Skip to main content
PMC home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2014 Apr 30.
Published in final edited form as: Depress Anxiety. 2012 May 14;29(9):797–806. doi: 10.1002/da.21924

Comorbidity and disease burden in the National Comorbidity Survey Replication (NCS-R)

Anne M Gadermann 1, Jordi Alonso 2, Gemma Vilagut 2, Alan M Zaslavsky 1, Ronald C Kessler 1,*
PMCID: PMC4005614  NIHMSID: NIHMS566979  PMID: 22585578

Abstract

Background

Disease burden estimates rarely consider comorbidity. Using a recently developed methodology for integrating information about comorbidity into disease burden estimates, we examined the comparative burdens of 9 mental and 10 chronic physical disorders in the National Comorbidity Survey Replication (NCS-R).

Methods

Face-to-face interviews in a national household sample (n = 5,692) assessed associations of disorders with scores on a visual analog scale (VAS) of perceived health. Multiple regression analysis with interactions for comorbidity was used to estimate these associations. Simulation was used to estimate incremental disorder-specific effects adjusting for comorbidity.

Results

74.9% of respondents reported one or more disorders. 73.8–98.2% of respondents with disorders reported having at least one other disorder. The best-fitting model to predict VAS scores included disorder main effects and interactions for number of disorders. Adjustment for comorbidity reduced individual-level disorder-specific burden estimates substantially, but with considerable between-disorder variation (.07–.69 ratios of disorder-specific estimates with and without adjustment for comorbidity). Four of the five most burdensome disorders at the individual level were mental disorders based on bivariate analyses (panic/agoraphobia, bipolar disorder, PTSD, major depression) but only two based on multivariate analyses, adjusting for comorbidity (panic/agoraphobia, major depression). Neurological disorders, chronic pain conditions, and diabetes were the other most burdensome individual-level disorders. Chronic pain conditions, cardiovascular disorders, arthritis, insomnia, and major depression were the most burdensome societal-level disorders.

Conclusions

Adjustments for comorbidity substantially influence estimates of disease burden, especially those of mental disorders, underlining the importance of including information about comorbidity in studies of mental disorders.

Keywords: bias, burden of illness, confounding factors, cost of illness, epidemiology, global burden of disease, health valuation, visual analog scale (VAS)

INTRODUCTION

Estimates of comparative disease burden are valuable for health policy planning and decision-making.[1–3] Disease burden is typically estimated using a condition-specific severity weight based on expert ratings of the relative burdens of different conditions.[3, 4] However, this approach does not take comorbidity into account and has been severely criticized for failing to do so.[5] This is a nontrivial issue, as methodological research has shown that condition-specific severity weights vary significantly depending on the presence-absence of comorbidity[6] and epidemiological studies show that comorbidity is a pervasive feature of common chronic mental and physical disorders.[7, 8] A statistical approach that allows comorbidity to be taken into consideration in estimating disease burden was recently developed,[9] but has not up to now been used to derive estimates of comparative disease burden for common mental and physical disorders in the US. Such estimates are presented in the current report based on data collected in the US National Comorbidity Survey Replication (NCS-R).[10]

METHODS

The sample

The NCS-R design has been described elsewhere.[10] In brief, the NCS-R is a national face-to-face household survey of adults (ages 18+) designed to study prevalence and correlates of DSM-IV disorders. The survey was administered 2001–3 and had two parts. Part I included an assessment of core DSM-IV mental disorders administered to all respondents (n = 9,282). Part II included questions about mental disorders of secondary interest, physical disorders, disease burden, and other correlates administered to all Part I respondents who met lifetime criteria for any Part I disorder plus a roughly one-in-three probability sub-sample of other Part I respondents (n = 5,692). The Part I-II response rate was 70.9%. The Part II sample, which is the basis of the current report, was weighted to adjust for differential probabilities of selection and under-sampling of Part I respondents without DSM-IV disorders. A weight was also used to adjust for small discrepancies between sample and population Census data on socio-demographic and geographic variables. Verbal informed consent was obtained from participants and these procedures were approved by the Institutional Review Board of Harvard Medical School. A more detailed discussion of NCS-R sampling and weighting procedures is presented elsewhere.[10]

Measures

Mental disorders

Mental disorders were assessed with Version 3.0 of the WHO Composite International Diagnostic Interview (CIDI), a fully-structured interview designed to generate diagnoses of common mental disorders according to the criteria of both the ICD-10 and DSM-IV systems.[11,12] DSM-IV criteria are used here. The nine mental disorders included here are major depressive episode or dysthymia, bipolar disorder I-II, panic/agoraphobia (either panic disorder or agoraphobia without a history of panic disorder), specific phobia, social phobia, generalized anxiety disorder, post-traumatic stress disorder (PTSD), alcohol abuse with or without dependence (i.e., abuse assessed without taking the presence vs. absence of dependence into consideration), and drug abuse with or without dependence. Most of these were assessed in Part I but some (PTSD and, in some countries, substance use disorders) in Part II. WMH clinical reappraisal studies showed that CIDI diagnoses of these disorders have generally good concordance with diagnoses based on blinded clinician-administered reappraisal interviews.[13] We focus on mental disorders present at some time in the 12 months before interview.

Chronic physical disorders

Physical disorders were assessed with a standard chronic conditions checklist based on the list in the US National Health Interview Survey.[14,15] Respondents were asked to report whether they ever had a series of symptom-based conditions (e.g., chronic headaches) and whether a health professional ever told them they had a series of silent conditions (e.g., hypertension) and whether episodic conditions were still present in the 12 months before interview. Methodological research has shown that checklists like this yield more accurate reports than estimates derived from responses to open-ended questions.[16,17] In addition, methodological studies in the US and UK have documented good concordance between such condition reports and medical records.[18–20] The reports were grouped into ten categories to maximize comparability with previous studies[3]: arthritis, cancer, cardiovascular disorders (heart attack, heart disease, hypertension, stroke), chronic pain conditions (chronic back or neck pain, other chronic pain conditions), diabetes, frequent or severe headaches or migraines, chronic insomnia, neurological disorders (multiple sclerosis, Parkinson’s, epilepsy, seizure disorders), digestive disorders (stomach or intestinal ulcer, irritable bowel disorder), and respiratory disorders (seasonal allergies, asthma, COPD, emphysema). We focus on disorders present at any time in the 12 months before interview.

