Abstract
In the South African context, resource constraints often preclude the comprehensive assessment of large numbers of people for the likelihood of Adult Attention-Deficit/Hyperactivity Disorder (ADHD). Primary screening through a self-report measure may be useful to stream at-risk individuals towards diagnostic assessment services, as well as being useful in population and workplace based research. The present study set out, first, to investigate the usefulness of a self-report ADHD scale to identify at-risk individuals, and, second, to provide preliminary prevalence estimates for Adult ADHD, guided by Diagnostic and Statistical Manual of Mental Disorders (5th ed.; DSM-5) criteria, in a South African workplace sample. Workers in full-time skilled employment (N = 1,917), aged 18–44, completed a self-report Adult ADHD scale, and participated in an interview with a psychologist. Their scale data, using three different scoring and interpretation systems, were subjected to statistical analysis. Favourable internal reliability and positive predictive validity were found. Different interpretation systems provided different prevalence estimations: using DSM-5 criteria, a total prevalence estimate of 3.3 % was calculated (attention deficit subtype = 0.9%, hyperactivity-impulsivity subtype = 1.0%, and combined subtype = 1.4%). The positive predictive validity found with this sample suggests that this scale can be used constructively in research or screening contexts to identify at-risk individuals. Furthermore, preliminary prevalence estimates for Adult ADHD, guided by DSM-5 criteria, are now available for a South African workplace sample.
Attention-Deficit/Hyperactivity Disorder (ADHD) is a condition characterised by a persistent pattern of inattention and/or hyperactivity-impulsivity that interferes with functioning or development (American Psychiatric Association [APA], 2013). The inattentive or hyperactive-impulsive symptoms are present before the age of 12 years, are present in two or more settings (e.g., at home, school, work, or in other activities), and demonstrably interfere with social, academic, or occupational functioning. ADHD is always a clinical diagnosis. The two main classification systems currently in use for diagnosis are the Diagnostic and Statistical Manual of Mental Disorders (5th ed.; DSM-5; APA, 2013) and International Classification of Diseases, Tenth Revision (ICD-10) (World Health Organization [WHO], 1993).
ADHD is not limited to the early developing years; between 50% and 66% of childhood cases persist into adulthood (Barkley, Fischer, Smallish, & Fletcher, 2002; Ebejer et al., 2012; Faraone, Biederman, & Mick, 2006; Lara et al., 2009). Adult ADHD (A-ADHD) is associated with significant functional impairment (Kessler, Lane, Stang, & Van Brunt, 2009). Inattentive symptoms are more common in adults than symptoms of hyperactivity or impulsivity (Asherson, 2005; Biederman, Mick, & Faraone, 2000). Comorbid disorders are common in adults with ADHD (APA, 2013; Asherson, 2005; Friedrichs, Igl, Larsson, & Larsson, 2012; Hesson & Fowler, 2018), with substance use, mood, and anxiety disorders most prominent (Fayyad et al., 2007; Gjervan, Torgersen, Nordahl, & Rasmussen, 2012; Kessler, Berglund et al, 2005; McGough et al., 2005; Nierenberg et al., 2005). A-ADHD is further associated with lower levels of education and employment (Ebejer et al., 2012; Fayyad et al., 2007; Gjervan et al., 2012; Küpper et al., 2012). ADHD has a significant impact on work (Adamou et al., 2013; Biederman et al., 2006; de Graaf et al., 2008). Adults with ADHD are at increased risk of accidents and workplace injuries, particularly traffic accidents (Küpper et al., 2012), with their chances of workplace accidents-injuries up to double that of controls (OR = 2; Kessler et al., 2009).
Most of the currently available epidemiology reports are based on Diagnostic and Statistical Manual of Mental Disorders (4th ed.; DSM-IV) criteria, with wide variation across studies (Simon, Czobor, Bálint, Mészáros, & Bitter, 2009). One meta-regression analysis reported a pooled prevalence (i.e., the percentage of cases that are present in a population at a given time) of 2.5% (Simon et al., 2009). The WHO’s World Mental Health Survey (WMHS; conducted across 10 countries) suggested a population-based prevalence of 3.4% (1.2%–7.3%, dependent on country) for ADHD among adults (Fayyad et al., 2007). Higher rates were observed among men (4.1%, vs. 2.7% for women), while the rates remained fairly stable across ages 18–44 (Fayyad et al., 2007). Cross-national differences in prevalence estimates are noteworthy, with lower prevalence in lower income countries (M = 1.9%) compared with higher income countries (M = 4.2 %; Fayyad et al., 2007). Recent figures from the developed world ranged from 2.9% in Canada (Hesson & Fowler, 2018) to 7.1% in France (Moulin et al., 2018). Some of the differences across studies could potentially be attributed to the different age ranges used, as well as their different diagnostic focus (e.g., attention deficit with or without hyperactivity).
