Evolution of Clinical Complexity, Treatment Burden, Health Care Use, and Diabetes-Related Outcomes Among Commercial and Medicare Advantage Plan Beneficiaries With Diabetes in the U.S., 2006–2018

Abstract

OBJECTIVE

To characterize trends in clinical complexity, treatment burden, health care use, and diabetes-related outcomes among adults with diabetes.

RESEARCH DESIGN AND METHODS

We used a nationwide claims database to identify enrollees in commercial and Medicare Advantage plans who met claims criteria for diabetes between 1 January 2006 and 31 March 2019 and to quantify annual trends in clinical complexity (e.g., active health conditions), treatment burden (e.g., medications), health care use (e.g., ambulatory, emergency department [ED], and hospital visits), and diabetes-related outcomes (e.g., hemoglobin A_1c [HbA_1c] levels) between 2006 and 2018.

RESULTS

Among 1,470,799 commercially insured patients, the proportion with ≥10 active health conditions increased from 33.3% (95% CI 33.1–33.4) in 2006 to 38.9% (38.8–39.1) in 2018 (P = 0.001) and the proportion taking three or more glucose-lowering medications increased from 11.6% (11.5–11.7) to 23.1% (22.9–23.2) (P = 0.007). The proportion with HbA_1c ≥8.0% (≥64 mmol/mol) increased from 28.0% (27.7–28.3) in 2006 to 30.5% (30.2–30.7) in 2015, decreasing to 27.8% (27.5–28.0) in 2018 (overall trend P = 0.04). Number of ambulatory visits per patient per year decreased from 6.86 (6.84–6.88) to 6.19 (6.17–6.21), (P = 0.001) while ED visits increased from 0.26 (0.257–0.263) to 0.29 (0.287–0.293) (P = 0.001). Among 1,311,903 Medicare Advantage enrollees, the proportion with ≥10 active conditions increased from 51.6% (51.2–52.0) to 65.1% (65.0–65.2) (P < 0.001); the proportion taking three or more glucose-lowering medications was stable at 16.6% (16.3–16.9) and 18.1% (18.0–18.2) (P = 0.98), and the proportion with HbA_1c ≥8.0% increased from 17.4% (16.7–18.1) to 18.6% (18.4–18.7) (P = 0.008). Ambulatory visits per patient per year remained stable at 8.01 (7.96–8.06) and 8.17 (8.16–8.19) (P = 0.23), but ED visits increased from 0.41 (0.40–0.42) to 0.66 (0.66–0.66) (P < 0.001).

CONCLUSIONS

Among patients with diabetes, clinical complexity and treatment burden have increased over time. ED utilization has also increased, and patients may be using ED services for low-acuity conditions.

Introduction

Diabetes is one of the most common and costly chronic diseases in the U.S., affecting 37.1 million adults or 14.7% of the U.S. adult population (1). It negatively impacts patient health and quality of life, is among the leading causes of death, and puts financial and logistical strain on the health care system (1–5). Diabetes is also an important driver of health care spending, as direct medical costs of diabetes total more than $2.4 billion annually, accounting for 17% of U.S. health care expenditures (6). Many people living with diabetes struggle with high medication costs (7,8), polypharmacy (9,10), and complexity of medical care (11). To improve the health outcomes of people living with diabetes and reduce the costs of their care, payors—including insurance companies, employers, and Accountable Care Organizations (ACOs)—have developed targeted care management programs focused on patients at highest risk for high health care use and poor outcomes (12–15). Such programs require detailed understanding of both current and historic patterns of patients’ clinical complexity, pharmacotherapy use, health care use, and diabetes-related outcomes.

Currently, there is a need for more data regarding national trends in medical complexity and health care use among patients with diabetes. The Centers for Disease Control and Prevention United States Diabetes Surveillance System (USDSS) (16) provides valuable information about diabetes trends and is the most comprehensive publicly available source of national diabetes data. However, USDSS data lacks granular information on comorbidities, medication use, health care use, and other important markers of clinical complexity (16–18) that are needed to fully capture the burden of diabetes and its management. National surveys, including the National Health and Nutrition Examination Survey (NHANES), can similarly provide cross-sectional information on diabetes epidemiology and management (19,20) but lack granularity, completeness, and longitudinal information. As a result, there are gaps in understanding of how people with diabetes are managing their disease, what comorbidities they have, what complications they are experiencing, and how many health care services they are using over time.