Health valuations

Respondents were asked to make a health valuation after all physical and mental disorders had been assessed by using a 0-to-100 visual analog scale (VAS), where 0 represents the worst possible health a person can have and 100 represents perfect health. The rating task was to rate their own overall health (combining physical and mental health into a single summary evaluation) during the past 30 days. The recall period for the VAS (30-days) is different from the recall period for disorders (12-months) because we wanted to include effects not only of active disorders but also of recent disorders that might still have important effects on health valuations (e.g., a recent heart attack). Previous research has reported high levels of reliability and validity for VAS measures in health valuation studies.[21–24]

Analysis methods

It is important to recognize that health valuation analyses are designed to estimate presumed causal effects of disorders. This is typically done using causal modeling techniques, but based on the realization that it is never possible to justify causal assertions about such effects unequivocally from the non-experimental data that are necessarily used as the basis of such evaluations. Provisional estimates of this sort are nonetheless useful as guides to policymakers in making resource allocation decisions. With that caveat in mind, a series of regression models was used to estimate joint predictive associations of disorders with VAS scores controlling for age and sex.

As the sample was too small to allow each of the logically possible multivariate disorder profiles to be a separate predictor, models necessarily made simplifying assumptions about the appropriate predictors to include in the model to describe comorbidity. The first multivariate model (M1) assumed additivity by including a separate predictor for each disorder without interactions. M2 included a series of predictors for number of disorders (e.g., one predictor for having exactly one disorder, another for exactly two, etc.) without terms for types of disorders. M3 included 19 predictors for types of disorders (i.e., one for each of the 9 mental and 10 physical disorders) in addition to predictors for overall number of disorders. The number-of-disorders dummies in M3 represented aggregate patterns of comorbidity assumed to be independent of types. M4 allowed the effects of type to vary as a linear function of number of other disorders (i.e., 19 dummies for the main effects of disorders, additional dummies for number of disorders, and 19 interaction terms between type and a continuous measure of number of disorders).

The skewed distribution of VAS scores made ordinary least squares (OLS) regression both biased and inefficient. We addressed this problem in two ways. First, a two-part modeling approach[25] was used where a Part I logistic regression equation[26] predicted having a VAS score of 100 versus 0–99 in the total sample and a Part II linear regression equation then predicted scores in the 0–99 range. Individual-level predicted scores were estimated by multiplying predicted values based on the two equations. A problem with this approach is that non-random variance in prediction errors can lead to bias even when sophisticated transformation methods are used.[27] A second approach, generalized linear models (GLMs), addressed that problem by pre-specifying functional forms of associations and error structures in one-part models. Such models can sometimes fit highly skewed data better than two-part models.[28–30]

A number of different one-part and two-part models were considered. We selected the best one using standard empirical model comparison procedures.[31] The GLMs with (i) a square root functional form and independent error structure and (ii) a linear functional form and independent error structure were the two best-fitting models. (Results are not reported, but are available on request.) Due to the simpler interpretation of the linear specification, that was the one used in further analyses. M4, the model that allowed the effects of comorbidity to vary by type of disorder, was the best-fitting model among those considered. This is a model of intermediate complexity in that it allows interactions to vary across disorders but not across particular pairs or higher numbers of disorders. Although this is unlikely to be the optimal interaction model one would arrive at in analyzing a sample of immense size, the fact that it provides the best fit across the range of models considered here suggests that it is a useful first approximation.

A complication in this model, though, is that the coefficients associated with individual disorders have no intuitive interpretation in the presence of interactions. We addressed this problem with a commonly-used simulation approach for models of this sort. This approach began by generating two estimates of predicted VAS scores for each respondent for each simulation. The first estimate was based on the coefficients of the predictors in M4, while the second estimate was based on a revision of this approach that assumed that one particular focal disorder was absent for all respondents. The first estimate of predicted VAS was then subtracted from the second and the sum of differences across respondents was divided by the number of respondents with the focal disorder to estimate the average incremental (i.e., net of all other disorders) individual-level change in VAS scores associated with that disorder. This difference was then interpreted as the incremental effect of the focal disorder (i.e., the effect holding all other disorders constant) in predicting the outcome. The individual-level disorder-specific estimate was then projected to the societal level (i.e., the effect on the mean VAS score) by multiplying it by the disorder prevalence estimate. This process was then repeated for each disorder. This approach generated a single burden estimate for each disorder even when that disorder was involved in a number of significant comorbidity interactions, thus resolving the problem of obtaining unitary disorder-specific estimates of comparative burden in the presence of comorbidity.

Two technical aspects of the model estimation process are noteworthy. First, the simulation approach treats the VAS as an interval scale. This assumption has been called into question[32,33] and nonlinear monotonic transformations have been proposed to approximate interval scale properties.[34] However, strong linear associations have been found between health state values based on VAS scores and ordinal[35] or partially-metric[36] scaling methods. As a result, and given that we explored a number of different nonlinear transformations of the VAS in the GLMs, we treated the VAS as an interval scale in the current analysis. Second, because the NCS-R sample design featured weighting and clustering, all regression analyses used the Taylor series linearization method[37] implemented in the SUDAAN software system[38] to generate design-based standard errors. Standard errors of simulated estimates were obtained using the method of Jackknife Repeated Replications[37] implemented in a SAS macro.[39] Statistical significance was consistently evaluated using two-sided .05 level tests.

RESULTS

Prevalence of disorders and comorbidity

The majority of respondents (74.9%) reported having one or more disorders in the 12 months before interview. Comorbidity was found to be the norm for both mental and physical disorders, as shown by the fact that 73.8–98.2% of respondents with any given disorder reported having at least one other disorder. (Table 1) Co-occurrence of three or more disorders is the typical pattern, as the mean number of additional disorders among respondents with any given disorder is in the range 2.0–4.6. Broad patterns of comorbidity are fairly similar for mental and physical disorders, with medians and inter-quartile ranges (IQR; 25th–75th percentiles) of the percent of cases with comorbidity equal to 93.5% (90.9–95.6%) for mental and 90.3% (85.6–92.1%) for physical disorders. Comparable statistics for mean numbers of other disorders are 3.6 (3.3–4.5) for mental and 2.8 (2.4–3.1) for physical disorders.

Table 1.

