Deciding When to Put Grandma in the Nursing Home

Abstract

For caregivers of persons with dementia, estimating when that person should be placed in long-term care is difficult. Health care providers also find it hard to give an exact time as to when the person should be placed. Using data from 197 caregivers working with the Dementia Care Services Project in North Dakota, we show that asking the caregiver about their inclination to place can be equated to asking them for a specific time to place (κ = .616). Using the probability density function of time to place we were able to translate it into inclination. This inclination is easier information for the caregiver to provide and places fewer burdens on the caregiver and patient. It also provides the health care provider with a measure of time to help advise caregivers and recommend interventions and provide service organizations with measures of cost savings to support the impact of outreach and intervention.

Introduction

Dementia is a degenerative neurological disorder with no known cure and the potential for enormous health and long-term care (LTC) costs. Being a caregiver for a person with dementia can be difficult and time intensive and is associated with poorer emotional, mental, and physical health. ^1,2 Primary care providers are often the health care home for these patients and their caregivers. Caregiving in rural areas is even more difficult due to social isolation and longer travel times, which may result in poorer access to services. ³ A difficult decision for many caregivers is whether and when to place their person with dementia in an LTC facility. ⁴ This decision is complicated by social circumstances, preferences, and values and is greatly affected by caregiver stress, with higher levels being linked to greater likelihood of placing persons with dementia in LTC. ^{5 –7}

Several studies have found that caregiver support programs can substantially alleviate caregiver burden, stress, and depressive symptoms ^{8 –10} and thereby reduce LTC placement and associated costs. ^8,11 Caregivers often seek advice from their primary care providers about the timing or need for LTC. Giving advice on this difficult decision could be done with better accuracy and confidence with information and tools. While most persons with dementia become institutionalized at some point, many are living in the community and delaying LTC placement through caregiver and patient support could result in significant health care savings that could be directed to other needs.

An increasingly important issue in conducting studies involving dementia caregivers and LTC placement is the means for measuring direct indicators (ie, the actual time before placement) and indirect indicators of LTC placement, such as caregivers’ inclination and/or estimated timing for LTC placement of persons with dementia. There are several ways to measure caregivers’ levels of inclination and timing for placing persons with dementia in LTC. One retrospective method is to contact caregivers of persons with dementia who were in fact placed in LTC and ask them how much time had transpired between dementia symptom onset or diagnosis and LTC placement. ¹² A prospective method to use as part of surveys or interviews is to measure caregivers’ level of likelihood for LTC placement (eg, yes/no item; Likert-type scale ranging from 1 to 5) at one or multiple points in time. ¹³ A second prospective method is to ask caregiver respondents about their estimated amount of time, defined in number of months/years, until they will likely have to place their person with dementia in LTC.

The purposes of this article are 2-fold: to compare dementia caregiver response completion rates between the likelihood to place and time to place questions and to suggest a method of translating likelihood to place answers into realistic estimations of time to place. Specifically, we use a Likert-type scale scoring of likelihood of LTC placement and a probability density function to estimate time to placement among dementia caregivers enrolled in a statewide caregiver support program in North Dakota.

Comparing Inclination and Time

The North Dakota Dementia Care Services Project (DCSP) provides information, resources, and referrals to caregivers of persons with dementia through phone and in-person visits. As part of the visits, caregivers in the DCSP were asked 2 questions about LTC placement, one referring to the likelihood of placing the person with dementia and one referring to the approximate timing of the placement.

• Inclination: On a scale of 1 to 5 (with 5 being most likely and 1 is not likely) how likely are you to place the person you are calling about?

• Time: How soon would you place the person you are calling about? (0-6 months, 7-12 months, 13-18 months, and more than 18 months).

Krista Headland, Alzheimer’s Association Western North Dakota Regional Center director, stated that the caregivers they work with are more willing to give their intention to place (measured in likelihood) rather than specify a time. It places no tangible expectations on themselves. They also realize that placement is in the future and they cannot know for sure when it will happen as it can depend on many other factors that are beyond their control (Krista.Headlund@ALZ.org, e-mail, December 13, 2011).

The number of responses to these 2 questions increased from 752 (by 382 caregivers) in December 2010, to 1769 (by 840 caregivers) in October 2011. During that time, the response rate for the question regarding time to place ranged from 60% to 72%. People were consistently more likely to answer the question on likelihood to place, with 67% to 74% providing responses.

