Int.J. Behav. Med. DOI 10.1007/s12529-016-9538-y

The Association between Educational Attainment and Patterns of Emergency Department Utilization among Adults with Sickle Cell Disease C. R. Jonassaint 1 & M. C. Beach 2 & J. A. Haythornthwaite 2 & S. M. Bediako 4 & M. Diener-West 3 & J. J. Strouse 2 & S. Lanzkron 2 & G. Onojobi 1 & C. P. Carroll 2 & C. Haywood Jr. 2

# International Society of Behavioral Medicine 2016

Abstract Purpose Patients with low educational attainment may be at increased risk for unplanned health care utilization. This study aimed to determine what factors are related to emergency department (ED) visits in hopes of guiding treatments and early interventions. Methods At two medical centers in the Mid-Atlantic United States, 258 adults with sickle cell disease aged 19–70 years participated in a retrospective study where we examined whether education level is independently associated with ED visits after accounting for other socioeconomic status (SES) variables, such as pain and disease severity and psychosocial functioning. Results The data showed that patients without a high school education visited the ED three times as frequently as patients with post secondary education. Controlling for poverty and employment status decreased the effect of education on ED visits by 33.24 %. Further controlling for disease severity and/ or psychosocial functioning could not account for the remaining association between education and ED visits, suggesting that education is independently associated with potentially avoidable emergency care.

* C. R. Jonassaint [email protected]

1

School of Medicine, University of Pittsburgh, 230 McKee Pl, Suite 600, Pittsburgh, PA 15213, USA

2

School of Medicine, Johns Hopkins University, Baltimore, MD, USA

3

Johns Hopkins Bloomberg School of Public Health, Baltimore, MD, USA

4

Department of Psychology, University of Maryland, Baltimore County, Baltimore, USA

Conclusions Early interventions addressing disparities in academic performance, especially for those children most at risk, may lead to improved long-term health outcomes in this population. Keywords Sickle cell disease . Socioeconomic status . Education . Health care utilization . Health disparities . Chronic illness . Emergency care . Blood disorders

Introduction Sickle cell disease (SCD) is an inherited blood disorder that primarily affects people of African descent. One in 500 African Americans in the USA has SCD and an additional one in 14 carry the trait. However, SCD is much more common in other parts of the world, particularly in malariaendemic countries. Worldwide, there are approximately 300, 000 babies born with sickle cell disease annually [1]. Although the medical treatment has drastically improved, life expectancy for sickle cell patients is approximately 30 years below that of their healthy peers [2]. In SCD, painful episodes of vaso-occlusive events (Bpain crises^) can cause long-term organ damage and increase risk of death [3]. Sickle cell pain crises also result in a disproportionately high use of health care resources, as indexed by number of emergency department (ED) visits, hospitalizations, and hospital length of stay [4–6]. Adult patients experiencing an average of almost three hospital encounters a year [5]. The percentage of SCD-related ED visits that result in hospitalization ranges from 29 to 40 % [7, 8], and one of every 100 hospitalizations results in death [9]. A 2010 study estimated that SCD care in the USA cost $2.4 billion annually [4]. Thus, it is critical to prevent symptoms and decrease the incidence of

Int.J. Behav. Med.

acute pain crises that can lead to unplanned health care utilization. It is a minority of patients that frequent the ED for potentially preventable symptoms and account for a majority of the health care resources in SCD [5, 10, 11]. Most patients rarely visit the ED while a small number of patients frequently require emergency care. Little is known about what contributes to the high frequency of ED visits and hospitalizations among selected patients [12] and, despite advances in medical treatment, health care utilization and the cost of care for patients with SCD may be increasing [8, 9]. Identifying patients who are at high risk for disease-related episodes and hospitalizations is important for improving patient health outcomes and quality of life, which will, in turn, lead to reduced health care costs. The lack of attention to academic achievement in SCD has been a missed opportunity for improving the lives of patients with the disease [13–15]. Frequent hospitalizations and other health consequences of SCD can make the process of completing an education even more challenging. New intervention approaches that address educational attainment may help improve the quality of life for patients with SCD [13, 14]. With the aim of identifying predictors of health care utilization in SCD, investigators used pilot data to conclude that patients who did not finish high school were hospitalized almost five times as often as patients with a college level degree [16]. Despite a body of evidence showing that low education increases risk of hospitalization in the general population [17], no published studies have evaluated this association in patients with SCD. Therefore, the primary aim of the current study is to determine the association between education and ED utilization among adults with SCD, independent of potential confounders. Specifically, we hypothesized that less education is independently associated with greater ED utilization. Prior evidence suggests that multiple indices of socioeconomic status (SES) [5, 18–21], disease severity [12, 22, 23], and psychosocial factors [24, 25] are related to health care utilization. Therefore, contingent upon confirmation of our hypothesized education-utilization association, a secondary aim of this study is to identify the extent to which other SES indicators, markers of disease severity, and psychosocial variables account for any observed association between education and ED utilization.

