General Hospital Psychiatry 37 (2015) 60–66
Contents lists available at ScienceDirect
General Hospital Psychiatry journal homepage: http://www.ghpjournal.com
Health screening, counseling, and hypertension control for people with serious mental illness at primary care visits Sharat P. Iyer, M.D., M.S. a,b,⁎, Alexander S. Young, M.D., M.S.H.S. c,d a
VISN3 Mental Illness Research, Education and Clinical Center, James J. Peters VA Medical Center, 130 W Kingsbridge Road, Room 6A-44, Bronx, NY 10468, USA Department of Psychiatry, Icahn School of Medicine at Mount Sinai, 1 Gustave L. Levy Place, Box 1230, New York, NY 10029, USA Desert Pacific Mental Illness Research, Education and Clinical Center, West Los Angeles VA Medical Center, 11301 Wilshire Blvd, Building 210, Los Angeles, CA 90073, USA d Department of Psychiatry and Biobehavioral Sciences, University of California, Los Angeles, 10920 Wilshire Blvd, Suite 300, Los Angeles, CA 90024, USA b c
a r t i c l e
i n f o
Article history: Received 11 October 2013 Revised 19 September 2014 Accepted 7 October 2014 Keywords: Primary care Serious mental illness Health screening Health counseling Hypertension
a b s t r a c t Objective: This study sought to determine if primary care visits for people with serious mental illness (SMI) demonstrate different rates of basic physical health services compared to others, and to determine factors associated with differing rates of these measures in people with SMI. Method: The study used 2005–2010 visit-level primary care data from the National Ambulatory Medical Care Survey and the National Hospital Ambulatory Medical Care Survey. The provision of health counseling, receipt of any diagnostic or screening test, measurement of blood pressure or weight and evidence of hypertension control were assessed, adjusting for identified patient, provider and visit-level factors. Results: After adjustment for covariates, we found no significant differences between visits for people with SMI and those without for any outcome. Probability of blood pressure measurement and diagnostic or screening testing significantly increased over time. Conclusion: The lack of significant differences found here might be due to adjustment for covariates, a focus only on primary care visits, the use of visit-level data or evolution over time. Mortality differences for people with SMI may be attributable to those not receiving primary care, self-management of disease or subsets of the population requiring targeted interventions. Published by Elsevier Inc.
1. Introduction Over the last two decades, there has been increasing concern about the physical health of people with serious mental illness (SMI). People with SMI have an age-adjusted risk of mortality 2 to 2.5 times higher than the general population, with the primary causes of excess mortality being cardiovascular disease and cancer, and with rates increasing over time [1–6]. One important reason for this excess mortality may be a lack of sufficient primary care assessment and management, such as basic physical health screening, health counseling, and management of chronic medical conditions. Prior studies using Medicaid claims data, 1990s Veterans Administration data and United Kingdom primary care data have shown lower rates of basic health screening, health counseling and immunization after adjusting for basic factors [7–13]. However, Daumit et al. [14] used nationally representative data from 1993 to 1998 and showed in an unadjusted comparison between those with SMI and those without that rates of health counseling by primary care physicians were not significantly different. No study has thus far used nationally representative data to assess primary care for people with SMI since
⁎ Corresponding author at: James J. Peters, VA Medical Center (MIRECC-OOMH), 130 W. Kingsbridge Road, Room 6A-44, Bronx, NY 10468, USA. Tel.: +1 718 584 9000x3705. E-mail addresses: [email protected] (S.P. Iyer), [email protected] (A.S. Young). http://dx.doi.org/10.1016/j.genhosppsych.2014.10.003 0163-8343/Published by Elsevier Inc.
adoption of screening guidelines in 2004 [15,16]. Furthermore, no study has controlled for the wide variety of patient-level, providerlevel and system-level factors that may influence receipt of services for people with SMI. Thus, this study sought to use nationally representative data after 2005 to determine if people with SMI have different rates of basic physical health assessment, health counseling, and hypertension control in primary care settings, compared to the general population, after adjusting for a broad array of potential confounders. Furthermore, this study sought to determine what factors might contribute to differing rates of screening, health counseling and hypertension control in people with SMI. The ultimate goal was to identify potential targets for intervention to improve primary care for people with SMI, in order to influence their high rates of mortality. 2. Methods 2.1. Study design This study is a cross-sectional analysis of data from two annual US national health care surveys conducted by the Centers for Disease Control and Prevention, the National Ambulatory Medical Care Survey (NAMCS) and the National Hospital Ambulatory Medical Care Survey (NHAMCS). NAMCS contains data obtained from office-based
S.P. Iyer, A.S. Young / General Hospital Psychiatry 37 (2015) 60–66
outpatient medical providers. NHAMCS includes data from both emergency departments and hospital-based outpatient clinics. In both of these studies, specially trained interviewers complete or assist providers in completing Patient Record Forms (PRFs) for patient visits. These visit-level data are the unit of analysis for both surveys. Both studies use a multistage probability design, with probability samples collected within each stage. For NAMCS, these stages are geographical probability sampling units (PSUs), physician practices and patient visits; for NHAMCS, the stages are PSUs, hospitals, clinics and patient visits. The NAMCS data set includes sample weights for each reported visit to correlate each data point with a representative number of patient visits. These sample weights have been adjusted since 2003 to allow for single sampling stage analysis using ultimate cluster design. Prior studies have shown that these data sets can be used successfully to evaluate care for people with SMI [14,17]. In this study, we combined NAMCS data and office-based outpatient NHAMCS data — emergency department data were excluded. NAMCS and outpatient NHAMCS data are similar in structure, thus allowing the combining of these data sets. Only data from 2005 to 2010 were combined, ensuring the analysis covers data collected after the publication of screening guidelines in 2004. NAMCS and NHAMCS surveys are revised yearly, including minor changes in variable coding; all variables of interest were recoded as possible to reflect their status in 2010. Singly-imputed race and ethnicity variables were provided in the NAMCS and NHAMCS data sets themselves. Inclusion criteria for the study were visits in which consumers were at least 16 years old and in which the visit provider was indicated on the PRF to be the consumer's primary care provider. 2.2. Study variables The five primary outcome variables were evidence in a visit of the provision of any type of health counseling, the provision any diagnostic or screening test, the measurement of weight, the measurement of blood pressure and evidence of blood pressure readings below 140/90 in consumers diagnosed with hypertension. Identification of provision of health counseling was based on selection of a check box labeled as such on the PRF, or of any of several related subcategory checkboxes. Identification of provision of any diagnostic or screening test was based on any of the following being selected on the PRF: examination of the breast, foot, retina, pelvis, rectum or skin; screening for depression; any imaging test; any laboratory test; biopsy; sexually transmitted disease testing; Papanicalou testing; any scope procedure; spirometry; or urine testing. Identification of blood pressure or weight measurement was based on the presence or absence of a measurement on the PRF. Hypertension diagnosis was based on a check box labeled as such on the PRF. The primary predictor was diagnosis of SMI, defined as schizophrenia and related primary psychotic disorders (ICD-9 295.xx) or mood disorders including major depressive disorder and bipolar disorder (ICD-9 296.xx) but excluding single major depressive episodes (ICD-9 296.2x) and mild, moderate, or remitted recurrent major depression (ICD-9 296.30-296.32, 296.35-296.36). Each PRF allows for the coding of up to three ICD-9 diagnoses per visit. Covariates of interest were adapted from the Andersen Behavioral Model of Health Services Utilization, which has been used in prior studies evaluating health screening in primary care, as well as existing literature on barriers to adoption of screening for people with SMI [18–24]. Covariates used in this analysis were age (including a squared term to account for nonlinear relationships), sex, race, ethnicity, location in (or not in) a metropolitan statistical area (MSA), US geographical region (with NAMCS-defined categories of Northeast, Midwest, South and West), total number of chronic medical conditions, the Deyo modification of the Charlson comorbidity index, total number of medications, receipt of antipsychotic medications, use of tobacco, number of prior visits in the last year, having seen a physician during the patient visit, primary reason for presenting for the visit, primary type of insurance used for
