Diabetes, gaps in care coordination, and preventable adverse events

Abstract

Objectives:

Compare the frequency of self-reported of gaps in care coordination and self-reported preventable adverse events among adults with versus without diabetes.

Study Design:

Cross-sectional study of REasons for Geographic And Racial Differences in Stroke (REGARDS) study participants ≥65 years of age who completed a survey on healthcare experiences in 2017–2018 (N=5,634).

Methods:

We analyzed the association of diabetes with self-reported gaps in care coordination and preventable adverse events. Gaps in care coordination were assessed using 8 validated questions. Four self-reported adverse events were studied (drug-drug interactions, repeat medical tests, emergency department visits, and hospitalizations). Respondents were asked if they thought these events could have been prevented with better communication among providers.

Results:

Overall, 1,724 (30.6%) participants had diabetes. Among participants with and without diabetes, 39.3% and 40.7% reported any gap in care coordination, respectively. The adjusted prevalence ratio (aPR) for any gap in care coordination for participants with versus without diabetes was 0.97 (95% confidence interval [95%CI]: 0.89–1.06). Any preventable adverse event was reported by 12.9% and 8.7% of participants with and without diabetes, respectively. The aPR for any preventable adverse event for participants with versus without diabetes was 1.22 (95%CI: 1.00–1.49). Among participants with and without diabetes, the aPRs for any preventable adverse event associated with any gap in care coordination were 1.53 (95%CI: 1.15–2.04) and 1.50 (95%CI: 1.21–1.88), respectively (p-value comparing aPRs: 0.922).

Conclusions:

Interventions to improve quality of care for diabetes patients could incorporate patient-reported gaps in care coordination to aide in preventing adverse events.

Precis:

We examined the association of diabetes with self-reported gaps in care coordination and self-reported preventable adverse events using data from a national sample of older adults.

Given their high burden of diabetes-related organ damage and chronic comorbid conditions^1–3, patients with diabetes may receive care from healthcare providers in different specialties.^{1, 4, 5} While receiving care from multiple healthcare providers may be clinically appropriate, patients’ health information is not always shared among providers.^{6, 7} When providers do not communicate with each other, they may not coordinate their evaluations or treatments, which could result in duplicate tests, drug-drug interactions, excess procedures, and avoidable emergency department visits and hospitalizations.^8–11

Patients are aware of gaps that may occur in their care coordination^{10, 12, 13} and of adverse healthcare-related events that they perceive could have been prevented with better care coordination, herein referred to as preventable adverse events.¹⁰ However, little is known about patient-reported gaps in care coordination or preventable adverse events among patients with diabetes. It is unclear whether reporting gaps in care coordination are associated with preventable adverse events among patients with diabetes. If patients with diabetes are more likely to report gaps in care coordination and preventable adverse events compared to those without diabetes, this would represent undesirable care processes which are potentially modifiable.

The aim of this study was to determine whether adults with diabetes are more likely to report gaps in care coordination or, separately, preventable adverse events, versus those without diabetes. We also determined whether self-reported gaps in care coordination are associated with self-reported preventable adverse events among adults with diabetes. To accomplish these aims, we analyzed data from the REasons for Geographic And Racial Differences in Stroke (REGARDS) study.

Methods

Study Population

REGARDS is a population-based cohort study that recruited 30,239 Black and White adults ≥45 years of age from the 48 contiguous US states and the District of Columbia between January 2003 and October 2007. Black adults and adults residing in the Southeastern US were oversampled by design.¹⁴ Participants completed a computer-assisted telephone interview (CATI) and an in-home study visit at baseline and in 2013–2016, and completed follow-up phone interviews every 6 months to identify potential stroke and myocardial infarction (MI) events that were adjudicated by experts based on medical records. Between August 2017 and November 2018, REGARDS study participants were invited to complete a survey on experiences with healthcare during one of their follow-up interviews. The Institutional Review Boards at all participating institutions approved the REGARDS study, and all participants provided written informed consent.

