John B. Carlin

Multiple imputation for missing data in epidemiological and clinical research: potential and pitfalls

Most studies have some missing data. Jonathan Sterne and colleagues describe the appropriate use and reporting of the multiple imputation approach to dealing with them

Does bullying cause emotional problems? A prospective study of young teenagers

Abstract Objectives: To establish the relation between recurrent peer victimisation and onset of self reported symptoms of anxiety or depression in the early teen years. Design: Cohort study over two years. Setting: Secondary schools in Victoria, Australia. Participants: 2680 students surveyed twice in year 8 (aged 13 years) and once in year 9. Main outcome measures: Self reported symptoms of anxiety or depression were assessed by using the computerised version of the revised clinical interview schedule. Incident cases were students scoring ≥12 in year 9 but not previously. Prior victimisation was defined as having been bullied at either or both survey times in year 8. Results: Prevalence of victimisation at the second survey point in year 8 was 51% (95% confidence interval 49% to 54%), and prevalence of self reported symptoms of anxiety or depression was 18% (16% to 20%). The incidence of self reported symptoms of anxiety or depression in year 9 (7%) was significantly associated with victimisation reported either once (odds ratio 1.94, 1.1 to 3.3) or twice (2.30, 1.2 to 4.3) in year 8. After adjustment for availability of social relations and for sociodemographic factors, recurrent victimisation remained predictive of self reported symptoms of anxiety or depression for girls (2.60, 1.2 to 5.5) but not for boys (1.36, 0.6 to 3.0). Newly reported victimisation in year 9 was not significantly associated with prior self report of symptoms of anxiety or depression (1.48, 0.4 to 6.0). Conclusion: A history of victimisation and poor social relationships predicts the onset of emotional problems in adolescents. Previous recurrent emotional problems are not significantly related to future victimisation. These findings have implications for how seriously the occurrence of victimisation is treated and for the focus of interventions aimed at addressing mental health issues in adolescents. What is already known on this topic Being bullied is a common experience for many young people Victimisation is related to depression and, to a lesser extent, anxiety, loneliness, and general self esteem Debate remains as to whether victimisation precedes the onset of emotional problems or whether young people with emotional problems “invite” victimisation What this study adds A history of victimisation predicts the onset of anxiety or depression, especially in adolescent girls Previous recurrent emotional problems are not significantly related to future victimisation Reduction in bullying in schools could have a substantial impact on the emotional wellbeing of young people

Cannabis use and mental health in young people: cohort study

Abstract Objective: To determine whether cannabis use in adolescence predisposes to higher rates of depression and anxiety in young adulthood. Design: Seven wave cohort study over six years. Setting: 44 schools in the Australian state of Victoria. Participants: A statewide secondary school sample of 1601 students aged 14-15 followed for seven years. Main outcome measure: Interview measure of depression and anxiety (revised clinical interview schedule) at wave 7. Results: Some 60% of participants had used cannabis by the age of 20; 7% were daily users at that point. Daily use in young women was associated with an over fivefold increase in the odds of reporting a state of depression and anxiety after adjustment for intercurrent use of other substances (odds ratio 5.6, 95% confidence interval 2.6 to 12). Weekly or more frequent cannabis use in teenagers predicted an approximately twofold increase in risk for later depression and anxiety (1.9, 1.1 to 3.3) after adjustment for potential baseline confounders. In contrast, depression and anxiety in teenagers predicted neither later weekly nor daily cannabis use. Conclusions: Frequent cannabis use in teenage girls predicts later depression and anxiety, with daily users carrying the highest risk. Given recent increasing levels of cannabis use, measures to reduce frequent and heavy recreational use seem warranted. What is already known on this topic Frequent recreational use of cannabis has been linked to high rates of depression and anxiety in cross sectional surveys and studies of long term users Why cannabis users have higher rates of depression and anxiety is uncertain Previous longitudinal studies of cannabis use in youth have not analysed associations with frequent cannabis use What this study adds A strong association between daily use of cannabis and depression and anxiety in young women persists after adjustment for intercurrent use of other substances Frequent cannabis use in teenage girls predicts later higher rates of depression and anxiety Depression and anxiety in teenagers do not predict later cannabis use; self medication is therefore unlikely to be the reason for the association

