Temporal Associations Between Depressive Symptoms and Self-reported Sexually Transmitted Disease Among Adolescents FREE

SEVERAL STUDIES in adolescents and adults have reported associations between mental health problems and sexually transmitted diseases (STDs).^{1 - 9} Compared with nondepressed youth, depressed children and adolescents have poor social relationships¹⁰ and may be more emotionally reactive in peer relationships,¹¹ factors that may contribute to increased STD risk behavior. Individuals with severe mental illness may engage in high-risk sexual behaviors as a result of poor impulse control, impaired judgment, and lack of knowledge or understanding of the risks.¹² In addition, affective disorders are associated with substance use,^{13 - 14} which is related to increased sexual risk behavior and STDs.^{15 - 18}

Depressed individuals may also use sexual risk behaviors as part of a coping response to their depressive symptoms. Research has suggested that men, in particular, tend to purposively turn their attention away from depressive symptoms and onto neutral or pleasant activities,^{19 - 20} such as sexual intercourse. Less is known about the coping responses of adolescents but some research suggests that by early to middle adolescence, boys exhibit distractive coping styles that are similar to those of adult men.^{21 - 22}

Most previous studies on mental health problems and STDs have focused on specialized populations, such as psychiatric patients,^{2 ,4} homeless and runaway youth,^{1 ,23} and patients in STD clinics.⁸ In addition, most research to date has been cross-sectional^{8 - 9 ,24} and therefore has not addressed the temporal nature of the relationship between mental health problems and acquisition of STDs. There are psychological sequelae of receiving an STD diagnosis,^{25 - 29} so the possibility that mental health problems follow rather than precede an STD diagnosis must be considered. Previous research has demonstrated that physical illness is a predictor of depression; in one prospective study of adolescents, physical disease and functional impairment at baseline were identified as risk factors for major depressive disorder 1 year later.³⁰

These findings are consistent with the integrative theory of depression,³¹ which draws on empirical findings and contemporary causal theories of depression in proposing that environmental and personal mechanisms interact to produce an episode of depression. Specifically, the theory postulates that eliciting events lead to a heightened state of self-focused attention that, in turn, results in a variety of affective, behavioral, and cognitive transformations that exacerbate dysphoria and induce and maintain a depressive episode. The diagnosis of an STD may represent an antecedent event that initiates the depression-producing process. Alternatively, an STD may be the result of the behavior changes that follow the evoking event and increased self-focused attention. It will be important to elucidate the relationship between depression and STD to design effective prevention programs for both of these pressing public health problems.

The objective of this study was to explore the reciprocal associations between depressive symptoms and diagnosis of STD using data from the National Longitudinal Study of Adolescent Health (Add Health),³² a longitudinal study of a national sample of adolescents. We hypothesized that depressive symptoms would be associated with self-reported STD diagnosis in the following year and that, among those without clinically significant depressive symptoms at baseline, reporting an STD diagnosis would be associated with clinically significant depressive symptoms at the end of the year. We anticipated that sex differences would exist in these associations. In addition, based on our previous analyses,⁹ we hypothesized that substance use would mediate the association between depressive symptoms and later STD diagnosis in boys but not girls.

Details of the design and collection of data in Add Health are described elsewhere.^{32 - 33} Students in the 7th through 12th grades from a systematic, random sample of 80 high schools and 52 associated feeder schools across the United States completed baseline in-home interviews (Wave 1). A total of 20 745 students completed the baseline 90-minute in-home interview between April and December 1995. Approximately 1 year later, 13 568 adolescents completed a second in-home interview (Wave 2). For the analytic sample, the mean (SD) interval between Wave 1 and Wave 2 interviews was 11.1 (1.7) months. Adolescents who were in the 12th grade at baseline or who were part of the subsample with disabilities were not interviewed in Wave 2. Each case was given a weight based on the sampling design such that the sample would be representative of all US 7th through 12th grade students.³⁴ The analyses herein include data from adolescents who participated in both Wave 1 and Wave 2 interviews, who reported consistently on separate questions that they had sexual intercourse by Wave 2 and that they had sexual intercourse at least once during the interval between the 2 interviews, and for whom sample weights were available³⁵ (N = 4738). The study protocol was approved by the University of North Carolina (Chapel Hill) institutional review board on research involving human subjects.

