Abstract
Background:
Electronic health records (EHRs) and claims records are widely used in defining type 2 diabetes mellitus (T2DM) complications across different types of health care encounters.
Objective:
This study investigates whether using different EHR encounter types to define diabetes complications may lead to different results when examining associations between diabetes complications and their risk factors in patients with T2DM.
Research Design:
The study cohort of 64,855 adult patients with T2DM was created from EHR data from the Research Action for Health Network (REACHnet), using the Surveillance Prevention, and Management of Diabetes Mellitus (SUPREME-DM) definitions. Incidence of coronary heart disease (CHD) and stroke events were identified using International Classification of Diseases (ICD)-9/10 codes and grouped by encounter types: (1) inpatient (IP) or emergency department (ED) type, or (2) any health care encounter type. Cox proportional hazards regression was used to estimate associations between diabetes complications (ie, CHD and stroke) and risk factors (ie, low-density lipoprotein cholesterol and hemoglobin A1c).
Results:
The incidence rates of CHD and stroke in all health care settings were more than twice the incidence rates of CHD and stroke in IP/ED settings. The age-adjusted and multivariable-adjusted hazard ratios for incident CHD and stroke across different levels of low-density lipoprotein cholesterol and hemoglobin A1c were similar between IP/ED and all settings.
Conclusion:
While there are large variations in incidence rates of CHD and stroke as absolute risks, the associations between both CHD and stroke and their respective risk factors measured by hazard ratios as relative risks are similar, regardless of alternative definitions.
Keywords: type 2 diabetes mellitus, complications, electronic health records, stroke, coronary heart disease
Diabetes complications are clinically defined as the injurious effects of hyperglycemia on human microvascular and macrovascular systems.1–5 Diabetes complications play a major role in the heavy economic and disease burdens of diabetes, both in the United States and globally.6,7 In the United States, the total estimated direct medical cost of type 2 diabetes mellitus (T2DM) and its related complications was reported at $237 billion in 2017 while the global economic burden due to T2DM was reported at $1.3 trillion in 2015.8–10
Electronic health records (EHRs) and claims data are widely used to define T2DM complications in observational studies, based on the information on diagnoses, labs, procedures, medications, biomarkers, and health care settings.11–13 Research may use EHR or claims data from one or many health care settings based on their study objectives (eg, costing health services in cost-effectiveness vs. examining associations between outcomes and risk factors). Thus diabetes complication rates vary across different studies, which may be partially attributable to types of encounters used to define them.14–16 Because patients usually have multiple diagnoses in different encounters, the first diagnosis of diabetes complications has usually been used to identify a patient’s complication status and the encounter type of complication identification.14,17 In the literature, the ways in which researchers identify T2DM complications in different health care settings or encounters differ across studies.15,18–32 While the studies that analyzed the development or incidence of T2DM complications usually used diagnosis in inpatient (IP) settings and emergency departments (EDs).16,19,21,22,24,25,32 Many studies that include T2DM complication histories collect diagnosis data from all health care encounters.15,18,20,23,26–31
There is a gap in understanding whether disease identification when using EHRs and claims data18–34 may differ by type of health care setting where diagnosis records are retrieved. In addition, coronary heart disease (CHD) and stroke are 2 important diabetes complications among patients with T2DM.3,35 Thus, this study aims to assess rates of CHD and stroke in patients with T2DM based on different encounter types in EHRs and assess relative and absolute risks of CHD and stroke.
METHODS
Data and Study Population
EHR data on patients with T2DM in the LEAD (Louisiana Experiment Assessing Diabetes outcomes) cohort study were obtained from the Research Action for Health Network (REACHnet), including EHR data for the time period between January 1, 2013, and October 31, 2017. Clinical data from REACHnet are conformed to the PCORnet Common Data Model (https://pcornet.org/pcornet-common-data-model/). These measures were also previously used in the Louisiana Experiment Assessing Diabetes (LEAD) outcomes study.14,17,36 The definition of T2DM in the present study was formulated according to the Surveillance Prevention, and Management of Diabetes Mellitus (SUPREME-DM) definitions as follows: (1) ≥ 1 of the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes and 10th Revision, Clinical Modification (ICD-10-CM) codes for T2DM associated with IP encounters; (2) ≥ 2 ICD codes associated with outpatient encounters on different days within 2 years; (3) combination of ≥ 2 of the following associated with outpatient encounters on different days within 2 years: (1) ICD codes; (2) fasting glucose level ≥ 126 mg/dL; (3) 2-hour glucose level ≥ 200 mg/dL; (4) random glucose ≥ 200 mg/dL; (5) hemoglobin A1c (HbA1c) ≥ 6.5%; and (6) prescription for an antidiabetic medications.37 The total population of the present study included 64,855 patients with T2DM. The baseline characteristics were collected starting from the first recorded date of the T2DM diagnosis and were limited to 1 year before the date of T2DM diagnosis. The study and analysis plan were approved by Tulane University (IRB#906810).
Measurements of Health Care Settings
The encounter type variable in the PCORnet Common Data Model was used to identify health care settings for each patient encounter. The health care settings were categorized as: (1) IP/ED; and (2) all health care settings. IP/ED encounters included IP visits, ED visits, emergency admit to IP, institutional stay, observation stay, or institutional consult. All health care settings included IP/ED, outpatient encounters (all types of ambulatory visits), and any other encounter type that cannot be categorized as the encounter types mentioned above.
Measurements of Coronary Heart Disease and Stroke
The measurements of CHD and stroke in this study are consistent with the LEAD outcomes study.14,17 The ICD-9-CM and ICD-10-CM codes were used to identify CHD (ICD-9-CM codes 410-414, 429.2, ICD-10-CM codes I20-I25), and stroke (ICD-9-CM codes 430-436, ICD-10-CM codes I60-I66).
