Current Medical Research & Opinion 0300-7995 doi:10.1185/03007995.2014.936930
Vol. 30, No. 10, 2014, 1991–2000
Article ST-0439.R1/936930 All rights reserved: reproduction in whole or part not permitted
Curr Med Res Opin Downloaded from informahealthcare.com by University of Washington on 12/02/14 For personal use only.
Original Article Severe hypoglycemia rates and associated costs among type 2 diabetics starting basal insulin therapy in the United States
Michael L. Ganz
Abstract
Evidera, Lexington, MA, USA
Neil S. Wintfeld Novo Nordisk Inc., Plainsboro, NJ, USA
Qian Li Yuan-Chi Lee Elyse Gatt Evidera, Lexington, MA, USA
Joanna C. Huang Novo Nordisk Inc., Plainsboro, NJ, USA Address for correspondence: Michael L. Ganz MS, PhD, Evidera, Retrospective Observational Studies, 430 Bedford Street, Lexington, 02420, USA. [email protected] Keywords: Costs and cost analysis – Diabetes mellitus, type 2 – Hypoglycemia – Insulin – Prevalence studies Accepted: 16 June 2014; published online: 15 July 2014 Citation: Curr Med Res Opin 2014; 30:1991–2000
Objectives: To derive current real-world data on the rates and costs of severe hypoglycemia (SH) for people with type 2 diabetes mellitus (T2D) who have initiated basal insulin therapy and to examine differences in SH rates and costs stratified by history of prior SH events. Methods: We used a nation-wide electronic health records database that included encounter and laboratory data, as well as clinical notes, to estimate the rates and costs of SH events among adults with T2D who initiated basal insulin between 2008 and 2011. Unadjusted and regression-adjusted rates and quarterly costs were calculated for all patients as well as stratified by history of a SH event before starting basal insulin and history of a SH event during the basal insulin titration period. Results: We identified 7235 incident cases of basal insulin use among patients with T2D who did not use insulin during the previous 12 months. Regression-adjusted incidence and total event rates were 10.36 and 11.21 per 100 patient-years, respectively. A history of SH events during the pre-index baseline and post-index titration periods were statistically significantly associated with both the incidence and total event rates (p50.01). Regression-adjusted total healthcare and diabetes-related costs were statistically significantly (p50.01) higher in those quarters when a SH event occurred than in those quarters without any SH events ($3591 vs. $487 and $3311 vs. $406, respectively). A history of previous SH or SH events during the titration period were not statistically significantly associated with costs. Conclusions: These results suggest that the real-world burden of SH is high among people with T2D who start using basal insulin and that history of previous SH events, both before starting insulin and during the insulin titration period, influences future SH. These results can also provide insights into interventions that can prevent or delay SH. These results should, however, be interpreted in light of the key limitations of our study: not all SH events may have been captured or coded in the database, data on filled prescriptions were not available, and the post-titration follow-up period could have been divided into time units other than quarters (3 month blocks) resulting in potentially different conclusions. Further real-world studies on the frequency and costs of SH, using methods to identify as many SH events as possible, can allow healthcare providers to make more informed decisions on the risks and benefits of basal insulin therapy in T2D patients.
Introduction The World Health Organization (WHO) has classified diabetes as a global epidemic1. Approximately 285 million (6.4%) adults worldwide have diabetes2, with the US responsible for almost 9% of those cases (25.6 million adults, ! 2014 Informa UK Ltd www.cmrojournal.com
Severe hypoglycemia among patients starting basal insulin Ganz et al.
1991
Curr Med Res Opin Downloaded from informahealthcare.com by University of Washington on 12/02/14 For personal use only.
Current Medical Research & Opinion Volume 30, Number 10
October 2014
corresponding to 11.3% of the US adult population)3. By the year 2030 the number of adults with diabetes is expected to grow substantially to almost 439 million globally and to almost 36 million in the US2. Reflecting its prevalence and associated morbidity, diabetes is responsible for $116 billion in annual direct medical costs (people with diabetes incur 2.3 times higher costs than otherwise similar people without diabetes) and for $58 billion annually in disability, lost work productivity and premature mortality in the US3. Hypoglycemia, a deficiency of glucose in the bloodstream, is a serious condition related to the use of antidiabetic medications and is a significant cause of morbidity and mortality in people with diabetes. Approximately 20% of diabetics who are treated with antidiabetic medications are expected to have symptoms related to hypoglycemia4. Severe hypoglycemia (SH), or hypoglycemia that requires the assistance of a third party, if unrecognized and untreated, can result in coma, seizures, or death5. Recent studies have estimated, not withstanding how SH was defined, that between 3.5% (annual)6 and 30.4% (first 6 months)7 of people with diabetes have experienced at least one SH while receiving antidiabetic therapy. The occurrence of SH is frequently followed by treatment adjustments involving the glycemic target, which can result in long-term diabetes complications, increased use of healthcare services, and substantial costs. The mean cost per hypoglycemic event, including inpatient, outpatient, and emergency department costs, has been reported to range from $1049 (2004 USD) to $1331 (2007 USD) per event8,9. Understanding the epidemiology and cost implications of SH is important for being able to project future demands on the public health system and for being able to evaluate the cost effectiveness of current and future treatments for diabetes. Despite this, there are relatively little recent real-world data on the rates of and costs associated with SH, especially for people with type 2 diabetes mellitus (T2D) who have initiated basal insulin therapy. Our objectives were, therefore, to estimate the real-world incidence and total SH event rates and the healthcare costs among people with (T2D) who have initiated basal insulin in US.
Materials and methods Data source We analyzed encounters and services that occurred between January 2008 and December 2011 using data from Humedica Inc.’s Real-time Longitudinal Clinical Data patient-level database. The Humedica database contains data abstracted from electronic health records (EHR) integrated with claims, prescription, and practice 1992
Severe hypoglycemia among patients starting basal insulin Ganz et al.
management data. The database reflects the real-world patient care experiences, including laboratory results and radiology reports, physician and nurse notes, prescriptions written, procedures, diagnoses, and other details of a patient’s office visit and/or hospital stay. The database includes information from approximately 30 million patients derived from 70 hospitals, 435 outpatient clinics, and 17,600 physicians in 38 states.
Patient selection We selected patients who had a T2D diagnosis using standard definitions in the context of EHR and administrative databases10,11. In particular, at least one of the following criteria had to be satisfied for a patient to qualify as a T2D patient: (1) International Classification of Diseases, Ninth Revision (ICD-9-CM) diagnosis codes of 250.x0 or 250.x2; (2) one or more prescription orders for a non-insulin antidiabetic drug (NIAD); (3) two consecutive fasting blood glucose levels of 126 mg/dl; or (4) glycated hemoglobin (HbA1c) 7.0%. Patients had to have started basal insulin without having used any type of insulin for at least the previous 12 months (we refer to a patient’s first prescription of basal insulin as the ‘index date’). Patients had to be at least 18 years old as of their index date and have at least 12 months of continuous eligibility in the database before and after their index date. Patients were further excluded if they were pregnant or had a diagnosis of Type 1, secondary, or gestational diabetes (ICD-9-CM codes 250.x1, 249.xx, and 648.0 or 648.8, respectively).
Study measures Patient measures included demographic and clinical characteristics such as age, sex, race, geographic region, smoking status, the type of insulin patients started, history of visiting an endocrinologist, use of oral antidiabetic drugs (OADs) and other medications, and other healthcare resource use during the 12 months prior to starting basal insulin. The presence of key comorbid conditions associated with diabetes outcomes was also assessed during the 12 months prior to starting basal insulin. A complete list of the baseline characteristics are presented in Table 1 (a list of the diabetes-related comorbid conditions is in Appendix 1). Severe hypoglycemia events SH events were defined as events requiring medical attention with the appropriate SH diagnosis codes attached to outpatient, inpatient, or emergency department visits or by a recorded glucose level of 40 mg/dL. The presence of diagnosis codes 251.0x, 251.1x, 251.2x, or 250.3x on different days or the presence of diagnosis codes 251.8x without codes 259.8x, 272.7, 681.xx, 682.xx, 686.9x, 707.1x– www.cmrojournal.com ! 2014 Informa UK Ltd
Current Medical Research & Opinion
Curr Med Res Opin Downloaded from informahealthcare.com by University of Washington on 12/02/14 For personal use only.
Table 1. Demographic and pre-index baseline clinical characteristics of type 2 diabetes mellitus patients initiating basal insulin (N ¼ 7235). Age (%) 18–24 25–34 35–44 45–54 55–64 65–74 75 Mean Age (SD) Sex (%) Male Female Race (%) White Black Asian/Other/Unknown Region (%) Northeast Midwest South West Other/Unknown Index Insulin (%) Glargine NPH Detemir History of Diabetes-Related Comorbidities (%) History of Healthcare Use (%) Endocrinologist Visits Inpatient Hospitalization Outpatient Services Emergency Department Use History of Medication Use (%) Metformin Sulfonylurea Other OADs Lipid Lowering Drugs Antihypertensive Drugs
0.43 1.87 7.37 18.87 29.51 35.44 6.52 60.82 (11.65) 49.30 50.70 63.25 19.09 17.66 8.06 50.68 29.38 11.82 0.06 77.24 5.86 16.90 80.17 13.74 9.01 4.34 5.47 36.66 38.06 25.82 32.51 51.93
SD, standard deviation; NPH, neutral protamine Hagedorn; OADs, oral antidiabetic drugs.
707.9x, 709.3x, 730.0x–730.2x, or 731.8x indicated a SH event12. Humedica uses state of the art natural language processing technology to extract data from physician, radiology, and pathology notes for over 8.5 million patients. We used these notes, available for two-thirds of the patients, to identify SH events not otherwise captured by diagnosis codes by searching for evidence of hypoglycemic events that occurred without immediate medical attention, but had serious sequelae (such as paresthesias, coma, loss of consciousness, and need for glucagon) that required the assistance of another person. SH on consecutive days was considered a single SH event unless interrupted by at least one day without evidence of SH, in which case each run of consecutive days of SH were considered separate SH events. Healthcare costs The Humedica database provides information on healthcare utilization at the encounter and service level, but it does not contain cost information. We, therefore, ! 2014 Informa UK Ltd www.cmrojournal.com
Volume 30, Number 10
October 2014
obtained costs from external sources. Inpatient stays recorded in the Humedica database were matched to similar hospitalizations in the National Inpatient Sample based on Clinical Classification Scheme (CCS) codes13 and geographic region to obtain predicted hospitalization costs. Costs for other encounters (outpatient, office, and emergency visits), except those that happened during an inpatient stay, were estimated by linking those encounters to the Medicare fee schedule using procedure codes. Encounters and services that could not be linked (26% of all encounters and services, which were predominantly laboratory and non-durable medical equipment services) were ignored. Although data were available on prescriptions that were written, which were used to determine inclusion eligibility, no data were available on filled prescriptions, so we were not able to include pharmacy costs in these analyses. Further details on these methods are provided in Appendix 2.
