Association of Fragmented Readmissions and Electronic Information Sharing With Discharge Destination Among Older Adults

Key Points Question What is the association between fragmented readmissions (ie, readmission to a different hospital than patient was discharged from) and health information exchange (HIE) and discharge destination? Findings In this cohort study of 275 189 admission-readmission pairs of Medicare beneficiaries, fragmented readmissions were associated with a 10% higher odds of discharge to a skilled nursing facility and a 22% lower odds of discharge home with home health vs nonfragmented readmissions. In fragmented readmissions with shared HIE, beneficiaries had 9% to 15% higher odds of discharge home with home health vs fragmented readmissions in which information sharing was not available. Meaning These findings indicate that fragmented readmissions may be associated with less favorable discharge destinations; however, HIE may mitigate information discontinuity present in fragmented readmissions.

EXPOSURES Same hospital vs fragmented readmissions and presence of the same health information exchange (HIE) at the admission and readmission hospitals vs no information shared between the admission and readmission hospitals.

MAIN OUTCOMES AND MEASURES
The main outcome was discharge destination following the readmission, including home, home with home health, skilled nursing facility (SNF), hospice, leaving against medical advice, or dying.Outcomes were examined for beneficiaries with and without Alzheimer disease using logistic regressions.

Introduction
Returning home after leaving the hospital is a major goal for most patients, particularly older adults.
However, more than 40% of Medicare beneficiaries require postacute care following a hospitalization, 1 a number that increased by 50% between 1996 and 2010 2 and has continued to increase since. 3[7][8] Repeated hospitalizations, admissions to postacute care, and readmissions present opportunities for interhospital fragmentation of care, wherein a patient's admission and readmission are to different hospitals.6][17] Because a patient's medical history and record of previous hospitalization may not be available during a fragmented readmission, effective transitions of care may be challenging and may lead to higher rates of nonhome discharge than would be observed in a nonfragmented readmission.
Systems that can bridge the information discontinuity 18 present in fragmented readmissions, namely shared electronic health records (EHRs) and health information exchanges (HIEs), may mitigate some of the negative effects of fragmented readmissions.Previous work has shown that HIEs may be associated with reduced costs, 19 less duplication of procedures and imaging, 20 and lower risk of subsequent readmissions [21][22][23] and fragmented readmissions, 22 among other outcomes. 24e association of information sharing with outcomes may differ across patient populations.If a patient can easily communicate their health history, the effect of outside sources of information on their clinical course may be lessened.However, if patients are not able to communicate their health history, such as is the case among some older adults with Alzheimer disease, electronic information sharing may have an outsize effect on outcomes.Previous work in the outpatient setting has shown that higher outpatient continuity of care resulted in fewer hospitalizations and emergency department visits among older adults with Alzheimer disease, 25 a benefit that may extend to continuity of care with a hospital as well.
In this study, we analyzed the associations between fragmented readmissions and discharge destination among a sample of Medicare beneficiaries.We also assessed whether electronic health information sharing was associated with a patient's odds of various discharge destinations, stratified by Alzheimer disease status.

Data Sources
This cohort study used the 2018 Medicare Provider Analysis and Review file, which was linked to the 2018 Medicare Master Beneficiary Summary and the Chronic Conditions Segment files to obtain additional beneficiary information.Hospital characteristics were obtained from the American Hospital Association (AHA) Annual Survey 26 and the AHA Information Technology (IT) Supplement

Primary Exposure: Fragmented Readmission
We first categorized all admission-readmission pairs as fragmented or as same hospital/ nonfragmented.If the admission and readmission hospitals had different identification numbers in the Medicare data, the readmission was fragmented.If the identification numbers were the same, the readmission was determined to be to the same hospital/nonfragmented, which served as the comparison group for the primary analysis.

