National Hospitalization Trends and the Role of Preventable Hospitalizations among Centenarians in the United States (2000–2009)

Increases in life expectancy mean that an unprecedented number of individuals are reaching centenarian status, often with complex health concerns. We analyzed nationally representative hospital admissions data (200–2009) from the National Inpatient Study (NIS) for 52,618 centenarians (aged 100–115 years, mean age 101.4). We predicted length of stay (LOS) via negative binomial models and total inflation adjusted costs via fixed effects regression analysis informed by descriptive data. We also identified hospitalizations due to ambulatory care-sensitive conditions defined by AHRQ Prevention Quality Indicators. Mean LOS decreased from 6.1 to 5.1 days, while over the same time period the mean total adjusted charges rose from USD 13,373 to USD 25,026 in 2009 dollars. Black, Hispanic, Asian, or other race centenarians had higher cost stays compared to White, but only Black and Hispanic centenarians had significantly greater mean length of stay. Comorbidities predicted greater length of stay and higher costs. Centenarians admitted on weekends had higher costs but shorter length of stay. In total, 29.4% of total costs were due to potentially preventable hospitalizations for total charges (2000–2009) of USD 341.8M in 2009 dollars. Centenarian hospitalizations cost significantly more than hospitalization for any other group of elderly in the U.S.


Introduction
Centenarians are one of the fastest growing demographic age groups in the United States. In 2010, there were 53,364 people aged 100 and older in the U.S. and this number jumped to 72,197 in 2014 [1,2]. Individuals are living longer with chronic diseases, meaning that an unprecedented number of individuals are reaching centenarian status with complex health concerns. Little research has addressed the hospitalization experiences of centenarians. It is particularly important to elucidate the role of preventable hospitalizations since these could indicate a lack of adequate primary care for this particularly vulnerable population and are costly to both the individual and the healthcare system.
Heart disease, Alzheimer's disease, stroke, cancer, and influenza and pneumonia were the leading causes of death for centenarians in 2014, with heart disease having an outsize effect accounting for 34.6% of centenarian deaths [2]. A study of participants in the Health and Retirement Survey participants, who reached the age of 100 in 2010, showed that only 23% reached that age with no chronic disease [3]. The vast majority of centenarians are living with chronic conditions as they become the oldest of the old [4,5], and these conditions are likely to multiply over time. A study looking at people aged 70-115 years being treated in the VA for cancer found that 70% had hypertension, over 50% had hyperlipidemia, 40% had heart disease, 25% had diabetes, and 21% had osteoarthritis, with the oldest old having the greatest prevalence of comorbidities [6]. An increase in the number of hospitalizations of centenarians in the U.S. has been observed between 2004 and 2008, and 96.9% of those admitted had mild or moderate comorbidities [7]. Evidence indicates that centenarians are likely to be living with multiple chronic conditions, potentially complicated by frailty and changes to mental status, that if not adequately addressed in an ambulatory care setting could lead to preventable hospitalizations.
Preventable hospitalizations are defined as potentially avoidable admissions to the hospital for an acute condition or worsening chronic condition that can be managed in an ambulatory care setting and can be used as a proxy for access and quality of ambulatory care systems [8]. These preventable hospitalizations are defined according to the Agency for Healthcare Research and Quality (AHRQ) and include diabetes, perforated appendix, chronic obstructive pulmonary disease (COPD) and asthma, hypertension, heart failure, dehydration, bacterial pneumonia, urinary tract infections, angina without procedure, and lower extremity amputations among patients with diabetes. Interestingly, a study using a subsample of Medicare claims data found that per capita spending for Medicare recipients in 2011 peaked at the age of 96 and then declined for beneficiaries 97 and above. This was also reflected in inpatient hospitalization costs to Medicare where individuals 95 and older cost less than the 75-94 age groups [9]. This may indicate the costs for preventable hospitalizations may actually be lower among centenarians compared to a 65-99 aged group. However, any costs associated with preventable hospitalizations among the centenarian group reflects an unused cost saving for the healthcare system, as this group is insured primarily through Medicare-federal old age insurance-and Medicaid-state and federally funded insurance for low-income populations.
There is a disparity in the racial and ethnic makeup of the centenarian population in the U.S., with 82.5% identifying as White in the 2010 census compared to 72.4% in the total U.S. population. Among centenarians in 2010, 5.8% identity as Hispanic or Latino compared to 16.3% in the total population. In total, 12.2% were Black or African American, 2.5% Asian, and 2.3% some other race or more than one race in 2010 [1]. Racial and ethnic minorities are less likely to survive to become centenarians. A 2001-2009 study of the National Inpatient Study (NIS) data showed that while potentially preventable hospitalization rates declined over that time period overall, Black or African American adults had over twice the rate of potentially preventable hospitalizations compared to the white non-Hispanic population and the rates of preventable hospitalization were consistently higher among the Hispanic population as well [8]. There is some question as to whether survivorship into the oldest age group may attenuate differences in health outcomes by race and ethnicity [10], but data are scanty. Specifically, among the 65 plus population in 2014 using Medicare claims data 15% of White, 18% of Black, and 17% of Hispanic patient's hospitalizations were potentially preventable [11]. This data indicates that disparities in health outcomes persist into the oldest old age groups.