Twelve-month prevalence estimates of physical and mental disorders in the Part II NCS-R (n = 5,692)

Prevalence % with at least
one other
disorder		 Mean number of
other disorders
among
those with any
% (SE) % (SE) Est (SE)
I. Physical
  Arthritis 27.3 (0.9) 87.8 (1.1) 2.4 (0.1)
  Cancer 6.6 (0.5) 90.6 (1.8) 2.6 (0.1)
  Cardiovascular disorders 23.8 (0.6) 84.2 (1.6) 2.3 (0.1)
  Chronic pain conditions 22.5 (0.9) 91.9 (1.0) 2.8 (0.1)
  Diabetes 6.1 (0.4) 92.8 (2.2) 2.8 (0.2)
  Digestive disorders 3.3 (0.2) 94.8 (1.6) 3.9 (0.2)
  Headaches or migraines 12.7 (0.6) 89.9 (1.5) 3.1 (0.1)
  Insomnia 14.0 (0.6) 92.1 (1.5) 3.2 (0.1)
  Neurological disorders 1.8 (0.2) 85.6 (5.7) 2.8 (0.3)
  Respiratory disorders 33.7 (1.2) 73.8 (1.4) 2.0 (0.1)
II. Mental
  Alcohol abusea 3.1 (0.3) 83.6 (2.6) 2.7 (0.2)
  Bipolar disorder 2.4 (0.2) 94.6 (1.3) 4.6 (0.2)
  Drug abusea 1.4 (0.2) 91.4 (4.3) 3.1 (0.3)
  Generalized anxiety disorder 2.7 (0.2) 98.2 (0.9) 3.8 (0.2)
  Major depressive episode 8.9 (0.4) 93.5 (1.0) 3.6 (0.2)
  Panic disorder/agoraphobia 3.7 (0.2) 93.7 (2.7) 4.5 (0.2)
  Post-traumatic stress disorder 3.6 (0.3) 96.5 (1.2) 4.4 (0.2)
  Social phobia 7.2 (0.3) 90.4 (1.3) 3.5 (0.1)
  Specific phobia 9.2 (0.5) 91.4 (1.3) 3.4 (0.1)
Open in a new tab
a

With or without dependence

We examined bivariate comorbidity with cross-tabulations. Of the 171 (i.e., 19x18/2) pairs of disorders, 87.7% are positive (median OR = 2.3, range of OR = 1.0–32.0) and 75.3% of these are statistically significant. (Detailed results are not shown, but are available on request.) One-third of the negative associations (median OR = 0.8, range of OR = 0.1–0.9) are statistically significant, virtually all of them involving associations of alcohol-drug abuse with physical disorders.

The distribution of VAS scores

Consistent with previous general population surveys,[40,41] VAS scores are highly skewed. (Detailed results are not reported, but are available on request.) The mean on the 0–100 scale is 82.3 and the standard deviation is 17.5. Fewer than 5% of respondents have VAS scores below 50, while 12.3% have scores of 100 and an additional 12.5% have scores in the range 91–100. The median (IQR) among respondents with scores less than 100 is 85 (75–90).

The individual-level predictive associations of conditions with VAS scores

Bivariate regression models (M0) find all 19 conditions are associated with decrements in perceived health. (Table 2) Four of the five disorders with the strongest associations are mental disorders (panic/agoraphobia, bipolar disorder, PTSD, and major depression), with decrements in VAS scores of 14.0–16.5. The one physical disorder in the top five is digestive disorders, with a decrement of 14.1.

Table 2.

Bivariate and multivariate associations of type-number of disorders with VAS scores (n = 5,692)

M0
Bivariatea 		M1 Multivariate
types of disordersb 		M2 Multivariate
number of disordersc 		M3 Multivariate
types and number
of disordersd
Est (SE) Est (SE) Est (SE) Est (SE)
I. Physical
  Arthritis −9.4* (0.8) −4.5* (0.7) −4.2* (0.9)
  Cancer −1.4 (1.4) 1.0 (1.0) 1.3 (1.0)
  Cardiovascular disorders −9.2* (0.9) −5.4* (0.8) −5.2* (1.0)
  Chronic pain conditions −12.1* (1.0) −6.7* (0.8) −6.2* (0.8)
  Diabetes −10.3* (2.3) −5.9* (1.8) −5.5* (1.7)
  Digestive disorders −14.1* (2.6) −5.0 (2.6) −4.3 (2.6)
  Headaches or migraines −10.8* (0.8) −3.7* (0.7) −3.3* (0.7)
  Insomnia −12.7* (1.3) −6.2* (1.2) −5.6* (1.1)
  Neurological disorders −12.4* (3.4) −8.6* (2.5) −8.2* (2.6)
  Respiratory disorders −3.2* (0.8) −0.4 (0.7) −0.2 (0.7)
II. Mental
  Alcohol abusee −7.3* (1.1) −1.9 (1.2) −1.6 (1.4)
  Bipolar disorder −16.3* (1.5) −3.8* (1.6) −3.2* (1.6)
  Drug abusee −9.7* (1.8) −3.1 (1.9) −2.9 (1.9)
  Generalized anxiety disorder −11.3* (0.9) −3.2* (1.3) −2.6 (1.5)
  Major depressive episode −14.0* (1.1) −5.9* (0.8) −5.5* (0.8)
  Panic disorder/agoraphobia −16.5* (1.4) −5.8* (1.3) −5.3* (1.4)
  Post-traumatic stress disorder −15.0* (1.4) −3.6* (1.2) −3.0* (1.3)
  Social phobia −10.7* (1.1) −2.5* (1.2) −2.1 (1.4)
  Specific phobia −9.8* (0.8) −2.4* (0.8) −2.0* (1.0)
III. Number of disorders
  Exactly 1 −2.4* (0.5)
  Exactly 2 −5.3* (0.9) 1.3 (1.1)
  Exactly 3 −10.3* (0.7) 0.3 (1.2)
  4 or more −21.8* (1.3) −2.5 (2.3)
Open in a new tab
*

Significant at the .05 level, two-sided test.

a

Nineteen models were estimated with one condition at a time.

b

A separate dummy variable predictor for each of the 19 conditions.

c

A separate dummy variable predictor for having exactly one of the 19 conditions, exactly two of the 19 conditions, etc.

d

The predictors in M3 include all those in M1 and M2 with the exception that the predictor for having exactly one disorder is omitted. As a result of this omission, the 19 coefficients associated with types of disorders should be interpreted as applying to pure disorders (i.e., respondents with exactly one disorder compared to those with no disorders).

e

With or without dependence

All coefficients become weaker in the multivariate additive model (M1), but this is especially true for bipolar disorder and PTSD, which are associated with among the highest VAS scores in bivariate models but not in the multivariate model. Further analysis (detailed results are not reported, but are available on request) showed that these extreme reductions for bipolar disorder and PTSD are largely due to strong comorbidities with major depression and panic disorder. Major depression is the only mental disorder that remains among those with the largest decrements in VAS scores (5.9), although panic/agoraphobia has nearly as large a decrement (5.8). Neurological disorders (8.6), chronic pain conditions (6.7), insomnia (6.2), and diabetes (5.9) are the other disorders in the top five. The model that considers only number of disorders (M2) shows a strong monotic association between number of disorders and decrease in VAS scores. However, this model fits the data less well than the additive model that considers only types of disorders (M1) (χ219 = 1784.7, p < .001, AIC = 48,789.9 for M1 vs. χ24 = 399.2, p < .001, AIC = 49,077.6 for M2), a result that is consistent with the fact that an evaluation of slope differences in M1 shows the assumption of equal slopes across disorders implicit in M2 to be inconsistent with the data (χ218 = 344.4, p < .001).

