Understanding the quality of diabetes care in Japan: a systematic review of the literature

Abstract

Background

Among chronic diseases, diabetes is a frequent focus of performance measurement. Disease-specific indicators based on evidence-based clinical guidelines have been used to evaluate the quality of care. There are worldwide efforts to improve the quality of diabetes care. Measuring the gap is an essential first step toward improving the quality of care.

Methods

In order to better understand the status of quality of diabetes care in Japan, a country with a universal healthcare system, we performed a literature search looking for all studies reporting on quality indicators. In this review, we summarized the studies that have looked at the status of the quality of diabetes care over the last decade.

Results

There were a total of 6 studies that reported on process including HbA1c, blood pressure, lipid screening, retinopathy and nephropathy screening and intermediate clinical measures which included percentages of patients reaching targets for HbA1c, blood pressure and LDL-C. Overall, the process measures continue to improve, however the clinical intermediate outcome measures remain suboptimal.

Conclusion

Despite the improvement in diabetes related process measures, there is limited data on clinical measures. It is necessary to shed more light on the assessment of the quality of diabetes care.

Introduction

The quality of diabetes care over the last decade has improved but remains suboptimal [1]. The current guidelines and clinical recommendations acknowledge and call for interventions to improve the quality of diabetes care worldwide [2–4]. This is true for all major geographies including Japan where the prevalence of diabetes is 7.9%; [5]; more than one-quarter of Japan’s adult population may have pre-diabetes or diabetes [6]. There are 7.9 million people with diabetes between the age of 20 and 79 based on the report by the International Diabetes Federation. The healthcare expenditure is estimated at 23.5 billion dollars and is considered to be the fifth highest in the world [7]. Since medical spending is upsoaring in Japan due to population aging, it is important to understand whether this high healthcare expenditure is associated with good quality of care.

Good metabolic control is essential to prevent diabetes complications. Despite the many guidelines, there is a huge gap between the recommended targets and what can be achieved in clinical settings [1, 8–10]. There are a few studies that have focused on the quality of diabetes care collecting data from different sources. Comprehensive measurement of quality indicators could be challenging. It is important to learn how quality is defined and measured across divergent health systems. Most countries suffer from a lack of standardization on the definition of quality of diabetes care. Japan, a high-income country with a Bismarckian social insurance system can shed some light on the strength and weakness of quality definition and measurement in such system. It would be interesting to learn how a country with universal health coverage like Japan has performed on quality metrics over the years [11].

Measuring the “gap” is an essential first step toward improving the quality of primary care. To better understand the status of quality of diabetes care in Japan, we performed a literature search looking for all studies reporting on quality indicators during the last decade in Japan.

Methods

Systematic review methodology

Data sources and searches

We searched the MEDLINE (PubMed) from 2009 to 2019 using the Keywords “diabetes”, “quality” and” Japan” between 01-01-2009 and 12-31-2019 for prospective and retrospective observational cohorts looking at the quality of diabetes care in Japan. Non-English-language articles were included when identified in references.

Study selection

Original studies were included if they met the following inclusion criteria: (1) non-randomized observational studies examining the quality of diabetes care in Japan; and (2) had at least one clinical outcome measure related to diabetes defined as rate any of the following: rates of achieving A1c target of < 53 mmol/mol (7%), BP target of < 130/80 mmHg, LDL-C < 3.98 mmol/l, albuminuria, retinopathy and neuropathy or at least a process outcome measure defined by A1c testing, blood pressure monitoring at every visit, yearly lipid profile monitoring, annual eye exam and annual nephropathy screening based on the recommendation from the Japan Diabetes Society [12]. Studies were excluded if they involved an intervention, were limited to abstracts or case reports. One investigator (JM) independently examined all titles and abstracts and obtained full texts of potentially relevant papers.

