Osteoporosis International

, Volume 22, Issue 4, pp 1133–1143

The impact of substance abuse on osteoporosis screening and risk of osteoporosis in women with psychotic disorders


    • Maryland Psychiatric Research CenterUniversity of Maryland School of Medicine
  • C. S. Myers
    • National Institute on Drug Abuse, Intramural Research Program
  • M. T. Abrams
    • The Hilltop Institute, University of Maryland, Baltimore County
  • S. Feldman
    • Maryland Psychiatric Research CenterUniversity of Maryland School of Medicine
  • J. Park
    • The Hilltop Institute, University of Maryland, Baltimore County
  • R. P. McMahon
    • Maryland Psychiatric Research CenterUniversity of Maryland School of Medicine
  • J.-C. Shim
    • Department of Psychiatry and Clinical Trial Center, Busan Paik HospitalInje University
Original Article

DOI: 10.1007/s00198-010-1294-2

Cite this article as:
Kelly, D.L., Myers, C.S., Abrams, M.T. et al. Osteoporos Int (2011) 22: 1133. doi:10.1007/s00198-010-1294-2



Review of the 1-year prevalence of screening for osteoporosis and of osteoporosis or idiopathic fracture in Maryland Medicaid administrative records found that screening rates did not differ among women in the control population, women with psychosis, and women with major mood disorders, but were reduced compared to controls in women with substance use disorder, with or without psychosis. Prevalence of osteoporosis was increased compared to controls in women with major mood disorders or women over 55 dually diagnosed with psychosis and substance use disorder.


Osteoporosis is a major public health concern. Substance abuse and psychosis may be risk factors, however, frequency of screening and disease risk in women with psychotic disorders and substance use disorder (SUD) remains unknown.


This study examined rates (FY 2005) of osteoporosis screening and disease risk in Medicaid enrolled women aged 50 to 64 (N = 18,953). Four diagnostic groups were characterized: (1) psychosis, (2) SUD, (3) major mood disorder, and (4) controls. The interaction of psychosis and SUD on screening and disease prevalence of osteoporosis was tested.


The prevalence of osteoporosis across the entire population was 6.7%. Four percent of those without an osteoporosis diagnosis received osteoporosis screening with no notable differences between psychosis and controls. Those with SUD, however, had a significant reduction in screening compared to controls (OR = 0.61, 95% CI = 0.40–0.91, p = 0.016). Women with a major mood disorder were more likely to have osteoporosis in their administrative record (OR = 1.32, 95% CI = 1.03–1.70, p = 0.028) compared to controls. Those who were dually diagnosed (SUD and psychosis) in the oldest ages (55–64 years) had a markedly higher prevalence of osteoporosis compared to controls (OR = 6.4 CI = 1.51–27.6, p = 0.012), whereas this interaction (SUD and psychosis) was not significant in the entire population over age 49.


Osteoporosis screening in the Medicaid population is significantly lower for women with SUD, after adjusting for age, race, and Medicaid enrollment category. The prevalence of osteoporosis appears markedly elevated in those with major mood disorders and those over age 55 dually diagnosed with schizophrenia and SUD.


Bone mineral densityMedicaidOsteoporosisPsychosisSchizophreniaScreeningSubstance abuse


People who suffer from schizophrenia and other psychotic disorders have higher co-occurring physical disorders than the general population, with approximately 70% of people with psychosis having at least one medical comorbidity and 33% having three or more comorbid health disorders [1]. Dixon et al. [2] also reported that in people with schizophrenia, a greater number of current medical problems contributed to worse perceived physical health status, more severe psychiatric symptoms, and greater likelihood of a history of a suicide attempt. People with psychotic disorders who abuse substances are at an even greater risk of general physical health problems [3]. Approximately 50% of people with schizophrenia have substance abuse disorders [4], and Mackell et al. [5] reported that substance abuse was one of the most significant predictors of poor physical health in people with schizophrenia. They found that preventive care for physical health is suboptimal in psychotic disorders and that those who receive less physical health care and substance abuse treatment reported the lowest scores on quality of life measures.

Women with schizophrenia and other psychotic disorders represent an underserved population with different treatment and physical health needs than men with these disorders. For example, women generally present with psychiatric symptoms at an older age, may have differences in metabolism of some medications [6], have more depressive symptoms, and often have different side effects with pharmacologic treatment as compared to men such as increased prolactin from dopamine antagonists [7]. Additionally, compared to men, women have distinct physical health needs, including gynecologic, fertility and reproductive health care, hormonal changes associated with menopause, and higher rates of breast cancer and osteoporosis [6]. In a medical system in which general physical health care, care for health issues specific to women, mental health care, and substance abuse treatment have remained fragmented, physical health treatment for persons with schizophrenia—particularly gender specific care for women—is often overlooked.