The coefficients associated with number of disorders are no longer significant individually, but are as a set (χ23 = 10.6, p = .014), in the model that includes information about both number and types of disorders (M3). A more complex model found that linear interactions between types and number of disorders are significant as a set (χ218 = 112.5, p < .001), making this the best-fitting model among those considered here (AIC = 48,752.3 vs. 48,777.9–49,077.6 for the other models). However, only three disorder-specific interactions were individually significant out of 18 (associations of arthritis and panic/agoraphobia with VAS increasing in the presence of comorbidity and the association of insomnia with VAS decreasing in the presence of comorbidity). (Detailed results are not reported, but are available on request.) Based on the fact that the vast majority of disorder-specific interactions were not significant in M4, we focused on M3 as our preferred model.

Given the presence of interactions in M3, the coefficients associated with types of disorders in that model have to be interpreted as the associations of pure disorders (i.e., disorders occurring to respondents with no other disorders) with VAS scores (compared to scores of respondents with no disorders). It is instructive to compare these coefficients to those in the bivariate models (M0), as this shows that the associations involving pure disorders are less than half as large as those involving overall disorders for all mental disorders and five physical disorders. These results indicate that comorbidity accounts for the major part of the associations with VAS scores for most disorders considered here. Only one of the five pure disorders with the strongest associations is a mental disorder – major depression, with a VAS decrement of 5.5 – while panic/agoraphobia has a somewhat lower decrement (5.3). The pure physical disorders in the top five are neurological disorders (8.2), chronic pain conditions (6.2), insomnia (5.6), and diabetes (5.5).

The coefficients associated with number of disorders in M3 can be interpreted as non-additive effects of comorbidity. Comorbid clusters made up of exactly two disorders are estimated to have VAS decrements 1.3 less than the sum of the pure-disorder decrements, while the VAS decrements associated with comorbid clusters made up of exactly three disorders are estimated to be 0.3 less than the sum of the pure-disorder decrements. By far the largest non-additive effects of comorbidity, though, are associated with clusters of four or more comorbid disorders, where the decrements are estimated to be 2.5 more than the sum of the pure-disorder decrements. This is referred to as a super-additive interaction. As noted above, these three number-of-disorders coefficients are significant as a set even though none of them is significant individually.

Simulated individual-level associations

Simulations of the disorder-specific associations with VAS scores based on M4 show that all disorders considered here other than cancer are associated with decrements in VAS scores. (Table 3) The median (IQR) disorder-specific VAS decrement is 3.7 (2.5–5.8). Two of the disorders with the largest (top 5) VAS decrements are mental disorders: panic/agoraphobia and major depression. The three others are neurological disorders, chronic pain conditions, and diabetes. Of the 171 differences between pairs of 19 disorders, 29.2% are statistically significant at the .05 level. (Detailed results of these comparisons are not shown, but are available on request.)

Table 3.

Simulated individual-level condition-specific estimates based on the best-fitting regression model

Simulated
estimatea 		Simulated estimate/
Bivariate estimateb
Est (SE) Est
I. Physical
  Arthritis −4.9* (0.7) 0.52
  Cancer 0.9 (1.0) −0.64
  Cardiovascular −5.7* (0.8) 0.62
  Chronic pain −6.8* (0.9) 0.57
  Diabetes −5.8* (1.7) 0.56
  Digestive −4.8 (2.5) 0.34
  Headache or migraine −3.7* (0.7) 0.34
  Insomnia −5.7* (1.2) 0.44
  Neurological −8.5* (2.4) 0.69
  Respiratory disease −0.2 (0.7) 0.07
II. Mental
  Alcohol abusec −1.9 (1.1) 0.26
  Bipolar disorder −3.3 (1.8) 0.20
  Drug abusec −3.0 (1.8) 0.31
  Generalized anxiety disorder −2.9* (1.4) 0.26
  Major depressive episode −6.2* (1.0) 0.45
  Panic disorder/agoraphobia −6.6* (1.5) 0.40
  Posttraumatic stress disorder −3.3* (1.3) 0.22
  Social phobia −2.4* (1.3) 0.22
  Specific phobia −2.5* (0.8) 0.25
Open in a new tab
*

Significant at the .05 level, two-sided test

a

Association of the disorder with mean change in VAS scores among respondents with the disorder net of the other disorders.

b

The ratio of the estimate in the first column divided by the M0 estimate in Table 2.

c

With or without dependence

The effects of comorbidity on these estimates can be seen in the fact that the final condition-specific estimates are, on average, only about one-third as large as the estimates based on the bivariate model, with a median (IQR) ratio of .34 (.24–.49). This reduction is considerably more pronounced for mental disorders [.26 (.22–.31)] than physical disorders [.48 (.34–.57] with the exceptions of the opposite-sign effect for cancer and an especially low ratio for respiratory disorders (.07). The greater reductions for mental than physical disorders are comparable to the pattern found above in the comparison of pure-disorder coefficients with bivariate coefficients. Indeed, the simulated estimates are very similar to, although typically somewhat higher than, the pure disorder coefficients. That the simulated coefficients are higher, not lower, than the pure-disorder coefficients means that the super-additive interaction involving profiles with four or more comorbid disorders overwhelms the sub-additive interactions involving profiles with two or three comorbid disorders.