These diagnoses were recorded in the course of routine patient care by a patient’s treating clinicians. The T2DM diagnosis date is defined as the date of the first documented diabetes mellitus diagnosis. While the data lack the histories of CHD and stroke, our approach helped to ensure that the new cases of diagnoses of CHD and stroke occurred after T2DM diagnosis date as: (1) No T2DM complication diagnosis before the T2DM diagnosis date in any health care settings; (2) at least 1 T2DM complication diagnosis in IP/ED or any other health care settings after the T2DM diagnosis date. A cohort member with both CHD and stroke was measured respectively according to each T2DM complications. The duration of follow-up for each cohort member (person-years) was tabulated from the date of the first documented diabetes mellitus diagnosis to the date of diagnosis of the outcome, death of IPs or October 31, 2017. The incidence rate as the absolute risk was measured as the number of new cases of CHD and stroke per person in the study population over the total person-years for the cohort.
Statistical Analyses
Cox proportional hazards regression was used to estimate hazard ratios (HRs) as relative risk measures for incident CHD and stroke by different levels of low-density lipoprotein (LDL) cholesterol and HbA1c, respectively. LDL cholesterol was evaluated as categories [ < 70, 70–99, 100–129 (reference group), 130–159, 160–189, ≥ 190 mg/dL].38 HbA1c was evaluated as categories [ < 6.0, 6.0–6.5 (reference group), 6.5–7.0, 7.0–7.9, 8.0–8.9, 9.0–9.9, and ≥ 10%]. LDL cholesterol levels and HbA1c levels were included in the models as dummy variables. The patients with diagnoses of CHD or stroke only in outpatient settings/any other encounter were excluded in the analysis of relative risk for incident CHD and stroke in IP/ED encounters. In general, all proportionality assumptions were appropriate (Appendix 1, Supplemental Digital Content 1, http://links.lww.com/MLR/B971). All analyses were first carried out adjusting for age, and further for LDL cholesterol (if according to HbA1c levels) or HbA1c (if according to LDL cholesterol levels), sex, race, body mass index, systolic blood pressure, high-density lipoprotein cholesterol, triglycerides, estimated glomerular filtration rate, smoking, insurance type, use of antihypertensive drugs, use of diabetes medications, and use of lipid-lowering agents (Appendix 2, Supplemental Digital Content 1, http://links.lww.com/MLR/B971). These data elements were collected starting from the date of the diabetes mellitus diagnosis (baseline). All statistical analyses were performed by using and SAS for Windows, version 9.4 (SAS Institute Inc., Cary, NC).
RESULTS
Baseline characteristics of the study population are presented in Table 1. The mean age and SD of the T2DM population were 66.8 (12.1) years old. Overall, 52.3% of the study population were women, and 11.6% of the population were current smokers. At baseline, the mean (SD) percentage of HbA1c was 7.3% (1.6%), which is higher than the goal of 7% for diabetes patients. In addition, the mean (SD) of systolic and diastolic blood pressure were 133.4 (13.0) and 75.1 (8.1) mmHg, respectively. At baseline, the mean (SD) of LDL cholesterol, and high-density lipoprotein cholesterol were 97.8 (32.3) and 44.0 (12.4) mg/dL, respectively. The mean (SD) of triglycerides was 144.0 (88.3) mg/dL, and the mean (SD) of estimated glomerular filtration rate was 68.4 (22.7) mL/min/ 1.73 m2. Last, the percentages of the population who have used lipid-lowering, antidiabetic, and antihypertensive drugs were 60.2%, 71.2%, 76.9%, respectively.
TABLE 1.
Characteristics of Type 2 Diabetes Mellitus Cohort (N = 64,855)
| Mean (SD) |
|||
|---|---|---|---|
| Demographics | Patients at Risk in All Settings | Patients at CHD Risk in IP/ED Settings Only | Patients at Stroke Risk in IP/ED Settings Only |
| Age (y) | 66.8 (12.1) | 66.2 (12.1) | 66.1 (12.0) |
| Body mass index (kg/m2) | 32.8 (7.5) | 32.8 (7.5) | 33.0 (7.5) |
| Sex (%) | |||
| Man | 47.8 | 47.0 | 47.7 |
| Woman | 52.3 | 53.0 | 52.3 |
| Race (%) | |||
| White | 40.0 | 41.0 | 40.2 |
| Nonwhite | 60.0 | 59.0 | 59.8 |
| Current smoker (%) | 11.6 | 11.0 | 11.1 |
| Insurance type (%) | |||
| Private | 44.9 | 46.4 | 47.0 |
| Medicare | 41.9 | 40.2 | 39.7 |
| Medicaid | 3.4 | 3.5 | 3.4 |
| Self-pay | 5.4 | 5.4 | 5.4 |
| Other | 4.4 | 4.5 | 4.5 |
| Clinical or biomarkers HbA1c (%) | 7.3 (1.6) | 7.3 (1.6) | 7.3 (1.6) |
| Blood pressure (mm Hg) | |||
| Systolic | 133.4 (13.0) | 133.4 (13.1) | 133.3 (12.9) |
| Diastolic | 75.1 (8.1) | 75.5 (8.2) | 75.5 (8.0) |
| LDL cholesterol (mg/dL) | 97.8 (32.3) | 98.8 (32.1) | 98.6 (32.1) |
| HDL cholesterol (mg/dL) | 44.0 (12.4) | 44.1 (12.5) | 44.0 (12.4) |
| Triglycerides (mg/dL) | 144.0 (88.3) | 144.0 (89.4) | 144.5 (89.6) |
| Estimated GFR (mL/min/1.73 m2) | 68.4 (22.7) | 69.2 (22.5) | 69.2 (22.5) |
| Medications (%) | |||
| Hypolipidemic drugs | 60.2 | 58.5 | 58.5 |
| Antidiabetic drugs | 71.2 | 70.3 | 70.3 |
| Antihypertensive drugs | 76.9 | 75.8 | 75.8 |
CHD indicates coronary heart disease; GFR, glomerular filtration rate; HbA1c, hemoglobin A1c; HDL, high-density lipoprotein; IP/ED, inpatient/emergency department; LDL, low-density lipoprotein.
