Trends in Prolonged Hospitalizations in the United States from 2001 to 2012: A Longitudinal Cohort Study

Abstract

Background

Health policy debate commonly focuses on frequently hospitalized patients, but less research has examined trends in long-stay patients, despite their high cost, effect on availability of hospital beds, and physical and financial implications for patients and hospitals.

Methods

Using the National Inpatient Sample, a nationally-representative sample of acute care hospitalizations in the United States, we examined trends in long-stay hospitalizations from 2001–2012. We defined long stays as those 21 days or longer and evaluated characteristics and outcomes of those hospitalizations, including discharge disposition and length of stay and trends in hospital characteristics. We excluded patients under 18, and those with primary psychiatry, obstetric or rehabilitation diagnoses and weighted estimates to the U.S. population.

Results

Prolonged hospitalizations represented only 2% of hospitalizations, but approximately 14% of hospital days and incurred estimated charges of over $20 billion dollars annually. Over time, patients with prolonged hospitalizations were increasingly younger, male, and of minority status, and these hospitalizations occurred more frequently in urban, academic hospitals. Inhospital mortality for patients with prolonged stays progressively decreased over the 10-year period from 14.5 to 11.6% (p<0.001 for trend in grouped years), even accounting for changes in demographics and comorbidity.

Conclusions

The profile of patients with prolonged hospitalizations in the United States has changed, although their impact remains large, as they continue to represent one out of every 7 hospital days. Their large number of hospital days and expense increasingly falls upon urban academic medical centers, which will need to adapt to this vulnerable patient population.

Introduction

Attention is increasingly focused on high-cost patients in the current healthy policy dialogue.^1,2 Much of this debate centers on frequently-hospitalized patients, who comprise a large proportion of health care costs. As of yet, little research has examined trends in long-stay patients, despite their high cost, effect on availability of hospital beds, and physical and financial implications for patients and hospitals.

Analyses examining long-stay hospitalizations, defined as over 21 days, between 1980 and 1990 found that these hospitalizations represented 5% of all discharges from acute hospitals, but utilized 25% of hospital days.³ With a definition of 30 days, in a Canadian study, 40% of acute hospital days were consumed by the 5% of prolonged hospitalizations.⁴ The implication for subsequent health care spending is not fully understood, as at least one study shown that patients with frequent high-cost hospitalizations in one year may not have sustained high-cost services.⁵ The importance of prolonged hospitalizations has persisted despite introduction of diagnosis-related groups DRGs in 1983 in an attempt to reduce the federal government’s exposure to high hospital costs, which ultimately decreased lengths of stay.⁶

To better understand the landscape of long-stay hospitalizations in the modern era, we examined medical and surgical patients with stays over 21 days using the National Inpatient Sample (NIS) from 2001–2012. The NIS provides a nationally-representative sample of hospitalizations from across the United States.

Methods

Design

We analyzed data from the 2001–2012 National Inpatient Sample (NIS) for a longitudinal cohort study. The NIS is the largest all-payer inpatient care database currently available in the United States.⁷ The NIS samples from the State Inpatient Databases participating in the Healthcare Cost and Utilization Project (HCUP), which is a family of healthcare databases sponsored by the Agency for Healthcare Research and Quality (AHRQ). The NIS approximates a 20%-stratified sample of discharges from U.S. community hospitals, excluding rehabilitation and long-term acute care hospitals. The NIS is a sample of inpatient discharges from American community hospitals, defined as "all non–Federal, short–term, general, and other specialty hospitals, excluding hospital units of institutions." Included among community hospitals are specialty hospitals, as well as public hospitals and academic medical centers. Starting in 2005, the American Hospital Association included long term acute care facilities with average lengths–of–stay less than 30 days in the definition of community hospitals. Institutional Review Board exemption was granted by Beth Israel Deaconess Medical Center.

