Diabetes Ther (2018) 9:1125–1141 https://doi.org/10.1007/s13300-018-0407-3 ORIGINAL RESEARCH Characteristics, Treatment Patterns, and Economic Outcomes of Patients Initiating Injectable Medications for Management of Type 2 Diabetes Mellitus in Japan: Results from a Retrospective Claims Database Analysis . . . . Shuichi Suzuki Urvi Desai Alena Strizek Jasmina Ivanova . . . Viviana Garcia-Horton Zhihong Cai Luke Schmerold Xinyue Liu Magaly Perez-Nieves Received: January 11, 2018 / Published online: April 16, 2018 The Author(s) 2018 provider, using the proxy setting of index ABSTRACT medication [hospital (including outpatient departments) for specialists; clinic for general Introduction: This study’s objective was to practitioner (GP)]. describe characteristics, treatment patterns, and Results: Of the 2683 patients included (mean economic outcomes of type 2 diabetes mellitus age: 50 years, 67% male), 1879 (70%) initiated (T2DM) patients initiating injectable antidia- injectable antidiabetics with specialists and 804 betic medications in Japan. (30%) with GPs. The specialist cohort had a Methods: Adults (C 18 years) with T2DM, C 2 signiﬁcantly greater comorbidity burden, but claims for injectable antidiabetics between 1 lower HbA1c levels during baseline, and was August 2011 and 31 July 2015 (ﬁrst claim = in- more likely to receive intensiﬁed treatment at dex date), no evidence of type 1 diabetes mel- index than the GP cohort. Almost 40% of litus, B 1 claim for insulin, no claims for GLP- patients (almost 30% of GP cohort) did not use 1RA before index, and continuous enrollment antidiabetics during baseline; the remaining for 6 months before (baseline) and 12 months patients received oral medications, primarily after index (follow-up) were selected from the from GPs. During follow-up, patients used the Japan Medical Center Database. Patient charac- index medication for approximately 7 months. teristics and outcomes during the baseline and Independent of specialist vs. GP setting, follow-up periods were described overall and by patients received antidiabetics and medications for T2DM-related comorbidities and complica- Enhanced content To view enhanced content for this tions during the baseline and follow-up periods article go to https://doi.org/10.6084/m9.ﬁgshare. from the same provider, primarily GPs. The overall average healthcare costs were ¥350,404 S. Suzuki (&) Z. Cai during baseline and ¥1,856,727 during follow- Eli Lilly Japan, K.K., Kobe, Japan up. e-mail: Suzuki_shuichi@lilly.com Conclusions: In Japan, most T2DM patients initiated injectable antidiabetics with specialists U. Desai Analysis Group, Inc., Boston, MA, USA vs. GPs. There were considerable differences in characteristics of patients treated by specialists A. Strizek M. Perez-Nieves vs. GPs. After initiation, injectable antidiabetics Eli Lilly and Company, Indianapolis, IN, USA were largely prescribed by GPs. Future research J. Ivanova V. Garcia-Horton L. Schmerold should evaluate the factors associated with dif- X. Liu ferent provider practices and communication Analysis Group, Inc., New York, NY, USA 1126 Diabetes Ther (2018) 9:1125–1141 channels between specialists and GPs to initiating injectable antidiabetic medications in improve patient management. Japan. In addition, a better understanding of Funding: Eli Lilly and Co. the roles and relationships between the differ- ent types of Japanese health professionals treating T2DM patients is required to inform Keywords: Costs; Injectable antidiabetic; future policy decisions, particularly given the Resource use; Specialist vs. GP setting; recent government initiatives to limit specialist Treatment patterns; Type 2 diabetes care to supporting and advising of non-special- ists when necessary . To the best of our knowledge, however, no INTRODUCTION study to date has provided a comprehensive assessment of the treatment patterns and clini- Type 2 diabetes mellitus (T2DM), the most common form of diabetes, affected about 7.6% cal as well as economic outcomes among T2DM patients initiating injectable antidiabetic medi- of adults aged 20–79 years in Japan in 2015  and accounted for approximately 4% of the cations—overall and by provider type—in Japan. For example, using a hospital-based national healthcare expenditure in 2014 . database, Ikeda et al.  found varied treatment The healthcare costs are particularly high among patients who develop diabetes-related and healthcare service utilization patterns among Japanese patients with T2DM who ini- complications, as demonstrated by a Japanese claims data study in 2016 . As such, it is tiated insulin therapy (long-acting, pre-mixed, rapid-acting, or a combination). However, only important to effectively manage patients with T2DM in clinical practice. Japanese guidelines medium- and large-sized administrative hospi- tals were included, and outpatient clinics were for the management of T2DM recommend beginning with lifestyle modiﬁcations such as not considered. In a different study, Had- jiyianni et al.  evaluated medication use and diet and exercise therapy, followed by treat- ment with oral antidiabetic agents and/or economic outcomes among patients initiating injectable therapy if the glycemic control target basal insulin in Japan and found that continu- ous use of basal insulin during the year after is not achieved . Injectable therapies may include the start of an insulin therapy (often initiation was associated with lower medical resources and costs. However, the study was basal insulin as the ﬁrst step of insulin therapy) or glucagon-like peptide-1 (GLP-1) receptor limited to basal insulin initiators and did not consider other injectable therapies. Addition- agonists . The Japanese treatment guidelines recommend a stepwise escalation of treatment ally, neither study evaluated practice patterns and outcomes by type of provider prescribing (in terms of either dose or use of more than one therapeutic agent) along the course and pro- insulin treatment. A few studies that did assess the variation between prescription patterns of gression of the disease. However, unlike other general physicians (GPs) and specialists treating treatment guidelines such as from the American T2DM patients in Japan did not evaluate the Diabetes Association (ADA) and the European treatment transitions leading up to and out- Association for the Study of Diabetes (EASD), which recommend use of injectable medica- comes following insulin treatment. For exam- ple, in the insurance-based database, Hidaka tions among patients with a more advanced disease stage , the recommendations regard- et al.  found that