Secondary Exposure: Type of Information Sharing
Among fragmented readmissions, we categorized information sharing based on the availability of electronic health information sharing at the admission and readmission hospitals based on the AHA IT Supplement. 27The AHA IT Supplement asks hospitals: "Please indicate your level of participation in a state, regional, and/or local [HIE] or health information organization [HIO]."Answer options were "do not know," "HIE/HIO is not operational in my area," "HIE/HIO is operational…but we are not participating," or "HIE/HIO is operational…and we are participating." 27If the hospital answered that HIE was operational in its area and it was participating in any AHA IT Supplement between 2017 and 2018, it was considered to have HIE capabilities for the purpose of this analysis.If a hospital had different or missing answers across the 2 years of data, we used the answer reflecting the hospital's highest level of information sharing.We took this approach because each year of AHA IT Supplement data had more than 20% missingness in response to this question, but only 3.4% of hospitals that responded to the AHA IT Supplement survey had missing responses for both years.Because not all hospitals that responded to the AHA Annual Survey responded to the AHA IT Supplement survey, 17.1% of admission-readmission pairs in our analytic sample had missing HIE information (eTable 1 in Supplement 1).Justification for this approach is that it is unlikely for a hospital to move from participating in HIE to not participating in HIE during this time.
To determine which HIE a hospital participated in, we examined answers to the question: "Which of the following national health information networks does your hospital participate in?" Potential responses include "your EHR vendor's network…" and "other," which offered a free-text JAMA Network Open | Geriatrics response option.We used these answers to categorize information sharing between different admission and readmission hospitals as participated in the same HIE, participated in different HIEs, 1 or both hospitals did not participate in an HIE, or 1 or both hospitals had missing HIE participation data.
In this analysis, we compared admission-readmission pairs in which both hospitals participated in the same HIE with admission-readmission pairs in which information sharing was not available because the admission and readmission hospitals had different HIEs, the admission and/or readmission hospital did not participate in HIE, or HIE data were missing; these 3 groups were combined to make the reference group.Because a beneficiary could have multiple admissionreadmission pairs in this analysis, they could have a pair in more than 1 category of information sharing.

Outcome: Discharge Destination Following the Readmission
The outcome of interest in this analysis was the discharge destination following the readmission.
Potential discharge destinations were home, skilled nursing facility (SNF), home with home health, hospice, left against medical advice, died during the readmission, or other.Hospice could be either inpatient or home hospice.Other included discharge to law enforcement or psychiatric facilities.

Covariates
Our analysis considered beneficiary demographic and clinical characteristics, as well as characteristics of the readmission hospital.Demographic characteristics included the beneficiary's age, sex, and race and ethnicity (Black, White, or other).Race and ethnicity was identified in the Master Beneficiary Summary and uses Social Security Administration data, 31 which is based on selfreport.Other included Hispanic, Asian, North American Native, other, and unknown.Race and ethnicity were included as covariates in this analysis because patients from minoritized racial and ethnic groups may have disparities in where they are discharged to after hospitalization due to bias in the health care system.Baseline clinical characteristics included a frailty score, Charlson Comorbidity Index score, and the diagnosis-related group of the readmission.The frailty score ranged from 0 to 1, with higher scores indicating more frailty.It was calculated using a deficit accumulation model using 93 claims-based variables.The frailty score was originally developed and validated using ICD-9 codes and has been updated with ICD-10 codes. 32,33admission hospital characteristics included the number of hospital beds (<200, 200-399, Ն400), hospital ownership (government, church, nonprofit, for profit), hospital type (general medical/surgical, other), urban/rural status (metropolitan, micropolitan, rural), and whether the hospital was a teaching hospital.Urban/rural status was identified via the Rural Urban Commuting Area Code 34 of the hospital.Hospitals were classified as teaching hospitals if they reported that they had programs accredited by the American Council of Graduate Medical Education, the American Osteopathic Association, or the Council of Teaching Hospitals or if they were affiliated with a medical school.Hospitals were categorized as either general medical/surgical or other, which included specialty hospitals. 26