Simulated societal-level associations

Societal-level estimates of disorder-specific burden were derived by multiplying simulated individual-level estimates by disorder prevalence to arrive at estimated associations of disorders with changes in mean population-level VAS scores. (Table 4) The median (IQR) value of the population-level estimates is .17 (.08–.55). Major depression is the only mental disorder in the top five population-level estimates (0.6). The other four are chronic pain conditions (1.5), cardiovascular disorders (1.4), arthritis (1.3), and insomnia (0.8). These associations are all less than one-tenth of a standard deviation on the VAS scale (which, as noted above, has a standard deviation of 17.5). The top conditions from a societal perspective are dominated by high-prevalence conditions that have intermediate individual-level burden, although two of the top five, chronic pain conditions and major depression, also have high individual-level burdens.

Table 4.

Simulated societal-level condition-specific estimates of burden based on the best-fitting regression model (n = 5,692)

Simulated
estimatea
Est (SE)
I. Physical
  Arthritis −1.3* (0.2)
  Cancer 0.1 (0.1)
  Cardiovascular disorders −1.4* (0.2)
  Chronic pain conditions −1.5* (0.2)
  Diabetes −0.4* (0.1)
  Digestive disorders −0.2 (0.1)
  Headaches or migraines −0.5* (0.1)
  Insomnia −0.8* (0.2)
  Neurological disorders −0.2* (0.1)
  Respiratory disorders −0.1 (0.3)
II. Mental
  Alcohol abuseb −0.1 (0.0)
  Bipolar disorder −0.1 (0.0)
  Drug abuseb −0.0 (0.0)
  Generalized anxiety disorder −0.1 (0.0)
  Major depressive episode −0.6* (0.1)
  Panic disorder/agoraphobia −0.2* (0.1)
  Posttraumatic stress disorder −0.1* (0.1)
  Social phobia −0.2 (0.1)
  Specific phobia −0.2* (0.1)
III. Any condition
  Physical disorders −5.5* (0.5)
  Mental disorders −1.6* (0.1)
  Any disorder −7.0* (0.5)
Open in a new tab
*

Significant at the .05 level, two-sided test

a

Estimated change in the population-level VAS mean associated with eradication of the disorder net of other disorders.

b

With or without dependence

DISCUSSION

The above results are limited in a number of ways. The first limitation is the one noted above as an inherent feature of the health valuation process: that health valuation analyses are designed to estimate presumed causal effects of disorders even though we recognize that causal assertions about such effects cannot be justified unequivocally from the non-experimental data that are necessarily used as the basis of such evaluations. Implicit causal interpretations consequently have to be recognized as provisional guides rather than definitive documentations.

Within the context of that overarching limitation, four more technical limitations of our analysis are noteworthy. First, only a small set of common disorders was included in the analysis. Second, diagnoses of chronic physical disorders were based on self-reports rather than medical records or objective tests. Third, VAS scores were based on evaluations of respondents’ own perceived health, whereas previous studies typically were based on expert VAS ratings of hypothetical illness scenarios. It would be useful for a future study to use parallel methods to obtain ratings based on both self-reports from a representative community sample and independent expert ratings of the illness burdens of that same sample across a wider range of disorders than considered here. For this to be done, though, the expert ratings should rate the disability of the entire patient based on a multivariate disorder profile rather than the disability of a particular condition averaged across patients. Such an approach would allow for a methodological evaluation of the effects of self-ratings versus expert ratings on estimates of disease burden taking comorbidity into consideration. Fourth, information on within-disorder variation in severity was not taken into consideration. The evaluation and analysis of severity are complex issues that lend themselves to no simple solutions, but clearly warrant consideration in future refinements of the methodology of evaluating disease burden.

As noted in the section on assessment, the analysis examined the burdens of 12-month disorders on 30-day health valuations. This discrepancy in time frames was created by design to estimate the current (i.e., past 30 days) effects of recent (i.e., past 12 months) chronic-recurrent conditions that might be either in or out of episode at the time of interview. The extent to which results would have been different if the time frames had been made the same is unclear. The highly skewed distribution of VAS scores and non-additive effects of comorbid conditions might also be seen as limitations in that they could have led to instability of results. Finally, while estimates might be accurate for the overall adult population, comparative ratings might be quite different in particular population subgroups.

In line with previous studies, our results show that comorbidity is the norm among chronic conditions[7, 8, 42, 43] and that the vast majority of the mental and physical disorders considered here are associated with decrements in perceived health.[9] A possible explanation for cancer being the exception is that psychological adaptation might have occurred for individuals with cancer, leading to a recalibration of internal standards for perceived health.[44, 45] We also went beyond previous studies and found that adjusting for comorbidity had an enormous effect on estimates of disorder burden. Indeed, disorder-specific decrements on VAS scores were, on average, only about one-third as large after than before adjusting for comorbidity. Furthermore, the joint effects of comorbid conditions on VAS scores were significantly non-additive, which means that simple linear regression models would have been inappropriate to control for the effects of comorbid conditions in interpreting the coefficients associated with specific disorders. Both of these results are consistent with the cross-national findings of Alonso et al.[9] Although a small number of previous studies also examined effects of comorbidity,[46–48] they focused largely on comorbidity between pairs of disorders. We found, in comparison, that super-additive effects of comorbidity were limited to comorbid clusters involving sets of at least four disorders. Based on this result, future research on comorbidity and illness burden should be especially concerned with this kind of high comorbidity, which has been referred to elsewhere as multimorbidity.[49]

At the individual level, neurological disorders, chronic pain conditions, panic/agoraphobia, major depression, and diabetes were estimated to be associated with the greatest decrements in VAS scores after adjusting for comorbidity. Three of these five (neurological, panic/agoraphobia, major depression) were in the top five across all developed countries in the Alonso et al. analysis,[9] but the other two (i.e., chronic pain conditions and diabetes) are higher in the US than in other developed countries, while two conditions in the top five in the Alonso et al. analysis (insomnia and digestive disorders) were not in the top five in the NCS-R. The neurological disorders considered here included epilepsy and seizure disorders, Parkinson’s disease, and multiple sclerosis, all of which have been associated with high disability in previous studies.[50, 51] Chronic pain conditions, panic/agoraphobia, and major depression have also been associated with high disability in previous studies.[52, 53] The evidence in previous studies is more mixed, though, regarding the burden of diabetes on functioning, which is sometimes estimated to be relatively low[54] and at least partially explained by comorbid mental disorders.[55] It is consequently striking that the results of the current study, which arguably included more thorough controls for comorbidity than any previous study of diabetes, found diabetes to be associated with among the highest decrements in VAS at the individual level. This is especially true in light of evidence that the prevalence of diabetes has reached epidemic proportions in many countries.[56]

With regard to societal-level burdens, the rank-ordering of disorders found here was close to the one found in a range of developed countries by Alonso et al.[9] Indeed, the three disorders ranked as the most severe at the societal level (chronic pain conditions, cardiovascular disorders, and arthritis) were ranked in exactly the same order in both studies. As in previous studies that compared the rank-ordering of individual-level and societal-level estimates,[57–59] the disorders estimated to be most burdensome at the societal level were dominated by high-prevalence conditions with intermediate individual-level severities, although two of the five disorders with the highest societal-level burdens (chronic pain conditions and major depression) also had high individual-level coefficients.