Table 2 presents the incidence rates of CHD and adjusted HRs by HbA1c levels. The incidence rates derived from all health care settings were 2.87 times the incidence rates of CHD in IP/ED encounters, which is 74.48 versus 25.94 cases per 1000 person-years. Across different levels of baseline HbA1c [ < 6.0, 6.0–6.5 (reference group), 6.5–7.0, 7.0–7.9, 8.0–8.9, 9.0–9.9, and ≥ 10%], the age-adjusted and multivariable-adjusted HRs for CHD show a U-shape pattern in associations between CHD and HbA1c. The baseline HbA1c 6.0–6.5 was associated with the lowest risk for CHD. Similar HR estimates were found, between IP/ED encounters and all health care settings. Specifically, significant differences of age-adjusted HR estimates were found in HbA1c levels, including <6.0, 9.0–9.9, and ≥ 10%, which were 1.22 (1.08–1.38), 1.21 (1.03–1.42), and 1.45 (1.24–1.70), respectively. However, with multivariable adjustment, a significant difference between adjusted HR estimates was only for HbA1c level ≥ 10, which was 1.34 (1.14–1.58).
TABLE 2.
Incidence Rates of CHD by Type of Health Care Encounters and Associations Between CHD and Baseline HbA1c
| Health Care Settings | Baseline HbA1c (%) | ||||||
|---|---|---|---|---|---|---|---|
|
| |||||||
| < 6.0 | 6.0–6.5 | 6.5–7.0 | 7.0–7.9 | 8.0–8.9 | 9.0–9.9 | ≥ 10 | |
| CHD cases diagnosed in all health care settings | |||||||
| No. patients | 9726 | 12,961 | 11,641 | 14,537 | 7371 | 3898 | 4721 |
| No. cases | 1777 | 2165 | 2080 | 2952 | 1512 | 794 | 735 |
| Person-years | 20,863 | 31,637 | 29,924 | 38,589 | 18,873 | 10,077 | 11,363 |
| Incidence rate | 85.17 | 68.43 | 69.51 | 76.50 | 80.11 | 78.79 | 64.68 |
| Age-adjusted HR (95% CI) | 1.25 (1.18–1.33) | 1.00 | 1.06 (1.00–1.12) | 1.23 (1.16–1.30) | 1.40 (1.31–1.49) | 1.50 (1.30–1.54) | 1.41 (1.30–1.54) |
| Multivariable-adjusted HR (95% CI) | 1.15 (1.08–1.23) | 1.00 | 1.03 (0.97–1.09) | 1.11 (1.05–1.18) | 1.24 (1.16–1.32) | 1.38 (1.27–1.51) | 1.40 (1.28–1.53) |
| CHD cases diagnosed only in inpatient and emergency encounters | |||||||
| No. patients | 8609 | 11,458 | 10,163 | 12,446 | 6350 | 3398 | 4320 |
| No. cases | 660 | 662 | 602 | 861 | 491 | 294 | 334 |
| Person-years | 19,389 | 29,536 | 27,894 | 35,859 | 17,539 | 9399 | 10,885 |
| Incidence rate | 34.04 | 22.41 | 21.58 | 24.01 | 27.99 | 31.28 | 30.68 |
| Age-adjusted HR (95% CI) | 1.53 (1.37–1.70) | 1.00 | 1.00 (0.90–1.12) | 1.19 (1.07–1.32) | 1.47 (1.31–1.66) | 1.81 (1.58–2.08) | 2.05 (1.79–2.34) |
| Multivariable-adjusted HR (95% CI) | 1.31 (1.17–1.46) | 1.00 | 1.00 (0.90–1.12) | 1.10 (0.99–1.22) | 1.31 (1.16–1.48) | 1.63 (1.41–1.87) | 1.87 (1.63–2.15) |
| Difference in age-adjusted HR (95% CI) | 1.22 (1.08–1.38)* | — | 0.95 (0.84–1.08) | 0.97 (0.86–1.09) | 1.05 (0.92–1.21) | 1.21 (1.03–1.42)* | 1.45 (1.24–1.70)* |
| Difference in multivariable-adjusted HR (95% CI) | 1.14 (1.00–1.29) | — | 0.97 (0.86–1.10) | 0.99 (0.88–1.12) | 1.06 (0.92–1.22) | 1.18 (1.00–1.39) | 1.34 (1.14–1.58)* |
The unit of incidence rate is the number of new cases per 1000 person-years.
P < 0.05.
CHD indicates coronary heart disease; CI, confidence interval; HbA1c, hemoglobin A1c; HR, hazard ratio.
Table 3 presents the incidence rates of CHD and adjusted HRs by LDL cholesterol levels. The incidence rates derived from all health care settings were almost 3 times the incidence rates of CHD in IP/ED encounters, which is 74.48 versus 25.94 cases per 1000 person-years in total. In the level of LDL cholesterol ≥ 190 mg/dL, the incidence rates derived from all health care settings were 1.89 times the incidence rates of CHD in IP/ED encounters. Across different levels of LDL cholesterol [70, 70–99, 100–129 (reference group), 130–159, 160–189, ≥ 190 mg/dL], the age-adjusted and multivariable-adjusted HRs for CHD show a U-shape pattern in associations between CHD and LDL cholesterol. The baseline LDL cholesterol 130–159 was associated with the lowest risk for CHD when adjusting for age, and the baseline LDL cholesterol 100–129 was associated with the lowest risk for CHD when adjusting for all covariates. Similar HR estimates were found, between IP/ED encounters and all health care settings. Specifically, significant differences in both age-adjusted and multivariable-adjusted HR estimates were only found for LDL cholesterol ≥ 190 mg/dL, which were 1.50 (1.05–2.13) and 1.55 (1.01–2.39), respectively.
