A Change in Perspective: From Dual Diagnosis to Multimorbidity

National VHA data from fiscal year 2012 included patients with any mental disorder based on International Classification of Diseases, ninth revision, (ICD-9) codes 290–319. Patients who had at least three mental health specialty clinic visits were included in order to represent psychiatrists’ typical caseloads. Three visits were used as a somewhat arbitrary cutoff to identify patients actively involved in treatment, rather than just assessment.

Measures of multimorbidity were chosen among four domains. Direct indicators were based first on simple calculations of the total number of ICD-9 identified psychiatric, substance use, and medical diagnoses out of a set of 10 psychiatric disorder clusters, seven substance use conditions, and 24 medical disorders. [A table listing the 10 psychiatric disorder clusters is available in an online supplement to this report.] Psychiatric diagnostic clusters include posttraumatic stress disorder (PTSD), anxiety disorders, major depressive disorder (MDD), other depressive disorders, bipolar disorder, schizophrenia, and personality disorders. Substance use disorders included dependence on or abuse of any one of seven substances: alcohol, cannabis, cocaine, opioids, sedatives/hypnotics/anxiolytics, amphetamines, and hallucinogens. General medical diagnoses were identified through ICD-9 codes representing common serious chronic medical problems making up the Charlson Comorbidity Index (11), such as hypertension, diabetes mellitus, and chronic obstructive airway disease. These measures were used, in turn, to construct five direct indicators of multimorbidity: the average total number of medical diagnoses or mental disorders (psychiatric or substance use) per veteran, the average number of mental disorders per veteran, the proportion of patients with co-occurring disorders, the proportion with more than one mental disorder or medical diagnosis, and the proportion with more than one mental disorder.

Measures of psychotropic medication complexity include the total number of psychotropic prescription fills during the year for any of six classes of medication, including antidepressants, antipsychotics, benzodiazepines/sedatives/hypnotics, lithium, stimulants, and anticonvulsants/, lithium, stimulants, and anticonvulsants/mood stabilizers [see online supplement]; the average number of classes of psychotropic medications filled per patient; the proportion of patients given a prescription for more than three classes of psychotropic medications; and the proportion given a prescription for at least one opioid medication.

Socioeconomic measures included average income, the proportion with incomes less than $7,000, the proportion with incomes less than $25,000 (the poverty level for an individual income supporting a family of four), and the proportion with a history of recent homelessness, as indicated by receipt of services from a specialized Veterans Affairs (VA) or non-VA homeless service program or a V-60 ICD-9 code.

The final set of indicators addressed general medical indicators, including the proportion with any pain diagnosis (12), obesity (body mass index >30), tobacco use disorder, or a diagnosis of insomnia.

Descriptive analyses were used to compare veterans with any psychiatric or substance use disorder or an index diagnosis of each of four frequent disorders (PTSD, MDD, schizophrenia, and alcohol use disorder), as well as to compare veterans with one diagnosis to those with more than one. For economy of presentation, index psychiatric diagnoses were limited to PTSD, MDD, schizophrenia, and alcohol use disorders because of their relatively high frequency. These index diagnoses were examined to compare differences in multimorbidity indicators among these disorders. However, to be consistent with the multimorbidity paradigm, these index diagnoses were not taken to be the primary, principal, or most important diagnoses. Instead they were considered to be neutral tags with no intended priority over other presenting diagnoses or problems. No statistical tests were used because of the descriptive goals of the study. Additionally, because the sample is very large, statistical significance testing would not be meaningful, because small and clinically insignificant differences may have statistical significance.

The study was approved by the institutional review board committee of the VA Connecticut Healthcare System. A waiver of informed consent was obtained because the study used administrative data and there were no patient identifiers included. All analyses were conducted using SAS statistical software, version 9.2.

The sample included 843,583 veterans with a mental disorder and at least three mental health visits in fiscal year 2012. Among these, 405,592 (48.1%) were diagnosed as having a nonexclusive diagnosis of PTSD, 225,987 (26.8%) had MDD, 64,984 (7.7%) had schizophrenia, and 188,634 (22.4%) had alcohol use disorder (Table 1). The mean±SD age of veterans in the sample was 54.2±14.1 years; 89.7% (N=756,693) were men, 73.4% (N=626,782) were white, 22.0% (N=185,588) were black, and 16.0% (N=134,973) were Hispanic.

Among veterans with any mental disorder, 30.6% were diagnosed as having a co-occurring psychiatric and substance use disorder. By using the new model of multimorbidity, 77.6% were diagnosed as having more than one mental disorder, and each veteran had an average of 2.68±1.52 mental disorders; 94.6% of veterans had more than one mental disorder or medical diagnosis and an average of 4.61±2.27 total psychiatric or medical diagnoses.