These results underscore the importance of including information about comorbidity in studies of the burden of common mental and physical disorders. This was found to be especially true, though, for mental disorders, arguing against a focus on pure disorders in epidemiological studies designed to evaluate the effects of mental disorders on functioning as well as in studies designed to evaluate the effects of treatment in reducing the impairments associated with mental disorders. Our results also highlight the importance of panic/agoraphobia and major depression as especially important components of the burden of common mental disorders, suggesting that assessments of these syndromes should be included routinely in clinical and epidemiological studies of other mental disorders.

Acknowledgments

The National Comorbidity Survey Replication (NCS-R) is supported by the National Institute of Mental Health (NIMH; U01-MH60220) with supplemental support from the National Institute on Drug Abuse (NIDA), the Substance Abuse and Mental Health Services Administration (SAMHSA), the Robert Wood Johnson Foundation (RWJF; Grant 044780), and the John W. Alden Trust. Additional support for the preparation of this work was provided by the Mental Health Burden Study (NIMH; HHSN271200700030C). Collaborating NCS-R investigators include Ronald C. Kessler (Principal Investigator, Harvard Medical School), Kathleen Merikangas (Co-Principal Investigator, NIMH), James Anthony (Michigan State University), William Eaton (The Johns Hopkins University), Meyer Glantz (NIDA), Doreen Koretz (Harvard University), Jane McLeod (Indiana University), Mark Olfson (New York State Psychiatric Institute, College of Physicians and Surgeons of Columbia University), Harold Pincus (University of Pittsburgh), Greg Simon (Group Health Cooperative), Michael Von Korff (Group Health Cooperative), Philip S. Wang (NIMH), Kenneth Wells (UCLA), Elaine Wethington (Cornell University), and Hans-Ulrich Wittchen (Max Planck Institute of Psychiatry; Technical University of Dresden). The views and opinions expressed in this report are those of the authors and should not be construed to represent the views of any of the sponsoring organizations, agencies, or U.S. Government. A complete list of NCS publications and the full text of all NCS-R instruments can be found at http://www.hcp.med.harvard.edu/ncs. Send correspondence to ncs@hcp.med.harvard.edu.

The NCS-R is carried out in conjunction with the World Health Organization World Mental Health (WMH) Survey Initiative. We thank the staff of the WMH Data Collection and Data Analysis Coordination Centres for assistance with instrumentation, fieldwork, and consultation on data analysis. These activities were supported by the National Institute of Mental Health (R01 MH070884), the John D. and Catherine T. MacArthur Foundation, the Pfizer Foundation, the US Public Health Service (R13-MH066849, R01-MH069864, and R01 DA016558), the Fogarty International Center (FIRCA R03-TW006481), the Pan American Health Organization, Eli Lilly and Company, Ortho-McNeil Pharmaceutical, Inc., GlaxoSmithKline, and Bristol-Myers Squibb. A complete list of WMH publications can be found at http://www.hcp.med.harvard.edu/wmh/.

Footnotes

Disclosure: Dr. Kessler has been a consultant for GlaxoSmithKline Inc., Sanofi-Aventis, Kaiser Permanente, Merck & Co., Inc., Ortho-McNeil Janssen Scientific Affairs, Pfizer Inc., Shire Pharmaceuticals, SRA International, Takeda Global Research & Development, Transcept Pharmaceuticals, Inc., Wyeth-Ayerst, Plus One Health Management; has served on advisory boards for Eli Lilly & Company, Mindsite, Wyeth-Ayerst, Appliance Computing II, Ortho-McNeil Janssen Scientific Affairs, and Plus One Health Management; and has had research support for his epidemiological studies from Analysis Group, Inc., Bristol-Myers Squibb, Eli Lilly & Company, EPI-Q, Ortho-McNeil Janssen Scientific Affairs, Pfizer, Inc., Sanofi-Aventis Gruoupe, Shire US, Inc., GlaxoSmithKline, and Walgreens Co.. He owns stock in Datastat, Inc. The remaining authors declare that no competing interests exist.

