Association between Hospital Admissions and Healthcare Association between Hospital Admissions and Healthcare Provider Communication for Individuals with Sickle Cell Disease Provider Communication for Individuals with Sickle Cell Disease

Objective: To test the hypothesis that caregivers ’ or adult participants ’ low ratings of provider communication are associated with more hospital admissions among adults and children with sickle cell disease (SCD), respectively. Secondarily, we determined whether there was an association between the caregivers ’ or participants ’ health literacy and rating of providers ’ communication. Methods: Primary data were collected from participants through surveys between 2014 and 2016, across six sickle cell centers throughout the U.S. In this cross-sectional cohort study, 211 adults with SCD and 331 caregivers of children with SCD completed surveys evaluating provider communication using the Consumer Assessment of Healthcare Providers and Systems (CAHPS), healthcare utilization, health literacy, and other sociodemographic and behavioral variables. Analyses included descriptive statistics, bivariate analyses, and logistic regression. Results: Participants with better ratings of provider communication were less likely to be hospitalized (odds ratio (OR) = 0.54, 95% con ﬁ dence interval (CI) = [0.35, 0.83]). Positive ratings of provider communication were associated with fewer readmissions for children (OR = 0.23, 95% CI = [0.09, 0.57]). Participants with better ratings of provider communication were less likely to rate their health literacy as lower (regression coe ﬃ cient (B) = − 0.28, 95% CI = [ − 0.46, − 0.10]). Conclusions: Low ratings of provider communication were associated with more hospitalizations and readmissions in SCD, suggesting the need for interventions targeted at improving patient-provider communication which could decrease hospitalizations for this population.


Introduction
Sickle cell disease (SCD) is a hereditary disorder of hemoglobin within the red blood cells, affecting over 100,000 Americans, many of whom face health disparities [1-4.]With improvements in care, SCD has become a chronic disease that affects both children and adults [5].Despite improvements in SCD management, significant challenges persist, including a significant burden of hospitalizations [4,[6][7][8][9], costs over $900,000 by the age of 45 years [10], poor adherence to preventive care, including missing appointments [11][12][13][14][15][16][17], and lack of providers with knowledge and expertise in the disease [18-20.]A poor provider-patient relationship, an important indicator of patient experience and healthcare quality, can contribute to these challenges.Inadequate provider communication has been shown to lead to unfavorable outcomes in other diseases such as high blood pressure, anxiety, pain in postoperative situations, and problem and symptom resolution [21,22].
Literature focusing on the provider-patient relationship in SCD is sparse, and only in adults [23-25.]In 2009, Haywood et al. evaluated
provider communication with sociodemographic factors and levels of trust in the medical profession.This study was done among adults with SCD at a single center and demonstrated that better ratings of provider communication were associated with older patient age, lower household income, and less frequent hospital utilization.The researchers also showed that lower ratings of provider communication were associated with lower levels of trust towards the medical profession.In 2014, Haywood et al. performed a multi-center study to compare the proportion of individuals with SCD that rated their provider communication as poor as compared to a U.S. sample of African American adult patients.Their study revealed that individuals with SCD rated provider communication as poor significantly more often than the national sample of African Americans.They also demonstrated that younger individuals with SCD and those with a higher education were more likely to rate their communication with their provider lower than similar individuals from the national sample.Prior studies that investigated provider communication only included adults with SCD, and only one evaluated associations with social determinants of health [23-25.]The present multi-center study of individuals with SCD across the U.S. tested the hypothesis that poor perceived provider communication is associated with increased hospitalizations in children and adults with SCD.The study was part of the Mid-South Clinical Data Research Network (CDRN) [26], which enrolled thousands of participants with different chronic diseases.We used two psychometrically validated composites from the Consumer Assessment of Healthcare Providers and Systems (CAHPS) to measure provider communication and shared decision-making [27-32.]We evaluated provider-patient communication within the framework of other contextual factors, such as social and behavioral determinants of health, given the high burden of social determinants of health in SCD and the influence of these factors on health utilization and outcomes [33,34].

Methods
This project was part of the Mid-South CDRN [26], funded by the Patient-Centered Outcomes Research Institute (PCORI).The Mid-South CDRN survey tool was designed to obtain uniform information across cohorts with obesity, coronary heart disease and SCD.The Institutional Review Boards of the participating sites approved all study procedures and informed consent was obtained from all participants.2) received care at one of the six contributing centers, and (3) had a diagnosis of SCD (of any phenotype) or were parents/caregivers of children with SCD.Individuals with SCD and their caregivers were recruited either by their health care providers during clinic visits or by using flyers in clinics.Participants completed surveys on computer tablets, or by paper-and-pencil.Members of the research team were present for questions.Participants' time was compensated with a gift card upon completion of the survey.