Study Sample

The NIS includes data on primary and secondary discharge diagnoses according to the International Classification of Diseases, 9^th revision, Clinical Modification (ICD-I-CM).⁸ For this analysis, we defined long stays as any hospitalization documented in the NIS among patients 18 years of age or older that was 21 days or longer. Of note, to evaluate a primarily medical/surgical cohort, hospitalizations among patients discharged to psychiatric facilities or those with psychiatric, obstetric, or rehabilitation diagnoses according to HCUP Clinical Classifications Software categorization scheme were excluded.

Study variables

Patient characteristics collected in NIS include age, sex, race/ethnicity, and primary payer. Hospital characteristics collected include U.S. census region, teaching status, urban or rural location, and hospital bed size. Chronic comorbidities were classified using Comorbidity Software maintained by HCUP to produce an Elixhauser Comorbidity Index score.⁹ Outcome and utilization measures examined include length-of-stay, in-hospital death, disposition, and the cost of hospitalization. Total hospital costs for each hospitalization were calculated by converting total hospital charges to costs using the HCUP cost-to-charge ratio, which is based on hospital account reports from the Centers for Medicare & Medicaid Services (CMS).¹⁰ All hospital costs are reported in 2012 dollars, with inflation adjusted using the Gross Domestic Product (GDP) Price Index from the U.S. Department of Commerce, Bureau of Economic Analysis with 2012 as the index base.

Statistical Analysis

Patient and hospital characteristics are reported as means with standard errors for continuous variables and proportions for binary or categorical variables. Non-normally distributed outcomes were reported as medians with interquartile ranges. To account for the complex sampling design in NIS, all data were weighted to reflect national estimates, and sample sizes and weighted percentages are reported. We then estimated a logistic regression model for mortality with the year as the primary covariate of interest, adjusting for age, sex, and number of comorbidities. In the case of incomplete data, we performed the regression analysis with complete case analysis. Odds ratios (ORs) and 95% confidence intervals (CIs) are reported. Data were analyzed with SAS 9.4 (SAS Institute, Inc., Cary, NC) and SUDAAN (RTI International, Research Triangle Park, NC) to account for the complex sampling design.

Results

Overall, an estimated 5.8 million long-stay hospitalizations in the U.S. occurred during this 10-year period, representing approximately 2% of all hospitalizations, excluding obstetric, psychiatric, rehabilitation and pediatric hospitalizations. As seen in figure 1, these accounted for approximately 14.6% of all acute hospital days annually. Consistent with national trends in length of stay, both the number of long stays and the percentage of hospital days occupied by long stay patients declined over the study period. The total charges for these hospitalizations exceeded $20 billion annually, although charges declined from $28.2 billion between 2001–2004 to $23.4 billion between 2009 and 2012.

Figure 1.

Number of prolonged-stay admissions and proportion of total hospital days occupied, 2001–2012

Patient Demographics

Table 1 provides demographic characteristics of prolonged-stay hospitalized patients in the U.S. over the 10 years included in these analyses. Over time, such patients were generally younger, more likely to be male, and less likely to be Caucasian. The change in the age distribution for the prolonged-stay patients represented an increase in patients between the ages of 45 and 64, and a decrease in those between the ages of 75–84. Insurance coverage also changed, with more patients covered by Medicaid or without insurance, and fewer patients with Medicare as in Prolonged Hospitalizations their primary insurance coverage. Over the ten years, long-stay patients also had increasing degrees of chronic illness as reflected in the number of comorbidities documented.

Table 1.

