Occurrence of comorbidities in newly diagnosed type 2 diabetes patients and their impact after 11 years’ follow-up

The burden of type 2 diabetes is growing, not only through increased incidence, but also through its comorbidities. Concordant comorbidities for type 2 diabetes, such as cardiovascular diseases, are considered expected outcomes of the disease or disease complications, while discordant comorbidities are not considered to be directly related to type 2 diabetes and are less extensively addressed under diabetes management. Here we show that the combination of concordant and discordant comorbidities appears frequently in persons with diabetes (75%). Persons with combined comorbidities visited family physicians more than persons with discordant, concordant or no comorbidity (17.3 ± 10.2, 11.6 ± 6.5, 8.7 ± 6.8, 6.3 ± 6.6 visits/person/year respectively, p < 0.0001). The risk of death during the study period was highest in persons with combined comorbidities and discordant only comorbidities (HR = 33.4; 95% CI 12.5–89.2 and HR = 33.5; 95% CI 11.7–95.8), emphasizing the contribution of discordant comorbidities to the outcome. Our study is unique as a long-term follow-up of an 11-year cohort of 9725 persons with new-onset type 2 diabetes. The findings highlight the contribution of discordant comorbidity to the burden of the disease. The high prevalence of the combination of both concordant and discordant comorbidities, and their appearance before the onset of type 2 diabetes, indicates a continuum of morbidity.


Characteristics of persons with concordant and discordant comorbidities. Diseases were
grouped into concordant and discordant comorbidities according to the bodily systems affected (see Supplementary Tables S1 and S2 online). Persons with a combination of both concordant and discordant comorbidities, referred to here as "combined comorbidity", were more prevalent (n = 7316; 75%) compared to persons with concordant only or discordant only comorbidities (16.6% and 2.5%, respectively). No comorbidity was reported in 549 (5.6%) persons. Men were more represented than women in all groups of comorbidities (Table 1). Persons aged 45-65 years at the diagnosis of type 2 diabetes were most highly represented in all comorbidity groups. Jews were dominant in the group of discordant comorbidities only (70%) and Arabs were dominant in the group of concordant comorbidities only (58%). Persons without an exemption from national health insurance payments were dominant in all groups of comorbidities ( Table 1). The number of comorbidities per person was highest in the group of persons with combined comorbidities, compared to persons with concordant only or discordant only comorbidities (7.3 ± 3.6, 2.8 ± 1.6, 2.0 ± 1.3 respectively), with a gap remaining between the number of comorbidities before and after the onset of type 2 diabetes ( Table 1). The distribution of the number of diseases between groups was different; in the groups of persons with concordant only or discordant only comorbidities, more persons had few diseases, while in the group of combined comorbidities, there was a broader range of diseases (Fig. 2).
A multinomial logistic regression model was designed to assess the risk of having any type of comorbidity of type 2 diabetes, considering the parameters of gender, age at onset of diabetes type 2, ethnic origin, smoking and exemption from national health insurance payments ( Table 2). The reference group was persons without any comorbidity. Compared to persons aged 30-45 years at the onset of type 2 diabetes, those aged 45-65 years were at higher risk for having combined comorbidities [OR = 4.9; 95% CI 3.9-6.2), as were those aged 65 years or more [ Comorbidities and utilization of health services. 'Visits to family physicians' was the most utilized health service during the study period, compared to visits to consultants and hospitalizations. Visits to family physicians were performed more by persons with combined comorbidities compared to persons with concordant only or discordant only comorbidities and persons without comorbidities (17.3 ± 10.2, 11.6 ± 6.5, 8.7 ± 6.8 and 6.3 ± 6.6 visits/person/year respectively, p < 0.0001) (see Supplementary Table S3 online). The trend of annual use of health services presents differences between groups throughout the study period. Persons with combined comorbidity used them more than persons from other groups, and persons without comorbidity used them less than others; the differences between the groups were maintained over time (Figs. 3A-C). A downward trend in annual visits was observed among persons without comorbidities. Differences in hospitalization were steady throughout most of the study period with a tendency to rise in all groups, except in persons without comorbidity (Fig. 3C).

Comorbidities and mortality.
During the study period, 1785 (18%) persons died. The rate of death was highest among persons with combined comorbidity, followed by persons with discordant only comorbidities and concordant only comorbidities, and lowest among persons without comorbidity (23%, 11.5%, 4.3% and 0.7% respectively, p < 0.0001) ( Table 1).
Persons who died were more often older, Jews, smokers, and had more comorbidities than those who survived (see Supplementary Table 4 (Table 3). No differences in risk of death were observed between Arabs and Jews, smokers and non-smokers, or between persons having or not having an exemption from national health insurance payments. We plotted the Kaplan-Meier curve to describe survival probabilities by comorbidity groups during the study period (Fig. 4). The steepest slope was observed in the group of persons with combined comorbidities, while the group of persons with discordant comorbidity only had a steeper slope than that of persons with concordant comorbidity only. The graph describing survival probabilities in the group of persons with no comorbidity was remarkably stable throughout the follow-up.