61
the visit, type of practice office setting (with three categories of government-funded or federally qualified health center, private provider or clinic, and HMO, hospital-based, or faculty clinic), clinical use of electronic medical records (EMRs) and year of visit. For the outcome of weight measurement, a diagnosis of obesity was also included as a covariate, and for the outcome of blood pressure measurement, a diagnosis of hypertension was included. 2.3. Analysis This study used logistic regressions and accounted for study design using provided sampling weights. Analyses were performed using STATA 13 (College Station, TX, USA). Correlations between all predictors in the model, including all covariates and SMI diagnosis, were assessed using Pearson correlations for continuous variables and Cramer's phi for categorical variables, with no correlations higher than 0.5. Weighted univariate analysis with adjusted Wald tests was performed to compare visits for people with SMI to those without for all covariates and outcomes. Weighted unadjusted and adjusted logistic regressions were performed on the outcomes of interest, with adjustment for the covariates listed above. We used interaction effects in order to determine what factors may affect the outcomes of interest specifically in visits for people with SMI. To screen for interaction effects of interest, separate weighted adjusted logistic regressions were performed with one interaction effect included at a time for each of the identified covariates. Covariates with interaction effects demonstrating absolute coefficient t scores greater than 1.5 were identified as being potentially meaningful for each outcome of interest. Final models were then created for each outcome of interest including interaction effects for all potentially meaningful factors. Predictive margins for diagnosis of SMI were calculated using Taylor series approximations in order to demonstrate the predicted differences in outcomes for each significant interaction effect between visits for people with SMI and those without. Specification error link tests were performed for all adjusted models. 3. Results 3.1. Study population description We identified 86,901 visits meeting inclusion criteria, with 1133 visits identified as including diagnoses of SMI, representing 2.14 billion and 14.9 million nationwide visits, respectively. We identified 31,547 visits as meeting inclusion criteria and noting a diagnosis of hypertension, representing 825 million nationwide visits. Table 1 provides a description of the sample, comparing primary care visits for people with SMI to those without, after accounting for survey weighting. Factors are grouped by domains of the Andersen Behavioral Model. Visits for people with SMI, as compared to those without SMI, included individuals that were significantly younger, had more medications and were more likely to be prescribed antipsychotic medication. There were significant differences between visits for people with SMI and those without across insurance type, office setting, number of visits in the last year and reason for visit. 3.2. Impact of SMI on outcomes of interest Table 2 demonstrates the unadjusted and adjusted effect of SMI diagnosis on the outcomes of interest. The covariates for the adjusted models were all covariates noted in the methods above. In the unadjusted analyses, SMI diagnosis is associated with a significant increase in the odds of receiving health counseling and a significant decrease in the odds of receiving any diagnostic testing, blood pressure measurement and weight measurement. SMI diagnosis is not associated with hypertension control. However, the significance of the association between SMI diagnosis and all outcome findings are lost after adjusting for
62
S.P. Iyer, A.S. Young / General Hospital Psychiatry 37 (2015) 60–66
Table 1 Comparison of characteristics between primary care visits for people with and without SMI Non-SMI % or Mean Environmental factors Region of US Northeast Midwest South West MSA Office setting Government-funded Private hospital HMO, or faculty Predisposing factors Mean age Female Nonwhite race Hispanic ethnicity Mean total number of chronic conditions Mean number of medications Receipt of antipsychotic medication
SMI 95% CI
% or mMean
Adjusted Wald F test
P value
95% CI
18.2% 26% 36.8% 19% 17.3%
[15.1%, 21.2%] [21.7%, 30.4%] [32.7%, 40.9%] [15.7%, 22.4%] [9.1%, 25.5%]
20.9% 26.6% 36.1% 16.4% 14.5%
[13.5%, 28.3%] [17.5%, 35.8%] [25.9%, 46.3%] [11.4%, 21.4%] [5.9%, 23.2%]
3.7% 86.3% 10%
[2.9%, 4.5%] [84.5%, 88.2%] [8.4%, 11.6%]
9.1% 77.7% 13.1%
[5.8%, 12.5%] [72.3%, 83.1%] [9.3%, 17%]
53.27 60.7% 17.1% 11.5% 1.58 3.11 1.5%
[52.58, 53.96] [59.8%, 61.6%] [15.1%, 19.1%] [9.6%, 13.5%] [1.53, 1.64] [2.99, 3.22] [1.4%, 1.7%]
44.45 58.6% 18.4% 9.8% 1.7 3.70 46.6%
[42.20, 46.70] [52.3%, 64.8%] [12.9%, 23.8%] [6%, 13.6%] [1.57, 1.83] [3.43, 3.98] [40.1%, 53.1%]
0.43
.735
1.14 8.97
.286 b.001
60.12 0.50 0.25 0.83 3.28 22.20 58.35
b.001 .481 .619 .363 .070 b.001 b.001
17.84
b.001
4.70
.003
0.88 0.24
.348 .625
Enabling factors Type of insurance Medicaid Medicare Private or other Uninsured Number of past visits in the past year 0 visits 1–2 visits 3–5 visits 6+ visits Seen by a physician EMR
9.7% 29.6% 56% 4.7%
[8.8%, 10.7%] [27.8%, 31.3%] [54.1%, 57.8%] [4.2%, 5.2%]
24.7% 26.2% 40.7% 8.4%
[19.8%, 29.5%] [20%, 32.5%] [33.7%, 47.7%] [4.8%, 11.9%]
9.1% 29.5% 33.7% 27.6% 95.1% 43.2%
[8.6%, 9.7%] [28.5%, 30.6%] [32.8%, 34.6%] [26.3%, 29%] [94.3%, 95.9%] [39.6%, 46.8%]
6.3% 22.9% 32.5% 38.3% 93.9% 45.1%
[3.5%, 9%] [18.4%, 27.5%] [26.6%, 38.3%] [32.9%, 43.8%] [91.1%, 96.7%] [36.8%, 53.4%]
Need-related factors Reason for visit Preventive care Routine chronic problem New problem, flare-up or surgery Mean Charlson index score Tobacco use
13.94
b.001
19.5% 32.9% 47.7% 0.30 18.6%
[18.2%, 20.7%] [31.1%, 34.6%] [46.4%, 48.9%] [0.29, 0.32] [17.5%, 19.7%]
8.9% 59.3% 31.7% 0.282 46.7%
[4.9%, 13%] [51.9%, 66.8%] [26.3%, 37.2%] [0.16, 0.40] [38%, 55.4%]
0.11 40.65
.736 b.001
Outcomes Health counseling Diagnostic or screening test Blood pressure Weight Controlled hypertension
41.9% 59.8% 90.1% 87.4% 73.7%
[39.7%, 44.1%] [57.9%, 61.7%] [89.1%, 91.2%] [86.3%, 88.6%] [72.8%, 74.6%]
50.6% 49.6% 79.3% 78.6% 74.9%
[42.2%, 59.1%] [41.9%, 57.4%] [70.5%, 88.1%] [69.7%, 87.6%] [69.6%, 80.2%]
4.11 5.56 5.01 3.25 0.20
.043 .019 .025 .071 .656
confounding factors. All adjusted models demonstrated significance and all specification link tests were nonsignificant, demonstrating no misspecification. 3.3. Factors associated with the outcomes of interest in visits for people with SMI Table 3 demonstrates the effect of potentially meaningful factors on the outcomes of interest in visits for people with SMI diagnoses, as compared to those without, grouped by Andersen Behavioral Model domain. The covariates for the adjusted models were all covariates noted in the methods above, as well as potentially meaningful interaction effects. Compared to visits for people without SMI, visits for people with SMI showed an overall significantly higher odds over time (i.e., from years 2005 to 2010) of provision of any diagnostic or screening test and of blood pressure assessment, as shown in Figs. 1 and 2. In 2010, visits for people with SMI had higher predicted probabilities of provision of diagnostic or screening tests (with SMI=72.8%, without SMI= 60.0%, difference P= .034) and of provision of blood pressure assessment (with SMI=97.1%, without SMI=90.9%, difference P= .002)
compared to visits for people without SMI. Visits for people with SMI had lower predicted probabilities of receipt of blood pressure measurement in 2005 (with SMI=73.9%, without SMI=93.9%, difference P= .048). Compared to visits for people without SMI, visits for people with SMI demonstrated significantly higher odds of provision of health counseling and of hypertension control if there were six or more visits (as compared to no visits) by that individual to the same provider in the past year. Compared to visits for people without SMI, visits for people with SMI had lower odds of provision of weight measurement for those with private insurance (as compared to Medicaid) and for those with an obesity diagnosis. Similarly, visits for people with SMI had lower odds of provision of health education as the Charlson index score increased, as seen in Fig. 3. However, compared to visits for people without SMI, visits for people with SMI had higher odds of provision of weight measurement as the number of medications increased, as seen in Fig. 4. Fig. 5 shows the comparison of the effect of age on hypertension control between visits for people with and without SMI. Compared to visits for people without SMI, visits for people with SMI demonstrated significantly lower predicted probabilities of hypertension control at lower and higher ends of