Statistical analyses We defined three study periods: the 12 months prior to the index date (‘pre-index baseline’ period), the 16-week period immediately following the index date (‘post-index titration’ period) during which we assumed patients learned how to correctly use their insulin and made adjustments to their dose, and the remaining time until the end of the eligibility in the database (‘post-titration follow-up’ period). Means and standard deviations were calculated for continuous variables, including healthcare costs during the first year of the post-titration follow-up period, and frequency distributions were reported for categorical variables; differences in mean costs between patients who did and did not experience SH events were assessed using the two-tailed t test (differences with p-values less than 0.05 were considered statistically significant). Incidence rates were calculated by dividing the total number of first SH events by the total number of patient-years observed from the end of the titration period until the first SH event for patients who eventually experienced a SH event or to the end of the post-titration follow-up period for patients who never experienced a SH event. Total SH event rates were calculated by dividing the total number of SH events that occurred during the post-titration follow-up period by the total number of available patient-years. Exact confidence intervals around these rates were computed using the Poisson distribution. We further stratified the crude incidence and total event rates by history of SH during the pre-index baseline and post-index titration periods to assess how these rates varied according to exposure to prior SH events. The association between the SH rates and history of any SH events during the pre-index baseline and post-index titration periods were estimated using negative binomial regression Severe hypoglycemia among patients starting basal insulin Ganz et al.
1993
Curr Med Res Opin Downloaded from informahealthcare.com by University of Washington on 12/02/14 For personal use only.
Current Medical Research & Opinion Volume 30, Number 10
October 2014
models adjusting for patient demographic and clinical characteristics. We assessed the association of SH events with healthcare costs (overall and for diabetes-related encounters and services) during the entire post-titration period using a multiple generalized linear regression model (gamma distribution and a log-link function), adjusting for patient demographic and clinical characteristics, to address the skewed distribution of costs (we did not adjust for history of any SH events during the pre-index baseline and post-index titration periods because those factors were not statistically significantly associated with costs in our preliminary analyses). Because we wanted to estimate the proximal effect of SH events on costs and because of the variable lengths of follow-up, we estimated the association of the occurrence of an SH event in a given quarter with that quarter’s costs (standard errors were corrected for multiple observations per patient). We applied the method of recycled predictions to the negative binomial regression results to compute the adjusted incidence and total SH event rates conditional on experiencing any SH events during the pre-index baseline and post-index titration periods and to the generalized linear regression results to compute adjusted healthcare costs conditional on experiencing any SH events during the pre-index baseline and post-index titration periods (associated standard errors were computed using the delta method)14,15. Recycled predictions of scenarios of interest were obtained by averaging the regressionadjusted estimates of the outcome variable for each individual after assigning each individual to membership in each category of the scenario of interest while using the observed values for the other variables. For example, to compute the adjusted effect of experiencing a SH during the pre-index baseline period, we calculated predicted costs for each individual after assigning the SH variable ^ 0 and after assigning the SH variable to to zero to obtain C i 1 ^ . The average costs associated with not one to obtain C i ^0 0 ¼ PN C experiencing a SH are calculated as C i¼1 i /N and the average costs associated with experiencing a SH are ^ 1 /N. The incremental effect of 1 ¼ PN C calculated as C i¼1 i 0. 1 C experiencing an SH is therefore DC=C
Approximately three-quarters of the patients were 55 years old or older (mean age 61), about 50% were female, almost two-thirds were white, and most patients resided in the South or Midwest (Table 1). A vast majority (80%) of patients experienced at least one diabetes-related comorbid condition in the year before they started basal Patients with a diagnosis of T2D from 1 January 2008 to 31 December 2011 N = 425,327
Patients with a basal insulin prescription after continuous enrollment start N = 67,086
Patients with at least 12 months continuous enrollment prior to the index date N = 30,621
Exclude patients with any insulin use prior to the index date N = 19,474
Age ≥18 as of the index date N = 19,381
Exclude diagnosed type 1 diabetes N = 16,654
Exclude diagnosed secondary diabetes N = 16,435
Exclude diagnosed gestational diabetes N = 16,234
Results We identified 7235 T2D patients who were new users of basal (detemir, glargine, or neutral protamine Hagedorn) insulin (i.e., did not use any insulin during the previous 12 months) who satisfied our inclusion and exclusion criteria (Figure 1), with 568 of those patients (7.85%) experiencing at least one SH event (1245 SH events in total) during the post-titration follow-up period (548 patients [1189 SH events] based on the encounter data and 20 patients [56 SH events] based on the clinical notes). 1994
Severe hypoglycemia among patients starting basal insulin Ganz et al.
Exclude pregnant patients N = 16,172
Patients were continuously enrolled for ≥ 16-week titration period + 12-month follow-up period N = 7235
Figure 1. Sample selection flowchart.
www.cmrojournal.com ! 2014 Informa UK Ltd
! 2014 Informa UK Ltd www.cmrojournal.com
5.91 (5.76–6.06) 9.00 (8.87–9.12)
37.59 (36.73–38.47) 90.69 (89.58–91.80)
4.11 (4.07–4.14) 7.86 (7.81–7.91)
9.15 (8.93–9.38) 11.54 (11.38–11.70)
October 2014
In parentheses are 95% confidence intervals. SH, severe hypoglycemia.
4.17 (4.13–4.20) 7.92 (7.87–7.97) 24.64 (24.06–25.23) 71.28 (70.41–72.16) 4.63 (4.59–4.67) 9.69 (9.64–9.75)
No (N ¼ 7064) Yes (N ¼ 171) Rate Ratio [Yes:No] No (N ¼ 7026) Yes (N ¼ 209)
Volume 30, Number 10
Incidence Rate (Per 100 Patient-Years) Total Event Rate (Per 100 Patients-Years)
All (N ¼ 7235)
The mean post-titration follow-up length was 1.8 years. Because follow-up times differed by history of SH events during the pre-index baseline (1.72 vs. 1.78 years for patients with and without baseline SH events, p ¼ 0.08) and the post-index titration (1.66 vs. 1.78 years for patients with and without titration SH events, p50.01) periods, we present results for the SH incidence and total event rates stratified by SH histories during the baseline and titration periods. As shown in Table 2, the crude incidence and total SH event rates during the entire post-titration follow-up period were 4.63 (95% CI 4.59–4.67) and 9.69 (95% CI 9.64–9.75) per 100 patient-years, respectively (the rates for the first post-titration follow-up year were similar at 5.02 and 9.33 per 100 patient-years; not shown). These rates differed substantially by a patient’s history of SH events during the pre-index baseline and the post-index titration periods and indicated that not only is a history of previous SH events associated with SH events during the post-titration period, but that the history of SH events during the titration period has a stronger association with post-titration SH events than the pre-index history of SH events. The incidence rate for patients with a history of SH events during the pre-index baseline period was 5.91 (95% CI 5.76–6.06) times higher than those without a SH history but was 9.15 (95% CI 8.93–9.38) times higher for those with a history of SH events during the post-index titration period than those without a history of SH. Similarly, the total SH event rate for patients with a history of SH events during the pre-index baseline period was 9.00 (95% CI 8.87–9.12) times higher than those without a SH history but was 11.54 (95% CI 11.38–11.70) times higher for those with a history of SH events during the post-index titration period than those without a history of SH. The incidence and total event rates computed for the first post-titration follow-up year, as well as the impacts of histories of SH events during the pre-index baseline and post-index titrations periods, were similar to the rates for the entire post-titration follow-up period (not shown). Table 3 displays the results of both of the multivariable negative binomial regression models for the SH incidence and total SH event rates. The risk of a SH event subsequent to titration was positively and significantly associated with history of SH events during the pre-index baseline period (coefficients of 1.82 and 1.77 for SH incidence rates and total events rates, respectively; both
Table 2. Crude severe hypoglycemia event incidence and total events rates during the post-titration follow-up period.
Curr Med Res Opin Downloaded from informahealthcare.com by University of Washington on 12/02/14 For personal use only.
Incidence and total severe hypoglycemia event rates
History of SH During Pre-Index Baseline Period
History of SH During Post-Index Titration Period
insulin (the most prevalent comorbid conditions included hypertension [61%], dyslipidemia or hyperlipidemia [54%], neuropathy [21%], nephropathy or nephrosis [13%], and obesity [11%]; see Appendix 1).
Rate Ratio [Yes:No]
Current Medical Research & Opinion
Severe hypoglycemia among patients starting basal insulin Ganz et al.
1995
Current Medical Research & Opinion Volume 30, Number 10
October 2014
Table 3. Multivariable negative binomial regression results for incident and total SH events during the post-titration follow-up period. Incident SH Events Coefficient
Curr Med Res Opin Downloaded from informahealthcare.com by University of Washington on 12/02/14 For personal use only.