To address these gaps, we use a national administrative claims database to examine two longitudinal cohorts of patients: one with commercial (predominantly employer-sponsored) insurance and one with Medicare Advantage insurance. We focus specifically on trends in clinical complexity, treatment burden, health care use, and diabetes-related outcomes. These trends offer important insights into a population that is a major driver of health care use and spending, and this information can guide private and public payors, ACOs, and integrated health care delivery systems as they invest in the analytic and human infrastructures necessary to support diabetes management programs and people living with diabetes.

Research Design and Methods

Study Design

We retrospectively analyzed de-identified administrative claims and linked laboratory data from OptumLabs Data Warehouse (OLDW), which contains longitudinal health information on enrollees of large U.S. commercial and Medicare Advantage health plans. It includes a diverse mixture of ages, ethnicities, and geographical regions (21). This study was exempt from review by the Mayo Clinic Institutional Review Board because it used de-identified data.

Study Population

We identified all adults (age ≥18 years) who met claims criteria for type 1 or type 2 diabetes between 1 January 2006 and 31 March 2019 (Supplementary Figs. 1 and 2). The diagnosis of diabetes was established with use of validated Healthcare Effectiveness Data and Information Set (HEDIS) criteria (22), which required that patients have one or more inpatient or two or more outpatient evaluation and management (E&M) visits with a diagnosis of diabetes on different days or a fill for one or more glucose-lowering medications other than metformin within a 12-month period. Once the date of confirmed diabetes was determined (if the patient met HEDIS criteria on the basis of two outpatient visits, the second of the two was counted as the HEDIS date), we anchored the enrollment around this index date and created a 365-day baseline period before the index date to ascertain diabetes incidence. Patients who had 365 days of enrollment prior to their index date (i.e., had a “clean period” without diabetes-related claims) were counted as “incident” cases, while patients who did not were counted as “prevalent” cases. Both incident and prevalent cases were included. Patients were then required to have 365 days of consecutive coverage following the index date on which they first met HEDIS criteria. Consecutive coverage was defined as enrollment in an included health insurance plan with no more than a 45-day gap.

Patients were censored on disenrollment from the health plan or 31 March 2019—whichever came first. Patients who were enrolled in an insurance plan for a portion of a calendar year were included in the cohort for that calendar year if they met all other inclusion criteria. Diabetes type was ascertained as previously described (23).

Outcomes

We examined outcomes representing four domains: clinical complexity, treatment burden, health care use, and diabetes-related outcomes.

Clinical Complexity

Measures of clinical complexity included patient age and the total number of health conditions that were actively managed in each calendar year. Diagnoses were ascertained with use of all ICD-9 and -10 codes in ambulatory, emergency department (ED), and hospital E&M claims. A condition was considered “actively managed” if the condition was billed for on an E&M claim in an ambulatory, ED, or hospital setting during that calendar year. For all diagnoses, ICD codes were mapped to their corresponding Clinical Classifications Software (CCS) categories, with ICD-9 codes mapping to CCS (24) and ICD-10 codes mapping to the beta version of the CCS (25) that was used for ICD-10 mapping prior to the development of the Clinical Classification Software Refined (CCSR) (as the cohort was built prior to CCSR availability). We assigned each ICD-9 and -10 code a comorbidity category derived from the CCS (24) and (if applicable) a diabetes complication category using the Diabetes Complications Severity Index (DCSI) (26,27).

Treatment Burden

Measures of treatment burden included the numbers of glucose-lowering, lipid-lowering, and cardiovascular medication classes filled per patient per year and the total number of medication classes (for any indication) filled per patient per year. We classified medications by therapeutic class, using a granular classification scheme for glucose-lowering, lipid-lowering, and cardiovascular medications (Supplementary Table 1) and the generic therapeutic class designation for all other medications. Each medication was assigned to only one class.

Health Care Use

Measures of health care use were the numbers of ambulatory, ED, and hospital visits per patient per year. ED visits were defined as visits to an ED that did not result in an inpatient admission or observation. Hospital visits included both inpatient admission and observation stays.