S.P. Iyer, A.S. Young / General Hospital Psychiatry 37 (2015) 60–66
63
Table 2 Impact of the diagnosis of SMI on the provision of health counseling, of any diagnostic or screening test, of weight measurement, of blood pressure measurement and on hypertension control
Unadjusted models SMI, OR [95% CI] N F df_r P Adjusted models SMI, OR [95% CI] N F df_r P Link test
Health counseling
Any diagnostic or screening test
Weight
Blood pressure
Hypertension control
1.42⁎ [1.02, 1.97] 85,829 4.41 3095 .036
0.66⁎⁎ [0.49, 0.89] 86,901 7.33 3106 .007
0.53⁎ [0.32, 0.89] 86,901 5.86 3106 .016
0.42⁎⁎⁎ [0.25, 0.70] 86,901 11.20 3106 .001
1.07 [0.80, 1.42] 77,157 0.19 2730 .660
1.20 [0.78, 1.84] 42,358 18.92 2648 0.000 .932
0.71 [0.30, 1.64] 42,770 12.71 2654 0.000 .473
0.51 [0.21, 1.25] 42,770 16.98 2654 0.000 .708
0.81 [0.55, 1.20] 42,770 17.46 2654 0.000 .755
0.91 [0.43, 1.89] 16,930 6.58 1997 0.000 .526
OR=odds ratio; 95% CI=95% confidence interval; N=number of observations in sample; F=overall model F statistic; df_r=degrees of freedom; P=overall model P value; Link test=specification error link test P value. ⁎ Pb.05. ⁎⁎ Pb.01. ⁎⁎⁎ Pb.001.
the age spectrum. Regional differences were noted between visits for people with SMI and those without for receipt of health education and weight measurement. 4. Discussion This study sought to determine if people with SMI seen in primary care settings have different rates of receiving health counseling, basic physical health screening and hypertension control compared to the general population, using nationally representative data after publication of screening guidelines. After adjusting for potential confounders, we found no significant differences between visits for people with SMI and those without for any of the outcomes of interest. For the outcome of receipt of health counseling, this supports earlier work using the same data set [14]. However, these findings otherwise seem to contradict prior evidence of low rates of physical health screening and poorer performance on process quality measures for people with SMI. This may be in part due to insufficient control for confounders in prior studies, as previously identified differences might be due to unmeasured clinical, environmental, or socioeconomic factors associated with SMI diagnoses. In addition, prior studies have not always focused exclusively on primary care provider visits. However, this finding may also be explained by the increasing probability over time of providing blood pressure measurement and diagnostic or screening tests in visits for people with SMI compared to those without. Further investigation is needed to understand how this occurred and might include mechanisms such as diffusion of evidence, adoption of guidelines, improving education of primary care practitioners or direct interventions. Elucidating the means by which this change occurred may help guide future interventions to improve primary care for this population. The lack of significant differences in receipt of basic services is important to understand in juxtaposition to the disparity in mortality for people with SMI. As this study only examined care provided at primary care visits, it does not address people with SMI who do not or cannot present to primary care. It also does not address potential disparities in self-management of chronic diseases between visits. In addition, it is unknown if primary care services moderate the excess mortality risk in this population. Furthermore, if the mortality disparity for people with SMI is due to factors inherent to their illnesses, then this vulnerable population may require more services than the general population, rather than only equivalent services. The findings of significantly higher probabilities in 2010 of provision of blood pressure measurement and diagnostic and screening tests for people with SMI compared to those
without are therefore reassuring. However, we did not find similar differences in weight measurement or health counseling. Understanding differences between these services in this population can help guide future primary care-based interventions for people with SMI. A number of different factors affected the receipt of basic services and hypertension control in visits for people with SMI, as compared to visits for others, and may point to avenues for intervention development. Lower probabilities of hypertension control at lower and higher ends of the age spectrum may point to a need for targeted interventions focused on these age groups, in conjunction with existing efforts to develop integrated service models for new-onset mental illness and geriatric populations. The higher probability of health counseling and controlled hypertension seen with higher numbers of visits over the last year may indicate a benefit to care models allowing higherfrequency, lower-intensity primary care contacts for people with SMI. The lower probability of weight measurement in primary care visits for people with SMI holding private insurance, as compared to Medicaid, may indicate a need for improved dissemination of guidelines beyond public settings. However, the lower probability of health counseling in visits for people with higher Charlson index scores and SMI diagnoses and the lower probability of weight measurement in visits for people obesity and SMI diagnoses are particularly concerning. These findings, as well as findings of regional differences in outcome, require further study. This study has a number of limitations, particularly because the data sets used provide visit-level data only. Some identified confounding patient-level and provider-level factors identified in the Andersen Model and the literature could not be included in this analysis. In particular, the NAMCS and NHAMCS data sets have insufficient capture of consumer socioeconomic status and health knowledge, and of provider education and attitudes toward people with SMI. These omitted variables may have affected the analyses performed, potentially masking significant differences between visits for people with and without SMI. Furthermore, the variables used are based on provider report or CDC staff chart abstraction and are limited by the necessary simplicity of the PRF. Variables on the PRF often do not have precise definitions and are open to interpretation. Therefore, the recorded variables may not fully represent the consumer's status or the content of the visit, such as the actual content of health counseling provided, or the distinction between blood pressure and weight measurement for screening versus diagnostic assessment. The determination of SMI diagnosis relies on a related diagnosis being listed in three available spaces for diagnosis on the PRF. Providers may not have coded an SMI diagnosis if it was not a
64
S.P. Iyer, A.S. Young / General Hospital Psychiatry 37 (2015) 60–66
Table 3 Factors associated with provision of health counseling, of any diagnostic or screening test, of weight measurement and of blood pressure measurement and factors associated with hypertension control, in visits for people with SMI and with results in odds ratios [95% confidence intervals] Potentially meaningful interactions
Health counseling
Environmental factors Years since 2005 Region of US Northeast (baseline) Midwest 1.13 [0.51, 2.49] South 2.12⁎ [1.00, 4.48] West 1.35 [0.49, 3.68] MSA Office setting Government-funded clinic (baseline) Private clinic HMO, hospital, faculty clinic Predisposing factors Hispanic Age Age squared Antipsychotic prescription Number of chronic conditions Number of meds Enabling factors Insurance Medicaid (baseline) Medicare Private insurance Uninsured Number of visits in the past year 0 visits (baseline) 1–2 visits 3–5 visits 6+ visits Saw a physician EMR Need-related factors Reason for visit Preventative care (baseline) Chronic problem New problem or flare-up Charlson score Tobacco use Diagnosis of obesity N F df_r P Link test
Any diagnostic or screening test Blood pressure 1.33⁎ [1.02, 1.72]
Weight
Normal blood pressure if hypertensive
1.96⁎⁎ [1.27, 3.01]
5.28⁎ [1.22, 22.77] 1.08 [0.24, 4.90] 4.49 [0.92, 22.02] 3.16 [0.69, 14.43]
0.27 [0.07, 1.03] 0.46 [0.08, 2.64]
0.34 [0.10, 1.11]
1.38⁎⁎ [1.14, 1.68] 1.00⁎⁎ [0.99, 1.00] 1.18 [0.42, 3.27] 1.37⁎ [1.02, 1.84]
2.09 [0.88, 4.96] 1.42 [0.58, 3.48] 0.87 [0.23, 3.34]
0.28 [0.06, 1.26] 0.72 [0.17, 3.06] 0.17 [0.03, 1.06]
0.14 [0.02, 1.24] 0.11⁎ [0.02, 0.66] 0.14 [0.02, 1.06]
3.62 [0.78, 16.75] 2.95 [0.78, 11.17] 5.01⁎ [1.30, 19.41]
0.86 [0.13, 5.58] 8.62 [0.92, 80.68] 7.06⁎ [1.35, 36.88] 0.79 [0.18, 3.50] 2.01 [0.53, 7.59]
0.61⁎⁎ [0.42, 0.87] 0.89 [0.49, 1.62]
42,358 15.11 2648 .000 .780
0.54 [0.26, 1.10]
1.09 [0.35, 3.38] 3.16⁎ [1.08, 9.28]
0.25 [0.03, 1.91] 1.63 [0.13, 20.59] 1.33 [0.57, 3.10]
42,770 16.97 2654 .000 .867
0.81 [0.21, 3.20]
42,770 14.71 2654 .000 .509
0.16⁎⁎ [0.04, 0.58] 42,770 14.14 2654 .000 .448
16,930 7.25 1997 .000 .459
N=number of observations in sample; F=overall model F statistic; df_r=degrees of freedom; P=overall model P value; Link test=specification error link test P value. ⁎ Pb.05. ⁎⁎ Pb.01.
