Joseph S. Wislar

Response Rates and Nonresponse Errors in Surveys

A S PERHAPS THE MOST WELL KNOWN OF ALL social science methodologies, survey research findings are commonly reported in the professional literature. Indeed, many biomedical journals routinely publish articles that report results from high-quality survey studies. Discerning unbiased survey findings can be challenging, however, to both editors and readers, necessitating some reflection as to the multiple sources of error in surveys and how to assess them. In this Viewpoint, we provide an overview of one of these sources of survey error—nonresponse bias—and briefly consider approaches for estimating, evaluating, and reporting it.

Childhood abuse and the use of inhalants: differences by degree of use.

OBJECTIVES Using two existing ethnographic studies of drug-involved adults, this study evaluates the association between child-abuse victimization and levels of involvement in inhalant use. METHODS Historical accounts of childhood exposure to physical or sexual abuse were compared among nonusers of inhalants (n = 197), light inhalant users (n = 64), and heavy inhalant users (n = 24). Crude and adjusted odds ratios were used to compare informants with no history of inhalant use with those having a history of light inhalant use and those having a history of heavy inhalant use. RESULTS Heavy inhalant use was associated with history of any child abuse (adjusted odds ratio [OR] = 4.6) and physical abuse (adjusted OR = 3.8). Light inhalant use showed no association with child-abuse history. CONCLUSIONS Child abuse may be an important correlate of extensive involvement in inhalant use. The findings invite speculation with respect to a hypothetical causal role for child abuse in the etiology of inhalant use. The lack of support for causality in this study underscores the need for replication and more carefully designed longitudinal research.

Effects on Alcohol Use and Anxiety of the September 11, 2001, Attacks and Chronic Work Stressors: A Longitudinal Cohort Study

OBJECTIVES We hypothesized that chronic stressors associated with an everyday social role (work) would interact with a traumatic macrosocial stressor (the terrorist attacks of September 11, 2001) in predicting mental health status (during the fall of 2001). METHODS We used mail surveys returned as part of wave 3 of a workplace cohort study, both before and after September 11, 2001, to assess decision latitude, sexual harassment, generalized workplace abuse, psychological distress, and alcohol use. We also used regression analyses to assess the main effect of September 11 and interactions between September 11 and stressors, after control for baseline mental health. RESULTS The main effect of September 11 on elevated alcohol use was significant for women but not for men. For women, work stressors significantly interacted with experiencing the events of September 11 to affect alcohol use and anxiety outcomes. CONCLUSIONS Women experiencing chronic work stressors were most vulnerable to elevated psychological distress and alcohol use after September 11, 2001.

Accuracy of Parent Mental Health Service Reporting: Results From a Reverse Record-Check Study

OBJECTIVE To examine the feasibility of conducting a reverse record-check study to validate parent reporting on child mental health service use and to examine the accuracy of parent reports of child mental health services. METHOD Information about child service use was abstracted from medical records and subsequently compared with reports provided by caregivers in telephone interviews. A sample of children using outpatient psychiatric services was compared with a sample of children using outpatient orthopaedic services. Rates and correlates of successful caregiver tracking and of service use reporting accuracy were explored. RESULTS Caregivers of nearly 30% of all index children were contacted and interviewed. Parent reports of lifetime mental health service use were more accurate than reports of lifetime receipt of orthopaedic services. Elapsed time between survey interview and last treatment episode was negatively associated with reporting accuracy. Number of clinic visits was positively associated with reporting accuracy. Preliminary findings suggested that questions about mental health services may be considered sensitive by parents whose children use them. CONCLUSIONS Reverse record-check studies based on telephone interviews are potentially problematic with a sample drawn from a large, inner-city medical center. Findings suggest that memory difficulties may be a more important correlate of reporting accuracy than response editing (social desirability). Potential discomfort with disclosure of mental health service use does not result in service use underreporting.

The Impact of Interviewer Characteristics on Drug Use Reporting by Male Juvenile Arrestees

We evaluated the importance of interviewer and subject effects on cocaine and marijuana use disclosure in a sample of over 3,000 male juvenile arrestees. Analyses evaluated the viability of Social Attribution and Conditional Social Attribution models of interviewer effects. The viability of alternative models was investigated in the context of comparative analyses excluding and including statistical adjustments for the clustering of responses by interviewers. Interviewer effects were more salient in models predicting marijuana disclosure than in models predicting cocaine disclosure. Logistic regression analyses provided support for Social Attribution and Conditional Social Attribution models of interviewer effects. Models suggested large interviewer cluster effects. Cluster adjustment altered interpretation of effects for both cocaine and marijuana. Subject race/ethnicity effects were salient in models predicting disclosure for both drugs, but were especially large in models predicting cocaine disclosure.