Diabetes-Related Outcomes

Measures of diabetes-related outcomes were 1) number of actively managed complications of diabetes, categorized according to the DCSI (26,27) as detailed in Supplementary Table 2; 2) incidence of severe hypoglycemia and severe hyperglycemia, as defined in Supplementary Table 2; and 3) mean annual hemoglobin A_1c (HbA_1c) level. If a patient had multiple HbA_1c tests in a calendar year, the patient’s average HbA_1c level for that calendar year was used in calculating cohort-level metrics. Laboratory results are available for a subset of individuals within OLDW based on data sharing agreements between OptumLabs and commercial laboratories.

Independent Variables

Patient age, sex, race/ethnicity, annual household income, U.S. region of residency, and health plan (commercial vs. Medicare Advantage) were ascertained from enrollment files as of the index date.

Statistical Analysis

Descriptive data were summarized as means (SD) or counts (percentages), as appropriate, and 95% CIs were calculated. Unadjusted trend analysis was completed with the Cuzick test (28) to determine trend over calendar years between 2006 and 2018. SAS Enterprise Guide, version 7.13 (SAS Institute, Cary, NC), Stata, version 16 (StataCorp, College Station, TX), and Microsoft Excel, version 2102 (Microsoft, Redmond, WA), were used for data management and analysis.

Results

We identified 1,470,799 patients with diabetes (5,005,963 person-years of observation) in the commercial plan cohort and 1,311,903 patients (4,506,300 person-years) in the Medicare Advantage cohort. All patients had a minimum of 1 year of follow up. Median duration of follow-up was 2.57 years (interquartile range 1.68–4.33) in the commercial plan cohort and 2.82 years (1.83–4.32) in the Medicare Advantage cohort, and mean (SD) duration of follow-up was 3.41 (2.44) and 3.44 (2.29) years, respectively. For the commercial plan cohort, mean age was 51.3 (10.0) years, 44.6% were female, 65.1% were non-Hispanic White, and 14.8% had annual household income <$40,000. For the Medicare Advantage cohort, mean age was 69.8 (8.7) years, 54.2% were female, 59.1% were non-Hispanic White, and 40.7% had annual household income <$40,000. Additional demographics details are available in Table 1.

Table 1.

Cohort characteristics

	Commercial plan	Medicare Advantage plan
Cohort size (N)	1,470,799	1,311,903
Age, years
Mean (SD)	51.3 (10.0)	69.8 (8.7)
18–44	355,088 (24.1)	16,252 (1.2)
45–64	1,040,266 (70.7)	220,283 (16.8)
65–74	71,060 (4.8)	665,303 (50.7)
≥75	4,386 (0.3)	410,065 (31.3)
Sex
Female	656,368 (44.6)	710,646 (54.2)
Male	814,431 (55.4)	601,257 (45.8)
U.S. region
Midwest	382,903 (26.0)	294,334 (22.4)
Northeast	123,180 (8.4)	233,162 (17.8)
South	765,726 (52.1)	668,961 (51.0)
West	196,745 (13.4)	115,203 (8.8)
Other/unknown	2,245 (0.2)	243 (0.02)
Annual household income
<$40,000	21,7965 (14.8)	533,571 (40.7)
$40,000–$74,999	375,012 (25.5)	357,721 (27.3)
$75,000–$124,999	381,386 (25.9)	195,222 (14.9)
$125,000–$199,999	172,462 (11.7)	51,040 (3.9)
≥$200,000	172,462 (11.7)	17,435 (1.3)
Other/unknown	229,950 (15.6)	156,914 (12.0)
Race/ethnicity
White	957,498 (65.1)	774,677 (59.1)
Black	203,106 (13.8)	255,385 (19.5)
Hispanic	191,715 (13.0)	159,095 (12.1)
Asian	63,162 (4.3)	45,452 (3.5)
Other/unknown	55,318 (3.8)	77,294 (5.9)
Diabetes type
Type 1	91,882 (6.3)	41,069 (3.1)
Type 2	137,8917 (93.8)	1,270,834 (96.9)
Diabetes onset
Incident	32,932 (7.6)	17,710 (4.6)
Prevalent	406,103 (93.8)	363,816 (95.4)

Data are reported based on patient characteristics on their index date and are reported as counts (percentages) unless otherwise specified.