REFERENCES

  • 1.Lopez AD, Mathers CD. Inequalities in health status: findings from the 2001 Global Burden of Disease study. In: Matlin S, editor. The Global Forum Update on Research for Health. Volume 4. London: Pro-Brook Publishing Limited; 2007. pp. 163–175. [Google Scholar]
  • 2.Murray CJ, Lopez AD. Evidence-based health policy--lessons from the Global Burden of Disease Study. Science. 1996;274:740–743. doi: 10.1126/science.274.5288.740. [DOI] [PubMed] [Google Scholar]
  • 3.Murray CJL, Lopez AD, Mathers CD, et al. The Global Burden of Disease 2000 Project: Aims, Methods and Data Sources. Geneva: World Health Organization; 2001. [Google Scholar]
  • 4.World Health Organization. The Global Burden of Disease: 2004 Update. Geneva: World Health Organization; 2004. [Google Scholar]
  • 5.Boyd CM, Weiss CO, Halter J, et al. Framework for evaluating disease severity measures in older adults with comorbidity. J Gerontol A Biol Sci Med Sci. 2007;62:286–295. doi: 10.1093/gerona/62.3.286. [DOI] [PubMed] [Google Scholar]
  • 6.Moussavi S, Chatterji S, Verdes E, et al. Depression, chronic diseases, and decrements in health: results from the World Health Surveys. Lancet. 2007;370:851–858. doi: 10.1016/S0140-6736(07)61415-9. [DOI] [PubMed] [Google Scholar]
  • 7.Fortin M, Soubhi H, Hudon C, et al. Multimorbidity's many challenges. BMJ. 2007;334:1016–1017. doi: 10.1136/bmj.39201.463819.2C. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.van den Akker M, Buntinx F, Metsemakers JF, et al. Multimorbidity in general practice: prevalence, incidence, and determinants of co-occurring chronic and recurrent diseases. J Clin Epidemiol. 1998;51:367–375. doi: 10.1016/s0895-4356(97)00306-5. [DOI] [PubMed] [Google Scholar]
  • 9.Alonso J, Vilagut G, Chatterji S, et al. Including information about co-morbidity in estimates of disease burden: results from the World Health Organization World Mental Health Surveys. Psychol Med. 2011;41:873–886. doi: 10.1017/S0033291710001212. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Kessler RC, Berglund P, Chiu WT, et al. The US National Comorbidity Survey Replication (NCS-R): design and field procedures. Int J Methods Psychiatr Res. 2004;13:69–92. doi: 10.1002/mpr.167. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Kessler RC, Üstün TB. The World Mental Health (WMH) Survey Initiative Version of the World Health Organization (WHO) Composite International Diagnostic Interview (CIDI) Int J Methods Psychiatr Res. 2004;13:93–121. doi: 10.1002/mpr.168. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Kessler RC, Üstün TB, editors. The WHO World Mental Health Surveys: Global Perspectives on the Epidemiology of Mental Disorders. New York, NY: Cambridge University Press; 2008. [Google Scholar]
  • 13.Haro JM, Arbabzadeh-Bouchez S, Brugha TS, et al. Concordance of the Composite International Diagnostic Interview Version 3.0 (CIDI 3.0) with standardized clinical assessments in the WHO World Mental Health surveys. Int J Methods Psychiatr Res. 2006;15:167–180. doi: 10.1002/mpr.196. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Center for Disease Control and Prevention. Health, United States, 2004. Atlanta, GA: National Center for Health Statistics; 2004. [Google Scholar]
  • 15.Schoenborn CA, Adams PF, Schiller JS. Summary health statistics for the U.S. population: National Health Interview Survey, 2000. Vital Health Stat. 2003;10:1–83. [PubMed] [Google Scholar]
  • 16.Baker M, Stabile M, Deri C. What do self-reported, objective, measures of health measure? J Hum Resour. 2001;39:1067–1093. [Google Scholar]
  • 17.Knight M, Stewart-Brown S, Fletcher L. Estimating health needs: the impact of a checklist of conditions and quality of life measurement on health information derived from community surveys. J Public Health Med. 2001;23:179–186. doi: 10.1093/pubmed/23.3.179. [DOI] [PubMed] [Google Scholar]
  • 18.Edwards WS, Winn DM, Kurlantzick V, et al. Evaluation of National Health Interview Survey diagnostic reporting. Vital Health Stat. 1994;2:1–116. [PubMed] [Google Scholar]
  • 19.Baker M, Stabile M, Deri C. What Do Self-reported, Objective Measures of Health Measure? Cambridge, MA: National Bureau of Economic Research; 2001. [Google Scholar]
  • 20.Revicki DA, Rentz AM, Dubois D, et al. Gastroparesis Cardinal Symptom Index (GCSI): development and validation of a patient reported assessment of severity of gastroparesis symptoms. Qual Life Res. 2004;13:833–844. doi: 10.1023/B:QURE.0000021689.86296.e4. [DOI] [PubMed] [Google Scholar]
  • 21.de Boer AG, van Lanschot JJ, Stalmeier PF, et al. Is a single-item visual analogue scale as valid, reliable and responsive as multi-item scales in measuring quality of life? Qual Life Res. 2004;13:311–320. doi: 10.1023/B:QURE.0000018499.64574.1f. [DOI] [PubMed] [Google Scholar]
  • 22.Harrison MJ, Boonen A, Tugwell P, et al. Same question, different answers: a comparison of global health assessments using visual analogue scales. Qual Life Res. 2009;18:1285–1292. doi: 10.1007/s11136-009-9546-3. [DOI] [PubMed] [Google Scholar]
  • 23.McCormack HM, Horne DJ, Sheather S. Clinical applications of visual analogue scales: a critical review. Psychol Med. 1988;18:1007–1019. doi: 10.1017/s0033291700009934. [DOI] [PubMed] [Google Scholar]
  • 24.Miller MD, Ferris DG. Measurement of subjective phenomena in primary care research: the Visual Analogue Scale. Fam Pract Res J. 1993;13:15–24. [PubMed] [Google Scholar]
  • 25.Duan N, Manning WG, Morris CN, et al. Choosing between the sample-selection model and the multi-part model. J Bus Econ Stat. 1984;2:289. [Google Scholar]
  • 26.Hosmer DW, Lemeshow S. Applied Logistic Regression. Second Edition. New York, NY: Wiley & Sons; 2001. [Google Scholar]
  • 27.Manning SC. Configuring compliance: a professional fit. J Am Health Inform Manag Assoc. 1998;69:36–38. [PubMed] [Google Scholar]
  • 28.Manning WG, Mullahy J. Estimating log models: to transform or not to transform? J Health Econ. 2001;20:461–494. doi: 10.1016/s0167-6296(01)00086-8. [DOI] [PubMed] [Google Scholar]
  • 29.McCullagh P, Nelder JA. Generalized Linear Models. 2nd Edition. London: Chapman & Hall; 1989. [Google Scholar]
  • 30.Mullahy J. Much ado about two: reconsidering retransformation and the two-part model in health econometrics. J Health Econ. 1998;17:247–281. doi: 10.1016/s0167-6296(98)00030-7. [DOI] [PubMed] [Google Scholar]
