Abstract
Background:
Care management has demonstrated improvements in quality of care for patients with complex care needs. The extent to which these interventions benefit race/ethnic minority populations is unclear.
Objectives:
To characterize race/ethnic differences in the longitudinal control of clinical outcomes for patients with complex care needs enrolled in Care Management Plus, a health information technology-enabled care coordination intervention.
Research Design:
Multilevel models of repeated observations from clinical encounters before and after program enrollment for six Oregon and California primary care clinics.
Subjects:
A total of 18,675 clinic patients were examined. We estimated multilevel models for 1,481 and 5,320 care-managed individuals with repeated hemoglobin A1c and blood pressure measurements, respectively.
Measures:
Primary outcomes were changes over time for two clinical markers of health status for complex care patients: (1) hemoglobin A1c for patients with diabetes, and (2) mid-BP (average systolic and diastolic blood pressure).
Results:
We found significant reductions in A1c for patients with previously uncontrolled A1c (pre-period slope, b=1.03 [0.83, 1.24]; post-period slope, b=−0.63 [−0.91, −0.35]). For mid-BP we found increasing unconditional pre-period trajectories (b=3.52 [2.39, 4.64]) and decreasing post-period trajectories (b=−5.21 [−5.70, −4.72]). We also found the trajectories of A1c and mid-BP were not statistically different for Black, Latino, and White patients.
Conclusions:
These analyses demonstrate some promising results for intermediate clinical outcomes for underrepresented patients with complex chronic care needs. It remains to be seen whether these health care system delivery redesigns yield long-term benefits for patients, such as improvements in function and quality of life.
Keywords: racial/ethnic differences, complex care patients, multimorbidity, hemoglobin A1c, blood pressure, multilevel modeling
INTRODUCTION
Chronic health conditions have emerged as the dominant health care burden in the US and abroad over the last century.1–3 Importantly, the frequency of multiple co-occurring chronic disease (multimorbidity) for middle-aged and older individuals is increasing,4 bringing to bear questions of how best to care for adults with complex care requirements and disrupt the early progression of chronic disease pathology to impairment, functional limitation, and disability.5–7 These trends dictate a need for improved health care system coordination to enable greater responsiveness and improved communication between patients and providers. Adults with multimorbidity and their caregivers face much greater challenges to manage often competing treatment plans and prescription medication changes from various providers and specialists.3,8–11
Accordingly, some of the most promising and successful chronic disease interventions involve multi-professional care teams to manage the complexities associated with multimorbidity and facilitate patient communication across providers and health care systems.8,12 Although there is substantial heterogeneity in care delivery programs designed to help patients with complex care needs, many interventions rely on several key components to streamline and improve the quality of care patients receive. These include developing care plans with patients, using care managers to coordinate multiple providers and prescription medications, and engaging patients in effective self-management of their chronic conditions.13
This study examines clinical outcomes for underrepresented racial/ethnic patient populations prior to and following the implementation of the Care Management Plus (CMP)14—a program that leverages the use of care managers, team-based care, and a health information technology platform to focus on patients with multiple co-occurring chronic conditions. CMP uses a team-based approach and employs a care manager as the central conduit for communication with patients, caregivers, and other providers. Care managers make use of information technology when examining electronic medical records to stratify patient based on risk, create and follow care plans, manage communication, and receive training in flexible protocols and patient engagement methods. Information technology tools enable care managers to document, track, and respond to patient needs by leveraging protocols and disease-specific guidelines to inform treatment decisions.
CMP allows for access to electronic health records, best practices via evidence-based treatment guidelines to devise patient worksheets, and efficient communication with patients, caregivers, and other specialists providing care. Physicians in participating primary care clinics were encouraged to refer any complex patient (multiple comorbidities, frailty) to CMP. Previous results have shown effectiveness of CMP for complex care populations. Specifically, complex care patients have fewer hospitalizations, lower costs, and overall improved health in clinics with CMP compared to similar populations in control clinics.15
Underrepresented racial and ethnic adults are at greatest risk of various poor health outcomes, including diabetic complications due to poor glycemic control. In addition, persons from minority racial and ethnic backgrounds have, on average, lower rates of receiving glycemic tests at recommended intervals.16 Intensive care management tools—like those available with CMP—may be effective in assisting individuals with vast health burdens by communicating clear, concrete chronic disease self-management goals. Actionable targets, such as regular clinical reminders for A1c tests as they become due, and regular contact with a more accessible care manager embedded within a primary care team may provide sufficient coordination of information within clinics to enable more successful outcomes for individuals with complex care needs.