Data extraction and quality indicators

For all eligible studies, we extracted publication year, the number of the total study cohort, the number of cases with the eligible outcome, the source of the data and the year they were collected, information on hemoglobin A1c (HbA1c) testing, blood pressure monitoring at every visit, yearly lipid profile monitoring, annual eye exam and annual nephropathy screening. The Japan Diabetes Society guidelines do not state the recommended frequency of HbA1c testing [12]. Hence, we have used 4 HbA1c testing during the study period as an indicator of good performance, based on the American Diabetes Association (ADA) recommendation of HbA1c testing every 3 months for unstable patients [13]. In Japanese guidelines, no specific frequency of HbA1c testing is recommended, whereas titration of treatment with regular monitoring of HbA1c is recommended; there is a consensus among specialists in Japan that HbA1c testing should be conducted every 3 months or more frequently [4]. For clinical outcome measures, we collected data on rates of achieving HbA1c target of < 53 mmol/mol (7%), BP target of < 130/80 mmHg, LDL-C < 3.98 mmol/l, monitoring albuminuria, retinopathy and neuropathy. Because of the paucity and heterogeneity of literature, we did not conduct data synthesis. Quality appraisal of included studies was done using the Quality Assessment Tool for Observational Cohort and Cross-Sectional Studies [14].

Results

Literature search

Initial search led to 634 abstracts. After the initial review, we identified 9 articles, 2 were excluded because they did not contain data on metrics limited to Japan and one was excluded because it involved an intervention. In total, 6 studies encompassing 10 cohorts were included in the final analysis [15–22]. (Fig. 1). All but one study reported on process quality indicators [17]. The quality of the data was mostly fair for all studies (Table 1).

Fig. 1.

Summary of evidence search and selection

Table 1.

Quality appraisal of included studies

Study first author	Quality rating (Good, Fair, or Poor)	Additional comments
Tomio	Fair	A1C be overestimated as the patients were limited to those who received medical care every month during the study period
Tanaka 2016	Good	17% drop out Large representative sample healthy worker effect underestimation of quality indicators: 1—inappropriately included as patients with T2D 2- 42% not on medications
Ozaki	Fair	Findings limited to samples from 4 clinics participating clinics were primary care clinics with a few physicians, physicians’ specialty and patients’ characteristics might have influenced the result
Tanaka 2019	Fair- Good	< 10% drop out Hospitalization excluded Limited to those with relatively high Socioeconomic status, aged 20–69 years; therefore, the quality of care for elderly patients was beyond the scope of the study
Sugiyama	Good	Those with hospitalization excluded. Also those who did not receive antidiabetic medication from previous year, in order to investigate the quality of care not immediately after the diagnosis
Yokoyama	Fair	Selection bias Based on provider records

Study characteristics

One study used data derived from national health insurance claim data among the elderly [15] and two studies used data from health insurance claims data managed by the Japan Medical Data Center (JMDC) Claims Database [16, 18]. One study analyzed the National Database of health insurance claims [19] and another study used medical claims and medical records [20].

We have summarized the different types of databases in Japan in Table 2; the source and the captured population. The National Database of Health Insurance Claims and Specific Health Check-ups of Japan (NDB) is a national representative database of claims and has recently become available and has been used in one study [19]. The National Health Insurance on the other hand is a type of insurance for self-employed and unemployed citizens under 75 years old, run by municipalities [15]. The Japan Medical Data Center (JMDC) claims database contains data from several health insurance societies for employees of large companies and their families [16, 18].

Table 2.

Database for the Japanese population used for assessing the quality of diabetes care

Database	Data source	Captured populations
Medical claims data, medical chart reviews, and patient surveys	Electronic Medical Records, EMR	Limited to samples from clinics using electronic medical records
Japan Medical Data Center (JMDC)	Employment based: clinical data recorded by a group of doctors	Limited to participating providers Well-processed data, easy to use since it contains ICD-10 and ATC classification. It contains ledger of the insured
The National Health insurance (NHI)	Self-employed and unemployed citizens under 75 years old	Limited to each municipality By contract, it can connect with other municipality data like death certificate, address information, and long-term care insurance claims data
The National Database (NDB)	All medical and pharmaceutical claims are collected electronically from hospitals, clinics and pharmacies by each insurer	National representative, 98.4% of claims from hospital/clinics and 99.9% from pharmacies It does not contain ledger of the insured. May often follow a patient when they transfer to another insurance The government plans to link with long-term care insurance claims data

Results in quality indicators

Process measures

Results on process measures are shown in Table 3.

Table 3.