experiences of people with SCD within the health care system. The local university institutional review boards approved this study. To be eligible for the parent study, patients had to be (1) adults aged 18 years or older or adolescents aged 15–17 years and receiving comprehensive sickle cell care at the Johns Hopkins Hospital (JHH) or Howard University Hospital (HUH); (2) diagnosed with SCD and at least one of the sickle hemoglobinopathies Hb SS, Hb SC, Hb S/beta-plus-thalassemia (Hb Sβ+), or Hb S/beta-zero-thalassemia (Hb Sβ0); (3) present for a routine follow-up appointment in a SCD clinic at either JHH or HUH; (4) willing to adhere to study procedures without plans to move out of the area for 3 years; and (5) English speaking. For our analysis, only patients 19 years of age or older were included to eliminate any patients who were still attending high school and had not yet had the opportunity to pursue post secondary education. Eligible patients participated in an audio self-interview and completed a battery of questionnaires which included the patient’s demographic characteristics, psychosocial functioning, pain and disease severity, health care utilization, and experiences of care. Clinical history data and health care utilization at the patient’s recruitment site (JHH or HUH) for the previous 12-month period were abstracted from medical records. Methods and Measurements Education The primary independent variable was the patient’s selfreported level of education. Participants were categorized into one of the three following educational level categories based on their response: (1) BNo High School^—did not finish high school or is still working on completing high school; (2) BHigh School^—has completed high school or a high school equivalent; or (3) BPost-secondary^—has completed a post high school degree or is currently enrolled in post high school education. Individuals 19 years of age or older who had not graduated high school were assumed to either have dropped out of school, failed, or missed at least 1 year of high school. Three patients were currently enrolled in school but had not yet finished high school at the time of the study (ages 20, 25, and 28 years). To test our assumption that 19 years was an appropriate age of cut-off, we conducted sensitivity analyses by repeating all analyses using only those patients who were 25 years of age or older.

Methods

Socioeconomic Status

Study Design and Setting

Data were collected on employment status (i.e., employed or unemployed, on disability, and retired) and medical insurance type (i.e., private, public, or uninsured). Participants were also asked to report their household income and number of people living in their household. This information was used to

Our data come from adults with SCD receiving ambulatory care at two urban hospital centers and participating in an ongoing federally funded cohort study evaluating the

Int.J. Behav. Med.

calculate poverty status based on the US Department of Health and Human Services 2011 poverty guidelines [26].

Dependent Variable ED Visits in the Previous Year

Disease Severity Several measures were used to model disease severity. Hemoglobin genotypes were determined from medical records and categorized into three groups: Hb SS/Hb Sβ0, Hb SC, or Hb Sβ+. The number of lifetime SCD-related comorbidities experienced by the patient was determined by summing the presence of medical diagnoses for the following conditions: acute chest syndrome, avascular necrosis, creatinine levels greater than 1.3, pulmonary hypertension, or iron overload. The number of non-SCD-related comorbidities was determined by summing the presence of medical diagnoses for the following conditions: diabetes, hypertension, HIV, hepatitis B, or hepatitis C. Finally, we also examined whether participants had any psychiatric diagnoses in their medical records, such as depression or bipolar disorder. In addition to the data from medical records, participants also provided information regarding their disease severity. Participants were asked to report whether they experienced daily chronic pain and, if so, what the severity of their pain was like on a Bgood day,^ as well as the number of days in a typical week that were good days. Daily chronic pain was treated as a yes (1)/no (0) binary response variable. Pain rating was based on a scale of 0 (no pain) to 10 (pain as bad as you can imagine), and number of good days in a typical week was a 0 to 7 response variable. Patients without daily chronic pain were coded as missing. Psychosocial Factors All patients were administered and completed three validated instruments that assessed one of the following psychosocial variables of interest: self-efficacy, life stress, and depression. Self-efficacy was measured using the Sickle Cell Self Efficacy Scale SCSES; [27], a 9-item scale that assesses a patient’s perceived ability to manage disease symptomatology and engage in day-to-day activities. Total scores range from a possible 9 (low selfefficacy) to 45 (high self-efficacy). Life stress was measured using the Urban Life Stress Scale SCSES; [27], a 21-item scale that assesses a person’s level of stress associated with various aspects of life, including money, housing, job condition, family matters, neighborhood environment, crime, and violence. Total scores range from 0 (low stress) to 84 (high stress). Depressive symptoms were measured using the 10-item Center for Epidemiological Studies Depression Scale CES-D [28] that has possible total scores ranging from 0 to 30 points. Scores of 10 or greater are indicative of clinically significant depression in the general US population.

Trained staff evaluated the medical records and abstracted the number of ED visits recorded for each participant at the study site during the previous 12-month period. This measure does not reflect utilization that may have occurred at other hospitals.

Analysis The analyses included (1) the bivariate association between our primary independent variable, education level, and ED visits; (2) the bivariate associations between education level and each of our potential mediators (SES, disease severity, and psychosocial factors); (3) the bivariate associations between our potential mediators and ED visits; and finally, (4) the multivariable association of education level and the potential mediators with ED visits. In this final step, we measured the indirect effect (amount of mediation) by subtracting the direct effect of education level on ED visits measured in the multivariable model adjusted for mediators (result of steps 2 and 3), from the total effect of education level on ED visits measured in simple, unadjusted models (result of step 1). To accomplish these four steps, we first tested differences in our study measures, covariates, potential mediators, and outcome variables by education level using the Pearson’s chi-squared test for categorical variables or the Fisher exact test when expected cell sizes were below n = 5, one-way analysis of variance for continuous variables, and negative binomial regressions for count variables. For interpreting chisquared contingency table and Fisher exact test results, statistically significant tests were followed with post hoc pairwise comparisons using logistic regressions. For regression analyses, education level was modeled as two dummy variables with post high school as the referent group. In our negative binomial regressions, the composite null hypotheses for both parameters were tested using the Wald test. Next, we used a Poisson regression to test the association between study measures and ED visits during the 12-month period, but due to overdispersion (large standard deviation (SD) relative to the mean), a negative binomial model was further investigated. Results are reported as unadjusted incidence rate ratio (IRR) and 95 % confidence interval (95 % CI) for number of events during the 12-month period. Finally, we tested whether SES, disease severity, or psychosocial factors mediated the association between education level and ED visits using multivariable negative binomial regressions. Variables that were associated with education level and ED visits at the P < 0.10 level were considered as potential mediators in the multivariable models.