History of SH Events during Pre-Index Baseline Period History of SH Events during Post-Titration Follow-Up Period Age (vs. 18–24) 25–34 35–44 45–54 55–64 65–74 75 Sex (vs. Female) Male Race (vs. White) Black Asian/Other/Unknown Region (vs. Northeast) Midwest South West Index Insulin (vs. Glargine) NPH Detemir Diabetes-Related Comorbidities Metformin Use Sulfonylurea Use Use of Other OADs Use of Lipid-Lowering Drugs Use of Antihypertensive Drugs Endocrinologist Visits Inpatient Hospitalization Outpatient Services Emergency Department Use Constant Number of Patients Predicted Rates: Overall History of SH During Pre-Index Baseline No Yes History of SH During Post-Index Titration No Yes
1.82*** 2.85***
Total SH Events
95% CI
Coefficient
(1.35, 2.29) (2.33, 3.37)
1.77*** 2.05***
95% CI (1.28, 2.25) (1.53, 2.58)
0.10 0.25 0.17 0.14 0.23 0.54
(1.86, 1.66) (1.88, 1.37) (1.76, 1.42) (1.72, 1.44) (1.35, 1.81) (1.07, 2.15)
0.42 0.27 0.20 0.05 0.27 0.38
(1.36, 2.21) (1.95, 1.42) (1.85, 1.46) (1.59, 1.69) (1.37, 1.91) (1.30, 2.06)
0.06
(0.26, 0.15)
0.12
(0.08, 0.33)
0.27** 0.09
(0.00, 0.53) (0.39, 0.21)
0.30** 0.21
(0.03, 0.57) (0.52, 0.09)
0.25 0.16 0.11
(0.16, 0.66) (0.61, 0.28) (0.62, 0.39)
0.04 0.47** 0.42*
(0.36, 0.44) (0.91, 0.04) (0.91, 0.07)
0.25 0.25* 0.10 0.26** 0.03 0.05 0.20 0.17 0.06 0.41** 0.18 0.22 1.97** 7231
(0.17, 0.66) (0.53, 0.04) (0.39, 0.18) (0.51, 0.01) (0.29, 0.23) (0.31, 0.21) (0.46, 0.06) (0.08, 0.41) (0.25, 0.37) (0.08, 0.75) (0.32, 0.68) (0.20, 0.64) (0.32, 3.61)
0.43** 0.08 0.08 0.35*** 0.01 0.13 0.11 0.16 0.23 0.10 0.00 0.20 1.43* 7231
(0.04, 0.82) (0.36, 0.20) (0.36, 0.20) (0.61, 0.10) (0.25, 0.26) (0.39, 0.13) (0.36, 0.15) (0.08, 0.40) (0.07, 0.54) (0.23, 0.44) (0.51, 0.50) (0.21, 0.61) (0.18, 3.05)
10.36
(6.76, 13.95)
11.21
(8.81, 13.61)
6.60 40.67
(5.20, 7.99) (20.18, 61.67)
8.27 48.33
(7.16, 9.38) (25.46, 71.19)
5.35 92.45
(4.59, 6.11) (42.66, 142.24)
8.13 63.26
(7.10, 9.15) (30.86, 95.67)
CI, confidence interval; NPH, neutral protamine Hagedorn; OADs, oral antidiabetic drugs; SH, severe hypoglycemia. ***p50.01. **p50.05. *p50.10.
p50.01) as well as with history of SH events during the post-titration period (coefficients of 2.85 and 2.05 for SH incidence rates and total events rates, respectively; both p50.01) These results were consistent with the unadjusted analyses. Other factors were also associated with SH incidence and total event rates. A history of metformin use during the pre-index baseline period was negatively associated both incidence and total SH events and history of inpatient hospital use during the pre-index baseline period was positively associated with total SH events. NPH insulin use was positively associated with incidence rates and insulin detemir was negatively associated with total SH event rates.
1996
Severe hypoglycemia among patients starting basal insulin Ganz et al.
The regression-adjusted SH incidence and total event rates computed from the coefficients in Table 3 are slightly larger than the crude rates (10.36 [95% CI 6.67–13.95] and 11.21 [95% CI 8.81–13.61], respectively) and were also sensitive to histories of SH events in prior periods as was the case in the unadjusted analyses as shown at the bottom of Table 3.
Costs associated with severe hypoglycemia Each SH-related event cost, on average, $2636 (95% CI $1746–3525; results not shown). Patients who experienced a SH event during the first post-titration follow-up year www.cmrojournal.com ! 2014 Informa UK Ltd
Current Medical Research & Opinion
Volume 30, Number 10
October 2014
Table 4. Total and diabetes-related healthcare cost during the first post-titration follow-up year. All (N ¼ 7235)
Curr Med Res Opin Downloaded from informahealthcare.com by University of Washington on 12/02/14 For personal use only.
Mean Total Healthcare Costs Inpatient Outpatient Total Mean Total Diabetes-Related Costs Inpatient Outpatient Total
Experienced 1 Severe Hypoglycemic Event During the First Post-Titration Follow-up Year Yes (N ¼ 353)
No (N ¼ 6882)
p Value
$3538 ($3082, $3994) $1705 ($1648, $1763) $5244 ($4765, $5722)
$12,540 ($9781, $15,298) $2793 ($2430, $3155) $15,332 ($12,397, $18,267)
$3076 ($2621, $3532) $1650 ($1592, $1707) $4726 ($4249, $5203)
50.01 50.01 50.01
$2833 ($2406, $3261) $422 ($405, $438) $3255 ($2823, $3688)
$10,522 ($7993, $13,051) $803 ($680, $925) $11,325 ($8748, $13,901)
$2439 ($2011, $2867) $402 ($386, $418) $2841 ($2408, $3274)
50.01 50.01 50.01
In parentheses are 95% confidence intervals.
had higher healthcare costs during that year than those who did not experience a SH event ($15,332 vs. $4726 for all costs and $11,325 vs. $2841 for T2D-related costs, both p50.01; Table 4). Healthcare costs were statistically significantly higher in those quarters in which a SH event occurred than in those quarters without any SH events (coefficient of 1.80, p50.01; corresponding to predicted quarterly adjusted costs of $7199 [95% CI $5949–$8449] and $1192 [95% CI $1114–$1270], respectively; Table 5) as were T2D-related healthcare costs (coefficient of 2.11, p50.01; corresponding to predicted quarterly adjusted costs of $6070 [95% CI $4773–$7366] and $737 [95% CI $667–$807], respectively; Table 5). Older age was also associated with higher quarterly costs, as was a history of antihypertensive (and antihyperlipidemia, for T2Drelated costs) drugs and any inpatient hospital and emergency department use during the pre-index baseline period. A history of using metformin or other OADs (other than sulfonylureas) was associated with lower quarterly healthcare costs.
Discussion This study contributes to the literature on SH incidence and total event rates and costs associated with SH in T2D patients, especially those who have initiated basal insulin therapy, by providing current data on SH rates and costs. In addition to updating the literature on the rates of and costs associated with SH events, our results also present new information on how history of previous SH events, both before starting insulin and during the insulin titration period, impacts SH incidence and total SH event rates for patients who have initiated basal insulin. In particular, we found that patients who experienced SH events during either the pre-index baseline or the postindex titration periods were more likely to experience an SH event during the post-titration follow-up period (and experience more events) than patients who did not experience SH events during the pre-index baseline or ! 2014 Informa UK Ltd www.cmrojournal.com
post-index titration periods. This suggests that prior SH experience, whether or not on insulin treatment, might be predictive of SH responses to insulin. Moreover we found that the impact of SH events during the titration period on SH rates after the titration period was stronger than the impact of SH events during the baseline period. While this is not unexpected (since SH events during and subsequent to titration are related to basal insulin use) this information may still be clinically useful: while prior SH events as a whole predict future risk, it suggests that SH during the period prior to insulin use is less important for predicting SH response to insulin than subsequent reactions during either the introduction of insulin (titration) or while on maintenance insulin treatment. Our findings for the SH incidence rate (4.63 per 100 patient-years for the entire follow-up period and 5.02 per 100 patient-years for the first 12 months of follow-up) were broadly consistent with previously reported data on the proportion of patients using basal insulin who had at least one SH-related claim during a 12-month period, which ranged from 3.2% to 8.47%6,8,9,16–19 and our estimate of the average cost of an SH event ($2636) was also consistent with previously reported estimates, which ranged from $1049 to $16788,9. Our findings for the total SH event rate (9.69 per 100 patient-years for the entire follow-up period and 9.33 per 100 patient-years for the first 12 months of follow-up) were less consistent with previously reported data, which ranged from 10 to 75 per 100 patient-years, with half of those rates ranging from 10 to 23 per 100 patient-years6,8,9,16,17,19. Differences in how these studies defined SH events (using health insurance claims data) and in the proportion of patients who were also using bolus insulin are partly responsible for the wide range in rates reported. Previous studies that contained predominantly basal or basal-only therapy patients (similar to our analysis) reported annual event rates that were closer to, but still higher than, our results (12–63 events per 100 patient-years; 75th percentile, 23 per 100 patient years)6,9,19. Our study, unlike the previous studies cited above, focused on all types of T2D patients who started Severe hypoglycemia among patients starting basal insulin Ganz et al.
1997
Current Medical Research & Opinion Volume 30, Number 10
October 2014
Table 5. Generalized linear regression results for total healthcare and diabetes-related costs during the entire post-titration follow-up period. All Healthcare Costs
Curr Med Res Opin Downloaded from informahealthcare.com by University of Washington on 12/02/14 For personal use only.