Statistical Analysis
Descriptive statistics, including χ 2 tests, t tests, and analysis of variance, were used to compare clinical and demographic characteristics between admission-readmission pairs across categories of information sharing; these were used to inform covariate selection for multivariable models.
Covariates were checked for multicollinearity using the collin function in Stata (StataCorp LLC).
Significance was set at a 1-sided P < .05.
To evaluate whether fragmented vs same hospital/nonfragmented readmissions were associated with discharge destination, we performed unadjusted and adjusted logistic regressions.
Each discharge destination was evaluated using dummy variables.Regression analyses were adjusted for patient demographic characteristics, clinical characteristics, and readmission hospital JAMA Network Open | Geriatrics characteristics.A fully adjusted model using all covariates was also created.Hospital referral region random effects were included in each model to account for geographic variation in SNFs and other discharge destination availability.Regression analyses were stratified a priori by Alzheimer disease status.
To examine the association between information sharing across the admission and readmission hospitals with discharge destination, we repeated unadjusted and adjusted logistic regressions comparing fragmented readmissions in which the admission and readmission hospitals participated in the same HIE with fragmented readmissions in which there was no information sharing available.
As above, robust SEs clustered at the hospital level were used, and regression analyses were stratified a priori by Alzheimer disease status.
We also completed several sensitivity analyses.First, we included being admitted through the emergency department vs another route of admission as a covariate.Second, we removed admission-readmission pairs with missing HIE information from the composite reference group to account for potential misclassification bias introduced by these pairs.To estimate the associations of information sharing compared with nonfragmented readmissions, we compared same HIE and no information shared admission-readmission pairs with nonfragmented readmissions.Additionally, to test whether there was a differential association between information sharing and the type of hospice discharge, we examined home hospice and facility-based hospice as separate discharge destinations.All analyses were performed using SAS, version 9.4 (SAS Institute) and Stata, version 17 statistical software.The data analysis was completed between November 1, 2021, and October 31, 2022.

Sample Development
The 2018 Medicare sample had 8 316 909 claims (Figure 1).After we removed noninpatient claims, interhospital transfers, and observations missing beneficiary identification numbers, we created admission-readmission pairs (n = 2 507 483).After we limited the sample to 30-day readmissions and index admissions for the initial diagnoses of interest and removed pairs in which beneficiaries were listed as deceased after their index admission, 275 189 pairs (268 768 unique patients; mean [SD] age, 78.9 [9.0] years; 54.1% female and 45.9% male; 12.2% Black, 82.1% White, and 5.7% other race and ethnicity) remained in the final sample (Figure 1; Table ).

Hospital and Patient Characteristics
Overall, 31.6% of pairs had fragmented readmissions.Compared with beneficiaries with fragmented readmissions that shared an HIE and those that did not, beneficiaries with same hospital/ nonfragmented readmissions tended to be older (mean [SD] age, 78.9 [9.0] vs 77.9 [8.8] and 78.3 [8.7] years; P < .001)and were more likely to be White (83.2% vs 80.0% and 79.6%; P < .001)vs Black (11.5% vs 14.8% and 13.5%) and other race and ethnicity (5.2% vs 5.1% and 6.7%).Distribution by sex between same hospital/nonfragmented and fragmented readmissions that shared and did not share an HIE was 54.7%, 51.7, and 53.5% female patients and 45.3%, 48.3%, and 46.5% male patients, respectively.Beneficiaries with fragmented readmissions to hospitals that shared an HIE were more likely to be in a metropolitan area (94.9% vs 84.7%; P < .001)and to have a readmission to a teaching hospital (79.3% vs 66.6%; P < .001)than beneficiaries with fragmented readmissions to hospitals that did not share an HIE (Table ).Beneficiary demographic and clinical characteristics as well as readmission hospital characteristics by subcategories of hospitals that did not share an HIE (ie, different HIEs, no HIE, HIE information missing) are shown in eTable 1 in Supplement 1.

Information Sharing and Discharge Destination
Among beneficiaries with fragmented readmissions, 14.3% were readmitted to a different hospital that shared an HIE with the admission hospital.In fully adjusted models, beneficiaries without Alzheimer disease and fragmented readmissions in which the admission and readmission hospital shared the same HIE had 14% lower odds of being discharged home (AOR, 0.86; 95% CI, 0.80-0.92)but 9% higher odds of being discharged home with home health (AOR, 1.09; 95% CI, 1.04-1.16)compared with beneficiaries with fragmented readmissions where the admission and readmission hospitals did not share an HIE (Figure 2; eTable 4 in Supplement 1).Beneficiaries with Alzheimer disease similarly had 15% higher odds of discharge home with home health when readmitted to a different hospital that shared an HIE with the index hospital (AOR, 1.15; 95% CI, 1.01-1.32)compared with hospitals that did not share HIEs.Additionally, beneficiaries with AD had 23% lower odds of dying during the readmission if the admission and readmission hospitals shared an HIE (AOR, 0.77; 95% CI, 0.61-0.98)compared with fragmented readmissions where the hospitals did not share an HIE (Figure 2; eTable 5 in Supplement 1).

