Application of computer-assisted interviews to sexual behavior research

Collection of sensitive data with the use of video-enhanced, computer-assisted, self-administered interviews (V-CASI) has the potential to reduce interview bias and improve the validity of the study. The purpose of this study was to compare responses to sensitive questions elicited by V-CASI and by face-to-face interview (FTFI) methods. Women attending a New Orleans, Louisiana, public family planning or sexually transmitted disease clinic from July 1995 to July 1996, diagnosed with a Chlamydia trachomatis infection responded to eight close-ended behavioral questions (four socially undesirable, two socially desirable, and two neutral behaviors) using both FTFI and V-CASI techniques in a randomized crossover design. Of the 280 women included, the mean age was 23 years, 95 percent were African American, and 71 percent felt comfortable using computers. While kappa scores indicated good-to-excellent agreement between interview techniques, women tended to admit to socially undesirable behaviors more often on V-CASI compared with FTFI. Thirty percent of the women gave a discrepant response between V-CASI and FTFI toward social desirability. Women who reported a socially undesirable behavior in V-CASI (i.e., more than two sex partners and infrequent condom usage) were more likely to have a discrepant response. Utilization of the same logistic regression model to predict condom use yielded different results when data from V-CASI were used compared with data from FTFI. The V-CASI technique can reduce social desirability bias and improve validity in research requiring information on sensitive sexual behaviors.

Reproducibility of oral contraceptive histories and validity of hormone composition reported in a cohort of US women

Histories of oral contraceptive (OC) use were reported by 116,686 women aged 25 to 42 years in the Nurses' Health Study II on a self-administered questionnaire accompanied by a color photo booklet of all OC preparations ever marketed in the US. To evaluate the quality of this information, we compared the responses of a randomly selected sample of 215 participants with their data from a subsequent, detailed telephone interview using a structured life events calendar. Agreement for a history of ever having used OC was high between the two methods (exact agreement 99%). Reported durations of use were equivalent (mean duration 42.7 months by telephone interview and 44.6 months by questionnaire). The Spearman correlation for duration of use calculated from the two methods was 0.94 (p < 0.0001). For a subset of women for whom we were able to obtain OC prescription records, the medical record confirmed the use of an identical or equivalent brand in 75% of intervals of reported use. Acceptably valid OC histories were obtained with a self-administered questionnaire.

The validity of self-reported condom use among adolescents

BACKGROUND

Research and public health interventions designed to reduce the risk of sexually transmitted diseases (STDs) often are based on self-reported condom use. Yet, validation of self-reported condom use, in particular with adolescents, has rarely been described in the literature.

METHODS

Baseline data were obtained from 540 adolescents, 13-21 years of age, enrolled in a 1-year longitudinal study of health beliefs, sexual behaviors, and STD acquisition. Of the 445 participants reporting to be sexually active, 404 (90.8%) agreed to a complete physical examination, including a genital examination, with STD screening after completing the self-administered written questionnaire. Participants' written self-report of condom use was compared to histories obtained by clinicians and laboratory diagnosis of acute STDs to assess validity of written self-report.

RESULTS

Complete data were available for 321 females and 77 males of whom 52 females and 5 males had laboratory evidence of 63 infections. Although three individuals who had STDs reported to be consistent users of condoms, a significant association (P < 0.05) was found between those who reported more frequent condom use with the last two partners and the absence of STDs.

CONCLUSION

In this group of adolescents, self-report of condom use with the last two partners was associated with the absence of an acute STD. This finding suggests that self-reported condom use is a valid indicator of risk for STDs, with implication for those working with adolescents clinically and in research contexts.

