Predictive Ability of an Illness Severity Measure: Implications for Nursing Research