Surveys
The full details of the surveys are described elsewhere [9,35], but are briefly described here.Various stakeholders, including individuals with SCD, helped design the survey tools and selected the final questions.Question domains included patient-reported healthcare experience (e.g.provider communication, shared decision-making), perceived health literacy, healthcare utilization (e.g.hospitalizations, readmissions and clinic appointment adherence), and social and behavioral determinants of health.We combined some categories of survey responses for ease of interpretation within the regression analyses.Caregivers responded about themselves for educational attainment, difficulty paying bills, and marital status, and answered about their child for the other questions.

Healthcare experience -CAHPS measures
In this study, we evaluated perceived provider communication which encompasses the ability of the provider to gather information to facilitate accurate diagnosis, counsel appropriately, give therapeutic instructions, and establish caring relationships with their patients [36].We selected four questions related to provider communication from the CAHPS Clinician & Group Survey Version 3.0 [37], using a 4-point scale from 'Never' (1) to 'Always' (4).The surveys cover topics such as how well providers communicate with their patients, provide care, and the helpfulness of staff, all important topics to patients and for which patients are the best information source.The four questions we selected were: did your provider (1) explain things in a way that was easy to understand, (2) listen carefully to you, (3) show respect for what you had to say, and (4) spend enough time with you or your child.CAHPS surveys are widely used and extensively validated measurement tools to elicit patient reports about their healthcare experiences.Shared decision making (SDM) has been defined as: 'an approach where clinicians and patients share the best available evidence when faced with the task of making decisions, and where patients are supported to consider options, to achieve informed preferences [38].'We selected two questions about shared decisionmaking based on the three relevant items in the CAHPS Patient-Centered Medical Home Survey [30] -(1) did your provider talk with you about the pros and cons of each choice for treatment or health care and (2) did your provider ask which choice you thought was best for you or your child, answered on a 4-point scale from 'Definitely yes' (1) to 'Definitely no' (4).For each of the two CAHPS domains, provider communication and shared decision-making, a composite score for each respondent was computed by averaging the individual item responses within a domain.Composite scores were calculated if at least half of the items in a domain were answered.
Health literacybrief health literacy screening Health literacy, or the ability to understand, communicate, and act on health information, was evaluated using the Brief Health Literacy Screening [39,40].Inadequate health literacy can be determined from one or a combination of all three of these questions [39,40].Responses of 'somewhat' or better for the question 'How confident are you filling out medical forms by yourself?' has been used to define 'good' health literacy [39].Caregivers responded about their health literacy, not their child's.

Social and behavioral determinants of health
Depressive symptoms were measured by the Patient Health Questionnaire (PHQ-2 [41]), Participants rated their social supports using the ENRICHD (Enhancing Recovery in Coronary Heart Disease) Social Support Inventory (ESSI [42]).Low support has been defined as 2 or more items ≤2, or 2 or more items ≤3 and an adjusted overall score ≤18 [43].Participants and caregivers rated spirituality using a single item 'how spiritual or religious do you consider yourself (or your child) to be,' from very (1) to not at all (4).Based on the distribution of the responses and for ease of analysis, we dichotomized the variable into 'very' spiritual (option 1) and 'not very' spiritual (options 2-4).Social determinants of health included sex, race, ethnicity, educational attainment, difficulty paying bills, and marital status.
Healthcare utilizationmissed clinic appointments, hospitalizations, and readmissions Adults with SCD and caregivers of children selfreported missed clinic appointments, hospitalizations, and readmissions within the past year.Readmissions were defined as being admitted to the hospital twice in a 30-day period.

