Attention-Deficit/Hyperactivity Disorder Symptom Expression: A Comparison of Individual Age at Onset Using Item Response Theory

Attention-Deficit/Hyperactivity Disorder Symptom Expression: A Comparison of Individual Age at Onset Using Item Response Theory

Hugo Peyre, MD, MPH; Nicolas Hoertel, MD, MPH; Samuele Cortese, MD, PhD; Eric Acquaviva, MD, PhD; Pierre De Maricourt, MD; Frédéric Limosin, MD, PhD; and Richard Delorme, MD, PhD

Attention-deficit/hyperactivity disorder (ADHD) is a childhood-onset behavioral disorder that is associated with a wide range of functional impairments1 and has an estimated worldwide prevalence exceeding 5% in school-age children.2 Impairing symptoms of ADHD have been estimated to persist in up to 65% of cases diagnosed in childhood.3

According to the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5),4 an individual must present with symptoms of inattention and/or hyperactivity-impulsivity associated with some impairment before the age of 12 years, compared with 7 years in the previous version of the DSM. Indeed, almost as soon as DSM-IV was published, the ADHD age at onset criterion became a subject of intense debate.5 Numerous authors have supported revising the current age at onset criterion by extending the upper limit to 125,6 and even 18 years,1,7-12 mainly for 2 reasons. First, no study comparing subjects with a DSM diagnosis of ADHD and an age at onset of 7 years or less to those with an age at onset after 7 years has reported differences in terms of course, severity, or treatment response.5 Second, prospective studies have highlighted that a substantial proportion of ADHD subjects make errors in reporting the age at onset of ADHD. In fact, previous studies12,13 suggest that nearly one-half of the ADHD participants with an age at onset of 7 years or less reported an age at onset after 7 years when reassessed retrospectively. Thus, extending the age at onset criterion would limit the risk of retrospective recall and reduce false-negatives.5 However, the psychometric properties of the ADHD symptoms might depend on the age at onset. It is therefore critical to generate empirical evidence to support the extension of the age at onset to 12 years4 or even 18 years, as previously suggested.1,7-12 Therefore, in this study, we considered the following ages at onset: ≤ 7 years, > 7 years and ≤ 12 years, and > 12 years and ≤ 18 years. For ease of reference, we will indicate these ages at onset as early age at onset (EAO), late age at onset (LAO), and very late age at onset (VLAO), respectively, while recognizing that this is non-official labeling.

In the present study, we used an item response theory (IRT)-based approach that has been conducted in various studies to determine the psychometric properties of DSM-IV symptoms of several mental disorders.14-16 A 2-parameter IRT model provides information regarding the point on a latent trait at which an item has a 50% probability of endorsement (item severity parameter) and how rapidly an item’s probability of endorsement changes across the latent trait (item discrimination parameter). Advantages of using an IRT-based approach over other statistical methodologies include the possibility of examining the likelihood that a particular symptom will be endorsed at a particular level of severity. Controlling for overall symptom severity among groups is critical because it is unclear whether any differential symptom expression reported in the literature5 is due to true phenomenological differences between age at onset groups or whether such differences are reflective of greater overall symptom severity in one group versus another.

With the use of a large nationally representative sample, the National Epidemiologic Survey on Alcohol and Related Conditions (NESARC), our aim was to compare item severity and discrimination parameters (ie, differential item functioning [DIF]) of ADHD symptoms between the EAO and LAO groups and between the EAO and VLAO groups along 2 latent ADHD traits (inattention and hyperactivity-impulsivity). We hypothesized that there is no significant difference in the psychometric properties of DSM-IV ADHD symptoms between age at onset groups.

• This study provides psychometric support to the change in the age at onset criterion introduced in DSM-5 and further suggests that the age at onset could be extended to 18 years without modifying the psychometric properties of attention-deficit/hyperactivity disorder symptoms.
• Considerations beyond these statistical properties should be taken into account before considering further modifications of the maximum age at symptom onset.