Despite progress in health care delivery innovations, it is unclear whether benefits of care management programs accrue equitably to underrepresented racial and ethnic individuals. Indeed, disparities in health care for race/ethnic minorities have not abated since these were first documented.17,18 For the most vulnerable persons—the elderly, those of low socioeconomic status, low health literacy, or underrepresented racial and ethnic background groups—the presence of multiple chronic diseases have the potential to complicate treatment decisions, delay health-seeking behavior, and hasten poor health outcomes.18–20
There are several gaps in knowledge about care coordination and health outcomes for racial and ethnic minorities with complex care needs. Previous studies have focused on the broad delivery of care coordination programs. Less focus has been paid to examining outcomes for racial and ethnic individuals enrolled in care management programs. In particular, for chronic conditions that disproportionately affect racial and ethnic minority populations—such as diabetes and cardiovascular disease21,22—there is less emphasis placed on tracking efforts to manage these conditions for these populations. Further, the potential for these innovative care delivery platforms to reduce racial and ethnic disparities in health care20,23 has not been assessed empirically.
Few studies have examined the performance of clinical interventions for various racial and ethnic groups in concert. Moving beyond examining one or two racial groups not only advances our understanding of the applicability of intensive care management programs to individuals from various backgrounds, but it also yields important information regarding relative changes in outcomes for underrepresented background groups compared to White patient populations. This study seeks to provide insight in these areas by estimating trends in clinical outcomes resulting from the implementation of CMP for enrolled non-Latino White, non-Latino Black, and Latino individuals with complex and chronic health conditions.
METHODS
Sample
Complex care patients were identified using an algorithm based on diagnosis lists from six clinics in Oregon and Northern California who participated in the CMP intervention. The algorithm is based on the patient’s comorbidity score, problem list, or key combinations of chronic diseases that were identified a priori to include patients with a broad range of chronic illnesses. For example, older patients with frailty, or patients with multiple co-occurring diseases were identified for care management. Additional details of the CMP program are provided elsewhere.15
Our study consisted of a total patient population of 18,675. In order to estimate trajectories of biophysical measures, at least three repeated observations are necessary.24 There were 1,481 patients with sufficient number of HbA1c measurements for longitudinal analysis. In addition, we analyzed 5,320 individuals with sufficient number of repeated blood pressure measurements to estimate trends of mid-BP. We analyzed clinical databases for one year prior and one year post CMP implementation in clinics, from 2009 to 2011. We excluded patients from analyses who did not meet criteria for complex care, as well as patients who did not have a clinical encounter before and after implementation of CMP in their clinic. For patients who had multiple measurements of clinical outcomes for the same encounter date, we included the averaged value. We were also sensitive to the ramping-up period of the CMP intervention in clinics, since it is highly unlikely that clinic operations would transform from one day to the next. Consequently, we defined post-intervention clinical outcome measurements to occur after a six week run-in period. We linked CMP electronic health record data to medical claims records in order to have access to demographic and diagnostic information, procedure codes, and insurance status.
Measures
Primary outcome measures
We examined longitudinal changes in two primary outcome (continuous) measures: (1) hemoglobin A1c for patients with diabetes, and (2) mid-BP. For patients with diabetes, A1c tests are recommended every six months, and benchmarks for glycemic control target <6% or <6.5%. Randomized controlled trials establish an association between greater risk of diabetic complications and all-cause mortality for individuals with sustained A1c levels at or above 8%.23,25 Mid-BP was used in order to incorporate diastolic and systolic blood pressure measurements jointly.26
Additional covariates
We derived race and ethnicity from electronic health record data supplemented by claims records. Mutually-exclusive categories were constructed for Non-Latino White (White), Non-Latino Black (Black), and Latino individuals. We gave precedence to individuals who were classified as Latino in either the health record or claims data in the construction of the mutually-exclusive categories. This means that individuals were only classified as Black or White if they were identified as such in either clinic or claims databases, and were not identified in these databases as Latino.
Because we anticipate socio-demographic factors are associated with health status and the control of biophysical indicators of health, we include important demographic and socioeconomic covariates in the adjusted analyses: female gender (binary), age (continuous), and Medicaid insurance coverage (binary) to provide a proxy measure for low socioeconomic status. Documented Medicaid status is incomplete in the data due to clinic differences in documenting Medicaid coverage for patients seen in Oregon under the Oregon Medicaid program (Oregon Health Plan) and those seen in California under their state plan (Medi-Cal). Although we provide information on the percent of the study population that has documented Medicaid coverage in our database, all of the enrolled clinics in the study serve a majority low socioeconomic status population. An indicator variable for uncontrolled diabetes—high baseline A1c defined as ≥8% at the baseline—was used to examine the main effect of disease severity at the baseline. The interaction terms between high baseline A1c and race/ethnicity (binary) were used to examine differential effects for disease severity by race/ethnicity.