We analyzed data from REGARDS study participants who were active in the cohort at the time the survey on experiences with healthcare was administered and who agreed to complete the survey (eAppendix Figure 1). We restricted the analysis to participants who were ≥65 years of age at the time of the survey, who reported having ≥2 healthcare visits with ≥2 healthcare providers in the 12 months prior to the survey (because those with only 1 visit to 1 provider are not at risk for problems with care coordination), who reported having a regular healthcare provider, and who reported seeing that provider in the 6 months prior to the survey. We excluded participants who did not complete the second CATI and in-home study visit and who were missing data on self-reported diabetes or self-reported use of glucose-lowering medication. Lastly, among participants who did not meet our definition of diabetes as defined in the next section, we excluded participants who were missing data on blood glucose levels, had a fasting blood glucose that was ≥126 mg/dL, or whose non-fasting blood glucose was ≥200 mg/dL. These participants were excluded because they could have undiagnosed diabetes and may not have been receiving the same care for the condition as participants whose diabetes was diagnosed. In total, 5,634 participants were included in the current analysis.

Diabetes

Diabetes was defined by a self-reported diagnosis of diabetes, self-reported use of glucose-lowering medication, or identification of a glucose-lowering medication on a medication inventory conducted at the second REGARDS in-home study visit. Participants taking metformin without other glucose-lowering medication who did not report a prior diagnosis of diabetes were not considered to have diabetes as these participants could have been taking metformin for pre-diabetes or weight loss.^{15, 16} Participants not meeting the above definition were considered to not have diabetes.

Gaps in Care Coordination and Preventable Adverse Events

We analyzed seven gaps in care coordination (eAppendix Table 1) and four preventable adverse events (eAppendix Table 2) using data from the survey on experiences with healthcare.¹⁰ The seven gaps in care coordination were defined based on eight questions. Six of the questions assessed participants’ perceptions of the coordination of their care in the past 6 months (e.g., “In the last 6 months, when you visited your personal doctor for a scheduled appointment, how often did he or she have your medical records or other information about your care?” [Never, Sometimes, Usually, or Always]). Two questions assessed participants’ overall perception of communication between their healthcare providers (e.g., “In general, do you think the doctors that you see communicate with each other about your care?” [Yes, No, or I Don’t Know]). The four preventable adverse events included (1) A medical test that was repeated because the results of the first test, conducted previously, was not available, (2) A drug-drug interaction due to doctors prescribing medications that did not go well together, (3) An emergency department visit the participant felt would have been preventable with better care coordination, (4) A hospital admission the participant felt would have been preventable with better care coordination.

Potential Confounders

We analyzed data on potential confounders including age, annual household income, body mass index (BMI), hypertension, dyslipidemia, chronic kidney disease (CKD), atrial fibrillation, and peripheral artery disease, which were assessed at the second CATI and second in-home study visit. We also analyzed data on gender, race, educational attainment, and region of residence which were assessed at the REGARDS study baseline. History of MI and stroke were defined using data from the baseline and second CATIs and study visits supplemented by adjudicated outcome events. We analyzed data on the number of ambulatory visits and ambulatory providers that participants reported having in the 12 months prior to completing their survey on experiences with healthcare. Among participants with diabetes, we also analyzed data on the duration of diabetes and insulin use defined using information collected at the second CATI and in-home study visit. Finally, we analyzed data on the number of prescription medications taken in the 2 weeks prior to each participant’s second in-home study visit for an explanatory analysis. eAppendix Table 3 shows the definitions of variables mentioned above. eAppendix Figure 2 shows a schematic of the study design.

Statistical Analysis

We calculated summary statistics for participant characteristics, and the proportion of participants who reported gaps in care coordination and preventable adverse events, overall and separately for those with and without diabetes. Three Poisson regression models with robust variance estimators were used to calculate prevalence ratios (PRs) and 95% confidence intervals (CIs) for reporting ≥1 gap in care coordination among participants with versus without diabetes. Model 1 was unadjusted. Model 2 included adjustment for age, gender, race, educational attainment, annual household income, and region of residence. Model 3 included adjustment for variables in model 2 and the following clinical variables: BMI, hypertension, dyslipidemia, CKD, history of MI, history of stroke, atrial fibrillation, and peripheral artery disease. We also calculated the count of gaps in care coordination reported by participants. The distribution of the count of gaps was overdispersed with a large number of zero values, so we used marginalized zero-inflated Poisson regression models with adjustment as described above to calculate the ratio of the mean count among participants with versus without diabetes.¹⁷