more patients with T2DM were treated by GPs than by specialists, and ing the use of injectable medications and treatment pathway in Japan are not clear. With prescription patterns between them were dif- ferent. Speciﬁcally, the specialists were more the increasing prevalence of T2DM , as well as the lack of speciﬁc recommendations likely to prescribe insulin (19% vs. 6%) and use intensiﬁed treatment approaches (e.g., triple regarding the use of injectable medications in combination therapy of biguanides ? DPP-4 Japan, there is a need for characterizing the treatment pathways followed in practice and inhibitors ? sulfonylureas) compared with GPs. Another study using administrative claims data subsequent outcomes in T2DM patients Diabetes Ther (2018) 9:1125–1141 1127 from the Japan Medical Data Center also information on enrollment history, patient reported that patients treated in academic hos- demographics, medical and prescription drug pitals were more likely to receive aggressive claims, and costs. Laboratory values from treatment, including higher rates of insulin annual health check-ups for people therapy, compared with those treated in clinics aged C 40 years, as well as the size of the med- . A different study surveyed more than ical facilities, are also available. 15,000 patients from clinics and hospitals in The study sample consisted of beneﬁciaries Japan and found that patients treated by dia- with T2DM who initiated injectable antidia- betes specialists were more likely to initiate betic medications (basal insulin, rapid insulin, insulin than those treated by general practi- basal-bolus therapy, pre-mixed insulin, or GLP- tioners . 1 receptor agonists) between 1 August 2011 and To address this gap in the literature, using 31 July 2015. The date of the ﬁrst observed de-identiﬁed retrospective data for commer- claim during this time period was deﬁned as the cially insured patients in Japan, the objectives study index date, the 6-month period prior to of this study were to: (1) understand the index date as the baseline period, and the demographic and clinical characteristics of 12-month period following the index date as T2DM patients initiating injectable medica- the follow-up period. Beneﬁciaries were identi- tions, (2) describe antidiabetic treatment pat- ﬁed as having T2DM if they met any of the terns, clinical outcomes, and all-cause and following conditions during the 2 years before T2DM-related medical resource use and costs and 1 year after the index date: (1) at least two during the 6 months before and 12 months diagnoses for T2DM [International Classiﬁca- after initiation of injectable medications, and tion of Diseases (ICD) 10 code: E11.x or E14.x] (3) provide a better understanding of the types or (2) at least one diagnosis of T2DM and at of providers prescribing the oral antidiabetic least one prescription for a non-insulin antidi- medications and medications for T2DM-related abetic medication . Further, beneﬁciaries comorbidities prior to and following initiation were required to have at least two additional of injectable antidiabetic medications. The claims for the index medication after the index outcomes were reported for the overall cohort date, be at least 18 years old, and have contin- as well as for cohorts stratiﬁed by type of med- uous enrollment with JMDC throughout the ical provider, specialist, or GP, using as a proxy baseline and follow-up periods (to ensure the setting in which the index medication was availability of complete pharmacy and medical prescribed [hospital (including outpatient care information). Beneﬁciaries with evidence departments) for specialist or clinic for GP]. of type 1 diabetes mellitus (T1DM; ICD-10: E10.x) before the index date or during the fol- low-up period not accompanied by oral antidi- METHODS abetic use were excluded. Additionally, to increase the likelihood of capturing the ‘true’ Data and Sample Selection initiators of injectable medications (as opposed to those using these medications for acute rea- This study was conducted using data from the sons), beneﬁciaries with more than one claim Japan Medical Data Center (JMDC) [13, 14]. The for insulin during the baseline period or with JMDC database contains de-identiﬁed adminis- any claim for an antidiabetic injectable medica- trative claims data for beneﬁciaries of the Kenpo tion other than insulin prior to the index date health insurance system who are employed by were excluded. middle-to-large size companies in Japan and The data do not contain information about their dependents. The JMDC database currently provider specialty. However, this information includes over 4 million unique beneﬁciaries can be approximated based on the setting in under age 75 and contains administrative which care was provided. Speciﬁcally, in Japan, claims for services provided between 1 January primary care is typically provided in clinics 2005 and 31 March 2016. Data contain (deﬁned as a facility with less than 20 beds), 1128 Diabetes Ther (2018) 9:1125–1141 whereas specialist care is provided in hospital antidiabetic medications, proportions of (deﬁned as a facility with at least 20 beds) patients using monotherapy vs. combination [11, 15]. Consistent with these deﬁnitions, therapy were described. Combination therapy patients in the ﬁnal analytic sample were strat- was deﬁned as the concurrent use of at least two iﬁed into two cohorts depending on the setting medications, each from a different class of of the index medication: the specialist cohort, antidiabetics, such that the days’ supply for the consisting of patients whose index medication medications overlap by at least 28 days, and was prescribed in a hospital setting, and the GP there was a new prescription for each of the cohort, consisting of patients whose index overlapping antidiabetic therapies after the start medication was prescribed in a clinic. The hos- of concurrent use. Combination therapy cap- pital setting included in- and outpatient medi- tured the use of a combination of oral medica- cal claims as well as pharmacy claims associated tions, orals and injections, or injections and with a facility with at least 20 beds. thus potentially included the use of insulin. All This article is based on previously collected patients using at least one class of antidiabetic data and does not involve any new studies of medication and not meeting the criteria for human or animal subjects performed by any of combination therapy were considered as using the authors. monotherapy. Furthermore, among patients using both antidiabetic medications and medications for Patient Characteristics