Statistical analysis
Study data were collected, de-identified, and managed using the REDCap electronic data capture tools hosted at Vanderbilt University [44].Surveys were excluded from analyses for missing data about age, site, and sex.We first used descriptive statistics to summarize demographics, social and behavioral determinants of health, and other questions.Means, standard deviations, and ranges were used for continuous variables, medians and ranges for count variables, and frequencies and proportions for categorical variables.Next, we reported descriptive statistics (mean, standard deviation, and frequency distribution) for the CAHPS items and conducted confirmatory factor analysis (CFA) to examine if the six items could be grouped into the two composites, provider communication and shared decision making, as hypothesized.
We explored bivariate associations among provider communication, shared decision-making, and potential risk factors, i.e. social determinants of health (sex, age, race, ethnicity, education level, marital status, household size, ability to pay bills), depressive symptoms, health literacy, social support, and spirituality.Bivariate associations were examined using either Pearson correlation or t-test as appropriate.We created logistic regression models for the outcome measure of hospital admissions, readmissions, and missed appointments.We also created linear regression models using each of the three health literacy items as outcome measures.In each regression model, provider communication, shared decision-making, and all the risk factors were included as predictors.Initially, models were created for all participants but given that adults and children with SCD have important differences in outcome measures, we also conducted regression analyses for adults and children separately.For binary outcomes (i.e.hospital admission, readmission, and missed appointment), race and ethnicity were not included as predictors in the regression models because there were too few non-African American or Hispanic participants when examining adults and children with SCD separately.Given that 14% of respondents (only 2% among adults but 21% among caregivers of children) did not report their education level, we considered missing education level as a valid response category in order to retain a larger sample size when conducting analyses in the full sample and in the pediatric patient sample.We excluded such cases in the regression models for adult patients, given the small proportion (2%) of missing data.

Demographics
A total of 573 individuals with SCD (adults and caregivers of children with SCD) completed the surveys at a single clinic visit.After excluding surveys with missing data, our final sample for analysis included 211 adults with SCD and 331 caregivers of children with SCD (n = 542).We oversampled our population to accommodate nonresponses and exceeded our projected sample size of 450.Table 1 shows the distribution of socio-demographic characteristics and summarizes scores for depressive symptoms, social support, and health literacy for adults and pediatric patients.

Socio-demographic variables vary among adults with SCD and children with SCD (as reported by their caregivers)
Forty-five percent of the total sample reported it was 'somewhat' to 'very difficult' to pay monthly bills.About 42% of the total sample rated themselves as 'very' spiritual or religious.Most adults and caregivers rated their health literacy as 'good' (75%).The mean score on the PHQ-2 for depression in adults (1.46, SD=1.55) was higher than what caregivers reported for children (0.84, SD=1.26).This is very similar to what we previously reported [9].

Patient-reported experiences about provider communication were positive
As shown in Table 2, the majority (74% to 85%) of the total sample answered 'Always' to the four items about provider communication.About 63% of respondents reported 'Definitely yes' to the two items about shared decision-making.

Hypothesized composite structure for provider communication and shared decisionmaking fit well
The two-factor CFA model showed excellent model fit, with the Root Mean Squared Error of Approximation (RMSEA) = 0.031, Comparative Fit Index (CFI) = 0.999, and Tucker-Lewis Index (TLI) = 0.999, suggesting that the measures have good construct validity and composite scores should be computed as planned.An RMSEA < 0.06, a CFI > 0.95, and a TLI > 0.95 indicate good fit [48-50.]The factor loading of the six items on the two factors (domains) is shown in Table 2.

Better experience of provider communication was associated with a lower likelihood of hospital admission and readmission as well as fewer problems reported by patients with learning about their medical conditions
Bivariate associations among provider communication, shared decision making, and patient characteristics are shown in Supplementary Tables 1 and 2. Better experience with provider communication was associated with higher level of shared decision-making (p < 0.0001) and better social support (mean difference of provider communication score = 0.15, 95% confidence interval (CI) = [0.04,0.27]).Female patients reported better experience with shared decision-making than males (mean difference = 3.32, 95% CI = [0.08,0.38]).Results from regression analyses are shown in Tables 3 and 4.
In the full sample, patients with better experience of provider communication were less likely to be hospitalized (odds ratio (OR) = 0.54, 95% CI = [0.35,0.83]).When analyzed separately, the result was not statistically significant for adult patients but still held true for children with SCD (OR = 0.54, 95% CI = [0.31,0.94]).Hospital readmission was not associated with provider communication or shared decision-making in the full sample.However, among children with SCD, patients whose caregivers had better experience of provider communication were less likely to have readmissions (OR = 0.23, 95% CI = [0.09,0.57]) while those with better experience of shared decision-making were more likely to have readmissions (OR = 2.38, 95% CI = [1.04,5.43]).Neither provider communication nor shared decision-making were significantly associated with the likelihood of missing appointments.Regarding health literacy, for the full sample, patients/caregivers with better experience of provider communication were less likely to have problems learning about their medical conditions because of difficulty understanding written information (regression coefficient (B) = −0.28,95% CI = [−0.46,−0.10]).When children and caregivers were analyzed separately, this result held true among adults (B = −0.36,95% CI = [−0.70,−0.02]) but not among caregivers of children with SCD.

