Introduction

Increasing national emphasis has been placed on high-value, high-quality patient care, with a focus on reducing hospital readmissions, cost, and waste. When patient care is divided among multiple organizations and providers, fragmentation of care occurs. Fragmentation of care in the outpatient setting has been associated with a range of negative patient outcomes, including increased morbidity,1, 2 more duplicate medications and more drug interactions,3, 4 redundant imaging tests,5 as well as more frequent admissions to the hospital.6,7,8,9 Interhospital fragmentation of care, when a patient is readmitted to a different hospital than the one they were previously discharged from, has not been widely studied, and its effects on patient outcomes are not well defined.

Among the few existing studies, interhospital care fragmentation has been reported to be associated with decreased patient satisfaction,10 longer length-of-stay,11,12,13,14 increased likelihood of discharge to a care facility,15 increased costs,16, 17 and increased mortality both in the hospital and following discharge from a fragmented readmission.13, 14, 18,19,20 These studies have been conducted in diverse populations with variable definitions of care fragmentation, and no previous efforts have been made to synthesize this information.

The purpose of this review was to determine if there was an association between fragmentation of hospital care and mortality, hospital length-of-stay, cost, and readmission risk in adult patients. It is essential for health professionals, patients, and other stakeholders to understand the impact of interhospital care fragmentation on patient outcomes in order to more appropriately evaluate the value and quality of care provided, as well as the outcomes patients experience. This systematic review sheds light on this area and identifies important gaps in the evidence around the impact of care fragmentation.

Methods

Registration, Protocol, and Disclosures

This systematic review was registered with Prospero (CRD42018094849) and adheres to PRISMA guidelines. The authors declare that they do not have any conflicts of interest.

Study Search and Selection

We searched MEDLINE, the Cochrane Library, EMBASE, and the Science Citation Index for English articles published in peer-reviewed journals through April 30, 2018. The following terms were used in our search string: “continuity of patient care,” “patient readmission,” “outcomes,” “fragmentation of care,” “fragmented care,” and “discontinuity of care.” A full example of our search string is included in Appendix 1. Eligible studies were cohort, case control, and cross-sectional studies, conducted in adults that reported an association between fragmentation of care and one or more of the following patient outcomes: mortality, hospital length of stay, cost, and healthcare utilization. Experimental studies, studies on pediatric patients, and qualitative studies were excluded.

One author (ST) reviewed each title and abstract for inclusion. Then two authors (ST and KS) reviewed the selected abstracts and selected studies for full text review. Then, both authors selected a set of studies to be included and disagreements were resolved by consensus. The reference list of the included studies was also hand-reviewed to identify potential additional articles. Both authors then screened these articles for inclusion.

Data Extraction and Outcomes

From the included articles, we recorded the number of patients included in each study, patient characteristics, location of the study or data source, and the definition of care fragmentation used. For each paper, we extracted the odds ratio (OR) or adjusted odds ratio (AOR) for each outcome, when reported. If OR/AOR were not available, we contacted the corresponding author of the article or calculated them from the available data. Each author was contacted up to 3 times over a 4-week period via e-mail. Given that the patient outcomes were heterogeneous, we also abstracted additional data based on the specific outcome; for instance, for studies examining mortality, the mortality follow-up period was recorded. This was not extracted for the studies examining length-of-stay or cost.

Risk of Bias Assessment

We used the Newcastle-Ottawa Scale to assess the quality and risk of bias in each study.21 The scale includes eight questions that address participant selection, the comparability of the cohorts and outcomes. For the selection domain, we assessed if the exposed cohort was representative of the study’s population (1 star) or not representative (0 stars), whether the non-exposed cohort was drawn from the same community as the exposed cohort (1 star) or not (0 stars), how the exposure was ascertained (secure record or interview 1 star, written self-report, no description; or other 0 stars), and if it was demonstrated that the outcome of interest was not present at the start of the study (yes 1 star; no 0 stars). Next, we examined whether the study controlled for age, sex, and marital status (1 star) and/or other factors (1 star), or if the cohorts were not comparable based on the design or analysis plan of the research (0 stars). Finally, we evaluated if the outcome was assessed by independent blind assessment or record linkage (1 star), or self-report/other (0 stars), if time to follow-up was sufficient for the outcome to occur (yes 1 star; no 0 stars), and how adequate the follow-up of cohorts was—complete (1 star), less than 20% lost to follow-up/those lost to follow up were not different from those followed (1 star), or follow up rate less than 80%, no description of those lost, or no statement (0 stars). The total number of stars was counted: a “good” quality study had 3–4 stars for selection, 1–2 stars in comparability, and 2–3 stars in outcome/exposure, a “fair” study had 2 stars for selection, 1–2 stars for comparability, and 2–3 stars in outcome/exposure, and a “poor” quality study had 0–1 star in selection domain or 0 stars in comparability or 0–1 stars in outcome/exposure.