Induced abortion and risk for breast cancer: reporting (recall) bias in a Dutch case-control study

BACKGROUND

In general, no association has been found between spontaneous abortion (naturally occurring termination of a pregnancy) and the risk for breast cancer. With respect to induced abortion (termination of a pregnancy by artificial means), the results have been more inconclusive. A positive association was found in five studies, no association was found in six studies, and a negative association was found in the only cohort study. It is thought that part of the inconsistency of the reported results may be attributable to reporting (recall) bias, since all but two studies on induced abortion used the case-control design and were based only on information obtained from study subjects. In comparison with breast cancer case patients, healthy control subjects may be more reluctant to report on a controversial, emotionally charged subject such as induced abortion. Thus, differential underreporting may be a cause of spurious associations in case-control studies.

PURPOSE

Our goal was threefold: 1) to evaluate the relationship between a history of induced or spontaneous abortion and the risk for breast cancer in a Dutch population-based, case-control study; 2) to examine reporting bias by comparing risks between two geographic areas (i.e., western regions and southeastern regions in The Netherlands that differ in prevalence of and attitudes toward induced abortion); and 3) to compare reporting bias in data on induced abortion with reporting bias in data on oral contraceptive use.

METHODS

Data analyzed in this study were obtained from 918 women (20-54 years of age at diagnosis) who were diagnosed with invasive breast cancer during the period from 1986 through 1989 and had been initially enrolled in a population-based, case-control study investigating oral contraceptive use and breast cancer risk. The women resided in one of four geographic areas that were covered by Regional Cancer Registries: two western regions (Amsterdam and West) and two southeastern regions (East and Eindhoven). Each case patient was pair-matched, on the basis of age (within 1 year) and region, with a control subject who was randomly selected from municipal registries that fully covered the Dutch population. Both the case patients and the control subjects were interviewed at home by the same trained interviewer, who used a structured questionnaire. Reporting bias was examined indirectly by comparing risks between the western and the southeastern regions of the country, which differ in the prevalence of and attitude toward induced abortion. Multivariate conditional logistic regression methods for individually matched case-control studies were used to estimate relative risks (RRs). Reported P values are two-sided.

RESULTS AND CONCLUSION

Among parous women, a history of induced abortion was associated with a 90% increased risk for breast cancer (adjusted RR = 1.9; 95% confidence interval [CI] = 1.1-3.2). Among nulliparous women, no association between induced abortion and breast cancer was found. Neither among parous women nor among nulliparous women was a history of spontaneous abortion related to the risk for breast cancer. The association between induced abortion and breast cancer was stronger in the southeastern regions of the country, which have a predominantly Roman Catholic population, than in the western regions (adjusted RR = 14.6 [95% CI = 1.8-120.0] versus adjusted RR = 1.3 [95% CI = 0.7-2.6], respectively; test of difference between regions, P = .017), suggesting reporting bias. Support for reporting bias as an explanation for the regional differences was also found in data supplied by both study subjects and their physicians on the use of oral contraceptives. In comparison with physicians, control subjects in the southeastern regions underreported the duration of their oral contraceptive use by 6.3 months more than control subjects in the western regions (P = .007)...

Analysis of failure in medical abortion

Medical abortion opens a new choice to women wishing to terminate a pregnancy. Increasingly, providers in the developing and developed world will begin to offer this option. Yet, the nomenclature and concepts used for measuring failure of surgical abortion are not directly adaptable because of important differences inherent in the method and in the way it is offered in a given setting. We propose that failures in medical abortion should be defined as a surgical intervention (whether vacuum aspiration or dilatation and curettage) performed for any reason. Such instances may be further classified into three types: user choice interventions, provider choice or error interventions, and true drug failures requiring intervention. Further description and examples of each type are given.

World population in a grid of spherical quadrilaterals

"We report on a project that converted subnational population data to a raster of cells on the earth. We note that studies using satellites as collection devices yield results indexed by latitude and longitude. Thus it makes sense to assemble the terrestrial arrangement of people in a compatible manner. This alternative is explored here, using latitude/longitude quadrilaterals as bins for population information.... The results to date of putting world boundary coordinates together with estimates of the number of people are described. The estimated 1994 population of 219 countries, subdivided into 19,032 polygons, has been assigned to over six million five minute by five minute quadrilaterals covering the world."