Methods
For this analysis we used nationally representative hospital admissions data (2000-2009) from the NIS. There is some evidence that Medicaid expansion reduced preventable hospitalizations [12], so we chose a time period for analysis before the implementation of the Affordable Care Act in 2010. In addition, the latest census data available describing the centenarian population in the U.S. is the 2010 census. Furthermore, it is key to analyze historical data in order to understand trends over time. The NIS was developed for the Healthcare Cost and Utilization Project (HCUP) and is the largest publicly available allpayer inpatient healthcare database with discharge-level healthcare data. It was designed to produce regional and national estimates of inpatient utilization, access, charges, quality, and outcomes. We included data for all centenarians who had an in-patient hospitalization in that time period, regardless of admissions source.

Measures
Our primary outcome measures were as follows: (1) total charges in USD and (2) length of stay in days. Charges were adjusted for general medical price change using the Personal Consumption Expenditure Health Price Index [13] with 2009 as reference year and were natural log transformed prior to analysis. Charges are for the hospital cost of producing the services and do not include physician charges associated with a hospital stay. Length of stay was coded as days between admission and discharge.
Age was measured in years. Sex was coded as female = 1, male = 0. Race codes included identification as 1 = White (reference group), 2 = African American, 3 = Hispanic, 4 = Asian or Pacific Islander, and 5 = Other. Number of recorded diagnoses was used as an indicator of multimorbidities. Whether an admission occurred over the weekend (1 = yes, 0 = no) was coded as this may impact the length of stay. The NIS coded whether the patient died during their hospital stay and this was included in the models as a covariate (1 = dead, 0 = alive) in order to determine whether there was a cost differential for individuals who died in the hospital during that stay. Whether the patient was admitted from the emergency department (1 = yes, 0 = no) was coded to understand if it was an emergent hospitalization. We coded whether an individual was hospitalized due to potentially preventable conditions defined by AHRQ guidelines as a primary diagnosis (1 = yes, 0 = no), as defined in the introduction above. Finally, we coded the admission year (2000-2009).