T2DM-related complications and comorbidities, the proportions of patients for whom their most Patient characteristics were evaluated over the recent medications were prescribed in a hospital 6-month baseline period prior to the index date or clinic setting were described. For the baseline or at index date. These included demographics period, the last medication prescribed for each (age and gender), type of index medication (i.e., patient before their index date was used; for the basal insulin only, rapid insulin only, basal and follow-up period, the last medication observed rapid insulin, pre-mixed insulin only, GLP-1 during the follow-up period for each patient receptor agonists only, and multiple), Charlson was used. In addition, the proportions of comorbidity index (CCI) , and presence of patients who had the same provider for the T2DM-related complications and comorbidities most recent use of any antidiabetic and T2DM- (e.g., retinopathy, neuropathy, and congestive related medication were reported for both time heart failure). periods. In addition, among patients with data from annual health check-ups prior to or at the index Healthcare Resource Use and Costs date, the test results for speciﬁc measures were Metrics of all-cause and T2DM-related health- reported. These included body mass index care resource use as well as associated costs were (BMI), HbA1c levels, fasting blood sugar levels, evaluated during the baseline and follow-up and blood pressure levels [including systolic periods. Speciﬁcally, proportions of patients blood pressure (SBP) and diastolic blood pres- with an in- or outpatient visit during the sure (DBP)]. respective period were reported. Additionally, among those with at least one inpatient visit Outcomes the mean duration of inpatient stay was repor- ted. Similarly, among those with an outpatient Medication Use visit, the mean number of outpatient visits Use of antidiabetic medications and other select during the relevant period were reported. Fur- medications for T2DM-related complications thermore, healthcare costs including total costs, and comorbidities (antihypertensives, statins, total medical costs, medical costs by hospital or antiplatelet agents, overall and by medica- department (inpatient and outpatient), and tion class) was evaluated during the baseline pharmacy costs during the respective period and follow-up periods. In addition, for were reported. The costs represent payments for Diabetes Ther (2018) 9:1125–1141 1129 Fig. 1 Sample selection and resulting patient counts. having T2DM based upon having at least two diagnoses GLP-1 RA glucagon-like peptide-1 receptor agonists, GP for T2DM (ICD-10 codes E11.x and E14.x) or having at general practitioner, T1DM type 1 diabetes mellitus, least one diagnosis for T2DM and at least one claim for an T2DM type 2 diabetes mellitus. Patients were identiﬁed as OAD; T1DM was identiﬁed using ICD-10 code E10.x medical services and prescription drugs. The were reported for age, all-cause healthcare costs associated with medical care received resource use, and all cost variables. include the insurance payment and patient Results were reported for the overall cohort copayment amount. Costs were inﬂated to ﬁrst as well as stratiﬁed by specialist vs. GP provider semi-annual 2016 Japanese yen (¥) using the type at the time of initiation. For the stratiﬁed medical care component of the Japanese Con- analyses, comparisons between the two cohorts sumer Price Index (CPI) . T2DM-related were conducted using Wilcoxon rank sum tests healthcare resource use and costs include claims for continuous variables and chi-squared tests with a diagnosis for T2DM, retinopathy, neu- for categorical variables. Statistical signiﬁcance ropathy/diabetic foot, nephropathy, cardiovas- was deﬁned as p \ 0.05. cular disease, congestive heart failure, peripheral vascular disease, stroke, hyperten- RESULTS sion, dyslipidemia, and severe hypoglycemic event. Patient Characteristics Statistical Analyses A total of 2683 patients were included in the ﬁnal analytic sample, with 1879 patients (70%) Patient characteristics, medication use, health- in the specialist cohort and 804 patients (30%) care resource use, and healthcare costs were in the GP cohort (Fig. 1). described using means and standard deviations For the overall population, the mean age at (SD) for continuous variables (e.g., age, CCI) index was 50 years and 67% were male and numbers and proportions for categorical (Table 1). At the last annual health check-up variables (e.g., gender, comorbidity rates). In prior to initiating injectable antidiabetic medi- addition, the median and interquartile range cation, among patients with available data (see 1130 Diabetes Ther (2018) 9:1125–1141 Table 1 Patient baseline characteristics Index medication setting Overall Specialist GP P value (n = 2683) (n = 1879) (n = 804) Demographics Age on the index date Mean (SD) 50.0 (10.4) 50.4 (10.8) 49.1 (9.6) 0.0012* Gender (male), n (%) 1799 (67.1%) 1227 (65.3%) 572 (71.1%) 0.0032* Index medication category, n (%) \0.0001* Basal insulin only (not mixed) 774 (28.8%) 349 (18.6%) 425 (52.9%) Rapid insulin only 779 (29.0%) 701 (37.3%) 78 (9.7%) Basal and rapid insulin 549 (20.5%) 488 (26.0%) 61 (7.6%) Pre-mixed insulin only 192 (7.2%) 108 (5.7%) 84 (10.4%) GLP-1 receptor agonists only 351 (13.1%) 202 (10.8%) 149 (18.5%) Multiple index medications (excluding basal and 38 (1.4%) 31 (1.6%) 7 (0.9%) rapid insulin) Index medication route of administration, n (%) \0.0001* Pen only 2126 (79.2%) 1330 (70.8%) 796 (99.0%) Vial only 322 (12.0%) 315 (16.8%) 7 (0.9%) Both pen and vial 235 (8.8%) 234 (12.5%) 1 (0.1%) CCI, mean (SD) 1.6 (2.2) 1.9 (2.4) 1.0 (1.4) \0.0001* T2DM-related complications and comorbidities, n (%) Microvascular conditions Retinopathy 17 (0.6%) 13 (0.7%) 4 (0.5%) 0.5611 Neuropathy/diabetic foot 8 (0.3%) 8 (0.4%) 0 (0.0%) 0.1150 Nephropathy 158 (5.9%) 126 (6.7%) 32 (4.0%) 0.0060* Macrovascular conditions Cardiovascular disease 385 (14.3%) 310 (16.5%) 75 (9.3%) \0.0001* Congestive heart failure 261 (9.7%) 226 (12.0%) 35 (4.4%) \0.0001* Peripheral vascular disease 282 (10.5%) 212 (11.3%) 70 (8.7%) 0.0462* Stroke 201 (7.5%) 155 (8.2%) 46 (5.7%) 0.0227* T2DM-related comorbidities Hypertension 1092 (40.7%) 763 (40.6%) 329 (40.9%) 0.8796 Dyslipidemia 1274 (47.5%) 845 (45.0%) 429 (53.4%) \0.0001* Severe hypoglycemic event 55 (2.0%) 47 (2.5%) 8 (1.0%) 0.0117* Diabetes Ther (2018) 9:1125–1141 1131 Table 1 continued Index medication setting Overall Specialist GP P value (n = 2683) (n = 1879) (n = 804) HbA1c levels (%), n (%) 1273 (47.4%) 840 (44.7%) 433 (53.9%) Mean (SD) 8.9 (2.3) 8.8 (2.3) 9.1 (2.2) 0.0165* Systolic blood pressure (mmHg), n (%) 1446 (53.9%) 960 (51.1%) 486 (60.4%) Mean (SD) 130 (18.0) 131 (18.5) 127 (16.8) 0.0003* Diastolic blood pressure (mmHg), n (%) 1446 (53.9%) 960 (51.1%) 486 (60.4%) Mean (SD) 80 (11.7) 81 (12.0) 79 (10.9) 0.0016* Fasting blood sugar levels (mg/dl), n (%) 1158 (43.2%) 766 (40.8%) 392 (48.8%) Mean (SD) 187 (72.3) 183 (70.9) 194 (74.4) 0.0052* BMI (kg/m ), n (%) 1447 (53.9%) 961 (51.1%) 486 (60.4%) Mean (SD) 26.7 (5.6) 27 (5.8) 27 (5.3) 0.6930 SD standard deviation, GLP-1 glucagon-like peptide-1, CCI Charlson comorbidity index, T2DM type 2 diabetes mellitus, HbA1c hemoglobin A1c, BMI body mass index *Statistically signiﬁcant at p \ 0.05; p values estimated using chi-squared tests for categorical variables and Wilcoxon rank- sum tests for continuous variables Table 1), the mean HbA1c level was 8.9%, the p = 0.0052) as well as higher systolic blood mean BMI was 27 kg/m , and the mean fasting pressure (131 vs. 127 mmHg; p = 0.0003) and blood sugar level was 187 mg/dl. In terms of the diastolic blood pressure (81 vs. 79 mmHg; type of index medications, 29% of patients used p = 0.0016), compared with the GP cohort basal insulin only, 29% rapid insulin only, and (Table 1). Patients in the specialist cohort were 21% basal-bolus treatment (both basal and more likely to receive rapid insulin (37%) or rapid insulin). Thirteen percent of patients used basal-bolus treatment (26%) compared with only GLP-1 receptor agonists as their index basal insulin only (19%), whereas patients in medication. the GP cohort were more likely to receive basal There were several differences in patient insulin only (53%) or GLP-1 receptor agonists characteristics between the specialist and GP only (19%). cohorts (Table 1). On average, patients in the specialist cohort were slightly older (50 vs. Medication Use 49 years, p = 0.0012) and had a higher comor- bidity burden compared with patients in the GP Baseline Period cohort as measured by the CCI, speciﬁc During the 6-month baseline period, 39% of the microvascular and macrovascular conditions, overall cohort did not use any antidiabetic and other T2DM-related comorbidities. How- medication, 52% used monotherapy, and 9% ever, at the last annual health check-up prior to used combination therapy (Fig. 2a). Nearly half injectable initiation, among patients with (54%) of the patients received their most recent available data, those in the specialist cohort had prescription for a non-injectable antidiabetic slightly, yet statistically signiﬁcantly lower from a GP; only 6% received it from specialists baseline HbA1c (8.8% vs. 9.1%, p = 0.0165) and (Table 2). The most commonly used classes of fasting blood glucose (183 vs. 194 mg/dl, oral antidiabetic medications were dipeptidyl- 1132 Diabetes Ther (2018) 9:1125–1141 during the baseline period from a specialist. Furthermore, patients in the specialist cohort were less likely to have used statins during the baseline (26% vs. 33%, p \ 0.0001), but had similar rates of use of antihypertensives and antiplatelet agents as the patients from the GP cohort (Table 2). In addition, for patients in the specialist cohort, the most recent medications for T2DM-related comorbidities prior to injectable initiation were generally prescribed by the same provider as those prescribing oral antidiabetic agents (i.e., GPs), suggesting that multiple physicians may be involved in the treatment decisions at initiation of injectable medications (Table 2). Follow-up Period On average, patients in the overall cohort used their index medication for 214 days (as approximated by the medication’s days of sup- ply) during the 12-month follow-up period and just over half used a combination therapy of antidiabetics (including their index injectable antidiabetic) (Fig. 2b). Nearly 70% of Fig. 2 T2DM-related medication use by cohort. T2DM patients received their most recent prescription type 2 diabetes mellitus, GP general practitioner. *Signif- for an oral antidiabetic from a GP, although icant difference between the specialist and GP cohorts, as 70% initiated the ﬁrst injectable therapy with a determined by p \ 0.05. a Baseline period medication use. specialist. As in the baseline period, the most b Follow-up period medication use frequently used oral agents during the follow- up period were DPP-4 inhibitors (55%), bigua- nides (49%), and sulfonylureas (41%). Approx- peptidase-4 (DPP-4) inhibitors (46%), sulfony- imately 66% of patients had medication for lureas (42%), and biguanides (36%). Approxi- T2DM-related comorbidities. Most patients mately half of the overall cohort also used (94%) had their most recent antidiabetic medi- medications for select T2DM-related complica- cation and medication for T2DM-related tions and comorbidities. These T2DM-related comorbidities prescribed by the same provider medications were most commonly (94%) pre- (Table 3). scribed by the same provider as that for antidi- Patients in the specialist cohort were more abetic medications (primarily GPs; Table 2). likely to have monotherapy during the When evaluating the medication use by 12-month follow-up period compared with index medication setting, patients in the spe- patients in the GP cohort (49% vs. 39%, cialist cohort were less likely to have oral p \ 0.0001) (Fig. 2b). Additionally, patients in antidiabetic use during the 6-month baseline the specialist cohort were signiﬁcantly less period compared with patients from the GP likely to use certain types of oral antidiabetic cohort (57% vs. 70%, p \ 0.0001) and were medications, such as sulfonylureas, biguanides, generally less likely to have used the various antidiabetic combinations, and SGLT2 inhibi- classes of antidiabetic medications (Table 2). tors, than patients from the GP cohort (Table 3). Additionally, only 9% of patients who initiated Furthermore, while on average patients used injectable medication with a specialist had their index injectable medications for nearly received their most recent oral medication 7 months, the mean duration of use was shorter Diabetes Ther (2018) 9:1125–1141 1133 Table 2 Baseline period medication use Index medication setting Overall Specialist GP P value (n = 2683) (n = 1879) (n = 804) T2DM-related drug use, n (%) Use of antidiabetic medications, overall and by class Oral antidiabetic medications Any oral antidiabetic medications 1626 (60.6%) 1064 (56.6%) 562 (69.9%) \0.0001* DPP-4 inhibitors 1228 (45.8%) 787 (41.9%) 441 (54.9%) \0.0001* Sulfonylureas 1137 (42.4%) 712 (37.9%) 425 (52.9%) \0.0001* Biguanides 968 (36.1%) 608 (32.4%) 360 (44.8%) \0.0001* Alpha-glucosidase