Coefficient Experienced a SH During Quarter Age (vs. 18–24) 25–34 35–44 45–54 55–64 65–74 75 Sex (vs. Female) Male Race (vs. White) Black Asian/Other/Unknown Region (vs. Northeast) Midwest South West Index Insulin (vs. Glargine) NPH Detemir Diabetes-Related Comorbidities Metformin Use Sulfonylurea Use Use of Other OADs Use of Lipid-Lowering Drugs Use of Antihypertensive Drugs Endocrinologist Visits Inpatient Hospitalization Outpatient Services Emergency Department Use Constant Number of Observations Number of Patients Predicted Quarterly Costs: Experienced a SH During Quarter No Yes
95% CI
Diabetes-Related Costs Coefficient
95% CI
1.80***
(1.63, 1.97)
2.11***
(1.90, 2.32)
0.93** 1.31*** 1.46*** 1.64*** 1.87*** 1.78***
(0.22, 1.65) (0.48, 2.15) (0.86, 2.06) (1.04, 2.24) (1.27, 2.46) (1.16, 2.40)
0.99* 1.79*** 1.81*** 2.02*** 2.21*** 2.11***
(0.16, 2.14) (0.63, 2.95) (0.90, 2.72) (1.11, 2.93) (1.31, 3.11) (1.18, 3.05)
0.08 0.18** 0.04
(0.21, 0.05) (0.01, 0.34) (0.23, 0.32)
0.08
(0.26, 0.09)
0.25** 0.12
(0.04, 0.46) (0.26, 0.49)
0.38*** 0.48*** 0.24*
(0.57, 0.20) (0.72, 0.24) (0.49, 0.02)
0.41*** 0.41** 0.21
(0.67, 0.15) (0.75, 0.08) (0.57, 0.16)
0.27*** 0.10 0.11 0.45*** 0.11 0.14 0.06 0.19*** 0.00 0.99*** 0.17* 0.49*** 5.75*** 55,051 7231
(0.46, 0.09) (0.24, 0.05) (0.31, 0.09) (0.59, 0.30) (0.27, 0.04) (0.35, 0.06) (0.09, 0.20) (0.06, 0.33) (0.19, 0.19) (0.83, 1.15) (0.02, 0.36) (0.32, 0.67) (5.14, 6.36)
0.40*** 0.15 0.29** 0.54*** 0.20* 0.12 0.19* 0.26*** 0.06 1.06*** 0.06 0.76*** 4.95*** 55,051 7231
(0.66, 0.14) (0.35, 0.06) (0.54, 0.04) (0.74, 0.34) (0.42, 0.02) (0.39, 0.16) (0.02, 0.41) (0.08, 0.45) (0.20, 0.33) (0.85, 1.27) (0.25, 0.37) (0.54, 0.97) (4.03, 5.87)
$1192 $7199
(1114, $1270) ($5949, $8449)
$737 $6070
($667, $807) ($4773, $7366)
CI, confidence interval; NPH, neutral protamine Hagedorn; OADs, oral antidiabetic drugs; SH, severe hypoglycemia. ***p50.01. **p50.05. *p50.10.
using basal insulin (i.e., their first type of insulin was basal insulin); we also conditioned our analyses on history of previous SH events, which helps to further elaborate the risk factors for SH events once patients have started basal insulin and reached their maintenance phase. Although our overall results for SH incidence and total SH events rates are consistent with, and contribute to, the literature on SH, we must mention a number of important limitations and other factors of our analytic design that should be kept in mind while interpreting our results. Using claims and other administrative data to study the epidemiology and cost implications of SH events is a challenging task. Although SH events, by definition, require the assistance of a third party, not all SH events result in an interaction with a healthcare provider and therefore do 1998
Severe hypoglycemia among patients starting basal insulin Ganz et al.
not appear in such databases. In addition, SH events might not appear in those databases if the interaction did not result in a billable service or if an appropriate SH-related diagnosis code was not used. We addressed this limitation by using EHR data, which may be more likely to capture relevant events in patients’ lives than claims data, and by exploiting the clinical notes to identify SH events that did not result in an interaction with a healthcare provider. Although we did identify some additional cases of SH using these clinical notes, their use did not, however, dramatically improve our ability to identify additional SH events nor did those additional events dramatically change the incidence or total SH event rates. The marginal number of additional SH events identified using clinical notes may imply that such data, or our www.cmrojournal.com ! 2014 Informa UK Ltd
Curr Med Res Opin Downloaded from informahealthcare.com by University of Washington on 12/02/14 For personal use only.
Current Medical Research & Opinion
methods, were not sensitive enough to detect SH events that did not result in contact with the healthcare system or that SH events almost always result in contact with the healthcare system (as EHR systems become more available, as they become used more often, and as the methods for extracting information from those systems become more sophisticated, our ability to identify SH, and other, events that did not result in contact with the healthcare system may increase). Our ability to identify SH events and healthcare costs was further limited by the extent to which patients received all of their care from providers and facilities that provided data to the Humedica database. Furthermore, costs are also likely to be underestimated as a result of excluding pharmacy costs, as well as encounters and service that we could not link to external cost data. However because costs are not likely to be differentially underestimated by SH status, except for pharmacy costs, which would be expected to be higher for patients who have experienced SH, we have most likely underestimated, rather than overestimated, the impact of SH events on quarterly costs. Finally, although we divided the post-titration follow-up period into quarters (3 month blocks) to facilitate the repeated measures analysis and because quarters represent meaningful financial units, we recognize that other meaningful time units could have been used and could have resulted in different conclusions. Our result that SH is associated with higher costs for care delivered temporally close to the SH events themselves (i.e., in the same quarter in our analyses) are important because it implies that patients who experience SH events are not always more expensive than patients who do not experience SH — they are just more expensive when they actually experience those events. These findings can also inform economic evaluations of competing therapies with different SH profiles and further imply that interventions that can prevent or delay SH can be associated with significant economic benefits even within rather short time horizons. Further real-world studies on the frequency and costs of SH, using methods to identify as many SH events as possible, can allow healthcare providers to make more informed decisions on the risks and benefits of basal insulin therapy in T2D patients.
Conclusions These results contribute to the literature not only by providing additional evidence that the real-world burden of SH is high among people with T2D who start using basal insulin but also by presenting new information on how history of previous SH events, both before starting insulin and during the insulin titration period, impacts SH incidence and total SH event rates for patients who have initiated basal insulin. Prior events, whether or not while on insulin treatment, are predictive of SH responses to ! 2014 Informa UK Ltd www.cmrojournal.com
Volume 30, Number 10
October 2014
insulin; in particular, it appears that SH during the period prior to insulin use is less important for predicting SH response to insulin than subsequent reactions during either the introduction of insulin (titration) or while on maintenance insulin treatment. These results should be of interest to clinicians and to those who evaluate the clinical and cost effectiveness of current and future treatments for diabetes.
Transparency Declaration of funding This research was supported by Novo Nordisk Inc. Declaration of financial/other relationships M.L.G. and Q.L. have disclosed that they are employees of Evidera, which received research funds from Novo Nordisk Inc. to conduct this study and were so at the time this research was conducted. Y.-Q.L. and E.G. have disclosed that they were also employees of Evidera. N.S.W. and J.C.H. have disclosed that they are employees of Novo Nordisk Inc. and are also stockholders of Novo Nordisk Inc. CMRO peer reviewers on this manuscript have no relevant financial or other relationships to disclose.
References 1. Diabetes Fact Sheet No. 312. World Health Organization, 2013. Available at: http://www.who.int/mediacentre/factsheets/fs312/en/ [Last accessed 18 July 2013] 2. Shaw JE, Sicree RA, Zimmet PZ. Global estimates of the prevalence of diabetes for 2010 and 2030. Diabetes Res Clin Pract 2010;87:4-14 3. Centers for Disease Control and Prevention. 2011 National Diabetes Fact Sheet: national estimates and general information on diabetes and prediabetes in the United States. 2011. Available at: http://www.cdc.gov/diabetes/ pubs/factsheet11.htm [Last accessed 8 May 2013] 4. The Diabetes Control and Complications Trial Research Group. The effect of intensive treatment of diabetes on the development and progression of longterm complications in insulin-dependent diabetes mellitus. N Engl J Med 1993;329:977-86 5. Lin YY, Hsu CW, Sheu WH, et al. Risk factors for recurrent hypoglycemia in hospitalized diabetic patients admitted for severe hypoglycemia. Yonsei Med J 2010;51:367-74 6. Baser O, Wei W, Baser E, Xie L. Clinical and economic outcomes in patients with type 2 diabetes initiating insulin glargine disposable pen versus exenatide BID. J Med Econ 2011;14:673-80 7. Zhang Q, Menditto L. Incremental cost savings 6 months following initiation of insulin glargine in a Medicaid fee-for-service sample. Am J Ther 2005;12:337-43 8. Bullano MF, Fisher MD, Grochulski WD, et al. Hypoglycemic events and glycosylated hemoglobin values in patients with type 2 diabetes mellitus newly initiated on insulin glargine or premixed insulin combination products. Am J Health Syst Pharm 2006;63:2473-82 9. Fabunmi R, Nielsen LL, Quimbo R, et al. Patient characteristics, drug adherence patterns, and hypoglycemia costs for patients with type 2 diabetes mellitus newly initiated on exenatide or insulin glargine. Curr Med Res Opin 2009;25:777-86 10. Brixner DI, McAdam-Marx C, Ye X, et al. Six-month outcomes on A1C and cardiovascular risk factors in patients with type 2 diabetes treated with exenatide in an ambulatory care setting. Diabetes Obes Metab 2009;11:1122-30
Severe hypoglycemia among patients starting basal insulin Ganz et al.
1999
Current Medical Research & Opinion Volume 30, Number 10
October 2014
16. Bullano MF, Al-Zakwani IS, Fisher MD, et al. Differences in hypoglycemia event rates and associated cost–consequence in patients initiated on longacting and intermediate-acting insulin products. Curr Med Res Opin 2005;21:291-8 17. Davis SN, Wei W, Garg S. Clinical impact of initiating insulin glargine therapy with disposable pen versus vial in patients with type 2 diabetes mellitus in a managed care setting. Endocr Pract 2011;17:845-52 18. Pawaskar M, Zagar A, Sugihara T, Shi L. Healthcare resource utilization and costs assessment of type 2 diabetes patients initiating exenatide BID or glargine: a retrospective database analysis. J Med Econ 2011;14: 16-27 19. Xie L, Wei W, Pan C, et al. A real-world study of patients with type 2 diabetes initiating basal insulins via disposable pens. Adv Ther 2011;28:1000-11
Curr Med Res Opin Downloaded from informahealthcare.com by University of Washington on 12/02/14 For personal use only.
11. Pawaskar M, Li Q, Hoogwerf BJ, et al. Metabolic outcomes of matched patient populations initiating exenatide bid versus insulin glargine in an ambulatory care setting. Diabetes Obes Metab 2012;14:626-33 12. Ginde AA, Blanc PG, Lieberman RM, et al. Validation of ICD-9-CM coding algorithm for improved identification of hypoglycemia visits. BMC Endocr Disord 2008;8:4 13. Overview of the Nationwide Inpatient Sample (NIS). Healthcare Costs and Utilization Project (HCUP). 2013. Available at: http://www.hcup-us.ahrq.gov/nisoverview.jsp [Last accessed 30 June 2014] 14. Graubard BI, Korn EL. Predictive margins with survey data. Biometrics 1999;55:652-9 15. Kleinman LC, Norton EC. What’s the risk? A simple approach for estimating adjusted risk measures from nonlinear models including logistic regression. Health Serv Res 2009;44:288-302
2000
Severe hypoglycemia among patients starting basal insulin Ganz et al.
www.cmrojournal.com ! 2014 Informa UK Ltd
Vol. 30, No. 10, 2014, 1991–2000
Article ST-0439.R1/936930 All rights reserved: reproduction in whole or part not permitted
Curr Med Res Opin Downloaded from informahealthcare.com by University of Washington on 12/02/14 For personal use only.