Onset of adolescent eating disorders: population based cohort study over 3 years

Abstract Objective: To study the predictors of new eating disorders in an adolescent cohort. Design: Cohort study over 3 years with six waves. Subjects: Students, initially aged 14-15 years, from 44 secondary schools in the state of Victoria, Australia. Outcome measures: Weight (kg), height (cm), dieting (adolescent dieting scale), psychiatric morbidity (revised clinical interview schedule), and eating disorder (branched eating disorders test). Eating disorder (partial syndrome) was defined when a subject met two criteria for either anorexia nervosa or bulimia nervosa according to the Diagnostic and Statistical Manual of Mental Disorders, fourth edition (DSM-IV). Results: At the start of the study, 3.3% (29/888) of female subjects and 0.3% (2/811) of male subjects had partial syndromes of eating disorders. The rate of development of new eating disorder per 1000 person years of observation was 21.8 in female subjects and 6.0 in male subjects. Female subjects who dieted at a severe level were 18 times more likely to develop an eating disorder than those who did not diet, and female subjects who dieted at a moderate level were five times more likely to develop an eating disorder than those who did not diet. Psychiatric morbidity predicted the onset of eating disorder independently of dieting status so that those subjects in the highest morbidity category had an almost sevenfold increased risk of developing an eating disorder. After adjustment for earlier dieting and psychiatric morbidity, body mass index, extent of exercise, and sex were not predictive of new eating disorders. Conclusions: Dieting is the most important predictor of new eating disorders. Differences in the incidence of eating disorders between sexes were largely accounted for by the high rates of earlier dieting and psychiatric morbidity in the female subjects. In adolescents, controlling weight by exercise rather than diet restriction seems to carry less risk of development of eating disorders. Key messages Adolescent females who diet at a severe level are 18 times more likely to develop an eating disorder than those who do not diet, and those who diet at a moderate level are five times more likely to develop an eating disorder High levels of psychiatric morbidity in females increase the risk of developing eating disorders by sevenfold Around two thirds of new cases of eating disorder arise in females who have dieted moderately The predominance of eating disorders in females is largely explained by the higher rates of earlier dieting and psychiatric morbidity Daily exercise seems to be a less risky strategy for controlling weight in adolescents

Depression, anxiety, and smoking initiation: a prospective study over 3 years.

OBJECTIVES This report considers the extent to which depression and anxiety predict smoking onset in adolescence. METHODS A 6-wave cohort design was used to study a sample of 14- and 15-year-old students (n = 2032) drawn from 44 secondary schools in the state of Victoria, Australia. The students were surveyed between 1992 and 1995 with a computerized questionnaire that included a 7-day retrospective diary for tobacco use and a structured psychiatric interview. RESULTS Experimental smokers were 29 times more likely than non-smokers to make a transition into daily use in the subsequent 6 months. Depression and anxiety, along with peer smoking, predicted initiation of experimental smoking. Specifically, depression and anxiety accentuated risks associated with peer smoking and predicted experimentation only in the presence of peer smoking. CONCLUSIONS The finding that experimental smoking is an overwhelmingly strong predictor of later daily smoking focuses attention on smoking initiation. Depressive and anxiety symptoms are associated with higher risks for initiation through an increased susceptibility to peer smoking influences. Promoting the psychological well-being of adolescents and addressing perceived interpersonal benefits of smoking may play a role in the prevention of adolescent tobacco use.

Cannabis use and mental health in younger people: Cohort study.

Abstract Objective: To determine whether cannabis use in adolescence predisposes to higher rates of depression and anxiety in young adulthood. Design: Seven wave cohort study over six years. Setting: 44 schools in the Australian state of Victoria. Participants: A statewide secondary school sample of 1601 students aged 14-15 followed for seven years. Main outcome measure: Interview measure of depression and anxiety (revised clinical interview schedule) at wave 7. Results: Some 60% of participants had used cannabis by the age of 20; 7% were daily users at that point. Daily use in young women was associated with an over fivefold increase in the odds of reporting a state of depression and anxiety after adjustment for intercurrent use of other substances (odds ratio 5.6, 95% confidence interval 2.6 to 12). Weekly or more frequent cannabis use in teenagers predicted an approximately twofold increase in risk for later depression and anxiety (1.9, 1.1 to 3.3) after adjustment for potential baseline confounders. In contrast, depression and anxiety in teenagers predicted neither later weekly nor daily cannabis use. Conclusions: Frequent cannabis use in teenage girls predicts later depression and anxiety, with daily users carrying the highest risk. Given recent increasing levels of cannabis use, measures to reduce frequent and heavy recreational use seem warranted. What is already known on this topic Frequent recreational use of cannabis has been linked to high rates of depression and anxiety in cross sectional surveys and studies of long term users Why cannabis users have higher rates of depression and anxiety is uncertain Previous longitudinal studies of cannabis use in youth have not analysed associations with frequent cannabis use What this study adds A strong association between daily use of cannabis and depression and anxiety in young women persists after adjustment for intercurrent use of other substances Frequent cannabis use in teenage girls predicts later higher rates of depression and anxiety Depression and anxiety in teenagers do not predict later cannabis use; self medication is therefore unlikely to be the reason for the association

