Care trajectories of chronically ill older adult patients discharged from hospital: a quantitative cross-sectional study using health insurance claims data

The design of this study is an observational cross-sectional study using health insurance claims data in the period of January 2015–July 2016. Quantitative measures were used in order to gain insight into the healthcare utilization of older adults after their hospital admission and describe the numbers of readmissions and deaths.

We included subjects insured by the largest Dutch insurance company (30% market share) with at least one claimed hospital admission in 2015. We selected all patients who were living in one of the company’s care-office regions in 2015. Hence, we were able to analyze the full chain of healthcare, including hospital care, general practitioners care, community nursing, medications, and long-term care. We attempted to define frail older adult patients from our dataset as they are most at risk for adverse events.

Research done by Makai et al. has shown that age, sex, medication use, and being incontinent are predicting variables for frailty when analyzing claims data [8]. We therefore included patients if they were aged 75 and up, used medications from at least two pharmacy-based cost groups [9], which are defined by The National Health Care Institute [10] and used incontinence materials in 2015.

Of the 375,044 insurees that were 75 years or older in 2015, 247,529 (66%) were clinically admitted to a hospital. With the additional criteria, we identified 21,258 (8.7%) as likely to be frail. This is much lower than the percentages of frailty found in the literature [11] and underpins the belief that the best way to identify frail older adults is by performing a comprehensive geriatric assessment [12, 13]. Thus, in order to not confuse existing literature we abstain from using the term ‘frail’ for our subjects and will refer to them as chronically ill older adult patients.

We identified 21,258 hospitalized chronically ill older adult patients. They were clinically admitted at least once in that year to one of 96 different organizations: 87 hospitals and 9 private clinics. In order to obtain a more homogenous group, we also conducted all analyses in a subsample of our selection. We chose older adults clinically admitted with a hip fracture, as this is a typical acute event requiring hospitalization and rehabilitation.

The insurance company’s database contains all claims data and underlying care provided at the level of healthcare activities and performance codes of their insurees. The data also includes information about the healthcare providers, as well as the 4-digit postal codes, age, sex, and marital status of insurees.

Data on socioeconomic status (SES) of 2014 were retrieved from the Social and Cultural Planning Office (SCP) and linked by four digital postal codes to the claims data. SCP calculated social economic status scores based on information regarding education, income, and position in the labor market [14]. The scores resemble the social status of a postal code region, relative to the other regions in the Netherlands.

For each insuree, all claims under the healthcare law and long-term care act were recovered. Claims under the social support act were not included (these are reimbursed through municipalities). We identified the last day of a patient’s first admission in 2015. These included clinical admissions, intensive care unit (ICU), and emergency room (ER) contacts in a hospital or independent treatment centre. The claims database does not include ICD-10 codes. In order to determine what the underlying reason for admission was, the healthcare insurance company provided us with a reference table where each diagnosis-related group (DRG) was linked to a certain chapter of the ICD-10 codes. We selected all healthcare activities that were registered within a 180 day period after their discharge. As there are thousands of different healthcare activities, we clustered the data according to predefined meaningful categories [15] .

In order to describe the number of deaths and readmissions for different combinations of aftercare, we labeled chronically ill older adult patients according to the type(s) of care they received in the 90 days after admission: geriatric rehabilitation (GR), community nursing (CN), a short stay in a nursing home (ST), and long-term care (LTC). Geriatric rehabilitation (GR) is short-term and intensive care given by a multidisciplinary team either in a hospital ward or a nursing facility [16]. A person is eligible for GR when they can return home after this treatment. CN is nursing care at home for patients with an illness or disability. The aim is to ensure that the older adults can stay at home for as long as possible. A short stay (ST) is a temporary stay in a nursing home where people after hospitalization can stay for up to 12 weeks. Patients are eligible for an ST if they need more care than can be provided in the home situation [17]. LTC is intensive care provided at home or in a facility and is only for patients that will need permanent care 24/7. Patients with LTC pay an income-bound contribution to the costs [18].

We chose these four categories, as these are the types of usual aftercare that we believe might ward off readmissions. It was possible for each chronically ill older adult patient to be labeled with more than one category, resulting in multiple combinations of aftercare. For the six most frequently occurring combinations of aftercare, the group without any of these four types and the remaining patients (others) we identified subject characteristics and the readmission and mortality rates in the 90 days and 90–180 days after admission.