Demographic Characteristics of Long-Stay Hospitalizations in the US, 2001–2012¹

	2001–2004 (n=438,821)	2005–2008 (n=419,137)	2009–2012 (n=335,424)	p-value2
Age	64.6 (0.3)3	63.6 (0.3)	62.6 (0.2)	<0.001
Female (n=1,193,171)	216,264 (49.3%)	197,787(47.2%)	148,703(44.3%)	<0.001
Race/ethnicity (n=989,655)
White	228,601 (65.9%)	219,587 (65.7%)	191,869 (62.5%)	0.004
Black	64,869 (18.6%)	59,287 (17.6%)	60,201 (19.7%)
Hispanic	34,186 (9.7%)	35,340 (10.4%)	31,714 (10.4%)
Other	20,056 (5.9%)	21,224 (6.3%)	22,721 (7.4%)
Primary payer (n=1,190,792)
Medicare	259,711 (59.3%)	237,246 (56.7%)	178,766 (53.5%)	<0.001
Medicaid	55,245 (12.7%)	58,891 (14.1%)	53,908 (16.1%)
Private	92,736 (21.2%)	87,216 (20.9%)	72,406 (21.6%)
No insurance	18,444 (4.2%)	21,313 (5.1%)	18,635 (5.6%)
Other	11,716 (2.6%)	13,665 (3.2%)	10,894 (3.2%)
Chronic Comorbidities (n=1,193,382)	2.54 (0.03)	2.78 (0.03)	3.63 (0.04)	<0.001

¹Individual variable numbers differ due to missing data.

²P values refer to comparison between time periods

³Continuous variables presented as mean (SE), categorical as unweighted N (weighted %)

The most common principal discharge diagnoses for the long-stay patients are seen in Table 2. Consistent with its acuity, septicemia was the most frequent single diagnosis. The top ten most frequent principal diagnoses encompassed approximately 40% of all long-stay hospitalizations over the time period studied.

Table 2.

Most Prevalent Primary Diagnoses among Long-Stay Hospitalizations in the US, 2001–2012

Diagnosis	Sample Size	Weighted Size	Percent of total weighted sample
Septicemia	101249	483902	8.5
Respiratory failure; insufficiency; arrest	54982	260763	4.6
Pneumonia (excluding TB and STDs)	52127	247878	4.4
Complication of device; implant or graft	48942	233092	4.1
Acute cerebrovascular disease	48069	229040	4.0
Congestive heart failure, non-hypertensive	45320	215831	3.8
Complications of surgical procedures or medical care	40043	190535	3.3
Acute myocardial infarction	35308	167955	3.0
Diabetes mellitus with complications	28059	133453	2.3
Intracranial injuries	26238	125043	2.2
Secondary malignancies	26028	124059	2.2
Leukemia	24464	117562	2.1
Intestinal obstruction without hernia	24265	115307	2.0
Coronary atherosclerosis and other heart disease	22687	107803	1.9
Aspiration pneumonitis; food/vomitus	21629	102691	1.8
Acute and unspecified renal failure	21385	101932	1.8

Discharge Outcomes

As seen in Table 3, the average length of stay decreased significantly but modestly over time, from 27.0 to 26.8 days. The average charge per hospitalization increased significantly despite the slightly shorter length of stay. This increase coincided with a higher number of procedures performed. Of perhaps greatest importance, in-hospital mortality decreased over the 12-year period, both crudely and adjusted for age, sex, and comorbidity. Compared with the reference years of 2001–2004, the adjusted odds ratios for in-hospital death were OR 0.91 (95% CI 0.88, 0.93) in 2005–2008 and OR 0.75 (95% CI 0.73, 0.78) in 2009–2012. Discharge disposition also changed, with fewer patients being discharged to home without support, and more to post-acute care facilities or with home health services.

Table 3.

In-Hospital Outcomes among Long-Stay Hospitalizations in the US, 2001–2012

	2001–2004	2005–2008	2009–2012	p-value
Discharges (Sample N, weighted N)	438,821 (2,071,191)	419,137 (1,986,495)	335,424 (1,631,677)	---
Length of Stay1	27.0 (33.0)	27.0 (32.9)	26.8 (32.5)	0.01
Cost/discharge	$54,492 ($53,765)	$59,709 ($55,692)	$62,7253($57,477)	<0.001
Disposition
Routine	103,363 (23.7%)	87,999 (20.9%)	68,132 (20.4%)	<0.001
Transfer to short- term hospital	14,612 (3.4%)	14,083 (3.4%)	11,758 (3.5%)
Other transfer2	184,257 (42.3%)	184,679 (44.2%)	155,849 (46.5%)
Home health care	68,609 (15.7%)	73,953 (17.7%)	58,483 (17.4%)
AMA	1,475 (0.3%)	1,547 (0.4%)	1,262 (0.4%)
Died	63,332 (14.5%)	55,780 (13.3%)	38,872 (11.6%)
Discharged alive but destination unknown	600 (0.1%)	638 (0.1%)	515 (0.2%)
Procedures per Hospitalization	5.0 (3.7)	5.22 (4.8)	5.6 (5.6)	<0.001