Analyses
We estimated unconditional and adjusted multilevel models (i.e., growth curve trajectory models) of our two primary outcome variables, hemoglobin A1c and mean arterial blood pressure. Multilevel models allow for the simultaneous examination of individual-level and clinic-level factors in estimating clinical outcome trends.27,28 We estimate three-level multilevel models to investigate within-individual change in clinical outcomes over time, accounting for between-individual differences, and variation between clinic sites. We considered non-linear specifications of multilevel models, and chose linear models based on best fit by smallest Akaike Information Criterion (AIC).
Model 0 (M0) estimates trajectories for the unconditional multilevel model. Model 1 (M1) adjusts for the primary variables of interest, racial and ethnicity group. Model 2 (M2) traces the trajectories of each clinical outcome after adjusting for a determined set of covariates. We also explore interaction terms to investigate whether the effect of high baseline A1c differed by race and ethnicity group in Model 3 (M3). In building M2 and M3, race variables were included because they were of interest. However, additional variables were tested as main effects and interactions with time and included when p<.10. If a covariate demonstrated a significant slope prior to the intervention but not following the intervention (or vice versa), the covariate was included in both the pre and post slope specifications of the equation. We report the estimated model coefficients along with the associated 95% confidence intervals. We used SAS version 9.329 and HLM version 6.0830 to conduct analyses.
RESULTS
We identified 1,481 patients with multiple measurements of A1c (Table 1). Of these, 55% were White, 5% Black, and 40% Latino. Mean A1c at the start of the intervention was 7.4%. We also identified 5,320 patients with repeated measurements of blood pressure, of which 67% were White, 3% Black, and 30% Latino. The average mid-BP at the intervention start was 101.3 (mean diastolic/systolic blood pressure was 127/76).
Table 1.
Demographic Characteristics for Patients by Race and Ethnicity Group
| Total (N=18,675) | White Non-Latino (N=15,468) | Black Non-Latino (N=440) | Latino (N=2,767) | |||||
|---|---|---|---|---|---|---|---|---|
|
| ||||||||
| Female, n (%) | 11,410 | (61.0) | 9,306 | (60.1) | 273 | (62.0) | 1,831 | (66.1) |
| Mean age (SD) | 52.3 | (17.9) | 51.9 | (17.5) | 55.4 | (17.7) | 53.8 | (19.5) |
| Medicaid documented, n (%) | 3,507 | (18.7) | 3,201 | (20.6) | 117 | (26.5) | 189 | (6.8) |
| Multiple chronic conditions, n(%) | 3,975 | (21.2) | 3,177 | (20.5) | 151 | (34.3) | 647 | (23.3) |
| Patients with diabetes, n(%) | 2,638 | (14.1) | 1,596 | (10.3) | 103 | (23.4) | 939 | (33.9) |
| Patients with hypertension, n(%) | 1,780 | (9.8) | 1,088 | (7.2) | 78 | (18.6) | 614 | (23.4) |
| Patients with hemoglobin A1c measurements in both pre- and post-intervention periods | n=1,481 | n=823 | n=70 | n=588 | ||||
| Number of HbA1c measurements per person, mean(SD) | 4.4 | (1.8) | 4.1 | (1.6) | 4.6 | (1.9) | 4.8 | (1.9) |
| Baseline HbA1c, mean(SD) | 7.4 | (1.5) | 7.3 | (1.3) | 7.6 | (1.9) | 7.4 | (1.5) |
| Baseline HbA1c ≥ 8%, n(%) | 389 | (26.2) | 210 | (25.5) | 22 | (31.4) | 157 | (26.7) |
| Patients with blood pressure measurements in both pre- and post-intervention periods |
n=5,320 | n=3,566 | n=165 | n=1,589 | ||||
| Number of BP measurements per person, mean(SD) | 16.4 | (10.8) | 16.3 | (11.1) | 19.6 | (11.0) | 16.2 | (10.0) |
| Pre-intervention BP measurements | ||||||||
| Systolic, mean(SD) | 127.0 | (13.8) | 125.8 | (13.3) | 129.9 | (15.9) | 129.5 | (14.4) |
| Diastolic, mean(SD) | 75.6 | (7.8) | 75.0 | (7.9) | 76.1 | (8.2) | 76.8 | (7.2) |
| Mid-BP, mean(SD) | 101.3 | (9.6) | 100.4 | (9.5) | 103.0 | (10.7) | 103.1 | (9.4) |
| Post-intervention BP measurements | ||||||||
| Systolic, mean(SD) | 126.9 | (13.0) | 125.7 | (12.4) | 130.5 | (14.1) | 129.2 | (13.9) |
| Diastolic, mean(SD) | 75.0 | (7.3) | 74.6 | (7.5) | 75.6 | (7.7) | 75.8 | (6.9) |
| Mid-BP, mean(SD) | 101.0 | (8.9) | 100.2 | (8.7) | 103.1 | (9.4) | 102.5 | (9.0) |
| Elevated BP (>135/85) at intervention start | 1,342 | (25.2) | 803 | (22.5) | 63 | (38.1) | 476 | (29.9) |
Table 2 presents the unconditional and adjusted multilevel model results for A1c, and Figure 1 presents the trajectories of A1c in graphical form. The general trend for all patients is fairly robust to model specification (M0 to M3). We found an elevated mean intercept of A1c for all patients (in M0, b=6.69 [6.60, 6.78]) with a decreasing trend in the period prior to CMP implementation (in M0, b=−0.21 [−0.32, −0.11]) and a non-significant decrease in the post-period slope (in M0, b=0.18 [0.07, 0.29]). The trajectory models for A1c included an indicator variable for patients exhibiting poor glycemic control at the outset (baseline A1c ≥ 8%). We found that these patients, by definition, exhibited a higher baseline intercept, and also demonstrated an increased pre-period slope (in M0, b=1.03 [0.83, 1.24]), and a decreased post-period slope (in M0, b=−0.72 [−0.92, −0.52]).
