 (95% CI: $6077 to $9489) for those with BMIs of 25 to 29.9 kg/m2, and $7885 (95% CI: $5941 to $9829) for those with BMIs ≥ 30 kg/m2 (Figure 4).

image

Figure 2. Cumulative costs of prescription drugs shown by year of follow-up and BMI.

Download figure to PowerPoint

image

Figure 3. Cumulative costs of outpatient services shown by year of follow-up and BMI.

Download figure to PowerPoint

image

Figure 4. Cumulative costs of inpatient care shown by year of follow-up and BMI.

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Research Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

Analyses of the relationship between BMI and healthcare costs have been performed from a variety of perspectives, including those of the US healthcare system (9) (10) (11) (12) (13), private-sector business (14), managed care organizations (15) (16), and individual patients (17) (18) (19) (20). With the exception of recent modeling studies (17) (19) (20), however, this research has focused on the relationship between BMI and “current” healthcare costs. How “future” healthcare costs vary in relation to BMI has received relatively little attention.

In this study, we examined the relationship between BMI and future healthcare costs over a 9-year period in a cohort of 1286 members of KPNW, a large not-for-profit health-maintenance organization located in the Pacific Northwest region of the United States. Study subjects were selected from among the respondents to a 1990 health survey (which included questions on height and weight) that was mailed to a random sample of KPNW members. We excluded from consideration all survey respondents who smoked cigarettes or had a history of CHD, stroke, HIV, or cancer. Survey respondents meeting all inclusion criteria were stratified according to their BMI in 1990, and levels of use and cost over the period 1990 through 1998 were compared among subjects with BMIs of 25 to 29.9 kg/m2 and ≥30 kg/m2 relative to those with BMIs of 20 to 24.9 kg/m2.

We found that persons with BMIs ≥30 kg/m2 had 36% higher annual healthcare costs, including 105% higher pharmacy costs and 39% higher costs of primary-care visits. Subjects with BMIs of 25 to 29.9 kg/m2 had 37% higher pharmacy costs and 13% higher visit costs; overall healthcare costs were 10% higher, although not significantly different from those of persons with BMIs of 20 to 24.9 kg/m2. Because there were significant differences across BMI strata in the demographic characteristics of study subjects, we also examined the relationship between BMI and healthcare costs in strata defined on the basis of age and sex. In these analyses, healthcare costs generally increased with BMI, although the magnitude of increase varied for men versus women and according to age.

In a recent cross-sectional study, Quesenberry et al. (15) reported that persons with BMIs of 25 to 29.9 kg/m2 had overall medical care costs that were no higher than those of persons with BMIs of 20 to 24.9 kg/m2. The authors speculated that this might have reflected confounding by wasting diseases (e.g., cancer and AIDS) associated with both low BMI and high healthcare costs. We similarly observed no significant difference in total healthcare costs between these BMI strata. However, we do not believe that our finding is caused by confounding, because we excluded all persons who had cancer or AIDS at the beginning of the study period. Statistical outliers, however, seem to have greatly influenced our results. Four subjects in our sample had cumulative costs of care in excess of $175,000 (∼10 times the overall mean); three of these were in the 20 to 24.9 kg/m2 stratum. Excluding these four subjects, persons with BMIs of 25 to 29.9 kg/m2 had costs that were 1.19 (95% CI: 1.03 to 1.38) times higher than those in the lowest stratum; the corresponding figure for subjects with BMIs ≥ 30 kg/m2 was 1.52 (1.28 to 1.80).

In designing this study, we attempted to eliminate potential sources of bias and confounding in the sample selection process. Therefore, we excluded subjects who reported that they smoked cigarettes or had a history of cancer or AIDS, because these are associated with both low BMI and high medical care costs. We further excluded subjects with preexisting CHD and stroke. However, we did not exclude those with type 2 diabetes, hypertension, or hypercholesterolemia, which have been characterized as “physiological and metabolic effects of obesity” (21). We deemed it inappropriate to control for these diseases because they are critical links in the pathway by which excess body weight leads to increased risks of CHD and stroke, among other important diseases. Although we believe that our approach is valid, some of the differences in healthcare costs that we observed were due to differences across BMI strata in the prevalence of these diseases at baseline. We note, however, that similar findings were obtained in the subgroup of subjects who did not have type 2 diabetes, hypertension, or hypercholesterolemia at baseline. Among these subjects (n = 911), total cost ratios for those with BMIs of 25 to 29.9 kg/m2 and ≥30 kg/m2 were 0.99 (95% CIs: 0.80 to 1.25) and 1.34 (95% CIs: 1.02 to 1.74), respectively.

Some additional limitations of our study should be noted. For one, we calculated BMI using self-reported height and weight. Although earlier research suggests that survey respondents generally overstate their height and understate their weight (22) (23), to what extent this occurred in our patient population is not known. If persons who are overweight tend to understate their weight to a greater degree than others (22) (23), then this would impart a conservative bias to our findings. We also note that only 56% of health plan members responded to the mail survey that was used to identify the study cohort. The impact of possible response bias on our findings is unknown.

We also are uncertain whether our findings can be generalized to other settings. KPNW has a predominately white, employed, and suburban membership; thus, our findings may not be applicable to patient populations with substantially different demographic and/or socioeconomic characteristics. In addition, because KPNW is a not-for-profit, group-model health maintenance organization located in the Pacific Northwest, practice patterns and/or costs of care may differ from those of health plans with different corporate structures, organizational forms, or geographic representation.

Finally, although BMI is a widely used measure of the degree of adiposity, it is not without limitations. In particular, BMI overestimates adiposity among persons with increased muscularity (e.g., caused by regular resistance training) and underestimates adiposity among elderly persons as a consequence of age-related declines in lean tissue mass. Accordingly, use of BMI may have led us to understate the relationship between adiposity and future healthcare costs. We note, however, that for the age ranges examined in this study (i.e., 35 to 64 years), the correlation between BMI and more accurate measures of adiposity (e.g., densitometry) exceeds 90% (24).

The role of excess body weight as a risk factor for various chronic disease conditions has been examined in prospective cohort studies. To the best of our knowledge, our study is the first to examine the relationship between BMI and future healthcare costs. Our results indicate that future healthcare costs are higher for persons who are overweight, especially those with BMIs ≥ 30 kg/m2. Accordingly, our findings may be of interest to clinicians, public health officials, and others interested in the long-term consequences of obesity.

Acknowledgments

  1. Top of page
  2. Abstract
  3. Introduction
  4. Research Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

Funding for this research was provided by Knoll Pharmaceutical Company, (Mount Olive, NJ). We thank John Edelsberg for helpful comments on the design of this study, Lisa J. McGarry and Amy K. O'Sullivan for technical assistance, and Matt Nguyen and G. Rhys Williams for support of this project.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Research Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References