METHOD

NESARC Sample

Data were drawn from the 2004-2005 NESARC, the second wave that followed the Wave 1 NESARC, conducted in 2001-2002 and described in detail elsewhere.17,18 The Wave 1 NESARC was a nationally representative survey of the population of the United States conducted by the US Census Bureau under the direction of the National Institute on Alcohol Abuse and Alcoholism.19 The target population included the civilian noninstitutionalized population, aged 18 years and older, residing in the United States. Face-to-face personal interviews were conducted with 43,093 individuals. The overall survey response rate was 81%. Blacks, Hispanics, and young adults (aged 18-24 years) were oversampled19 because these groups have been underrepresented in previous comorbidity surveys in the United States.17 Data were weighted at the individual and household levels to adjust for oversampling and nonresponse on demographic variables and to be representative of the US civilian population based on the 2000 census. Excluding respondents not eligible for the Wave 2 interview (eg, because they were deceased or were mentally or physically impaired), the Wave 2 response rate was 86.7%, reflecting 34,653 completed Wave 2 interviews.20 The cumulative response rate at Wave 2 was 70.2%. As in Wave 1, the Wave 2 NESARC data were weighted to reflect design characteristics of the NESARC survey and to account for oversampling.20 These weights were not used in the current analyses. The research protocol, including written informed consent procedures, received full ethical review and approval from the US Census Bureau and the Office of Management and Budget.17,18

ADHD Clinical Assessment

ADHD diagnosis was assessed according to the DSM-IV criteria (except the age at onset criterion, which was set at 18 years for the purpose of this study) by using the Alcohol Use Disorder and Associated Disabilities Interview Schedule-IV (AUDADIS-IV). This instrument is a valid and reliable fully structured diagnostic interview designed for use by professional interviewers who are not clinicians.18 Specifically, participants were asked 20 symptom questions operationalizing the 18 DSM-IV ADHD criteria.21 Consistent with DSM-IV criteria, lifetime and childhood AUDADIS-IV diagnosis of ADHD required the respondent to meet the symptom threshold of 6 or more DSM-IV symptoms of either inattention or hyperactivity-impulsivity or both that have persisted for at least 6 months to a degree that the symptoms interfere significantly with social, school, or work functioning. Test-retest reliability for DSM-IV ADHD was good (κ = 0.71).22 Internal consistency reliability of the ADHD symptom items was excellent (Cronbach α = 0.89).22

Data Analysis

We first converted the 20 questions corresponding to the diagnostic criteria of ADHD into the 18 items following DSM-IV criteria. The 2 questions “often very active when not supposed to be” and “often feel restless” were converted into 1 criterion, as were the items “often interrupt people” and “often interrupt activities that had already started.” Comparisons of the prevalence of each symptom across age at onset groups were performed using χ2 tests. Second, we conducted confirmatory factor analyses in both dimensions of ADHD and in the 3 age at onset groups separately to determine if the symptoms met the unidimensionality assumption. Factorial analyses were calculated using the Mplus statistical software program.23 Because ADHD symptoms are assessed as 2 separate dimensions in the DSM (ie, inattention and hyperactivity-impulsivity), all analyses in the present study were performed separately for these 2 dimensions.

Item Response Theory

A 2-parameter IRT model was conducted on each dimension separately (ie, inattention and hyperactivity-impulsivity) and on each age at onset group.24 The IRT model estimates parameters describing the relationship between the probability of an item response (eg, endorsement of the item “easily distracted from play or work”) and an individual’s level of the latent trait (in this example, the inattentive dimension). The 2-parameter IRT model estimates (1) a severity parameter to determine the point along the latent trait at which a symptom has a probability above 50% to be endorsed25,26 and (2) a discrimination parameter to describe how rapidly the probability of observing the item changes across increasing levels of the latent trait15 (ie, the slope of the item response function). The severity parameter is reflective of the likelihood that a given symptom will occur at a given severity level, whereas the discrimination parameter allows identification of whether a given symptom is a good or poor indicator of the latent trait.27 Once IRT parameters for each item were estimated, Spearman rank order correlation coefficient was used to evaluate the ordering of the severity and discrimination parameters between the different groups, for each latent trait.

