The organization of patient care within hospitals has been shown to be associated with patient mortality, as well as with emotional exhaustion and job dissatisfaction among patient caregivers (nurses). These results derive from surveys of organizations (hospitals) in which the ultimate and primary sampling units are individuals (nurses). Nurses are sampled and surveyed regarding their individual background characteristics and social life feelings, plus organizational attributes of the hospitals in which they work. Responses are aggregated to the hospital level where they are linked with organizational data such as comorbidity-adjusted atient death rates. This survey sampling strategy minimizes non-response at the organizational level; because hospitals are not required to provide the sampling frame for nurses, they (e.g., their administrative or corporate leadership) are not able to "opt out" of the study. However, the sheer number of interviews required to represent reliably most if not all hospitals in a state places a great burden on a survey organization. A result is response rates *at the individual level* that may only reach 50%, as in the 1999 survey, which nonetheless had sufficient numbers of responses within hospitals for an 88% response rate across the 210 acute care hospitals in Pennsylvania. A second NIH-funded survey is currently in the field, replicating the Pennsylvania sample and extending it to nurses (hence hospitals) in California and New Jersey. The specific aims of this application are (1) to conduct an intensive "double sample" of 1,000 non-respondents; and (2) to use this sample to either validate the representativeness of the far larger survey or to suggest weights and models for missingness at random that can be used in the analysis of these survey data. Results from the 1999 study have received much attention in the medical and nursing communities since they bear on how organizations--hospitals--may adjust and adapt to improve the health care of the general public. But there is inevitably some skepticism attendant to high rates of non-response: Might estimates of hospital characteristics be biased by differential characteristics of nurses who are and are not willing to evaluate the organizations in which they work? Being able to compare the responses of those inclined to participate under "normal" survey conditions with those motivated to respond after initial non-response can and should bolster confidence in the findings from these large-scale organizational studies.
Principal Investigator
Abstract
Funded By
Award Dates
-