prominent part of the visit's content or the consumer's presentation, or if multiple other diagnoses were more prominent. Thus, not all visits for people with SMI may have been identified in this study, and some analyses may have been underpowered. People with SMI likely present to primary care late in the course of their chronic medical conditions, due to insurance coverage, selfstigma, and cognitive impairment. However, direct assessment of the severity of physical illnesses was not available in this data set in sufficient granularity. While prior studies have used the Charlson index to adjust for severity of medical comorbidity using NAMCS and NHAMCS data, this index may have been an imperfect proxy in this population and may have underestimated true medical severity. The ultimate goal of this study was to address the disparity in chronic physical health mortality for people with SMI. The findings of this study did not demonstrate a similar disparity in basic services provided in primary care for this population, and found improvements in provision of certain services over time. Nevertheless, certain subsets of visits for people with SMI demonstrated lower probabilities of receipt of
services, including visits for people with higher medical severity, lower visit frequency, older or younger age and private insurance. Determining how mortality risk correlates with these subsets of people with SMI, as well as how lack of primary care may affect risk of death for people with SMI, will help to better identify the precise means to reduce the excess mortality of this vulnerable population. Acknowledgments This material is the result of work supported with resources and the use of facilities at the James J. Peters VA Medical Center, Bronx, NY, and the West Los Angeles VA Medical Center, Los Angeles, CA. Additional support was provided by the Mount Sinai Conduits Institutes for Translational Sciences. Assistance for this article was provided by Kenneth Wells, M.D., M.P.H., and Lisa Dixon, M.D., M.P.H., who reviewed the manuscript and contributed to the discussion, and Alan Weinberg, M.S., who contributed to the analytic plan. Funding for Dr. Parameswaran was provided by the Robert Wood Johnson Foundation Clinical Scholars
.4
.6
.8
1
65
.2
Probability of diagnostic test or screening
S.P. Iyer, A.S. Young / General Hospital Psychiatry 37 (2015) 60–66
0
1
2
3
4
5
Years since 2005 Non-SMI
SMI
Fig. 1. Change in probability of diagnostic or screening testing over time, comparing visits with and without and SMI diagnosis, after adjustment for covariates.
Fig. 3. Change in probability of receipt of health education by Charlson index score, comparing visits with and without and SMI diagnosis, after adjustment for covariates.
program and the US Department of Veterans Affairs. The views expressed in this article are those of the authors and do not necessarily reflect the position or policy of the Department of Veterans Affairs or the United States government.
[10] Kaplowitz RA, Scranton RE, Gagnon DR, Cantillon C, Levenson JW, Sesso HD, et al. Health care utilization and receipt of cholesterol testing by veterans with and those without mental illness. Gen Hosp Psychiatry 2006;28: 137–44. [11] Desai MM, Rosenheck RA, Druss BG, Perlin JB. Receipt of nutrition and exercise counseling among medical outpatients with psychiatric and substance use disorders. J Gen Intern Med 2002;17:556–60. [12] Roberts L, Roalfe A, Wilson S, Lester H. Physical health care of patients with schizophrenia in primary care: a comparative study. Fam Pract 2007;24:34–40. [13] Osborn DP, Baio G, Walters K, Petersen I, Limburg H, Raine R, et al. Inequalities in the provision of cardiovascular screening to people with severe mental illnesses in primary care: cohort study in the united kingdom thin primary care database 2000–2007. Schizophr Res 2011;129:104–10. [14] Daumit GL, Pratt LA, Crum RM, Powe NR, Ford DE. Characteristics of primary care visits for individuals with severe mental illness in a national sample. Gen Hosp Psychiatry 2002;24:391–5. [15] American Diabetes Association, American Psychiatric Association, American Association of Clinical Endocrinologists, North American Associat0069on for the Study of Obesity. Consensus development conference on antipsychotic drugs and obesity and diabetes. Diabetes Care 2004;27:596–601. [16] Marder SR, Essock SM, Miller AL, Buchanan RW, Casey DE, Davis JM, et al. Physical health monitoring of patients with schizophrenia. Am J Psychiatry 2004;161:1334–49. [17] Rost K, Hsieh YP, Xu S, Menachemi N, Young AS. Potential disparities in the management of schizophrenia in the United States. Psychiatr Serv 2011;62:613–8. [18] Andersen RM. Revisiting the behavioral model and access to medical care: does it matter? J Health Soc Behav 1995;36:1–10. [19] Brennenstuhl S, Fuller-Thomson E, Popova S. Prevalence and factors associated with colorectal cancer screening in Canadian women. J Womens Health (Larchmt) 2010;19:775–84. [20] Rahman SM, Dignan MB, Shelton BJ. A theory-based model for predicting adherence to guidelines for screening mammography among women age 40 and older. Int J Cancer Prev 2005;2:169–79.
References
1 .9 .8 .7 .6 .5
Probability of blood pressure measurement
[1] Saha S, Chant D, McGrath J. A systematic review of mortality in schizophrenia: is the differential mortality gap worsening over time? Arch Gen Psychiatry 2007;64:1123–31. [2] Brown S, Kim M, Mitchell C, Inskip H. Twenty-five year mortality of a community cohort with schizophrenia. Br J Psychiatry 2010;196:116–21. [3] Crump C, Winkleby MA, Sundquist K, Sundquist J. Comorbidities and mortality in persons with schizophrenia: a Swedish national cohort study. Am J Psychiatry 2013;170:324–33. [4] Hoang U, Stewart R, Goldacre MJ. Mortality after hospital discharge for people with schizophrenia or bipolar disorder: retrospective study of linked English hospital episode statistics, 1999–2006. BMJ 2011;343:d5422. [5] Talaslahti T, Alanen HM, Hakko H, Isohanni M, Hakkinen U, Leinonen E. Mortality and causes of death in older patients with schizophrenia. Int J Geriatr Psychiatry 2012;27:1131–7. [6] Bushe CJ, Taylor M, Haukka J. Mortality in schizophrenia: a measurable clinical endpoint. J Psychopharmacol 2010;24:17–25. [7] Essock SM, Covell NH, Leckman-Westin E, Lieberman JA, Sederer LI, Kealey E, et al. Identifying clinically questionable psychotropic prescribing practices for Medicaid recipients in New York State. Psychiatr Serv 2009;60:1595–602. [8] Morrato EH, Druss B, Hartung DM, Valuck RJ, Allen R, Campagna E, et al. Metabolic testing rates in 3 state Medicaid programs after FDA warnings and ADA/APA recommendations for second-generation antipsychotic drugs. Arch Gen Psychiatry 2010;67:17–24. [9] Druss BG, Rosenheck RA, Desai MM, Perlin JB. Quality of preventive medical care for patients with mental disorders. Med Care 2002;40:129–36.