inhibitors 516 (19.2%) 334 (17.8%) 182 (22.6%) 0.0034* Thiazolidinediones 412 (15.4%) 262 (13.9%) 150 (18.7%) 0.0019* Meglitinides 114 (4.2%) 84 (4.5%) 30 (3.7%) 0.3846 Antidiabetic combinations 98 (3.7%) 50 (2.7%) 48 (6.0%) \0.0001* SGLT2 inhibitors 10 (0.4%) 7 (0.4%) 3 (0.4%) 1.0000 Setting for most recent use of antidiabetic medications Type of provider prescribing oral antidiabetic medication Hospital only 169 (6.3%) 166 (8.8%) 3 (0.4%) \0.0001* Clinic only 1442 (53.7%) 883 (47.0%) 559 (69.5%) \0.0001* Multiple 5 (0.6%) 15 (0.8%) 0 (0.0%) 0.0082* Drug use related to other select complications and comorbid conditions, n (%) Any drug use 1304 (48.6%) 881 (46.9%) 423 (52.6%) 0.0066* Antihypertensives 988 (36.8%) 691 (36.8%) 297 (36.9%) 0.9351 Antiplatelet agents 290 (10.8%) 213 (11.3%) 77 (9.6%) 0.1789 Statins 747 (27.8%) 480 (25.5%) 267 (33.2%) \0.0001* Patients with the same provider for medications , n (%) Patients with medications for T2DM-related comorbidities 1111 (41.4%) 729 (38.8%) 382 (47.5%) \0.0001* and oral antidiabetics Patients with the same provider 1042 (93.8%) 677 (92.9%) 365 (95.5%) \0.0001* DPP-4 dipeptidyl peptidase-4, SGLT2 sodium-glucose co-transporter 2, T2DM type 2 diabetes mellitus *Statistically signiﬁcant at p \ 0.05; p values estimated using chi-squared tests for categorical variables and Wilcoxon rank- sum tests for continuous variables The most recent medication prescriptions observed during the baseline period were used 1134 Diabetes Ther (2018) 9:1125–1141 Table 3 Follow-up period medication use Index medication setting Overall Specialist GP P value (N = 2683) (N = 1879) (N = 804) Days of index medication use during the follow-up period, 214 (104) 195 (104) 260 (89) \0.0001* mean (SD) T2DM-related drug use, n (%) Use of antidiabetic medications, overall and by class Oral antidiabetic medications Any oral antidiabetic medications 2197 (81.9%) 1504 (80.0%) 693 (86.2%) 0.0002* DPP-4 inhibitors 1466 (54.6%) 1008 (53.6%) 458 (57.0%) 0.1136 Sulfonylureas 1093 (40.7%) 689 (36.7%) 404 (50.2%) \0.0001* Biguanides 1321 (49.2%) 895 (47.6%) 426 (53.0%) 0.0111* Alpha-glucosidase inhibitors 673 (25.1%) 455 (24.2%) 218 (27.1%) 0.1125 Thiazolidinediones 377 (14.1%) 252 (13.4%) 125 (15.5%) 0.1447 Meglitinides 278 (10.4%) 210 (11.2%) 68 (8.5%) 0.0343* Antidiabetic combinations 95 (3.5%) 43 (2.3%) 52 (6.5%) \0.0001* SGLT2 inhibitors 43 (1.6%) 23 (1.2%) 20 (2.5%) 0.0170* Any non-index injectable antidiabetic medications Basal insulin (not mixed) 366 (13.6%) 299 (15.9%) 67 (8.3%) \0.0001* Rapid insulin 403 (15.0%) 324 (17.2%) 79 (9.8%) \0.0001* Pre-mixed insulin 119 (4.4%) 95 (5.1%) 24 (3.0%) 0.0170* GLP-1 receptor agonists 95 (3.5%) 74 (3.9%) 21 (2.6%) 0.0886 Setting for most recent use of antidiabetic medications Type of provider prescribing injectable antidiabetic medication Hospital only 667 (24.9%) 653 (34.8%) 14 (1.7%) \0.0001* Clinic only 1981 (73.8%) 1194 (63.5%) 787 (97.9%) \0.0001* Multiple 35 (1.3%) 32 (1.7%) 3 (0.4%) 0.0054* Type of provider prescribing oral antidiabetic medication Hospital only 359 (13.4%) 355 (18.9%) 4 (0.5%) \0.0001* Clinic only 1812 (67.5%) 1125 (59.9%) 687 (85.4%) \0.0001* Multiple 26 (1.0%) 24 (1.3%) 2 (0.2%) 0.0127* Drug use related to other select complications and comorbid conditions, n (%) Any drug use 1766 (65.8%) 1256 (66.8%) 510 (63.4%) 0.0879 Antihypertensives 1375 (51.2%) 1009 (53.7%) 366 (45.5%) 0.0001* Diabetes Ther (2018) 9:1125–1141 1135 Table 3 continued Index medication setting Overall Specialist GP P value (N = 2683) (N = 1879) (N = 804) Antiplatelet agents 423 (15.8%) 334 (17.8%) 89 (11.1%) \0.000* Statins 1011 (37.7%) 700 (37.3%) 311 (38.7%) 0.4845 Patients with the same provider for medications , n (%) Patients with medications for T2DM-related comorbidities 1725 (64.3%) 1221 (65.0%) 504 (62.7%) 0.2557 and any antidiabetics Patients with the same provider 1621 (94.0%) 1140 (93.4%) 481 (95.4%) 0.6819 Patients with medications for T2DM-related comorbidities 1698 (63.3%) 1197 (63.7%) 501 (62.3%) 0.4936 and injectable antidiabetics Patients with the same provider 1499 (88.3%) 1028 (85.9%) 471 (94.0%) 0.0642 Patients with medications for T2DM-related comorbidities 1438 (53.6%) 995 (53.0%) 443 (55.1%) 0.3073 and oral antidiabetics Patients with the same provider 1344 (93.5%) 919 (92.4%) 425 (95.9%) 0.0607 DPP-4 dipeptidyl peptidase-4, SGLT2 sodium-glucose co-transporter 2, GLP-1 glucagon-like peptide-1, T2DM type 2 diabetes mellitus *Statistically signiﬁcant at p \ 0.05; p values estimated using chi-squared tests for categorical variables and Wilcoxon rank- sum tests for continuous variables The most recent medication prescriptions observed during the follow-up period were used for the specialist cohort than for the GP cohort days for patients with at least one visit was also (195 vs. 260 days). Despite initiating an higher (23 vs. 7, p = 0.0023). Nearly 90% of all injectable antidiabetic medication with a spe- patients had at least one outpatient visit during cialist, 64% of patients in the specialist cohort the 6-month baseline period; the mean number received their most recent injectable antidia- of outpatient visits was 7. betic medication from a GP. Additionally, most Overall, the average healthcare costs were patients in the specialist cohort received their ¥350,404 during the baseline period, 75% of most recent antidiabetic medications and other which were attributable to T2DM (Fig. 3b). The T2DM-related medications from the same pro- average total and medical costs were signiﬁ- vider (Table 3). cantly greater for patients in the specialist vs. the GP cohort (total costs: ¥418,894 vs. ¥190,338; medical costs: ¥352,022 vs. ¥121,545; Healthcare Resource Use and Costs all p \ 0.05); this difference was mainly driven by some extreme medical costs observed for the Baseline Period specialist cohort. However, the difference in Overall, 12% of patients had an inpatient visit average pharmacy costs between the two during the 6-month baseline period. A signiﬁ- cohorts was small (¥66,873 vs. ¥68,792, cantly higher proportion of the specialist cohort p = 0.0007). had an inpatient visit than the GP cohort (15% Most of the observed resource use and costs vs. 3%, p \ 0.0001) during the baseline period were attributable to T2DM (Fig. 3a, c). (Fig. 3a), and the average number of inpatient 1136 Diabetes Ther (2018) 9:1125–1141 Fig. 3 Baseline period healthcare resource use and costs by and GP cohorts, as determined by p \ 0.05. a Healthcare cohort. T2DM type 2 diabetes mellitus, GP general resource use. b All-cause healthcare costs. c T2DM-related practitioner. *Signiﬁcant