TABLE 3.
Incidence Rates of CHD by Type of Health Care Encounters and Associations Between CHD and Baseline LDL Cholesterol
| Health Care Settings | Baseline LDL Cholesterol (mg/dL) | |||||
|---|---|---|---|---|---|---|
|
| ||||||
| < 70 | 70–99 | 100–129 | 130–159 | 160–189 | ≥ 190 | |
| CHD cases diagnosed in all health care settings | ||||||
| No. patients | 12,127 | 24,745 | 18,413 | 7059 | 1881 | 630 |
| No. cases | 3057 | 4908 | 2781 | 897 | 276 | 96 |
| Person-years | 23,886 | 63,094 | 50,239 | 18,187 | 4495 | 1425 |
| Incidence rate | 127.98 | 77.79 | 55.36 | 49.32 | 61.40 | 67.37 |
| Age-adjusted HR (95% CI) | 1.93 (1.83–2.03) | 1.28 (1.22–1.34) | 1.00 | 0.94 (0.87–1.01) | 1.17 (1.03–1.32) | 1.37 (1.12–1.68) |
| Multivariable-adjusted HR (95% CI) | 1.30 (1.20–1.41) | 1.07 (1.01–1.13) | 1.00 | 1.09 (1.00–1.18) | 1.41 (1.23–1.63) | 1.74 (1.37–2.21) |
| CHD cases diagnosed only in inpatient and emergency encounters | ||||||
| No. patients | 10,050 | 21,360 | 16,577 | 6473 | 1702 | 582 |
| No. cases | 980 | 1523 | 945 | 311 | 97 | 48 |
| Person-years | 21,604 | 58,571 | 47,422 | 17,303 | 4254 | 1349 |
| Incidence rate | 45.36 | 26.00 | 19.93 | 17.97 | 22.80 | 35.58 |
| Age-adjusted HR (95% CI) | 1.93 (1.77–2.11) | 1.20 (1.10–1.30) | 1.00 | 0.96 (0.84–1.09) | 1.24 (1.01–1.53) | 2.05 (1.54–2.74) |
| Multivariable-adjusted HR (95% CI) | 1.17 (1.01–1.36) | 0.99 (0.90–1.10) | 1.00 | 1.15 (0.99–1.32) | 1.61 (1.26–2.05) | 2.70 (1.88–3.87) |
| Difference in age-adjusted HR (95% CI) | 1.00 (0.90–1.11) | 0.93 (0.85–1.03) | — | 1.02 (0.88–1.18) | 1.06 (0.83–1.35) | 1.50 (1.05–2.13)* |
| Difference in multivariable-adjusted HR (95% CI) | 0.90 (0.76–1.07) | 0.93 (0.83–1.04) | — | 1.05 (0.89–1.24) | 1.14 (0.86–1.50) | 1.55 (1.01–2.39)* |
The unit of incidence rate is the number of new cases per 1000 person-years.
P < 0.05.
CHD indicates coronary heart disease; CI, confidence interval; HR, hazard ratio; LDL, low-density lipoprotein.
Table 4 presents the incidence rates of stroke and adjusted HRs by HbA1c levels. The incidence rates derived from all health care settings were > 5 times the incidence rates of stroke in IP/ED settings, which is 42.33 versus 7.75 cases per 1000 person-years. Across different levels of baseline HbA1c [ < 6.0, 6.0–6.5 (reference group), 6.5–7.0, 7.0–7.9, 8.0–8.9, 9.0–9.9, and > 10%], the age-adjusted and multivariable-adjusted HRs for stroke also show a U-shape pattern in associations between stroke and HbA1c. The baseline HbA1c 6.0–6.5 was associated with the lowest risk for stroke in IP/ED encounters, and baseline HbA1c 6.5–7.0 was associated with the lowest risk for stroke in all health care settings. Similar HR estimates were found, between IP/ED encounters and all health care settings. Specifically, significant differences in both age-adjusted HR estimates were found in HbA1c levels, including <6.0 and ≥ 10%, which were 1.36 (1.12–1.65) and 2.16 (1.72–2.73), respectively. After adjusting for multivariable, the significant differences in adjusted HR estimates were lower than the differences of age-adjusted HR estimates in HbA1c level <6.0 and ≥ 10%, which were 1.28 (1.05–1.56) and 1.95 (1.53–2.48), respectively.
TABLE 4.