0
1
2
3
4
5
Years since 2005 Non-SMI
SMI
Fig. 2. Change in probability of blood pressure measurement over time, comparing visits with and without and SMI diagnosis, after adjustment for covariates.
Fig. 4. Change in probability of weight measurement by number of medications, comparing visits with and without and SMI diagnosis, after adjustment for covariates.
66
S.P. Iyer, A.S. Young / General Hospital Psychiatry 37 (2015) 60–66
Fig. 5. Change in probability of hypertension control over the age spectrum, comparing visits with and without and SMI diagnosis, after adjustment for covariates.
[21] Mangurian C, Giwa F, Shumway M, Fuentes-Afflick E, Perez-Stable EJ, Dilley JW, et al. Primary care providers' views on metabolic monitoring of outpatients taking antipsychotic medication. Psychiatr Serv 2013;64:597–9. [22] Parameswaran SG, Chang C, Swenson AK, Shumway M, Olfson M, Mangurian CV. Roles in and barriers to metabolic screening for people taking antipsychotic medications: a survey of psychiatrists. Schizophr Res 2013;143:395–6. [23] McDonell MG, Kaufman EA, Srebnik DS, Ciechanowski PS, Ries RK. Barriers to metabolic care for adults with serious mental illness: provider perspectives. Int J Psychiatry Med 2011;41:379–87. [24] Ronsley R, Raghuram K, Davidson J, Panagiotopoulos C. Barriers and facilitators to implementation of a metabolic monitoring protocol in hospital and community settings for second-generation antipsychotic-treated youth. J Can Acad Child Adolesc Psychiatry 2011;20:134–41.
Contents lists available at ScienceDirect
General Hospital Psychiatry journal homepage: http://www.ghpjournal.com
Health screening, counseling, and hypertension control for people with serious mental illness at primary care visits Sharat P. Iyer, M.D., M.S. a,b,⁎, Alexander S. Young, M.D., M.S.H.S. c,d a
VISN3 Mental Illness Research, Education and Clinical Center, James J. Peters VA Medical Center, 130 W Kingsbridge Road, Room 6A-44, Bronx, NY 10468, USA Department of Psychiatry, Icahn School of Medicine at Mount Sinai, 1 Gustave L. Levy Place, Box 1230, New York, NY 10029, USA Desert Pacific Mental Illness Research, Education and Clinical Center, West Los Angeles VA Medical Center, 11301 Wilshire Blvd, Building 210, Los Angeles, CA 90073, USA d Department of Psychiatry and Biobehavioral Sciences, University of California, Los Angeles, 10920 Wilshire Blvd, Suite 300, Los Angeles, CA 90024, USA b c
a r t i c l e
i n f o
Article history: Received 11 October 2013 Revised 19 September 2014 Accepted 7 October 2014 Keywords: Primary care Serious mental illness Health screening Health counseling Hypertension
a b s t r a c t Objective: This study sought to determine if primary care visits for people with serious mental illness (SMI) demonstrate different rates of basic physical health services compared to others, and to determine factors associated with differing rates of these measures in people with SMI. Method: The study used 2005–2010 visit-level primary care data from the National Ambulatory Medical Care Survey and the National Hospital Ambulatory Medical Care Survey. The provision of health counseling, receipt of any diagnostic or screening test, measurement of blood pressure or weight and evidence of hypertension control were assessed, adjusting for identified patient, provider and visit-level factors. Results: After adjustment for covariates, we found no significant differences between visits for people with SMI and those without for any outcome. Probability of blood pressure measurement and diagnostic or screening testing significantly increased over time. Conclusion: The lack of significant differences found here might be due to adjustment for covariates, a focus only on primary care visits, the use of visit-level data or evolution over time. Mortality differences for people with SMI may be attributable to those not receiving primary care, self-management of disease or subsets of the population requiring targeted interventions. Published by Elsevier Inc.
1. Introduction Over the last two decades, there has been increasing concern about the physical health of people with serious mental illness (SMI). People with SMI have an age-adjusted risk of mortality 2 to 2.5 times higher than the general population, with the primary causes of excess mortality being cardiovascular disease and cancer, and with rates increasing over time [1–6]. One important reason for this excess mortality may be a lack of sufficient primary care assessment and management, such as basic physical health screening, health counseling, and management of chronic medical conditions. Prior studies using Medicaid claims data, 1990s Veterans Administration data and United Kingdom primary care data have shown lower rates of basic health screening, health counseling and immunization after adjusting for basic factors [7–13]. However, Daumit et al. [14] used nationally representative data from 1993 to 1998 and showed in an unadjusted comparison between those with SMI and those without that rates of health counseling by primary care physicians were not significantly different. No study has thus far used nationally representative data to assess primary care for people with SMI since
⁎ Corresponding author at: James J. Peters, VA Medical Center (MIRECC-OOMH), 130 W. Kingsbridge Road, Room 6A-44, Bronx, NY 10468, USA. Tel.: +1 718 584 9000x3705. E-mail addresses: [email protected] (S.P. Iyer), [email protected] (A.S. Young). http://dx.doi.org/10.1016/j.genhosppsych.2014.10.003 0163-8343/Published by Elsevier Inc.
adoption of screening guidelines in 2004 [15,16]. Furthermore, no study has controlled for the wide variety of patient-level, providerlevel and system-level factors that may influence receipt of services for people with SMI. Thus, this study sought to use nationally representative data after 2005 to determine if people with SMI have different rates of basic physical health assessment, health counseling, and hypertension control in primary care settings, compared to the general population, after adjusting for a broad array of potential confounders. Furthermore, this study sought to determine what factors might contribute to differing rates of screening, health counseling and hypertension control in people with SMI. The ultimate goal was to identify potential targets for intervention to improve primary care for people with SMI, in order to influence their high rates of mortality. 2. Methods 2.1. Study design This study is a cross-sectional analysis of data from two annual US national health care surveys conducted by the Centers for Disease Control and Prevention, the National Ambulatory Medical Care Survey (NAMCS) and the National Hospital Ambulatory Medical Care Survey (NHAMCS). NAMCS contains data obtained from office-based
S.P. Iyer, A.S. Young / General Hospital Psychiatry 37 (2015) 60–66
outpatient medical providers. NHAMCS includes data from both emergency departments and hospital-based outpatient clinics. In both of these studies, specially trained interviewers complete or assist providers in completing Patient Record Forms (PRFs) for patient visits. These visit-level data are the unit of analysis for both surveys. Both studies use a multistage probability design, with probability samples collected within each stage. For NAMCS, these stages are geographical probability sampling units (PSUs), physician practices and patient visits; for NHAMCS, the stages are PSUs, hospitals, clinics and patient visits. The NAMCS data set includes sample weights for each reported visit to correlate each data point with a representative number of patient visits. These sample weights have been adjusted since 2003 to allow for single sampling stage analysis using ultimate cluster design. Prior studies have shown that these data sets can be used successfully to evaluate care for people with SMI [14,17]. In this study, we combined NAMCS data and office-based outpatient NHAMCS data — emergency department data were excluded. NAMCS and outpatient NHAMCS data are similar in structure, thus allowing the combining of these data sets. Only data from 2005 to 2010 were combined, ensuring the analysis covers data collected after the publication of screening guidelines in 2004. NAMCS and NHAMCS surveys are revised yearly, including minor changes in variable coding; all variables of interest were recoded as possible to reflect their status in 2010. Singly-imputed race and ethnicity variables were provided in the NAMCS and NHAMCS data sets themselves. Inclusion criteria for the study were visits in which consumers were at least 16 years old and in which the visit provider was indicated on the PRF to be the consumer's primary care provider. 