difference between the specialist healthcare costs Follow-up Period The average total and medical costs during During the 12-month follow-up period, 55% of the 12-month follow-up period were signiﬁ- the overall cohort had at least one inpatient cantly higher for patients in the specialist visit, and almost all patients had at least one cohort vs. GP provider cohort (total costs: outpatient visit (Fig. 4a). The average number of ¥2,344,988 vs. ¥715,628; medical costs: outpatient visits in the overall cohort for ¥2,117,343 vs. ¥503,204; all p \ 0.0001; phar- patients with at least one visit was 16. The macy costs: ¥227,645 vs. ¥212,424, p = 0.0131) average healthcare costs were ¥1,856,727 during (Fig. 4b). the follow-up period (Fig. 4b). Most of the resource use and costs during the The stratiﬁed analyses showed that a signiﬁ- follow-up period were attributable to T2DM for cantly greater proportion of patients treated by all cohorts (Fig. 4a, c). specialists had an inpatient visit (73% vs. 12%, p \ 0.0001) during the follow-up period com- DISCUSSION pared with the GP cohort (Fig. 4a), and this difference remained signiﬁcant when the index In this claims database study of T2DM patients month was excluded (51% vs. 12%, p \ 0.0001). initiating injectable antidiabetic medications in The average number of inpatient days for Japan, most (70%) of the patients initiating an patients with at least one visit was also signiﬁ- injectable antidiabetic medication did so under cantly higher for patients in the specialist specialist care as opposed to GP. Independent of cohort (41 vs. 16 days, p \ 0.0001). Diabetes Ther (2018) 9:1125–1141 1137 Fig. 4 Follow-up period healthcare resource use and costs and GP cohorts, as determined by p \ 0.05. a Healthcare by cohort. T2DM type 2 diabetes mellitus, GP general resource use. b All-cause healthcare costs. c T2DM-related practitioner. *Signiﬁcant difference between the specialist healthcare costs the provider type, a large proportion of patients burden prior to treatment initiation may have (almost 40%) did not use any antidiabetic triggered a more intensiﬁed treatment medication in the 6-month period prior to ini- approach, such as short-term intensive insulin tiating injectable medications, suggesting that therapy . Indeed, in the DAWN Japan study, in Japan, a substantial proportion of patients the physicians reported that they would con- initiate injectable medications as the ﬁrst-line sider initiating insulin at HbA1c levels C 8.7%, treatment. These ﬁndings are similar to those and patients reported that insulin was recom- reported by Kohro et al.  using the same mended when average HbA1c levels were database as the present study. Speciﬁcally, in C 9.6% . their study of ﬁrst-line antidiabetic medica- In the current study, the average HbA1c tions, the authors found that nearly 20% of all levels at the last annual health check-up before T2DM patients initiated insulin as the ﬁrst the start of injectable antidiabetic medication antidiabetic medication. The proportions were were slightly higher for those treated by the GPs even higher among those treated in academic (9.1% GP vs. 8.8% specialist), suggesting that hospitals vs. clinics (* 40% vs. * 10% in 2011) the GPs may use less intensive approaches to . Despite the relatively young age (50 years managing T2DM. In this study, patients treated on average), the population had high average by specialists were more likely to receive inten- HbA1c (8.9%) and fasting glucose levels as well siﬁed treatment regimens involving rapid insu- as high rates of T2DM-related comorbidities and lin only (37% vs. 10%) or basal-bolus treatment complications before initiating injectable medi- (26% vs. 8%) and less likely to receive basal cations; this suggests that the high disease insulin only (19% vs. 53%) compared with 1138 Diabetes Ther (2018) 9:1125–1141 those treated by GPs. Additionally, over half of also consistent with the previously mentioned the patients used a combination of antidiabetic study by Hadjiyianni et al. . In particular, medications, including additional oral antidia- 55% of all patients had at least one hospital- betic medications, during the follow-up period. ization in the year after injectable initiation; the These results are consistent with previous stud- average all-cause costs during the 12-month ies in Japan as well as other countries and follow-up period were ¥1856,727, approxi- highlight the considerable variation in physi- mately 90% of which were attributable to T2DM cian practice patterns. For example, Ikeda et al. and related complications and comorbidities.  studied Japanese patients with T2DM who The economic burden was even higher among initiated insulin therapy in a hospital setting those initiating the injectable medication with and also concluded that only 11% of the a specialist. Speciﬁcally, nearly three-quarters patients were treated with long-acting insulin (73%) of the patients in the specialist cohort and that treatment with rapid insulin was the had an inpatient visit during the 12-month most common therapy (47% of initiators). follow-up period compared with 12% in the GP Additionally, patients treated with more inten- cohort, and among those with an inpatient siﬁed treatments, including with rapid insulin, visit, the average number of inpatient days was also tended to be older, have lower HbA1c, and signiﬁcantly higher for patients in the specialist experience more comorbid conditions. A survey cohort. Consequently, patients in the specialist of patients treated by general practitioners and cohort had higher medical costs relative to diabetes specialists in Japan and a different those in the GP cohort. However, differences in study using the same database as the present pharmacy costs across the two cohorts were study also found a similar pattern [11, 12]. Such small. Taken together, our ﬁndings suggest that ﬁndings could be attributed to the fact that GPs the considerable increase in the proportion of in Japan often have limited resources to facili- patients with an inpatient visit during the fol- tate initiation of injectable antidiabetic regi- low-up period was the major driver of the cost mens. For example, in the aforementioned increase for patients in the specialist cohort. It is DAWN Japan study, nearly 55% of the non- possible that the higher rates of hospital visits specialist clinicians reported that ‘‘[they] do not are related to continued care management in a have staff (nurse, pharmacists) who can assist hospital setting because of the complexity of with explanations’’ and that ‘‘[it] is difﬁcult to the