¹Continuous outcomes (LOS, cost/discharge, and procedures) all reported as median (IQR). P-value reported for log-transformed outcome variable.

²Includes skilled nursing facility, intermediate care, and another type of facility

Hospital Characteristics

As seen in table 4, we observed a pronounced shift in the type of hospital where long hospital stays occurred, with a significant increase over the ten years toward urban teaching hospitals and away from both rural hospitals and non-academic hospitals. We observed no significant change in the geographic distribution of patients.

Table 4.

Hospital Characteristics of Long-Stay Hospitalizations in the US, 2001–2012

	2001–2004 (n=438,821)	2005–2008 (n=419,137)	2009–2012 (n=335,424)	p-value
Hospital location/teaching status (n=1,187,886)
Rural	30,988 (7.4%)	25,880 (6.2%)	16,277 (4.9%)	<0.0001
Urban non-teaching	170,126 (38.1%)	155,345 (36.7%)	103,732 (31.6%)
Urban teaching	237,693 (54.6%)	237,601 (57.1%)	210,244 (63.5%)
Hospital size1 (n=1,187,886)
Small	40,906 (8.9%)	35,683 (8.0%)	31,095 (9.2%)	0.27
Medium	98,130 (22.4%)	88,703 (21.4%)	65,483 (19.9%)
Large	299,771 (68.7%)	294,440 (70.6%)	233,675 (70.8%)
Hospital region (n=1,193,382)
Northeast	116,444 (27.1%)	109,370 (26.7%)	84,301 (25.4%)	0.45
Midwest	75,323 (18.0%)	66,102 (16.4%)	58,609 (17.1%)
South	175,133 (38.4%)	162,625 (37.9%)	129,858 (38.8%)
West	71,921 (16.5%)	81,040 (18.9%)	62,656 (18.7%)

¹NIS Definition of hospital size varies by region, urban or rural location and teaching status.¹⁸

Discussion

Over the past 10 years, the profile of a prolonged hospitalization in the United States has changed, both for patients and the hospitals where they are hospitalized. Their charges and procedures per hospitalization have increased, with a concomitant drop in their length of stay and inpatient mortality. These changes may have important implications for hospitals in the current health care environment.

The demographic characteristics of patients with prolonged hospitalizations changed significantly, with a move toward younger, minority men with public insurance. This finding is consistent with a single-center study which demonstrated that Medicaid insurance and younger age were independent risk factors for length of stay over 21 days.¹¹ Of note, some of these demographic trends were also apparent in the population of all hospitalized patients. The reason for this demographic shift may be found in increasing burden of comorbidities for younger patients. In frequently readmitted patients, multiple comorbidities increase the risk for recurrent hospitalization, and it seems likely that similar processes act to prolong hospitalizations as well.¹² Long stays, however, may expose patients to particular financial risk, possibly reflected in the recent rise in medical bankruptcy,¹³ and a shift toward individuals with public insurance, who may have the fewest outside resources, suggests that this rise is apt to continue.

The decrease in in-hospital mortality for patients with prolonged hospitalizations was notable over the ten-year period studied. This decline accords with the overall decrease in hospital mortality over the past ten years despite increases in hospitalizations, specifically in the Medicare population.^14,15 It does not appear to be primarily related to the changing age of the patients, as the association persisted with adjustment, and may reflect overall improvements in medical care for the very sickest hospitalized patients. Of note, the decline in in-hospital mortality was accompanied by a decrease in length of stay, and hence we cannot exclude the possibility that patients most likely to die are increasingly being discharged beforehand, although the absolute drop in mortality was large relative to the modest decrease in length of hospitalization.