2.2. Study variables The five primary outcome variables were evidence in a visit of the provision of any type of health counseling, the provision any diagnostic or screening test, the measurement of weight, the measurement of blood pressure and evidence of blood pressure readings below 140/90 in consumers diagnosed with hypertension. Identification of provision of health counseling was based on selection of a check box labeled as such on the PRF, or of any of several related subcategory checkboxes. Identification of provision of any diagnostic or screening test was based on any of the following being selected on the PRF: examination of the breast, foot, retina, pelvis, rectum or skin; screening for depression; any imaging test; any laboratory test; biopsy; sexually transmitted disease testing; Papanicalou testing; any scope procedure; spirometry; or urine testing. Identification of blood pressure or weight measurement was based on the presence or absence of a measurement on the PRF. Hypertension diagnosis was based on a check box labeled as such on the PRF. The primary predictor was diagnosis of SMI, defined as schizophrenia and related primary psychotic disorders (ICD-9 295.xx) or mood disorders including major depressive disorder and bipolar disorder (ICD-9 296.xx) but excluding single major depressive episodes (ICD-9 296.2x) and mild, moderate, or remitted recurrent major depression (ICD-9 296.30-296.32, 296.35-296.36). Each PRF allows for the coding of up to three ICD-9 diagnoses per visit. Covariates of interest were adapted from the Andersen Behavioral Model of Health Services Utilization, which has been used in prior studies evaluating health screening in primary care, as well as existing literature on barriers to adoption of screening for people with SMI [18–24]. Covariates used in this analysis were age (including a squared term to account for nonlinear relationships), sex, race, ethnicity, location in (or not in) a metropolitan statistical area (MSA), US geographical region (with NAMCS-defined categories of Northeast, Midwest, South and West), total number of chronic medical conditions, the Deyo modification of the Charlson comorbidity index, total number of medications, receipt of antipsychotic medications, use of tobacco, number of prior visits in the last year, having seen a physician during the patient visit, primary reason for presenting for the visit, primary type of insurance used for
61
the visit, type of practice office setting (with three categories of government-funded or federally qualified health center, private provider or clinic, and HMO, hospital-based, or faculty clinic), clinical use of electronic medical records (EMRs) and year of visit. For the outcome of weight measurement, a diagnosis of obesity was also included as a covariate, and for the outcome of blood pressure measurement, a diagnosis of hypertension was included. 2.3. Analysis This study used logistic regressions and accounted for study design using provided sampling weights. Analyses were performed using STATA 13 (College Station, TX, USA). Correlations between all predictors in the model, including all covariates and SMI diagnosis, were assessed using Pearson correlations for continuous variables and Cramer's phi for categorical variables, with no correlations higher than 0.5. Weighted univariate analysis with adjusted Wald tests was performed to compare visits for people with SMI to those without for all covariates and outcomes. Weighted unadjusted and adjusted logistic regressions were performed on the outcomes of interest, with adjustment for the covariates listed above. We used interaction effects in order to determine what factors may affect the outcomes of interest specifically in visits for people with SMI. To screen for interaction effects of interest, separate weighted adjusted logistic regressions were performed with one interaction effect included at a time for each of the identified covariates. Covariates with interaction effects demonstrating absolute coefficient t scores greater than 1.5 were identified as being potentially meaningful for each outcome of interest. Final models were then created for each outcome of interest including interaction effects for all potentially meaningful factors. Predictive margins for diagnosis of SMI were calculated using Taylor series approximations in order to demonstrate the predicted differences in outcomes for each significant interaction effect between visits for people with SMI and those without. Specification error link tests were performed for all adjusted models. 3. Results 3.1. Study population description We identified 86,901 visits meeting inclusion criteria, with 1133 visits identified as including diagnoses of SMI, representing 2.14 billion and 14.9 million nationwide visits, respectively. We identified 31,547 visits as meeting inclusion criteria and noting a diagnosis of hypertension, representing 825 million nationwide visits. Table 1 provides a description of the sample, comparing primary care visits for people with SMI to those without, after accounting for survey weighting. Factors are grouped by domains of the Andersen Behavioral Model. Visits for people with SMI, as compared to those without SMI, included individuals that were significantly younger, had more medications and were more likely to be prescribed antipsychotic medication. There were significant differences between visits for people with SMI and those without across insurance type, office setting, number of visits in the last year and reason for visit. 3.2. Impact of SMI on outcomes of interest Table 2 demonstrates the unadjusted and adjusted effect of SMI diagnosis on the outcomes of interest. The covariates for the adjusted models were all covariates noted in the methods above. In the unadjusted analyses, SMI diagnosis is associated with a significant increase in the odds of receiving health counseling and a significant decrease in the odds of receiving any diagnostic testing, blood pressure measurement and weight measurement. SMI diagnosis is not associated with hypertension control. However, the significance of the association between SMI diagnosis and all outcome findings are lost after adjusting for
62
S.P. Iyer, A.S. Young / General Hospital Psychiatry 37 (2015) 60–66
Table 1 Comparison of characteristics between primary care visits for people with and without SMI Non-SMI % or Mean Environmental factors Region of US Northeast Midwest South West MSA Office setting Government-funded Private hospital HMO, or faculty Predisposing factors Mean age Female Nonwhite race Hispanic ethnicity Mean total number of chronic conditions Mean number of medications Receipt of antipsychotic medication
SMI 95% CI
% or mMean
Adjusted Wald F test
P value
95% CI
18.2% 26% 36.8% 19% 17.3%
[15.1%, 21.2%] [21.7%, 30.4%] [32.7%, 40.9%] [15.7%, 22.4%] [9.1%, 25.5%]
20.9% 26.6% 36.1% 16.4% 14.5%
[13.5%, 28.3%] [17.5%, 35.8%] [25.9%, 46.3%] [11.4%, 21.4%] [5.9%, 23.2%]
3.7% 86.3% 10%
[2.9%, 4.5%] [84.5%, 88.2%] [8.4%, 11.6%]
9.1% 77.7% 13.1%
[5.8%, 12.5%] [72.3%, 83.1%] [9.3%, 17%]
53.27 60.7% 17.1% 11.5% 1.58 3.11 1.5%
[52.58, 53.96] [59.8%, 61.6%] [15.1%, 19.1%] [9.6%, 13.5%] [1.53, 1.64] [2.99, 3.22] [1.4%, 1.7%]
44.45 58.6% 18.4% 9.8% 1.7 3.70 46.6%
[42.20, 46.70] [52.3%, 64.8%] [12.9%, 23.8%] [6%, 13.6%] [1.57, 1.83] [3.43, 3.98] [40.1%, 53.1%]
0.43
.735
1.14 8.97
.286 b.001
60.12 0.50 0.25 0.83 3.28 22.20 58.35
b.001 .481 .619 .363 .070 b.001 b.001
17.84
b.001
4.70
.003
0.88 0.24
.348 .625
Enabling factors Type of insurance Medicaid Medicare Private or other Uninsured Number of past visits in the past year 0 visits 1–2 visits 3–5 visits 6+ visits Seen by a physician EMR
9.7% 29.6% 56% 4.7%
[8.8%, 10.7%] [27.8%, 31.3%] [54.1%, 57.8%] [4.2%, 5.2%]
24.7% 26.2% 40.7% 8.4%
[19.8%, 29.5%] [20%, 32.5%] [33.7%, 47.7%] [4.8%, 11.9%]
9.1% 29.5% 33.7% 27.6% 95.1% 43.2%
[8.6%, 9.7%] [28.5%, 30.6%] [32.8%, 34.6%] [26.3%, 29%] [94.3%, 95.9%] [39.6%, 46.8%]
6.3% 22.9% 32.5% 38.3% 93.9% 45.1%
[3.5%, 9%] [18.4%, 27.5%] [26.6%, 38.3%] [32.9%, 43.8%] [91.1%, 96.7%] [36.8%, 53.4%]
Need-related factors Reason for visit Preventive care Routine chronic problem New problem, flare-up or surgery Mean Charlson index score Tobacco use