Incidence Rates of Stroke by Type of Health Care Encounters and Associations Between Stroke and Baseline HbA1c
| Health Care Settings | Baseline HbA1c (%) | ||||||
|---|---|---|---|---|---|---|---|
|
| |||||||
| < 6.0 | 6.0–6.5 | 6.5–7.0 | 7.0–7.9 | 8.0–8.9 | 9.0–9.9 | ≥ 10 | |
| Stroke cases diagnosed in all health care settings | |||||||
| No. of patients | 9726 | 12,961 | 11,641 | 14,537 | 7371 | 3898 | 4721 |
| No. cases | 1282 | 1580 | 1372 | 1997 | 1015 | 530 | 513 |
| Person-years | 25,337 | 38,077 | 35,930 | 47,438 | 23,413 | 12,164 | 13,471 |
| Incidence rate | 50.60 | 41.49 | 38.19 | 42.10 | 43.35 | 43.57 | 38.08 |
| Age-adjusted HR (95% CI) | 1.22 (1.13–1.31) | 1.00 | 0.96 (0.89–1.03) | 1.10 (1.03–1.17) | 1.23 (1.14–1.33) | 1.42 (1.28–1.56) | 1.39 (1.25–1.54) |
| Multivariable-adjusted HR (95% CI) | 1.10 (1.03–1.19) | 1.00 | 0.96 (0.90–1.04) | 1.06 (0.99–1.13) | 1.15 (1.06–1.25) | 1.33 (1.20–1.47) | 1.28 (1.15–1.42) |
| Stroke cases diagnosed only in inpatient and emergency encounters | |||||||
| No. patients | 8697 | 11,606 | 10,484 | 12,847 | 6526 | 3465 | 4371 |
| No. cases | 253 | 225 | 215 | 307 | 170 | 97 | 163 |
| Person-years | 23,824 | 35,852 | 34,071 | 44,552 | 22,016 | 11,455 | 12,860 |
| Incidence rate | 10.62 | 6.28 | 6.31 | 6.89 | 7.72 | 8.47 | 12.67 |
| Age-adjusted HR (95% CI) | 1.65 (1.38–1.98) | 1.00 | 1.02 (0.85–1.23) | 1.15 (0.97–1.37) | 1.47 (1.20–1.79) | 1.77 (1.39–2.25) | 3.00 (2.44–3.70) |
| Multivariable-adjusted HR (95% CI) | 1.41 (1.18–1.70) | 1.00 | 1.05 (0.87–1.27) | 1.14 (0.96–1.36) | 1.37 (1.12–1.68) | 1.61 (1.26–2.06) | 2.50 (2.01–3.10) |
| Difference in age-adjusted HR (95% CI) | 1.36 (1.12–1.65)* | — | 1.07 (0.88–1.31) | 1.05 (0.87–1.27) | 1.19 (0.96–1.48) | 1.25 (0.97–1.62) | 2.16 (1.72–2.73)* |
| Difference in multivariable-adjusted HR (95% CI) | 1.28 (1.05–1.56)* | — | 1.09 (0.89–1.33) | 1.08 (0.89–1.30) | 1.19 (0.95–1.49) | 1.21 (0.93–1.59) | 1.95 (1.53–2.48)* |
The unit of incidence rate is the number of new cases per 1000 person-years.
P < 0.05.
CI indicates confidence interval; HbA1c, hemoglobin A1c; HR, hazard ratio.
Table 5 presents the incidence rates of stroke and adjusted HRs by LDL cholesterol levels. The incidence rates derived from all health care settings were > 5 times the incidence rates of stroke in IP/ED encounters, which is 46.74 versus 8.87 cases per 1000 person-years in total. For LDL cholesterol ≥ 190 mg/dL, the incidence rates from all health care settings were 2.63 times the incidence rates of stroke in IP/ED encounters. Across different levels of LDL cholesterol [ < 70, 70–99, 100–129 (reference group), 130–159, 160–189, ≥ 190 mg/dL], the age-adjusted and multivariable-adjusted HRs for stroke show a U-shape pattern in associations between stroke and LDL cholesterol. The baseline LDL cholesterol 70–99 was associated with the lowest risk for stroke in IP/ED encounters, and the baseline LDL cholesterol 130–159 was associated with the lowest risk for stroke in all health care settings. Similar HR estimates were found between IP/ED encounters and all health care settings. Specifically, significant differences of age-adjusted HR estimates were found in LDL cholesterol levels 130–159 and ≥ 190 mg/dL, which were 1.27 (1.02–1.59) and 1.97 (1.23–3.16), respectively. After adjusting for multivariable, the significant differences in adjusted HR estimates were found in LDL cholesterol levels 160–189 and ≥ 190 mg/dL, which were 1.54 (1.02–2.35) and 2.19 (1.19–4.01), respectively.
TABLE 5.
Incidence Rates of Stroke by Type of Health Care Encounters and Associations Between Stroke and Baseline LDL Cholesterol
| Health Care Settings | LDL Cholesterol (mg/dL) | |||||
|---|---|---|---|---|---|---|
|
| ||||||
| < 70 | 70–99 | 100–129 | 130–159 | 160–189 | ≥ 190 | |
| Stroke cases diagnosed in all health care settings | ||||||
| No. patients | 12,127 | 24,745 | 18,413 | 7059 | 1881 | 630 |
| No. cases | 2034 | 3324 | 2050 | 627 | 183 | 71 |
| Person-years | 35,207 | 77,537 | 56,234 | 20,313 | 5015 | 1523 |
| Incidence rate | 57.77 | 42.87 | 36.45 | 30.87 | 36.49 | 46.62 |
| Age-adjusted HR (95% CI) | 1.29 (1.22–1.38) | 1.05 (1.00–1.11) | 1.00 | 0.90 (0.82–0.98) | 1.08 (0.93–1.26) | 1.48 (1.16–1.87) |
| Multivariable-adjusted HR (95% CI) | 0.99 (0.90–1.10) | 0.95 (0.88–1.01) | 1.00 | 0.95 (0.86–1.05) | 1.15 (0.97–1.37) | 1.56 (1.18–2.06) |
| Stroke cases diagnosed only in inpatient and emergency encounters | ||||||
| No. patients | 10,459 | 21,943 | 16,703 | 6567 | 1740 | 584 |
| No. cases | 366 | 522 | 340 | 135 | 42 | 25 |
| Person-years | 32,725 | 72,917 | 53,224 | 19,477 | 4810 | 1478 |
| Incidence rate | 11.18 | 7.16 | 6.39 | 6.93 | 8.73 | 16.91 |
| Age-adjusted HR (95% CI) | 1.42 (1.22–1.64) | 1.00 (0.87–1.14) | 1.00 | 1.14 (0.94–1.40) | 1.49 (1.08–2.05) | 2.91 (1.94–4.37) |
| Multivariable-adjusted HR (95% CI) | 0.89 (0.69–1.13) | 0.84 (0.71–0.99) | 1.00 | 1.21 (0.96–1.51) | 1.77 (1.21–2.59) | 3.40 (1.99–5.82) |
| Difference in age-adjusted HR (95% CI) | 1.09 (0.93–1.29) | 0.95 (0.82–1.10) | — | 1.27 (1.02–1.59)* | 1.38 (0.96–1.96) | 1.97 (1.23–3.16)* |
| Difference in multivariable-adjusted HR (95% CI) | 0.89 (0.69–1.17) | 0.88 (0.74–1.06) | — | 1.27 (1.00–1.63) | 1.54 (1.02–2.35)* | 2.19 (1.19–4.01)* |
The unit of incidence rate is the number of new cases per 1000 person-years.