We calculated PRs and 95% CI for reporting ≥1 preventable adverse event associated with diabetes using three Poisson regression models with robust variance estimators and adjustment for potential confounders as described above. Adults with diabetes may take more prescription medications than those without diabetes. In an explanatory model, we calculated the PR and 95% CI for reporting ≥1 preventable adverse event associated with diabetes, including adjustment for the potential confounders in model 3 and number of prescription medications being taken. We considered the number of prescription medications to be a potential variable in the causal pathway because one mechanism by which gaps in care coordination may cause harm is through lack of communication about prescribing, which can lead to drug-drug interactions.^{18, 19} Few participants had >1 preventable adverse event, so we did not compare the count of adverse events among those with versus without diabetes.

Among participants with and without diabetes, we calculated the PR and 95% CI for reporting ≥1 preventable adverse event associated with reporting ≥1 versus 0 gaps in care coordination. The models included adjustment for covariates as described above, a main effect for diabetes, and an interaction term between diabetes and gaps in care coordination, which was used to assess effect modification by diabetes status. In exploratory analyses, we compared the report of gaps in care coordination and preventable adverse events, and the association between gaps in care coordination and preventable adverse events among participants with diabetes using versus not using insulin.

Results

Participant Characteristics

The proportions of participants who were Black, who had an annual household income <$25,000, and who had comorbid conditions were higher in those with versus without diabetes, while the proportion who were college graduates was lower in participants with diabetes (Table 1). Compared to participants without diabetes, those with diabetes were taking more prescription medications. Participants with diabetes reported a higher number of health care providers and healthcare visits in the 12 months preceding the survey on experiences with healthcare compared to their counterparts without diabetes. Among participants with diabetes, the median duration of diabetes was 11 years and 28% were using insulin.

Table 1.

Characteristics of participants included in the current study, overall and among those with and without diabetes, separately.

Characteristic	Overall N = 5,634	Without diabetes N = 3,910	With diabetes N = 1,724	p-value
Age, y, mean ± SD	73.0 ± 6.6	73.2 ± 6.7	72.6 ± 6.3	0.002
Women, N (%)	3,109 (55.2)	2,166 (55.4)	943 (54.7)	0.642
Black, N (%)	1,919 (34.1)	1,085 (27.8)	834 (48.4)	<0.001
Annual income <$25,000, N (%)	1,076 (20.3)	652 (17.7)	424 (26.4)	<0.001
Educational attainment, N (%)				<0.001
Less than high school	321 (5.7)	183 (4.7)	138 (8.0)
High school	1,214 (21.6)	773 (19.8)	441 (25.6)
Some college	1,446 (25.7)	954 (24.4)	492 (28.5)
College graduate or above	2,653 (47.1)	2,000 (51.2)	653 (37.9)
Region of residence, N (%)				0.010
Stroke Bucklea	1,251 (22.2)	843 (21.6)	408 (23.7)
Stroke Beltb	1,827 (32.4)	1,241 (31.7)	586 (34.0)
Other region of the US	2,556 (45.4)	1,826 (46.7)	730 (42.3)
Body mass index, N (%)				<0.001
<25 kg/m2	1,324 (23.6)	1,122 (28.7)	202 (11.8)
25 kg/m2 to <30 kg/m2	2,090 (37.2)	1,510 (38.7)	580 (33.8)
≥30 kg/m2	2,206 (39.3)	1,273 (32.6)	933 (54.4)
Hypertension, N (%)	3,855 (68.4)	2,441 (62.5)	1,414 (82.0)	<0.001
Dyslipidemia, N (%)	3,470 (62.7)	2,231 (57.7)	1,239 (74.3)	<0.001
Chronic kidney disease, N (%)	2,148 (39.9)	1,379 (36.5)	769 (48.1)	<0.001
Myocardial infarction, N (%)	831 (15.2)	493 (13.0)	338 (20.5)	<0.001
Stroke, N (%)	333 (5.9)	187 (4.8)	146 (8.5)	<0.001
Atrial fibrillation, N (%)	609 (11.3)	413 (10.9)	196 (12.0)	0.241
Peripheral artery disease, N (%)	196 (3.5)	122 (3.1)	74 (4.3)	0.033
Number of ambulatory providers,c N (%)				<0.001
2 providers	1,790 (31.8)	1,313 (33.6)	477 (27.7)
3 providers	1,730 (30.7)	1,236 (31.6)	494 (28.6)
4 providers	1,042 (18.5)	690 (17.6)	352 (20.4)
≥ 5 providers	1,072 (19.0)	671 (17.2)	401 (23.3)
Number of ambulatory visits,c median (25th −75th percentiles)	5 (3 – 8)	5 (3 – 8)	6 (4 – 9)	<0.001
Duration of diabetes, years, median (25th −75th percentiles)	-	-	11 (6 – 20)	-
Insulin use, N (%)	-	-	353 (28.0)	-
Number of prescription medications, median (25th - 75th percentiles)	5 (3 – 8)	4 (2 – 6)	7 (5 – 10)	<0.001