13.94
b.001
19.5% 32.9% 47.7% 0.30 18.6%
[18.2%, 20.7%] [31.1%, 34.6%] [46.4%, 48.9%] [0.29, 0.32] [17.5%, 19.7%]
8.9% 59.3% 31.7% 0.282 46.7%
[4.9%, 13%] [51.9%, 66.8%] [26.3%, 37.2%] [0.16, 0.40] [38%, 55.4%]
0.11 40.65
.736 b.001
Outcomes Health counseling Diagnostic or screening test Blood pressure Weight Controlled hypertension
41.9% 59.8% 90.1% 87.4% 73.7%
[39.7%, 44.1%] [57.9%, 61.7%] [89.1%, 91.2%] [86.3%, 88.6%] [72.8%, 74.6%]
50.6% 49.6% 79.3% 78.6% 74.9%
[42.2%, 59.1%] [41.9%, 57.4%] [70.5%, 88.1%] [69.7%, 87.6%] [69.6%, 80.2%]
4.11 5.56 5.01 3.25 0.20
.043 .019 .025 .071 .656
confounding factors. All adjusted models demonstrated significance and all specification link tests were nonsignificant, demonstrating no misspecification. 3.3. Factors associated with the outcomes of interest in visits for people with SMI Table 3 demonstrates the effect of potentially meaningful factors on the outcomes of interest in visits for people with SMI diagnoses, as compared to those without, grouped by Andersen Behavioral Model domain. The covariates for the adjusted models were all covariates noted in the methods above, as well as potentially meaningful interaction effects. Compared to visits for people without SMI, visits for people with SMI showed an overall significantly higher odds over time (i.e., from years 2005 to 2010) of provision of any diagnostic or screening test and of blood pressure assessment, as shown in Figs. 1 and 2. In 2010, visits for people with SMI had higher predicted probabilities of provision of diagnostic or screening tests (with SMI=72.8%, without SMI= 60.0%, difference P= .034) and of provision of blood pressure assessment (with SMI=97.1%, without SMI=90.9%, difference P= .002)
compared to visits for people without SMI. Visits for people with SMI had lower predicted probabilities of receipt of blood pressure measurement in 2005 (with SMI=73.9%, without SMI=93.9%, difference P= .048). Compared to visits for people without SMI, visits for people with SMI demonstrated significantly higher odds of provision of health counseling and of hypertension control if there were six or more visits (as compared to no visits) by that individual to the same provider in the past year. Compared to visits for people without SMI, visits for people with SMI had lower odds of provision of weight measurement for those with private insurance (as compared to Medicaid) and for those with an obesity diagnosis. Similarly, visits for people with SMI had lower odds of provision of health education as the Charlson index score increased, as seen in Fig. 3. However, compared to visits for people without SMI, visits for people with SMI had higher odds of provision of weight measurement as the number of medications increased, as seen in Fig. 4. Fig. 5 shows the comparison of the effect of age on hypertension control between visits for people with and without SMI. Compared to visits for people without SMI, visits for people with SMI demonstrated significantly lower predicted probabilities of hypertension control at lower and higher ends of
S.P. Iyer, A.S. Young / General Hospital Psychiatry 37 (2015) 60–66
63
Table 2 Impact of the diagnosis of SMI on the provision of health counseling, of any diagnostic or screening test, of weight measurement, of blood pressure measurement and on hypertension control
Unadjusted models SMI, OR [95% CI] N F df_r P Adjusted models SMI, OR [95% CI] N F df_r P Link test
Health counseling
Any diagnostic or screening test
Weight
Blood pressure
Hypertension control
1.42⁎ [1.02, 1.97] 85,829 4.41 3095 .036
0.66⁎⁎ [0.49, 0.89] 86,901 7.33 3106 .007
0.53⁎ [0.32, 0.89] 86,901 5.86 3106 .016
0.42⁎⁎⁎ [0.25, 0.70] 86,901 11.20 3106 .001
1.07 [0.80, 1.42] 77,157 0.19 2730 .660
1.20 [0.78, 1.84] 42,358 18.92 2648 0.000 .932
0.71 [0.30, 1.64] 42,770 12.71 2654 0.000 .473
0.51 [0.21, 1.25] 42,770 16.98 2654 0.000 .708
0.81 [0.55, 1.20] 42,770 17.46 2654 0.000 .755
0.91 [0.43, 1.89] 16,930 6.58 1997 0.000 .526
OR=odds ratio; 95% CI=95% confidence interval; N=number of observations in sample; F=overall model F statistic; df_r=degrees of freedom; P=overall model P value; Link test=specification error link test P value. ⁎ Pb.05. ⁎⁎ Pb.01. ⁎⁎⁎ Pb.001.
the age spectrum. Regional differences were noted between visits for people with SMI and those without for receipt of health education and weight measurement. 4. Discussion This study sought to determine if people with SMI seen in primary care settings have different rates of receiving health counseling, basic physical health screening and hypertension control compared to the general population, using nationally representative data after publication of screening guidelines. After adjusting for potential confounders, we found no significant differences between visits for people with SMI and those without for any of the outcomes of interest. For the outcome of receipt of health counseling, this supports earlier work using the same data set [14]. However, these findings otherwise seem to contradict prior evidence of low rates of physical health screening and poorer performance on process quality measures for people with SMI. This may be in part due to insufficient control for confounders in prior studies, as previously identified differences might be due to unmeasured clinical, environmental, or socioeconomic factors associated with SMI diagnoses. In addition, prior studies have not always focused exclusively on primary care provider visits. However, this finding may also be explained by the increasing probability over time of providing blood pressure measurement and diagnostic or screening tests in visits for people with SMI compared to those without. Further investigation is needed to understand how this occurred and might include mechanisms such as diffusion of evidence, adoption of guidelines, improving education of primary care practitioners or direct interventions. Elucidating the means by which this change occurred may help guide future interventions to improve primary care for this population. The lack of significant differences in receipt of basic services is important to understand in juxtaposition to the disparity in mortality for people with SMI. As this study only examined care provided at primary care visits, it does not address people with SMI who do not or cannot present to primary care. It also does not address potential disparities in self-management of chronic diseases between visits. In addition, it is unknown if primary care services moderate the excess mortality risk in this population. Furthermore, if the mortality disparity for people with SMI is due to factors inherent to their illnesses, then this vulnerable population may require more services than the general population, rather than only equivalent services. The findings of significantly higher probabilities in 2010 of provision of blood pressure measurement and diagnostic and screening tests for people with SMI compared to those
without are therefore reassuring. However, we did not find similar differences in weight measurement or health counseling. Understanding differences between these services in this population can help guide future primary care-based interventions for people with SMI. A number of different factors affected the receipt of basic services and hypertension control in visits for people with SMI, as compared to visits for others, and may point to avenues for intervention development. Lower probabilities of hypertension control at lower and higher ends of the age spectrum may point to a need for targeted interventions focused on these age groups, in conjunction with existing efforts to develop integrated service models for new-onset mental illness and geriatric populations. The higher probability of health counseling and controlled hypertension seen with higher numbers of visits over the last year may indicate a benefit to care models allowing higherfrequency, lower-intensity primary care contacts for people with SMI. The lower probability of weight measurement in primary care visits for people with SMI holding private insurance, as compared to Medicaid, may indicate a need for improved dissemination of guidelines beyond public settings. However, the lower probability of health counseling in visits for people with higher Charlson index scores and SMI diagnoses and the lower probability of weight measurement in visits for people obesity and SMI diagnoses are particularly