Osteoporosis and bone mineral density loss are significant physical health issues whose prevalence is highest in postmenopausal women. Osteoporosis and associated fractures are a significant public health concern because of related mortality, morbidity, disability, and diminished quality of life [810]. According to the National Osteoporosis Risk Assessment survey, 5–20% of women aged 50 years and older have osteoporosis [8]. Osteoporosis is about four times more prevalent in women as compared to men [911]. People with schizophrenia are considered a high-risk group for developing bone loss and osteoporosis, due in part to the illness itself and to multiple associated risk factors such as poor diet, low exercise, early menopause, and antipsychotic associated hyperprolactinemia [1216]. There is also evidence that osteoporosis and bone loss is greater in people who are smokers and who are substance abusing, both common in persons with psychotic disorders [17]. Despite the many risk factors for osteoporosis present in women with psychosis, the frequency of osteoporosis screening in this population has not been studied nor have there been studies of how the comorbidity of substance abuse effects screening and the risk of osteoporosis in this population [15].

This study was carried out using administrative Medicaid data from the State of Maryland to assess the impact of psychotic disorders and/or substance abuse on the rate of screening for osteoporosis or bone mineral density changes and the prevalence of osteoporosis in women with schizophrenia and other psychotic disorders compared to women without such disorders.

The primary aims of this study were: (1) to examine the rates of screening for osteoporosis in a Medicaid population among women with psychotic disorders, substance use disorders and nonpsychotic major mood disorders, compared to controls without substance use disorder and (2) to assess the prevalence of osteoporosis among the same groups of women.


Sample selection

This study included a cross-sectional sample from State of Maryland Medicaid database records for fiscal year (FY) 2005 (July 2004–June 2005), a recent representative year prior to the initiation of Medicare part D. This administrative sample included Children’s Health Insurance Program (CHIP) data from the State of Maryland. Medicaid/CHIP (hereafter: Medicaid) is an important context in which to evaluate somatic and coordination of care issues for women and persons with severe mental illness because the program covers disproportionate shares of both populations compared to other insurance plans [18].

This study included women between the ages of 50 and 64 years with 12 months of continuous Medicaid enrollment and excluded women who were receiving Medicaid benefits for pregnancy and family planning services alone. Selection of the main study population from Medicaid records is summarized in Table 1. For this study, the presence of psychosis was defined by International Classification of Diseases (ICD) 9 codes 295, 297, or 298, and the presence of substance use disorder (SUD) was defined by the following codes: ICD-9 codes 291–2, 303–5 (except 305.1), 760.7, 965.0, 967, 968.5, 970.0, and 980.0 or Diagnosis Related Group codes 433–437 and procedural codes (local and national) for identifying methadone treatment or other outpatient substance abuse therapy. Major mood disorder was defined by ICD-9 codes 296 or 300.4 without the occurrence of psychosis or SUD; and finally, controls were identified as having no psychotic or major mood disorder and no record of SUD. From these definitions, each woman in the study could be assigned to one or both of the following diagnoses: psychosis or SUD or in the absence of those major diagnostic labels, the remaining subjects were placed either in a major mood disorder category or a control group (i.e., a group without psychosis, SUD, or major mood disorders). These diagnostic labels (and all of the other variables reviewed for this study) were assigned using FY 2005 Medicaid records only; no other years or other sources of administrative or clinical data were reviewed.
Table 1

Medicaid study population for fiscal year 2005a



All Medicaid enrollees


Women only


12 months of Medicaid enrollment


Age 19–64 years


Exclude pregnant womenb


Exclude pharmacy assistancec


Exclude family planningc


Older adults: age 50 to 64 years


Those with no record of osteoporosisd in the study year


aThe fiscal year runs from July 2004 to June 2005

bAny indication of pregnancy during the year led to an exclusion

cLimited Medicaid benefits

dOsteoporosis exclusions = 1,218; and pathologic bone fracture in absence of an osteoporosis diagnosis (ICD-9 = 733.1) exclusions = 56. Absence of an osteoporosis diagnostic record for women with such bone fractures may reflect osteoporosis-related or some other pathology; it is included here to increase sensitivity regarding the detection of osteoporosis, though it simultaneously decreases specificity for that disease.

