Models to predict need for nursing home level of care in community-dwelling older adults with dementia

Population: Community dwelling older adults aged 65 years and older with dementia

Outcomes: 2, 5, and 10 year risk and median time to needing nursing home level of care. The outcome of nursing home level of care was defined as one of the following 3 items:

  1. ≥3 ADL dependencies (including bathing/showering, getting in/out of bed, dressing, toileting, and walking across the room),
  2. ≥2 ADL dependencies and proxy report that the individual wanders or cannot be left alone,
  3. eating dependency (e.g., needing help cutting up food).

In general, dependency with an ADL was defined by needing help with the task.

Scroll to the bottom for more detailed information.

Risk Calculator
1. Was the information obtained from the patient or via a surrogate?

2. What is your patient’s age?





3. What is your patient’s biological sex?
4. Which of the following ADL and IADL dependencies does your patient have?

Dependency with an ADL/IADL means that the patient requires help performing the specific task and cannot perform it independently. Note: If your patient has 3 ADL dependencies or eating dependency at baseline, they would already be classified as nursing home level of care.








5. What is your patient’s driving status?


6. Does your patient have issues with incontinence? In other words, has your patient lost any amount of urine beyond their control?
7. Was your patient able to recall the following items:
• What is the day of the week?

• What is the month of the year?

• What is the year?


Information about the models

We created 2 models based on whether interview responses were from the individual (self-respondent model) or a proxy (typically a spouse/family member) (proxy model). Proxy interviews are generally conducted for participants who could not participate in the interview (e.g., physical/cognitive reasons).

Self-respondent model

  • The self-respondent model was developed in 2,312 community-dwelling individuals with dementia not classified as nursing home level of care at baseline who were interviewed in the Health and Retirement study from 1998-2016 (mean age 83 years, 71% female, 13% Black).
  • Participants were classified as having dementia using a validated algorithm which has shown high accuracy in validation studies (e.g., AUC of 0.93 in training data and 0.84 in validation data).
  • Information on type of dementia was not available in the Health and Retirement Study.
  • This index was internally validated using bootstrapping. External validation was performed in a cohort of community-dwelling older adults with dementia not considered nursing home level of care at baseline who were interviewed in the National Health and Aging Trends Study from 2011-2019 (n = 1,327).

Discrimination: This risk calculator sorts patients who are classified as nursing home level of care earlier from patients who were still not classified as nursing home level of care correctly 64% of the time (as measured by the integrated area under the receiver operating characteristic curve which is a weighted average of the AUC values at all event times).

64%

Calibration: The model was well calibrated across all risk ranges at the 2, 5, and 10-year time points.

Proxy model

  • The proxy model was developed in 1,015 community-dwelling individuals with dementia not classified as nursing home level of care at baseline who were interviewed in the Health and Retirement study from 1998-2016 (mean age 80 years, 68% female, 10% Black).
  • Participants were classified as having dementia using a validated algorithm which has shown high accuracy in validation studies (e.g., AUC of 0.93 in training data and 0.84 in validation data).
  • Information on type of dementia was not available in the Health and Retirement Study.
  • This index was internally validated using bootstrapping. External validation was performed in a cohort of community-dwelling older adults with dementia not classified as nursing home level of care at baseline who were interviewed in the National Health and Aging Trends Study from 2011-2019 (n = 385).

Discrimination: This risk calculator sorts patients who are classified as nursing home level of care earlier from patients who were still not classified as nursing home level of care correctly 72% of the time (as measured by the integrated area under the receiver operating characteristic curve which is a weighted average of the AUC values at all event times).

72%

Calibration: The model was well calibrated across all risk ranges at the 2, 5, and 10-year time points.

Citation: Manuscript accepted.

DISCLAIMER

The information provided on ePrognosis is designed to complement, not replace, the relationship between a patient and his/her own medical providers. ePrognosis was created with the support of the Division of Geriatrics at the University of California San Francisco. However, its content is strictly the work of its authors and has no affiliation with any organization or institution. This web site does not accept advertisements. If you reproduce the material on the website please cite appropriately. For feedback and questions regarding the site please email Sei Lee, MD (sei.lee@ucsf.edu), Alex Smith, MD (aksmith@ucsf.edu) or Eric Widera, MD (eric.widera@ucsf.edu).