Int.J. Behav. Med.

We tested three different regression models. Model 1 included education level and the covariates age, sex, hemoglobin genotype, and site (JHH or HUH); model 2 added all SES mediators to model 1; and model 3 added all disease severity and psychosocial mediators to model 2. We compared the IRRs for the education-level parameter, comparing no high school to post high school in model 1 (total effect) to the same parameter in each subsequent model (direct effect). The IRR differences comparing no high school to post high school were presented as percentage changes with bootstrapped 95 % CI calculated using 100 repetitions.

Missing Data Approximately 10 % of our total sample was missing data on at least one of our psychosocial variables. Further, more than 10 % of patients had missing data on household income level. Thus, to maintain equal sample sizes across all analyses, missing values were replaced with the mean value of their respective education level group. For categorical variables with missing data, missing values were coded and tested as a dummy variable in the model. This allowed us to estimate an effect for participants with missing values for the variable of interest. In sensitivity analyses, our multivariable models were repeated in a sample with complete data only. Statistical significance was considered at a threshold of P < 0.05 based on two-tailed tests. Multicollinearity was evaluated in our models by calculating the tolerance and variance inflation factor (VIF) statistics for each variable. The criteria for collinearity were tolerance values less than 0.1 and VIF values greater than 10. Improvement in model fit across models was evaluated using −2 * log likelihood change scores, Akaike information criterion, and Bayesian information criterion. All statistical analyses were performed using Stata® 11 software (College Station, TX).

Results Characteristics of Study Participants The study included 258 participants, 144 from JHH and 114 from HUH. Only 31 (12 %) participants had not completed high school, and 107 (41 %) completed high school but did not pursue post high school education. There were 120 (46 %) participants who either had a post high school degree or were enrolled in post high school education at the time of their baseline interview. The mean age of the participants was 35.9 (range, 19–70 years). Participants from the two study sites were of similar age, sex, and education level (P > 0.05).

Differences by Education Level There were no differences by gender or age in educational level (Table 1). The no high school group had six times greater odds (odds ratio [OR] = 5.93, 95 % CI = 2.36–14.97) of being in poverty than the post high school group, while the high school group had three times greater odds (OR = 3.24; 95 % CI = 1.75–6.01). Further, both the no high school and high school groups were less likely to be employed (OR = 0.18; 95 % CI = 0.07–0.50 and OR = 0.19; 95 % CI = 0.10–0.35, respectively) and more likely to be on disability (OR = 3.81, 95 % CI = 1.67–8.66 and OR = 4.32, 95 % CI = 2.45–7.62, respectively) compared to the post high school group. No participant with less than a high school education carried private medical insurance coverage; this group was more likely to have public insurance, such as Medicare or Medicaid (OR = 7.6; 95 % CI = 2.18–26.71). Hemoglobin genotype and number of SCD-related, nonSCD-related, or psychiatric comorbidities did not differ by education level. For the self-reported measures of disease severity and psychosocial functioning, the high school group reported more daily chronic pain than the post high school group (OR = 2.07; 95 % CI = 1.22–3.54), and both the no high school and high school groups reported more depressive symptoms on the CES-D. Education level was not related to self-efficacy and life stress. Education and ED Visits Education level was negatively associated with ED visits; during the previous 12-month period, participants in the no high school and high school groups experienced a higher mean number of ED visits than the post secondary group (Tables 1 and 2). Table 2 shows the incidence rate of ED visits was significantly higher in the high school group (IRR = 2.34, 95 % CI = 1.44–3.79) and the no high school group (IRR = 3.23, 95 % CI = 1.59–6.56) compared to the post secondary groups, respectively. Although the no high school group had, on average, 1.2 more ED visits than the high school group, the relative difference in rate of ED visits between these two groups was not statistically significant (IRR = 1.38; 95 % CI = 0.68–2.80). As shown in Table 2, older age was associated with a lower rate of ED visits. Among SES indicators, participants in poverty had an ED visit rate 2.5 times as great as participants not in poverty. For employment status, the disability/retired group had a higher ED visit rate than the employed participants. Medical coverage was also associated with health care utilization; those with public insurance had twice the rate of ED visits than privately insured participants. For indices of disease severity, having more SCD-related comorbidities was related to more frequent ED visits (χ2(2, 258) = 7.65, P = 0.03). Participants with one non-SCD

Int.J. Behav. Med. Table 1

Demographic and clinic variables by education level for N = 258 patients with sickle cell disease Education

Variable

No HS (n = 31)

High school (n = 107)

Post sec (n = 120)

P valuea

Age, mean (SD)

37.48 (13.29)

37.11 (11.67)

34.48 (11.29)

0.178

Male Socioeconomic status

19 (61.29)

44 (41.12)

53 (44.17)