Screening and diagnostic definitions for osteoporosis

Screening events for osteoporosis were defined by the presence of Healthcare Common Procedure Coding System, Current Procedural Terminology, or ICD-9 procedure codes specific to dual energy X-ray absorptiometry (DXA) or other bone density assessment procedures or osteoporosis screening specifically (code lists appear as footnote to Table 2). The diagnosis for osteoporosis is based on the presence of at least one claim during the year with the ICD-9 code (733.0) for that disease. That claim could occur anywhere in the person’s administrative medical record and need not be the incident diagnosis, e.g., it may alternatively represent a secondary report by a primary care physician delivering follow-up or other care. To increase sensitivity to unlabeled osteoporosis, we further isolated persons who during the year had experienced a bone fracture of pathologic origin (ICD-9 code = 733.1) suggesting possible bone disease that could include osteoporosis. Because the latter may not be specific to osteoporosis, this proxy indicator is used only in sensitivity analyses.
Table 2

Characteristics and prevalence of osteoporosis screening or osteoporosis by diagnosis groupa, Medicaid women age 50–64 years (n = 18,953)b


PSY (n = 1,898)

PSY/SUDb (n = 245)

SUD (n = 1,121)

MMD (n = 704)

Control (n = 14,985)

Age, years (mean, s.d.)

56 (n)

4.3 (%)

54 (n)

3.7 (%)

55 (n)

3.8 (%)

56 (n)

4.1 (%)

57 (n)

4.3 (%)















































 Long-term care











 Dual eligible











 Families and children






















Screening and diagnoses

 Received osteoporosis screeningd











 Osteoporosis (ICD-9 = 733.0)











 Pathologic bone fracture/no record of osteoporosis (ICD-9 = 733.1)











Annual counts and rates reported for 2005

PSY psychosis only, SUD substance use disorder, PSY/SUD both psychosis and substance use disorder

aExcludes women receiving pregnancy/Family planning services only

bPSY/SUD group was not included as a primary group in the analysis but is shown separately here for descriptive purposes

cEighty-nine percent in this category are women with cervical or breast cancer, the remaining few are aliens with emergencies (4%) and persons receiving partial benefits from both Medicare and Medicaid (7%)

dCurrent Procedure Terminology code in the following list: 76075, 76076, 77080, 77079, 76070, 76071, 76078, 76977, 77075, 77076, 77078, 77079, 77083, 78350, 78351, and 77080–77083 or ICD-9 code is 88.98 or V82.81 or Healthcare Common Procedure Coding System Code = G0130

Statistical analysis

Regression modeling

Logistic regressions [19, 20] were used to examine the odds of osteoporosis screening (or of the disease itself in separate modeling) during Maryland’s fiscal year 2005 (July 2004 through June 2005) as a function of diagnostic grouping, controlling for age, race, and Medicaid eligibility category. The main statistical model used for this investigation was as follows:
$$ {\hbox{Ln}}\left( {{\hbox{odds}}\;{\hbox{of}}\;{\hbox{screening}}\;{\hbox{or}}\;{\hbox{disease}}} \right) = {\beta_0} + {\beta_1} \times {\hbox{Psychosis}} + {\beta_2} \times {\hbox{SUD}} + {\beta_3} \times {\hbox{Mood}} + {\beta_4} \times {\hbox{Control}} + {\beta_5} \times {\hbox{Age}} + {\beta_6} \times {\hbox{Race}} + {\beta_7} \times {\hbox{Elig}} + \varepsilon $$
Beta coefficients 1 through 4 captured the individual diagnostic effects using binary flags for the presence or absence of each diagnosis with potential overlap (simultaneity) between psychosis and SUD. Equation 1 presumes that the effects of co-occurring psychosis and SUD are additive, an assumption evaluated by testing whether adding a psychosis × SUD interaction term contributes significance to model fit. Use of the interaction term allowed direct assessment of how the common comorbidity between psychosis and SUD altered the prevalence of screening (or osteoporosis). The presence of a significant interaction term in the regression modeling conducted here is consistent with a synergistic (i.e., more than just additive) effect between comorbid psychosis and SUD. Absence of such a significant interaction term is instead consistent with individual psychosis and SUD effects that are simply additive when they co-occur. An alternative model specification was also tested as follows:
$$ {\hbox{Ln}}\left( {{\hbox{odds}}\;{\hbox{of}}\;{\hbox{screening}}\;{\hbox{or}}\;{\hbox{disease}}} \right) = {\beta_0} + {\beta_1}\times{\hbox{Diagnosis}}\;{\hbox{Category}} + {\beta_2} \times {\hbox{Age}} + {\beta_3} \times {\hbox{Race}} + {\beta_4} \times {\hbox{Elig}} + \varepsilon $$

Parameters in Eq. 2 are as they are in Eq. 1 except that β1 represents the effect of categorical changes among five mutually exclusive diagnostic groupings (i.e., psychosis only, psychosis and SUD together, SUD alone, and mood disorder; with controls as the referent). For this alternative specification, the psychosis and SUD together group is analogous to the interaction term for model, but this latter model has the disadvantage of not testing whether the In(odds) for cases with both psychosis and SUD is different from the sum of the individual effects of psychosis and SUD.

