A new risk assessment tool may accurately predict 3-year survival after a dementia diagnosis, thereby enabling dementia patients and their families to engage in shared decision making and advance care planning.
Investigators used a large Swedish database to follow during a 10-year period more than 50,000 people aged 65 years or older who had been diagnosed with incident dementia.
Slightly more than 40% of patients died during the follow-up period. The median survival time was roughly 5 years. On the basis of analysis of sociodemographic, medical, and cognitive factors, the researchers developed two simple prediction calculators, one for use in the primary care setting and the other for use in specialized memory care clinics.
The final model for primary care settings included age, sex, cognitive ability, and comorbidities; the model for memory care centers included all these factors plus dementia subtype.
“Conversations about survival in dementia are important for patients but are not easy to carry out,” senior author Sara Garcia-Ptacek, MD, PhD, postdoctoral fellow, researcher at the Department of Neurobiology, Care Sciences and Society (NVS) at Karolinska Institutet, and neurologist at General South Hospital, Stockholm, Sweden, told.
“However, when done correctly, these conversations are very valuable for patients, and we hope this tool will help initiate this difficult type of conversation,” she said.
The study was published online December 16 in Neurology.
Several recommendations, including one from the American Geriatrics Society and the Lancet Commissions’ Dementia Prevention, Intervention, and Care, published in 2017, emphasize the importance of talking to families and patients about the patient’s needs and wishes toward the end of life and incorporating information on life expectancy into clinical decisions; however, clinicians “appear to encounter several barriers in this process,” the authors write.
One barrier is difficulty in incorporating information about the patient’s life expectancy into these conversations, as well as difficulty in discussing prognosis — a challenge that may be caused, at least in part, by the lack of a proper prediction model, they suggest.
To address this problem, the researchers utilized data from the large Swedish Dementia Registry (SveDem), which monitors the diagnosis, treatment, and care of people with dementia in Sweden. Participants were drawn from 829 Swedish healthcare centers that included memory clinics and primary care settings.
All participants in the current analysis, which was conducted from 2007 to 2015 (n = 50,076; 59.4% women; mean age at diagnosis, 81.6 years; interquartile range [IQR], 76.5 – 86.0 years), had late-onset dementia, defined as dementia that begins at age 65 years or older. Survival was assessed for up to 9.7 years, until 2016.
Alzheimer disease was the most common dementia type, followed by dementia of unspecified cause, mixed dementia, and vascular dementia (31.8%, 20.6%, 20.4%, and 20.0%, respectively).
Most patients (90.4%) were living at home.
The investigators collected data regarding socioeconomic factors, including age, sex, and living situation; comorbidities, based on the Charlson Comorbidity Index (CCI); medication use; cognitive status, based on the Mini–Mental State Examination; dementia type; and date of death.
To create the risk predictors, they used Cox proportional hazards regression models to analyze these factors.
The estimated median survival time following diagnosis was 4.8 years (IQR, 2.6 – 7.6 years).
By the end of the 9.7-year follow-up period, 40.6% of patients (n = 20,828) had died. More died in primary care than in specialized settings (n = 10,863 vs 9965, respectively). However, the risk for mortality was higher in those diagnosed at memory clinics, after correction for the included variables (hazard ratio, 1.15; 95% confidence interval [CI], 1.11 – 1.19).
A forward selection procedure found that the most common predictors (in order) in the memory clinic setting were higher age at diagnosis, lower global cognition, more comorbidities, having non-Alzheimer dementia, male sex, living alone, and taking a higher number of drugs.
In the primary care setting, the most common predictors (in order) were higher age at diagnosis, more comorbidities, lower global cognition, male sex, living alone, and taking a higher number of drugs. Dementia subtype was not included in the primary care model because it was often unspecified in that setting.
The final 3-year risk calculator for primary care settings included age, sex, cognitive status, and comorbidities. For specialized memory clinics, it also included dementia subtype.
The risk tool divides 3-year survival probability into six categories, ranging from low probability of survival (<0.40) to high probability (0.50 – 1.00).
The investigators reported that their model yielded c indexes for men and women of 0.71 (95% CI, 0.70 – 0.72) and 0.72 (95% CI, 0.71 – 0.73), respectively, in the memory clinic setting. In the primary care setting, the c indexes for men and women were 0.70 (95% CI, 0.69 – 0.71) and 0.71 (95% CI, 0.70 – 0.71), respectively.
“This is substantially higher than the c index based on age and sex only, which was 0.65 (95% CI 0.64 – 0.65),” they state.
The models also showed “good calibration” with “strong agreement between predicted and observed survival curves, indicating good accuracy.”
“Some dementia types progress faster and have higher mortality,” Garcia-Ptacek commented.
“Comorbidity plays a part. For example, patients with vascular dementia can be expected to have high cardiovascular comorbidity, including strokes, which would contribute to lower survival,” she continued.
She noted, however, that some differences were apparent after controlling for comorbidity, suggesting that the “pathophysiology of some particular dementia disorders might just be especially aggressive.”
One study limitation is that the acceptability of the model was not tested with patients, she said.
Additionally, the model was “created for Swedish dementia patients and would require validation to extrapolate to other populations.”
Commenting on the study, Maria C. Carrillo, PhD, chief science officer of the Alzheimer’s Association, who was not involved with the study, said, “As the number of people living with Alzheimer’s and all dementia continues to grow, a better understanding of survival time in people with dementia, and better tools to calculate it, are needed for clinicians — and for researchers too, as we seek better treatments and effective preventions.”
The current study “is evidence that work is going on in this area,” she said.
She cautioned that because the study “was done in one country, in a relatively homogeneous study group, there may be limitations in the ability to generalize the results to other populations,” so to confirm the findings, more studies that test the model in “more diverse study groups” and in different healthcare systems are necessary.
Carillo noted that the Alzheimer’s Association supports “early and accurate diagnosis that is communicated quickly, clearly, and honestly, inclusion of the person living with dementia in the planning and decision-making process as long as possible, and referral to community-based support and services, such as those available through Alzheimer’s Association, and to the opportunity to participate in research studies.”
Garcia-Ptacek added, “We combined our joint team knowledge of dementia — clinical and epidemiological — and aimed to develop a useful tool, which will hopefully make a difficult conversation somewhat easier.”
The study was conducted with researchers from Radboud University Medical Center in the Netherlands and was financed with grants from the Swedish Research Council, the Swedish Research Council for Health, Working Life and Welfare (Forte), the Swedish Society for Medical Research, the Swedish Order of St John, Region Stockholm (ALF), the Foundation for Geriatric Diseases at Karolinska Institutet, the Loo and Hans Osterman Foundation for Medical Research, Radboud University Medical Center in the Netherlands, and Alzheimer Nederland. Garcia-Ptacek and coauthors and Carrillo report no relevant financial relationships.
Neurology. Published online December 16, 2019.