Clinical Complexity

Enrollees in both cohorts grew older and more clinically complex over time. As shown in Fig. 1, mean age of enrollees in the commercial cohort increased from 51.9 years (95% CI 51.9–52.0) in 2006 to 54.5 years (54.5–54.5) in 2018, and the proportion with ≥10 different categories of actively managed health conditions increased from 33.3% (33.1–33.4) to 38.9% (38.8–39.1) (P = 0.001). Mean number of different actively managed health conditions per patient increased from 8.59 (8.57–8.61) to 9.37 (9.35–9.39) (P = 0.002).

Figure 1.

Trends in complexity within the commercial and Medicare Advantage plan patient cohorts. All P values are for trends from 2006 to 2018. See Supplementary Tables 5 and 6 for 95% CIs around point estimates.

The mean age in the Medicare Advantage cohort increased from 70.1 years (95% CI 70.1–70.2) to 72.6 years (72.6–72.7) years (P < 0.001), and the percentage with ≥10 different categories of actively managed health conditions increased from 51.6% (51.2–52.0) to 65.1% (65.0–65.2) (P < 0.001). The mean number of categories of actively managed health conditions increased from 12.3 (12.2–12.4) to 15.2 (15.2–15.2) (P < 0.001).

Treatment Burden

Figure 2 summarizes changes in the pharmaceutical burden among patients with diabetes, with greater detail available in Supplementary Tables 3 and 4. Within the commercial cohort, the mean number of medication classes annually filled did not change significantly over time (5.15 [95% CI 5.14–5.16] to 7.18 [7.17–7.19], P = 0.07). The proportion of patients taking ≥10 unique medication classes increased from 12.6% (12.5–12.7) in 2006 to 24.8% (24.7–25.0) in 2009, but the proportion plateaued at this level (finishing at 26.9% [26.7–27.0] in 2018), and the overall trend was not statistically significant (P = 0.13). The proportion of patients taking three or more unique classes of glucose-lowering medications, however, increased from 11.6% (11.5–11.7) in 2006 to 23.1% (22.9–23.2) in 2018 (P = 0.007), and the mean number of glucose-lowering medication classes filled annually increased from 1.11 (1.11–1.11) to 1.59 (1.59–1.59) (P = 0.01). There was no increase in the proportion of patients taking three or more cardiovascular medication classes (20.4% [20.3–20.5] in 2006 vs. 32.6% [32.5–32.8] in 2018, P = 0.82) or in the mean number of classes taken (1.13 [1.13–1.13] vs. 1.77 [1.55–1.57], P = 0.31). The proportion of patients taking any lipid-lowering medication increased significantly (37.0% [36.9–37.2] in 2006 to 59.3% [59.2–59.5] in 2018, P = 0.001), but there was no change in the mean number of lipid-lowering medication classes taken over time (0.49 [0.49–0.49] vs. 0.67 [0.67–0.67], P = 0.81).

Figure 2.

Trends in pharmaceutical burden within the commercial and Medicare Advantage plan cohorts. All P values are for trends from 2006 to 2018. See Supplementary Tables 5 and 6 for 95% CIs around point estimates.

Medicare Advantage beneficiaries experienced a substantial and increasing pharmaceutical burden, with the percentage of patients taking ≥10 unique medication classes increasing from 37.6% (95% CI 37.3–38.0) in 2006 to 48.5% (48.4–48.6) in 2018 (P = 0.004). The mean number of medication classes filled annually increased from 8.31 (8.28–8.34) to 9.55 (9.54–9.56) (P = 0.004). There was no change in the proportion of patients taking three or more classes of glucose-lowering medications (16.6% [16.3–16.9] in 2006 to 18.1% [18.0–18.2] in 2018, P = 0.98) or in the mean number of glucose-lowering medication classes taken (1.49 [1.48–1.50] vs. 1.42 [1.42–1.42], P = 0.24). There was also no change in proportion of patients taking three or more classes of cardiovascular medications (54.2% [53.8–54.6] in 2006 to 53.9% [53.8–54.1] in 2018, P = 0.12) or in the mean number of cardiovascular medication classes filled (2.71 [2.69–2.73] vs. 2.79 [2.79–2.79], P = 0.70). There was, however, a significant increase in the proportion of patients taking any lipid-lowering medication (52.7% [52.3–53.1] in 2006 vs. 73.2% [73.1–73.3] in 2018, P < 0.001) and in the mean number of lipid-lowering medication classes filled (0.64 [0.64–0.65] to 0.83 [0.83–0.83]) (P = 0.002).