Original Article Severe hypoglycemia rates and associated costs among type 2 diabetics starting basal insulin therapy in the United States
Michael L. Ganz
Abstract
Evidera, Lexington, MA, USA
Neil S. Wintfeld Novo Nordisk Inc., Plainsboro, NJ, USA
Qian Li Yuan-Chi Lee Elyse Gatt Evidera, Lexington, MA, USA
Joanna C. Huang Novo Nordisk Inc., Plainsboro, NJ, USA Address for correspondence: Michael L. Ganz MS, PhD, Evidera, Retrospective Observational Studies, 430 Bedford Street, Lexington, 02420, USA. [email protected] Keywords: Costs and cost analysis – Diabetes mellitus, type 2 – Hypoglycemia – Insulin – Prevalence studies Accepted: 16 June 2014; published online: 15 July 2014 Citation: Curr Med Res Opin 2014; 30:1991–2000
Objectives: To derive current real-world data on the rates and costs of severe hypoglycemia (SH) for people with type 2 diabetes mellitus (T2D) who have initiated basal insulin therapy and to examine differences in SH rates and costs stratified by history of prior SH events. Methods: We used a nation-wide electronic health records database that included encounter and laboratory data, as well as clinical notes, to estimate the rates and costs of SH events among adults with T2D who initiated basal insulin between 2008 and 2011. Unadjusted and regression-adjusted rates and quarterly costs were calculated for all patients as well as stratified by history of a SH event before starting basal insulin and history of a SH event during the basal insulin titration period. Results: We identified 7235 incident cases of basal insulin use among patients with T2D who did not use insulin during the previous 12 months. Regression-adjusted incidence and total event rates were 10.36 and 11.21 per 100 patient-years, respectively. A history of SH events during the pre-index baseline and post-index titration periods were statistically significantly associated with both the incidence and total event rates (p50.01). Regression-adjusted total healthcare and diabetes-related costs were statistically significantly (p50.01) higher in those quarters when a SH event occurred than in those quarters without any SH events ($3591 vs. $487 and $3311 vs. $406, respectively). A history of previous SH or SH events during the titration period were not statistically significantly associated with costs. Conclusions: These results suggest that the real-world burden of SH is high among people with T2D who start using basal insulin and that history of previous SH events, both before starting insulin and during the insulin titration period, influences future SH. These results can also provide insights into interventions that can prevent or delay SH. These results should, however, be interpreted in light of the key limitations of our study: not all SH events may have been captured or coded in the database, data on filled prescriptions were not available, and the post-titration follow-up period could have been divided into time units other than quarters (3 month blocks) resulting in potentially different conclusions. Further real-world studies on the frequency and costs of SH, using methods to identify as many SH events as possible, can allow healthcare providers to make more informed decisions on the risks and benefits of basal insulin therapy in T2D patients.
Introduction The World Health Organization (WHO) has classified diabetes as a global epidemic1. Approximately 285 million (6.4%) adults worldwide have diabetes2, with the US responsible for almost 9% of those cases (25.6 million adults, ! 2014 Informa UK Ltd www.cmrojournal.com
Severe hypoglycemia among patients starting basal insulin Ganz et al.
1991
Curr Med Res Opin Downloaded from informahealthcare.com by University of Washington on 12/02/14 For personal use only.
Current Medical Research & Opinion Volume 30, Number 10
October 2014
corresponding to 11.3% of the US adult population)3. By the year 2030 the number of adults with diabetes is expected to grow substantially to almost 439 million globally and to almost 36 million in the US2. Reflecting its prevalence and associated morbidity, diabetes is responsible for $116 billion in annual direct medical costs (people with diabetes incur 2.3 times higher costs than otherwise similar people without diabetes) and for $58 billion annually in disability, lost work productivity and premature mortality in the US3. Hypoglycemia, a deficiency of glucose in the bloodstream, is a serious condition related to the use of antidiabetic medications and is a significant cause of morbidity and mortality in people with diabetes. Approximately 20% of diabetics who are treated with antidiabetic medications are expected to have symptoms related to hypoglycemia4. Severe hypoglycemia (SH), or hypoglycemia that requires the assistance of a third party, if unrecognized and untreated, can result in coma, seizures, or death5. Recent studies have estimated, not withstanding how SH was defined, that between 3.5% (annual)6 and 30.4% (first 6 months)7 of people with diabetes have experienced at least one SH while receiving antidiabetic therapy. The occurrence of SH is frequently followed by treatment adjustments involving the glycemic target, which can result in long-term diabetes complications, increased use of healthcare services, and substantial costs. The mean cost per hypoglycemic event, including inpatient, outpatient, and emergency department costs, has been reported to range from $1049 (2004 USD) to $1331 (2007 USD) per event8,9. Understanding the epidemiology and cost implications of SH is important for being able to project future demands on the public health system and for being able to evaluate the cost effectiveness of current and future treatments for diabetes. Despite this, there are relatively little recent real-world data on the rates of and costs associated with SH, especially for people with type 2 diabetes mellitus (T2D) who have initiated basal insulin therapy. Our objectives were, therefore, to estimate the real-world incidence and total SH event rates and the healthcare costs among people with (T2D) who have initiated basal insulin in US.
Materials and methods Data source We analyzed encounters and services that occurred between January 2008 and December 2011 using data from Humedica Inc.’s Real-time Longitudinal Clinical Data patient-level database. The Humedica database contains data abstracted from electronic health records (EHR) integrated with claims, prescription, and practice 1992
Severe hypoglycemia among patients starting basal insulin Ganz et al.
management data. The database reflects the real-world patient care experiences, including laboratory results and radiology reports, physician and nurse notes, prescriptions written, procedures, diagnoses, and other details of a patient’s office visit and/or hospital stay. The database includes information from approximately 30 million patients derived from 70 hospitals, 435 outpatient clinics, and 17,600 physicians in 38 states.
Patient selection We selected patients who had a T2D diagnosis using standard definitions in the context of EHR and administrative databases10,11. In particular, at least one of the following criteria had to be satisfied for a patient to qualify as a T2D patient: (1) International Classification of Diseases, Ninth Revision (ICD-9-CM) diagnosis codes of 250.x0 or 250.x2; (2) one or more prescription orders for a non-insulin antidiabetic drug (NIAD); (3) two consecutive fasting blood glucose levels of 126 mg/dl; or (4) glycated hemoglobin (HbA1c) 7.0%. Patients had to have started basal insulin without having used any type of insulin for at least the previous 12 months (we refer to a patient’s first prescription of basal insulin as the ‘index date’). Patients had to be at least 18 years old as of their index date and have at least 12 months of continuous eligibility in the database before and after their index date. Patients were further excluded if they were pregnant or had a diagnosis of Type 1, secondary, or gestational diabetes (ICD-9-CM codes 250.x1, 249.xx, and 648.0 or 648.8, respectively).
Study measures Patient measures included demographic and clinical characteristics such as age, sex, race, geographic region, smoking status, the type of insulin patients started, history of visiting an endocrinologist, use of oral antidiabetic drugs (OADs) and other medications, and other healthcare resource use during the 12 months prior to starting basal insulin. The presence of key comorbid conditions associated with diabetes outcomes was also assessed during the 12 months prior to starting basal insulin. A complete list of the baseline characteristics are presented in Table 1 (a list of the diabetes-related comorbid conditions is in Appendix 1). Severe hypoglycemia events SH events were defined as events requiring medical attention with the appropriate SH diagnosis codes attached to outpatient, inpatient, or emergency department visits or by a recorded glucose level of 40 mg/dL. The presence of diagnosis codes 251.0x, 251.1x, 251.2x, or 250.3x on different days or the presence of diagnosis codes 251.8x without codes 259.8x, 272.7, 681.xx, 682.xx, 686.9x, 707.1x– www.cmrojournal.com ! 2014 Informa UK Ltd
Current Medical Research & Opinion
Curr Med Res Opin Downloaded from informahealthcare.com by University of Washington on 12/02/14 For personal use only.
Table 1. Demographic and pre-index baseline clinical characteristics of type 2 diabetes mellitus patients initiating basal insulin (N ¼ 7235). Age (%) 18–24 25–34 35–44 45–54 55–64 65–74 75 Mean Age (SD) Sex (%) Male Female Race (%) White Black Asian/Other/Unknown Region (%) Northeast Midwest South West Other/Unknown Index Insulin (%) Glargine NPH Detemir History of Diabetes-Related Comorbidities (%) History of Healthcare Use (%) Endocrinologist Visits Inpatient Hospitalization Outpatient Services Emergency Department Use History of Medication Use (%) Metformin Sulfonylurea Other OADs Lipid Lowering Drugs Antihypertensive Drugs
0.43 1.87 7.37 18.87 29.51 35.44 6.52 60.82 (11.65) 49.30 50.70 63.25 19.09 17.66 8.06 50.68 29.38 11.82 0.06 77.24 5.86 16.90 80.17 13.74 9.01 4.34 5.47 36.66 38.06 25.82 32.51 51.93
SD, standard deviation; NPH, neutral protamine Hagedorn; OADs, oral antidiabetic drugs.
707.9x, 709.3x, 730.0x–730.2x, or 731.8x indicated a SH event12. Humedica uses state of the art natural language processing technology to extract data from physician, radiology, and pathology notes for over 8.5 million patients. We used these notes, available for two-thirds of the patients, to identify SH events not otherwise captured by diagnosis codes by searching for evidence of hypoglycemic events that occurred without immediate medical attention, but had serious sequelae (such as paresthesias, coma, loss of consciousness, and need for glucagon) that required the assistance of another person. SH on consecutive days was considered a single SH event unless interrupted by at least one day without evidence of SH, in which case each run of consecutive days of SH were considered separate SH events. Healthcare costs The Humedica database provides information on healthcare utilization at the encounter and service level, but it does not contain cost information. We, therefore, ! 2014 Informa UK Ltd www.cmrojournal.com
Volume 30, Number 10
October 2014
obtained costs from external sources. Inpatient stays recorded in the Humedica database were matched to similar hospitalizations in the National Inpatient Sample based on Clinical Classification Scheme (CCS) codes13 and geographic region to obtain predicted hospitalization costs. Costs for other encounters (outpatient, office, and emergency visits), except those that happened during an inpatient stay, were estimated by linking those encounters to the Medicare fee schedule using procedure codes. Encounters and services that could not be linked (26% of all encounters and services, which were predominantly laboratory and non-durable medical equipment services) were ignored. Although data were available on prescriptions that were written, which were used to determine inclusion eligibility, no data were available on filled prescriptions, so we were not able to include pharmacy costs in these analyses. Further details on these methods are provided in Appendix 2.
