Objectives: The learner will be able to: 1. Discuss and compare reliability, validity, and sensitivity/specificity of the Medical Outcome Study Short Form-36 Physical Function Subscale (SF-36PF) and the Barthel Index (BI). 2. Designate the psychometrics of each instrument relative to the prediction of long term outcomes in the SAH population.
Design: This study used a prospective comparative design to determine reliability, factor structure and validity of the SF-36PF and BI in the subarachnoid hemorrhage (SAH) population.
Population, Sample, Setting, Years: Eighty-nine male and female subjects admitted to the University of Pittsburgh Medical Center, Presbyterian University Hospital with a diagnosis of severe SAH from 6/99 through 4/2001 were recruited into this study. The sample was 76% female, 75 % Caucasian with a mean age of 53 (SD=12.9).
Concept or Variables Studied Together: Initial injury severity scores [Glasgow Coma Score (GCS), Hunt & Hess score (H&H), and Fisher grades] were collected upon admission. Glasgow Outcome Score (GOS), Modified Rankin Score (MRS), BI, and SF-36PF were assessed via telephone interview at 3 months (N=89), 6 months (n=80), and 12 months (n=58) post injury.
Methods: Data were prospectively collected as part of an on-going NIH funded study. Internal consistency was assessed using Cronbach's coefficient alpha. Post-dictive validity was assessed by correlating 3 month SF-36 PF and BI scores with indices of severity of injury. Correlations with 6, and 12 month MRS and GOS assessed short term and long term predictive validity of the scales. Confirmatory factor analysis (CFA) was conducted to assess dimensionality of the underlying constructs. Receiver operating characteristic (ROC) curve analysis evaluated the sensitivity and specificity of the SF-36PF and BI at 3 months in predicting 12 month outcomes. Outcomes were dichotomized with good outcomes defined as a GOS>=4, and poor outcomes as a GOS<=3. Multiple regression analysis identified the best model for predicting functional outcomes in this population. Findings: Neither scale correlated with any demographic variables. The SF-36 PF and the BI both had excellent internal consistency with alphas>=.95. Both scores correlated significantly with GCS (SF-36PF- r=.38, p=.00; BI- r=.30, p=.00). The SF-36PF correlated with the H&H (r=-.24, p=.03), while the BI did not ( r=-.17, p=.11). Neither tool correlated with Fisher grade (SF-36PF- r=-.18, p=.09; BI- r=-.10, p=.34). The 3 month SF-36PF had high short and long term predictive validity as compared to the GOS (6 mo. r=.45, p=.00; 12 mo r=.39, p=.00) and MRS (6 mo. r=-.52, p=.00; 12 mo. r=-.45, p=.00). The 3 month BI also had high short and long term predictive validity as compared to the GOS (6 mo. r=.51, p=.00; 12 mo r=.38, p=.00) and MRS (6 mo. r=-.53, p=.00; 12 mo. r=-.46, p=.00). CFA for both the SF-36PF and BI showed unidimensional factorial model has a good fit to the data. ROC curve analysis revealed that a score of 14 on the SF-36 PF and a score of 12 on the BI provided optimal levels of sensitivity and specificity in predicting good versus poor 12 month outcomes. Multiple linear regression identified the 3 month BI to be the better predictor of 12 month outcomes.
Conclusions: The SF-36 PF and BI show similar psychometric properties in the SAH population as compared to other populations. They both correlated well with clinical indicators of severity of injury. They both show excellent reliability and validity when compared to the GOS and MRS in this population. However, when predicting long term outcomes from 3 months post injury, the BI was more accurate.
Implications: Recovery from SAH is a slow process requiring high levels of resources. An accurate and reliable tool to predict outcomes after a SAH may help the nurse design a care plan that minimizes resource utilization while maximizing its effect on outcome.
Back to Analysis and Design of Measurement Scales
Back to The Advancing Nursing Practice Excellence: State of the Science