With regard to worker population samples, the WHO WMHS suggested a prevalence of 3.5% (1.2%–7.3%, dependent on country) for ADHD among adults (de Graaf et al., 2008). A-ADHD appears to be more common among males than females, and also to remain fairly stable across ages 18–44 (de Graaf et al., 2008). Higher rates were observed in developed countries than in developing nations (using World Bank classification; de Graaf et al., 2008). Workplace prevalence rates from studies in the United States report similar findings, from 2.9% (Biederman, 2005), to 4.2% (Kessler, Adler, Ames, Barkley, et al., 2005) and 4.4% (Kessler et al., 2006).
The DSM-5 estimates an adult population ADHD prevalence around 2.5% (APA, 2013). Only one study using DSM-5 criteria in a large sample field trial could be located in preparation of this paper: Matte et al. (2015) reported a prevalence of 3.55% among young adults in Brazil, using structured interviews. Inattentive symptoms were more prominent in their sample and were furthermore a better predictor of clinical impairment than were symptoms of hyperactivity/impulsivity. Little gender effect was found in their sample of 18–19 year olds (Matte et al., 2015).
Apart from direct interviewing, screening for A-ADHD in workplace studies most often used the WHO’s Adult ADHD Self-Report Scale (ASRS). The ASRS is a validated screening scale for DSM-IV A-ADHD (Kessler, Adler, Ames, Demler, et al., 2005). The scale consists of 18 items. The first six are referred to as ‘Set A’ and are used to identify the likelihood of ADHD. The next 12 items (‘Set B’) are considered additional cues and not used for determining diagnostic likelihood. The main limitation of the ASRS is that it is based on the now out-dated DSM-IV, and the items are not fully aligned to the updated criteria in the DSM-5.
Prevalence figures for A-ADHD in South Africa (SA) are not known (Schoeman & Liebenberg, 2017), although estimates have been developed for adolescents (Walker, Venter, van der Walt, & Esterhuyse, 2011).
In occupational and clinical contexts in SA, resource constraints often preclude the comprehensive assessment of large numbers of people – be they employees or patients – to determine likelihood of A-ADHD. Primary screening through a self-report measure could be extremely useful to stream at-risk individuals towards diagnostic assessment services. In such a context, however, optimal interpretation frameworks and cut-points for any such measure would be required. Furthermore, screening A-ADHD through self-report may also be useful for both population and workplace-based research.
To address the twin challenges of the lack of available workplace prevalence estimates, and the lack of a validated self-report screening instrument for DSM-5 criteria in the local context, the present study set itself two aims: first, to investigate the usefulness of a self-report A-ADHD scale to identify at-risk individuals; second, if the scale is shown to have acceptable psychometric properties, to provide preliminary prevalence estimates for A-ADHD – guided by DSM-5 criteria – in a SA workplace sample.
Methods
Participants
Workers in full-time employment were invited to complete the A-ADHD scale. Individuals were eligible for inclusion in the study if they met the following criteria, namely, were in the age range 18–44, gave informed consent, and had at least 9 years of formal schooling (to enable them to complete the English language survey unassisted). As one requirement for a diagnosis of ADHD is onset of symptoms in childhood, the assessment was limited to respondents in the age range of 18–44 years because of concerns about the accuracy of retrospective recall among older respondents (cf. Fayyad et al., 2007). The sample composition (N = 1,917) is presented in Table 1. Age by gender distribution is presented in Figure 1. No record was kept on how many individuals were invited, thus no response rate data is available. The participants were drawn from across different sites and occupational contexts. They were all considered ‘skilled labour’, as they all had some form of vocational training (either post school academic training or formal employer-provided training).
