Patterns of healthcare utilization among patients with sickle cell disease hospitalized with pain crises

Abstract Background Vaso‐occlusive crises (VOC) are the hallmark of sickle cell disease (SCD). Adults experiencing VOC often have high rates of unexpected healthcare utilization. We characterized prior and future healthcare utilization among adults hospitalized with VOC at an urban, academic medical center. Methods We identified 449 VOC hospitalizations among 63 patients from 2013 to 2016. Patients were categorized based on receiving established care at the medical center and prior utilization: (a) not established (n = 21); (b) newly established (n = 10); (c) established with low utilization in past 12 months (<4 VOC hospitalizations) (n = 22); and (d) established with high utilization in past 12 months (≥4 VOC hospitalizations) (n = 10). Patient and hospitalization characteristics and future utilization were compared across categories. Results Median age was 26 years (Q1 = 22, Q3 = 29) and 55.6% were female. Established patients with high prior utilization tended to have higher median pain scores at admission (10, P = .08). Thirty‐day readmissions were highest in established patients with high prior utilization (P = .06), but 30‐day clinic visits were highest in established patients with low prior utilization (P = .08). Adjusted linear regression found that newly established patients (β = −4.6, P < .01) and established patients with low prior utilization (β = −5.6, P < .01) had fewer VOC hospitalizations in the ensuing 12 months than established patients with high prior utilization. Conclusion Among patients with SCD hospitalized for VOC, there was heterogeneity in healthcare utilization, with persistence in utilization over time for some patients. Efforts are needed to shift care from the acute setting to the outpatient clinic, which may lead to improved outcomes.

and Hispanics of Caribbean ancestry [1]. Patients with SCD have one of the highest 30-day hospital readmission rates (>30%) of all chronic conditions, reflecting both the severity of their condition and potentially avoidable healthcare utilization [2][3][4]. Vaso-occlusive crisis (VOC), also referred to as "painful crisis," is the hallmark of SCD and the most common reason for hospitalization among patients with SCD [5,6]. In addition to VOC, patients with SCD also experience hemolytic anemia and organ system complications, which all contribute to morbidity, poor health-related quality of life, high healthcare utilization, and premature mortality. Although a widely used and validated measure of SCD severity does not exist, frequent VOCs and more SCD-related complications (ie, acute chest syndrome, pneumonia, avascular necrosis of bone) likely represent higher disease severity [7,8].
Current treatments for VOC are primarily palliative and may involve pain management with or without opioids, while treatment with hydroxyurea or blood transfusions may reduce VOC occurrence. More recently, the United States of America (USA) Food and Drug Administration (FDA) has approved additional drugs (eg, L-glutamine, crizanlizumab, voxelotor) to prevent VOC; however, their uptake into clinical practice is not well understood [9]. Most patients with SCD manage their pain at home and have infrequent emergency department (ED) and hospital visits [10]. However, healthcare utilization is skewed with the highest utilization and healthcare costs among patients with severe disease, such as those with more SCD-related complications [10]. When patients in VOC present acutely to EDs or clinicians' offices, they have usually exhausted home care options, which signals the need for more aggressive pain management, such as parenteral opioids delivered in the inpatient setting [11]. Unfortunately, patients with SCD face barriers to appropriate pain management due to clinician mistrust, stigmatization, or fear of drug use [12][13][14][15][16][17].
A subset of patients with SCD may experience persistent healthcare utilization over time, while others may have intermittent utilization reflecting changing disease or patient factors [18]. Types of healthcare utilization may also vary across patients or over time. For example, a patient who has routine clinic visits with their hematologist and a well-delineated home-based care plan may have better disease management and avoid ED visits or unexpected, prolonged hospital stays. However, evidence suggests that patients with SCD have worse health outcomes and access to fewer health resources compared to other diseases [19][20][21][22]. In addition, patients in racial/ethnic minority groups have worse access to effective primary care and are more likely to receive low-value services [23,24]. Previous research has shown that not receiving follow-up care, missing appointments, and not having a primary care physician were associated with more 30-day readmissions for patients with SCD [25][26][27]. Other factors associated with higher healthcare utilization in adults with SCD were SCD complications, mood disorders, fragmented care, and VOCs [18,25,26].
Our institution (Tufts Medical Center in the city of Boston, Massachusetts (MA), USA) identified patients with SCD as one of the patient populations with the highest 30-day readmissions rate, indicating an opportunity for quality improvement among a vulnerable group of patients who have historically faced disparities in their care. Identifying the subset of patients with SCD who are at risk for high healthcare utilization could enhance targeting of services and improve disease management. Therefore, this study aimed to create a cohort of patients with SCD hospitalized for VOC to describe the persistence of different types of healthcare utilization over time and identify factors associated with healthcare utilization. We hypothesized that patients with more severe disease and high prior utilization would have more healthcare utilization in the future.

Sample
We created a retrospective cohort of consecutive hospitalizations for

Data
The study database was designed to have three levels of information: patient, hospitalization, and hospital day. Hospital days were nested in hospitalizations, which were subsequently nested within patient.