P < 0.05.
CI indicates confidence interval; HR, hazard ratio; LDL-low-density lipoprotein.
DISCUSSION
This study assessed whether incidence rates of CHD and stroke differed across health care encounter settings due to differences in defining CHD and stroke for patients with T2DM. We found significant differences in incidence rates of CHD and stroke across different health care settings, with incidence rates of CHD and stroke in all health care settings being much greater than incidence rates of CHD and stroke in IP/ED settings.
The larger incidence rates of CHD and stroke and the number of patients with CHD and stroke in all health care settings are expected since patients with diabetes usually need regular screenings for macrovascular complications such as cardiovascular disease and stroke in outpatient settings.39 Clinically, established complications such as retinopathy and nephropathy would likely need ongoing, consistent monitoring, which may contribute to the significantly higher complication rates and the number of patients with complications.39
Distinguishing the incidence rates of diabetes complications by encounter types may provide valuable information to health care providers and payers who oversee and strategize on ways to allocate health care resources. Previous studies have concluded that the greatest share of costs are due to IP care for diabetes-related complication management.8 The acute complication events often occur in IP/ED settings and cause marked increase in costs, which might be more resource-intensive than events that are identified from encounters in outpatient settings because those recorded diagnoses in more urgent encounters are usually linked with specific treatment procedures and hospital admissions.
This study examines the association between CHD and stroke and LDL cholesterol and HbA1c levels among T2DM patients. Regardless of how incident CHD and stroke events were defined based on different health care settings, this study found U-shape trends of HRs for CHD and stroke across different levels of LDL cholesterol and HbA1c after adjusting for some confounding factors, respectively. The U-shaped patterns have been found in other studies on HbA1c level and mortality, HbA1c, and cardiovascular events, as well as LDL and mortality.40–43 In addition, the significant difference of multivariable-adjusted HRs was smaller than the corresponding level of difference of age-adjusted HRs, which indicates that the confounders may have made a great contribution to the HR differences. Although a high level of HbA1c was a risk factor to cardiovascular events, patients with HbA1C ≤ 6% were found to be more likely to experience a cardiovascular event and mortality of all causes than the group with HbA1C of > 6%–8%.40,41,43 Moreover, differing associations have been found between LDL cholesterol levels and the risk of CHD or stroke. In 1981, Gordon and colleagues’ study found an association between low LDL cholesterol and risk of stroke, Hlatky and Hulley42 also stated the U-shape pattern between LDL and incidence of stroke, however, Hlatky and Hulley42 provided other possible explanations for the association, such as it is due to chance or bias.35 Current clinical guidelines and previous studies support that lowering LDL cholesterol levels would benefit diabetes patients and reduce the risk of CHD events.44,45 However, Rist et al46 found that LDL cholesterol <70 mg/dL was associated with stroke among women. Our study also found the U-shape patterns and the association between low LDL cholesterol with the risk of CHD and stroke using real-world EHR data. Future studies could be benefit from analyzing the association between low LDL and incidence of CHD or stroke.
This study has several limitations. First, the present study only included CHD and stroke as outcomes, however, future studies would benefit from including other microvascular complications, such as retinopathy, nephropathy, or neuropathy among the T2DM population. Secondly, the incidence rate of CHD and stroke only includes health records from 2013 to 2017 in 2 large health systems in Southeast Louisiana, with the majority of patients being from one health system. Thus, our study may have limited generalizability due to the limited years of data and limited health systems included in the study. Even if a patient consistently used health care services in the 2 health systems included in our study, a patient may have a complication history before 2013, which is not accounted for in our measurement of incidence. Thus, the new complication cases identified in this study may not be the true incidence as defined in epidemiology. Furthermore, a patient may have a complication history in another health system that was not included in the study, which also is not accounted for in our measurement of incidence. Second, our data source does not include services that patients received in other health systems. The dates of T2DM diagnoses and the first event of diabetes-related CHD and stroke in the present study are derived from the current EHRs. Epidemiology study using national health survey data may include valid diabetes complication histories and measure the incidence of diabetes complications more accurately. For instance, Gregg et al47 found a declining trend in diabetes complications regarding stroke using national health survey data. The incidence rates of acute myocardial infarction in stroke was 5.3 cases per 1000 person-years in 2010. Despite the decreasing trend of stroke, our result using EHRs based on IP/ED encounters was still higher than the Gregg et al47 epidemiology findings, which may imply that the incidence of stroke could be overestimated using EHRs based on IP/ED encounters, let alone based on all health care settings. Future studies would benefit from using a chart review as the best validation for comparison, which would provide more accurate and specific complication histories and complication incidences among the T2DM population.
In conclusion, most T2DM patients with CHD and stroke had recorded diagnoses outside IP/ED settings, that in turn lead to significant differences in incidence rates of CHD and stroke when comparing IP/ED settings versus all health care settings. The preliminary findings on the association between both CHD and stroke and risk factors suggest that the relative risks in HRs of CHD and stroke among T2DM patients are similar when defining the CHD and stroke based on IP/ED versus all health care settings. Whether studies should consider the health care settings that the data come from depends on the study purposes as they relate to diabetes complications.