Response bias in a case-control study: analysis utilizing comparative data concerning legal abortions from two independent Swedish studies

Independent reports of legal abortions in two Swedish epidemiologic studies of breast cancer in young women, covering the same women and overlapping the same time period, have been compared in order to estimate a putative response bias. One study used case-control methods and obtained data by retrospective interviews from 317 cases and 512 controls. The other study was based on objectively documented information froma nationwide registry covering legally induced abortions. Analysis demonstrated a ratio between the odds ratios from the two studies of 1.5 (95 percent confidence interval 1.1-2.1) and an observed ratio of 22.4 (p less than 0.007) between underreporting of previous induced abortions among controls relative to overreporting among cases. This response bias may explain the tendency toward increased risk of breast cancer which, according to several case-control studies, appears to be associated with induced abortion.

Perinatal mortality rates do not contain what they purport to contain

Three cases of perinatal death, of which two needed to be included in and one excluded from national perinatal mortality statistics, were presented to 1004 specialist and trainee obstetricians in northern Belgium (ie, Flanders) and the Netherlands. Of the respondents (52%), 69% admitted that they would report none of the cases and 13% would report all of them. Overreporting occurred twice as often and underreporting ten times as often as correct reporting--only 6% would apply the current regulations for registration of perinatal mortality correctly in all three cases. Although the statutory regulations with regard to the cases were similar in the two countries, there were differences between Belgian and Dutch doctors in their reporting of these three cases. This indicates that Belgian and Dutch perinatal mortality statistics do not measure the same thing and that neither contains what it purports to contain. The findings cast doubts on the validity of using national perinatal mortality figures as indicators of perinatal health or perinatal care in and between European countries.

Impact of HIV infection on mortality and accuracy of AIDS reporting on death certificates

To assess the impact of HIV infection on mortality and the accuracy of AIDS reporting on death certificates, we analyzed data from 6704 homosexual and bisexual men in the San Francisco City Clinic cohort. Identification of AIDS cases and deaths in the cohort was determined through multiple sources, including the national AIDS surveillance registry and the National Death Index. Through 1990, 1518 deaths had been reported in the cohort and 1292 death certificates obtained. Of the 1292 death certificates, 1162 were for known AIDS cases, but 9% of the AIDS cases did not have HIV infection or AIDS noted on the death certificate. Only 0.7% of the decedents had AIDS listed as a cause of death and had not been reported to AIDS surveillance. AIDS and HIV infection was the leading cause of death in the cohort, with the highest proportionate mortality ratio (85%) and standardized mortality ratio (153 in 1987), and the largest number of years of potential life lost (32,008 years). The devastating impact of HIV infection on mortality is increasing and will require continued efforts to prevent and treat HIV infection.

The completeness of cancer registration in England: an assessment from the Oxford-FPA contraceptive study

The completeness of cancer registration in England for the period 1968-85 has been assessed in a cohort of 17,000 women who reported malignancies directly to the investigators. Of 325 cancers reported, 281 (86.5%) had been registered by mid-1987. Under-registration varied considerably between regional cancer registries. Eight (18%) of the 44 unregistered cancers were treated in private hospitals. Under-registration also varied considerably with cancer site: only 8% of 150 breast cancers were not registered, and at sites accounting for 79% of all tumours, under-registration was less than 15%; however, 40% of melanomas (20 cases) and 50% of lung cancers (6 cases) were not registered. Of 281 registered tumours, only 219 (78%) were notified to the investigators from the NHSCR at Southport, with a median lag-time of 2.5 years since diagnosis. There has been a tendency for notification of registered cancers to the investigator to become more prompt but less complete.