^aThe stroke buckle includes the coastal plains region of North Carolina, South Carolina, and Georgia.

^bThe stroke belt includes the remaining parts of North Carolina, South Carolina and Georgia, and Alabama, Mississippi, Tennessee, Arkansas, and Louisiana.

^cIn the 12 months preceding the survey on experiences with healthcare.

Gaps in Care Coordination

eAppendix Table 4 shows the frequency of each of the 7 gaps in care coordination, overall and in those with and without diabetes, separately. Among participants with diabetes, 39.3% reported ≥1 gap in care coordination compared to 40.7% in those without diabetes (Table 2). In the fully adjusted model, diabetes was not associated with reporting ≥1 gap in care coordination (adjusted PR [aPR]: 0.97; 95% CI: 0.89–1.06). Diabetes was also not associated with the number of gaps in care coordination (Ratio: 0.93; 95% CI: 0.84–1.03; eAppendix Table 5).

Table 2.

Prevalence ratios for ≥1 self-reported gap in care coordination associated with diabetes status.

	Without diabetes N = 3,910	With diabetes N = 1,724
Participants with ≥1 gap in care coordination, N (%)	1,590 (40.7)	677 (39.3)
Prevalence ratio (95% CI)
Model 1	1 (Ref)	0.97 (0.90–1.04)
Model 2	1 (Ref)	0.94 (0.87–1.01)
Model 3	1 (Ref)	0.97 (0.89–1.06)

Model 1 is unadjusted.

Model 2 includes adjustment for age, gender, race, educational attainment, annual household income, and region of residence.

Model 3 includes adjustment for variables in Model 2 and body mass index, hypertension, dyslipidemia, chronic kidney disease, history of myocardial infarction, history of stroke, atrial fibrillation, and peripheral artery disease.

Preventable Adverse Events

Participants with diabetes were more likely to report a drug-drug interaction, any emergency department visit and any hospitalization versus those with diabetes (eAppendix Table 6). The proportion of participants who reported ≥1 preventable adverse event was 12.9% among those with diabetes and 8.7% in those without diabetes (Table 3). After multivariable adjustment for all potential confounders (Model 3), the prevalence of reporting ≥1 preventable adverse event was higher for participants with versus without diabetes (aPR [95% CI]: 1.22 [1.00–1.49]). After further adjustment for number of prescription medications (explanatory model), the aPR for reporting ≥1 preventable adverse event in adults with versus without diabetes was 1.10 (95% CI: 0.89–1.35).

Table 3.

Prevalence ratios for ≥1 self-reported preventable adverse event associated with diabetes status.

	Without diabetes N = 3,910	With diabetes N = 1,724
Participants with ≥1 preventable adverse event, N (%)	339 (8.7)	223 (12.9)
Prevalence ratio (95% CI)
Model 1	1 (Ref)	1.49 (1.27–1.75)
Model 2	1 (Ref)	1.32 (1.11–1.57)
Model 3	1 (Ref)	1.22 (1.00–1.49)
Explanatory model	1 (Ref)	1.10 (0.89–1.35)

Model 1 is unadjusted.