Both of the dependent variables (odds of osteoporosis screening or disease) were analyzed with separate regression models. Both were dichotomous, i.e., they reflect either their presence or absence of the variable of interest. Modeling for both dependent variables used the same populations except for the fact that persons with osteoporosis (n = 1,218), or those with osteoporosis or idiopathic bone fracture (n = 1,264), were excluded from the screening analyses so as to focus on primary prevention. These models assumed a linear effect of age in years, which was treated as a continuous variable; models incorporating a quadratic (age-squared) term for age were also evaluated to test whether the estimated effects of diagnostic grouping were potentially altered by how the adjustment for age was specified.

Additional modeling

Post menopausal (age 55–64)

Similar follow-up analyses examined the age group of 55–64 years—the oldest decade in which large numbers of women rely on Medicaid (rather than Medicare) as their principal source of health care coverage—in order to focus upon a slightly more homogenous sample in which the percentage of menopausal women will be increased (compared to the full sample ages 50–64 years).

Pathologic bone fracture

Similar follow-up analyses also were carried out by altering our definition of osteoporosis to include pathologic bone fracture in the absence of any diagnostic record of osteoporosis. Though such inclusion decreases specificity (i.e., it may mistakenly capture other disease such as bone cancer), it simultaneously may increase our sensitivity to cases of osteoporosis that otherwise may not be visible in the Medicaid record. This expanded definition of osteoporosis increased our osteoporosis case counts by 4.6% (increased n by 56 cases).

Defining race and Medicaid eligibility

Subject race as recorded in the administrative record was grouped in three mutually exclusive categories (Black, White, or “Other”, the latter group including those whose race was unrecorded). White was used as the reference category for logistic regression modeling.

The eligibility variable was composed of five mutually exclusive groups: families and children (temporary cash assistance or medically needy categories), dually eligible (enrolled dually in Medicaid and Medicare), long-term care (extended facility-based care for aged, blind, disabled groups), disabled (as determined by state or federal programs), and all other enrollment categories (other). The families and children category was used as the reference for the regression models. The other category was small (n = 160), but of potential importance as it is composed mostly (89%) of women enrolled in Medicaid because they have a history of breast or cervical cancer, conditions with treatments that increase the risk for osteoporosis [21]. The Medicaid eligibility variable is an important covariate in these analyses as a proxy of general health status: for example, individuals are classified as disabled due to severe morbidity (often from mental illness). The status of dual eligibility for Medicare and Medicaid is distinctive from the other eligibility categories because in addition to marking disability, it further marks limited Medicaid coverage levels. Specifically, part B of the Medicare program is likely to cover the majority of preventive screening for osteoporosis and thus the Medicaid record (the source of data for this review) would not reveal such screening.

All analyses were carried out using SAS® 9.1 (Cary, NC, USA). p values <0.05 were considered significant. This study was submitted to the National Institute on Drug Abuse Institutional Review Board (IRB), the University of Maryland IRB, and the Department of Health and Mental Hygiene (DHMH) IRBs and determined exempt due to the nature of the existing and deidentified data.


Descriptive data

The sample is presented and described in Table 2. The overall observed prevalence of psychosis is 11.3% (2,143/18,953), higher than the general population [22], but not unexpected for an older Medicaid population with continuous enrollment [23]. The fact that the population has continuous enrollment likely correlates with chronic disability status that can result from a serious mental disorder. The fact that they are older also means they are more likely to be engaged in Medicaid because of chronic disability since the program otherwise favors enrollment for pregnant women and children. The prevalence of a major mood disorder (3.7%, 704/18,953) is likely lower because it, per se, does not carry the same level of morbidity and inability to work as severe psychoses. Overall SUD prevalence in the population was 7.2% (1,366/18,953). If limited to those with psychotic disorders, only the prevalence of SUD was 11.4% (245/2,143), and for those without psychosis or major mood disorder the prevalence was 6.9% (1,121/16,106).