Differential Item Functioning

DIF analysis was conducted using the software IRTLRDIF version 2.028 to examine whether the ADHD symptom function was similar between the different age at onset groups, following a method previously described by McBride et al.29 The DIF approach compares 2 groups (the reference group and the focal group) and utilizes information about the measurement properties of the set of items simultaneously.14 Concerning the current analyses, the focal groups comprised individuals with LAO and VLAO, whereas the reference group included those with EAO. Analyses were conducted iteratively to determine which item function differed across age at onset groups and which item was DIF free. To explore for DIF, the discrimination and severity parameters for each age at onset group were constrained to be equal across all 9 criteria of each ADHD dimension. For each item, the likelihood ratio test statistic (G2 test; df = 2) was calculated. The aim was to compare the model with the parameter estimates constrained to be equal between the reference and the focal group with a model that frees the parameters to be estimated separately between the 2 groups. If the omnibus test was significant, likelihood ratio tests (G2 test; df = 1) were then conducted to identify if the DIF was present in the discrimination, the severity, or both parameters of the item. Due to multiple comparisons implemented in this study, we set α at .05 and used the Benjamini-Hochberg procedure to adjust P values for all tests with 1 degree of freedom.30,31 Small differences in severity between groups could be statistically significant but may not be clinically meaningful.14 Thus, it was decided a priori that only differences higher than 0.25 in symptom severity, which can be interpreted as one-quarter of the “standard unit difference between the value of the (underlying) trait necessary to have a 50-50 chance of responding positively in one group compared to another,”32(pp405-406) would be considered as clinically meaningful. Minimum sample size for DIF analyses is usually considered in the range of 100-200 subjects per group.33 On the basis of a simulation study, Scott et al34 recommended a minimum of 200 participants per group to ensure adequate performance (ie, 80% power). Finally, the test information function (TIF) and the standard error of measurement (SEM; equal to the inverse square root of the TIF) were estimated in each group and on each latent trait. The TIF is a graphic representation of the total quantity of information yielded by a set of items at each latent trait level. The SEM is related to the reliability of the measurement (the SEM is equal to the square root of 1 minus reliability; eg, a SEM of 0.55 is equal to an internal consistency of 0.70).35 The TIF and the SEM represent the information and precision of a set of items across different levels of a latent trait.

RESULTS

Sample

Of the 807 individuals with a lifetime DSM-IV diagnosis of ADHD at Wave 2, we identified 364 participants (45.1%) with an EAO, 252 (31.2%) with an LAO, and 148 (18.3%) with a VLAO. Data on age at onset were missing for 43 participants (5.3%) with a lifetime diagnosis of ADHD.

Endorsement Rates

The endorsement rates of all symptoms of ADHD were greater than 30.7% in each group (Table 1). Using the Benjamini-Hochberg procedure, P values for χ2 tests comparing prevalence rates among groups were found to be nonsignificant except for the item “get up from seat” (63.81% in EAO, 59.76% in LAO, and 47.30% in VLAO; P = .0026). The endorsement rates were notably different between the EAO and VLAO groups (χ2 test = 13.69; P = .002; χ2 test comparing LAO and VLAO = 6.44; P = .04).

Factor Analyses

Confirmatory factor analysis was performed to examine the dimensional properties of the symptoms of ADHD in both dimensions of ADHD symptoms separately (Table 2). For both dimensions and the 3 age at onset groups, the χ2 values were significant. However, the χ2 statistic is known to be highly sensitive to large sample sizes and may have overestimated the lack of fit of the structural model in this study.36 The other fit indices (comparative fit index, Tucker-Lewis index, and root mean square error of approximation) indicated an adequate fit to the data.