Statistical Analysis
Sample descriptive statistics were estimated separately for men and women and are presented in Table 1. The most common primary diagnoses, average number of comorbidities, payer types, and demographics for the centenarian patients are presented in the descriptive table. Total costs were inflation-adjusted to 2009 USD-equivalents prior to analysis. They were also first natural log transformed (adding 1 so that the zero point would be equivalent in transformed and untransformed analyses). Models for ln(Total Charges + 1) evaluated the predictor of interest (potentially preventable hospitalizations) and adjusted for the following potential confounders: age, sex, dummy coded race, length of stay, number of diagnoses as an indicator of multimorbidities, weekend admission, admission from emergency department, whether the patient died during the hospital stay, and dummy coded year (to capture non-linear secular trends).
Models for length of stay evaluated the predictor of interest (potentially preventable hospitalizations or not) and adjusted for the following potential confounders: age, sex, dummy coded race, length of stay, number of diagnoses as an indicator of multimorbidities, weekend admission, admission from emergency department, whether the patient died during the hospital stay, and dummy coded year (to capture non-linear secular trends).
Based on the results of Hausman specification tests, both outcomes were estimated via fixed effects models to address unobserved heterogeneity. Natural log transformed total costs were modeled via fixed effects regression and length of stay was estimated using a fixed effects negative binomial model to allow for overdispersion. Both models adjusted for clustering at the level of the state. Table 1 presents descriptive statistics for our sample (n = 52,618; aged 100-115 years; mean age 101.4 years) separately by men and women on all study variables. As reflected in the general population of centenarians, the number of women in the sample greatly outweigh the number of men. Among the centenarians that were hospitalized between 2000 and 2009, 75.8% identified as White, 13.6% as Black, 5.8% as Hispanic, 2.3% as Asian and Pacific Islander, and 2.4% identified as other race. The primary payer for hospital stays was Medicare with only 4% of the men and 0.7% of the women self-paying. Government sponsored insurance (Medicare and Medicaid) was the primary payer in 90.7% of the centenarian hospitalizations. On average, the centenarians had eight diagnoses upon admission to the hospital, demonstrating the high level of multimorbidity typical for this age group [14]. Table 2 presents the 10 most common overall diagnoses by sex. Nine of the top 10 diagnoses were common to men and women. Pneumonia was the most frequent diagnosis, followed by congestive heart failure and urinary tract infections. Gastrointestinal hemorrhage was a leading diagnosis among female but not male centenarians, while intercranial hemorrhage was in the top 10 for males only.  Table 3 shows the total and mean adjusted charges per centenarian hospitalizations in 2009 USD by year and by potentially preventable hospitalizations. While the mean centenarian hospitalization costs was USD 22,308 over the years 2000-2009, the range was a low of USD 18,833 in 2000 to a high of USD 25,027 in 2009. The total national charges for centenarian hospitalizations in this time period was USD 116 billion, of which USD 34.2 billion was due to a potentially preventable hospitalization as a primary diagnosis at admittance to the hospital. Mean adjusted total charge per hospitalization over this time period was highest for centenarians identifying as other race and ethnicity while the lowest mean charges were for Asian or Pacific Islander centenarians. On average, the total adjusted charges per hospitalization were lower for female centenarians than for males. Results from the fixed-effects regression model for ln (total charges + 1) are shown in Table 4. Total charges were higher on average for older centenarians (p ≤ 0.05) and lower for women than men (p ≤ 0.0001). Compared to Whites, total charges were significantly higher for Black, Hispanics, Asian and Pacific Islanders, and other race centenarians. Length of stay (p ≤ 0.0001) and number of diagnoses (p ≤ 0.0001), weekend (p ≤ 0.05) or emergency department admission (p ≤ 0.0001) was associated with significantly higher total charges. In contrast, total costs were lower for individuals who died during their hospital stay and those centenarians admitted with an ambulatory care sensitive diagnosis. Costs (in adjusted dollars) were significantly higher than in 2000 for every year except 2001. Table 5 shows the adjusted total charges per hospital stay for each of the top 10 primary diagnoses for each year. Even after adjustment to 2009 dollars, there is still a distinct upward trend in total charges for each diagnosis, consistent with the results of the multivariable regression model. Diagnoses with an asterisk in Table 4 indicate potentially preventable hospitalizations.

Length of Stay
Results from the fixed-effects negative binomial regression model for length of stay are shown in Table 6. Length of stay was slightly but significantly shorter among older centenarians. Compared with White centenarians, Black (p ≤ 0.0001) and Hispanic (p ≤ 0.05) centenarians had longer average stays. Individuals with multimorbidities had significantly longer stays, whereas individuals admitted on weekends or who died during hospitalization had shorter stays. Lengths of stay were significantly shorter than in 2000 for every year except 2001 (p ≤ 0.0001).  Table 7 shows mean length of stay for each of the top 10 primary diagnoses for each year. There is still a distinct trend toward shorter stays for each diagnosis, consistent with the results of the multivariable negative binomial model. The rows with asterisks indicate ambulatory care sensitive diagnoses and associated length of hospital stay for centenarians.