Age at menarche based on recall information

Censored likelihood methods which use recalled dates yield accurate estimates of the age at menarche in an urban population of black adolescent girls. The mean age at menarche is 12.44 (SE 0.08) years for a sample of 272 black adolescents in Philadelphia, PA. The methodology we describe is more efficient than logit or probit methods, which are not designed to utilize recall information. Alternative methods which use recall are biased if subject memory is influenced by length of time since menarche. Our methods also provide information on the nature of subject recall that may be useful in further analysis.

Maternal recall of symptoms associated with childhood deaths in rural east Africa

Verbal autopsies (VA) are frequently used to determine causes of death for individuals for whom there is no reliable clinical information regarding the terminal illness. VA interviews are used to note key symptoms and signs recalled by relatives of the deceased and diagnoses ascribed according to the symptom complexes. The VA technique assumes that individual disease entities have discrete symptom complexes and that these can be accurately recognized and recalled by the interviewees. We have examined the accuracy with which specific symptoms are recalled over time by mothers or normal guardians of 491 children who died on the paediatric wards of two district hospitals in East Africa. Kwashiorkor, measles, trauma, generalized convulsions and neonatal tetanus were all reported with a high degree of accuracy for children who died of these conditions and had low false positive rates for children without these conditions. Recall was similar within 1 month of death compared to recall after 6 months for most symptoms and signs except neonatal tetanus where false positive reports by mothers increased with time since death. Symptoms and signs commonly used to describe malaria, respiratory tract and diarrhoea-related deaths were reported by mothers to have been present during the terminal illness in 43% of cases where these features were absent. Recall abilities differed between the two communities studied for some symptoms and signs highlighting the importance of such studies in every setting where VA are applied.

Measurement error in the Current Population Survey: a nonparametric look

"This article utilizes an exact match file between the 1978 March [U.S.] Current Population Survey and administrative records from the Social Security Administration to analyze errors in the reporting of annual income using nonparametric methodology.... Three new findings are of interest: there is higher measurement error in cross-sectional samples than in panels. The negative relationship between measurement error and earnings is driven largely by overreporting among low earners. Median response errors are not related to earnings."

Lack of evidence for transmission of human immunodeficiency virus through vaginal intercourse

Research in determining risks for human immunodeficiency virus (HIV) transmission is confounded by many issues. They include lack of clarity or specificity in terminology used, respondents misunderstanding of questions, and lying. The base rate of lying (or social desirability responding) by itself is sufficient to account for the small percentage of Americans and Europeans claiming "heterosexual" transmission from partners not known to be intravenous drug users. This study integrates the physiological and epidemiological data on risk factors for HIV transmission with the psychological literature on the frequency of anal intercourse and of lying (in this case to researchers and clinicians about risk factors). When these factors are considered, intravenous and anal activities remain the only clear vectors for HIV transmission. Research suggesting that spermicidals are more effective at inactivating HIV than condoms are at physically containing HIV is also noted.

Epidemiological studies of chronic disease: maladjustment of observed mortality rates

Age adjustment of observed mortality and morbidity rates is not a substitute for age-specific analysis. Measures of association between potential causal factors and adjusted mortality rates are functions of the particular adjustment procedure and the choice of reference population. We exhibit here the wide variation in simple correlation statistics that occurs with eight adjustment methods and three reference populations. We then generalize these results to the multivariate situation showing an example in which there is coherent structure for the associations between predictors and mortality. This is contrasted with another example in which no such meaningful pattern exists. Studies are cited that could have been improved by greater attention to the underlying structure of age-adjusted rates. Age adjustment of total observed rates yields meaningless numbers that are useful for comparative purposes only. Total observed rates have substantive meaning but provide useful etiological clues primarily when supported by analyses of appropriate age-specific data.