IRT Item Parameters

The 2-parameter IRT models were conducted on 2 separate latent traits corresponding to the DSM-IV dimensions, inattention and hyperactivity-impulsivity (Table 3).

IRT analysis of the inattention dimension. We found no significant DIF on IRT parameters. The ranking of IRT parameters was similar between EAO and LAO groups (Spearman correlation coefficients were 0.82 for severity parameters [differences of rank (dor) < 4] and 0.63 for discrimination parameters [dor > 3 only for the items “lose things” (dor = 4) and “avoid things requiring concentration” (dor = 5)]) and EAO and VLAO groups (Spearman correlation coefficients were 0.93 for severity parameters [dor < 4] and 0.53 for discrimination parameters [dor > 3 only for the item “lose things” (dor = 6)]).

IRT analysis of the hyperactivity-impulsivity dimension. There was no significant DIF on IRT parameters. Spearman correlation coefficients were 0.92 and 0.68 for severity parameters and 0.73 and 0.65 for discrimination parameters, for EAO versus LAO and EAO versus VLAO, respectively. Differences of IRT parameters rank were < 4, except for the discrimination parameters of the item “get up from seat” for EAO versus VLAO (dor = 4).

Test Information Function

The TIF and the SEM for the latent traits of both the inattention and hyperactivity-impulsivity dimensions are presented in Figure 1. For both latent traits, the SEM curves indicate that the measurement in the range for which there was acceptable SEM (ie, SEM < 0.55) was similar across groups.

DISCUSSION

The present study sought to examine measurement properties of ADHD symptoms according to age at onset. The endorsement rates of the symptoms of ADHD were not significantly different across age at onset groups except for the symptom “get up from seat” (hyperactivity-impulsivity dimension), which was more frequent in participants with an EAO than in those with a VLAO. This result could be accounted for by the fact that the EAO group showed a significantly higher level of the latent trait hyperactivity-impulsivity than the VLAO group (Wald test = 6.95; P = .009). Another explanation could be that this item performs differently in EAO and VLAO groups. Although the DIF test was not significant, the discrimination parameter was numerically lower in the VLAO group, suggesting a better discrimination for the EAO compared to the VLAO group. Moreover, the sample size of the VLAO group was smaller than the size previously recommended (ie, 200 subjects per group34). Therefore, the power to detect this difference may have been insufficient in our study. Results from the current study support that the psychometric properties of the ADHD symptoms in participants with EAO are remarkably similar to those with LAO and VLAO. Indeed, in both dimensions of ADHD (inattention and hyperactivity-impulsivity), no differential item functioning was significant, and the ordering of the severity and discrimination parameters was similar across age at onset groups.

Our study indicates that only 45.1% of the participants who met the DSM-IV symptoms of ADHD reported an age at onset before 7 years. This result should be considered in light of the important clinical implications of extending the age at onset in the diagnostic criteria.6 First, the extension of the age at onset criterion from 7 years to 18 years would result in a doubling of the ADHD prevalence in our sample.6 Second, although this extension may have the potential to reduce bias recall and thus the number of false-negatives,12,13 it may also lead to an increase in the number of false-positives.37 Although our study would support an extension of the age at onset criterion up to 18 years based on the psychometric properties of the ADHD symptoms, considerations beyond statistical properties must be taken into account for psychiatry nosology. An extension of the age at onset criterion up to 18 years raises other issues, especially for a disorder that is conceptualized as a neurodevelopmental disorder emerging during childhood. The revision of the current ADHD age at onset criterion by extending the upper limit to age 12, as recently acknowledged in DSM-5, is prudent (by limiting an increase of false-positives) and preserves the notion that ADHD emerges during early childhood.6 Our study provides psychometric support to the choice of the DSM-5 committee.