Sensitivity Analysis
When we included emergency department origin as a covariate, removed fragmented admissionreadmission pairs that had missing HIE data, and compared fragmented admission-readmission pairs that shared an HIE and those that did not with same hospital/nonfragmented admissionreadmission pairs, the trends were similar to those seen in the main analysis (eTables 6-10 in Supplement 1).When hospice discharge was separated into home and facility-based hospice, beneficiaries without Alzheimer disease with fragmented readmissions where an HIE was shared had 15% lower odds of discharge to home hospice (AOR, 0.85; 95% CI, 0.74-0.97)but 15% higher odds of discharge to a facility-based hospice (AOR, 1.15; 95% CI, 1.02-1.30)after adjusting for demographic, clinical, and hospital characteristics (eTable 11 in Supplement 1).

Discussion
In this cohort study, fragmented readmissions were common (>30% of readmissions), and the association between these fragmented readmissions and discharge destination was notable.For older adults with and without Alzheimer disease, compared with readmission at the same hospital, a fragmented readmission was associated with higher odds of discharge to an SNF by 6% to 10%, lower odds of discharge home with home health and to hospice by 22% to 26% and 5% to 14%, respectively, and higher odds of leaving against medical advice between 64% and 85%.
There are patient-, hospital-, and system-level reasons why a patient may have a fragmented readmission.At the patient level, fragmented readmissions may be influenced by a patient's location or geography, [35][36][37] their family's wishes, acuity of their illness, and dissatisfaction with previous care. 38A patient's diagnosis and illness severity, the quality of care, need for specialist care, and insurance payer can affect whether they return to their previous hospital.Previous work has shown that at the hospital and health care system level, density of hospitals and ambulance use are also associated with fragmented readmissions. 39,40tably, more than 85% of patients with a fragmented readmission were readmitted to hospitals that did not share an HIE with the admission hospital.While information sharing via HIE in fragmented readmissions was associated with discharge destination, it was not enough to completely mitigate the association between fragmentation and discharge destination in most cases.
One exception to the trend was home health; beneficiaries with and without Alzheimer disease who had fragmented readmissions in which both the admission and readmission hospitals shared an HIE had 9% to 15% increased odds of going home with home health compared with fragmented readmissions in which the admission and readmission hospitals did not share an HIE.This association may be due to the electronically available health information; for example, accessing notes from a previous hospitalization or from outpatient visits may provide information about a patient's goals of care 41,42 (ie, to avoid discharge to an SNF) or their baseline functional status. 43agmented readmissions were associated with 24% higher odds of dying during the readmission in beneficiaries without Alzheimer disease, but there was no statistically significant difference in the odds of dying during the readmission in beneficiaries with Alzheimer disease.When those fragmented readmissions occurred at hospitals that shared an HIE, beneficiaries with Alzheimer disease had a 23% lower odds of dying during the readmission compared with fragmented readmissions to hospitals that did not share an HIE (AOR, 0.77; 95% CI, 0.61-0.98).Older adults with Alzheimer disease may be particularly sensitive to poor care coordination, so the observation of a positive association with in-hospital mortality when information exchange is present is encouraging.
While our data set does not define HIE pairs beyond hospitals, 1 potential reason for this observation may be that information exchange is not limited to hospitals; HIEs and other electronic information exchange systems also exist between the inpatient and outpatient settings and between hospitals and long-term-care facilities, which could further facilitate care coordination.
These results have implications beyond an individual patient and discharge destination.Beyond the paramount importance of being able to honor a patient's wishes, different discharge destinations add different stressors to patients and the health care system.Discharges to SNFs, for example, are associated with higher costs 3,44 and an increased risk of subsequent readmission. 45,46Discharges to home with home health support are increasing, 47 and patients and payers may prefer these to inpatient postacute care facilities.To provide each patient with the best possible care, including preventing unnecessarily intense care, clinicians should be encouraged to consider the effect of fragmentation and to seek additional information about care delivered elsewhere.
Hospitals that have HIE functionality may be different than those that do not.Previous work has found that hospitals that are part of health care systems are more likely to participate in electronic information exchange, 48,49 as are hospitals in areas with a dominant EHR vendor, 50 larger hospitals, and nonprofit hospitals. 51Some of these characteristics may be associated with quality of care delivered, 52 the rate of fragmented readmissions, and discharge destination.

