Impact of bariatric surgery on health care utilization and costs among patients with diabetes

Abstract

Background: The effect of bariatric surgery on health care utilization and costs among individuals with type 2 diabetes remains unclear.

Objective: To examine health care utilization and costs in an insured cohort of individuals with type 2 diabetes after bariatric surgery.

Research design: Cohort study derived from administrative data from 2002 to 2008 from 7 Blue Cross Blue Shield Plans.

Participants: Seven thousand eight hundred six individuals with type 2 diabetes who had bariatric surgery.

Measures: Cost (inpatient, outpatient, pharmacy, and others) and utilization (number of inpatient days, outpatient visits, specialist visits).

Results: Compared with presurgical costs, the ratio of hospital costs (excluding the initial surgery), among beneficiaries who had any hospital costs, was higher in years 2 through 6 of the postsurgery period and increased over time [post 1: odds ratio (OR)=0.58; 95% confidence interval (CI), 0.50-0.67; post 6: OR=3.43; 95% CI, 2.60-4.53]. In comparison with the presurgical period, the odds of having any health care costs was lower in the postsurgery period and remained relatively flat over time. Among those with hospitalizations, the adjusted ratio of inpatient days was higher after surgery (post 1: OR=1.05; 95% CI, 0.94-1.16; post 6: OR=2.77; 95% CI, 1.57-4.90). Among those with primary care visits, the adjusted OR was lower after surgery (post 1: OR=0.80; 95% CI, 0.78-0.82; post 6: OR=0.66; 95% CI, 0.57-0.76).

Conclusions: : In the 6 years after surgery, individuals with type 2 diabetes did not have lower health care costs than before surgery.

Figure 1

Adjusted Odds of Any Costs and Adjusted Ratio of Mean Total Costs in Post-surgical Periods compared to Pre-surgical Period † Probability of having any positive costs ( ) ‡ Cost ratio conditional on having any positive costs ( ) Notes: Each post year refers to time after surgery. For example, “Post 1” is equivalent to the first year after surgery. The squares represent the point estimate for each year along with the corresponding confidence interval. The model controlled for morbidity level, Diabetes Complications Severity Index (DCSI), sex, health plan site, linear age, and obesity propensity. Y-axes are presented on logarithmic scales.

Figure 2

Adjusted Odds of a Hospitalization and Adjusted Ratio of Counts of Hospitalization in Post-surgical Periods compared to Pre-surgical Period † Probability of having no hospitalizations ( ) ‡ Count ratio conditional on having any hospitalization ( ) Notes: Each post year refers to time after surgery. For example, “Post 1” is equivalent to the first year after surgery. The squares represent the point estimate for each year along with the corresponding confidence interval. The model controlled for morbidity level, Diabetes Complications Severity Index (DCSI), sex, health plan site, linear age, and obesity propensity. We used the biased corrected percentile confidence intervals (rather than the normal theory confidence intervals) to account for the skewed nature of the data. Y-axes are presented on logarithmic scales. The odds ratio for post-year 1 is OR = 3.27 × 10⁻¹¹ (95% CI: 6.24 × 10⁻¹⁶, 1.72 × 10⁻⁶).

Figure 3

Adjusted Ratio of Counts of Specialist Visits in Post-surgical Periods compared to Pre-surgical Period † Probability of having no specialist visits ( ) ‡ Count ratio conditional on having any specialist visits ( ) Notes: Each post year refers to time after surgery. For example, “Post 1” is equivalent to the first year after surgery. The squares represent the point estimate for each year along with the corresponding confidence interval. The model controlled for morbidity level, diabetes severity index, sex, health plan site, linear age, and obesity propensity. We used the biased corrected percentile confidence intervals (rather than the normal theory confidence intervals) to account for the skewed nature of the data. Y-axes are presented on logarithmic scales.

Figure 4

Adjusted Ratio of Counts of Primary Care Visits in Post-surgical Periods compared to Pre-surgical Period † Probability of having no primary care visits ( ) ‡ Count ratio conditional on having any primary care visits ( ) Notes: Each post year refers to time after surgery. For example, “Post 1” is equivalent to the first year after surgery. The squares represent the point estimate for each year along with the corresponding confidence interval. The model controlled for morbidity level, diabetes severity index, sex, health plan site, linear age, and obesity propensity. We used the biased corrected percentile confidence intervals (rather than the normal theory confidence intervals) to account for the skewed nature of the data. Y-axes are presented on logarithmic scales.