At baseline and follow-up, the frequency of depressive symptoms in the past week was assessed using a modified version of the 20-item Center for Epidemiologic Studies Depression Scale (CES-D).³⁶ The CES-D has been used in several studies of adolescents^{37 - 38} and has demonstrated good reliability and validity in adolescent and adult samples.^{39 - 41} Internal consistency reliability is generally greater than .80^{36 ,39 - 41} and higher CES-D scores are correlated with diagnostic interview–validated depression.^{36 ,39 - 40} The depressive symptoms scale in this study (19 items; range of possible scores, 0-54) had high internal consistency reliability (Cronbach α coefficient⁴² = .87 at both baseline and follow-up). As previously reported,⁹ factor analysis of these items demonstrated the 4-factor structure that has been described for the CES-D.^{36 ,41} Scores predictive of major depressive disorder in adolescents have been identified (22 for males, 24 for females).³⁹ These scores were adapted proportionally to the modified scale (21 for males, 23 for females) and used to identify a "very high" level of depressive symptoms. The remaining scores were divided into equal groups with "low," "moderate," and "high" levels of depressive symptoms to create a 4-category variable.⁹ If at least 80% of the scale items had been completed, missing data were filled in using imputation with sample means.

At Wave 1 ("ever") and Wave 2 (for the interval between the 2 interviews), adolescents were asked individual questions about whether they had been told by a doctor or a nurse that they had each of the following STDs: chlamydia, gonorrhea, syphilis, HIV (human immunodeficiency virus) or AIDS (acquired immunodeficiency syndrome), genital herpes, genital warts, and trichomoniasis. Because of the low numbers of boys and girls reporting each STD, it was not possible to conduct analyses using the individual STD variables, so variables were constructed for self-reported history of 1 or more STDs at Wave 1 and for diagnosis of 1 or more STDs between Waves 1 and 2.

Alcohol use was measured at Wave 2 by the number of drinking days in the previous year, categorized as none, minimal (1-2 days in the past year to once per month or less), moderately frequent (2-3 days per month to 1-2 days per week), and very frequent (3-5 days per week to every day or almost every day). Marijuana use was determined at Wave 2 by the response to a question on whether the adolescent had tried or used marijuana since Wave 1.

Age in years at Wave 2 was determined from the month and year of birth and the month and year of the study interview and was categorized to correspond with the developmental stages of early, middle, and late adolescence (12-14, 15-17, and 18-21 years).⁴³ Gender was evaluated by the interviewer and confirmed by asking the adolescent if it was not apparent. Self-reported race or ethnicity was categorized as non-Hispanic white, non-Hispanic black, Hispanic, and other (included small numbers of adolescents reporting American Indian or Native American, Asian or Pacific Islander, and other race). To account for the effect of recent coital debut on acquiring an STD, adolescents who had never had sexual intercourse at baseline but became sexually active during the interval between Wave 1 and Wave 2 were distinguished from those who reported having been sexually active at baseline.

Data management and descriptive analyses were performed using SPSS for Windows version 10.0 (SPSS Inc, Chicago, Ill). Appropriate bivariate tests and multivariate logistic regression models using weighted data were analyzed with SUDAAN version 7.5.4 (Research Triangle Institute, Research Triangle Park, NC) to adjust for the design effects associated with the complex clustered sampling design used in Add Health.⁴⁴ The first set of logistic regression models (models A1-A3) evaluated whether baseline depressive symptoms independently predicted diagnosis of an STD within the following year. To account for the effect of a previous diagnosis of STD on the odds of future STDs, all models controlled for ever being diagnosed as having an STD at baseline. In model A1, the variable measuring baseline depressive symptoms was entered into the model. Model A2 adjusted for the demographic variables (age, race/ethnicity) and baseline sexual activity. To determine whether substance use mediated the association between depressive symptoms and STD diagnosis, the substance use variables significantly associated with both depressive symptoms and STD diagnosis on bivariate analysis were added in model A3. A variable functions as a mediator to the degree to which it accounts for the association between an independent variable (predictor) and a dependent variable (outcome).⁴⁵ According to Kraemer et al,⁴⁶ because the definition of a mediator suggests that it explains how and why the independent variable works to produce the outcome, a mediator should occur after the independent variable. Additionally, a mediator should be significantly associated with both the independent variable and the dependent variable and, when controlled for, cause a previously significant association between the independent and dependent variables to no longer be significant.⁴⁵