Our results should be considered in light of some limitations. A first limitation involves the retrospective nature of the assessment of lifetime ADHD symptoms. Our findings should be further examined in studies assessing prospectively current ADHD symptoms among youth of different ages. In addition, errors in reporting ADHD symptoms may have occurred in our study. Indeed, several studies indicate that retrospective report can lead to underestimation38 or overestimation39 of ADHD prevalence. However, reliability and validity of the retrospective report of ADHD symptoms have been shown as adequate in several studies.40,41 Second, symptoms of ADHD among participants with symptoms below a diagnostic threshold were not available, thus limiting the full examination of item functioning. Third, the data are cross-sectional, and important information about social contextual influences, cognitive development, and clinical course (eg, length of illness, medication regimen), which may shape the incidence of symptoms, was not available in NESARC.42,43 Future longitudinal studies should take the potential influence of these factors into account. Fourth, ADHD diagnosis was assessed following DSM-IV criteria, except for the age at onset criterion, which was set at 18 years for the purpose of this study. This could have led to decreases in the test-retest reliability of ADHD diagnosis and internal consistency reliability of ADHD symptom items. However, our findings suggest that symptom expression of ADHD was not affected by age at onset. Last, although NESARC has a nationally representative sample, it is uncertain how findings from the present study would be similar or different if enriched correctional or clinical samples were employed. In particular, information on institutionalized individuals, eg, those in the hospital or in prison (for whom ADHD symptoms expression might be different than that in the general population44) was unavailable, thus limiting the generalizability of our findings.45,46

Despite these limitations, the present study, using a methodology grounded in IRT, suggests that symptom expression of ADHD is not affected by age at onset. This study provides psychometric support to the age at onset criterion (ie, 12 years), recently modified in the DSM-5 and further suggests that the age at onset could be extended to 18 years without changing the psychometric properties of the diagnostic criteria.

Author affiliations: Assistance Publique Hôpitaux de Paris (APHP), Robert Debré Hospital, Child and Adolescent Psychiatry Department, Paris (Drs Peyre, Acquaviva, and Delorme); Cognitive Sciences and Psycholinguistic Laboratory, Ecole Normale Supérieure, Paris (Dr Peyre); Assistance Publique-Hôpitaux de Paris (APHP), Corentin Celton Hospital, Department of Psychiatry, Issy-les-Moulineaux; Paris Descartes University, PRES Sorbonne Paris Cité, Paris (Drs Hoertel and Limosin); INSERM UMR-894, Psychiatry and Neurosciences Center; Paris Descartes University, PRES Sorbonne Paris Cité, Paris (Drs Hoertel, Maricourt, and Limosin); INSERM UMR-669, Paris-Sud Innovation Group in Adolescent Mental Health; Paris Sud University, Villejuif (Dr Acquaviva); and Human Genetics and Cognitive Functions, Institut Pasteur, Paris (Dr Delorme), France; Phyllis Green and Randolph Cowen Institute for Pediatric Neuroscience, Child Study Center of the NYU Langone Medical Center, New York, New York (Dr Cortese); and Child Neuropsychiatry Unit, Department of Life Sciences and Reproduction, Verona University, Verona, Italy (Dr Cortese).

Potential conflicts of interest: Dr Cortese has served as scientific consultant for Shire (2009-2010). Dr Limosin has served as scientific consultant for Lundbeck and Eutherapy. Drs Peyre, Hoertel, Acquaviva, De Maricourt, and Delorme report no conflicts of interest.

Funding/support: None reported.

Additional information: The original data set relative for the National Epidemiologic Survey on Alcohol and Related Conditions (NESARC) is available from the National Institute on Alcohol Abuse and Alcoholism (http://www.niaaa.nih.gov).

