Frailty among critically ill patients in India: Prevalence and association with patient-centered outcomes

Our prospective registry-embedded observational study will evaluate the prevalence of frailty at ICU admission and test the association between admission frailty and ICU outcomes. We will include all adult critically ill patients admitted to the ICU during the study period. Our estimated sample size is 800 and study duration is 2 years.

Objectives:

1.     Quantify the prevalence of frailtyamong critically ill adults admitted to 4 ICUs.

2.     Quantify the association betweenfrailty and ICU mortality

3.     Determine factors associated withfrailty

ResearchQuestions:

Among critically ill adults admitted to 4intensive care units in India:

1.     What is the prevalence of frailty?

2.     Is there an association betweenfrailty and ICU mortality?

3.     What factors are associated withfrailty at ICU admission?

Inclusion Criteria:

-       Adult patients (≥ 18 years old)

-       Admitted to the intensive care unitfrom the emergency room/ or from the inpatient ward

Exclusion criteria: Lack ofinformed consent

StatisticalAnalysis:

Hypotheses: Frailtyis associated with worse outcomes, specifically higher ICU mortality.

Null hypothesis: There isno difference between the frail and non-frail in ICU mortality.

Alternate hypothesis:There is a difference between frail and non-frail critically ill patients inICU mortality.

Sample size and power:

Based on data from high-income countries, theprevalence of frailty is ranges from 25 to 30%.4,7 There is no dataon the prevalence of frailty among critically ill patients in India. One studyfrom the community documented a prevalence of 26%5. In feasibilitywork led by us (unpublished data), the total proportion of vulnerable (11%) andfrail patients (8%) was 19%. For the purposes of sample size calculation, wewill hence assume a prevalence of frailty of 10%.

In our setting, baseline mortality is20%.  For sample size calculation weassumed that the risk is 15% higher in the frail cohort (i.e. 35% would die).We set the significance level to 5% and the type 2 error at 10%.

Based on this framework, we estimated adesired sample size of 720 patients. Assuming a 10% loss to follow-up, weinflated our sample size to 800 patients.

Analytic approach:

We will report categorical variables aspercentages and continuous variables as mean ± SD or median and IQR (for skeweddata). We will describe the prevalence of frailty in our cohort. We willanalyse the association between frailty and mortality using a chi square testand develop a multivariable logistic regression model to adjust for potentialconfounders. All analyses will be two-tailed and will be performed using Stata14.

For analysis of risk factors associated withfrailty, candidate variables were chosen based on previous research or based onbiological and contextual probability. Stepwise logistic regression will beperformed to select independent predictors of frailty in this cohort.