Also important is the number of people who answered the same question (likelihood or time) at least 2 weeks after their initial answer (note that the DCSP staff reported that it takes at least 2 weeks before the caregiver can experience significant changes, hence make changes in their inclination to place). Having the follow-up information for a question is necessary to determine changes in the caregivers’ inclination for LTC placement over time. The percentage of caregivers providing 2 or more answers regarding time to place rose from 49% to 66% and likelihood rose from 59% to 69%.

In the 2010 to 2011 data, 197 caregivers twice provided both likelihood and time answers (there were 9 who ranked likelihood only and 1 who estimated time only) As likelihood was measured on a 5-point scale while time was estimated using a 4-point scale, a direct comparison of answers is not possible. However, the Spearman correlation between likelihood and time was .853 (P < .001) for the caregivers’ first answers and .945 (P < .001) for their final answers. These results suggest that caregivers answered both questions in a similar fashion.

The 197 likelihood and time answers can also be associated by how the caregivers’ answers changed between the first time they answered the question and the last time they answered, at least 2 weeks later. This change between their first and last answers measures how the caregiver changed their inclination and time toward placement. The answers can change in 3 directions. The caregiver can decide to wait longer to place the person with dementia, keep their likelihood or time toward placing the same, or decide the person with dementia should be placed sooner. These 3 types of changes can be used for either the likelihood or time questions.

Table 1 shows the number of caregivers who indicated a change in placement inclinations that was longer, the same, or sooner than they originally thought for both likelihood and time questions. Among the 197 caregivers, there was a high level of agreement (77%) between response changes for the likelihood to place and time to place questions. The Spearman correlation coefficient was .726 (P < .001) and κ = .616 (95% confidence interval: 0.520-0.712). Thus, these measures indicated a strong agreement between the caregivers’ changes in likelihood to place and time to place.

Table 1.

Agreement Between Caregivers’ Changes in Inclinations and Time to Place

Likelihood	Time			Stats
	Longer	Same	Sooner
Longer	29	4	0	Spearman corr = .726; P < .001; κ = .616 (.520 - .712)
Same	13	85	1
Sooner	3	25	37

Any differences that exist between the likelihood and time changes should be noted. The greatest difference is in the lower half of Table 1, where several caregivers’ likelihood answers indicated same (n = 13) or sooner (n = 25), compared with longer or same in their time answer. This difference may reflect inaccurate reporting by the caregiver. If they reported the same time answer (eg, within 7-12 months) twice or more for a year, they may have been reluctant to give their reply serious thought. A caregiver may have changed an answer because they think it is expected. It also may be that caregivers reflected on their answers between time and likelihood and choose the same end number assuming they were being consistent. There were 25 caregivers who did not change time to place but did change likelihood to place.

Differences in responses to the likelihood and time questions may also be due to time being measured in strict 6-month intervals while no specific time intervals were used for likelihood. The longest time estimation a caregiver could choose was over 18 months. This answer can represent up to 20 or more years and, thus, the time indicator loses sensitivity to measure long delays in placement. Likelihood does not have exact times associated with it and can reflect longer times.

Translating Inclination Into Time

If inclination is easier than time for caregivers to give, being able to translate it into time would be helpful for caregivers and health care providers. The first step we used in the translation was to match the reported value for likelihood to place with a percentage representing the probability that any caregiver in the universal population (theoretically, all caregivers) would place a person with dementia in LTC.

We assumed there was 20% between each likelihood value (uniformity in the universal population). For example, a likelihood score of 1 assumes that 10% of all persons with dementia would be placed in LTC, while a likelihood score of 4 assumes that 70% of all persons with dementia would be placed. The end points of the likelihood scale 0 (never place) and 6 (placed now) are implicitly associated with 0% and 100%, respectively (Figure 1).

Figure 1.

Matching likelihood to place with probability of placing.

The next step was matching specific times (representing years to place) to the probabilities. This can be done using previous studies that have estimated probability of placement over long periods of time. ^8,14 The midpoint can be determined by estimating when half the population of caregivers will place a person with dementia. Studies we found suggest a median time of 3 to 7 years to placement following dementia diagnosis. ^{8, 14 ,15} Making the assumption that caregivers come to DCSP immediately after diagnosis (in reality, some come before they have the diagnosis and some come to DCSP sometime after the diagnosis), the midpoint (likelihood score of 3; 50%) was matched to 5 years.

The far right end point (6 or placed) would be 0 years. The far left end point was estimated to be 20 years. This was considered a practical end point, meaning nearly all persons with dementia would be placed in LTC within 20 years of diagnosis. Additional estimated matching points include 90% of the persons with dementia (likelihood score of 5) are placed following 1 year or more after their diagnosis and 10% (likelihood score of 1) after 15 or more years. While these matching points can be adjusted, all of these estimates were consistent with other cumulative density functions ⁸ and deemed plausible or realistic by 4 members of the Minnesota/North Dakota Alzheimer’s Association for this population.