Health Care Use

For commercial enrollees, there was a significant decrease in annual ambulatory visits during the study period, decreasing from 6.86 visits per patient per year (95% CI 6.84–6.88) to 6.19 visits per patient per year (6.17–6.21) (P = 0.001). In contrast, the mean number of ED visits not resulting in hospitalization per patient per year increased from 0.26 (0.26–0.26) to 0.29 (0.29–0.29) (P = 0.001). The mean number of hospitalizations per patient per year decreased from 0.19 (0.19–0.19) to 0.11 (0.11–0.11) (P = 0.001).

For Medicare Advantage enrollees, there was no significant change in annual ambulatory care visits, with patients having an average of 8.01 ambulatory visits per year (95% CI 7.96–8.06) in 2006 compared with 8.17 (8.15–8.19) in 2018 (P = 0.23). However, similar to the commercially insured population, the mean number of ED visits not resulting in hospitalization per patient per year increased from 0.41 (0.40–0.42) to 0.66 (0.66–0.66), (P < 0.001). Simultaneously, mean hospitalizations per patient per year decreased from 0.41 (0.40–0.41) to 0.35 (0.34–0.35) (P = 0.002).

Diabetes-Related Outcomes

HbA_1c laboratory results were available for an average of 28.2% of commercial health plan enrollees in each given year. Among enrollees with available laboratory data, mean number of HbA_1c tests per enrollee per year ranged from 1.45 (2007) to 2.06 (2017). Mean annual HbA_1c was 7.49% (95% CI 7.49–7.49) or 58 mmol/mol (58.0–58.0). There was no change in the percentage of patients achieving HbA_1c <7.0% (<53 mmol/mol) (48.9% [48.6–49.2] in 2006 vs. 50.5% [50.2–50.8] in 2018, P = 0.13) or in mean HbA_1c level over time (7.47% [7.46–7.48] or 58 mmol/mol [57.9–58.1] in 2006 vs. 7.40% [7.39–7.41] or 57 mmol/mol [56.9–57.1] in 2018, P = 0.09) (Fig. 3). The percentage of patients with HbA_1c ≥8.0% (≥64 mmol/mol), however, increased from 28.0% (27.7–28.3) in 2006 to a peak of 30.5% (30.2–30.7) in 2015 and then declined back to 27.8% (27.5–28.0) in 2018. The overall trend showed a statistically significant increase in rates of poor control (P = 0.04).

Figure 3.

Trends in use and outcomes within the commercial and Medicare Advantage plan cohorts. All P values are for trends from 2006 to 2018. See Supplementary Tables 5 and 6 for 95% CIs around point estimates.

The number of actively managed diabetes complications remained stable over time. The proportion of patients with any active diabetes complication was 33.4% (95% CI 33.2–33.5) in 2006 and 32.8% (32.7–33.0) in 2018 (P = 0.08), and the mean number of complications active per year was 0.50 (0.50–0.50) and 0.49 (0.49–0.49), respectively (P = 0.14). Annual incidence of severe hyperglycemia was 1.12% (1.09–1.15) in 2006 and 0.82% (0.79–0.85) in 2018 (P = 0.10), and annual incidence of severe hypoglycemia was 2.88% (2.83–2.93) in 2006 and 1.72% (1.68–1.76) in 2018 (P = 0.11).