Statistical analyses We defined three study periods: the 12 months prior to the index date (‘pre-index baseline’ period), the 16-week period immediately following the index date (‘post-index titration’ period) during which we assumed patients learned how to correctly use their insulin and made adjustments to their dose, and the remaining time until the end of the eligibility in the database (‘post-titration follow-up’ period). Means and standard deviations were calculated for continuous variables, including healthcare costs during the first year of the post-titration follow-up period, and frequency distributions were reported for categorical variables; differences in mean costs between patients who did and did not experience SH events were assessed using the two-tailed t test (differences with p-values less than 0.05 were considered statistically significant). Incidence rates were calculated by dividing the total number of first SH events by the total number of patient-years observed from the end of the titration period until the first SH event for patients who eventually experienced a SH event or to the end of the post-titration follow-up period for patients who never experienced a SH event. Total SH event rates were calculated by dividing the total number of SH events that occurred during the post-titration follow-up period by the total number of available patient-years. Exact confidence intervals around these rates were computed using the Poisson distribution. We further stratified the crude incidence and total event rates by history of SH during the pre-index baseline and post-index titration periods to assess how these rates varied according to exposure to prior SH events. The association between the SH rates and history of any SH events during the pre-index baseline and post-index titration periods were estimated using negative binomial regression Severe hypoglycemia among patients starting basal insulin Ganz et al.
1993
Curr Med Res Opin Downloaded from informahealthcare.com by University of Washington on 12/02/14 For personal use only.
Current Medical Research & Opinion Volume 30, Number 10
October 2014
models adjusting for patient demographic and clinical characteristics. We assessed the association of SH events with healthcare costs (overall and for diabetes-related encounters and services) during the entire post-titration period using a multiple generalized linear regression model (gamma distribution and a log-link function), adjusting for patient demographic and clinical characteristics, to address the skewed distribution of costs (we did not adjust for history of any SH events during the pre-index baseline and post-index titration periods because those factors were not statistically significantly associated with costs in our preliminary analyses). Because we wanted to estimate the proximal effect of SH events on costs and because of the variable lengths of follow-up, we estimated the association of the occurrence of an SH event in a given quarter with that quarter’s costs (standard errors were corrected for multiple observations per patient). We applied the method of recycled predictions to the negative binomial regression results to compute the adjusted incidence and total SH event rates conditional on experiencing any SH events during the pre-index baseline and post-index titration periods and to the generalized linear regression results to compute adjusted healthcare costs conditional on experiencing any SH events during the pre-index baseline and post-index titration periods (associated standard errors were computed using the delta method)14,15. Recycled predictions of scenarios of interest were obtained by averaging the regressionadjusted estimates of the outcome variable for each individual after assigning each individual to membership in each category of the scenario of interest while using the observed values for the other variables. For example, to compute the adjusted effect of experiencing a SH during the pre-index baseline period, we calculated predicted costs for each individual after assigning the SH variable ^ 0 and after assigning the SH variable to to zero to obtain C i 1 ^ . The average costs associated with not one to obtain C i ^0 0 ¼ PN C experiencing a SH are calculated as C i¼1 i /N and the average costs associated with experiencing a SH are ^ 1 /N. The incremental effect of 1 ¼ PN C calculated as C i¼1 i 0. 1 C experiencing an SH is therefore DC=C
Approximately three-quarters of the patients were 55 years old or older (mean age 61), about 50% were female, almost two-thirds were white, and most patients resided in the South or Midwest (Table 1). A vast majority (80%) of patients experienced at least one diabetes-related comorbid condition in the year before they started basal Patients with a diagnosis of T2D from 1 January 2008 to 31 December 2011 N = 425,327
Patients with a basal insulin prescription after continuous enrollment start N = 67,086
Patients with at least 12 months continuous enrollment prior to the index date N = 30,621
Exclude patients with any insulin use prior to the index date N = 19,474
Age ≥18 as of the index date N = 19,381
Exclude diagnosed type 1 diabetes N = 16,654
Exclude diagnosed secondary diabetes N = 16,435
Exclude diagnosed gestational diabetes N = 16,234
Results We identified 7235 T2D patients who were new users of basal (detemir, glargine, or neutral protamine Hagedorn) insulin (i.e., did not use any insulin during the previous 12 months) who satisfied our inclusion and exclusion criteria (Figure 1), with 568 of those patients (7.85%) experiencing at least one SH event (1245 SH events in total) during the post-titration follow-up period (548 patients [1189 SH events] based on the encounter data and 20 patients [56 SH events] based on the clinical notes). 1994
Severe hypoglycemia among patients starting basal insulin Ganz et al.
Exclude pregnant patients N = 16,172
Patients were continuously enrolled for ≥ 16-week titration period + 12-month follow-up period N = 7235
Figure 1. Sample selection flowchart.
www.cmrojournal.com ! 2014 Informa UK Ltd
! 2014 Informa UK Ltd www.cmrojournal.com
5.91 (5.76–6.06) 9.00 (8.87–9.12)
37.59 (36.73–38.47) 90.69 (89.58–91.80)
4.11 (4.07–4.14) 7.86 (7.81–7.91)
9.15 (8.93–9.38) 11.54 (11.38–11.70)
October 2014
In parentheses are 95% confidence intervals. SH, severe hypoglycemia.
4.17 (4.13–4.20) 7.92 (7.87–7.97) 24.64 (24.06–25.23) 71.28 (70.41–72.16) 4.63 (4.59–4.67) 9.69 (9.64–9.75)
No (N ¼ 7064) Yes (N ¼ 171) Rate Ratio [Yes:No] No (N ¼ 7026) Yes (N ¼ 209)
Volume 30, Number 10
Incidence Rate (Per 100 Patient-Years) Total Event Rate (Per 100 Patients-Years)
All (N ¼ 7235)
The mean post-titration follow-up length was 1.8 years. Because follow-up times differed by history of SH events during the pre-index baseline (1.72 vs. 1.78 years for patients with and without baseline SH events, p ¼ 0.08) and the post-index titration (1.66 vs. 1.78 years for patients with and without titration SH events, p50.01) periods, we present results for the SH incidence and total event rates stratified by SH histories during the baseline and titration periods. As shown in Table 2, the crude incidence and total SH event rates during the entire post-titration follow-up period were 4.63 (95% CI 4.59–4.67) and 9.69 (95% CI 9.64–9.75) per 100 patient-years, respectively (the rates for the first post-titration follow-up year were similar at 5.02 and 9.33 per 100 patient-years; not shown). These rates differed substantially by a patient’s history of SH events during the pre-index baseline and the post-index titration periods and indicated that not only is a history of previous SH events associated with SH events during the post-titration period, but that the history of SH events during the titration period has a stronger association with post-titration SH events than the pre-index history of SH events. The incidence rate for patients with a history of SH events during the pre-index baseline period was 5.91 (95% CI 5.76–6.06) times higher than those without a SH history but was 9.15 (95% CI 8.93–9.38) times higher for those with a history of SH events during the post-index titration period than those without a history of SH. Similarly, the total SH event rate for patients with a history of SH events during the pre-index baseline period was 9.00 (95% CI 8.87–9.12) times higher than those without a SH history but was 11.54 (95% CI 11.38–11.70) times higher for those with a history of SH events during the post-index titration period than those without a history of SH. The incidence and total event rates computed for the first post-titration follow-up year, as well as the impacts of histories of SH events during the pre-index baseline and post-index titrations periods, were similar to the rates for the entire post-titration follow-up period (not shown). Table 3 displays the results of both of the multivariable negative binomial regression models for the SH incidence and total SH event rates. The risk of a SH event subsequent to titration was positively and significantly associated with history of SH events during the pre-index baseline period (coefficients of 1.82 and 1.77 for SH incidence rates and total events rates, respectively; both
Table 2. Crude severe hypoglycemia event incidence and total events rates during the post-titration follow-up period.
Curr Med Res Opin Downloaded from informahealthcare.com by University of Washington on 12/02/14 For personal use only.
Incidence and total severe hypoglycemia event rates
History of SH During Pre-Index Baseline Period
History of SH During Post-Index Titration Period
insulin (the most prevalent comorbid conditions included hypertension [61%], dyslipidemia or hyperlipidemia [54%], neuropathy [21%], nephropathy or nephrosis [13%], and obesity [11%]; see Appendix 1).
Rate Ratio [Yes:No]
Current Medical Research & Opinion
Severe hypoglycemia among patients starting basal insulin Ganz et al.
1995
Current Medical Research & Opinion Volume 30, Number 10
October 2014
Table 3. Multivariable negative binomial regression results for incident and total SH events during the post-titration follow-up period. Incident SH Events Coefficient
Curr Med Res Opin Downloaded from informahealthcare.com by University of Washington on 12/02/14 For personal use only.