0.135

<0.001

Poverty status, n (%) Not in poverty

10 (32.26)

50 (46.73)

91 (75.83)

15 (48.39)

41 (38.32)

23 (19.17)

6 (19.35)

16 (14.95)

6 (5.00)

Employed Unemployed

6 (19.35) 3 (9.68)

21 (19.63) 14 (13.08)

67 (55.83) 15 (12.50)

Disability/retired

22 (70.97)

72 (67.29)

38 (31.67)

0 (0.00)

9 (9.18)

35 (29.41)

Poverty Not reported Employment status, n (%)

Medical coverage, n (%) Private Public

<0.001

<0.001

26 (86.67)

78 (79.59)

59 (49.58)

4 (13.33) 1 (3.23)

11 (11.22) 9 (8.41)

25 (21.01) 1 (.83)

23 (74.19) 5 (16.13)

75 (70.09) 19 (17.76)

87 (72.50) 27 (22.50)

3 (9.68)

13 (12.15)

6 (5.00)

7 (22.58)

19 (17.76)

29 (24.17)

7 (22.58) 13 (41.94)

36 (33.64) 26 (24.30)

42 (35.00) 30 (25.00)

3+ No. of non SCD-related, n (%) 0

4 (12.90)

26 (24.30)

19 (15.83)

18 (58.06)

66 (61.68)

78 (65.00)

0.457

1 2 Psychiatric diagnosis, n (%)

8 (25.81) 5 (16.13) 4 (12.90)

30 (28.04) 11 (10.28) 22 (17.19)

35 (29.17) 7 (5.83) 11 (11.11)

0.413

19 (63.33) 3.32 (2.44) 3.72 (1.51)

69 (64.49) 3.22 (2.60) 3.97 (1.55)

56 (46.67) 2.30 (2.57) 4.75 (1.76)

0.018 0.013 0.331

18.82 (3.67) 17.96 (13.39) 10.22 (5.43) 3.58 (5.36)

18.91 (2.98) 18.11 (13.75) 10.01 (6.41) 2.38 (4.37)

19.53 (3.00) 15.74 (10.29) 7.96 (5.89) 1.07 (2.26)

0.249 0.327 0.028 <0.001

Other Missing Disease severityb Hemoglobin genotype, n (%) SS/Sb0 SC Sb+ Comorbidities No. of SCD-related, n (%) 0 1 2

Daily pain (y/n), n (%) Pain on Bgood day^, mean (SD) No. of Bgood days^, mean (SD) Psychosocial variables SCD self-efficacy score Urban life stress score CES-D score ED visits, mean (SD)c a

0.354

0.226

P value represents the Pearson’s test or Fisher’s exact test (where cell size <5) for categorical variables, F test for continuous variables, and the Wald test of the composite null hypotheses for count variables b

When comorbid conditions were tested individually, there were no education differences in prevalence

c

Utilization variables are total number of visits by type over the past 12 months and are presented as means and standard deviations

Int.J. Behav. Med. Table 2 Results of unadjusted univariate binomial regression models testing predictors of emergency department visits among adults with sickle cell disease

Emergency department visits Variable

IRR

95 % CI

P value

Age

0.97

(0.95, 0.99)

0.007

Male Socioeconomic status

0.83

(0.52, 1.33)

0.44

Post secondary High school

1 2.34

(1.00, 1.00) (1.44, 3.79)

. <0.001

No High school

3.23

(1.59, 6.56)

<0.001

Education level

Poverty status Not in poverty

1

(1.00, 1.00)

.

2.46 2.68

(1.50, 4.04) (1.31, 5.51)

<0.001 0.007

Employed Unemployed

1 1.31

(1.00, 1.00) (0.62, 2.76)

. 0.470

Disability Retired/other

2.96 0.84

(1.79, 4.90) (0.35, 1.99)

<0.001 0.69

Medical coveragea Private Public Other

1 2.13 1.14

(1.00, 1.00) (1.13, 4.02) (0.50, 2.60)

. 0.019 0.75

Missing Disease severity Hemoglobin genotype SS/Sb0 SC

1.92

(0.55, 6.67)

0.3

1 0.60

(1.00, 1.00) (0.33, 1.10)

. 0.097

Sb+ Comorbidities No. of SCD-related 0 1

1.52

(0.67, 3.44)

0.31

1 1.14

[1.00,1.00] [0.59,2.18]

. .70

2 3+ No. of non SCD-related 0 1 2 Psychiatric diagnosis (y/n) Daily pain (y/n)

1.54 1.88

[0.78,3.02] [0.91,3.89]

.21 .091

1 2.00 1.71 1.95 2.15

(1.00, 1.00) (1.20, 3.33) (0.76, 3.83) (1.03, 3.69) (1.35, 3.42)

. 0.008 0.19 0.042 0.001

1.08 0.87

(0.97, 1.20) (0.74, 1.02)

0.15 0.097

0.90 1.01 1.04

(0.83, 0.97) (0.99, 1.03) (1.00, 1.09)

0.005 0.39 0.061

Poverty Not reported Employment status

Pain on Bgood day^ No. of Bgood days^ Psychosocial variables SCD self-efficacy Urban life stress CES-D score a

ALL models test the univariate association between predictor and number of ED visits over a 12-month period

Int.J. Behav. Med.