The second set of analyses explored the possibility that the diagnosis of STD precedes rather than follows the development of depression. To identify the risk of the most clinically relevant level of depressive symptoms, we used the "very high" category of depressive symptoms as the outcome. To determine a change in depressive symptoms to the very high level, these analyses excluded adolescents who reported very high levels of depressive symptoms at baseline (n = 624, [212 boys and 412 girls]). Hierarchical models examined whether ever being diagnosed as having an STD between Wave 1 and Wave 2 was associated with a very high level of depressive symptoms at Wave 2, first controlling for depressive symptoms at baseline alone (model B1) and then adding demographic characteristics and virginity (model B2). Significance was defined at P<.05. Because of small cell sizes, we could not determine interaction terms with reliable estimates. All analyses were conducted separately for boys and girls.

Characteristics of the sample are shown in Table 1. The sample was almost equally male (n = 2232) and female (n = 2506) and predominantly non-Hispanic white and black. Mean (SD) age was 16.8 (1.4) years for males (range, 13-23 years) and 17.0 (1.4) years for females (range, 13-21 years). Almost half of boys (45.5%) and girls (48.2%) were aged between 15 and 17 years.

Most adolescents (71.4% of boys, 69.0% of girls) who had sexual intercourse in the year between Wave 1 and Wave 2 reported having already initiated sexual activity at baseline. Girls were almost twice as likely as boys to report a history of STD at baseline (5.4% vs 2.8%, P<.001) and to report being diagnosed as having an STD between Wave 1 and Wave 2 (6.1% vs 3.2%, P<.001). Eighty percent of the interval STDs were reported by adolescents who did not have a history of STD at baseline.

The median depressive symptoms score was slightly higher for girls than for boys at both Wave 1 (13 vs 10; Mann-Whitney U test, P<.001) and Wave 2 (12 vs 10; Mann-Whitney U test, P<.001). At both times, approximately 1 of every 10 boys and almost 1 of every 6 girls reported very high levels of depressive symptoms. More than 42% of adolescents with very high levels of depressive symptoms at Wave 2 had reported very high levels at Wave 1.

Baseline substance use was prevalent. Almost two thirds of boys and girls reported alcohol use in the year before the Wave 2 interview; more boys (8.7%) than girls (5.2%) reported very frequent alcohol use (P<.001). Almost half of both boys and girls had used marijuana at least once between Wave 1 and Wave 2.

For boys and girls, level of depressive symptoms at Wave 1 was associated with diagnosis of STD between Wave 1 and Wave 2 in unadjusted analyses. Compared with boys with low and moderate levels of depressive symptoms, boys with high and very high levels of depressive symptoms were more likely to have been diagnosed as having an STD between Waves 1 and 2 (1.3% low and 1.7% moderate vs 5.4% high and 6.5% very high; P<.005). Similarly, girls with high and very high levels of depressive symptoms were more likely than those with low and moderate levels of depressive symptoms to report an interval STD diagnosis (7.1% high and 9.4% very high vs 4.9% low and 4.6% moderate; P<.05).

Among boys, after adjusting for a history of STD at baseline, both high and very high levels of depressive symptoms remained independently associated with being diagnosed as having an STD within the following year (Table 2, model A1). This association was robust to the addition of the demographic and sexual debut variables (model A2). Compared with boys with a low level of depressive symptoms, those with high and very high levels at baseline had 3.92 times and 3.98 times, respectively, the odds of being diagnosed as having an STD by the 1-year follow-up interview.

For boys, very frequent alcohol use was associated with baseline level of depressive symptoms (P<.005) and diagnosis of interval STD (P<.05) on bivariate analysis. However, in the adjusted model (Table 2, model A3), the addition of the alcohol use variable did not substantially change the parameter estimates for the level of depressive symptoms, which remained significantly associated with STD diagnosis. Very frequent alcohol use was independently associated with STD diagnosis (OR, 3.11; 95% CI, 1.02-9.52), so the alcohol use variable was retained in model A3. Marijuana use was not significantly associated with the dependent variable, STD diagnosis, so further analyses assessing marijuana use as a mediator were not performed. For girls, Wave 1 level of depressive symptoms did not predict STD diagnosis by Wave 2 in either of the adjusted models (Table 2, models A1 and A2).