Due to heterogenous study populations and a small number of articles for several outcomes, we decided to not pool the results for each outcome in a meta-analysis; additional details are provided in the “Results” section.

Results

Study Characteristics

We identified 4682 unique abstracts from our literature search. Of these, 79 full texts were reviewed and 16 were included (Fig. 1). The most common reasons for exclusion were that studies analyzed outpatient continuity of care or fragmentation between physicians or emergency rooms (n = 4592), and that studies did not assess one of the patient outcomes of interest (n = 4056). The most common reason for exclusion of full-text articles reviewed was that they did not assess inpatient fragmentation (n = 30). Six additional articles were identified in the references of included articles and were also included in the review.

Figure 1
figure 1

PRISMA flowsheet for article selection.

Full size image

Of the 22 included studies, all but one were retrospective observational studies using secondary data. Eighteen studies reported mortality, 6 reported hospital length of stay, 4 reported hospital cost, and 3 reported hospital readmission rates. Some studies were included in multiple groups if they evaluated multiple outcomes. Duration of study follow-up ranged from 30 days to 9 years. Nineteen of the studies were in the USA, 2 in Canada, and 1 in Israel. Characteristics of each study are included in Tables 1, 2, 3, and 4. The 18 studies that evaluated patient mortality included 4,248,826 participants with a wide array of diagnoses as described in Table 1. Hence, we determined that study populations were too heterogenous to perform a meta-analysis for this outcome. The other outcomes were limited similarly by heterogenous study populations and a small number of articles in each category.

Table 1 Studies with Mortality as an Outcome
Full size table
Table 2 Studies with LOS as an Outcome
Full size table
Table 3 Studies with Cost/Charges as an Outcome
Full size table
Table 4 Studies with Readmissions as an Outcome
Full size table

Study Risk of Bias

All studies generally had a clear research question, study population, and appropriate exposures and outcomes. Most of the studies (86%) were characterized as good quality based on the Newcastle-Ottawa quality assessment tool and 14% were fair quality (Supplemental Table 1).

Mortality

Eighteen of the included articles utilized mortality as a primary endpoint (Table 1). All but two22, 23 defined care fragmentation as a readmission to a different hospital than the index admission. Seven studies examined in-hospital mortality, 6 studies examined mortality 30 days post-discharge, and 9 studies examined mortality follow-up periods 90 days or longer (some studies included multiple endpoints). Of the studies examining in-hospital mortality, 4 report fragmented readmissions was associated with 10% to two times higher mortality odds (AOR 1.11–2.1).11, 12, 15, 24 One study had variable results depending on the factors adjusted for in analyses,25 while two did not find significant associations between fragmentation and mortality.26, 27 Of the studies examining mortality post-discharge, 4 reported that fragmentation was associated with 6% to over three-fold higher odds for death within 30 days (AOR 1.06–3.62).20, 28,29,30 Among 6 studies, fragmentation was associated with no increased odds to five-fold higher odds for death beyond 30 days after a fragmented readmission (AOR 1.0–5.66).18, 19, 22, 23, 30, 31

Overall, 11 studies looked at mortality following a fragmented postoperative readmission, and 7 of these included cancer-related surgeries. Among the 5 studies that included only patients with cancer,15, 25, 30,31,32 3 showed post-operative fragmentation associated with 18% to five-fold higher mortality odds (AOR 1.18–5.66). In 4 studies that included both cancer-related and non-cancer-related surgeries,19, 20, 26, 27 only 2 found post-operative care fragmentation was associated with increased odds of mortality, but the increase was notable: 50 to 75% higher mortality odds (OR 1.57–1.75).