(95 % CI = 1.06–4.17); this represented a 33.24 % (95 % bootstrap CI = 8.17–58.31) decrease in the education coefficient after adjustment for SES variables. In model 3, after controlling for daily pain and depressive symptoms, the IRR for not having completed high school was further reduced to 1.94 (95 % CI = 0.97–3.87); however, this reflected a nonsignificant change (7.14 %; 95 % bootstrap CI = −11.02– 25.48), change in the education coefficient. In addition to the total effect of education, poverty and chronic pain were both significantly associated with ED visits in the final model. We repeated the multivariable analyses using a sample restricted to individuals aged 25 years and older (n = 212). Overall, there were no differences in the conclusions of our findings when we limited the sample to this older age group. Analyses were also repeated among those of all ages with complete data only (n = 213), and the results were consistent with findings using the total sample.

comorbidity visited the ED twice as often as participants without non-SCD comorbidities. Having a psychiatric diagnosis was also associated with a higher rate of ED visits (Table 2). Among self-reported measures of disease severity and psychosocial functioning, having daily chronic pain was associated with twice the rate of ED visits. Lower self-efficacy in self-care and having more depressive symptoms were also associated with a higher rate of ED visits. Mediational Analysis Table 3 presents the results of multivariable analyses of the direct and indirect effects of education on the number of ED visits using negative binomial regression. These results were similar to those found using the Poisson regression model. After adjustment for covariates, participants who did not complete high school visited the ED three times as often as participants with post high school education. A similar pattern was observed for participants who completed high school, and they visited the ED significantly more often than the post high school group. Comparing utilization between the two lower education groups, the increased rate for the no high school group in ED visits (IRR = 1.27, 95 % CI = 0.67–2.40) was not statistically significant from that of the high school group (data not shown). In the test of potential mediators, after controlling for SES variables, the IRR for not having completed high school in model 1 decreased from 3.15 (95 % CI = 1.63–6.07) to 2.10

Discussion Our study is the first to systematically examine the association between education level and ED utilization among patients with SCD. Consistent with our primary hypothesis, participants who had less education experienced greater rates of ED utilization during the previous 12 months. More specifically, participants who did not graduate from high school incurred a rate of ED visits three times as high as participants

Table 3 Incidence rate ratios (IRR) and 95 % confidence intervals for the association of education and potential mediation variables with number of emergency department (ED) visits among N = 258 adult patients with sickle cell disease Variables Education level Post secondary High school No high school Poverty Employment status Employed Unemployed Disability Medical coverage Private Public Daily pain (y/n) CES-D score
Model 1 IRR

95 % CI

P

1 2.48 3.15

. (1.57, 3.92) (1.63, 6.07)

<0.01 <0.01

871.84 903.82

.

Model 2 IRR

95 % CI

P

Model 3 IRR

95 % CI

P

1 1.67 2.10 1.92

. (0.99, 2.80) (1.06, 4.17) (1.14, 3.23)

. 0.05 0.03 0.01

1 1.53 1.94 2.11

. (0.92, 2.55) (0.97, 3.87) (1.26, 3.56)

. 0.10 0.06 0.01

1 1.30

. (0.61, 2.74)

. 0.50

1 1.15

. (0.53, 2.51)

. 0.72

1.93

(1.06, 3.52)

0.03

1.55

(0.84, 2.87)

0.16

1 0.66

. (0.32, 1.34)

. 0.25

(4.53, 56.84)

. (0.36, 1.49) (1.05, 2.65) (0.98, 1.06) (−7.04, 22.26)

. 0.39 0.03 0.44

30.69 873.24 933.64

1 0.73 1.67 1.02 7.61 865.41 932.84

All models included covariates age, sex, hemoglobin genotype, and study site; effects for Bmissing^ and Bother^ groups are not shown

Int.J. Behav. Med.

with post high school education. Our secondary aim was to examine the potential mediating role of other indices in explaining the relationship between education level and ED utilization. We found that the education–utilization relationship was only partially explained by socioeconomic factors (e.g., poverty), markers of disease severity (e.g., chronic pain), or psychosocial factors (e.g., depression)―some of the most commonly measured and well-established risk factors for health care utilization [24]. Therefore, these findings support the notion that among adults with SCD, there is a relationship between education level and health care utilization that is independent of several other important and well-documented risk factors. Other factors apart from education were also independent correlates of ED utilization. Chronic pain was found to be the only index of disease severity that correlated with ED utilization after adjustment for other significant risk factors. The importance of the association between chronic pain and increased utilization frequency, particularly ED visits, is consistent with other studies [22, 29]. Among the psychosocial variables examined, low selfefficacy showed the strongest association with more frequent ED visits, but was surprisingly only marginally related to lower educational attainment. Low self-efficacy has previously been correlated with more physician visits, as well as increased physical and psychological symptomatology, which may account for utilization patterns [30]. Although selfefficacy has been posited as an important factor in predicting utilization patterns in patients with SCD, the construct has received relatively little attention in the previous SCD research [24, 31]. To better understand health care utilization and disease outcomes in SCD, future research will need to include psychosocial measures in addition to clinical measures and take a more comprehensive approach to identifying high-risk patients. For instance, there are several other psychological and behavioral factors not measured in our study which may be important with respect to health care utilization, such as social support, health literacy, cognitive functioning, and self-management behaviors. Despite the importance of disease severity and psychosocial factors in SCD outcomes, we found that low education was related to increased ED utilization independent of these factors. Educational attainment has received very little attention as an independent risk factor for poor health outcomes in SCD, despite consistent evidence suggesting that students with SCD are particularly vulnerable for poor academic achievement and dropout [32–34]. We compared a group of individuals with SCD who were relatively well educated to other SCD samples [35] and found that the association of less education with increased ED utilization was robust to statistical adjustment for several important risk factors expected to have a higher proximal impact on health care utilization and outcomes. In