Among boys and girls without very high levels of depressive symptoms at baseline, diagnosis of STD between Wave 1 and Wave 2 was associated with an increase in depressive symptoms to very high levels at Wave 2. For boys, 31% of those reporting an interval diagnosis of STD had very high levels of depressive symptoms at Wave 2, compared with only 9% of boys without a diagnosis of STD between Waves 1 and 2 (P<.01). Compared with those without an interval STD diagnosis, twice as many girls with STD diagnoses reported very high levels of depressive symptoms at Wave 2 (28% vs 14%, P<.05). These associations remained significant in the adjusted logistic regression models for both boys and girls. Self-reported diagnosis of interval STD was associated with 3.3 (boys) and 2.4 (girls) times the odds of very high levels of depressive symptoms at Wave 2 (Table 3, model B2).

Despite efforts to reduce the prevalence of STDs, adolescents and young adults continue to demonstrate high rates of chlamydia, gonorrhea, and syphilis,⁴⁷ and are experiencing greater increases in the seroprevalence of herpes simplex virus 2 than adults.⁴⁸ Approximately 1 in 4 new STD diagnoses is in a teenager.⁴⁹ Dysthymia, major depressive disorder, and other psychiatric disorders also affect young people at alarming rates.^{37 ,50 - 53} In one community sample of older adolescents, nearly 30% reported at least 1 current symptom of a major depressive episode.⁵⁴

This longitudinal study suggests that the temporal relationship between depressive symptoms and STD is complex and gender specific. For boys, the association is bidirectional; boys with depressive symptoms were at increased risk for being diagnosed as having an STD by the 1-year follow-up and boys diagnosed as having an STD by the 1-year follow-up were at increased risk for reporting very high levels of depressive symptoms at the end of that year. Among girls, recent STD diagnosis was a risk factor for future depression, but depressive symptoms were not associated with being diagnosed as having an STD once models controlled for baseline history of STD.

Sex differences in the expression of psychological distress exist among adolescents.^{55 - 58} While girls with depressive disorders tend to exhibit classic symptoms of social withdrawal, diminished interest in pleasurable activities, and low libido,⁵⁹ boys with psychiatric symptoms are more prone to engage in acting-out behaviors,⁵⁵ which may include sexual behaviors that increase the risk of acquiring an STD. One explanation for this sex difference is that social sanctions against men appearing depressed may encourage boys to demonstrate nondepressive behaviors. Social pressures for boys and men not to display their emotions may also result in differential societal responses to depressive symptoms in men vs women.²⁰ In a study of roommates of college freshmen, roommates were more likely to rate a depressed man negatively than a depressed women.⁶⁰ In addition, several studies in adolescents^{21 - 22} and adults^{19 - 20} with depressive symptoms suggest that while females tend to use ruminative coping, males are more likely to distract themselves by engaging in pleasurable activities. Further research is needed to determine if sexual behavior is part of a distractive coping response to depressive symptoms in boys.

Previous cross-sectional analyses⁹ suggest that use of alcohol and marijuana may mediate the association between depressive symptoms and STD diagnosis among boys. However, although alcohol use every day or almost every day was an independent predictor of STD diagnosis in our study, neither alcohol nor marijuana use functioned as a mediator between baseline depressive symptoms and later STD diagnosis.

Cross-sectional studies have observed increased depressive symptoms in young women who report a history of STD^{9 ,24} and in adult women who are STD clinic patients.⁸ In our analyses of adolescent girls, while recent diagnosis of STD was a risk factor for depression, depressive symptoms did not predict future STD diagnosis once history of STD was entered into the models. To explore this finding further, we examined the interaction between baseline depressive symptoms and history of STD. The cell sizes were small and the interaction term did not achieve significance. However, there was a suggestion of effect modification; girls with a history of STD who had higher depressive symptoms at baseline were at greater risk of contracting an STD between Wave 1 and Wave 2 than might be expected from the individual independent effects of a history of STD and increased depressive symptoms. It is possible that for girls with increased depressive symptoms at Wave 1, the risk behaviors associated with STD preceded the Wave 1 interview. Alternatively, similar to this study, previous reports have noted high rates of both depressive symptoms^{3 ,8} and STD diagnosis⁸ in women. As a result, it may be difficult to detect increases in the prevalence of either condition among girls. Finally, while high rates of depressive symptoms may co-occur or follow the diagnosis of an STD, depression may not predict STD among girls in a general, nonclinical population. Factors such as poverty or other life stressors not included in these analyses may be important in determining whether high depressive symptoms are an important predictor of STD in certain young female subpopulations.

