Table 2.

Methods and Results of Sensitivity Analyses Performed

Variable of Interest	Methods Used to Perform Sensitivity Analyses	Results
ADLs	Each individual ADL score was compared to the summary score. Two different grouped levels of summary score were tested.	No significant improve in the model
Psychiatric diagnosis	Types of psychiatric diagnoses (eg, anxiety, psychosis) were compared to the number of diagnoses recorded (one, two, three or more vs none).	No significant improve in the model
Age	Categorical groupings in Table 1 were compared with continuous variable versus categorical groupings with the last category of ≥60 years.	No significant improve in the model
24-Hour residential setting	Entire analysis was re-run using only clients from 24-hour residential settings because their representation in the database was relatively complete.	All variables remained in the model except guardian and summary ADL score. Effect sizes were similar but slightly higher for all remaining variables in the model. C statistic = 0.723
Recent influenza vaccination	Characteristics of influenza vaccine were negative; women who did not receive a mammogram were analyzed and compared with women who did receive a mammogram and had influenza.	Less able to communicate (64% vs 79%) More likely to have a guardian (58% vs 46%) Less likely to have one or psychiatric diagnosis (54% vs 62%) More likely to have high ADL need (31% vs 19%) More likely to require sedation (22% vs 15%)