Supplementary Material
ACKNOWLEDGMENTS
The authors acknowledge the contributions of our partners in this study. The success of this study depended on their ongoing support and expertise. These partners include Ochsner Health System and the Ochsner Patient Research Advisory Board; Tulane Medical Center; University Medical Center New Orleans; Research Action for Health Network (REACHnet a PCORnet Clinical Research Network) and their multi-stakeholder Diabetes Advisory Groups; Pennington Biomedical Research Center; Blue Cross and Blue Shield of Louisiana; and our patient and community partners Patricia Dominick, Cathy Glover, and Peggy Malone.
Supported through a Patient-Centered Outcomes Research Institute (PCORI) Program Award (NEN-1508-32257).
All statements in this manuscript, including findings and conclusions, are solely those of the authors and do not necessarily represent the views of PCORI, its Board of Governors or the Methodology Committee.
Footnotes
The authors declare no conflict of interest.
Supplemental Digital Content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal’s website, www.lww-medicalcare.com.
REFERENCES
- 1.Fowler MJ. Microvascular and macrovascular complications of diabetes. Clin Diabetes. 2008;26:77–82. [Google Scholar]
- 2.Roglic G. WHO Global report on diabetes: a summary. Int J Noncommunicable Dis. 2016;1:3. [Google Scholar]
- 3.Stratton IM, Adler AI, Neil HA, et al. Association of glycaemia with macrovascular and microvascular complications of type 2 diabetes (UKPDS 35): prospective observational study. BMJ. 2000;321:405–412. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Alberti KG, Zimmet PF. Definition, diagnosis and classification of diabetes mellitus and its complications. Part 1: diagnosis and classification of diabetes mellitus. Provisional report of a WHO consultation. Diabet Med. 1998;15:539–553. [DOI] [PubMed] [Google Scholar]
- 5.Nathan DM. Long-term complications of diabetes mellitus. N Engl J Med. 1993;328:1676–1685. [DOI] [PubMed] [Google Scholar]
- 6.Deshpande AD, Harris-Hayes M, Schootman M. Epidemiology of diabetes and diabetes-related complications. Phys Ther. 2008;88:1254–1264. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Hong D, Si L, Jiang M, et al. Cost effectiveness of sodium-glucose cotransporter-2 (SGLT2) inhibitors, glucagon-like peptide-1 (GLP-1) receptor agonists, and dipeptidyl peptidase-4 (DPP-4) inhibitors: a systematic review. Pharmacoeconomics. 2019;37:777–818. [DOI] [PubMed] [Google Scholar]
- 8.Association AD. Economic costs of diabetes in the US in 2017. Diabetes Care. 2018;41:917–928. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Bommer C, Sagalova V, Heesemann E, et al. Global economic burden of diabetes in adults: projections from 2015 to 2030. Diabetes Care. 2018; 41:963–970. [DOI] [PubMed] [Google Scholar]
- 10.Riddle MC, Herman WH. The cost of diabetes care—an elephant in the room. Diabetes Care. 2018;41:929–932. [DOI] [PubMed] [Google Scholar]
- 11.Evans JA. Electronic medical records system. Google Patents; 1999.
- 12.Cox E, Martin BC, Van Staa T, et al. Good research practices for comparative effectiveness research: approaches to mitigate bias and confounding in the design of nonrandomized studies of treatment effects using secondary data sources: the International Society for Pharmacoeconomics and Outcomes Research Good Research Practices for Retrospective Database Analysis Task Force Report—Part II. Value Health. 2009;12:1053–1061. [DOI] [PubMed] [Google Scholar]
- 13.Häyrinen K, Saranto K, Nykänen P. Definition, structure, content, use and impacts of electronic health records: a review of the research literature. Int J Med Inform. 2008;77:291–304. [DOI] [PubMed] [Google Scholar]
- 14.Shen Y, Shi L, Nauman E, et al. Race and sex differences in rates of diabetic complications. J Diabetes. 2019;11:449–456. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Wharam JF, Zhang F, Eggleston EM, et al. Diabetes outpatient care and acute complications before and after high-deductible insurance enrollment: a Natural Experiment for Translation in Diabetes (NEXT-D) study. JAMA Intern Med. 2017;177:358–368. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Zhao Y, Campbell CR, Fonseca V, et al. Impact of hypoglycemia associated with antihyperglycemic medications on vascular risks in veterans with type 2 diabetes. Diabetes Care. 2012;35:1126–1132. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Shen Y, Shi L, Nauman E, et al. Inverse association between HDL (high-density lipoprotein) cholesterol and stroke risk among patients with type 2 diabetes mellitus. Stroke. 2019;50:291–297. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Pelletier EM, Shim B, Ben-Joseph R, et al. Economic outcomes associated with microvascular complications of type 2 diabetes mellitus. Pharmacoeconomics. 2009;27:479–490. [DOI] [PubMed] [Google Scholar]
- 19.Shi L, Ye X, Lu M, et al. Glycemic and cholesterol control versus singlegoal control in US veterans with newly diagnosed type 2 diabetes: a retrospective observational study. Diabetes Ther. 2015;6:339–355. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Schwab P, Saundankar V, Bouchard J, et al. Early treatment revisions by addition or switch for type 2 diabetes: impact on glycemic control, diabetic complications, and healthcare costs. BMJ Open Diab Res Care. 2016;4:e000099. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Chen Y-L, Weng S-F, Yang C-Y, et al. Long-term risk of stroke in type 2 diabetes patients with diabetic ketoacidosis: a population-based, propensity score-matched, longitudinal follow-up study. Diabetes Metab. 2017;43:223–228. [DOI] [PubMed] [Google Scholar]