Discussion
Our manuscript is one of the first to leverage a national research network of sickle cell centers to describe the relationship between patient experience (i.e.provider communication and shared decision making) and acute healthcare utilization (i.e.hospitalizations, readmissions, and missed appointments) in children and adults with SCD.Poor provider communication, as experienced by caregivers of children, was a significant predictor of higher hospitalizations and readmissions; however, these significant associations were not seen in adults.The reason we may see this difference in findings is that caregivers are usually at a heightened level of vigilance and anxiety, poor communication and relationship with providers may make them go to hospital more often when they see even a small sign of discomfort in their children.Adult patients, on the other hand, might be more experienced with SCD self-management and less likely to go to hospital even when their communication with providers is not as good.Also, there are differences between these two populations, including other potentially important factors that might contribute to making healthcare provider communication not significant in relation to healthcare utilization for adults.These might include a lack of primary care providers, mental health issues, and financial hardship which are all more prevalent in the adult population [9,51].In addition, while not statistically significant, the odds ratio for both adults and pediatrics were similar (pediatrics: 0.54 vs adults: 0.57).The lack of significance could be because of our smaller sample size of adults (pediatric: 330 vs. adults: 175).Provider communication may become a significant predictor if we had a larger sample of adults with SCD.Further research is needed on these differences.These findings are novel, have not been well described in the literature in any chronic diseases, and demonstrate the importance of good provider communication as experienced by patients and their caregivers in decreasing acute healthcare utilization.These findings demonstrate a need for interventions that are focused on improving poor patient-provider communication.
Paradoxically, children demonstrated lower healthcare utilization when their caregivers experienced poor shared decision-making with their providers.One potential reason for this finding is that there are inevitable reasons for hospitalization in children with SCD (e.g.fevers).Children with SCD with fevers can develop blood infections and sepsis and need to be monitored closely and given proper interventions.Caregivers who perceive that they can participate in good shared decision making with their providers may feel enabled to seek early interventions This may lead to increased hospitalization but prevents significant morbidity and mortality that can happen with delayed treatment.Others have shown that better disease knowledge is associated with higher healthcare utilization in children with SCD [52], and it is conceivable that better knowledge and shared decision- making yield a more activated caregiver who ends up at the hospital more often.
In our secondary analyses of patient experience and health literacy as associated with missed appointments, we found mixed results.Adults with better experience of provider communication were less likely to have problems learning about their medical conditions because of difficulty understanding written information.Having better health literacy could mean a better ability to assimilate the content of the provider communication.Shared decisionmaking was not significantly associated with health literacy.Neither provider communication nor shared decision-making were significantly associated with missed appointments.In our prior work, we found that 'forgetting', 'time not working' and 'not having transportation' were the most likely reasons for missing appointments [17].These reasons for missed appointments would not likely be affected by poor provider communication or shared decision making.
Shared decision-making and provider communication were positive and comparable to a national benchmark [53].Our results in the provider communication domain were comparable to the percentages of responding 'Always' (79% to 87%) to the provider communication questions (Table 2) reported in the 2016 national benchmark sample of the CAHPS Clinician & Group Adult Survey [53].For shared decisionmaking, while positive, the percentage of responding 'Definitely yes' to the shared decision making questions (Table 2) had more variation among the three items in the 2016 national benchmark sample (range, 43% to 80%) than the two items in our survey (both approximately 63%).Since the questions in our survey were somewhat different from those in the CAHPS PCMH (Patient Centered Medical Home) item set, results were not quite comparable.In our survey, two shared decision-making items from the CAHPS PCMH item set (i.e.'provider talked about reasons to take a medicine' and 'provider talked about reasons not to take a medicine') were combined into one item named 'provider talked about the pros and cons of each choice'.In addition, unlike in the CAHPS PCMH item set, respondents were not asked to skip the shared decisionmaking items in our survey if they previously reported that providers did not tell them that they had more than one choice for their (or their child's) treatment.
Certain limitations caution interpretation of our study findings.First, there can be recall bias from patient-reported healthcare utilization.While there may be difficulty recalling hospitalizations, these are disruptive events that people are more likely to remember.Better evaluation of the accuracy of selfreported hospitalizations as compared to electronic health record data is an area of future research.Interventions that are focused on improving poor patientprovider communication should be dependent on more precise data to support conclusions.Also, missing appointments were self-reported and likely under-reported by some participants, which may impact the associations between missing appointments and other variables, such as health literacy and provider communication.Second, while selection bias may have occurred from our convenience sample of participants who attended outpatient clinic visits, our participants were approached sequentially, without any selection for disease severity or social factors.The participants in this study are also from a sample that have a pattern of established outpatient care.There is a subpopulation of adults who are high utilizers who rarely, if ever, are seen in an outpatient setting.Further research including this population is needed.Third, there may be concerns of external validity as responses may differ in other areas of the U.S. that were not sampled.Fourth, caregivers may have answered questions about themselves or the child  patient inappropriately.However, research personnel who administered the surveys did not feel there was confusion about whom the questions were concerned with.Fifth, the wording of the questions asked in our survey was modified from the original shared decision-making domain of CAHPS, which may limit interpretation of these findings.Sixth, we were unable to assess disease severity.This can be a very important factor for perceptions about provider communication.Disease severity is an area of current SCD  research as genotype does not always predict severity of sequelae, but is an important component that will need to be considered when evaluating provider communication [54].Seventh, there is a possibility of response bias.Research coordinators were the ones primarily responsible for providing participants with tablets or paper-and-pencil versions of the survey and remained available for questions.In a very few cases, participants may have been handed a tablet by a member of the healthcare team.However, the survey was self-administered and not completed through an interview with the research or clinical staff.The informed consent form contained language that responses to the survey would not have impact on healthcare benefits received.All of this would limit some biases like social-desirability bias but may lead to other biases [55].Finally, other factors for which data were not collected (e.g. insurance coverage) could have contributed to admissions and readmissions and may have affected the significant relationships found with patient experience predictors.However, provider communication could be the root cause of many mediators (e.g.medication adherence, outpatient follow-up) and could ultimately lead to hospitalization.We were unable to test this pathway based on our available data.Future research would include a longitudinal study where the predictor (provider communication), mediators, and the outcome (hospitalization) are measured over time.Evaluating associations of communication with rates of patient adherence in other areas of their care such as filling prescriptions and taking medications would be interesting.Future research might also subsequently focus on developing an intervention to improve provider communication, then evaluate the causal relationship between provider communication and hospitalizations.Due to our cross-sectional survey data, we cannot fully exclude the possibility that our findings represent coincidental correlations, but the current study, unique in its focus on provider communication and patient/family experiences across the lifespan in SCD, lays the foundation for future research.