Summary of process measures

Author	Publication year	Data source	Total Number of patients	Year of data collection	HbA1c		Blood pressure		Lipid profile		Retinopathy outcome		Kidney outcome
Tomio (1)	2010	National health insurance claims	13650	2006-2007	≥4 A1C/ year	69.8%	NR		NR		Screening 1/year	20.8%	Urine albumin excretion /year	5.8%
Tanaka (2)	2016	National health insurance claims	11500	04-2010 to 03-2011	≥1/3month	49.2%	NR		≥ 1 lipid test /year	87.3%	Screening	32.8%	Urine albumin	12.3%
Ozaki (3)	2019	4 Primary care community clinics: medical claims data, medical chart reviews, and patient surveys.	197	2014-2015	NR		BP recorded at every visit	78.7%	NR		NR		Urine protein within last year	57.5%
Tanaka (4)	2019	Health insurance claims (Japan Medical data Center: JMDC)	1908	2007	≥1/3month	68.2%	NR		≥ 1 lipid test /year	75.5%	≥ 1 retinopathy screening/ Year	42%	Urine albumin ≥1 year	14%
			3060	2009		69.3%				83.3%		40.3%		20.3%
			8531	2011		69.2%				84.9%		38.5%		21.5%
			15222	2013		68.3%				84.9%		37.8%		23.3%
			32471	2015		68.9%				85.4%		38.7%		24.2%
Sugiyama (5)	2019	The National Database of Health Insurance Claims and Specific Health Check-ups of Japan (NDB)	4,154,452	04-2015 to 03- 2016	≥4/year	77.9%	NR		NR	NR	≥ 1 retinopathy screening/ Year	46.5%	Urine albumin ≥1 year	19.4%

Process measures: NR not reported, N total number of records, HbA1C hemoglobin A1c, BP Blood pressure, K Kidney outcome, E retinopathy outcome

1.Tomio J, Toyokawa S, Tanihara S, Inoue K, Kobayashi Y. Quality of care for diabetes patients using National Health Insurance claims data in Japan. J Eval Clin Pract. 2010;16(6):1164–9

2.Tanaka H, Tomio J, Sugiyama T, Kobayashi Y. Process quality of diabetes care under favorable access to healthcare: a 2-year longitudinal study using claims data in Japan. BMJ Open Diabetes Res Care. 2016;4(1):e000291

3.Ozaki M, Matsumura S, Iwamoto M, Kamitani S, Higashi T, Toyama M, et al. Quality of primary care provided in community clinics in Japan. J Gen Fam Med. 2019;20(2):48–54

4.Tanaka H, Sugiyama T, Ihana-Sugiyama N, Ueki K, Kobayashi Y, Ohsugi M. Changes in the quality of diabetes care in Japan between 2007 and 2015: A repeated cross-sectional study using claims data. Diabetes Res Clin Pract. 2019;149:188–99

5.Sugiyama T, Imai K, Ihana-Sugiyama N, Tanaka H, Yanagisawa-Sugita A, Sasako T, et al. Variation in process quality measures of diabetes care by region and institution in Japan during 2015–2016: An observational study of nationwide claims data. Diabetes Res Clin Pract. 2019;155:107,750

Glycemic control monitoring, lipid profile monitoring, and retinopathy screening

For glycemic control monitoring, an HbA1c testing of ≥ 1 every 3 months was considered to meet the standard of care based on the Japan clinical guidelines. If rates of HbA1c testing of less than once every 3 months were reported, they were not included in the table because they were considered sub-standard. Most studies defined ≥ 4 HbA1c testing per year as an indicator of good performance. The rate of HbA1c testing ranged from 52.9% to 69.8%. Lipid profile monitoring ranged from 75.5% to 85.9%. All studies reported screening for retinopathy. The rate of screening has improved over the years with 20.8% screening in 2006 but notably less than 50% in 2016 although these data represent different cohorts.

Renal quality indicators

Renal quality indicators differed significantly between studies. One or more urinary albumin excretion (UAE) tests and/or qualitative urine albumin tests during the study period were reported in one study [15]. While another study defined 2 indicators for nephropathy screening: one or more urine microalbumin excretion tests (as recommended by the JDS guidelines) and serum creatinine tests [16]. For a patient with diabetes not receiving pharmacotherapy, urinary protein (or microalbumin level) is examined during the previous year [20]. The performance of one or both of the following tests was considered appropriate: urine protein quantitative test or urine albumin quantitative test. For the urine protein test, patients on dialysis and/or with a diagnosis of end-stage kidney disease, for whom these tests were no longer recommended were excluded. In addition, the results of serum creatinine tests in a year were assessed [18, 19].