Screening rates for osteoporosis

Overall 6.7% of the full sample of 18,953 was screened for osteoporosis during FY 2005, but only 4.0% of the sample without an osteoporosis diagnosis or a pathologic bone fracture (n = 17,679) was screened during this time period. To focus upon true primary prevention (e.g., asymptomatic) screening rates, the sample without osteoporosis is the principal one reviewed with regard to screening. Logistic regression modeling for osteoporosis screening (see Table 3) demonstrated significant results (p < 0.0001), but fit statistics indicated only slightly above 1% of the variance was characterized by this modeling. This model found that the prevalence of screening was not lower in those with psychosis (p = 0.60) or major mood disorders (p = 0.86) compared to the control group, and no significant interaction was observed between psychosis and SUD (p = 0.50). However, people who have a SUD (with or without co-occurring psychosis) were 39% less likely than controls to be screened for osteoporosis (OR = 0.61, 95% CI = 0.40–0.91, p = 0.016) after adjusting for the covariates in this model. Those who are black were less likely to be screened than white subjects (OR = 0.67, 95% CI = 0.57–0.79, p < 0.0001). Patients enrolled in Medicaid eligibility as dually eligible, disabled, or “Other” were likely to be screened at higher rates than those in the families and children categories. The effect of age (or age2) was not statistically significant, possibly because of the restricted age range included in the sample, whether age was modeled with a linear or quadratic term did not substantially alter the results regarding relative risks (OR) comparing the diagnostic groups. Excluding the 56 persons with pathologic bone fracture from the models for osteoporosis screening did not change the results in any remarkable way.
Table 3

Logistic model results: predictors of osteoporosis screening, Medicaid women age 50–64 yearsa (n = 17,679), those with osteoporosis excluded from sample; screening rate = 4.2%



95% CI

Chi-square (df = 1)

p value

Diagnosis (OR relative to controls)






 Substance use disorder





 Major mood disorder





Race (OR relative to white)











Eligibility (OR relative to families and children)

 Long-term care




0. 27

 Dual eligible















Age (OR per one year increase in age)b





Psychosis × SUD interaction vs. control (OR = 1.75, 95% CI = 0.34–8.9; chi-square, 0.46; p = 0.50). Accordingly, the model presented here does not include this parameter

aExcludes women receiving pregnancy/family planning services only

bUsing age2 did not change this result

Mutually exclusive modeling

The mutually exclusive diagnostic group analysis (Eq. 2 in the methods) was similar regarding overall fit and individual effect sizes and probabilities to the analyses presented in Table 3, showing that the only diagnostic group different from the controls was the SUD group alone, for whom the odds of screening were decreased (OR = 0.58, 95% CI = 0.36–0.88, p = 0.011). The group with both psychosis and SUD did not demonstrate significantly reduced screening rates, though the point estimate was consistent with reduced odds compared to controls (OR = 0.82, 95% CI = 0.38–1.75, p = 0.61).

Analysis of ages 55–64 (postmenopausal)

For the subset of women with ages 55–64 years old (n = 11,237), odds ratios for diagnosis group were similar to those in the full 50–64-year-old sample, but none of the diagnostic group effects, including that for SUD (OR = 0.65, CI = 0.37–1.15, p = 0.14), were statistically significant. Other findings were analogous to previously described models except that the disabled and dual eligibility categories did not differ significantly from other eligibility groups (ps > 0.27).

Prevalence of osteoporosis

Overall osteoporosis/pathologic bone fracture prevalence was 6.7% during FY 2005, whereas the diagnosis of osteoporosis alone was found in a slightly smaller proportion of women included in this study (6.4%). Table 4 summarizes the logistic regression results using the osteoporosis diagnosis alone as the binary outcome of interest. The overall model was significant (p < 0.0001), but fit statistics indicated that only slightly above 4% of the variance was accounted for by this main model. This modeling revealed similar rates of osteoporosis between those with psychosis in the absence of SUD compared to controls (OR = 0.90, 95% CI = 0.73–1.12, p = 0.35) after adjustment for the other covariates. However, women with major mood disorders were significantly more likely to have osteoporosis compared to the control group (OR = 1.37, 95% CI = 1.070–1.76, p = 0.013). Compared to the families and children eligibility category, all other eligibility categories were associated with significantly heightened risk for osteoporosis—a finding that likely reflects the fact that overall morbidity is, by eligibility definition, heightened in these categories of Medicaid enrollees compared to the women who compose the vast majority of the families and children category. Increasing age was associated with increased risk of osteoporosis (OR = 1.040, 95% CI = 1.025–1.054, p < 0.0001), and black race was associated with markedly decreased risk (OR = 0.40, 95% CI = 0.35–0.45, p < 0.0001). No significant interaction between psychosis and SUD for this modeling was apparent (OR = 1.66, 95% CI = 0.57–4.89, p = 0.35). Choice of linear or quadratic terms when adjusting for age did not substantially alter relative risk (odds ratios) of osteoporosis among diagnostic groups. An age2 specification was significant, but no more so than the linear form of this variable (OR = 1.00034, 95% CI = 1.00021–1.00046, chi-square = 29.5, p < 0.0001). Pathologic bone fracture as an indicator of osteoporosis (i.e., as an indicator for the dependent variable for this analysis) did not alter the results in any remarkable way
Table 4