REFERENCES

1. Applegate B, Lahey BB, Hart EL, et al. Validity of the age-of-onset criterion for ADHD: a report from the DSM-IV field trials. J Am Acad Child Adolesc Psychiatry. 1997;36(9):1211-1221. PubMed doi:10.1097/00004583-199709000-00013

2. Polanczyk G, de Lima MS, Horta BL, et al. The worldwide prevalence of ADHD: a systematic review and metaregression analysis. Am J Psychiatry. 2007;164(6):942-948. PubMed doi:10.1176/appi.ajp.164.6.942

3. Faraone SV, Biederman J, Mick E. The age-dependent decline of attention deficit hyperactivity disorder: a meta-analysis of follow-up studies. Psychol Med. 2006;36(2):159-165. PubMed doi:10.1017/S003329170500471X

4. American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition. Arlington, VA: American Psychiatric Association; 2013.

5. Kieling C, Kieling RR, Rohde LA, et al. The age at onset of attention deficit hyperactivity disorder. Am J Psychiatry. 2010;167(1):14-16. PubMed doi:10.1176/appi.ajp.2009.09060796

6. Polanczyk G, Caspi A, Houts R, et al. Implications of extending the ADHD age-of-onset criterion to age 12: results from a prospectively studied birth cohort. J Am Acad Child Adolesc Psychiatry. 2010;49(3):210-216. PubMed

7. Barkley RA, Biederman J. Toward a broader definition of the age-of-onset criterion for attention-deficit hyperactivity disorder. J Am Acad Child Adolesc Psychiatry. 1997;36(9):1204-1210. PubMed doi:10.1097/00004583-199709000-00012

8. Faraone SV, Biederman J, Doyle A, et al. Neuropsychological studies of late onset and subthreshold diagnoses of adult attention-deficit/hyperactivity disorder. Biol Psychiatry. 2006;60(10):1081-1087. PubMed doi:10.1016/j.biopsych.2006.03.060

9. Faraone SV, Biederman J, Spencer T, et al. Diagnosing adult attention deficit hyperactivity disorder: are late onset and subthreshold diagnoses valid? Am J Psychiatry. 2006;163(10):1720-1729, quiz 1859. PubMed

10. Faraone SV, Kunwar A, Adamson J, et al. Personality traits among ADHD adults: implications of late-onset and subthreshold diagnoses. Psychol Med. 2009;39(4):685-693. PubMed doi:10.1017/S0033291708003917

11. Reinhardt MC, Benetti L, Victor MM, et al. Is age-at-onset criterion relevant for the response to methylphenidate in attention-deficit/hyperactivity disorder? J Clin Psychiatry. 2007;68(7):1109-1116. PubMed doi:10.4088/JCP.v68n0720

12. Todd RD, Huang H, Henderson CA. Poor utility of the age of onset criterion for DSM-IV attention deficit/hyperactivity disorder: recommendations for DSM-V and ICD-11. J Child Psychol Psychiatry. 2008;49(9):942-949. PubMed doi:10.1111/j.1469-7610.2008.01892.x

13. Barkley RA, Murphy KR, Fischer M. ADHD in Adults: What the Science Says. New York, NY: Guilford Press; 2010.

14. Strong DR, Kahler CW, Colby SM, et al. Linking measures of adolescent nicotine dependence to a common latent continuum. Drug Alcohol Depend. 2009;99(1-3):296-308. PubMed doi:10.1016/j.drugalcdep.2008.09.001

15. Weinstock LM, Strong D, Uebelacker LA, et al. Differential item functioning of DSM-IV depressive symptoms in individuals with a history of mania versus those without: an item response theory analysis. Bipolar Disord. 2009;11(3):289-297. PubMed doi:10.1111/j.1399-5618.2009.00681.x

16. Gelhorn H, Hartman C, Sakai J, et al. An item response theory analysis of DSM-IV conduct disorder. J Am Acad Child Adolesc Psychiatry. 2009;48(1):42-50. PubMed doi:10.1097/CHI.0b013e31818b1c4e

17. Grant BF, Stinson FS, Dawson DA, et al. Prevalence and co-occurrence of substance use disorders and independent mood and anxiety disorders: results from the National Epidemiologic Survey on Alcohol and Related Conditions. Arch Gen Psychiatry. 2004;61(8):807-816. PubMed doi:10.1001/archpsyc.61.8.807