The Validity of Juvenile Arrestees' Drug Use Reporting: A Gender Comparison

Few studies investigating the validity of drug use reporting among arrestees have included an adequate sample of females to study gender differences. This exploratory study examined gender differences in the validity of drug use reporting among juvenile arrestees, using a gender-matched sample (n = 6,377) drawn from the Drug Use Forecasting Program,1 1992-1996. Self-reported marijuana and crack and/or cocaine use was compared to urinalysis results to test gender differences in the accuracy of disclosure. Among urine positives, girls were more willing than boys to disclose past month and lifetime marijuana use. Gender was not a significant main effect for cocaine use reporting but interacted with race/ethnicity and family structure in predicting valid disclosure. Hispanic girls were significantly more likely than Hispanic boys to underreport recent cocaine use. Among arrestees from homes with one or no parents, girls were significantly more likely than boys to disclose recent cocaine use.

Sexual Harassment, Generalized Workplace Abuse and Drinking Outcomes: The Role of Personality Vulnerability:

Interpersonal stressors, including workplace harassment and abuse, have been associated with deleterious drinking outcomes. Studies have also shown that the relationship between drinking and stress can be moderated by personality characteristics. This study examines the roles that neuroticism and narcissism play in the perception of sexual harassment and generalized workplace abuse. Utilizing data from a two-wave study of university employees, we found that respondents high on neuroticism and narcissism were significantly more likely to report abusive or harassing work environments at both waves of the study. However, significant associations between harassment or abuse and deleterious drinking outcomes remained after controlling for personality characteristics. The implications of these results are discussed.

Youth suicide-related visits in an emergency department serving rural counties: Implications for means restriction

A large proportion of suicide attempts result in injuries requiring medical attention, but little is known about youth presenting to rural emergency departments (EDs) for suicide-related behavior. We conducted a chart review of youth receiving mental health (MH) evaluations during 1994 in an ED serving rural counties. Forty percent of the these visits were for suicide-related events. These youth were more likely to be older, female and white than youth receiving MH evaluations for reasons not related to suicide. No differences were found on recent substance use, non-suicidal aggression, or MH service contact when comparing suicidal to non-suicidal youth. Youth receiving MH evaluations were homogenous with respect to suicide risk factors. Chart reviews provided no evidence that means restriction education was provided. We discuss broader implications of these findings for the role of means restriction in EDs.

Can self-reported drug use data be used to assess sex risk behavior in adolescents?

To better understand and control the spread of sexually transmitted diseases among high-risk youth, we must first acquire reliable reports of sexual risk behavior. This study evaluates one potential method for validating such reports. We examined the association between marijuana and cocaine use reporting patterns and the number of reported recent sexual partners in a sample of juvenile arrestees/detainees. Using urinalysis to validate self-reported drug use, we categorized drug use reporting patterns into four groups: overreporters, underreporters, honest users, and honest nonusers. Analyses showed, in general, that overreporters reported more sexual partners than either underreporters or accurate reporters, suggesting that overreporters of drug use may also exaggerate sex partner reports. Findings suggest a new method for validating self-reported sexual behavior and provide a challenge to theories of juvenile delinquency.

The Utility of Debriefing Questions in a Household Survey on Drug Abuse

Since subjects are often uncomfortable disclosing sensitive information, questions gauging respondent reaction to survey questions about drug abuse may aid in the interpretation of responses. Such debriefing questions can be worded subjectively, i.e., asking subjects about their own reactions to questions, or projectively, i.e., asking subjects about their perceptions of the reactions of others to questions. Using data from a study employing both types of debriefing questions, we examined whether question wording made any difference. We evaluated the extent to which drug reporting was associated with respondent reaction to the survey and whether reactions varied by type of debriefing question. A factor analysis of debriefing questions appended to an experimental household survey on drug abuse yielded factors reflecting distinct subjective and projective dimensions. Analyses suggested that debriefing question wording does make a difference with respect to respondent reaction. Contrary to expectations, analyses suggested that those reporting drug use expressed less comfort on subjectively worded items and more comfort on projectively worded items. Subjects who self-administered their survey also reported lower levels of subjective comfort. The projective measures derived from analyses were associated with three demographic variables: Younger subjects reported higher levels of projective ease than older subjects. Those with some college reported higher levels of projective ease than those with the least amount of formal education. African-Americans reported lower levels of projective ease than White/Other subjects. Implications for the design, interpretation, and analysis of drug use surveys are considered.