treatment and higher comorbidity burden provide guidance and education on insulin in the specialist cohort. However, our ﬁndings injection to patients’’ compared with 1% and indicate that independent of the index medi- 7% of the specialists reporting the same con- cation setting, after initiating the cerns . With the limited number of diabetes injectable medication, routine follow-up care specialists in Japan (4760 as of August 2013) may transition to a clinic setting. Speciﬁcally, , providing adequate education and support 64% of patients who initiated injectable medi- systems to the GPs may help improve the cations in a hospital setting received their most management of T2DM patients. recent injectable medication during the follow- Regarding the outcomes after injectable ini- up period in a clinic setting. Most patients also tiation, we ﬁnd that patients continued treat- received their oral medications as well as med- ment with the index medication for ications for T2DM-related complications and approximately 7 months, independent of the comorbidities from the same providers, sug- setting in which the treatment was initiated. gesting that use of hospitals and both specialist These ﬁndings are consistent with previous and GPs for maintenance treatment of T2DM research that demonstrated that the probability and related conditions is limited. Given this of discontinuing basal insulin treatment within observation, an alternative explanation for the the ﬁrst year after initiation in Japan was low higher observed costs could be related to dif- (22%) . Despite this, the patients in our ferences in practice patterns by provider spe- sample had considerable resource use and costs cialty and/or other, unobserved patient in the year after treatment initiation, a ﬁnding characteristics. While this was not studied Diabetes Ther (2018) 9:1125–1141 1139 directly in the present study, Chin et al.  glucose control). Relatedly, while the methods studied the differences in resource utilization used to identify patients with T2DM are con- among older patients with diabetes in the US sistent with prior studies using the same data- who were cared for by endocrinologists, inter- base, the effect of alternative criteria for nists, family practitioners, and general practi- identifying T2DM patients on the study ﬁnd- tioners and concluded that patients treated by ings is not known . In addition, HIV/AIDS specialists (endocrinologists and internists) had information was not available in JMDC, so the higher utilization of diabetes-speciﬁc process- CCI may not be precise. Similarly, the specialty of-care measures and received a more costly of the treating physician was not directly style of care than patients treated by family observed in the data, and the proxy used to practitioners and general practitioners. Addi- approximate this attribute may not accurately tional research is warranted to understand how represent care provided by specialists in hospi- hospital- and clinic-based physicians commu- tals with \ 20 beds or by GPs in large hospitals. nicate treatment strategies for T2DM and The JMDC data on laboratory test results were T2DM-related complications/comorbidities, limited to tests performed as a part of the reg- including who makes the decisions to augment ular annual check-ups for beneﬁciaries aged at or switch treatments as patients’ conditions least 40 years and did not include information progress over time and whether patients are from tests performed for other reasons (e.g., referred to specialists in a timely fashion for while in the hospital). Further, the results were initiating more complicated treatment limited to tests performed at special facilities regimens. performing health check-ups or hospital/clinics To the best of our knowledge, this is the ﬁrst recommended by the beneﬁciaries’ insurance study to provide a comprehensive assessment of provider for their routine annual check-up. the characteristics and outcomes among Furthermore, ﬁndings are limited to people patients with T2DM initiating their ﬁrst aged under 75 years who have employer-spon- injectable medication with or without prior oral sored insurance in Japan and may not general- antidiabetic treatment in Japan, both before ize to other populations (e.g., may not be and after treatment initiation, using real-world representative of patients covered by other health insurance data. Additionally, the study insurance programs) as the JMDC database is provides important insight into the character- not representative of the national population in istics and outcomes of patients treated by spe- Japan. cialists (i.e., in a hospital inpatient or outpatient setting) compared with those treated by general CONCLUSIONS practitioners in Japan, which could help inform future policy decisions regarding the coordina- The study ﬁndings indicate that in Japan tion of care between different types of provi- patients with T2DM are more likely to initiate ders. This study particularly beneﬁts from the medications with specialists as opposed to GPs. large sample size of the JMDC claims database as Patients initiating injectable medications with well as from the availability of laboratory mea- specialists are older and have greater disease sures from annual health check-ups. Another severity prior to treatment initiation, but are advantage of the JMDC database is that it con- less likely to be using oral antidiabetic medica- tains a classiﬁcation of the medical facilities tions in the 6 months prior to initiating associated with each claim, which aided in the injectable medications than patients initiating construction of the two study cohorts to injectable medications with GPs. Additionally, approximate provider specialty. However, this patients initiating injectable medications with analysis was subject to limitations associated specialists are more likely to receive intensive with any inaccuracies or incompleteness of the treatment with rapid insulin or basal-bolus ICD-10 codes and ATC codes used to identify therapy on their index date than those initiat- diagnoses and medication use as well as lack of ing treatment in a clinic setting and are more clinical information (e.g., to assess blood 1140 Diabetes Ther (2018) 9:1125–1141 likely to have inpatient visits and higher employee of Analysis Group, Inc. at the time of healthcare costs both before and after treatment the study. initiation. However, the majority of patients Compliance with Ethics Guidelines. This initiating injectable antidiabetic medications article is based on previously collected data and with specialists receive subsequent does not