Length of Stay

Six of the included articles used length-of-stay(LOS) as an outcome (Table 2). All defined care fragmentation as a readmission to a different hospital than the index admission. The studies varied widely in how they reported LOS, with some reporting median or mean LOS and others reporting the hazard ratio for discharge on a given day. One study reported a longer mean LOS of less than 1 day in fragmented readmissions versus non-fragmented readmissions. (11.6 days vs. 11.0 days),12 and 2 studies reported a hazard ratio of 10–15% less for discharge on a given day in fragmented readmissions (HR 0.90–0.85).11, 33

Costs and Charges

Four of the included studies examined cost or charges as a primary or secondary outcome (Table 3). These studies varied in their timeframe of measurement: 2 examined in-hospital costs or charges, 1 examined costs for a year following cancer diagnosis, and 1 examined costs for 90 days following surgery. The differences in costs/charges between fragmented and non-fragmented readmissions ranged from $270 to $22,000. Only 1 found statistically significantly higher costs in patients with fragmented readmissions in adjusted models, with a median cost of a fragmented readmission of $8568 and the median cost of a non-fragmented readmission of $8298.17

Hospital Readmission

Three papers examined the association of care fragmentation and hospital readmission (Table 4). They found that patients with fragmented readmissions were between 16% and twice as likely to experience a third admission than patients who did not experience interhospital care fragmentation (AOR 1.16–2.3).23, 25, 34 Notably, one subgroup who accessed care at the VA and also held Medicare and Medicaid insurance was reported to have an AOR of 13.6 (95% CI 6.1–30.1) for a third admission following a fragmented readmission.23

Discussion

In this systematic review, we analyzed 22 studies examining patient outcomes during and following a fragmented hospital readmission. Overall, 11 of the 18 studies that examined patient mortality found increased odds of mortality during or following a fragmented readmission, 1 had mixed results, and 6 did not show a difference in the odds of mortality. Three of the 6 studies reporting length-of-stay found a longer length-of-stay or decreased hazard ratio for discharge in fragmented readmissions. Among studies examining cost/charges, 1 study reported higher costs during a fragmented readmission, 1 had mixed results, and 2 reported higher costs/charges during or after non-fragmented readmissions. Finally, all three papers that investigated the odds of a third hospital readmission following a fragmented readmission found higher odds of a third admission in patients who experienced fragmentation, compared to patients who had a non-fragmented readmission. Overall, this systematic review suggests that fragmented hospital readmissions contribute to increased mortality, longer length-of-stay, and increased risk of future readmission to the hospital.

Our systematic review adds to the previous reviews of the association between continuity of care and patient outcomes.2, 35 Previous reviews have focused on outpatients only or on continuity during transitions of care (i.e., inpatient to outpatient). Van Walraven et al.2 performed a systematic review of studies examining continuity of care, healthcare use, and patient satisfaction, primarily focused on continuity following hospital discharge and in the outpatient setting. They found that there were significant associations between improved provider continuity, decreased healthcare use, and increased patient satisfaction. Our review is the first to synthesize the literature on fragmented hospital readmissions and adds to the existing literature by including papers that address different types of continuity.23, 33

Our study is subject to limitations. Our definition of interhospital care fragmentation was narrow in an effort to make studies as comparable as possible, which may have excluded some studies looking at other types of hospital-based fragmentation (i.e., fragmentation across providers in a hospital or interhospital transfers). Notably, many of the included studies examined very specific populations (i.e., postoperative from cancer surgery), making the results less generalizable. Finally, because of the retrospective nature of most of the included studies, we are unable to know whether fragmentation was the cause or an effect of a patient’s poor outcomes.2 This is especially true regarding hospital readmissions—if the patient had a poor experience during their index admission, they may actively seek care elsewhere, leading to a fragmented readmission. If they remain dissatisfied following the readmission or their health has not improved, they may seek care at yet another hospital.

In conclusion, this systematic review suggests that interhospital care fragmentation may be a contributor to increased mortality, longer lengths-of-stay, and readmissions. These results support the need for improved care coordination and should increase provider awareness of the role of interhospital care fragmentation on our patients and hospitals. Interhospital care fragmentation should be considered when designing interventions to reduce duplicative care and waste, and when evaluating provider and hospital outcomes.