We found that among boys and girls who did not have very high levels of depressive symptoms at baseline, diagnosis of an STD by the 1-year follow-up was associated with very high levels of depressive symptoms at the end of that year. Our analyses used a score cutoff that has been correlated with a psychiatric diagnosis of depression on previous research using a similar depressive symptoms scale.³⁹ By restricting these analyses to adolescents who did not report a very high level of depressive symptoms at Wave 1 and controlling for level of depressive symptoms at Wave 1, we attempted to examine the association between STD diagnosis and very high depressive symptoms occurring later. However, because the STD diagnosis could have occurred at any time between Wave 1 and Wave 2, it is possible that the diagnosis could have been made after the development of very high depressive symptoms. Nonetheless, these findings support literature suggesting that there are psychological sequelae to being diagnosed as having an STD.^{25 - 29} Diagnosis of an STD may be the antecedent event that initiates the process leading to the depressive episode.³¹ In support of this contention, one cross-sectional study comparing the psychological morbidity among individuals infected vs those not infected with genital herpes found that nearly 1 in 4 participants with herpes reported emotional trauma attributable to the infection.²⁹ In our study, the small numbers of students reporting each STD precluded analysis by specific STD or by bacterial vs viral STD. It is likely that psychological sequelae related to the diagnosis of an incurable viral STD differ from those related to a curable bacterial STD. Psychological sequelae may also differ between the sexes depending on the physical consequences associated with a particular STD. For example, because girls with genital chlamydial infection may be informed of the possibility of subsequent infertility or ectopic pregnancy, they may be more likely than boys to experience psychological morbidity associated with being diagnosed as having this STD. However, in this study, we found that boys and girls who reported being diagnosed as having an STD were at increased risk of future depression. It is possible that the diagnosis of STD may not alone cause later clinical depression but that the STD may be an indicator of psychosocial problems that can lead to depression. It is important to use the time of STD diagnosis and follow-up as an opportunity for screening for depressive symptoms.

This study has several limitations. Although the depressive symptoms scale used in Add Health was derived from the CES-D, which has been shown to be valid and reliable in adolescent samples,^{39 - 41} the psychometric properties of the modified scale have not been established. However, most Add Health scale items are identical to the items in the CES-D and the results of our internal consistency reliability and factor analyses suggest that the modified scale performs similarly to the CES-D. Depressive symptoms were assessed for a 1-week period before each interview date; data on chronic or recurrent depressive symptoms between Wave 1 and Wave 2 were not available. We would expect to find only a stronger association between depressive symptoms and STD diagnosis among students for whom symptoms occur more frequently or last longer. Adolescents who are not in school, such as those who are homeless or runaways or who have been expelled, are at increased risk for both mental health problems and STDs.^{61 - 62} Because the data collected in Add Health are self-reported, adolescents may underreport or overreport depressive symptoms and STD history.⁶³ Studies have shown that the CES-D is well correlated with clinical depression diagnosed by interview.^{39 - 41} Adolescents may be able to report diagnosis of STD fairly accurately,⁶⁴ although a recent prospective study found that only 69% of adolescent girls with biologically confirmed STD accurately reported their positive STD status.⁶⁵ The prevalence of very high depressive symptoms in this study (13.0% of boys and girls combined) is comparable with the 12.1% of high school students identified as severely depressed on the CES-D in a previous study.³⁹ The proportion of sexually active adolescents reporting a history of STD diagnosis was approximately one half the rates of infection with gonorrhea and chlamydia noted in a study screening high school students.⁶⁶ Most students with infection in that study were asymptomatic and had not received STD screening in the previous year; among those with STD symptoms, not all sought care. The questions on STD in Add Health referred to being told by a doctor or nurse of the diagnosis, so the true prevalence of sexually transmitted infections in this sample is likely to have been underestimated. This may be especially true for boys, who are less likely than girls to receive preventive health care, including screening for STDs.^{65 ,67} Boys in this study were half as likely to report either a history of STD at Wave 1 or diagnosis of STD between Wave 1 and Wave 2. While it is possible that boys underreport having STDs compared with girls, a study of high school–based screening has reported similar findings with actual, as opposed to self-reported STD.⁶⁶ If boys were less likely than girls to report having STDs, our findings would be affected if the tendency to underreport differed by level of depressive symptoms. Systematic underreporting of STDs may result in misclassification bias and increased type II error.⁶⁵ Despite this possibility, we still found significant associations between depressive symptoms and self-reported STD. In addition, at the time that the Add Health Wave 1 and Wave 2 data were collected, the relatively insensitive methods used in STD testing (nonnucleic acid amplification) and limited screening in adolescents would result in further underestimation of STD in this study. The use of biological measures of STD incidence in future studies will be important and may strengthen the findings in this study. Finally, the date of the STD diagnosis was not elicited, so we were unable to determine the interval of time between this diagnosis and the subsequent report of very high levels of depressive symptoms.