Results were similar for Medicare Advantage beneficiaries. HbA_1c laboratory results were available for an average of 27.7% of the cohort in each calendar year. Among enrollees with available laboratory data, mean HbA_1c tests per enrollee per year ranged from 1.49 (2007) to 2.59 (2016). Mean annual HbA_1c was 7.11% (95% CI 7.11–7.11) or 54 mmol/mol (54.0–54.0). Mean HbA_1c was 7.05% (7.03–7.07) or 54 mmol/mol (53.7–54.3) in 2006 and 7.00% (6.99–7.01%) or 53 mmol/mol (52.9–53.1) in 2018 (P = 0.15). There was no change in percentage of patients achieving HbA_1c <7.0% (58.2% [57.3–59.1] in 2006 vs. 60.7% [60.5–60.9] in 2018, P = 0.35). The proportion of patients with HbA_1c ≥8.0% increased, rising from 17.4% (16.7–18.1) in 2006 to 18.6% (18.4–18.7) in 2018 (P = 0.008) after peaking at 21.2% (21.0–21.4) in 2014.

The proportion of patients with at least one actively managed diabetes complication increased from 59.1% (95% CI 58.7–59.5) to 65.1% (65.0–65.2) (P = 0.001), and the mean number of complications per patient increased from 1.09 (1.08–1.10) in 2006 to 1.27 (1.27–1.27) in 2018 (P = 0.002). Annual incidence of severe hyperglycemia was 1.00% (0.92–1.08) in 2006 and 0.89% (0.87–0.91) in 2018 (P = 0.26), and annual incidence of severe hypoglycemia was 4.99% (4.82–5.16) in 2018 and 3.28% (3.24–3.32) in 2018 (P = 0.08).

Conclusions

Effective population health management for chronic conditions requires granular data on both historic and current metrics of patients’ clinical complexity, treatment burden, health care use, and health outcomes. Our analysis of 13 years of claims data of enrollees in U.S. commercial and Medicare Advantage health plans revealed that these populations have increased in clinical complexity, experience high pharmaceutical treatment burden, exhibit worsening glycemic control, have stable or increasing rates of diabetes complications, and experience increasing rates of potentially preventable ED visits in the backdrop of stable or declining use of ambulatory care.

Multimorbidity was very common. By 2018, commercially insured beneficiaries averaged 9.37 (95% CI 9.35–9.39) actively managed health conditions and Medicare Advantage beneficiaries averaged 15.2 (15.2–15.2). This clinical complexity increases treatment burden (29), may diminish patient’s capacity for self-management (30), and predisposes patients to treatment-related severe hypoglycemia (23). Alongside multimorbidity, polypharmacy was also prevalent, with commercially insured patients taking an average of 7.18 unique drug classes (7.17–7.19) in 2018 (compared with 5.15 [5.14–5.16] in 2006) and Medicare Advantage beneficiaries taking an average of 9.55 (9.54–9.56) (up from 8.31 [8.28–8.34] in 2006). Polypharmacy poses multiple risks to patients, particularly those who are older and clinically complex, and pharmaceutical treatment comes with medical, logistical, and financial burdens (7,8,31,32). It is therefore imperative for clinicians to individualize therapy to ensure evidence-based and goal-concordant care for patients, especially those with significant comorbidities and limited life expectancy (33–37). As the population of patients with diabetes grows older and more complex, clinicians should regularly reassess patients’ glycemic and other treatment targets and adjust regimens to align with the anticipated benefits and potential harms of treatment.

Medicare Advantage enrollees had tighter glycemic control than commercial plan patients (mean HbA_1c 7.11% [95% CI 7.11–7.11] or 54 mmol/mol [54.0–54.0] vs. 7.49% [7.49–7.49] or 58 mmol/mol [58.0–58.0], respectively), which is consistent with previous studies demonstrating higher rates of intensive control among older adults and undertreatment of younger adults (19,20,38). Younger adults often have significant daytime responsibilities (e.g., work, education, child care) that may make it difficult for them to access ambulatory care during normal office hours. For patients with elevated HbA_1c levels, these barriers may inhibit timely adjustment of pharmacotherapies. Previous research suggests that adults diagnosed with diabetes earlier in life tend to have worse diabetes-related outcomes and higher rates of complications than adults diagnosed later in life, which may create a survival effect among older cohorts of patients with diabetes (39). Importantly, both the commercial and Medicare Advantage plan populations saw increasing rates of poor glycemic control (defined as HbA_1c ≥8.0%) over the study period, though this trend may be showing signs of improvement after peaking in 2014–2015. For the commercial plan population, rates of poor control peaked at 30.5% (95% CI 30.2–30.7%) in 2015 and decreased in each subsequent year to 27.8% (27.5–28.0) in 2018; for Medicare Advantage patients, rates of poor control peaked at 21.2% (21.0–21.4) in 2014 and fell to 18.6% (18.4–18.7) by 2018. The timing of these declines in HbA_1c roughly correspond to observed increases in the proportion of patients filling at least three classes of glucose-lowering medications and in the mean total number of glucose-lowering medication classes filled, suggesting that higher rates of treatment intensification and better adherence to prescribed treatment regimens may explain these positive trends. Therapeutic inertia has been identified as a major cause of suboptimal glucose control in type 2 diabetes, and clinicians should ensure that they initiate prompt, guideline-based escalation of therapy when indicated (40–42).

