Published Online: July 19, 2013
Hsien-Yen Chang, PhD; Thomas M. Richards, MS; Jeanne M. Clark, MD, MPH; Jonathan P. Weiner, DrPH; and Jodi B. Segal, MD, MPH
Background: Because laboratory test results are less available to researchers than claims data, a claims-based indicator of diabetes improvement would be valuable.
Objectives: To determine whether a decrease in medication use for diabetes parallels clinical improvement in glycemic control.
Study Design: This was a retrospective cohort study using up to 3.5 years of pharmacy and laboratory data from 1 private insurer. Data included 104 patients with diabetes who underwent bariatric surgery and had at least 1 glycated hemoglobin (A1C) test before and after surgery.
Methods: We assigned each A1C test to a 90-day interval before or after surgery. Medication availability was noted for the midpoint of the interval (on insulin, on oral medications, count of medications). Each subject could contribute 1 presurgery and up to 3 postsurgery observations. We recorded the changes in A1C test results and medication use from the presurgery to the postsurgery period. Using the A1C test as the reference standard, positive and negative predictive values of medication-based indicators were calculated.
Results: After bariatric surgery, A1C test values decreased by more than 1% and the count of unique medications decreased by 0.6. All 3 medication-based indicators had high positive predictive values (0.85) and low negative predictive values (0.20), and count of medications had better performance than the other indicators.
Conclusions: Without clinical information, a decrease in use of medications can serve as a proxy for clinical improvement. Validation of results in other settings is needed.
Am J Manag Care. 2013;19(7):572-578
Diabetes is a major health concern in the United States given its high prevalence and significant morbidity. Many interventions have been implemented to improve glycemic control, especially those targeting weight loss.1-5 Glycated hemoglobin (A1C) is the standard method of monitoring diabetes mellitus control, including responses to treatment.6-8 In health services research, clinical data often are not available to researchers, particularly in studies that use administrative claims for outcomes evaluation.9,10
Claims data have been used to describe the outcomes of interventions or as a method for risk stratification among individuals with diabetes.11-15 However, it is hard to know whether glycemic control has improved without access to laboratory test data. Therefore, in this study we determined whether a decrease in the use of medications for diabetes parallels clinical improvement in glycemic control among individuals with substantial weight loss. We used individuals with diabetes who had undergone bariatric surgery, as they are expected to have improvement in A1C after surgery.16-19 We hypothesized that a decrease in medication use from the presurgical to the postsurgical period would parallel an improvement in A1C values over the same period.
MATERIALS AND METHODS
This was a retrospective cohort study using up to 3.5 years of pharmacy and laboratory data (up to 1 year before and 2.5 years after bariatric surgery).
We accessed claims and laboratory data from 1 Blue Cross Blue Shield plan, as part of a large ongoing collaborative effort. We received data on all members who met at least 1 of the following inclusion criteria at any point during 2002 through 2006: (1) completed a health risk assessment that captured member height and body weight; (2) had an obesity diagnosis reported in aclaim; (3) had a paid or denied claim for bariatric surgery; (4) had a paid or denied claim for a medication for promoting weight loss; or (5) were more than 12 years of age and had a diagnosis of hyperlipidemia, type 2 diabetes, sleep apnea, gall bladder disease or surgery, or metabolic syndrome. These diagnoses were identified in the claims by Current Procedural Terminology codes, International Classification of Disease, Ninth Revision, Clinical Modification (ICD-9-CM) codes, and National Drug Codes or Diagnosis-Related Group codes.
We acquired (1) enrollment files from administrative data; (2) benefits information to determine medical and pharmacy coverage; (3) inpatient, outpatient, and pharmacy claims records containing ICD-9-CM diagnosis, Current Procedural Terminology codes, prescription National Drug Codes, and the starting date and length of prescriptions; and (4) laboratory records containing titles of test, test dates, and results in a subset of the population.
We restricted analyses to enrollees with bariatric surgery and diabetes. We considered patients to have had bariatric surgery if they had appropriate ICD-9-CM procedure codes or Current Procedural Terminology codes.20 For 4 of the ICD-9-CM codes, we required that a code for morbid obesity (ICD-9-CM 287.1) be present concurrently with the procedure code. We defined individuals as having type 2 diabetes if they had 1 relevant inpatient code or 2 outpatient codes separated by at least 30 days. The relevant codes were 250.xx, 648.0 (diabetes mellitus with pregnancy), 362.0 (diabetic retinopathy), or 266.41 (diabetic cataract). Individuals with only 250.x3 (type 1 diabetes) were not included. Additionally, any individual filling a prescription for a medication for treatment of hyperglycemia was included. If the prescription was for metformin alone, the individual was also required to have an ICD-9-CM code for diabetes for inclusion in this group.