To develop appropriate, effective STD prevention interventions, it will be important to further elucidate the sex differences in the association between depressive symptoms and future STD diagnosis and to identify the mechanisms and mediational pathways of the reciprocal association. Prospective research will also be important to develop risk indicators and to prevent the development of clinically significant depressive symptoms in adolescents who have been diagnosed as having STDs. Clinicians, including mental health professionals, need to recognize the temporal associations between depressive symptoms and STDs in adolescents. Although girls are more likely than boys to report depressive symptoms after early adolescence,¹⁰ our findings suggest that clinicians need to be particularly alert to boys with depressive symptoms, who are at increased risk for being diagnosed as having an STD. Clinical encounters for health maintenance, STD screening and treatment, contraception, and mental health treatment all provide opportunities for screening for both depressive symptoms and STDs.

Accepted for publication February 28, 2002.

Dr Shrier was supported in part by the Charles A. Janeway Award in Child Health Research (HD27805-08), Child Health Research Center, Children's Hospital, Boston, and by grant 5K23MH01845 from the National Institute of Mental Health, National Institutes of Health, Bethesda, Md.

This research is based on data from the Add Health project, a program project designed by J. Richard Udry (principal incestigator) and Peter Bearman, and funded by grant P01-HD31921 from the National Institute of Child Health and Human Development to the Carolina Population Center, University of North Carolina at Chapel Hill, with cooperative funding participation by the National Cancer Institute; the National Institute of Alcohol Abuse and Alcoholism; the National Institute on Deafness and Other Communication Disorders; the National Institute on Drug Abuse; the National Institute of General Medical Sciences; the National Institute of Mental Health; the National Institute of Nursing Research; the Office of AIDS Research, NIH; the Office of Behavior and Social Science Research, NIH; the Office of Director, NIH; the Office of Research on Women's Health, NIH; the Office of Population Affairs, DHHS (Department of Health and Human Services); the National Center for Health Statistics; Centers for Disease Control and Prevention, DHHS; the Office of Minority Health, Centers for Disease Control and Prevention, DHHS; the Office of Minority Health, Office of Public Health and Science, DHHS; the Office of the Assistant Secretary for Planning and Evaluation, DHHS; and the National Science Foundation. Persons interested in obtaining data files from The National Longitudinal Study of Adolescent Health should contact Add Health Project, Carolina Population Center, 123 W Franklin St, Chapel Hill, NC 27516-3997 (e-mail: addhealth@unc.edu).

While previous cross-sectional research has linked STDs and mental health problems, the temporal nature of the association between these 2 prevalent conditions has not been fully explored. This longitudinal study suggests a bidirectional association for boys; boys with depressive symptoms were at increased risk for reporting an STD diagnosis at 1-year follow-up and boys who self-reported a diagnosis of STD at 1-year follow-up were at increased risk for reporting very high levels of depressive symptoms at the end of that year. Among girls, self-report of a recent STD diagnosis was a risk factor for very high levels of depressive symptoms, but depressive symptoms did not predict reporting an STD diagnosis once models controlled for baseline history of STD. Medical and mental health clinicians need to screen sexually active adolescents for depressive symptoms and for STDs.

Corresponding author and reprints: Lydia A. Shrier, MD, MPH, Division of Adolescent/Young Adult Medicine, Children's Hospital, 300 Longwood Ave, Boston, MA 02115 (e-mail: lydia.shrier@tch.harvard.edu).