TABLE 2.
Multilevel Model Results for Care Management Plus (CMP) Patients: Trajectories of Hemoglobin A1c
| Hemoglobin A1c | Model 0 |
Model 1 |
Model 2 |
Model 3 |
||||
|---|---|---|---|---|---|---|---|---|
| (M0) | (M1) | (M2) | (M3) | |||||
|
| ||||||||
| Fixed effects | ||||||||
| Intercept | 6.688 | [6.597, 6.778] | 6.673 | [6.575, 6.771] | 6.677 | [6.577, 6.777] | 6.719 | [6.612, 6.826] |
| Baseline HbA1c ≥ 8 | 2.595 | [2.472, 2.718] | 2.595 | [2.472, 2.718] | 2.553 | [2.429, 2.677] | 2.394 | [2.226, 2.563] |
| Black | 0.166 | [−0.102, 0.435] | 0.185 | [−0.083, 0.453] | 0.024 | [−0.299, 0.347] | ||
| Latino | 0.038 | [−0.104, 0.179] | 0.043 | [−0.102, 0.187] | −0.048 | [−0.210, 0.114] | ||
| Age | −0.008 | [−0.012, −0.004] | −0.008 | [−0.012, −0.004] | ||||
| Baseline ≥ 8* black interaction | 0.500 | [−0.064, 1.064] | ||||||
| Baseline ≥ 8* Latino interaction | 0.331 | [0.078, 0.583] | ||||||
| Slope, 1 y pre | −0.214 | [−0.324, −0.105] | −0.265 | [−0.408, −0.121] | −0.269 | [−0413, −0.126] | −0.284 | [−0.438, −0.129] |
| Baseline HbA1c ≥ 8 | 1 033 | [0.827, 1.238] | 1.026 | [0.820, 1.231] | 1.068 | [0.859, 1.276] | 1.130 | [0.828, 1.433] |
| Black | 0.234 | [−0.187, 0.656] | 0.192 | [−0.230, 0.615] | −0.021 | [−0.547, 0.506] | ||
| Latino | 0.086 | [−0.104, 0.277] | 0.062 | [−0.130, 0.254] | 0.112 | [−0.112, 0.337] | ||
| Age | 0.008 | [0.001, 0.015] | 0.008 | [0.001, 0.015] | ||||
| Baseline ≥ 8* black interaction | 0.544 | [−0.334, 1.423] | ||||||
| Baseline ≥ 8* Latino interaction | −0.194 | [−0.618, 0.230] | ||||||
| Slope, 1 y post | 0.177 | [0.068, 0.286] | 0.125 | [−0.012, 0.262] | 0.128 | [−0.009, 0.265] | 0.106 | [−0.041, 0.253] |
| Baseline HbA1c ≥ 8 | −0,719 | [−0.918, −0.519] | −0.719 | [−0.918, −0.520] | −0.706 | [−0.907, −0.505] | −0.628 | [−0.907, −0.349] |
| Black | −0.220 | 1–0.657, 0.217] | −0.230 | [−0.668, 0.208] | 0.050 | [−0.497, 0.597] | ||
| Latino | 0.153 | [−0.034, 0.341] | 0.141 | [ — 0.049, 0.331] | 0.167 | [−0.058, 0.393] | ||
| Age | 0.003 | [−0.004, 0.010] | 0.003 | [−0.004, 0.010] | ||||
| Baseline ≥ 8* black interaction | −0.773 | [− 1.683, 0.136] | ||||||
| Baseline ≥ 8* Latino interaction | −0.098 | [−0.508, 0.312] | ||||||
|
| ||||||||
| Random Effects | Variance | Variance | Variance | Variance | ||||
|
| ||||||||
| Intercept | 0.625*** | 0.624*** | 0.613*** | 0.607*** | ||||
| Time slope, pre | 0.963*** | 0.953*** | 0.952*** | 0.949*** | ||||
| Time slope, post | 0.809*** | 0.798*** | 0.80*** | 0.793*** | ||||
| Level 1 E | 0.547*** | 0.547*** | 0.547*** | 0.546*** | ||||
| Level 3 intercept | 0.005* | 0.003 | 0.003* | 0 004* | ||||
| Akaike Information Criterion | 18,533.33 | 18,536.96 | 18,510.51 | 18,508.00 | ||||
Note:
P<0.05
P<0.001.