Despite increasing multimorbidity and high pharmaceutical burden, ambulatory clinic visits remained stable among Medicare Advantage enrollees and decreased among commercial plan enrollees. Simultaneously, we observed increased ED visits not resulting in hospitalization in both groups, accompanied by a decrease in hospital stays. Taken together, these findings suggest that patients may be increasingly using ED services for low-acuity conditions that do not require hospitalization, rather than managing them in the ambulatory setting. It is unlikely that this trend (increase in ED visits not resulting in hospitalization) is due to increased overall ED use, since if this were the case, the average number of hospitalizations would be expected to increase as well. Difficulty accessing ambulatory care may contribute to this trend, as patients who cannot access ambulatory care may seek emergency care either for issues that are best managed in an outpatient setting (e.g., medication refills) or for conditions that could have been prevented with appropriate outpatient care (e.g., cellulitis for unrecognized lower-extremity ulcers). Further research is necessary to ascertain specific factors causing these trends and to identify strategies to support patients in seeking timely ambulatory care.

While this study is, to our knowledge, the largest contemporary longitudinal assessment of diabetes management among insured patients in the U.S., it has important limitations. Our findings may not generalize to patients who lack insurance or who have public health plans, and these patients may have even greater burdens of comorbidities, diabetes complications, and barriers to care. Laboratory data were only available for roughly one in four enrollees in a given calendar year, limiting the completeness of laboratory-based measures (HbA_1c). Nevertheless, OLDW is a robust source of longitudinal health care use and laboratory data for the included population. Administrative claims data lack clinical detail, and as a result we lack granular data on disease severity and likely underestimate the prevalence of undercoded conditions such as obesity, debility, and cognitive impairment. Additionally, administrative databases may miss drugs provided as free samples or purchased out of pocket through Low-Cost Generics Programs (43), further underestimating treatment burden. However, prior research has shown good agreement between electronic health record– and OLDW-derived data, particularly over 12-month periods (44). Finally, our data cannot account for individualized goals of care and the role that patient values and preferences played in treatment decisions.

Diabetes management is increasingly driven by chronic care models with recognition of the importance of clinical informatics, decision support, supported self-management, proactive patient engagement, and integrated multidisciplinary care (45–49). Many payors have recognized the need for effective, accessible, and affordable chronic disease management programs that can meet patients’ increasingly complex care needs (12–15). These programs use advanced analytics to identify patients at elevated risk of high health care use and poor outcomes, and they include interventions that support medication adherence, promote timely screening examinations, provide education, and offer resources that support important lifestyle changes (48–52). Such programs have been shown to improve diabetes management and glycemic control (50,53–55), and our findings underscore the urgent need for expansion of accessible disease management programs to provide increasingly complex care for patients. Payors and health systems should also strive to reduce barriers to ambulatory care, such as high cost sharing and requirements for in-person visits. Telehealth visits for diabetes may also help improve access to ambulatory care, particularly for younger patients who would benefit from the flexibility and efficiency.

Conclusion

This longitudinal cohort study of 1,470,799 commercially insured and 1,311,903 Medicare Advantage beneficiaries with diabetes is the largest contemporary study examining population trends in diabetes clinical complexity, treatment burden, health care use, and health outcomes. We find evidence that care for the population of patients with diabetes is growingly increasingly complex, with patients having high comorbidity and pharmaceutical burdens. Despite therapeutic advances, trends in glycemic control and complications are mixed. Strategies to promote medication adherence, increase appropriate use of outpatient services, and overcome therapeutic inertia will play important roles in the increasingly complex care of patients with diabetes.