18. Grant BF, Harford TC, Dawson DA, et al. The Alcohol Use Disorder and Associated Disabilities Interview Schedule (AUDADIS): reliability of alcohol and drug modules in a general population sample. Drug Alcohol Depend. 1995;39(1):37-44. PubMed doi:10.1016/0376-8716(95)01134-K

19. Grant BF, Stinson FS, Dawson DA, et al. Co-occurrence of 12-month alcohol and drug use disorders and personality disorders in the United States: results from the National Epidemiologic Survey on Alcohol and Related Conditions. Arch Gen Psychiatry. 2004;61(4):361-368. PubMed doi:10.1001/archpsyc.61.4.361

20. Grant BF, Stinson FS, Hasin DS, et al. Prevalence, correlates, and comorbidity of bipolar I disorder and Axis I and II disorders: results from the National Epidemiologic Survey on Alcohol and Related Conditions. J Clin Psychiatry. 2005;66(10):1205-1215. PubMed doi:10.4088/JCP.v66n1001

21. Bernardi S, Faraone SV, Cortese S, et al. The lifetime impact of attention deficit hyperactivity disorder: results from the National Epidemiologic Survey on Alcohol and Related Conditions (NESARC). Psychol Med. 2012;42(4):875-887. PubMed doi:10.1017/S003329171100153X

22. Ruan WJ, Goldstein RB, Chou SP, et al. The Alcohol Use Disorder and Associated Disabilities Interview Schedule-IV (AUDADIS-IV): reliability of new psychiatric diagnostic modules and risk factors in a general population sample. Drug Alcohol Depend. 2008;92(1-3):27-36. PubMed doi:10.1016/j.drugalcdep.2007.06.001

23. Muthén L, Muthén B. Mplus: Statistical Analyses With Latent Variables. User’s Guide. Los Angeles, CA: Muthen and Muthen; 2007.

>24. Birnbaum A. Some latent trait models and their use in inferring an examinee’s ability. In: Lord FM, Novick MR. Statistical Theories of Mental Test Scores. New York, NY: Addison-Wesley; 1968:395-479.

25. Kahler CW, Strong DR. A Rasch model analysis of DSM-IV alcohol abuse and dependence items in the National Epidemiological Survey on Alcohol and Related Conditions. Alcohol Clin Exp Res. 2006;30(7):1165-1175. PubMed doi:10.1111/j.1530-0277.2006.00140.x

26. Saha TD, Stinson FS, Grant BF. The role of alcohol consumption in future classifications of alcohol use disorders. Drug Alcohol Depend. 2007;89(1):82-92. PubMed doi:10.1016/j.drugalcdep.2006.12.003

27. Hartman CA, Gelhorn H, Crowley TJ, et al. Item response theory analysis of DSM-IV cannabis abuse and dependence criteria in adolescents. J Am Acad Child Adolesc Psychiatry. 2008;47(2):165-173. PubMed doi:10.1097/chi.0b013e31815cd9f2

28. Thissen D. IRTLRDIF v.2.OB: Software for the Computation of the Statistics Involved in Item Response Theory Likelihood Ratio Tests for Differential Item Functioning. Chapel Hill, NC: University of North Carolina L. L. Thurstone Psychometric Laboratory; 2001.

29. McBride O, Strong DR, Kahler CW. Exploring the role of a nicotine quantity-frequency use criterion in the classification of nicotine dependence and the stability of a nicotine dependence continuum over time. Nicotine Tob Res. 2010;12(3):207-216. PubMed doi:10.1093/ntr/ntp196

30. Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J R Stat Soc, B. 1995;57(1):289-300.