Sample composition, and prevalence estimation of A-ADHD in a South African workplace sample, using the AASS-5, expressed in percentages.
ADHD: attention deficit hyperactivity disorder; AASS: Adult ADHD Symptom Scale; DSM-5: Diagnostic and Statistical Manual of Mental Disorders (5th ed.).
Gender and age distribution of sample.
Instruments
This study used a modified version of the ASRS, updated to be aligned to DSM-5 Criterion A (detailed in Psych Central, n.d.). The modified version was developed using international research (Grohol, 2019) which mainly originated from the global north. The present study added three items to reflect DSM-5 Criteria B to D to the modified scale, which in this study was termed the Adult ADHD Symptom Scale for DSM-5 (AASS-5) for easy reference. The AASS-5 thus consists of 21 items in question format and was formulated according to DSM-5 Criteria A to D. Items 1–18 correspond to DSM-5 Criterion A and are endorsed on a 4-point Likert-type scale (anchored with never and often). Items 19–21 correspond to Criteria B to D and are indicated as either yes or no. As far as could be ascertained, the validity of the AASS-5 in the local context has not yet been established. While updated to DSM-5 criteria, the modified version did not substantially change Set A, which thus allowed responses on the AASS-5 to be compared with those of the original ASRS. No other DSM-5 based self-report survey-type screening instruments could be located in preparation of this study.
The AASS-5 can be scored and interpreted in three ways:
By using the six ‘Set A’ items, as was done with the ASRS, and which would allow relative comparison with existing published data.
By interpreting the Total Scale scores (i.e., the sum of all responses across 18 items). Full interpretation possibilities can be found at https://psychcentral.com, where it is reported that sum totals of 26–33 suggests that ADHD is possible, and > 33 suggests that ADHD is likely.
By using DSM-5 criteria. The DSM-5 requires five checks for either inattention (AD; Items 1–9) or hyperactivity-impulsivity (HA; Items 10–18), or five checks for both to diagnose combined type (for Criterion A), as well as endorsement of Items 19–21 (Criteria B, C, and D). However, experience during the study indicated that Item 21 was interpreted very widely, so caution must be advised when this score is used to determine diagnostic likelihood.
Participants also completed socio-demographic details on their survey forms, namely age, gender, home language, and field of work. Field of work was categorised retrospectively, and entries that were blank or unclear, or fields that applied to less than 5% of the sample, were collapsed into an ‘other’ category. Furthermore, participants were invited for an interview to clarify their answers. To minimise discomfort to participants, the interviews were brief and aimed to achieve an indication of likelihood, rather than a comprehensive diagnosis, of A-ADHD. About 96% of those invited participated in the interviews, which were conducted by clinical psychologists. Interview outcomes (i.e., diagnostic likelihood) were recorded on the survey form, and entered into the database with the survey data, to maintain the anonymity of participants. Individuals with a high likelihood of A-ADHD or with other mental health concerns were offered referral to appropriate clinical service providers.
Procedure
This study employed a survey-type study design. Workers were invited to participate in the study during on-site visits by the research team, as part of existing routine health surveillance programmes. Participants were briefed before being invited to complete the AASS-5, and the return of a completed scale was implied as consent for their data to be used in the analysis (as explained in the briefing). Participation was anonymous (i.e., names were not recorded on the dataset) and was not part of any other work performance evaluation. The brief interviews were conducted on the same day that the scale was completed.
Ethical considerations
This research was part of a larger study and had been approved by the Health Research Ethics Committee of Stellenbosch University (# N18/03/039).
Data analysis
Psychometric properties were explored through a number of statistical techniques. Internal consistency was calculated using Cronbach’s alpha coefficient (for the total scale, as well as for the two subscales). Furthermore, given that the AASS-5 consists of two subscales, a factor analysis was conducted. Predictive validity was calculated by means of a receiver operating characteristic (ROC) curve analysis for the total scale scores and subscale scores. In addition, sensitivity and specificity data were also calculated. Furthermore, chi-square analyses were conducted using the cases identified according to the three scoring systems, and diagnostic cases that were identified by the psychologist interview.