On average, veterans received 17.31±27.64 prescriptions from 1.79±1.10 psychotropic medication classes. A total of 25.5% were prescribed medications from more than three classes, and 34.6% received an opioid prescription.

Socioeconomic measures were notable because 64.3% of veterans had incomes below the poverty line for a family of four and 14.8% were recently or currently homeless.

General medical indicators of multimorbidity showed 61.9% had a pain disorder, 45.0% were obese, and 27.0% had a tobacco use disorder. There was limited variation among the indicators for each of the index mental disorders, although there were some notable differences. Veterans with PTSD had a lower rate of past homelessness (9.8%) than those with other index diagnoses.

Among those with MDD, a higher proportion had more than one mental disorder (93.6%). Veterans with alcohol use disorders had the largest proportion with more than one mental disorder (96.3%), the largest number of mental disorders (4.01±1.76), and the largest number of medical diagnoses and mental disorders (5.85±2.49), compared the other three index diagnosis groups. The group with alcohol use disorder also had the highest proportion of veterans with incomes under the poverty level (73.5%), homelessness (31.6%), and tobacco use (45.9%). This group had a somewhat lower proportion of obesity (34.8%) than did the other three groups of index diagnoses.

Veterans diagnosed as having schizophrenia were less likely to have pain disorders (48.5%) and insomnia disorders (4.2%) than those with other index diagnoses. Compared with veterans with one diagnosis, those with more than one diagnosis had substantial increases in all multimorbidity indicators (Table 1).

The landmark National Comorbidity Study Replication (NCS-R) showed that almost one-half of U.S. adults with a mental disorder had more than one diagnosis and one-quarter had three or more diagnoses (3). However, the NCS-R did not address multimorbidity as experienced by clinicians in real-world practice. The study presented here used national VHA data to present multiple indicators of multimorbidity among VHA service users participating in active treatment, as defined as three or more visits during the fiscal year at specialty mental health clinics across the country. The traditional category of co-occurring disorders identified a substantially lower proportion of the patients than were identified using the multimorbidity indicators, and most VHA patients with psychiatric diagnoses had multiple medical and psychiatric diagnoses. Not only did the veterans in our study frequently have direct and indirect indicators of multimorbidity, but they were also prescribed substantial numbers and several different classes of psychotropic medications, as well as opiates; they experienced considerable socioeconomic stress; and they presented with many indicators of medical risk that could interact with psychiatric treatment. It is especially notable that there was limited variation among veterans across the selected index diagnostic categories and that veterans with more than one mental disorder differed from others on all indicators.

Under the multimorbidity paradigm, the challenge emerges of identifying which of the array of disorders and social dysfunctions to focus on while treating patients in a clinical setting. One way of triaging may be to choose the disorder or problem that causes the most suffering or dysfunction, regardless of whether it is psychological distress, vocational disability, relationship conflict, or legal entanglement. Another is to determine whether a single disorder may be driving other problems. Other options are to pick the disorder for which we have the most robust evidence for efficacy of a particular treatment, that costs the patient the least, that takes the shortest amount of time to generate a response, that is most feasible to implement, or that is most favored by the patient. Unfortunately, clinical guidelines, like clinical trials, are often based on studies that exclude comorbid conditions, offering minimal guidance to the real-world tapestry of multimorbidity (13).

The data presented here provide clear evidence from real-world practice that patients are most often more complex than can be captured by any single DSM diagnosis or pair of DSM diagnoses. This is not a new discovery for clinicians, but it is a reminder that focusing on a principal diagnosis may foster the narrowing of vision when applied to clinical practice. In his book How Doctors Think, Jerome Groopman identifies a tendency of physicians to cope with uncertainty by prematurely jumping to diagnostic conclusions based on internal “confirmation biases,” rather than by systematically weighing alternatives for assessment and treatment (14). The multimorbidity perspective reminds us that a shift from a monodiagnostic perspective to a multidimensional, multimorbidity perspective is one of the important challenges and opportunities we face. Billing and other administrative requirements of medical practice may require justification of a plan based on a simplified diagnostic classification, but this should be thought of as a preliminary sketch rather than a conclusive stopping point.

Several methodological limitations of this study deserve comment. The data presented here are based on available administrative VHA data of uncertain validity and do not address non-VA service use by veterans treated by VHA. Furthermore, this study used national VHA data where women and young people are underrepresented, which may limit the generalizability of our findings.

The data reviewed here show that patients receiving psychiatric specialty care are better understood from the multimorbidity perspective, which allows linkage with the medical mainstream while promoting a multifaceted perspective that widens rather than narrows the field of practice.