These results are interesting from the hospital perspective as well. Long-stay patients utilize almost 1 in every 7 hospital days, which inevitably influences the patient mix and the case mix for hospitals, with attendant repercussions. The decrease in length of stay of approximately five hours yields a saved day for every five hospitalizations, a possibly important source of additional capacity for full hospitals. However, the growing costs for this patient group, despite this modestly shorter length-of-stay, will similarly have important implications for accountable care organizations at financial risk for their care. At the same time, the increase in hospitalizations for patients more likely to have Medicaid or no insurance may confer increased financial risk for urban hospitals already operating with slim profit margins. Compounding this financial risk are increased discharges to other facilities for rehabilitation, a trend consistent with overall discharge disposition trends.¹⁶ In an age of increasingly capitated or accountable care, this trend may represent a more costly option for hospitals. Indeed, the increasing disposition to other facilities implies that the complete financial impact of these long-stay patients is even larger than we document here.

From the vantage point of the academic center, the patient landscape is also evolving. Patients with prolonged stays are increasingly hospitalized in urban teaching hospitals and away from non-urban community hospitals. The demographic shifts to teaching hospitals may reflect a changing role for academic medical centers nationally, with an increased consolidation of complex and costly patients in academic hospitals.¹⁷ This new role for academic medical centers may affect the educational opportunities for residents and medical students, as the educational value of caring for patients with prolonged stays differs from multiple patients with shorter stays, even as these long-stay patients may offer opportunities for new educational foci, such as value-based care and rehabilitation potential.

At a juncture of increasing health care costs, myriad readmission reduction programs with mixed success have been introduced. However, attention has rarely focused on long-stay patients. At this point, hospitals are increasingly changing their discharge processes, and introducing more comprehensive multidisciplinary rounds and hiring new complex discharge planning experts to help navigate complex discharges with inadequate insurance coverage. This resource allocation to complex patients, so called “hot spotting”, may offer gains in reducing length of stay for the types of patients reviewed here, although focused efforts specifically for these particularly vulnerable patients are likely necessary.²

Limitations

Although we were able to provide estimates of long-stay hospitalizations for the entire United States using the NIS, it does not provide access to clinical data on individual patients that might be useful in profiling them, such as the proportion of their long-stay days when they may have been clinically stable. For that reason, future studies among individual hospitals remain important, even if less generalizable. In addition, we do not have information on long-term outcomes for patients with prolonged hospitalizations with which to judge mortality rates or other clinical outcomes. Regarding the mortality data, the data regarding disposition is limited, and did not allow us assessment use of palliative care and hospice care during hospitalization or at discharge. The NIS has also only recently begun to provide information on readmission rates and hence we cannot yet examine trends in this important determinant of costs and quality.

Conclusions

In the United States, patients with prolonged hospitalizations consume 14–15% of all hospital days despite representing only 2% of hospitalizations. Over the twelve years studied, such patients have been characterized by lower age and increased reliance on Medicaid or no insurance. These hospitalizations confer over $20 billion of financial burden to the health care system annually. Large academic hospitals, increasingly the providers of care for these patients, should consider investing in systems to expedite discharge for patients to reduce length of stay and the associated financial burdens while continuing to find ways to lower their in-hospital mortality even further. In addition, this study also opens further areas of investigation of trends in the long-stay population, including trends in readmissions, as well as other long term health outcomes for this large group.

Highlights.

1. From 2001–2012, prolonged hospitalizations of over 21 days represented 2% of hospitalizations in US hospitalis, but 14% of hospital days.

2. Over the ten years studied, inpatient mortality for patients with prolonged hospitalizations decreased from 14.5% to 11.6%.

3. Long stay patients were increasingly younger and male, and of minority status, and were increasingly hospitalized at urban teaching hospitals.

Appendix 1

Flow Chart of Inclusion and Exclusion Criteria