Logistic model results: predictors of osteoporosis, Medicaid women age 50–64 yearsa (n = 18,953, disease prevalence = 6.4%)



95% CI

Chi-square (df = 1)

p value

Diagnosis (OR relative to controls)






 Substance use disorder





 Major mood disorder





Race (OR relative to white)











Eligibility (OR relative to families and children)

 Long-term care





 Dual eligible















Age (OR per one year increase in age)b





aExcludes women receiving pregnancy/family planning services only

bAge2 specification did not enhance the fit of the model or offer greater explanatory power regarding the dependent variable (OR = 1.00034, 95% CI = 1.00021–1.00046, chi-square = 29.5, p < 0.0001).

Psychosis × SUD interaction vs. controls (OR = 1.66, 95% CI = 0.57–4.89, chi-square = 0.85, p = 0.35). Accordingly, this parameter was dropped from the final modeling summarized above

Mutually exclusive modeling

The mutually exclusive diagnostic group analysis (Eq. 2 in the “Methods” section) was similar regarding the overall fit and individual effect sizes and probabilities to the results presented in Table 4. Again, only the mood disorder diagnosis was significantly increased compared to controls with regards to the odds of osteoporosis after adjusting for the covariates in the model (OR = 1.44, 95% CI = 1.11–1.86, p < 0.0001). The group defined as those with both psychosis and SUD was not different from the control group (p > 0.28). Age2 specification did not add to the overall fit or individual effect significance levels for that variable or other variables in the model.

Analysis of age 55–64 (postmenopausal)

Results of analyses restricted to those age 55–64 years were distinctive enough from the broader sample to comment about them in some detail. This older sample, despite the reduced size of 12,100 observations and only a slightly increased osteoporosis prevalence of 7.1%, led to a significant interaction term demonstrating that those with psychosis and SUD together are more likely to have a diagnosis of osteoporosis than controls (OR = 6.4, 95% CI = 1.51–27.6, p = 0.012). Such an interaction effect was not apparent in the full population studied suggesting that it may be unique to the slightly older subgroup, which presumably would include a higher proportion of post- or perimenopausal women. Other diagnostic effects in the age 55–64 sample were not significant except for the mood disorder group (OR = 1.42, 95% CI = 1.023–1.96, p = 0.036)—a finding that was consistent in magnitude and direction with results from the larger population (see Table 4). The prevalence of elevated osteoporosis risk in the dually diagnosed population is consistent across race and eligibility categories (data not shown). Other point estimates in the age 55–64 group were not remarkably different from the full population modeling.


To our knowledge, this is the first report to examine the impact of SUD on screening for osteoporosis and the frequency of osteoporosis diagnosis in women with schizophrenia and other psychotic disorders. Osteoporosis screening and treatment is essential as bone fracture is a significant clinical and economic concern. Twenty percent of people in the US sustaining an osteoporotic hip fracture die within 1 year, another 20% become permanently disabled, and 50% can no longer walk without assistance after the event [11]. These associated costs of osteoporosis-related fractures and repair are estimated to cost 17 billion dollars in the US alone [24].

Overall, we found a low rate of osteoporosis screening in the Medicaid population as a whole. Just 4% of our Medicaid sample of women aged 50 years and older without any record of osteoporosis received osteoporosis screening within a one-year period. It should be noted that such screening is not necessarily indicated for all women who are peri- or even postmenopausal. Instead the clinical focus for such screening is placed upon women with various risk factors especially including atraumatic fractures, and others such as smoking or the presence of diabetes [25]. From our administrative data, we could not accurately define the subset of women who had these clinical indicators. Many people do not begin screening until the age of 60 or 65; however, people with severe mental illness or substance abuse have many risk factors that indicate earlier screening age targets [11]. In one study of women over the age of 66 years, population screening rates were reported on the order of 9% to 20% (depending upon Medicaid or Medicare engagement and the presence or absence of diabetes) [26]. Such population screening rates seem consistent with the observations in our study given the fact that our population is younger and thus less likely to require or be referred for screening and additionally given the fact that the rates for the study just cited were based on a 2-year follow-up whereas ours were based on a single year follow-up.