Standardized interpretation of paediatric chest radiographs for the diagnosis of pneumonia in epidemiological studies

BACKGROUND Although radiological pneumonia is used as an outcome measure in epidemiological studies, there is considerable variability in the interpretation of chest radiographs. A standardized method for identifying radiological pneumonia would facilitate comparison of the results of vaccine trials and epidemiological studies of pneumonia. METHODS A WHO working group developed definitions for radiological pneumonia. Inter-observer variability in categorizing a set of 222 chest radiographic images was measured by comparing the readings made by 20 radiologists and clinicians with a reference reading. Intra-observer variability was measured by comparing the initial readings of a randomly chosen subset of 100 radiographs with repeat readings made 8-30 days later. FINDINGS Of the 222 images, 208 were considered interpretable. The reference reading categorized 43% of these images as showing alveolar consolidation or pleural effusion (primary end-point pneumonia); the proportion thus categorized by each of the 20 readers ranged from 8% to 61%. Using the reference reading as the gold standard, 14 of the 20 readers had sensitivity and specificity of > 0.70 in identifying primary end-point pneumonia; 13 out of 20 readers had a kappa index of > 0.6 compared with the reference reading. For the 92 radiographs deemed to be interpretable among the 100 images used for intra-observer variability, 19 out of 20 readers had a kappa index of > 0.6. CONCLUSION Using standardized definitions and training, it is possible to achieve agreement in identifying radiological pneumonia, thus facilitating the comparison of results of epidemiological studies that use radiological pneumonia as an outcome.

Lower respiratory infection and inflammation in infants with newly diagnosed cystic fibrosis

The nature and timing of lower respiratory infections in infants with cystic fibrosis is largely unknown1 because infants do not produce sputum and throat cultures may not predict lower respiratory pathogens.2 We performed a prospective cross sectional study of an unselected cohort of infants with cystic fibrosis in which bronchoalveolar lavage was used to determine lower respiratory infection and inflammation during the first three months of life. The state of Victoria, Australia (66000 births per year) has a cystic fibrosis screening programme, all patients being managed by one centre. Between February 1992 and September 1994 we recruited 45 (27 boys) of the 52 infants with newly diagnosed disease; 32 were identified by screening, 12 from meconium ileus, and one by failure to thrive, and all cases were confirmed by sweat testing. Sixteen infants had respiratory symptoms, and seven of them were receiving oral antibiotics when bronchoalveolar lavage was performed at a mean age of 2.6 (SD 1.6) months. Nine otherwise healthy infants (five boys) aged …

Is smoking associated with depression and anxiety in teenagers

OBJECTIVES An association of smoking with depression and anxiety has been documented in adult smokers. This study examines this association in a representative group of teenage smokers. METHODS A two-stage cluster sample of secondary school students in Victoria, Australia, were surveyed by using a computerized questionnaire, which included a 7-day retrospective diary for tobacco use and a structured psychiatric interview. RESULTS Subjects reporting high levels of depression and anxiety were twice as likely to be smokers after the potential confounders of year level, sex, alcohol use, and parental smoking were controlled for. Regular smokers were almost twice as likely as occasional smokers to report high levels of depression and anxiety. In a stratified analysis, an association between regular smoking and psychiatric morbidity was found in girls of all ages but for boys only in the youngest group. CONCLUSIONS The cross-sectional association is consistent with the use of smoking by teenage girls as self-medication for depression and anxiety. Therefore, future health promotional campaigns might consider strategies that attend to perceived psychological benefits of smoking.

Bias and efficiency of multiple imputation compared with complete-case analysis for missing covariate values.

When missing data occur in one or more covariates in a regression model, multiple imputation (MI) is widely advocated as an improvement over complete-case analysis (CC). We use theoretical arguments and simulation studies to compare these methods with MI implemented under a missing at random assumption. When data are missing completely at random, both methods have negligible bias, and MI is more efficient than CC across a wide range of scenarios. For other missing data mechanisms, bias arises in one or both methods. In our simulation setting, CC is biased towards the null when data are missing at random. However, when missingness is independent of the outcome given the covariates, CC has negligible bias and MI is biased away from the null. With more general missing data mechanisms, bias tends to be smaller for MI than for CC. Since MI is not always better than CC for missing covariate problems, the choice of method should take into account what is known about the missing data mechanism in a particular substantive application. Importantly, the choice of method should not be based on comparison of standard errors. We propose new ways to understand empirical differences between MI and CC, which may provide insights into the appropriateness of the assumptions underlying each method, and we propose a new index for assessing the likely gain in precision from MI: the fraction of incomplete cases among the observed values of a covariate (FICO).