The difference with the years to placement compared with both probability and likelihood is that years are not distributed uniformly (ie, they are skewed). For that reason, you cannot directly translate the probability of placing to the years to placement on a one-to-one basis. However, if the probability density function (a line where the area under it represents probability of a value on the x-axis) can be estimated using these initial points, you can estimate the number of years that correspond with each percentage and hence each likelihood value.

Figure 2 is an estimate of the probability density function for this distribution (based on the gamma distribution). Here, the horizontal axis represents time in years, and the vertical axis is the probability density function. This shows a skewed distribution with a median of 5, a left end start of 0, and a practical cutoff on the right of 20 (meaning the probability of placing someone after more than 20 years is very small).

Figure 2.

Probability density function of years to placement in long-term care (LTC).

The probability of placing is not directly read from the axes. Instead, to find the probability of placing sooner than 5 years, you would use calculus to integrate that curve from 0 to 5. In other terms, if every person with dementia is a dot under the line, half of the dots would be before 5 years and half would be after. That means that with 5 years as the median value, there is a 50% probability of placing before 5 years and a 50% probability after. Similarly, you could calculate that 41.1% of the dots are before 4 years (integrate the curve from 0 to 4) or 41.1% of the persons with dementia are placed before 4 years.

You can also find the year that corresponds with a percentage. For example, if the caregiver is not likely to place the person with dementia (value of 2), with a corresponding probability of placing being 30%, then you go to the right side of the probability density function (Figure 2) to where 30% of all persons with dementia are to the right or higher in years. The year that corresponds to marking where 30% of the persons with dementia are to the right of it is 8.089. The nomogram that translates likelihood into approximate time or years to place can now be completed (Figure 3).

Figure 3.

Nomogram of likelihood, probability, and years to placement in long-term care (LTC).

From this nomogram, the years potentially saved for not placing someone when they change from a high (more likely to place) to a low (less likely) value can be calculated as depicted in Table 2 (note that the likelihood score of 6 or “now placed” is included in Table 2 as caregivers may take the person with dementia from the LTC facility after receiving help from the DCSP).

Table 2.

Translation of Likelihood to Place Into Years Until Placement

Likelihood	Years	Likelihood	Years	Likelihood	Years	Likelihood	Years	Likelihood	Years
2 → 1	7
3 → 1	10	3 → 2	3
4 → 1	12	4 → 2	5	4 → 3	2
5 → 1	14	5 → 2	7	5 → 3	4	5 → 4	2
6 → 1	15	6 → 2	8	6 → 3	5	6 → 4	3	6 → 5	1

The years of placement potentially avoided can be changed into days and then matched with the costs for those days. For example, a caregiver who changed their likelihood from a 4 to a 2 potentially reduced the time to place the person with dementia in LTC by 5 years. As an estimate, the median cost for 1 day of LTC in North Dakota is $196 (range $151-$389). ¹⁶ This median daily cost can then be used to estimate the potential savings by lessening the inclination of placement by 5 years is $357 700 ([$196 per day] × [5 years] × [365 days]).

Summary

Advantages have been noted in asking caregivers questions about likelihood rather than time to place in LTC. Caregivers had answered the question of likelihood more readily than the time question throughout the data gathering time period (note that according to the DCSP staff, the increase in answers regarding time was largely due to care consultation workers pressing caregivers for answers). Posing both questions to respondents in the same interview may also have been confusing for some caregivers.

Furthermore, questions regarding inclination do not pin caregivers down to a concrete time to placement with limited range, making inclination a less specific, more comfortable question to answer; it is about how the caregiver is feeling about placement and not an estimated timetable. It represents a more sensitive distribution.

The decision to place a relative in LTC is complex and difficult emotionally for patients, caregivers, and families. It is a decision that impacts relationships, finances, and the health and well-being of the patient and others. Continuing research efforts are helping to inform our understanding of factors that predict the need for placement in LTC and the interventions that can be done to improve care and outcomes and potentially delay the decision to place in LTC. ¹⁷ The use of estimation of a caregiver’s inclination to place a person with dementia in LTC is more practical and easier to measure than time to placement. Inclination can be translated into time without the pressure of forcing the caregiver to give a time. By using this nomogram, changes in inclination can be used to estimate the timing and potential savings that result from delaying LTC placement.