In this study we found that, in the absence of substance use disorder, rates of screening for osteoporosis were comparable among women with schizophrenia and other psychosis, women with major mood disorders, and the control population (Table 3). One small study (n = 46 schizophrenia and n = 46 controls) in a VA population aged 45–64 found similar rates of screening in the two groups, but that people with schizophrenia had significantly lower rates than controls of medication prescription for prevention and treatment of osteoporosis [27]. In our study, people with SUD in the absence of other major psychiatric conditions had a 39% lower likelihood of receiving screening. From the administrative data we assembled, we could not determine whether this reduction in screening occurs because clinicians fail to recommend the test to this population, because patients fail to follow-up on that recommendation, or due to other factors. Nevertheless, our data indicate that SUD per se appears to correlate with reduced utilization of screening.

Our study did not find that the prevalence of diagnosed osteoporosis was elevated in the psychosis group compared to controls (Table 4). Other studies in similar populations have found higher rates of osteoporosis. For example, a recent study which used broadband ultrasound attenuation found that bone mass in women with schizophrenia under age 50 was significantly lower than healthy controls; however, broadband attenuation values were similar in groups in those over 50 years [15]. A study by our group found that bone mineral density as measured by DXA is also lower in middle-aged women with schizophrenia than in same-age healthy controls [14]. However, substance use was not examined in these reports. Though not observed as a significant finding in our full population age 50–64, our post hoc analysis of women age 55–64 found that SUD might significantly impact osteoporosis prevalence. For those oldest subjects, there was more than a 6-times increase (OR = 6.7) in the observed prevalence of osteoporosis for the group with the diagnosis of psychosis and SUD together, although the psychosis or substance use diagnoses alone demonstrated no such increased prevalence.

Alcohol and drugs of abuse have been associated with an increased risk of osteoporosis and fracture [28, 29], particularly in the older population [30]. In a Swedish study 20% of all patients with hip fracture during a 2-year period were in people with alcohol and drug abuse [31]. Furthermore, early menopause is associated with a greater risk of osteoporosis, and drug use has been found to be associated with a 2.6-fold increased risk for early menopause [32]. Amenorrhea from prolactin elevating antipsychotics has also been suggested to increase osteoporosis risk in women with schizophrenia and long-term cumulative exposure may explain the higher risk in the older ages. However, because we did not observe a significant risk of osteoporosis in the non-SUD psychotic disorders group, this may suggest that comorbid substance use plays a significant role and may partially explain the higher rates of bone loss seen in other studies where substance use was not reported [12, 16, 33]. Whatever the explanation, the convergence of both the diagnosis of schizophrenia and substance use in our subpopulation of women age 55–64 correlates with elevated risk of osteoporosis. Consistent with this apparent life-stage finding, we have recently completed another study using DXA in older (>50 years) patients and found that those who smoke or are alcohol dependent had a significantly greater bone mass loss [17].

Other interesting findings of the current study include the following. First, women of black race were significantly less likely to receive screening for osteoporosis, as has been reported previously [34, 35], yet were also less likely to be diagnosed with this disorder. There is evidence to suggest that black women are at a decreased risk for osteoporosis and our findings parallel these reports [36, 37]. It is not clear however if the decreased prevalence noted in our study is an artifact of decreased screening or truly represents a lower risk in this population. Regarding race, it is noteworthy that the apparent prevalence of major mood disorders in black women was substantially lower than in white women (209/9,690, 2.2% vs. 456/7968, 5.7%; Table 2). This project was not designed to evaluate diagnostic racial disparities, however, this finding is interesting in view of previously observed disparities in treating and possibly diagnosing mood disorders in minority populations [38]. Lastly, although prevalence of screening was similar among patients with mood disorders and controls (Tables 2 and 3), risk of osteoporosis was significantly elevated in those with mood disorders (see Table 4). People within this diagnostic category were 37% more likely to develop osteoporosis compared to the control group. This diagnostic group may be at increased risk of osteoporosis that may be related to other mechanisms than those associated with substance use or psychotic disorders [39]. Osteoporosis in women with mood disorders might be elevated due in part to hypercortisolemia and maintaining a hyperadrenergic state [40, 41], or increases in interleukin 6, one of the most potent known bone resorption factors. Additionally, recent evidence suggests that antidepressants also may contribute to risk of osteoporosis [42]. It should finally be noted here that reverse causality may be partially in play regarding this observed association, i.e., that osteoporosis may lead to depression because the disease is associated with deteriorating physical health.