Figure 1.
Trajectories of Hemoglobin A1c for Care Management Plus Patients (CMP) by Race and Ethnicity Group and Baseline A1c
When we look more closely at the trajectories for Black and Latino patients, the trajectories of A1c were not statistically different from White patients. Latino patients had pre-period slopes that were decreasing at a slightly slower rate relative to Whites (in M3 b=0.11 [−0.11, 0.34]), although these changes prior to CMP implementation were not statistically significant. On average, because there was a negative slope representing a decline in A1c post-intervention, a positive coefficient here means that Latinos had a higher-value slope than Whites indicating they declined at a slower rate than Whites. In the post-period, both Black and Latino patients demonstrated a (non-significant) slower rate of decline in A1c relative to White patients (in M3 for Latino, b=0.17 [−0.06, 0.39]; for Black, b=0.05 [−0.50, 0.60]). None of the interaction terms tested were statistically significant with the exception of Latino patients with baseline high A1c, who had a significantly higher initial level of A1c (intercept in M3, b=0.33 [0.08, 0.58]) relative to White patients.
Table 3 presents the unconditional and adjusted multilevel model results for mid-BP, and Figure 2 presents the trajectories of mid-BP in graphical form. The general trend for all patients is also robust to model specification (M0 to M2). On average, there was an increase in the slope of the trajectory of blood pressure for all patients in the pre-period (in M0, b=3.52 [2.39, 4.64]) and a decrease in the slope of the trajectory of blood pressure in the post-period (in M0, b=−5.21 [−5.70, −4.72]). Examining the racial and ethnic differences across M1 and M2, we also found a similar trend across model specifications. In the pre-period, Latino patients demonstrated a slower increase in the slope of mid-BP relative to White patients (in M2, b=−2.13 [−3.76, −0.51]). Black patients also demonstrated a slower trend in pre-period mid-BP, although this slope was not statistically different from White patients (in M2, b=−1.83 [−5.17, 1.51]). However, in the post-period, Latino and Black patients exhibited slower reductions in their slopes—they are less steep though comparable to White patients (in M2, Black b=0.79 [−2.73, 4.31]; Latino b=1.54 [−0.13, 3.21]).
TABLE 3.
Multilevel Model Results for Care Management Plus (CMP) Patients: Trajectories of Mid-BP
| Mid-BP (Average Systolic and Diastolic) | Model 0 |
Model 1 |
Model 2 |
|||
|---|---|---|---|---|---|---|
| (M0) | (M1) | (M2) | ||||
|
| ||||||
| Fixed effects | ||||||
| Intercept | 111,50 | [110.489, 112.512] | 111.471 | [110.318, 112.624] | 111.86 | [110.591, 113.130] |
| Black | 0.758 | [−1.476, 2.991] | 0.923 | [−1.320,3.165] | ||
| Latino | 0.445 | [−1.052, 1,941] | 0.517 | [−1.002, 2.036] | ||
| Female | −0.675 | [−1.426, 0.075] | ||||
| High risk | −0.112 | [−0.890, 0.666] | ||||
| Age | −0.034 | [−0.068, 0.000] | ||||
| Slope, 1 y pre | 3.516 | [2.389, 4.643] | 4.365 | [3.309, 5.420] | 4.295 | [3.221, 5.369] |
| Black | −2.084 | [−5.403, 1.235] | −1,830 | [−5.173, 1.513] | ||
| Latino | −2.289 | [−3.850, −0.729] | −2.131 | [−3.757, −0.505] | ||
| Age | −0.049 | [−0.102, 0.004] | ||||
| Slope, 1 y post | −5.208 | [−5.696, −4,721] | −5.878 | [−6.884, −4.872] | −5.864 | [−6.889, −4.840] |
| Black | 0.854 | [−2.656, 4.364] | 0.791 | [−2.727, 4.310] | ||
| Latino | 1.545 | [−0.083, 3.173] | 1.543 | [−0.127, 3.213] | ||
| Age | 0.018 | [−0.038, 0.073] | ||||
|
| ||||||
| Random Effects | Variance | Variance | Variance | |||
| Intercept 1 | 38.404*** | 38.292*** | 37.970*** | |||
| Time slope, pre | 31,062*** | 30.611*** | 30.579*** | |||
| Time slope, post | 52.238*** | 51.800*** | 51.792*** | |||
| Level 1 E | 114.159*** | 114.161*** | 114.153*** | |||
| Level 3 random effects | ||||||
| Intercept | 1.222* | 1.309* | 1.298* | |||
| Time slope, pre | 1.135*** | 0.098*** | 0.118*** | |||
| Time slope, post | 0.060*** | 0.069*** | 0.105*** | |||
| Akaike Information Criterion | 139,694.30 | 139,697.38 | 139,698.22 | |||
Note:
P<0.05
P<0.001.