Model 2 includes adjustment for age, gender, race, educational attainment, annual household income, and region of residence.

Model 3 includes adjustment for variables in Model 2 and body mass index, hypertension, dyslipidemia, chronic kidney disease, history of myocardial infarction, history of stroke, atrial fibrillation, and peripheral artery disease.

The explanatory model includes adjustment for potential confounders in Model 3 and accounts for the number of prescription medications taken in the 2 weeks prior to each participant’s second in-home study visit.

Gaps in Care Coordination and Preventable Adverse Events

Among participants with diabetes, the proportion who reported ≥1 preventable adverse event was higher among those with ≥1 versus 0 gaps in care coordination (Table 4). The proportion of participants without diabetes reporting ≥1 preventable adverse event was also higher among those reporting ≥1 versus 0 gaps in care coordination. After multivariable adjustment, the aPRs for reporting ≥1 preventable adverse event associated with reporting ≥1 gap in care coordination was 1.53 (95% CI: 1.15–2.04) among participants with diabetes and 1.50 (95% CI: 1.21–1.88) in those without diabetes (p-value comparing PRs = 0.922).

Table 4.

Prevalence ratios for ≥1 preventable adverse event associated with reporting ≥1 versus 0 gaps in care coordination among participants with and without diabetes, separately.

	Without diabetes		With diabetes
	0 gaps in care coordination	≥1 gap in care coordination	0 gaps in care coordination	≥1 gap in care coordination	p-valuea
N	2,320	1,590	1,047	677
Participants with ≥1 preventable adverse event, N (%)	168 (7.2)	171 (10.8)	109 (10.4)	114 (16.8)
Prevalence ratio (95% CI)
Model 1	1 (Ref)	1.49 (1.21–1.82)	1 (Ref)	1.62 (1.27–2.06)	0.598
Model 2	1 (Ref)	1.49 (1.21–1.84)	1 (Ref)	1.48 (1.15–1.92)	0.976
Model 3	1 (Ref)	1.50 (1.21–1.88)	1 (Ref)	1.53 (1.15–2.04)	0.922

Model 1 is unadjusted.

Model 2 includes adjustment for age, gender, race, educational attainment, annual household income, and region of residence.

Model 3 includes adjustment for variables in Model 2 and body mass index, hypertension, dyslipidemia, chronic kidney disease, history of myocardial infarction, history of stroke, atrial fibrillation, and peripheral artery disease.

^aP-value comparing prevalence ratios by diabetes status.

Insulin, Gaps in Care Coordination, and Preventable Adverse Events

Among participants with diabetes, there was no association between insulin use and reporting a gap in care coordination (aPR 0.88; 95% CI: 0.73–1.05; eAppendix Table 7) or reporting a preventable adverse event (aPR 1.03; 95% CI: 0.72–1.46; eAppendix Table 8). In fully adjusted models, the aPR for reporting ≥1 preventable adverse event associated with ≥1 versus no gaps in care coordination was 2.65 (95% CI: 1.45–4.83) in participants using insulin and 1.41 (95% CI: 1.02–1.96) in those with diabetes not using insulin (p-value comparing PRs = 0.069; eAppendix Table 9).

Discussion

In the current national study of Black and White adults ≥65 years of age receiving care from multiple healthcare providers, there was no difference in self-reported gaps in care coordination by diabetes status. However, compared to those without diabetes, participants with diabetes were more likely to report an adverse event they perceived could have been prevented with better care coordination. This difference appears to be explained, at least in part, by participants with diabetes taking more prescription medications, a risk factor for drug-drug interactions, versus those without diabetes. Participants with diabetes who reported any versus no gap in care coordination were more likely to report any preventable adverse event. The association between reporting any gap in care coordination and reporting any preventable adverse event appeared to be stronger among participants with diabetes using versus not using insulin.

In the current study, 39% of patients with diabetes reported a gap in care coordination. This proportion is quite high considering that most of the questions used to assess gaps in care coordination were restricted to the past 6 months. Improving diabetes quality of care has been a major goal in healthcare for decades, but quality improvement efforts have largely focused on increasing patient engagement in recommended screening tests and controlling intermediate clinical outcomes such as blood glucose levels.^{20, 21} Patients’ experiences of care coordination are not often considered in quality measures for diabetes.²² The high proportion of patients with diabetes who report problems with care coordination suggests a need for quality improvement efforts in this area of care.