Overall, the rate of screening for diabetic nephropathy increased from 5.8% in 2006 to 24.2% in 2015 [15, 18]. Screening for urine albumin remains very poor.

Clinical outcome measures

There was one study, the Japan Diabetes Clinical Data Management 40, (JDDM 40) that reported on clinical outcomes (Table 4). The data were based on patients’ records provided by physicians. Subjects were eligible if they had type 2 diabetes and attended one of the seventeen primary care medical clinics specializing in diabetes care, located in different areas in Japan. Those who had urinary albumin and serum creatinine measured as well as retinopathy and neuropathy evaluated within a year were registered in the study. In the JDDM40, the authors reported on the targets for blood glucose, blood pressure and lipids in 9956 subjects (average age 65 years, BMI 24.9 kg/m², BP 126/72 mmHg and HbA1c 53 mmol/mol (7.04%). Almost 30% of subjects had nephropathy, 26.4% had retinopathy and 27.7% had neuropathy [17]. Table 4 shows the proportions of patients achieving the target for HbA1c, BP, and LDL-C. A total of 20.8% achieved all three targets. As expected, the proportion free from any complication was highest at 53.2% in subjects achieving the 3 treatment targets [17].

Table 4.

Summary of clinical outcome measures

Author	Publication year	Data source	Total N	Year of data collection	HbA1c		Blood pressure		LDL-C		Retinopathy outcome		Kidney outcome
Yokoyama	2016	Primary care clinics/Japan Diabetes Clinical data management (JDDM)	13,039	2013	A1c < 7%	52.90%	BP < 130/80 mmHg	46.8%	< 120 mg/dl	65.5%	Retinopathy	26.4%	albuminuria	29.5%

Discussion

The rate of process measure testing in diabetes has improved in Japan over the years although it remains suboptimal. The proportion of patients who achieved the target for blood glucose, blood pressure and lipids remains suboptimal with 52.9% achieving HbA1c < 7%, 46.8% achieving BP < 130/80 mmHg and 65.5% achieving targets for lipids. Those proportions are similar to those reported in the US through NHANES [23]. The prevalence of microvascular disease is ~ 30%.

The rate of four HbA1c tests per year may be unnecessarily high for well-controlled persons and may be too frequent compared to the recommendations in other countries however that frequency is taken as usual in Japan. Process measures reporting has varied by prefecture and institutional characteristics. Factors that have likely contributed to the variation in the rate include insurance type, database source, age or gender distribution. There also seems to be regional differences in screening for retinopathy and nephropathy; for example, in the study by Sugiyama, retinopathy examinations were higher in prefectures with more population (e.g. Tokyo), whereas quality indicators for qualitative and quantitative urine examination were higher in northern regions (Hokkaido and Northeast). Similarly, differences were also noted by hospital type [19]. Disparity by who cared their diabetes also existed; Sugiyama et al. showed that patients were more likely to receive retinopathy monitoring when they receive antidiabetic medication in the facilities certified by the Japan Diabetes Society. Eye examination was notably low in all studies. These results suggest that the referral system between primary care doctors and ophthalmologists may not work effectively at the community level.

The rate of nephropathy screening was low. This could be due to many reasons. There is a strict requirement for urine albumin excretion tests in the reimbursement system, which allows testing only for patients with diabetes or those with early-stage diabetic nephropathy with suspected microalbuminuria. Results of individual studies may have been affected by selection bias. For example, in the JDDM40, there is likely a selection bias that might have influenced the results. The subjects included in the study were taken care of by general practitioners (GPs) specializing in diabetes and therefore the prevalence of subjects at target might be higher than the real-world practice. In contrast, the rate of microvascular complications could be higher since GPs receive a referral of subjects with more advanced disease [17]. It is also possible that this measure is under-reported specifically in clinics with ≥ 200 beds due to the comprehensive payment system. Although in the largest study available, those patients were excluded from the analysis [19].

When it comes to clinical metrics, it is important to note that we are unable to assess the quality metrics for meeting targets based on age. For example, a target of 7% for HbA1c is not recommended for all patients; the target when intensification of therapy is considered difficult is set at < 8.0%. Similarly, the target for elderly patients may be set even higher [12].