A limitation of our study is that our estimate of prevalence rates for osteoporosis comes from administrative records of treatment, rather than direct screening of the entire sample, and thus could potentially miss large numbers of undiagnosed cases. Thus, we would interpret the apparently similar prevalence of osteoporosis (Table 4) in women with SUD and controls, despite lower screening rates (Table 3), as a signal that osteoporosis is likely under-diagnosed in women with SUD. Prevalence of osteoporosis is higher in women with mood disorders compared with controls even while screening rates are comparable in both groups. This may reflect appropriate and targeted testing, or it might, again, suggest that the mood disorder group may have a true prevalence of osteoporosis that is even higher than reported, due to inadequate screening in this high risk group. Accordingly, Medicaid records are not as sensitive to prevalence as epidemiologic survey data that typically have higher potential to obtain untreated or otherwise undisclosed pathology. The apparent prevalence of substance use in this population of women age 50–64 was 7.2%, of whom 11% had comorbid diagnoses of psychosis. While the overall population rate (7.2%) is on par with epidemiologic studies including younger adults, the rate of substance use as a comorbidity to psychosis was only 11.4%—lower than the 50% lifetime prevalence anticipated [43]. In Medicaid reporting, the principal disease (psychosis) and its treatment may eclipse reporting of comorbid substance abuse because it is not relevant to reimbursement expectations. In addition, formally diagnosed SUD is known to be lower than the actual rates of such illness [44]. The Epidemiological Catchment Area study reported a prevalence of 6.1% for drug/alcohol abuse and dependence in the general population [45] and the most recent National Comorbidity Survey reported 3.8% for drug/alcohol abuse or dependence [46] similar to what we have observed here. In evaluating our reported prevalence of the comorbid substance abuse and psychosis, it is important to keep in mind that our derived rates are annual and not life-time prevalence figures. Our figures furthermore reflect late-life prevalence and therefore may seem lower than expected [41, 42]. While lifetime alcohol use rates have been reported to be as high as 77% in schizophrenia, at any point in time, current prevalence of alcohol abuse or dependence is generally reported to be around 20% [47, 48]. Occasional and lifetime use would not necessarily be included in administrative data as it, by design, emphasizes contemporary diagnostic and treatment events.

Another limitation to this study is the fact that the logistic regression model results predicted only 1% and 4% of the variance regarding screening or osteoporosis rates, respectively. Though significant, much remains unexplained regarding who received screening or who is at risk for disease. For example, we were unable to control for obesity, smoking status, diet, exercise, and other unknown factors in this analysis, but this type of information is difficult to gather in even a prospective design. Medication use, particularly antipsychotic exposure and prescriptions for treatment (e.g., biphosphonates) were not analyzed. Future work would benefit from considering such ancillary and direct treatment variables. Lastly, although the majority of people with schizophrenia and psychotic disorders in the USA are receiving Medicaid benefits [18], generalizability of these findings across groups is questionable because the sample size of the control group was considerably larger than the diagnostic groups, and some of the diagnostic groupings were quite small. Finally, it remains unclear if SUD leads to lower rates for all preventive health screening or if this is specific to DXA and other bone density scanning. There is little published literature on this topic in an SUD population and more research is needed in this area.

In conclusion, screening rates for osteoporosis in people with schizophrenia are low but do not differ from the general population of those on Medicaid. However, this population may be at greater risk if they are substance users/abusers, thereby increasing the potential need for screening for such persons. Additionally, screening rates for those with substance abuse problems in isolation also appear to be below those for most other Medicaid enrollees. Accordingly, it appears that efforts should be made to increase osteoporosis prevention and treatment in those with substance abuse problems and especially in those who also have comorbid psychosis [49, 50]. The higher rates of diagnosed osteoporosis among older women with major mood disorders indicate that this group, also, may require more extensive screening.


This study was supported by the Intramural Research Program, NIH, National Institute on Drug Abuse and the NIDA Residential Research Support Services Contract HHSN271200599091CADB. The authors wish to thank Cynthia Boddie-Willis, Susan Chen, and Nancy Svehla for their contributions to this work.

Conflicts of interest

Mr. Abrams is a frequent consultant to Maryland's Medicaid Program.

Copyright information

© International Osteoporosis Foundation and National Osteoporosis Foundation 2010