Analysis
Patients' entry into our cohort, defined by their first hospitalization at year; [33] this cut-off is slightly higher than some cut-offs of the number of VOCs or VOC-related healthcare utilization because our sample was more severe with at least one hospitalization in the study period [34].
We calculated the following types of 30-day utilization after discharge from the first hospitalization using hospital billing data: 30

Cohort development
Initially

Hospitalization characteristics
The median length of hospital stay was 7 days (Q1 = 4, Q3 = 10) for the first hospitalization in the study window (Table 2)

Future utilization
Analyses of future utilization were restricted to established and newly established patients (n = 42) (

DISCUSSION
We For most patients with SCD, healthcare utilization and hospitalizations rates are low [34]. In our sample of hospitalized patients with SCD, only a subset had persistent high utilization over time.
The availability heuristic may explain the inappropriate use of terms like "frequent flyer" because providers may more easily recall this subset of patients, despite their lack of representativeness of all patients with SCD [15,35]. Pain severity and frequency may be underestimated by clinicians in patients without high utilization, despite many patients still experiencing pain outside of the healthcare setting [36]. Although prior research shows that patients with more SCD complications typically constitute the subset with higher utilization [10,18], we found a relatively comparable number of complications across prior utilization groups. Given that our sample all experienced at least one VOC hospitalization, they may represent a generally more severe group, or there may be other factors influencing healthcare utilization [37].
As with other chronic health conditions, the management of SCD may be further complicated by the patient's mental health status. Over 50% of established patients in our sample had a documented affective disorder, which is comparable to rates reported in other SCD samples [38,39]. Affective disorders and the pain experience are deeply intertwined, with more pain reported among those with anxiety and depression, and pain leading to symptoms of anxiety and depression [39]. Patients with high prior utilization had nearly twice the rate of affective disorders than patients with low prior utilization; affective disorders were not reported for not established and newly established patients given the lack of documented medical history at the time of their first hospitalization. Prior research has similarly found that patients with SCD and mental health disorders have higher healthcare utilization than patients without mental health disorders [37,38].
Factors related to SES, insurance, and employment may also be related to healthcare utilization among patients with SCD [40].  [41]. Investing resources for patients to navigate the disability process for Medicare may improve access to affordable healthcare [42]. As a proxy measure for SES, we found that the ADI did not vary by prior utilization level. However, the median overall ADI score of 6.5 was higher than the statewide median of 5, providing further evidence of the SES challenges faced by many patients with SCD.
Other measures of SES, such as employment, housing, or income, were not routinely collected in the EMR, so could not be studied, despite their known relationship with healthcare utilization [40].
Research has suggested several approaches for reducing avoidable healthcare utilization in patients with SCD, such as the use of individual pain treatment plans when patients present to the ED with a VOC, and shorter time to opiate dosing [26,43]. Shifting care from the reactionary acute setting to proactive disease management provided in the clinic may also reduce avoidable utilization. Our analysis found similar rates of any type of 30-day healthcare utilization across established patients with low and high prior utilization, but a higher 30-day readmission rate in the high prior utilization group and a higher 30-day clinic visit rate in the low prior utilization patients. In addition, use of hydroxyurea or chronic transfusions may result in fewer VOC episodes, shorter VOC hospitalizations, fewer SCD-related hospitalizations, and lower opioid utilization [44][45][46][47]. However, we found the highest rate of hydroxyurea prescription in the established high prior utilization group. One explanation is patients with higher utilization may have more severe disease and thus, be more likely to have already been identified as potential candidates for treatment with hydroxyurea. In addition, having received a prescription for hydroxyurea does not mean patients were adherent, which we were unable to assess.
Further, patients of lower SES may experience cost-related medication non-adherence [48]. Prior research has shown wide variability in rates of hydroxyurea adherence, and those with higher healthcare utilization perceive hydroxyurea as less useful, which may lower adherence [49,50]. With regards to chronic transfusion, we had an insufficient number of patients receiving chronic transfusions to analyze its asso- We acknowledge this study's limitations. Although the sample size was small and some analyses may have been under powered, we were able to identify important factors that may influence high healthcare utilization and should be the focus of future, larger studies. Our sample size was further reduced by the exclusion of patients not established at our medical center from some analyses. This further highlights issues related to fragmented care for SCD. This study was conducted at a single medical center, which may reduce generalizability. It is possible that some utilization was missed and our results are underestimates of actual utilization. Patients presenting at other institutions may have both low prior and future hospitalizations; however, our inclusion of VOC hospitalizations at outside hospitals that were documented in the EMR should help reduce this bias. We did not collect information on laboratory markers, hematocrit or ferritin level, which could be associated with increased VOC and future healthcare utilization [8]. However, changes in these laboratory markers can be patient dependent and may require future research.
In conclusion, we created a cohort of patients with SCD hospitalized for VOC and described their healthcare utilization prior to and following hospitalization for an acute VOC episode. A subset of patients had high healthcare utilization that persisted over time. Differences in types of future utilization (eg, readmissions, clinic visits) represent an opportunity to shift care from the acute setting to the outpatient clinic setting. Issues related to mental health, access and adherence to hydroxyurea, and fragmented care may also contribute to avoidable, acute healthcare utilization. Although addressing these issues will be challenging, particularly given patient-reported mistrust and barriers to pain management, it should help improve disease management and result in better clinical outcomes.

ACKNOWLEDGMENTS
We would like to thank Ruth Ann Weidner and Daqin Mao from the

Institute of Clinical Research and Health Policy Studies at Tufts Med-
ical Center for their assistance with data management. We would also like to thank Megan Catalano, Annika Jain, and Raisa Masood for their assistance with data entry. This work was funded by the National Institute on Minority Health and Health Disparities (R21 MD011455-01) and the National Center for Advancing Translational Sciences (1KL2TR002545).

CONFLICT OF INTEREST
The authors do not have financial or commercial interests to disclose.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.