Prevalence estimates were computed for the total sample, as well as per gender, age group (using the categories of de Graaf et al., 2008), and field of work. The outcomes of the three different scoring/interpretation systems were explored by calculating the proportion of participants who would meet the threshold as an A-ADHD case according to each system’s format.
Results
Internal consistency
Favourable internal reliability was found for the full scale (Cronbach α = .881), as well as for the two subscales (AD: α = .826; HA: α = .807). In neither case did any deleted items improve α. Furthermore, there was only a marginal difference between English first language speakers (α = .900) and non-English first language speakers (α = .876).
When conducting a factor analysis, three factors with Eigenvalues > 1 could be extracted (explaining 47% of variance). However, all items loaded primarily on a single factor, and further analysis was considered unproductive and thus discontinued.
Predictive validity
The AASS-5 demonstrated positive predictive validity. The ROC analysis showed highly significant areas under the curve (full scale = .986, AD = .971, HA = .955). The curve is graphically presented in Figure 2. Sensitivity and specificity data can be found in Table 2. Optimal sensitivity and specificity appears to be around a total scale score > 26, which appears similar to previous reports (Psych Central, n.d.).
ROC curve of the AASS-5 and confirmed diagnosis.
Sensitivity and specificity data for the AASS-5 and A-ADHD.
ADHD: attention deficit hyperactivity disorder; AASS: Adult ADHD Symptom Scale.
Chi-square results are presented in Table 3, which demonstrate that all three scoring/interpretation formats had significant predictive validity.
Predictive validity outcomes using chi-square.
ASRS: ADHD Self-Report Scale; ADHD: attention deficit hyperactivity disorder; AASS: Adult ADHD Symptom Scale; DSM-5: Diagnostic and Statistical Manual of Mental Disorders (5th ed.).
Prevalence estimations
Prevalence estimates using the three scoring interpretation systems are presented in Table 1. Using the Set A criteria, the total sample suggested a 1.7% estimated prevalence rate, while the AASS-5 total scale score (using > 33 as cut-off, and with Item 21 indicating symptom interference) suggested a 2.1% estimated prevalence rate. When employing the formal DSM-5 criteria, an estimated prevalence rate of 3.3% was calculated.
There were no significant differences between the estimated prevalence of women and men on any of the three interpretation systems. This was found both when using identified cases (with chi-square) and when using total scale scores (with t tests). On visual inspection, older participants (30–44) indicated a somewhat lower prevalence than the younger participants (19–29), with hyperactivity-impulsivity appearing to account for the greater part of this difference. While there appeared to be large differences in prevalence estimates across different fields of work, unequal group sizes and lack of prospective classification may limit the value of this observation.
Discussion
This sample of full-time employed adults used the AASS-5 to provide self-reported endorsement of symptoms of attention deficit and hyperactivity-impulsivity.
The first aim of the study considered the usefulness of AASS-5 to identify at-risk individuals. In this regard, the AASS-5 demonstrated encouraging psychometric properties, with good internal consistency and predictive validity reported. The instrument, at least in this sample, was able to identify individual cases of likely A-ADHD with high accuracy. Practically, the English scale can be used with some confidence in the South African mixed language environment, provided that basic educational levels are present. Further studies are needed to consider at what level of education, especially where English is not the first language, the scale items become too difficult to elicit accurate self-report of symptoms.
In skilled workplace samples, the AASS-5 could productively be employed within a ‘screen-to-stream’ occupational health approach. Broad-based primary screening in the workplace would allow the streaming of at-risk individuals towards diagnostic and subsequent intervention services. This in turn could benefit both individuals and employers.
Two cautions need to be observed when implementing the AASS-5 for clinical screening or general research, however. First, Item 21, which enquires whether the symptoms interfere with, or reduce the quality of social or work life, was interpreted very differently across the sample. During debriefing, the interviewing psychologists reported that individuals’ responses were not always accurate in terms of the intention of the item. Some participants reported severe symptoms, but did not previously perceive it to interfere with their lives. Others reported very mild symptoms, but described severe reduction in quality of life, often due to other factors. There thus remained a need for face-to-face confirmation to clarify the severity of impairment due to the reported symptoms. Future use of the scale may need to include an expanded description or explanation in Item 21. Second, Item 19, requiring the presence of symptoms prior to age 12, appeared open to retrospective memory bias. During the interviews, older participants reported greater uncertainty of the accuracy of their recall of their childhood years.