A1C Test Value Calculation
We classified the dates of the laboratory tests into a 90-day period before surgery and a 90-day period after surgery (time zero). We required an individual to have pharmacy coverage throughout the entire 90-day period for his/her results to be evaluated in that period. We had only the month of the laboratory test result, so we arbitrarily assigned the middle of the month (day 15) to each record and then calculated the mean test result for each period. We limited observations to 360 days before and 900 daysafter surgery, and to enrollees with at least 1 presurgery and 1 postsurgery A1C value. We created pre-post test value pairs by matching any presurgery observation with any postsurgery observation from the same individual. We further restricted analyses to the latest presurgery observation and up to the 3 earliest postsurgery observations from the same individual.
Calculation of Drug in Hand
For a 90-day period, we used pharmacy claims to determine whether an enrollee had diabetes drugs in hand on the median day in that period. Diabetes drugs were classified into the following categories: alpha-glycosidase inhibitors, dipeptidyl peptidase inhibitors, exenatide, metformin, nateglinide, sulfonylureas, thiazolidinediones, and insulin. We calculated the number of unique medications as the count of medications from these categories that were in hand on the median day in a period.
Definition of the Change From Presurgery to Postsurgery Periods
We defined change as a 3-level categorical variable: “decrease,” “same,” and “increase” (Table 1). For example, an enrollee would be assigned to the decrease category if he/she switched from having any diabetes medication in the presurgery period to no medication in the postsurgery period, or had a drop in the count of unique medications, or had a postsurgery A1C test value lower than the presurgery value by at least 0.4%.
We further collapsed these 3 categories into a binary variable: “favorable” or “unfavorable” (Table 1). Observations in the decrease category were considered favorable while those in the increase category were unfavorable. For observations classified into the “same” category, the use of no diabetes medications in both periods was classified as favorable, whereas others were unfavorable. For “same” observations of A1C test values, postsurgery test values lower than the presurgery values,by any amount, were considered favorable; the rest were unfavorable.
Using the binary change in A1C test results as a gold standard, we calculated the positive predictive value and negative predictive value for 3 binary medication-based indicators. We also calculated the Spearman’s correlation coefficient for 2 continuous variables: change in the count of unique medications and change in the test value (a simple difference between the presurgery and postsurgery values).
Given the high prevalence of being a nonuser of oral medication/insulin or having a zero count of medications in both periods, we tested the results while restricting the analyses to enrollees who were active users in at least 1 period to identify whether these medication-based indicators performed better among active users (restricted set; eAppendix A, available at www.ajmc.com). In addition, we also tested the outcomes with the sample limited to 1 pre-post observation pair (the latest presurgery observation and the earliest postsurgery observation; 104 pairs in total; eAppendix B, available at www.ajmc.com). Furthermore, when calculating drug in hand, instead of using the median day only, we also investigated whether a longer period (15 days; from a week before to a week after the median day in a period) would result in differences in the performance of medication-based indicators; the results are shown in eAppendix C (available at www.ajmc.com).
The data were de-identified in accordance with the Health Insurance Portability and Accountability Act’s definition of a limited data set. The Johns Hopkins University Office of Research Subjects deemed the study to be exempt from federal regulations because the research activities were considered to be of minimal risk to subjects, since the subjects were not identifiable.
The Figure shows the flow chart for sample selection.
Characteristics of the Sample
There were 208 paired presurgery and postsurgery observations (A1C and medication use) from 104 enrollees when including up to the 3 earliest postsurgery observations (Table 2). The mean age of the 104 eligible enrollees was 47.9 years, and females accounted for 80% of the sample. Of 208 pairs in the presurgery period, the mean A1C test value was 7.5%. About half of the individuals used oral medications, 12% used insulin, the average number of unique medications was 1.1, and 41% were nonusers. We observed reductions in medication use and A1C in the postsurgery period: the mean A1C value decreased to 6.1%, only 19% had oral medications, only 3% had insulin, the average number of unique medications was 0.4, and close to 80% were nonusers.
PDF is available on the last page.