Bruggeman S, Cabanela R, Williams AR, McMurtry EK, Naessens JM; AcademyHealth. Meeting (2005 : Boston, Mass.).
Abstr AcademyHealth Meet. 2005; 22: abstract no. 4051.
Mayo Clinic Rochester, Health Care Policy and Research, 200 SW 1st St, Rochester, MN 55905 Tel. (507) 284-4582 Fax
RESEARCH OBJECTIVE: To evaluate reliability and validity of the IHI 19 item Safety Climate Survey using groups of healthcare providers at a large mid-western hospital as respondents. STUDY DESIGN: Questionnaires containing the 19 IHI Safety Climate Survey items were distributed via intranet to a random sample of 1476 hospital nursing staff including registered nurses, licensed practical nurses, and patient care assistants; 1106 consulting physicians with recent hospital experience; and 641 resident physicians who had participated at least one year in the Residency Program at the hospital. The response rate was 45.4%. Respondents answers were analyzed using principal component factor analyses and Item Response Theory (IRT). A screen test of the eigenvalues from the factor analysis was used to examine dimensionality. MULTILOG 7.03 software (Samejima's model) was used to calculate item parameters and produce Item Characteristic Curves (ICC) and Total Information Curves. POPULATION STUDIED: 1451 clinical survey respondents consisting of 733 nurses, 485 physicians and 232 residents at a large mid-western hospital. PRINCIPAL FINDINGS: After principal component factor analysis (varimax rotation) was done, a screen test identified 2 factors. The factors loaded on 13 of IHI questionnaire items, with one loading on both factors. The first factor loaded on 8 items, and the second loaded on 6, including the cross-loading. The maximum variance accounted for on both factors combined was 38.6%.MULTILOG parameter calculations found that each of these factors had 2 items with alphas less than 1.00, indicating a very poor discrimination of traits. For items on each of the factors, MULTILOG produced four beta values, almost all were negative. As a consequence, no Total Information was provided on the upper part of the trait distributions (> 1 SD) on these factors. CONCLUSIONS: It is not clear what exactly is being measured on either of these factors. The Total Information provided on either factor is poor, with information being concentrated at the lower ends of the trait distributions. Many of the 19 items on the IHI questionnaire also appear to be redundant, since 11 alphas are less than 1.00. This suggests that many items could be dropped from the questionnaire. IMPLICATIONS FOR POLICY, DELIVERY OR PRACTICE: Although measuring Safety Climate has become popular, the generalized use of non-validated instruments and unjustified confidence in the results of their use may lead to undesirable consequences including less safety for patients not more.
Publication Types:
Keywords:
- Blood Transfusion
- Climate
- Demography
- Factor Analysis, Statistical
- Humans
- Injections
- Nursing Staff, Hospital
- Questionnaires
- Research Design
- Safety
- methods
- hsrmtgs
UI: 103623514
From Meeting Abstracts