The current study adds to the existing literature by showing that patients with diabetes are aware of experiencing adverse events, and that they attribute many of those events to poor care coordination. The current study also suggests that participants with diabetes may be more likely to report a preventable adverse event, compared to their counterparts without diabetes, despite not experiencing more gaps in care coordination. Previous studies have shown that patients with diabetes take more medications²³ and have higher risk for any hospitalization versus those without diabetes.²⁴ Therefore, experiencing gaps in care coordination may be particularly hazardous to patients with diabetes. Consistently, the association of diabetes with higher report of preventable adverse events in the current analysis was attenuated and no longer statistically significant after adjustment for number of prescription medication being taken. This finding supports the inference that patients with diabetes on multiple medications need particular monitoring for gaps in communication among prescribers.

Using data from the REGARDS study, Kern et al. previously showed that adults who report a gap in care coordination are more likely to report preventable adverse events.¹⁰ The current study expands prior knowledge by showing that this association is present among adults with diabetes. The current study also suggests that the association between gaps in care coordination and preventable adverse events may be stronger among adults with diabetes taking versus not taking insulin. Patients using insulin have more diabetes-related complications than those with diabetes not using this medication.²⁵ Experiencing gaps in care coordination may have a more deleterious effect on the occurrence of preventable adverse events in adults with diabetes using versus not using insulin, a hypothesis that needs to be confirmed in future studies.

Previous quality improvement efforts for adults with diabetes have typically selected patients based on their hemoglobin A1c level or a recent hospitalization, which is appropriate, but may miss some opportunities for improvement.^26–28 A future intervention to improve care coordination for patients with diabetes might start with identifying those who perceive that their care is not well coordinated. Patient safety experts have shown that patients’ perceptions of their care often have merit.²⁹ Therefore, addressing patients’ concerns may help prevent future adverse events.

Strengths and Limitations

The current study has several strengths. The analysis included a large, national sample of patients with and without diabetes. Gaps in care coordination were assessed using previously validated questions. The current study also has several limitations. Diabetes was defined, in part, by self-report, which may have resulted in some misclassification. Gaps in care coordination and preventable adverse events were both defined using self-report of events occurring over long periods (i.e., 6 and 12 months). Thus, there is the potential for inaccurate recall. Additionally, some of the questions about gaps in care coordination referred to events occurring over the 6 months preceding the survey on experiences with healthcare, while questions about preventable adverse events included events occurring over the 12 preceding months. It is possible that preventable adverse events may have preceded the gaps in care coordination in some participants. Lastly, the outcomes in the current study were subjective. Participants may have understood the question about drug-drug interactions (i.e., medications not going well together) differently from one another. In addition, whether participants’ emergency department visits and hospitalizations could have been prevented with better coordinated care cannot be known for certain.

Conclusions

A high proportion of older adults with and without diabetes receiving care from multiple healthcare providers reported a problem with the coordination of their care. Adults with diabetes were more likely to report an adverse event that they attributed to poor care coordination compared to their counterparts without diabetes. The frequencies of these problems are quite high given decades of work to improve the quality of care for patients with diabetes. Whereas previous interventions to improve diabetes have typically identified patients for inclusion based on severity of illness or transitions in care, this work suggests that new interventions are needed, which would identify patients based on their experiences of care. Identifying and addressing patient-reported gaps in care coordination would be a novel strategy that may increase quality of care, patient satisfaction, and improve patient safety.

Take-Away Points:

In the current analysis of 5,634 adults ≥65 years old who completed a survey on their experiences with healthcare in 2017–2018:

• Gaps in care coordination were common among adults with and without diabetes.

• Participants with diabetes were more likely than those without diabetes to report an adverse event they felt was preventable with better care coordination.

• Reporting a gap in care coordination was associated with a higher likelihood of reporting a preventable adverse event among participants with and without diabetes, separately.

• Interventions to improve quality of care for diabetes could identify and address patient-reported gaps in care coordination to prevent adverse events.