addition, our analyses suggest that education may not only affect health care utilization but also many dimensions of health, including social, psychological, and medical outcomes, which are consequences of education that have gone unrecognized in this population. Identifying the psychological and behavioral risk factors for poor SCD outcomes will help us better understand and explain the association between less education and increased health care utilization. It is important to note that the association between education and SCD-related health outcomes is not unidirectional as has been modeled in our study, but rather, the causal pathway can also go in the opposite direction where SCD-related health outcomes directly influence a patient’s ability to obtain education. During early childhood in particular, students with SCD are at increased risk for low educational attainment due to multiple disease-related health complications and frequent hospitalizations [32–34, 36, 37]. More than one-third of children and adolescents with SCD will miss at least 1 month of classes each school year [32]. Further, children with SCD often have deficiencies in school readiness [38] or cerebral ischemia that may also limit achievement [14, 39]. Most students with SCD who have a history of stroke or silent cerebral infarcts will have a cognitive deficit or require special education, but even those with a normal magnetic resonance imaging (MRI) may show impairments [13, 14]. Therefore, in our study sample, low educational attainment may be a proxy for unrecognized or undocumented brain disease in our sample [40]. The early identification of students with SCD at risk for school failure is currently inadequate, and when these students are identified, they are not receiving the appropriate educational resources and support from their teachers and school administration [13, 41, 42]. Although frequent negative school experiences of young people living with SCD and the lack of adequate accommodations for these students are welldocumented [43], there are only two studies examining academic- or school-based interventions for students with SCD. The first, a randomized, controlled study showed that providing education about SCD and its management to the student with SCD, their teachers, and peers significantly decreased SCD-related school absences [44]. In contrast, the second study, a prospective cohort, found that a schoolbased program to increase individualized education plans, development, and educational resources (e.g., tutoring) for children with SCD and cerebral infarcts did not improve absenteeism or academic performance during the 2-year follow-up [14]. However, unlike the first study, the second study did not provide education or engage the student’s teachers or school administration. Specific components of school-based interventions for students with SCD may be necessary for effective education. Disclosure to teachers and peers of someone’s SCD status alone may worsen the student’s school experience [45], but combining this information with education about SCD and

Int.J. Behav. Med.

its management may prove to elicit the appropriate teacher and peer support that students with SCD need to be successful. School-based educational programs have shown to improve academic outcomes among low-income minority students without SCD [46]. Whether these programs are effective in children with SCD needs to be tested, as does whether improving educational achievement reduces health care utilization. Our study had some limitations that should be noted. First, our sample size may have been insufficient to detect important secondary effects of interest. Although there was sufficient statistical power to detect the effect of education, we were unable to appropriately test smaller effect sizes and differences between educational groups that may have been informative. Second, ED utilization data were collected retrospectively. Therefore, it is difficult to infer causality and, in some cases, associations may be due to reverse causation; it is possible that the levels of ED utilization may have impacted our independent measures. A third limitation of the study is that utilization data were only available from the two study sites and did not include visits to other EDs that patients may have made during the study period. Lastly, we have no measure of education quality. Participants who have completed high school may not all have equal levels of academic attainment due to differences in educational quality, leading to significant withingroup heterogeneity. However, it is important to note that the latter three limitations would tend to bias the education– utilization association toward the null; thus, our results are likely to be conservative estimates of this association.

Acknowledgements The primary author would like to thank Andrea Ball for her indispensible intellectual contributions and editing assistance in the preparation of this manuscript. Study data were collected and managed using REDCap electronic data capture tools hosted at The Johns Hopkins University. REDCap (Research Electronic Data Capture) is a secure, Web-based application designed to support data capture for research studies, providing (1) an intuitive interface for validated data entry; (2) audit trails for tracking data manipulation and export procedures; (3) automated export procedures for seamless data downloads to common statistical packages; and (4) procedures for importing data from external sources. Funding This study was funded by NHLBI grant no. 5R01HL088511-04 Dr. Haywood is funded by a career development award from the NHLBI: 5 K01 HL108832 02. Dr. Jonassaint was supported by grant number K12HS022989 from the Agency for Healthcare Research and Quality. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NHLBI or the Agency for Healthcare Research and Quality. Compliance with ethical standards Conflict of Interest The authors declare that they have no competing interests. Ethical Approval This study was approved by the Howard University and the Johns Hopkins Hospital Institutional Review Boards. All procedures performed in this study involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. Informed Consent Informed consent was obtained from all individual participants included in the study.

Conclusion References We found that higher educational attainment is associated with lower health care utilization, even after accounting for multiple confounders. Our study directly tested a number of the most relevant pathways (e.g., poverty, medical insurance coverage, disease severity, and psychosocial functioning) that could potentially explain this relationship. Future studies will need to identify, in prospective study designs, specifically how educational attainment exerts its effects on health outcomes in adulthood, with the aim of informing the design of school-based interventions for children and adolescents living with SCD. In this study, we have tried to capture emergent and potentially preventable health care utilization rather than prevention care such as outpatient visits. Indeed, for patients at high risk of ED visits, more frequent outpatient clinic visits to monitor symptoms and or more liberal use of outpatient pain treatment (i.e., day hospitals), consistently intervening early, before pain symptoms become severe and difficult to treat, may decrease preventable ED visits, subsequent hospitalizations, and ultimately, risk of death.