Limitations
This study has several limitations.A key limitation of the analysis is that hospital-level availability of HIE may not reflect clinician use of HIE.This limitation is common in observational studies using secondary data on HIE topics. 22,23,53,54In previous studies, a lack of clinically useful information 55,56 was 1 reason that providers reported accessing HIEs in only 10% to 20% of clinical encounters. 19,57wever, what may be clinically useful in the care of a patient at the beginning of an acute care episode may be different from what is useful for discharge planning.Specific investigations into how HIEs are used in care coordination vs clinical decision making could shed more light on this issue.As HIE use continues to expand, in part due to federal priorities, 58 it will be increasingly important to understand how and when clinicians use HIE and what effect the information contained in the HIE has on clinical decision making, care coordination, and patient outcomes.
Another limitation is how hospitals responded to the AHA IT Supplement survey.Despite including 2 years of data, several hospitals had missing answers to the questions around HIE

Question
What is the association between fragmented readmissions (ie, readmission to a different hospital than patient was discharged from) and health information exchange (HIE) and discharge destination?Findings In this cohort study of 275 189 admission-readmission pairs of Medicare beneficiaries, fragmented readmissions were associated with a 10% higher odds of discharge to a skilled nursing facility and a 22% lower odds of discharge home with home health vs nonfragmented readmissions.In fragmented readmissions with shared HIE, beneficiaries had 9% to 15% higher odds of discharge home with home health vs fragmented readmissions in which information sharing was not available.Meaning These findings indicate that fragmented readmissions may be associated with less favorable discharge destinations; however, HIE may mitigate information discontinuity present in fragmented readmissions.

Downloaded From: https://jamanetwork.com/ on 10/21/2023 from
2017 to 2018.The study was approved by the institutional review board of the Emory University School of Medicine.with a waiver of informed consent due to no more than minimal risk to human participants.The study followed the Strengthening the Reporting of Observational Studies in

eTable 2 .
Association of Fragmentation Status and Discharge Destination From Readmission, Medicare Beneficiaries Without Alzheimer Disease, 2018 eTable 3. Association of Fragmentation Status and Discharge Destination From Readmission, Medicare Beneficiaries With Alzheimer Disease, 2018 eTable 4. Association of Health Information Exchange Status and Discharge Destination From Readmission, Medicare Beneficiaries Without Alzheimer Disease, 2018 eTable 5. Association of Health Information Exchange Status and Discharge Destination From Readmission, Medicare Beneficiaries With Alzheimer Disease, 2018 eTable 6. Association of Health Information Exchange Status and Discharge Destination From Readmission, Medicare Beneficiaries Without Alzheimer Disease, 2018 eTable 7. Association of Health Information Exchange Status and Discharge Destination From Readmission, Medicare Beneficiaries With Alzheimer Disease, 2018 eTable 8. Association of Health Information Exchange (HIE) Status and Discharge Destination From Readmission, Missing HIE Information Pairs Removed, Medicare Beneficiaries With and Without Alzheimer Disease, 2018 eTable 9. Association of Health Information Exchange Status and Discharge Destination From Readmission Compared With Same Hospital/Nonfragmented Readmissions, Medicare Beneficiaries Without Alzheimer Disease, 2018 eTable 10.Association of Health Information Exchange Status and Discharge Destination From Readmission Compared With Same Hospital/Nonfragmented Readmissions, Medicare Beneficiaries With Alzheimer Disease, 2018 eTable 11.Association of Fragmented Readmissions and Discharge to Home Hospice or Facility-Based Hospice From Readmission, Medicare Beneficiaries, 2018