involve any new studies of human or injectable medication during the follow-up animal subjects performed by any of the period from the GPs, suggesting that GPs are authors. responsible for maintenance care as opposed to the prescribing specialists. Further research is Data Availability. The data used for the needed to understand the reasons behind the current study are not publicly available because prescribing patterns and subsequent outcomes they were provided by Japan Medical Data among T2DM patients in Japan, as well as to Center (JMDC) to Analysis Group, Inc., and the better understand the communication channels data license agreement does not permit sharing between GPs and specialists regarding patients’ of data sets with people external to the study treatments over time. team. Interested readers may request the data directly from JMDC. ACKNOWLEDGEMENTS Open Access. This article is distributed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International Funding. Research funding provided by Eli License (http://creativecommons.org/licenses/ Lilly and Co. to Analysis Group, Inc. Funding by-nc/4.0/), which permits any non- for publication charges was also provided by Eli commercial use, distribution, and reproduction Lilly and Co. in any medium, provided you give appropriate credit to the original author(s) and the source, Authorship. All authors made signiﬁcant provide a link to the Creative Commons license, contributions to the study design, analysis and and indicate if changes were made. interpretation of data, as well as preparing and reviewing the manuscript. All named authors meet the International Committee of Medical Journal Editors (ICMJE) criteria for authorship REFERENCES for this manuscript, take responsibility for the integrity of the work as a whole, and have given 1. International Diabetes Federation. IDF Diabetes ﬁnal approval to the version to be published. All Atlas, 7th Edn [Internet]; 2015 [cited 2017 July 27].: authors had full access to all of the data pre- http://www.diabetesatlas.org/across-the-globe. html. sented in this study and take complete respon- sibility for the integrity of the data and accuracy 2. National Health Expenditure 2014 [Internet]. Min- of the data analysis. istry of Health, Labour and Welfare; 2014 [cited Aug 9]. http://www.mhlw.go.jp/toukei/saikin/hw/k- Disclosures. Alena Strizek is an employee iryohi/14/dl/toukei.pdf. and stockholder of Eli Lilly and Co. Magaly 3. Fukuda H, Ikeda S, Shiroiwa T, Fukuda T. The effects Perez-Nieves is an employee and stockholder of of diagnostic deﬁnitions in claims data on health- Eli Lilly and Co. Shuichi Suzuki is an employee care cost estimates: evidence from a large-scale of Eli Lilly Japan; K.K. Zhihong Cai is an panel data analysis of diabetes care in Japan. Phar- macoeconomics. 2016;34(10):1005–14. employee of Eli Lilly Japan; K.K. Urvi Desai is an employee of Analysis Group, Inc. Jasmina Iva- 4. Japan Diabetes Society. Treatment guide for dia- nova is an employee of Analysis Group, Inc. betes 2014–2015 [Internet]; pp. 15–17. [cited 2016 Viviana Garcia-Horton is an employee of Anal- Sept 30]. http://www.jds.or.jp/modules/en/index. php?content_id=1. ysis Group, Inc. Xinyue Liu is an employee of Analysis Group, Inc. Luke Schmerold was an Diabetes Ther (2018) 9:1125–1141 1141 5. Inzucchi SE, Bergenstal RM, Buse JB, et al. Man- 13. Kimura S, Sato T, Ikeda S, Noda M, Nakayama T. agement of hyperglycemia in type 2 diabetes: a Development of a database of health insurance patient-centered approach. position statement of claims: standardization of disease classiﬁcations the american diabetes association (ADA) and the and anonymous record linkage. J Epidemiol. European Association for the Study of Diabetes 2010;20(5):413–9. (EASD). Diabetes Care. 2012;35:1364–79. 14. Japan Medical Data Center. http://www.jmdc.co.jp/ 6. Charvat H, Goto A, Goto M, et al. Impact of pop- . ulation aging on trends in diabetes prevalence: a meta-regression analysis of 160,000 Japanese adults. 15. International Health Care System Proﬁles. The Japa- J Diabetes Investig. 2015;6(5):533–42. nese Health Care System [Internet]. The Common- wealth Fund [cited 2017 Dec 11]. http://international. 7. Central Social Insurance Medical Council. General commonwealthfund.org/countries/japan/. assembly 346, function of general practitioner [In- ternet].2017. [cited 2017 Aug 9]. http://www.mhlw. 16. Quan H, Sundararajan V, Halfon P, et al. Coding go.jp/stf/shingi2/0000152389.html. algorithms for deﬁning comorbidities in ICD-9-CM and ICD-10 administrative data. Med Care. 8. Ikeda S, Crawford B, Sato M. Utilization patterns of 2005;43:1130–9. insulin therapy and healthcare services among Japanese insulin initiators during their ﬁrst year: a 17. Statistics Bureau, Ministry of Internal Affairs and descriptive analysis of administrative hospital data. Communications. Consumer price index [Internet]. BMC Health Serv Res. 2016;16(1):6. Statistics Japan [cited 2016 Sept 30]. http://www. stat.go.jp/english/data/cpi/index.htm. 9. Hadjiyianni I, Desai U, Suzuki S, et al. Basal insulin persistence, associated factors, and outcomes after 18. Ryan EA, Imes S, Wallace C. Short-term intensive treatment initiation: a retrospective database study insulin therapy in newly diagnosed type 2 diabetes. among people with type 2 diabetes mellitus in Diabetes Care. 2004;27(9):2282–3. Japan. Diabetes Ther. 2017;8(1):149–66. 19. Ishii H, Iwamoto Y, Tajima N. An exploration of 10. Hidaka H, Tokumoto K, Hori Y, Homma S, Inoue C. barriers to insulin initiation for physicians in Japan: Prescription and quality of diabetes care provided ﬁndings from the Diabetes Attitudes, Wishes and by specialists and general physicians in Japan -an Needs (DAWN) Japan study. PLoS ONE. analysis of insurance claim and health checkup 2012;7(6):e36361. data. J Japan Diab Soc. 2014;57(10):774–82. 20. Present number of medical specialists [Internet]. 11. Kohro T, Yamazaki T, Sato H, et al. Trends in Japanese Board of Medical Specialties. [cited 2017 antidiabetic prescription patterns in Japan from Aug 9]. Available from: http://www.japan-senmon- 2005 to 2011. Int Heart J. 2013;54(2):93–7. i.jp/hyouka-nintei/data/. 12. Arai K, Takai M, Hirao K, et al. Present status of 21. Chin MH, Zhang JX, Merrell K. Specialty differences insulin therapy for type 2 diabetes treated by gen- in the care of older patients with diabetes. Med eral practitioners and diabetes specialists in Japan: Care. 2000;38(2):131–40. third report of a cross-sectional survey of 15,652 patients. J Diabetes Investig. 2012;3(4):396–401.
– Springer Journals
Published: Apr 16, 2018