Infant feeding in Bolivia: a critique of the World Health Organization indicators applied to demographic and health survey data

Infant feeding is a multidimensional activity that can be described and analysed in many different ways. The World Health Organization (WHO) has recently issued recommended indicators for assessing infant feeding practices. This paper presents these indicators and demonstrates their applications using the 1989 Demographic and Health Survey (DHS) data for Bolivia. The results indicate that, although most Bolivian infants are breastfed and two-thirds are breastfed for > 1 year, supplementary feeding practices deviate considerably from international recommendations. Only 58% of infants < 4 months old are receiving breastmilk alone (the 'exclusive breastfeeding rate') and a similarly low percentage (54.7%) of 6-9 month olds are receiving the recommended combination of breast milk plus solid or semi-solid foods (the 'timely complementary feeding rate'). Furthermore, almost half of breastfed infants < 12 months old are also receiving bottle feeds. The infant feeding practices of city residents are least likely to conform to the infant feeding recommendations, while practices of mothers who have always lived in the country are most likely to be similar to the WHO guidelines. Mothers who have moved to the city since the age of 12 are most likely to be giving their infants other milks in addition to breast milk and to be bottle feeding their infants. The WHO infant feeding indicators provide a useful framework for quantifying infant feeding practices, and most of the indicators can readily be applied to DHS data. Nonetheless, improvements can be made in both the indicators themselves and the DHS questionnaire to improve reporting of internationally comparable infant feeding information.

Immaculate conceptions in sub-saharan Africa: exploratory analysis of inconsistencies in the timing of first sexual intercourse and first birth

Survey data from a number of developing countries show that a considerable proportion of parous women report that they had their first birth one or more years before they first had sexual intercourse. In this paper, I use data from eight African Demographic and Health Surveys to explore factors that contribute to cross-national differentials in the prevalence of these "immaculate conceptions." The results suggest that this data problem results not only from recall errors, but also from the fact that some respondents misinterpret the question on first sexual intercourse and report their age at the onset of intercourse with either their first or current husband rather than their age at sexual initiation.

Methodological issues in collecting data from traditional birth attendants

Information on the activities, practices and social context of pregnancy and delivery care provided by traditional birth attendants (TBA) is a critical requirement in planning, monitoring and evaluating maternal health programs in many countries. As a result of experimental studies in which such information was obtained by a variety of methods, and a review of alternative methodologies, a set of guidelines has been developed for the collection of such information. High-lighted are the need for good background knowledge on the local situation, involving TBAs themselves in design and collecting methods, a system of supervision to ensure adequate training and careful monitoring, and finally sharing the findings with the TBAs as well as with health officials.

On an average: the rural hospital in sub-Saharan Africa

The general impression of a heavy workload for rural hospitals in sub-Saharan Africa is illustrated by a quantification for 16 parameters, based on data in 120 annual reports from 40 hospitals. The figures presented in this study can be looked upon as “figures of reference” which hospitals can use for reflection on their own data.

Comparison with other studies is made possible by presenting the figures after conversion to an assumed population of 100 000. Marginal notes are made on the comparability of the data as presented by the hospitals.

Statistical methods to adjust for date and age misreporting to improve estimates of vital rates in Pakistan

Misreporting of dates and ages poses serious difficulties for the estimation of the age distribution and birth and death rates in many developing countries. The pervasiveness of these problems is illustrated with data from a well-designed on-going survey in Pakistan, the Pakistan Demographic Survey. Methods for reconciling discrepancies, based on the assumptions of constant misreporting and survivorship patterns, are presented. The reasoning behind these methods could be applied much more generally. Research into the cultural interpretations of age and dates, and the nature of possible biases, is called for.

Abridged life tables for Pakistan based on the 1971 Population Growth Survey

A new set of abridged life tables for Pakistan is presented. Data from the 1971 Population Growth Survey were first analysed to estimate the degree of completeness of the reporting of male and female deaths; female deaths were substantially more under-reported than male deaths. Age–sex specific mortality schedules were adjusted accordingly.

Life expectancy was around 50 years at birth, but increased by 8–9 years for those surviving the substantial risks of dying in the first year of life. No significant sex differential in mortality could be discerned from the life tables.