For both samples, means and standard deviations for age; median and ranges for SES, total 2015 spending, and number of morbidities; frequencies and percentages for sex, marital status, types of morbidities, and the three most frequently registered DRGs opened at admission, were calculated. We analyzed the six-month post-discharge healthcare utilization for all FE and FE admitted due to a hip fracture, for hospital care, GP care, long-term care, and community nursing care. We did this by calculating the number and percentage of chronically ill older adult patients with a certain type of treatment and by calculating the average number of treatments per user. To assess group differences, a T-test and a non-parametric equivalent were performed on all continuous variables. A Chi-square test was performed on all categorical variables.

All analyses were performed in SAS Enterprise Guide.

In Table 1, descriptive statistics for all included chronically ill older adult patients are presented. The average age at the time of admission was 83.8 years and more females than males were clinically admitted (62.1%). The most common morbidities were hypertension (88.1%), heart conditions (62.4%), and high cholesterol (54.8%). Of all 21,258 chronically ill older adult patients, 505 (2.4%) died during their first 2015 admission, resulting in a total of 20,753 discharged patients. A total of 12.5% died during admission or during the 90 days after discharge. The five most common reasons for a hospital admission was injury or poisoning (7.0%), ischemia, CABG, PCI (6.2%), cerebrovascular diseases (5.3), diseases of the digestive system (5.2%), and cardiac arrhythmia (5.5%).

Table 2 described the post-discharge healthcare utilization. During the six-month period after admission, 51.5% of all chronically ill older adult patients had a (non-ICU) readmission with a median number of hospital days of 12. Within the same time-frame, 12.3% of all patients were admitted to the ER with an average visit of 1.3 times (median 1). Almost half of all chronically ill older adult patients received care of a community nurse, whereas less than 5% received long-term home care. Up to 18% received care in a nursing home during the 6 months after discharge. Almost all patients visited their GP at least once, with a median of 7 visits, and in two thirds of the cases, the GP visited the patients at home (median of 3).

Significant differences between the total group of chronically ill older adult patients and chronically ill older adult patients with a hip fracture are: patients with a hip fracture were on average 3 years older, they had a higher healthcare spending in 2015, and more patients were readmitted with almost double the amount of hospital days and more rehabilitation days. Patients suffering from a hip fracture had significantly less outpatient visits, non-invasive diagnostics, and GP consultations compared to the total group of chronically ill older adult patients.

The 6 most frequent combinations of aftercare for GR, CN, ST, and LTC during the first 90 days are: CN, LTC, GR + CN, CN + LTC, GR, and ST. These groups are shown in Tables 3 and 4 together with the group of chronically ill older adult patients not receiving any of these four types and the remaining category ‘other’. Table 3 depicts the results for the total group of chronically ill older adult patients and Table 4 shows the results for the subgroup with hip fracture. The results from both tables will be discussed below.

Because we used data from one of the largest Insurance companies in the Netherlands with a 30% market share, we were able to analyze a large and representative group of the older adults in the Netherlands. Also, the fact that the company not only collected claims data on a DRG level but on healthcare activity level means that we can actually analyze the chain of healthcare in much greater detail. We were unable to include claims made under the social support act. This is healthcare reimbursed by municipalities which focuses on providing assistance in daily living (for instance grocery or meal services or cleaning support) in order to let the older adult live independently for as long as possible.

In 2015, an average of 22% of all persons aged 75 and over resided in a long-term care facility which is similar to but slightly higher than the 18% of the older adults we identified residing in nursing homes. Our inclusion criterion ‘using medications from at least two different pharmacy-based cost groups’ most likely lead to an under-sampling of older adults residing in long-term care facilities. This is because in those facilities, the medication costs are directly claimed with the facility and not with the insurance company. Also, by using pharmacy-based cost groups, not all drugs administered in the inpatient setting could be used to identify morbidities of the FE. These two reasons lead to a lower morbidity rate for cancer and dementia. This does not mean that a large part of our subjects does not suffer from these diseases; it means that we were unable to identify them.