History of SH Events during Pre-Index Baseline Period History of SH Events during Post-Titration Follow-Up Period Age (vs. 18–24) 25–34 35–44 45–54 55–64 65–74 75 Sex (vs. Female) Male Race (vs. White) Black Asian/Other/Unknown Region (vs. Northeast) Midwest South West Index Insulin (vs. Glargine) NPH Detemir Diabetes-Related Comorbidities Metformin Use Sulfonylurea Use Use of Other OADs Use of Lipid-Lowering Drugs Use of Antihypertensive Drugs Endocrinologist Visits Inpatient Hospitalization Outpatient Services Emergency Department Use Constant Number of Patients Predicted Rates: Overall History of SH During Pre-Index Baseline No Yes History of SH During Post-Index Titration No Yes
1.82*** 2.85***
Total SH Events
95% CI
Coefficient
(1.35, 2.29) (2.33, 3.37)
1.77*** 2.05***
95% CI (1.28, 2.25) (1.53, 2.58)
0.10 0.25 0.17 0.14 0.23 0.54
(1.86, 1.66) (1.88, 1.37) (1.76, 1.42) (1.72, 1.44) (1.35, 1.81) (1.07, 2.15)
0.42 0.27 0.20 0.05 0.27 0.38
(1.36, 2.21) (1.95, 1.42) (1.85, 1.46) (1.59, 1.69) (1.37, 1.91) (1.30, 2.06)
0.06
(0.26, 0.15)
0.12
(0.08, 0.33)
0.27** 0.09
(0.00, 0.53) (0.39, 0.21)
0.30** 0.21
(0.03, 0.57) (0.52, 0.09)
0.25 0.16 0.11
(0.16, 0.66) (0.61, 0.28) (0.62, 0.39)
0.04 0.47** 0.42*
(0.36, 0.44) (0.91, 0.04) (0.91, 0.07)
0.25 0.25* 0.10 0.26** 0.03 0.05 0.20 0.17 0.06 0.41** 0.18 0.22 1.97** 7231
(0.17, 0.66) (0.53, 0.04) (0.39, 0.18) (0.51, 0.01) (0.29, 0.23) (0.31, 0.21) (0.46, 0.06) (0.08, 0.41) (0.25, 0.37) (0.08, 0.75) (0.32, 0.68) (0.20, 0.64) (0.32, 3.61)
0.43** 0.08 0.08 0.35*** 0.01 0.13 0.11 0.16 0.23 0.10 0.00 0.20 1.43* 7231
(0.04, 0.82) (0.36, 0.20) (0.36, 0.20) (0.61, 0.10) (0.25, 0.26) (0.39, 0.13) (0.36, 0.15) (0.08, 0.40) (0.07, 0.54) (0.23, 0.44) (0.51, 0.50) (0.21, 0.61) (0.18, 3.05)
10.36
(6.76, 13.95)
11.21
(8.81, 13.61)
6.60 40.67
(5.20, 7.99) (20.18, 61.67)
8.27 48.33
(7.16, 9.38) (25.46, 71.19)
5.35 92.45
(4.59, 6.11) (42.66, 142.24)
8.13 63.26
(7.10, 9.15) (30.86, 95.67)
CI, confidence interval; NPH, neutral protamine Hagedorn; OADs, oral antidiabetic drugs; SH, severe hypoglycemia. ***p50.01. **p50.05. *p50.10.
p50.01) as well as with history of SH events during the post-titration period (coefficients of 2.85 and 2.05 for SH incidence rates and total events rates, respectively; both p50.01) These results were consistent with the unadjusted analyses. Other factors were also associated with SH incidence and total event rates. A history of metformin use during the pre-index baseline period was negatively associated both incidence and total SH events and history of inpatient hospital use during the pre-index baseline period was positively associated with total SH events. NPH insulin use was positively associated with incidence rates and insulin detemir was negatively associated with total SH event rates.
1996
Severe hypoglycemia among patients starting basal insulin Ganz et al.
The regression-adjusted SH incidence and total event rates computed from the coefficients in Table 3 are slightly larger than the crude rates (10.36 [95% CI 6.67–13.95] and 11.21 [95% CI 8.81–13.61], respectively) and were also sensitive to histories of SH events in prior periods as was the case in the unadjusted analyses as shown at the bottom of Table 3.
Costs associated with severe hypoglycemia Each SH-related event cost, on average, $2636 (95% CI $1746–3525; results not shown). Patients who experienced a SH event during the first post-titration follow-up year www.cmrojournal.com ! 2014 Informa UK Ltd
Current Medical Research & Opinion
Volume 30, Number 10
October 2014
Table 4. Total and diabetes-related healthcare cost during the first post-titration follow-up year. All (N ¼ 7235)
Curr Med Res Opin Downloaded from informahealthcare.com by University of Washington on 12/02/14 For personal use only.
Mean Total Healthcare Costs Inpatient Outpatient Total Mean Total Diabetes-Related Costs Inpatient Outpatient Total
Experienced 1 Severe Hypoglycemic Event During the First Post-Titration Follow-up Year Yes (N ¼ 353)
No (N ¼ 6882)
p Value
$3538 ($3082, $3994) $1705 ($1648, $1763) $5244 ($4765, $5722)
$12,540 ($9781, $15,298) $2793 ($2430, $3155) $15,332 ($12,397, $18,267)
$3076 ($2621, $3532) $1650 ($1592, $1707) $4726 ($4249, $5203)
50.01 50.01 50.01
$2833 ($2406, $3261) $422 ($405, $438) $3255 ($2823, $3688)
$10,522 ($7993, $13,051) $803 ($680, $925) $11,325 ($8748, $13,901)
$2439 ($2011, $2867) $402 ($386, $418) $2841 ($2408, $3274)
50.01 50.01 50.01
In parentheses are 95% confidence intervals.
had higher healthcare costs during that year than those who did not experience a SH event ($15,332 vs. $4726 for all costs and $11,325 vs. $2841 for T2D-related costs, both p50.01; Table 4). Healthcare costs were statistically significantly higher in those quarters in which a SH event occurred than in those quarters without any SH events (coefficient of 1.80, p50.01; corresponding to predicted quarterly adjusted costs of $7199 [95% CI $5949–$8449] and $1192 [95% CI $1114–$1270], respectively; Table 5) as were T2D-related healthcare costs (coefficient of 2.11, p50.01; corresponding to predicted quarterly adjusted costs of $6070 [95% CI $4773–$7366] and $737 [95% CI $667–$807], respectively; Table 5). Older age was also associated with higher quarterly costs, as was a history of antihypertensive (and antihyperlipidemia, for T2Drelated costs) drugs and any inpatient hospital and emergency department use during the pre-index baseline period. A history of using metformin or other OADs (other than sulfonylureas) was associated with lower quarterly healthcare costs.
Discussion This study contributes to the literature on SH incidence and total event rates and costs associated with SH in T2D patients, especially those who have initiated basal insulin therapy, by providing current data on SH rates and costs. In addition to updating the literature on the rates of and costs associated with SH events, our results also present new information on how history of previous SH events, both before starting insulin and during the insulin titration period, impacts SH incidence and total SH event rates for patients who have initiated basal insulin. In particular, we found that patients who experienced SH events during either the pre-index baseline or the postindex titration periods were more likely to experience an SH event during the post-titration follow-up period (and experience more events) than patients who did not experience SH events during the pre-index baseline or ! 2014 Informa UK Ltd www.cmrojournal.com
post-index titration periods. This suggests that prior SH experience, whether or not on insulin treatment, might be predictive of SH responses to insulin. Moreover we found that the impact of SH events during the titration period on SH rates after the titration period was stronger than the impact of SH events during the baseline period. While this is not unexpected (since SH events during and subsequent to titration are related to basal insulin use) this information may still be clinically useful: while prior SH events as a whole predict future risk, it suggests that SH during the period prior to insulin use is less important for predicting SH response to insulin than subsequent reactions during either the introduction of insulin (titration) or while on maintenance insulin treatment. Our findings for the SH incidence rate (4.63 per 100 patient-years for the entire follow-up period and 5.02 per 100 patient-years for the first 12 months of follow-up) were broadly consistent with previously reported data on the proportion of patients using basal insulin who had at least one SH-related claim during a 12-month period, which ranged from 3.2% to 8.47%6,8,9,16–19 and our estimate of the average cost of an SH event ($2636) was also consistent with previously reported estimates, which ranged from $1049 to $16788,9. Our findings for the total SH event rate (9.69 per 100 patient-years for the entire follow-up period and 9.33 per 100 patient-years for the first 12 months of follow-up) were less consistent with previously reported data, which ranged from 10 to 75 per 100 patient-years, with half of those rates ranging from 10 to 23 per 100 patient-years6,8,9,16,17,19. Differences in how these studies defined SH events (using health insurance claims data) and in the proportion of patients who were also using bolus insulin are partly responsible for the wide range in rates reported. Previous studies that contained predominantly basal or basal-only therapy patients (similar to our analysis) reported annual event rates that were closer to, but still higher than, our results (12–63 events per 100 patient-years; 75th percentile, 23 per 100 patient years)6,9,19. Our study, unlike the previous studies cited above, focused on all types of T2D patients who started Severe hypoglycemia among patients starting basal insulin Ganz et al.
1997
Current Medical Research & Opinion Volume 30, Number 10
October 2014
Table 5. Generalized linear regression results for total healthcare and diabetes-related costs during the entire post-titration follow-up period. All Healthcare Costs
Curr Med Res Opin Downloaded from informahealthcare.com by University of Washington on 12/02/14 For personal use only.
Coefficient Experienced a SH During Quarter Age (vs. 18–24) 25–34 35–44 45–54 55–64 65–74 75 Sex (vs. Female) Male Race (vs. White) Black Asian/Other/Unknown Region (vs. Northeast) Midwest South West Index Insulin (vs. Glargine) NPH Detemir Diabetes-Related Comorbidities Metformin Use Sulfonylurea Use Use of Other OADs Use of Lipid-Lowering Drugs Use of Antihypertensive Drugs Endocrinologist Visits Inpatient Hospitalization Outpatient Services Emergency Department Use Constant Number of Observations Number of Patients Predicted Quarterly Costs: Experienced a SH During Quarter No Yes
95% CI
Diabetes-Related Costs Coefficient
95% CI
1.80***
(1.63, 1.97)
2.11***
(1.90, 2.32)
0.93** 1.31*** 1.46*** 1.64*** 1.87*** 1.78***
(0.22, 1.65) (0.48, 2.15) (0.86, 2.06) (1.04, 2.24) (1.27, 2.46) (1.16, 2.40)
0.99* 1.79*** 1.81*** 2.02*** 2.21*** 2.11***
(0.16, 2.14) (0.63, 2.95) (0.90, 2.72) (1.11, 2.93) (1.31, 3.11) (1.18, 3.05)
0.08 0.18** 0.04
(0.21, 0.05) (0.01, 0.34) (0.23, 0.32)
0.08
(0.26, 0.09)
0.25** 0.12
(0.04, 0.46) (0.26, 0.49)
0.38*** 0.48*** 0.24*
(0.57, 0.20) (0.72, 0.24) (0.49, 0.02)
0.41*** 0.41** 0.21
(0.67, 0.15) (0.75, 0.08) (0.57, 0.16)
0.27*** 0.10 0.11 0.45*** 0.11 0.14 0.06 0.19*** 0.00 0.99*** 0.17* 0.49*** 5.75*** 55,051 7231
(0.46, 0.09) (0.24, 0.05) (0.31, 0.09) (0.59, 0.30) (0.27, 0.04) (0.35, 0.06) (0.09, 0.20) (0.06, 0.33) (0.19, 0.19) (0.83, 1.15) (0.02, 0.36) (0.32, 0.67) (5.14, 6.36)
0.40*** 0.15 0.29** 0.54*** 0.20* 0.12 0.19* 0.26*** 0.06 1.06*** 0.06 0.76*** 4.95*** 55,051 7231
(0.66, 0.14) (0.35, 0.06) (0.54, 0.04) (0.74, 0.34) (0.42, 0.02) (0.39, 0.16) (0.02, 0.41) (0.08, 0.45) (0.20, 0.33) (0.85, 1.27) (0.25, 0.37) (0.54, 0.97) (4.03, 5.87)
$1192 $7199
(1114, $1270) ($5949, $8449)
$737 $6070
($667, $807) ($4773, $7366)
CI, confidence interval; NPH, neutral protamine Hagedorn; OADs, oral antidiabetic drugs; SH, severe hypoglycemia. ***p50.01. **p50.05. *p50.10.