  • 31.Buntin MB, Zaslavsky AM. Too much ado about two-part models and transformation? Comparing methods of modeling Medicare expenditures. J Health Econ. 2004;23:525–542. doi: 10.1016/j.jhealeco.2003.10.005. [DOI] [PubMed] [Google Scholar]
  • 32.Krabbe PF, Stalmeier PF, Lamers LM, et al. Testing the interval-level measurement property of multi-item visual analogue scales. Qual Life Res. 2006;15:1651–1661. doi: 10.1007/s11136-006-0027-7. [DOI] [PubMed] [Google Scholar]
  • 33.Parkin D, Devlin N. Is there a case for using visual analogue scale valuations in cost-utility analysis? Health Econ. 2006;15:653–664. doi: 10.1002/hec.1086. [DOI] [PubMed] [Google Scholar]
  • 34.Krabbe PF. Thurstone scaling as a measurement method to quantify subjective health outcomes. Med Care. 2008;46:357–365. doi: 10.1097/MLR.0b013e31815ceca9. [DOI] [PubMed] [Google Scholar]
  • 35.Craig BM, Busschbach JJ, Salomon JA. Modeling ranking, time trade-off, and visual analog scale values for EQ-5D health states: a review and comparison of methods. Med Care. 2009;47:634–641. doi: 10.1097/MLR.0b013e31819432ba. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36.Krabbe PF, Salomon JA, Murray CJ. Quantification of health states with rank-based nonmetric multidimensional scaling. Med Decis Making. 2007;27:395–405. doi: 10.1177/0272989X07302131. [DOI] [PubMed] [Google Scholar]
  • 37.Wolter KM. Introduction to Variance Estimation. New York, NY: Springer-Verlag; 1985. [Google Scholar]
  • 38.Research Triangle Institute. SUDAAN: Professional Software for Survey Data Analysis. Research Triangle Park, NC: Research Triangle Institute; 2002. [Google Scholar]
  • 39.SAS Institute Inc. SAS/STAT® Software, Version 9.1 for Windows. Cary, NC: SAS Institute Inc; 2002. [Google Scholar]
  • 40.Johnson JA, Coons SJ. Comparison of the EQ-5D and SF-12 in an adult US sample. Qual Life Res. 1998;7:155–166. doi: 10.1023/a:1008809610703. [DOI] [PubMed] [Google Scholar]
  • 41.Johnson JA, Luo N, Shaw JW, et al. Valuations of EQ-5D health states: are the United States and United Kingdom different? Med Care. 2005;43:221–228. doi: 10.1097/00005650-200503000-00004. [DOI] [PubMed] [Google Scholar]
  • 42.Fortin M, Bravo G, Hudon C, et al. Prevalence of multimorbidity among adults seen in family practice. Ann Fam Med. 2005;3:223–228. doi: 10.1370/afm.272. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 43.Kessler RC, Greenberg PE, Mickelson KD, et al. The effects of chronic medical conditions on work loss and work cutback. J Occup Environ Med. 2001;43:218–225. doi: 10.1097/00043764-200103000-00009. [DOI] [PubMed] [Google Scholar]
  • 44.Sprangers MA, Schwartz CE. Integrating response shift into health-related quality of life research: a theoretical model. Soc Sci Med. 1999;48:1507–1515. doi: 10.1016/s0277-9536(99)00045-3. [DOI] [PubMed] [Google Scholar]
  • 45.Wensing M, Vingerhoets E, Grol R. Functional status, health problems, age and comorbidity in primary care patients. Qual Life Res. 2001;10:141–148. doi: 10.1023/a:1016705615207. [DOI] [PubMed] [Google Scholar]
  • 46.Fried LP, Bandeen-Roche K, Kasper JD, et al. Association of comorbidity with disability in older women: the Women's Health and Aging Study. J Clin Epidemiol. 1999;52:27–37. doi: 10.1016/s0895-4356(98)00124-3. [DOI] [PubMed] [Google Scholar]
  • 47.Rijken M, van Kerkhof M, Dekker J, et al. Comorbidity of chronic diseases: effects of disease pairs on physical and mental functioning. Qual Life Res. 2005;14:45–55. doi: 10.1007/s11136-004-0616-2. [DOI] [PubMed] [Google Scholar]
  • 48.Verbrugge LM, Lepkowski JM, Imanaka Y. Comorbidity and its impact on disability. Milbank Q. 1989;67:450–484. [PubMed] [Google Scholar]
  • 49.Diederichs C, Berger K, Bartels DB. The measurement of multiple chronic diseases--a systematic review on existing multimorbidity indices. J Gerontol A Biol Sci Med Sci. 2011;66:301–311. doi: 10.1093/gerona/glq208. [DOI] [PubMed] [Google Scholar]
  • 50.Jacoby A, Baker GA. Quality-of-life trajectories in epilepsy: a review of the literature. Epilepsy Behav. 2008;12:557–571. doi: 10.1016/j.yebeh.2007.11.013. [DOI] [PubMed] [Google Scholar]
  • 51.Singer MA, Hopman WM, MacKenzie TA. Physical functioning and mental health in patients with chronic medical conditions. Qual Life Res. 1999;8:687–691. doi: 10.1023/a:1008917016998. [DOI] [PubMed] [Google Scholar]
  • 52.Ormel J, Petukhova M, Chatterji S, et al. Disability and treatment of specific mental and physical disorders across the world. Br J Psychiatry. 2008;192:368–375. doi: 10.1192/bjp.bp.107.039107. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 53.Ormel J, VonKorff M, Ustun TB, et al. Common mental disorders and disability across cultures. Results from the WHO Collaborative Study on Psychological Problems in General Health Care. JAMA. 1994;272:1741–1748. doi: 10.1001/jama.272.22.1741. [DOI] [PubMed] [Google Scholar]
  • 54.Murray CJ, Lopez AD. Global mortality, disability, and the contribution of risk factors: Global Burden of Disease Study. Lancet. 1997;349:1436–1442. doi: 10.1016/S0140-6736(96)07495-8. [DOI] [PubMed] [Google Scholar]
  • 55.Das-Munshi J, Stewart R, Ismail K, et al. Diabetes, common mental disorders, and disability: findings from the UK National Psychiatric Morbidity Survey. Psychosom Med. 2007;69:543–550. doi: 10.1097/PSY.0b013e3180cc3062. [DOI] [PubMed] [Google Scholar]
  • 56.King H, Aubert RE, Herman WH. Global burden of diabetes: 1995–2025: prevalence, numerical estimates, and projections. Diabetes Care. 1998;21:1414–1431. doi: 10.2337/diacare.21.9.1414. [DOI] [PubMed] [Google Scholar]
  • 57.Andlin-Sobocki P, Jonsson B, Wittchen HU, et al. Cost of disorders of the brain in Europe. Eur J Neurol. 2005;12(Suppl 1):1–27. doi: 10.1111/j.1468-1331.2005.01202.x. [DOI] [PubMed] [Google Scholar]
  • 58.Saarni SI, Suvisaari J, Sintonen H, et al. Impact of psychiatric disorders on health-related quality of life: general population survey. Br J Psychiatry. 2007;190:326–332. doi: 10.1192/bjp.bp.106.025106. [DOI] [PubMed] [Google Scholar]
  • 59.Whiteford H. Unmet need: a challenge for governments. In: Andrews G, Henderson S, editors. Unmet Need in Psychiatry: Problems, Resources, Responses. Cambridge, UK: Cambridge University Press; 2000. pp. 8–10. [Google Scholar]

RESOURCES