Conclusions
Our results highlight that provider communication can be a powerful factor in predicting hospitalizations and readmissions.Interventions are needed to help improve patient-provider communication.These interventions have the potential to decrease costly hospitalizations and readmissions in SCD and could potentially translate to lowering acute healthcare utilization in other chronic diseases.Future research evaluating patient-provider communication in chronic diseases can improve our understanding of this important concept and how it affects healthcare utilization, morbidity, and mortality.
Between October 2014 and March 2016, we surveyed a convenience sample of adults with SCD (patients age ≥18 years) and caregivers of children with SCD (patients age < 18 years).Six sickle cell centers across the U.S. participated: Cincinnati Children's Hospital Medical Center, Lurie Children's Hospital of Chicago, University of Tennessee Health Science Center, St. Jude Children's Research Hospital, Vanderbilt University Medical Center, and the University of California San Francisco Benioff Children's Hospital Oakland.Survey participant inclusion criteria included: (1) ability to speak and read English, (

Table 1 .
Socio-demographics and standard measures for participants with sickle cell disease.Caregivers were asked to report for their children under 18 years, except for Highest degree or level of school completed, marital status, spirituality/religiosity, and difficulty paying monthly bills.
a b Percentages may not add up to 100% because of missing data.c UTHSC = University of Tennessee Health Science Center.

Table 2 .
Item descriptive statistics and CFA factor loadings, for provider communication and shared decisionmaking.

Table 3 .
Logistic regression models for hospital admission, readmission, and missed appointment.

Table 4 .
Linear regression model for each of the three health literacy items.
a B = regression coefficient.