Discussion
In this study of a large health maintenance organization, we followed persons from the time of onset of type 2 diabetes. At this time, most of the persons already exhibited other comorbidities. There were more individuals with both concordant and discordant comorbidities of type 2 diabetes than with either concordant or discordant comorbidities alone. The risk of all-cause death was highest in persons with combined comorbidity and in persons with discordant only comorbidities, indicating the contribution of discordant comorbidities to the outcome.  www.nature.com/scientificreports/ We defined the onset of type 2 diabetes as the starting point of our study, but we soon realized that the most common situation is that comorbidities, both concordant and discordant, appear before and after the onset of type 2 diabetes. The demonstration of prevalent comorbidities at the onset of diabetes corroborates previous reports 6,7,13 . This may be explained, at least in part, by the risk factors shared by type 2 diabetes and cardiovascular diseases, such as the metabolic syndrome 14 . The essence of type 2 diabetes may be a sequence of events in the development of comorbidities, rather than a definitive starting point.
The association of advanced age with comorbidities in persons with diabetes concurs with previous reports 3, 4 . In our study, older age was associated with risk for both comorbidity and death.
Our study sample is characterized by near equal gender distribution; men had a higher risk for comorbidity and for death. A previous study reported higher prevalence in women than men of comorbidity in general, but not specifically of diabetes 3 .
Exemption from national health insurance payments, indicating low SES, was not associated with comorbidity and its outcomes, unlike previous findings 3,4 . This difference may be attributed to the national health insurance in Israel, which includes primary care services and hospitalization free of charge.
Arab ethnic origin was associated with higher risk for all categories of comorbidity. Considering cardiovascular diseases, which is the dominant concordance comorbidity, the outcome is in line with previous reports in Israel, of generally higher rates of cardiovascular diseases among Arabs than among Jews [15][16][17] . A novel aspect of the current study is the report of a higher rate of discordant comorbidity in the Arab population in Israel.
In our study, persons with combined comorbidities of type 2 diabetes consistently used health services more frequently than did persons with only concordant or discordant comorbidities. 'Visits to family doctors' was the health service most used in the community, more so among persons with combined comorbidity. The same trend was observed for hospitalization. Interestingly, among persons without comorbidities, the use of health services declined during the follow-up period. This could be due to better health or self-care. Our findings corroborate a report, based on private insurance company claims, of higher health service costs among persons with combined comorbidities than among persons with either discordant or concordant diseases alone 18 . In contrast to that study, our study is based on data from the national insurance health system. A recent publication from the Netherlands also reported higher societal costs of persons with diabetes and lower utilities 19 .
The death rate was higher in persons with combined comorbidities of type 2 diabetes, compared to only concordant or discordant comorbidities, or no comorbidities. The risk of death was similar for persons with combined comorbidities and discordant only comorbidities, although persons with discordant comorbidities only had fewer diseases. This raises the issue of the contribution of specific diseases, which is not addressed in our study. This may be consequent to the inclusion of malignant diseases in our study. Comorbidity of malignant diseases in persons with diabetes has been previously reported, and partly explained by shared risk factors 20,21 .
Other parameters that were identified as carrying higher risk for death in persons with type 2 diabetes were age 45 years and older, and male gender, in addition to comorbidities. Our study did not find Arab background to be an independent risk factor for mortality. Previous studies reported higher rates of all-cause mortality among Arabs, particular with respect to cardiovascular diseases 22,23 . However, combined comorbidities as defined herein were not reported in those studies.
We have not found another community-based study that compared mortality rates by concordance of comorbidities of type 2 diabetes. Previous publications that reported mortality in persons with diabetes focused on trends of mortality, showing that the decline in persons with diabetes lags behind those without diabetes [24][25][26] .
The strengths of our study include the long-term follow-up from the year of diagnosis of type 2 diabetes, the considerable sample size, and the real-life data of persons in the community. The large sample enabled us to draw conclusions regarding specific modes of the disease. This study is based on robust, valid sources of information, including the Clalit Health Services (CHS) chronic disease register, which is comprehensive and continuously updated 27 with ongoing updated quality measures 28 .
Our study is limited by the crudeness of the data on type 2 diabetes, without accounting for specific diagnosis or severity of diseases, or for the individual level of control of type 2 diabetes or other chronic diseases. All-cause death was a study outcome, and not a specific cause of death. This is because the reason for death is not identified in the CHS registers. The analysis of death events using the Cox model was limited by the few cases recorded in the group of persons without comorbidities, which served as a reference group. As a result, the assumption of proportionality which underlies the Cox model was not met. Still, we consider it sensible to use the Cox model to examine the hazard as a function of the tested parameters during the study period.
Our data included services within the framework of public services. We do not have information on the use of private services, which could be relevant mainly for consultations. Our data did not include social information such as education or income. Instead, we used the exemption from national health insurance payments as an indicator of low SES. Since Israel operates a national health insurance system that provides services equally to all citizens, exemption from payment is an indirect indicator of SES. The health services assessed in our study are free of charge (visits to family physicians and hospitalization) or covered by a small copayment (visit to consultants). Medications are purchased at a subsidized price, which is adjusted for low SES and the number of medications. This health system thus narrows financial gaps for chronic patients. This could explain why SES was not found to be associated with the outcomes of our study.
In our databases, the validity of diagnosis of concordant diseases is high due to their inclusion in the Quality Measures program. Data related to discordant diseases may be less accurate, and could be under-reported. In our study, we grouped the chronic diseases in the list into systems. We tried to be most accurate in our grouping; however, some mis-grouping may have occurred. Malignant diseases were considered discordant diseases and may have contributed to the high rate of mortality. We chose not to exclude persons with malignant diseases, as in previous studies 9,12,29,30 , since this would necessitate excluding persons with other comorbidities as well.