1.

Williams TN, Weatherall DJ. World Distribution, Population Genetics, and Health Burden of the Hemoglobinopathies. Cold Spring Harbor Perspectives in Medicine. 2012;2(9). doi:10.1101/ cshperspect.a011692. 2. Hassell KL. Population estimates of sickle cell disease in the U.S. Am J Prev Med. 2010;38(4 Suppl):S512–21. doi:10.1016/j.amepre. 2009.12.022. 3. Platt OS, Brambilla DJ, Rosse WF, et al. Mortality in sickle cell disease. Life expectancy and risk factors for early death. N Engl J Med. 1994;330(23):1639–44. 4. Lanzkron S, Carroll CP, Haywood Jr C. The burden of emergency department use for sickle-cell disease: an analysis of the national emergency department sample database. Am J Hematol. 2010;85(10):797–9. doi:10.1002/ajh.21807. 5. Brousseau DC, Owens PL, Mosso AL, Panepinto JA, Steiner CA. Acute care utilization and rehospitalizations for sickle cell disease. JAMA. 2010;303(13):1288–94. doi:10.1001/jama.2010.378. 6. Wolfson JA, Schrager SM, Khanna R, Coates TD, Kipke MD. Sickle cell disease in California: sociodemographic predictors of emergency department utilization. Pediatr Blood Cancer. 2012;58(1):66–73. doi:10.1002/pbc.22979. 7. Yusuf HR, Atrash HK, Grosse SD, Parker CS, Grant AM. Emergency department visits made by patients with sickle cell

Int.J. Behav. Med.

8.

9. 10.

11.

12.

13.

14.

15.

16.

17. 18.

19.

20.

21.

22.

23.

24.

25.

26. 27.

disease: a descriptive study, 1999–2007. Am J Prev Med. 2010;38(4 Suppl):S536–41. doi:10.1016/j.amepre.2010.01.001. Lanzkron S, Haywood Jr C, Segal JB, Dover GJ. Hospitalization rates and costs of care of patients with sickle-cell anemia in the state of Maryland in the era of hydroxyurea. Am J Hematol. 2006;81(12):927–32. doi:10.1002/ajh.20703. Steiner CA, Miller JL. Sickle cell disease patients in U.S. hospitals, 2004. Rockville, MD: Agency for Health Care Quality; 2006. Epstein RM, Shields CG, Franks P, Meldrum SC, Feldman M, Kravitz RL. Exploring and validating patient concerns: relation to prescribing for depression. Ann Fam Med. 2007;5(1):21–8. Wolfson JA, Schrager SM, Coates TD, Kipke MD. Sickle-cell disease in California: a population-based description of emergency department utilization. Pediatr Blood Cancer. 2011;56(3):413–9. doi:10.1002/pbc.22792. Carroll CP, Haywood C, Fagan P, Lanzkron S. The course and correlates of high hospital utilization in sickle cell disease: evidence from a large, urban Medicaid managed care organization. Am J Hematol. 2009;84(10):666–70. doi:10.1002/ajh.21515. Herron S, Bacak SJ, King A, DeBaun MR. Inadequate recognition of education resources required for high-risk students with sickle cell disease. Arch Pediatr Adolesc Med. 2003;157(1):104. King A, Herron S, McKinstry R, et al. A multidisciplinary health care team's efforts to improve educational attainment in children with sickle-cell anemia and cerebral infarcts. J Sch Health. 2006;76(1):33–7. doi:10.1111/j.1746-1561.2006.00064.x. Smith KE, Patterson CA, Szabo MM, Tarazi RA, Barakat LP. Predictors of academic achievement for school age children with sickle cell disease. Adv Sch Ment Health Promot. 2013;6(1):5–20. doi:10.1080/1754730X.2012.760919. Jonassaint CR, Beach MC, Lanzkron S, Haywood Jr C. How is patient socioeconomic status related to health care utilization in sickle cell disease? Annual National Sickle Cell Disease Scientific Meeting. 2011. Arendt JN. In sickness and in health–till education do us part: education effects on hospitalization. Econ Educ Rev. 2008;27(2):161–72. Woods K, Karrison T, Koshy M, Patel A, Friedmann P, Cassel C. Hospital utilization patterns and costs for adult sickle cell patients in Illinois. Public Health Rep. 1997;112(1):44–51. Haque A, Telfair J. Socioeconomic distress and health status: the urban–rural dichotomy of services utilization for people with sickle cell disorder in North Carolina. J Rural Health. 2000;16(1):43–55. Telfair J, Haque A, Etienne M, Tang S, Strasser S. Rural/urban differences in access to and utilization of services among people in Alabama with sickle cell disease. Public Health Rep. 2003;118(1):27–36. Ellison AM, Bauchner H. Socioeconomic status and length of hospital stay in children with vaso-occlusive crises of sickle cell disease. J Natl Med Assoc. 2007;99(3):192–6. Aisiku IP, Smith WR, McClish DK, et al. Comparisons of high versus low emergency department utilizers in sickle cell disease. Ann Emerg Med. 2009;53(5):587–93. doi:10.1016/j. annemergmed.2008.07.050. Smith WR, Penberthy LT, Bovbjerg VE, et al. Daily assessment of pain in adults with sickle cell disease. Ann Intern Med. 2008;148(2):99–101. Reese FL, Smith WR. Psychosocial determinants of health care utilization in sickle cell disease patients. Ann Behav Med. 1997;19(2):171–8. Edwards CL, Scales MT, Loughlin C, et al. A brief review of the pathophysiology, associated pain, and psychosocial issues in sickle cell disease. Int J Beh Med. 2005;12(3):171–9. doi:10.1207/ s15327558ijbm1203_6. Start EH, Assistance EL-IPT. The 2011 Health & Human Services (HHS) Poverty Guidelines. Fed Regist. 2011;76(13):3637–8. Edwards R, Telfair J, Cecil H, Lenoci J. Reliability and validity of a self-efficacy instrument specific to sickle cell disease. Behav Res Ther. 2000;38(9):951–63.