Discussion
We set out to characterize hospitalization trends among centenarians in the decade prior to the implementation of the Affordable Care Act. Increases of life expectancy of 2.2 years per decade of life indicate that the proportion of individuals reaching centenarian status is likely to increase dramatically in the coming decades [15]. At the same time, there are mounting concerns about the burden of disease and associated costs that a large proportion of these individuals are likely to encounter in their second century of life [16].
Our analyses suggest that health disparities by race persist even among exceptional survivors. Compared with Whites, mean adjusted total charges were USD 4057 higher for Blacks, USD 12,962 more expensive for Hispanics, USD 15,189 more expensive for Asian or Pacific Islanders, and USD 8498 more expensive for individuals reporting other race. While Black, Hispanic, Asian or Pacific Islanders, and other race centenarians had higher cost stays compared to White, only Black and Hispanic centenarians had significantly longer mean hospital stays. This indicates that other race centenarians had other drivers besides length of stay for the higher cost per hospitalizations.
Comorbidities predicted both greater length of stay and higher costs. Centenarians admitted on weekends had higher costs but shorter lengths of stay per hospitalization. The length of the hospital stay for centenarians decreased on average from 2002 through 2009, but adjusted total charges increased over the same time period. Therefore, while centenarians were staying for a reduced length of time in the hospital per admission, the cost if that admission in adjusted dollars is actually increasing. This has real implications for Medicare, the primary payer for centenarian hospitalizations.
We also found evidence of significant sex differences in hospitalization charges, such that the average adjusted total charge of stay was USD 5070 higher for men than women, despite comparable lengths of stay. This may be explained in part by women being functionally younger at every age in the centenarian population [10]. This finding is consistent with general population hospitalization costs, where on average the hospital stay for women costs less than for men [17].
Centenarian hospitalizations cost significantly more than hospitalizations for other groups of older adults in the U.S. Centenarians (USD 25,027) had 2.7 times the average hospital costs of the 85 plus population (USD 9400) and 2.1 times the costs of the 65-84 (USD 11,900) population in 2009 [17]. One study in the U.K. has suggested that total healthcare costs among individuals aged 80 and above may be highest in the last year of life [18]. In the current analyses, approximately 10.8% of centenarians died in the hospital, so this is unlikely to drive the current findings alone. However, our results suggest that there are considerable sex differences in costs by discharge status (alive/dead). Specifically, total charges (2009 USD) for male centenarians were USD 25,000 when alive at discharge and USD 36,172 when deceased at discharge. Total charges (2009 USD) for female centenarians were USD 20,916 when alive at discharge and USD 24,746 when deceased at discharge. Differences in age-graded hospitalization charges among the oldest old is a topic worthy of further consideration in light of sex differences that are even greater among the centenarian population than individuals 85+.
Hospitalizations due to ambulatory care-sensitive conditions as primary diagnoses are a significant contributor to admissions and cost in the centenarian population. Approximately 29% of total hospitalization charges and 34% of all admissions were associated with potentially preventable hospitalizations for a total charge (2000-2009) of USD 34.2 billion in 2009 dollars.

Limitations
Application of this study globally may not be possible as there have been shown to be differences between the health status of U.S. centenarians and centenarians in other countries [19,20]. While this data is potentially informative, cross-national comparisons are difficult due to differences in health status, health systems, and policies.
The period of data analysis was selected to align with both with a period immediately prior to implementation of important relevant national policy as well as the most current data available on centenarians from the Census. Expanding the analyses to consider the period immediately following implementation of the ACA, which includes many pilot projects on medical care delivery such as Accountable Care Organizations, will thus be an important next step.
As centenarians remain so rare even among the older adult population, the granularity of data that we would wish for to understand the needs of this exceptional population, particularly at the community level, is unavailable. For this reason, we currently have few tools available to answer questions such as what proportion of centenarians is hospitalized in any given year, the proportion of centenarians with multiple hospitalizations, and what characteristics distinguish centenarians with potentially preventable diagnoses who are hospitalized from those who do not. To address these questions, data on both hospitalized and non-hospitalized centenarians would be needed.
This data, while showing a significant impact of preventable hospitalizations on cost among the centenarian population, does not allow us to ascertain the specific drivers behind these potentially preventable hospitalizations such as lack of access to care and/or the quality of the care provided.

Future Directions
Future analyses can use these historical data trends in order to understand the impact of changing policies and demographics on preventable hospitalizations and associated costs.

Conclusions
More than one third of centenarian hospitalizations are for potentially preventable ambulatory care diagnoses. Since government sponsored insurance (Medicare and Medicaid) is the primary payer in this population in over 90% of hospitalizations and the of people who survive into their 10th decade in the U.S. are increasing, this is an area to leverage potential savings through greater access to quality primary care for centenarians. It is also important to recognize that racial disparities in health outcomes persist even among exceptional survivors.