using basal insulin (i.e., their first type of insulin was basal insulin); we also conditioned our analyses on history of previous SH events, which helps to further elaborate the risk factors for SH events once patients have started basal insulin and reached their maintenance phase. Although our overall results for SH incidence and total SH events rates are consistent with, and contribute to, the literature on SH, we must mention a number of important limitations and other factors of our analytic design that should be kept in mind while interpreting our results. Using claims and other administrative data to study the epidemiology and cost implications of SH events is a challenging task. Although SH events, by definition, require the assistance of a third party, not all SH events result in an interaction with a healthcare provider and therefore do 1998
Severe hypoglycemia among patients starting basal insulin Ganz et al.
not appear in such databases. In addition, SH events might not appear in those databases if the interaction did not result in a billable service or if an appropriate SH-related diagnosis code was not used. We addressed this limitation by using EHR data, which may be more likely to capture relevant events in patients’ lives than claims data, and by exploiting the clinical notes to identify SH events that did not result in an interaction with a healthcare provider. Although we did identify some additional cases of SH using these clinical notes, their use did not, however, dramatically improve our ability to identify additional SH events nor did those additional events dramatically change the incidence or total SH event rates. The marginal number of additional SH events identified using clinical notes may imply that such data, or our www.cmrojournal.com ! 2014 Informa UK Ltd
Curr Med Res Opin Downloaded from informahealthcare.com by University of Washington on 12/02/14 For personal use only.
Current Medical Research & Opinion
methods, were not sensitive enough to detect SH events that did not result in contact with the healthcare system or that SH events almost always result in contact with the healthcare system (as EHR systems become more available, as they become used more often, and as the methods for extracting information from those systems become more sophisticated, our ability to identify SH, and other, events that did not result in contact with the healthcare system may increase). Our ability to identify SH events and healthcare costs was further limited by the extent to which patients received all of their care from providers and facilities that provided data to the Humedica database. Furthermore, costs are also likely to be underestimated as a result of excluding pharmacy costs, as well as encounters and service that we could not link to external cost data. However because costs are not likely to be differentially underestimated by SH status, except for pharmacy costs, which would be expected to be higher for patients who have experienced SH, we have most likely underestimated, rather than overestimated, the impact of SH events on quarterly costs. Finally, although we divided the post-titration follow-up period into quarters (3 month blocks) to facilitate the repeated measures analysis and because quarters represent meaningful financial units, we recognize that other meaningful time units could have been used and could have resulted in different conclusions. Our result that SH is associated with higher costs for care delivered temporally close to the SH events themselves (i.e., in the same quarter in our analyses) are important because it implies that patients who experience SH events are not always more expensive than patients who do not experience SH — they are just more expensive when they actually experience those events. These findings can also inform economic evaluations of competing therapies with different SH profiles and further imply that interventions that can prevent or delay SH can be associated with significant economic benefits even within rather short time horizons. Further real-world studies on the frequency and costs of SH, using methods to identify as many SH events as possible, can allow healthcare providers to make more informed decisions on the risks and benefits of basal insulin therapy in T2D patients.
Conclusions These results contribute to the literature not only by providing additional evidence that the real-world burden of SH is high among people with T2D who start using basal insulin but also by presenting new information on how history of previous SH events, both before starting insulin and during the insulin titration period, impacts SH incidence and total SH event rates for patients who have initiated basal insulin. Prior events, whether or not while on insulin treatment, are predictive of SH responses to ! 2014 Informa UK Ltd www.cmrojournal.com
Volume 30, Number 10
October 2014
insulin; in particular, it appears that SH during the period prior to insulin use is less important for predicting SH response to insulin than subsequent reactions during either the introduction of insulin (titration) or while on maintenance insulin treatment. These results should be of interest to clinicians and to those who evaluate the clinical and cost effectiveness of current and future treatments for diabetes.
Transparency Declaration of funding This research was supported by Novo Nordisk Inc. Declaration of financial/other relationships M.L.G. and Q.L. have disclosed that they are employees of Evidera, which received research funds from Novo Nordisk Inc. to conduct this study and were so at the time this research was conducted. Y.-Q.L. and E.G. have disclosed that they were also employees of Evidera. N.S.W. and J.C.H. have disclosed that they are employees of Novo Nordisk Inc. and are also stockholders of Novo Nordisk Inc. CMRO peer reviewers on this manuscript have no relevant financial or other relationships to disclose.
References 1. Diabetes Fact Sheet No. 312. World Health Organization, 2013. Available at: http://www.who.int/mediacentre/factsheets/fs312/en/ [Last accessed 18 July 2013] 2. Shaw JE, Sicree RA, Zimmet PZ. Global estimates of the prevalence of diabetes for 2010 and 2030. Diabetes Res Clin Pract 2010;87:4-14 3. Centers for Disease Control and Prevention. 2011 National Diabetes Fact Sheet: national estimates and general information on diabetes and prediabetes in the United States. 2011. Available at: http://www.cdc.gov/diabetes/ pubs/factsheet11.htm [Last accessed 8 May 2013] 4. The Diabetes Control and Complications Trial Research Group. The effect of intensive treatment of diabetes on the development and progression of longterm complications in insulin-dependent diabetes mellitus. N Engl J Med 1993;329:977-86 5. Lin YY, Hsu CW, Sheu WH, et al. Risk factors for recurrent hypoglycemia in hospitalized diabetic patients admitted for severe hypoglycemia. Yonsei Med J 2010;51:367-74 6. Baser O, Wei W, Baser E, Xie L. Clinical and economic outcomes in patients with type 2 diabetes initiating insulin glargine disposable pen versus exenatide BID. J Med Econ 2011;14:673-80 7. Zhang Q, Menditto L. Incremental cost savings 6 months following initiation of insulin glargine in a Medicaid fee-for-service sample. Am J Ther 2005;12:337-43 8. Bullano MF, Fisher MD, Grochulski WD, et al. Hypoglycemic events and glycosylated hemoglobin values in patients with type 2 diabetes mellitus newly initiated on insulin glargine or premixed insulin combination products. Am J Health Syst Pharm 2006;63:2473-82 9. Fabunmi R, Nielsen LL, Quimbo R, et al. Patient characteristics, drug adherence patterns, and hypoglycemia costs for patients with type 2 diabetes mellitus newly initiated on exenatide or insulin glargine. Curr Med Res Opin 2009;25:777-86 10. Brixner DI, McAdam-Marx C, Ye X, et al. Six-month outcomes on A1C and cardiovascular risk factors in patients with type 2 diabetes treated with exenatide in an ambulatory care setting. Diabetes Obes Metab 2009;11:1122-30
Severe hypoglycemia among patients starting basal insulin Ganz et al.
1999
Current Medical Research & Opinion Volume 30, Number 10
October 2014
16. Bullano MF, Al-Zakwani IS, Fisher MD, et al. Differences in hypoglycemia event rates and associated cost–consequence in patients initiated on longacting and intermediate-acting insulin products. Curr Med Res Opin 2005;21:291-8 17. Davis SN, Wei W, Garg S. Clinical impact of initiating insulin glargine therapy with disposable pen versus vial in patients with type 2 diabetes mellitus in a managed care setting. Endocr Pract 2011;17:845-52 18. Pawaskar M, Zagar A, Sugihara T, Shi L. Healthcare resource utilization and costs assessment of type 2 diabetes patients initiating exenatide BID or glargine: a retrospective database analysis. J Med Econ 2011;14: 16-27 19. Xie L, Wei W, Pan C, et al. A real-world study of patients with type 2 diabetes initiating basal insulins via disposable pens. Adv Ther 2011;28:1000-11
Curr Med Res Opin Downloaded from informahealthcare.com by University of Washington on 12/02/14 For personal use only.
11. Pawaskar M, Li Q, Hoogwerf BJ, et al. Metabolic outcomes of matched patient populations initiating exenatide bid versus insulin glargine in an ambulatory care setting. Diabetes Obes Metab 2012;14:626-33 12. Ginde AA, Blanc PG, Lieberman RM, et al. Validation of ICD-9-CM coding algorithm for improved identification of hypoglycemia visits. BMC Endocr Disord 2008;8:4 13. Overview of the Nationwide Inpatient Sample (NIS). Healthcare Costs and Utilization Project (HCUP). 2013. Available at: http://www.hcup-us.ahrq.gov/nisoverview.jsp [Last accessed 30 June 2014] 14. Graubard BI, Korn EL. Predictive margins with survey data. Biometrics 1999;55:652-9 15. Kleinman LC, Norton EC. What’s the risk? A simple approach for estimating adjusted risk measures from nonlinear models including logistic regression. Health Serv Res 2009;44:288-302
2000
Severe hypoglycemia among patients starting basal insulin Ganz et al.
www.cmrojournal.com ! 2014 Informa UK Ltd