28.

Kohout FJ, Berkman LF, Evans DA, Cornoni-Huntley J. Two shorter forms of the CES-D (Center for Epidemiological Studies Depression) depression symptoms index. J Aging Health. 1993;5(2):179–93. http://www.ncbi.nlm.nih.gov/pubmed/10125443. 29. Grant MM, Gil KM, Floyd MY, Abrams M. Depression and functioning in relation to health care use in sickle cell disease. Ann Behav Med. 2000;22(2):149–57. 30. Edwards R, Telfair J, Cecil H, Lenoci J. Self-efficacy as a predictor of adult adjustment to sickle cell disease: one-year outcomes. Psychosom Med. 2001;63(5):850–8. 31. Maxwell K, Streetly A, Bevan D. Experiences of hospital care and treatment seeking for pain from sickle cell disease: qualitative study. BMJ. 1999;318(7198):1585–90. 32. Schwartz LA, Radcliffe J, Barakat LP. Associates of school absenteeism in adolescents with sickle cell disease. Pediatr Blood Cancer. 2009;52(1):92–6. doi:10.1002/pbc.21819. 33. Wang W, Enos L, Gallagher D, et al. Neuropsychologic performance in school-aged children with sickle cell disease: a report from the Cooperative Study of Sickle Cell Disease. J Pediatr. 2001;139(3):391–7. doi:10.1067/mpd.2001.116935. 34. Day S, Chismark E. The cognitive and academic impact of sickle cell disease. J Sch Nurs. 2006;22(6):330–5. 35. Farber MD, Koshy M, Kinney TR. Cooperative study of sickle cell disease: demographic and socioeconomic characteristics of patients and families with sickle cell disease. J Chronic Dis. 1985;38(6):495–505. 36. Taras H, Potts-Datema W. Chronic health conditions and student performance at school. J Sch Health. 2005;75(7):255–66. doi:10. 1111/j.1746-1561.2005.00034.x. 37. van Heesch MM, Bosma H, Traag T, Otten F. Hospital admissions and school dropout: a retrospective cohort study of the 'selection hypothesis'. Eur J Public Health. 2012;22(4):550–5. doi:10.1093/ eurpub/ckr129. 38. Chua-Lim C, Moore RB, McCleary G, Shah A, Mankad VN. Deficiencies in school readiness skills of children with sickle cell anemia: a preliminary report. South Med J. 1993;86(4):397–402. 39. Schatz J, Brown RT, Pascual JM, Hsu L, DeBaun MR. Poor school and cognitive functioning with silent cerebral infarcts and sickle cell disease. Neurology. 2001;56(8):1109–11. 40. Vichinsky EP, Neumayr LD, Gold JI, et al. Neuropsychological dysfunction and neuroimaging abnormalities in neurologically intact adults with sickle cell anemia. JAMA. 2010;303(18):1823–31. doi:10.1001/jama.2010.562. 41. Mayes S, Wolfe-Christensen C, Mullins LL, Cain JP. Psychoeducational screening in pediatric sickle cell disease: an evaluation of academic and health concerns in the school environment. Child Health Care. 2011;40(2):101–15. doi:10.1080/ 02739615.2011.566465. 42. Dyson SM, Atkin K, Culley LA, Dyson SE, Evans H, Rowley DT. Disclosure and sickle cell disorder: a mixed methods study of the young person with sickle cell at school. Soc Sci Med. 2010;70(12): 2036–44. doi:10.1016/j.socscimed.2010.03.010. 43. Dyson SM, Atkin K, Culley LA, Dyson SE. The educational experiences of young people with sickle cell disorder: a commentary on the existing literature. Disabil Soc. 2007;22(6):581–94. doi:10. 1080/09687590701560196. 44. Koontz K, Short AD, Kalinyak K, Noll RB. A randomized, controlled pilot trial of a school intervention for children with sickle cell anemia. J Pediatr Psychol. 2004;29(1):7–17. 45. Dyson SM, Atkin K, Culley LA, Dyson SE, Evans H. Sickle cell, habitual dys-positions and fragile dispositions: young people with sickle cell at school. Sociol Health Illn. 2011;33(3):465–83. doi:10. 1111/j.1467-9566.2010.01301.x. 46. Muennig P, Schweinhart L, Montie J, Neidell M. Effects of a prekindergarten educational intervention on adult health: 37-year follow-up results of a randomized controlled trial. Am J Public Health. 2009;99(8):1431–7. doi:10.2105/AJPH.2008.148353.