- 22.Lai Y-J, Hu H-Y, Lee Y-L, et al. A retrospective cohort study on the risk of stroke in relation to a priori health knowledge level among people with type 2 diabetes mellitus in Taiwan. BMC Cardiovasc Disord. 2017;17:130. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Murphree DH, Arabmakki E, Ngufor C, et al. Stacked classifiers for individualized prediction of glycemic control following initiation of metformin therapy in type 2 diabetes. Comput Biol Med. 2018;103:109–115. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Pantalone KM, Misra-Hebert AD, Hobbs TM, et al. Effect of glycemic control on the Diabetes Complications Severity Index score and development of complications in people with newly diagnosed type 2 diabetes. J Diabetes. 2018;10:192–199. [DOI] [PubMed] [Google Scholar]
- 25.Shi Q, Liu S, Krousel-Wood M, et al. Long-term outcomes associated with triple-goal achievement in patients with type 2 diabetes mellitus (T2DM). Diabetes Res Clin Pract. 2018;140:45–54. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Zghebi SS, Rutter MK, Ashcroft DM, et al. Using electronic health records to quantify and stratify the severity of type 2 diabetes in primary care in England: rationale and cohort study design. BMJ Open. 2018;8:e020926. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Brixner D, Ermakova A, Xiong Y, et al. Clinical and economic outcomes of patients with type 2 diabetes on multiple daily injections of basal-bolus insulin (MDI) therapy: a retrospective cohort study. Clin Ther. 2019;41:303.e1–313.e1. [DOI] [PubMed] [Google Scholar]
- 28.Chapman D, Foxcroft R, Dale-Harris L, et al. Insights for care: the healthcare utilisation and cost impact of managing type 2 diabetes-associated microvascular complications. Diabetes Ther. 2019;10:575–585. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Lee DC. Age disparities among patients with type 2 diabetes and associated rates of hospital use and diabetic complications. Prev Chronic Dis. 2019;16:E101. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Winhusen T, Theobald J, Kaelber DC, et al. Medical complications associated with substance use disorders in patients with type 2 diabetes and hypertension: electronic health record findings. Addiction. 2019;114:1462–1470. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Wu H, Eggleston KN, Zhong J, et al. Direct medical cost of diabetes in rural China using electronic insurance claims data and diabetes management data. J Diabetes Investig. 2019;10:531–538. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Yoo H, Park I, Kim DJ, et al. Effects of sarpogrelate on microvascular complications with type 2 diabetes. Int J Clin Pharm. 2019;41: 563–573. [DOI] [PubMed] [Google Scholar]
- 33.Desai JR, Wu P, Nichols GA, et al. Diabetes and asthma case identification, validation, and representativeness when using electronic health data to construct registries for comparative effectiveness and epidemiologic research. Med Care. 2012;50:S30. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Navaneethan SD, Jolly SE, Schold JD, et al. Development and validation of an electronic health record–based chronic kidney disease registry. Clin J Am Soc Nephrol. 2011;6:40–49. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Fuller JH, Shipley MJ, Rose G, et al. Mortality from coronary heart disease and stroke in relation to degree of glycaemia: the Whitehall study. Br Med J (Clin Res Ed). 1983;287:867–870. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Pathak RD, Schroeder EB, Seaquist ER, et al. Severe hypoglycemia requiring medical intervention in a large cohort of adults with diabetes receiving care in US integrated health care delivery systems: 2005–2011. Diabetes Care. 2016;39:363–370. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Nichols GA, Desai J, Lafata JE, et al. Construction of a multisite DataLink using electronic health records for the identification, surveillance, prevention, and management of diabetes mellitus: the SUPREME-DM project. Prev Chronic Dis. 2012;9:E110. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Grundy S Risk Assessment and guidelines for the management of high blood cholesterol. Endotext. MDText. com Inc; 2015. [Google Scholar]
- 39.Buse JB, Caprio S, Cefalu WT, et al. How do we define cure of diabetes? Diabetes Care. 2009;32:2133–2135. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.Shao H, Fonseca V, Stoecker C, et al. Novel risk engine for diabetes progression and mortality in USA: building, relating, assessing, and validating outcomes (BRAVO). Pharmacoeconomics. 2018;36: 1125–1134. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Currie CJ, Peters JR, Tynan A, et al. Survival as a function of HbA1c in people with type 2 diabetes: a retrospective cohort study. Lancet. 2010;375:481–489. [DOI] [PubMed] [Google Scholar]
- 42.Hlatky MA, Hulley SB. Plasma cholesterol: can it be too low? Arch Intern Med. 1981;141:1132. [DOI] [PubMed] [Google Scholar]
- 43.Colayco DC, Niu F, McCombs JS, et al. A1C and cardiovascular outcomes in type 2 diabetes: a nested case-control study. Diabetes Care. 2011;34:77–83. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44.Sacks FM, Tonkin AM, Craven T, et al. Coronary heart disease in patients with low LDL-cholesterol: benefit of pravastatin in diabetics and enhanced role for HDL-cholesterol and triglycerides as risk factors. Circulation. 2002;105:1424–1428. [DOI] [PubMed] [Google Scholar]
- 45.Grundy SM, Stone NJ, Bailey AL, et al. 2018 AHA/ACC/AACVPR/ AAPA/ABC/ACPM/ADA/AGS/APhA/ASPC/NLA/PCNA guideline on the management of blood cholesterol: a report of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines. J Am Coll Cardiol. 2019;73:e285–e350. [DOI] [PubMed] [Google Scholar]
- 46.Rist PM, Buring JE, Ridker PM, et al. Lipid levels and the risk of hemorrhagic stroke among women. Neurology. 2019;92: e2286–e2294. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47.Gregg EW, Li Y, Wang J, et al. Changes in diabetes-related complications in the United States, 1990–2010. N Engl J Med. 2014;370:1514–1523. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.