Consequences of reducing nonresponse in a national telephone survey.

Critics of public opinion polls often claim that methodological shortcuts taken to collect timely data produce biased results. This study compares two random digit dial national telephone surveys that used identical questionnaires but very different levels of effort: a "Standard" survey conducted over a 5-day period that used a sample of adults who were home when the interviewer called, and a "Rigorous" survey conducted over an 8-week period that used random selection from among all adult household members. Response rates, computed according to AAPOR guidelines, were 60.6 percent for the Rigorous and 36.0 percent for the Standard study. Nonetheless, the two surveys produced similar results. Across 91 comparisons, no difference exceeded 9 percentage points, and the average difference was about 2 percentage points. Most of the statistically significant differences were among demographic items. Very few significant differences were found on attention to media and engagement in politics, social trust and connectedness, and most social and political attitudes, including even those toward surveys.

Chapter V. A comparison of CATI and non-CATI questionnaires.

It appears likely that a greater proportion of surveys will use the telephone as a medium of sampling and data collection in the future. As software developments proceed and computer hardware costs shrink, many of these will use CATI systems. This experiment can be used as a benchmark for the transition to CATI because it provides documentation on potential problems with changing modes. For most criteria, there are only small differences between CATI and non-CATI interviewing in this project. The criteria include response rates, reactions of the interviewer and respondent, and most health statistics of interest. There are, however, some exceptions to this finding of equivalence between methods. The first exception is the result that the average number of minutes per CATI interview exceeded that for non-CATI interviews. There also is some evidence that the interviewer variability estimates tend to be lower in CATI than non-CATI. Finally, there is evidence of lower skip error problems in the CATI interviews. The first of these results affects survey costs; the second and third, survey error. The first may be a function of software or hardware choice and thus can be addressed in new CATI designs. The second and third will be of benefit to all CATI systems in the future.

Chapter VI. Measurement of interviewer errors in the SRC Telephone Survey.

The most important finding of this section is that unusually low levels of interviewer effects were measured in the telephone survey. This result may be due to the stringent controls on interviewer behavior that were introduced in this study but were absent in past studies. These low interviewer variances inhibited attempts to explain interviewer variability on the health variables. Because there was little interviewer variability, the correlates among the monitoring data were weak, and the findings did not exhibit consistency over variables.

Chapter VII. Nonsampling bias and variance in the SRC Telephone Survey data.

This small exercise was a check on the nature of the effects of the experimental interviewing behaviors, one that attempted to dismiss the possibility that the increases in reporting were coming at the expense of greater interviewer variance. The findings presented are limited by the small number of statistics examined, but they suggest that the possibility that the experimental behaviors merely move errors from bias terms to variance terms is unlikely to be experienced. It is also unlikely that the experimental effects do not uniformly decrease interviewer variance and they appear to be reducing response bias. The measurement of the net effect of the treatments on mean square error requires estimates of the relative size of total response variance composed both of interviewer and respondent variability and response bias associated with the experimental procedures.

The effects of respondent rules on health survey reports.

Survey researchers believe that self reports, in general, are more accurate than reports obtained by proxy. This paper focuses on the reassessment of previous self/proxy comparisons and presents findings from a telephone adaptation of the National Health Interview Survey (NHIS) designed to investigate response error associated with self and proxy reports. Unlike previous studies in which the type of report is confounded with characteristics of the population home at the time of the interview, the design of this study (random allocation to self or proxy report) allows comparison of reports from similar populations. The results show that when self response is limited to a randomly selected respondent, the self respondents report fewer health events for themselves versus for others in their household.

Computer assisted telephone interviewing: effects on interviewers and respondents.

The effects of using a computer assisted telephone interviewing system on response distributions, interviewer behavior, and other nonsampling errors are measured through a survey experiment which randomly assigned cases either to a paper questionnaire or a CATI version. There were few differences in response distributions obtained in the two modes, some evidence of reduced interviewer variance among CATI cases, but little difference in response rates between the methods. Interviewers, who conducted interviews in both methods, tended to have no clear preference between them, but favored the CATI system for ease of following the questionnaire and the paper version for ease of making changes to prior answers.

Aortico-pulmonary paraganglioma: report of a case with ultrastructural study and review of the literature.

The light microscopic and ultrastructural features of an aortico-pulmonary paraganglioma (A-PP) are presented. The tumor was characterized by organoid clustering of neoplastic chief cells to form Zellballen. Argyrophilic granules were demonstrated within chief cell cytoplasm using a modified Grimelus technique. Ultrastructurally, three distinct cell types were present within the tumor: endothelial cells, pericytes and neoplastic chief cells. Membrane-bound neurosecretory granules were present and measured 100 to 2000 nm in diameter. "Light" and "dark" chief cells were less distinct than previously reported in other head and neck paragangliomas. Analysis of the 36 documented A-PP reported in the English literature reveals that the tumor has been either incompletely excised or has been considered unresectable in one-third of the cases. The reported surgical mortality is 9%, or approximately equal to the incidence of malignant behavior. The treatment of choice is surgical resection but when this is not possible, radiation may be a useful adjunct in therapy.