Methods
Design. This is a retrospective cohort study.

Participants.
The study included all adults aged 30-90 years on January 1, 2007, insured by CHS and residing within the northern district of Israel. Persons were included if they had a blood glucose test suggestive of new-onset type 2 diabetes in the year 2007, defined as a blood glucose level of ≥ 126 mg/dL (7.0 mmol/L) or a hemoglobin A1C of ≥ 6.5% (48 mmol/mol). Data were validated by the absence of a diagnosis of type 2 diabetes, or purchase of medications for type 2 diabetes during 2006. Study exclusion criteria were: gestational diabetes, insulin initiation at the diagnosis of type 2 diabetes and death recorded in the first year of the study. We excluded persons who died in the first year after inclusion so that the participants had at least one year of diabetes.
Israel operates a national health insurance scheme. Residents select one of four national health service providers, and an affiliated personal family physician 31 . CHS provides primary care to 70% of the inhabitants in the northern region of Israel. Primary care is provided in clinics by family physicians and consultants. Primary care clinics and health centers are dispersed throughout the captured area and accessible within a 30-min drive. Visits to family physicians are free of charge. Visits to consultants carry a low copayment. Admissions to hospitals are free of charge.
Persons were monitored until death or the end of the study period, December 31, 2017.
Source of data. CHS operates an integrated electronic medical and administrative file, based on the International Classification of Diseases (9th Revision) combined with ICPC coding. All visits to doctors, either in the community or in hospitals, as well as hospitalizations, are recorded in the CHS database. Chronic conditions recorded in the CHS central register are based on reports by family physicians and community-based specialists, and hospital discharge letters. Chronic diseases that take part in the Quality Measures program, such as type 2 diabetes, cardiovascular diseases, asthma, and chronic obstructive pulmonary disease are also cross-validated against medication possession records and laboratory data through an automated disease-specific process 27,32 . We grouped diseases into bodily systems for analysis (Supplementary Table S1 online).
Dates of death (but not causes) are updated through direct linkage to the Israel Ministry of Internal Affairs, using a unique national personal identification number.
Baseline sociodemographic data included age at onset of diabetes, gender and ethnicity (Jewish or Arabs). The only available measure indicating life style behavior was smoking. The cohort was classified into three age groups: 30-45, 45-65 and 65 years and older. Eligibility for exemption from national health insurance payments is given to persons with very low income or who are unemployed for a long period, and was considered an indicator of low SES, as in previous studies 33,34 .
Type 2 diabetes and comorbidities were documented by the date of recording. We classified comorbid diseases using the system suggested by Piett and Kerr and developed by Mangan et al. 8,10 (see Supplementary Table 2). Accordingly, concordant diseases share similar pathophysiology with type 2 diabetes and discordant diseases comprise the remaining pathologies. This classification was used in studies that evaluated quality of care, cost of care and self-care of type 2 diabetes 10,18,29,30 . Data on utilization of health services in the community, including visits to family physicians and consultants, were retrieved for all persons included in the cohort.
Main measures. Persons were monitored for 11 years from the diagnosis of type 2 diabetes, stratified into groups according to whether comorbidities were concordant, discordant, combined (concordant and discordant) or absent. The use of health services and mortality were compared between the groups according to sociodemographic variables, the type of comorbidity and the number of comorbidities.
Statistics. The data were analyzed using SAS version 9.4; p < 0.05 was considered significant. Categorical data were reported as numbers (%) and compared using the chi-square test. Multinomial logistic regression analysis was used to assess the risk of comorbidities of type 2 diabetes. A Cox regression model was used to identify factors that predict mortality. Continuous variables were reported as mean ± SD. Between-group comparisons regarding health services were performed using the Kruskal-Wallis test for non-parametric data.
Ethical approval was obtained from the CHS institutional ethics review board, with exemption from informed consent in data-based research using anonymous information (0137-17-COM2).
Statement: All methods were performed in accordance with the relevant guidelines.