Figure 2.
Trajectories of Mid-BP for Care Management Plus (CMP) Patients by Race and Ethnicity Group
DISCUSSION
This study examined racial and ethnic differences in clinical outcome trends following an intensive care management intervention. We estimated the trajectories of hemoglobin A1c and mid-BP for complex care patients enrolled in Care Management Plus across six clinics. In both unconditional and adjusted multilevel models, we found few differences between racial and ethnic groups in the changes of intermediate clinical outcomes following the implementation of CMP in the clinics where they sought care.
Based on our analyses we found that patients with poor glycemic control at the baseline (A1c ≥ 8%) were not improving in their measured A1c prior to the program, but did display significant reductions in A1c after participating in the program. Although there were gains for these patients with previously uncontrolled A1c, there were no discernible differences in the trajectories of A1c between racial and ethnic groups. Overall, we found that improved trajectories of A1c are achieved by all patients with poor glycemic control (A1c ≥ 8%). This may be due to system-wide improvements in communication and enhanced coordination of treatment, as well as increases in patient self-management brought on by care management tools. Still, it will be important to follow these patients over longer time horizons to ensure that these trends continue to reflect greater control of HbA1c, and less steep trajectories do not translate into significant gaps between Latino and White complex care patients.
For mid-BP, we found that the trajectory modestly increased leading up to program implementation and decreased afterward for all three race/ethnic background groups. It is noteworthy that while we did not find any statistically significant differences between race/ethnic groups either prior to or after implementation, there may be clinically significant gaps in both the initial and post-intervention period levels of blood pressure between White patients and Black and Latino patients that are not equalized or eliminated by the Care Management intervention. Even after the intervention, there were differences between the initial and post-period levels of blood pressure for underrepresented groups relative to Whites. Still, the moderate benefits in blood pressure control as evidenced by decreasing trajectory slopes appear to be attained by patients with complex care needs regardless of their racial and ethnic background group. In particular, these findings suggest that if left unintervened upon, the consequences of uncontrolled blood pressure may be greater for Latino and Black groups.
On balance, we found that the Care Management Plus program offers promising results for the observed population of complex care patients. Intermediate clinical outcomes of HbA1c for adult patients with uncontrolled diabetes, and mid-BP for adults with elevated blood pressure both show post-intervention improvements that may be of clinical relevance. Although we hypothesized that a general care coordination program that was not culturally tailored may elicit differential outcomes for underrepresented minority patients, we did not see any statistically significant differences between Latino, Black, and White complex care patients after Care Management Plus enrollment.
In general, these results are in line with the balance of literature that finds quality improvement tools such as tracking and reminder systems to improve some clinical outcomes for racial/ethnic patients are generally effective.31 Further, studies that evaluate the effects of different brands of care management find some improvements in patient satisfaction, perceived health, and processes of care for older patient populations with multiple health problems,31–33 particularly for programs with substantial in-person contact such as a dedicated care manager functioning as part of the primary care team.11
This study has several limitations that are worth noting. First, this study was conducted on a limited number of clinics willing to roll-out the Care Management Plus intervention. More to the point, there were comparatively few Black complex care patients available for analysis across the six clinics—although analyzing multiple measurements of clinical outcomes over time mitigated concerns of overly small sample sizes. Although these clinics are not representative of primary clinics across the US, they provide insight on whether these tools yield benefits for the control of important clinical outcomes for underrepresented racial and ethnic groups. In future, it will be important to test Care Management Plus on a larger clinical population, including a more robust sample of Black patients with complex chronic conditions. Relatedly, this study was conducted on a clinical population actively seeking care. Results may be different for adults with complex care needs who are not seen regularly in clinics, or those individuals who are not being captured by the health care system at all.
Second, we excluded patients with insufficient number of repeated blood pressure and HbA1c measurements from the analyses. Bias could be introduced should Latino or Black patients be disproportionately excluded. To explore this issue, we compared the relative proportion of included/excluded White, Black, and Latino patients (not shown). Although excluded patients did vary significantly by race/ethnicity, it was not in the worrisome direction. That is, proportionally more White patients were excluded from analyses because of insufficient number of repeated observations compared to Latino or Black patients.
Third, this study relied on electronic health record and medical claims databases. While these data are not broad and exhaustive—our analyses would have greatly benefited from examining various other socioeconomic and demographic factors—these data provide important information on the responsiveness of important clinical markers to a health system intervention that is gaining broad appeal in current health care reform experiments. This study will contribute to the national conversation around demonstrations of alternate models of health care organization (such as the Patient-Centered Medical Home), by providing additional knowledge around health technology-enabled versions of care management.