Our study has some limitations. We are unable to review the quality of care throughout the country as we are limited by the databases. The NDB and other claims provide a good overview of process measures however there are no datasets to overview the clinical outcome measures. All studies are cross-sectional in nature, which limits understanding of causality. As in any review, the quality is limited by the qualities of each individual study. However as mentioned in our quality assessment, most studies were at least fair in quality. Lastly, the number of articles that met the criteria for our review is small highlighting the need for more research on diabetes-related quality metrics in Japan.

Compared to other countries, the studies reporting on quality indicators are limited in Japan and the setting of quality indicators has been uncommon until recently. Therefore, one of the major strengths of this review is that it is extensive and includes the National Database which is unique and includes a large number of subjects which are likely to be representative of subjects with type 2 diabetes and covers almost all medical care covered by the universal health coverage. In the absence of a nation-wide diabetes registry system, which is commonly maintained in European countries, in Japan, this constitute a major strength to this study. This is not the case for clinical outcome measures which are harder to assess on the national level. We, therefore, expanded our search to a longer period, but we found only two studies (one in Japanese and the other in English) whose quality appeared to be poorer [21, 22]. One study included 498 employees in a company within the Sanyo Electric Group Health Insurance Association. Using questionnaires and health insurance claims 87% had at least one HbA1c within a year period. However, the results did not provide numerators or denominators of the proportions. There was no adjustment to see causal relationship [22]. The other study was limited by a high risk of selection bias. It included 1333 Tokyo Women’s Medical University medical records and reported 70% of periodic eye exam defined as an eye exam performed within less twice the interval set by ophthalmologist [21].

Factors contributing to suboptimal quality of diabetes care could be related to the healthcare system, the healthcare provider, the patient or a combination those factors. Primary care plays an important role in the healthcare system [24]. The Japan Diabetes Outcome Intervention Trial-2 (J-DOIT2) a large-scale, randomized controlled study has successfully improved the quality-of-care indicators. The J-DOIT2 study included a multifaceted intervention that targeted both the providers and the patients [25]. Over a 1-year period, the intervention group consisting of primary care physicians received a monthly report of their care quality with feedback. At the end of the study, and in comparison to a control group, there was a statistically significant difference in the quality of care score between the two groups.

It is important to note that the quality of diabetes care is not only evaluated by clinical metrics or process measures. Previous studies showed that poor health-related quality of life is associated with high mortality risk in the general population and in individuals with diabetes, regardless of the presence or absence of severe diabetic complications such as diabetic nephropathy or requiring hemodialysis [26, 27]. This is relevant because dealing with diabetes distress can present a psychological obstacle to effective self-management behaviors and therefore to achieving good glycemic [28, 29]. A significant positive association between high diabetes distress and all-cause mortality in Japanese men with diabetes has been found [30]. Similarly, higher physical component summary and mental component summary scores and any single 1-item scores were associated with a lower risk of all-cause mortality in individual with type 2 diabetes in Japan [31].

Compared to other countries, the health insurance system in Japan is characterized by a universal insurance scheme. Coverage is provided either through insurance programs managed by employers or through the National Health Insurance, NHI. Those who do not qualify for employer management insurance programs are covered under the NHI plan [32]. The benefit package and payment systems are uniform nationally. Despite increase in medical spending for chronic disease, it has been suggested that increase in medical spending for management of diabetes is offset by an increase in quality. For example, it has been shown that increase in diabetes related medical expenditures has been associated with lower modifiable risk of all-cause mortality in Japan [33].

In conclusion, the quality of diabetes care in Japan assessed based on process measures has improved over the years. Despite the high rate of process measures achieved, there was scant evidence about intermediate clinical outcome measures and clinical outcome measures remain suboptimal. There are many opportunities to improve the quality of diabetes care overall. Future measures should focus on improving clinical outcomes and reducing quality gaps.

Author contributions

JM designed the study, performed the literature search, collected and analyzed the data and drafted the initial manuscript. TS, HT, MO and RAG critically reviewed the work and provided substantial comments. All the authors approved the final version of the manuscript.

Compliance with ethical standards

Conflict of interest

Authors Takehiro Sugiyama, Hirokazu Tanaka, Mitsuru Ohsugi and Robert Gabbay declare they have no conflict of interest. The authors declare no potential conflict of interest. Author Joanna Mitri received honoraria for consulting for Novo Nordisk and Dairy Farmers, and receives research support from the National Dairy Council, KOWA, Inc., the NIH and the JDRF.

Ethical standards

This article does not contain any studies with human or animal subjects performed by any of the authors.