Higher Prevalence Estimates of Substance Use Disorders With DSM-5 Versus DSM-IV Criteria Among U.S. Nonelderly Adults With Substance Use: The Role of DSM-IV Diagnostic Orphans

In this issue of the Journal, Compton and colleagues have investigated an important issue, namely, whether national rates of past-year adult substance use disorders (SUD) were higher using DSM-IV or DSM-5 criteria, and if so, what changes in DSM-5 criteria accounted for the differences (1). A particular focus was whether the addition of a new criterion, craving, to DSM-5 was responsible for the increased rates. The Compton et al. study has many strengths. The significance of understanding differences in U.S. national prevalences of SUD using DSM-IV or DSM-5 criteria was made clear. The sample of 18–64-year old household residents was large and nationally representative. Four substances (alcohol, cannabis, cocaine and misused prescription opioids) were included. The year of the data collection (2021) enabled a much more recent look at the issue than several earlier studies that showed little difference in rates between DSM-IV and DSM-5 criteria (2).

Compton and colleagues found that prevalences of alcohol, cannabis, prescription opioids, and cocaine-related SUDs were 1.6–2.6 times higher with DSM-5 criteria than with DSM-IV criteria (DSM-IV abuse or DSM-IV dependence). Removing craving from the DSM-5 criteria set while keeping the same diagnostic thresholds reduced the differences in rates between DSM-IV and DSM-5. However, prevalence estimates of DSM-5 alcohol, cannabis, prescription opioids, and cocaine-related SUDs remained 1.3–2.1 times higher than corresponding DSM-IV SUD prevalence. Therefore, the addition of craving to the DSM-5 SUD criteria contributed to the differences in rates but did not explain the differences entirely. Compton and colleagues also found that omitting the craving criterion from the DSM-5 SUD criteria set resulted in loss of many mild DSM-5 SUD cases but very few moderate to severe ones. The authors concluded that relative to DSM-IV, DSM-5 detects mild cases that might benefit from prevention and early intervention.

Many earlier studies comparing within-subject diagnostic results using different nomenclatures calculated the level of within-participant agreement (kappa) on the diagnoses, including studies of DSM-IV and DSM-5 SUD (3–5). We therefore computed the kappas (chance-corrected agreement) for DSM-IV and DSM-5 SUD diagnoses in the 2021 NSDUH public use data set within the same age range as Compton et al. As shown in Table 1, agreement between DSM-IV and DSM-5 SUD diagnoses ranged from fair to excellent. Agreement between DSM-IV and DSM-5 SUD diagnoses was poorest with the DSM-5 threshold of 2+ criteria, which includes the mild as well as moderate and severe DSM-5 cases. Agreement between DSM-IV and DSM-5 SUD diagnoses was best across all four substances with the DSM-5 threshold of 4+, which excludes the mild cases. This suggests that cases that were mild in DSM-5 (two or three criteria) did not receive a SUD diagnosis when using DSM-IV criteria, consistent with the findings of Compton et al.

To better understand the SUD cases that were missed by DSM-IV, we went back to the reasons for the changes between DSM-IV and DSM-5. A major change was to move from a two-disorder framework in DSM-IV (abuse and dependence) to a single-disorder framework in DSM-5. One of the main reasons for this change was that in DSM-IV, individuals could have one or two dependence criteria but no abuse criteria and therefore no SUD diagnosis. These individuals were termed “diagnostic orphans” (6–8), a term indicating no DSM-IV diagnosis among symptomatic individuals. Previously, diagnostic orphans accounted for much of the discrepancy when DSM-5 SUD rates were higher than DSM-IV, including samples from large genetics studies (4), clinical settings (5), primary care patients in the Veterans Administration (9), and adolescents (3). Because the DSM-5 rates remained 1.3–2.1 times higher in the Compton et al. study even after craving was removed from the criteria set, we explored whether diagnostic orphans (symptomatic individuals with no DSM-IV SUD diagnosis) accounted for the remaining discrepancies in rates between DSM-IV and DSM-5 SUD. Accordingly, among NSDUH 2021 participants ages 18–64, we identified the cases of DSM-5 alcohol, cannabis, cocaine, and prescription opioid use disorders that were undiagnosed in DSM-IV and found the following.

These findings for AUD, CUD, POUD, and CoUD indicate that the DSM-5 transition to a single-disorder SUD framework of graded severity accomplished an important goal, namely, providing a way to diagnose SUDs among individuals who were symptomatic but undiagnosed “orphans” in DSM-IV. By capturing these individuals diagnostically, DSM-5 rates were increased and the remaining discrepancies between DSM-IV and DSM-5 SUD rates not accounted for by the addition of the craving criterion were explained.

Several future directions are possible for this area of research. First, participants older than 65 years of age were omitted from all analyses. While this is an age group with historically low rates of substance use, these rates have recently been increasing. Therefore, comparisons of DSM-IV and DSM-5 SUD rates in older adults are needed, especially since SUD poses additional risks in older adults (e.g., falls, adverse interactions with medications for medical problems). Further information on discrepancies in rates is also needed by sex, race/ethnicity, by psychiatric comorbidity and by comorbidity with chronic pain, which can complicate the diagnosis of substance use disorders (10).