Fourth, we were concerned that by virtue of examining a clinical population with great health burdens, improvements in clinical outcomes may be expected, regardless of any care management program intervening on the population. Because of this concern, we conducted sensitivity analyses on a subsample of the most at-risk patients (i.e., those whose mean blood pressure was elevated above a specified level—135/85—in the pre-intervention period), and found that this group yielded very similar results to all care-managed patients in the final specified models, mitigating these concerns. Finally, given the nature of our data, it was necessary to specify the hierarchical linear models using three different nesting structures of the data: one for the individual changes within persons over time, another for the differences between individuals, and a final level to model the variation between patients seeking care at the different facilities. While we would have preferred the inclusion of clinic characteristics to model the variation between clinics we were unable to do so because of the limited number of clinics at our disposal. In future, comparing the clinical outcomes across a greater number of clinics implementing Care Management Plus will augment our capacity to explore contextual clinical factors that may explain differences in time-varying clinical outcomes between individual clinics that undoubtedly operate very differently.
This study was able to determine that Care Management Plus, a multipronged intervention that involves both, embedded care manager as well as a suite of tools to improve care for adults with complex care needs, elicits similar trends for glycemic and blood pressure outcomes for White, Black, and Latino adults. These analyses demonstrate promising results for the use of Care Management Plus and similar programs to improve intermediate clinical outcomes for vulnerable populations with complex chronic care needs. It remains to be seen whether these health care system delivery redesigns yield sustained benefits for underrepresented ethnic patients, and improvements in more distal outcomes, such as improvements in function and quality of life.
Acknowledgement of funding:
This study was funded by the Medical Research Foundation of Oregon (MRF-1217, Quiñones PI) and the American Diabetes Association (ADA 7-13-CD-08, Quiñones PI). Dr. Quiñones was also supported by the Summer Institute on Mentoring Researchers in Latino Health Disparities at San Diego State University (NIH/NHLBI R25HL105430). The views expressed in this article are those of the authors and do not necessarily represent the views of the Medical Research Foundation or the American Diabetes Association. An earlier version of this paper was presented at the Health Disparities Research at the Intersection of Race, Ethnicity, and Disability national conference in Washington, D.C. on April 25–26, 2013.
REFERENCES
- 1.Marengoni A, Rizzuto D, Wang H, Winblad B, Fratiglioni L. Patterns of chronic multimorbidity in the elderly population. J Am Geriatr Soc. 2009;57(2):225–230. [DOI] [PubMed] [Google Scholar]
- 2.Diederichs C, Berger K, Bartels DB. The measurement of multiple chronic diseases—a systematic review on existing multimorbidity indices. The Journals of Gerontology Series A: Biological Sciences and Medical Sciences. 2011;66(3):301–311. [DOI] [PubMed] [Google Scholar]
- 3.Schoen C, Osborn R, Squires D, Doty M, Pierson R, Applebaum S. New 2011 survey of patients with complex care needs in eleven countries finds that care is often poorly coordinated. Health Aff. 2011;30(12):2437–2448. [DOI] [PubMed] [Google Scholar]
- 4.Quiñones AR, Liang J, Bennett JM, Xu X, Ye W. How does the trajectory of multimorbidity vary across black, white, and Mexican Americans in middle and old age? The Journals of Gerontology Series B: Psychological Sciences and Social Sciences. 2011;66(6):739–749. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Vogeli C, Shields AE, Lee TA, et al. Multiple chronic conditions: Prevalence, health consequences, and implications for quality, care management, and costs. J Gen Intern Med. 2007;22 Suppl 3:391–395. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Tinetti ME, Fried T. The end of the disease era. Am J Med. 2004;116(3):179–185. [DOI] [PubMed] [Google Scholar]
- 7.Verbrugge LM, Jette AM. The disablement process. Soc Sci Med. 1994;38(1):1–14. [DOI] [PubMed] [Google Scholar]
- 8.Wagner EH. The role of patient care teams in chronic disease management. BMJ: British medical journal. 2000;320(7234):569. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Grumbach K, Bodenheimer T. Can health care teams improve primary care practice? JAMA: the journal of the American Medical Association. 2004;291(10):1246. [DOI] [PubMed] [Google Scholar]
