Using Statistical Models to Predict Worsening Health among Older People with Disabilities

• Download Summary
• Español (pdf)
• Audio Recording (mp3)

What was the research about?

Many older adults in the United States are living with long-term, or chronic, health problems and disabilities. These health problems and other personal traits, like people’s age or gender, can put people at risk for worse health and disabilities as they age. Some people must move into nursing homes to get the care they need. Being able to predict worsening health could help patients and doctors make healthcare decisions.

In this study, the research team developed statistical models that used Medicare data to identify traits that made an older person with disabilities more likely to have the following worsening health events:

• Becoming more disabled within two years
• Moving into a nursing home within two years
• Dying within two years

What were the results?

According to the statistical models, people with the following traits were more likely to become more disabled, move into a nursing home, or die within two years:

• Being 85 or older
• Being male
• Not living with a spouse
• Having only Medicare insurance
• Having Alzheimer’s disease or dementia
• Having diabetes
• Being unable to do daily activities

The research team found that risk scores calculated using models that included these traits could predict if a person would have one of the worsening health events within two years.

Who was in the study?

The research team looked at Medicare data for 21,264 people ages 65 and older. Among these people, 56 percent were ages 65 to 74, 34 percent were 75 to 84, and 10 percent were 85 and older. Also, 56 percent were women, 82 percent were white, 8 percent were black, and 7 percent were Hispanic.

What did the research team do?

First, the research team conducted group interviews with doctors and people with and without disabilities. The interviews were about experiences with the healthcare system, problems getting care, and how having disabilities affects care. The interviews helped the team learn what traits to include in the statistical models.

Next the research team looked at the traits of people who became more disabled, moved into a nursing home, or died within two years. The team used these details to create the models. The models allowed the team to create a risk score that can be used to predict a person’s risk of having one of the worsening health events within two years.

What were the limits of the study?

The results only apply to people who are 65 and older and have Medicare. The models only showed that having certain traits increases the chance of becoming more disabled, moving into a nursing home, or dying. The models didn't show that the traits caused these events.

Future research could explore how well the statistical models work to identify older patients most in need of care to prevent worsening health.

How can people use the results?

Doctors could use statistical models to see which patients are at risk for worsening health. Older people with disabilities and their doctors could consider using risk scores to help manage care.

Objective

To develop statistical models to predict two-year risk of functional deterioration, institutionalization, and death among Medicare beneficiaries ages 65 and older with disabilities

Study Design

Researchers used mixed methods to develop statistical models for identifying risk factors among older people with disabilities for three health outcomes: functional deterioration, institutionalization, and death.

In the first phase of the study, researchers conducted focus groups with clinicians and people with and without disabilities to discuss experiences with the healthcare system, barriers to health care, and how disability influences care. From the focus groups, the team identified risk factors that people with disabilities believe are important for predicting functional deterioration, institutionalization, and death.

In the second phase of the study, researchers analyzed data from 21,264 people in the Medicare Current Beneficiary Study (MCBS) ages 65 and older. Among beneficiaries, 56% were ages 65–74, 34% were 75–84, and 10% were 85 and older. Also, 56% were female, 82% were white, 8% were black, and 7% were Hispanic.

Researchers modeled the association between risk factors, such as sociodemographic characteristics or health conditions, and the three health outcomes of interest for people in the study sample. The researchers derived models from 60% of the MCBS sample and validated them with the remaining 40%. Using risk factors identified in the models, the researchers developed a risk score to predict the chance of experiencing functional deterioration, institutionalization, or death within two years.

Two patient advocates were co-investigators on the research team.

Results

In all, 19 risk factors predicted functional deterioration, institutionalization, and death among Medicare beneficiaries with disabilities: male gender, being age 85 or older; education below a high school level; being black or Hispanic; not living with a spouse; having only Medicare insurance; using a proxy for healthcare decisions; being in a more disabled subgroup at baseline; and having Alzheimer’s disease or dementia, angina pectoris or coronary heart disease, diabetes, emphysema or asthma or COPD, a mental or psychiatric disorder, Parkinson’s disease, stroke or brain hemorrhage, complete or partial paralysis, hearing impairment, vision impairment, or myocardial infarction.

The model predicting functional deterioration included 16 of these 19 risk factors, the model predicting institutionalization included 11, and the model predicting death included 15 (all statistically significant at the p<0.05 level). The following risk factors were included in all three models:

• Male gender
• Being age 85 or older
• Not living with a spouse
• Having only Medicare insurance
• Being in a more disabled subgroup at baseline
• Having Alzheimer’s disease or dementia
• Having diabetes

The models produced risk scores that discriminated between patients with no change in disability versus functional deterioration over time, no change versus institutionalization, no change versus death, functional deterioration and institutionalization, functional deterioration versus death, and institutionalization versus death (M-index = 0.687).

Limitations

The results of the study only apply to Medicare beneficiaries ages 65 and older. The study only identified predictive risk factors, not causal relationships, with functional deterioration, institutionalization, or death.

Conclusions and Relevance

Statistical models may help clinicians identify patients who are at risk for worsening health outcomes. Older people with disabilities and their clinicians could monitor risk scores to assist in disability management.

Future Research Needs

Future research could assess the effect of using predictive models to identify older patients in greatest need of interventions to prevent future deterioration.