The second aim of the study was to provide preliminary prevalence estimates. The different scoring systems presented different estimates. According to the WHO’s Set A criteria, the prevalence estimate of 1.7% was lower that what is typically reported for worker samples generally (i.e., M = 3.5; de Graaf et al., 2008), but similar to reports from low-income countries (i.e., M = 1.9; Fayyad et al., 2007). Although SA is considered an upper-middle income country according to updated World Bank classification (The World Bank, 2018), poor levels of education in the general population and high rates of unemployment are often reported (Statistics South Africa, 2016, 2017). Furthermore, A-ADHD is associated with lower levels of education and employment uptake (Fayyad et al., 2007; Gjervan et al., 2012; Küpper et al., 2012). Therefore, individuals with A-ADHD may not be present in workplace surveys in the numbers that are found in population samples. The current sample had at least 9 years of formal schooling and a relatively high level of skill. The sample may thus have had a bias towards recruitment and retention of individuals who potentially experienced lesser interference from attention deficit or hyperactivity in their occupational performance. The sample may therefore have inherently excluded more severe forms of A-ADHD.
In contrast to the Set A figures, using DSM-5 criteria indicated a prevalence estimation of 3.3%, which although somewhat higher than the original DSM-5 projection (i.e., 2.5%; APA, 2013), was comparable to other worker samples (de Graaf et al., 2008) and the interview findings of Matte et al. (2015) using the same criteria. The difference between the two scoring system outcomes is most likely because of the changed diagnostic threshold in the DSM-5 (Regier et al., 2013).
The incidence of hyperactivity appeared slightly higher than that of attention deficit, which stands in contrast to the published literature (which is mainly based on DSM-IV criteria; Asherson, 2005; Biederman et al., 2000). However, most of the cases of likely A-ADHD presented with combined type, and the actual numbers of the AD or HA subtypes (and differences) are very small. Furthermore, it could be hypothesised that most of the sample were employed in practical fields of work, which may be more accommodating of HA symptoms, and thus allow them to function productively in their workplaces.
Previously reported gender differences (de Graaf et al., 2008; Fayyad et al., 2007) were not observed, which may be a factor of the fields of work covered by the study. The absence of significant gender differences was also observed by Matte et al. (2015), who found differing gender prevalence’s only at reduced diagnostic thresholds.
Furthermore, in contrast to earlier findings using the same age categories (de Graaf et al., 2008; Fayyad et al., 2007), older participants indicated somewhat lower prevalence estimates than the younger participants, with hyperactivity-impulsivity appearing to account for the greater part of the difference. The total estimated prevalence using DSM-5 criteria was almost double for the younger group compared with the older group. Given the skilled work contexts, the age difference could possibly reflect attrition of individuals more severely affected by A-ADHD, in that those people may exit these particular occupational contexts due to difficulty in maintaining required levels of work performance.
The study has a number of limitations. First, the current workforce sample may be too highly educated to generalise results to all SA worker, or community, samples. Second, the process of confirming diagnoses may not have been rigorous enough: the interviews were of necessity brief, and relied on adult self-report, without collateral confirmation. Future studies would need to ensure more equal group sizes to enable meaningful comparisons and include collateral information to confirm the accuracy and severity of self-reported symptoms. It may further be useful to include measures of workplace functioning, to determine the effect of any reported symptoms on work performance.
Conclusion
The positive predictive validity found with this sample suggests that the AASS-5 scale can be used constructively in research or screening contexts to identify self-reported at-risk individuals. While the scale is useful to indicate likelihood of diagnosis, it should never be used to confirm A-ADHD, which remains a clinical diagnosis. Furthermore, preliminary prevalence estimates for A-ADHD, guided by DSM-5 criteria, are now available for a SA workplace sample. It appears that A-ADHD is not that uncommon in this worker sample, which could guide policymakers when considering the allocation of resources for screening and support in the workplace.
Footnotes
Acknowledgements
The author acknowledges the support of Marilize Fourie, Nazneen Firfirey, Jarred Martin, and Ruby Arrenbrecht in conducting the clinical interviews.
Declaration of conflicting interests
The author declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author received no financial support for the research, authorship, and/or publication of this article.