- 10.Stille CJ, Jerant A, Bell D, Meltzer D, Elmore JG. Coordinating care across diseases, settings, and clinicians: A key role for the generalist in practice. Ann Intern Med. 2005;142(8):700–708. [DOI] [PubMed] [Google Scholar]
- 11.Peikes D, Chen A, Schore J, Brown R. Effects of care coordination on hospitalization, quality of care, and health care expenditures among Medicare beneficiaries. JAMA: the journal of the American Medical Association. 2009;301(6):603–618. [DOI] [PubMed] [Google Scholar]
- 12.Quiñones A, O’Neil M, Saha S, Freeman M, Henry S, Kansagara D. Interventions to improve minority health care and reduce racial and ethnic disparities. Vol 05–225. Washington DC: Department of Veterans Affairs: VA Evidence-based Synthesis Program Reports; 2011:05–225. [PubMed] [Google Scholar]
- 13.Berry-Millett R, Bodenheimer TS. Care management of patients with complex health care needs. Synth Proj Res Synth Rep. 2009;(19). pii: 52372. doi(19):52372. Epub 2009 Dec 16. [PubMed] [Google Scholar]
- 14.Dorr DA, Wilcox A, Burns L, Brunker CP, Narus SP, Clayton PD. Implementing a multidisease chronic care model in primary care using people and technology. Disease Management. 2006;9(1):1–15. [DOI] [PubMed] [Google Scholar]
- 15.Dorr DA, Wilcox AB, Brunker CP, Burdon RE, Donnelly SM. The effect of technology-supported, multidisease care management on the mortality and hospitalization of seniors. J Am Geriatr Soc. 2008;56(12):2195–2202. [DOI] [PubMed] [Google Scholar]
- 16.Gary TL, McGuire M, McCauley J, Brancati FL. Racial comparisons of health care and glycemic control for African American and white diabetic adults in an urban managed care organization. Disease Management. 2004;7(1):25–34. [DOI] [PubMed] [Google Scholar]
- 17.Smedley BD, Stith AY, Nelson AR. Unequal treatment: Confronting racial and ethnic disparities in health care (with CD). National Academies Press; 2009. [PubMed] [Google Scholar]
- 18.Williams DR, Jackson PB. Social sources of racial disparities in health. Health Aff. 2005;24(2):325–334. [DOI] [PubMed] [Google Scholar]
- 19.Braveman PA, Egerter SA, Mockenhaupt RE. Broadening the focus: The need to address the social determinants of health. Am J Prev Med. 2011;40(1):S4–S18. [DOI] [PubMed] [Google Scholar]
- 20.Starfield B, Shi L, Macinko J. Contribution of primary care to health systems and health. Milbank Q. 2005;83(3):457–502. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Quiñones AR, Liang J, Ye W. Differences in diabetes mellitus onset for older black, white, and Mexican Americans. Ethn Dis. 2013;23(3):310–315. [PMC free article] [PubMed] [Google Scholar]
- 22.Hajjar I, Kotchen TA. Trends in prevalence, awareness, treatment, and control of hypertension in the united states, 1988–2000. JAMA: the journal of the American Medical Association. 2003;290(2):199–206. [DOI] [PubMed] [Google Scholar]
- 23.Kirkman MS, Briscoe VJ, Clark N, et al. Diabetes in older adults. Diabetes Care. 2012;35(12):2650–2664. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Curran PJ, Obeidat K, Losardo D. Twelve frequently asked questions about growth curve modeling. Journal of Cognition and Development. 2010;11(2):121–136. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Moghissi ES, Korytkowski MT, DiNardo M, et al. American association of clinical endocrinologists and American Diabetes Association consensus statement on inpatient glycemic control. Diabetes Care. 2009;32(6):1119–1131. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Allen NB, Siddique J, Wilkins JT, et al. Blood pressure trajectories in early adulthood and subclinical atherosclerosis in middle age. JAMA. 2014;311(5):490–497. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Diez-Roux AV. Multilevel analysis in public health research. Annu Rev Public Health. 2000;21:171–192. [DOI] [PubMed] [Google Scholar]
- 28.Raudenbush SW. Hierarchical linear models: Applications and data analysis methods. Vol 1. Sage; 2002. [Google Scholar]
- 29.SAS Institute Inc. SAS 9.1.3 [computer software]. . 2002-2004.
- 30.Raudenbush SW, Bryk AS, Congdon R. HLM 6 for windows [computer software]. . 2004.
- 31.Counsell SR, Callahan CM, Clark DO, et al. Geriatric care management for low-income seniors. JAMA: the journal of the American Medical Association. 2007;298(22):2623–2633. [DOI] [PubMed] [Google Scholar]
- 32.Boult C, Reider L, Frey K, et al. Early effects of “Guided care” on the quality of health care for multimorbid older persons: A cluster-randomized controlled trial. J Gerontol A Biol Sci Med Sci. 2008;63(3):321–327. [DOI] [PubMed] [Google Scholar]
- 33.Shea S, Weinstock RS, Teresi JA, et al. A randomized trial comparing telemedicine case management with usual care in older, ethnically diverse, medically underserved patients with diabetes mellitus: 5 year results of the IDEATel study. Journal of the American Medical Informatics Association. 2009;16(4):446–456. [DOI] [PMC free article] [PubMed] [Google Scholar]