31. Thissen D, Steinberg L, Kuang D. Quick and easy implementation of the Benjamini-Hochberg procedure for controlling the false positive rate in multiple comparisons. J Educ Behav Stat. 2002;27(1):77-83. doi:10.3102/10769986027001077

32. Steinberg L, Thissen D. Using effect sizes for research reporting: examples using item response theory to analyze differential item functioning. Psychol Methods. 2006;11(4):402-415. PubMed doi:10.1037/1082-989X.11.4.402

33. Zumbo BD. A Handbook on the Theory and Methods of Differential Item Functioning (DIF): Logistic Regression Modeling as a Unitary Framework for Binary and Likert-Type (Ordinal) Item Scores. Ottawa, Ontario, Canada: Directorate of Human Resources Research and Evaluation, Department of National Defense; 1999.

34. Scott NW, Fayers PM, Aaronson NK, et al; EORTC Quality of Life Group and the Quality of Life Cross-Cultural Meta-Analysis Group. Differential item functioning (DIF) analyses of health-related quality of life instruments using logistic regression. Health Qual Life Outcomes. 2010;8(1):81. PubMed

35. Weiss DJ, Davison ML. Test theory and methods. Annu Rev Psychol. 1981;32(1):629-658. doi:10.1146/annurev.ps.32.020181.003213

36. Bollen KA. Structural Equations With Latent Variables. New York, NY: Wiley; 1989.

37. Frances A. Increasing the age at onset for ADHD? Am J Psychiatry. 2010;167(6):718, author reply 718-719. PubMed doi:10.1176/appi.ajp.2010.10010097

38. Barkley RA, Fischer M, Smallish L, et al. The persistence of attention-deficit/hyperactivity disorder into young adulthood as a function of reporting source and definition of disorder. J Abnorm Psychol. 2002;111(2):279-289. PubMed doi:10.1037/0021-843X.111.2.279

39. Mannuzza S, Klein RG, Klein DF, et al. Accuracy of adult recall of childhood attention deficit hyperactivity disorder. Am J Psychiatry. 2002;159(11):1882-1888. PubMed doi:10.1176/appi.ajp.159.11.1882

40. Epstein JN, Kollins SH. Psychometric properties of an adult ADHD diagnostic interview. J Atten Disord. 2006;9(3):504-514. PubMed doi:10.1177/1087054705283575

41. Ward MF, Wender PH, Reimherr FW. The Wender Utah Rating Scale: an aid in the retrospective diagnosis of childhood attention deficit hyperactivity disorder. Am J Psychiatry. 1993;150(6):885-890. PubMed

42. Peyre H, Hoertel N, Cortese S, et al. Long-term effects of ADHD medication on adult height: results from the NESARC [letter]. J Clin Psychiatry. 2013;74(11):1123-1124. PubMed doi:10.4088/JCP.13l08580

43. Peyre H, Hoertel N, Hatteea H, et al. Adulthood self-reported cardiovascular risk and ADHD medications: results from the 2004-2005 National Epidemiologic Survey on Alcohol and Related Conditions [letter]. J Clin Psychiatry. 2014;75(2):181-182.

44. Ginsberg Y, Hirvikoski T, Lindefors N. Attention deficit hyperactivity disorder (ADHD) among longer-term prison inmates is a prevalent, persistent and disabling disorder. BMC Psychiatry. 2010;10(1):112. PubMed doi:10.1186/1471-244X-10-112

45. Hoertel N, Le Strat Y, Blanco C, et al. Generalizability of clinical trial results for generalized anxiety disorder to community samples. Depress Anxiety. 2012;29(7):614-620. PubMed doi:10.1002/da.21937

46. Hoertel N, Le Strat Y, Lavaud P, et al. Generalizability of clinical trial results for bipolar disorder to community samples: findings from the National Epidemiologic Survey on Alcohol and Related Conditions. J Clin Psychiatry. 2013;74(3):265-270. PubMed doi:10.4088/JCP.12m07935

Editor’s Note: We encourage authors to submit papers for consideration as a part of our Early Career Psychiatrists section. Please contact Marlene P. Freeman, MD, at [email protected].