concerning. These findings, as well as findings of regional differences in outcome, require further study. This study has a number of limitations, particularly because the data sets used provide visit-level data only. Some identified confounding patient-level and provider-level factors identified in the Andersen Model and the literature could not be included in this analysis. In particular, the NAMCS and NHAMCS data sets have insufficient capture of consumer socioeconomic status and health knowledge, and of provider education and attitudes toward people with SMI. These omitted variables may have affected the analyses performed, potentially masking significant differences between visits for people with and without SMI. Furthermore, the variables used are based on provider report or CDC staff chart abstraction and are limited by the necessary simplicity of the PRF. Variables on the PRF often do not have precise definitions and are open to interpretation. Therefore, the recorded variables may not fully represent the consumer's status or the content of the visit, such as the actual content of health counseling provided, or the distinction between blood pressure and weight measurement for screening versus diagnostic assessment. The determination of SMI diagnosis relies on a related diagnosis being listed in three available spaces for diagnosis on the PRF. Providers may not have coded an SMI diagnosis if it was not a
64
S.P. Iyer, A.S. Young / General Hospital Psychiatry 37 (2015) 60–66
Table 3 Factors associated with provision of health counseling, of any diagnostic or screening test, of weight measurement and of blood pressure measurement and factors associated with hypertension control, in visits for people with SMI and with results in odds ratios [95% confidence intervals] Potentially meaningful interactions
Health counseling
Environmental factors Years since 2005 Region of US Northeast (baseline) Midwest 1.13 [0.51, 2.49] South 2.12⁎ [1.00, 4.48] West 1.35 [0.49, 3.68] MSA Office setting Government-funded clinic (baseline) Private clinic HMO, hospital, faculty clinic Predisposing factors Hispanic Age Age squared Antipsychotic prescription Number of chronic conditions Number of meds Enabling factors Insurance Medicaid (baseline) Medicare Private insurance Uninsured Number of visits in the past year 0 visits (baseline) 1–2 visits 3–5 visits 6+ visits Saw a physician EMR Need-related factors Reason for visit Preventative care (baseline) Chronic problem New problem or flare-up Charlson score Tobacco use Diagnosis of obesity N F df_r P Link test
Any diagnostic or screening test Blood pressure 1.33⁎ [1.02, 1.72]
Weight
Normal blood pressure if hypertensive
1.96⁎⁎ [1.27, 3.01]
5.28⁎ [1.22, 22.77] 1.08 [0.24, 4.90] 4.49 [0.92, 22.02] 3.16 [0.69, 14.43]
0.27 [0.07, 1.03] 0.46 [0.08, 2.64]
0.34 [0.10, 1.11]
1.38⁎⁎ [1.14, 1.68] 1.00⁎⁎ [0.99, 1.00] 1.18 [0.42, 3.27] 1.37⁎ [1.02, 1.84]
2.09 [0.88, 4.96] 1.42 [0.58, 3.48] 0.87 [0.23, 3.34]
0.28 [0.06, 1.26] 0.72 [0.17, 3.06] 0.17 [0.03, 1.06]
0.14 [0.02, 1.24] 0.11⁎ [0.02, 0.66] 0.14 [0.02, 1.06]
3.62 [0.78, 16.75] 2.95 [0.78, 11.17] 5.01⁎ [1.30, 19.41]
0.86 [0.13, 5.58] 8.62 [0.92, 80.68] 7.06⁎ [1.35, 36.88] 0.79 [0.18, 3.50] 2.01 [0.53, 7.59]
0.61⁎⁎ [0.42, 0.87] 0.89 [0.49, 1.62]
42,358 15.11 2648 .000 .780
0.54 [0.26, 1.10]
1.09 [0.35, 3.38] 3.16⁎ [1.08, 9.28]
0.25 [0.03, 1.91] 1.63 [0.13, 20.59] 1.33 [0.57, 3.10]
42,770 16.97 2654 .000 .867
0.81 [0.21, 3.20]
42,770 14.71 2654 .000 .509
0.16⁎⁎ [0.04, 0.58] 42,770 14.14 2654 .000 .448
16,930 7.25 1997 .000 .459
N=number of observations in sample; F=overall model F statistic; df_r=degrees of freedom; P=overall model P value; Link test=specification error link test P value. ⁎ Pb.05. ⁎⁎ Pb.01.
prominent part of the visit's content or the consumer's presentation, or if multiple other diagnoses were more prominent. Thus, not all visits for people with SMI may have been identified in this study, and some analyses may have been underpowered. People with SMI likely present to primary care late in the course of their chronic medical conditions, due to insurance coverage, selfstigma, and cognitive impairment. However, direct assessment of the severity of physical illnesses was not available in this data set in sufficient granularity. While prior studies have used the Charlson index to adjust for severity of medical comorbidity using NAMCS and NHAMCS data, this index may have been an imperfect proxy in this population and may have underestimated true medical severity. The ultimate goal of this study was to address the disparity in chronic physical health mortality for people with SMI. The findings of this study did not demonstrate a similar disparity in basic services provided in primary care for this population, and found improvements in provision of certain services over time. Nevertheless, certain subsets of visits for people with SMI demonstrated lower probabilities of receipt of
services, including visits for people with higher medical severity, lower visit frequency, older or younger age and private insurance. Determining how mortality risk correlates with these subsets of people with SMI, as well as how lack of primary care may affect risk of death for people with SMI, will help to better identify the precise means to reduce the excess mortality of this vulnerable population. Acknowledgments This material is the result of work supported with resources and the use of facilities at the James J. Peters VA Medical Center, Bronx, NY, and the West Los Angeles VA Medical Center, Los Angeles, CA. Additional support was provided by the Mount Sinai Conduits Institutes for Translational Sciences. Assistance for this article was provided by Kenneth Wells, M.D., M.P.H., and Lisa Dixon, M.D., M.P.H., who reviewed the manuscript and contributed to the discussion, and Alan Weinberg, M.S., who contributed to the analytic plan. Funding for Dr. Parameswaran was provided by the Robert Wood Johnson Foundation Clinical Scholars
.4
.6
.8
1
65
.2
Probability of diagnostic test or screening
S.P. Iyer, A.S. Young / General Hospital Psychiatry 37 (2015) 60–66
0
1
2
3
4
5
Years since 2005 Non-SMI
SMI
Fig. 1. Change in probability of diagnostic or screening testing over time, comparing visits with and without and SMI diagnosis, after adjustment for covariates.
Fig. 3. Change in probability of receipt of health education by Charlson index score, comparing visits with and without and SMI diagnosis, after adjustment for covariates.
program and the US Department of Veterans Affairs. The views expressed in this article are those of the authors and do not necessarily reflect the position or policy of the Department of Veterans Affairs or the United States government.
[10] Kaplowitz RA, Scranton RE, Gagnon DR, Cantillon C, Levenson JW, Sesso HD, et al. Health care utilization and receipt of cholesterol testing by veterans with and those without mental illness. Gen Hosp Psychiatry 2006;28: 137–44. [11] Desai MM, Rosenheck RA, Druss BG, Perlin JB. Receipt of nutrition and exercise counseling among medical outpatients with psychiatric and substance use disorders. J Gen Intern Med 2002;17:556–60. [12] Roberts L, Roalfe A, Wilson S, Lester H. Physical health care of patients with schizophrenia in primary care: a comparative study. Fam Pract 2007;24:34–40. [13] Osborn DP, Baio G, Walters K, Petersen I, Limburg H, Raine R, et al. Inequalities in the provision of cardiovascular screening to people with severe mental illnesses in primary care: cohort study in the united kingdom thin primary care database 2000–2007. Schizophr Res 2011;129:104–10. [14] Daumit GL, Pratt LA, Crum RM, Powe NR, Ford DE. Characteristics of primary care visits for individuals with severe mental illness in a national sample. Gen Hosp Psychiatry 2002;24:391–5. [15] American Diabetes Association, American Psychiatric Association, American Association of Clinical Endocrinologists, North American Associat0069on for the Study of Obesity. Consensus development conference on antipsychotic drugs and obesity and diabetes. Diabetes Care 2004;27:596–601. [16] Marder SR, Essock SM, Miller AL, Buchanan RW, Casey DE, Davis JM, et al. Physical health monitoring of patients with schizophrenia. Am J Psychiatry 2004;161:1334–49. [17] Rost K, Hsieh YP, Xu S, Menachemi N, Young AS. Potential disparities in the management of schizophrenia in the United States. Psychiatr Serv 2011;62:613–8. [18] Andersen RM. Revisiting the behavioral model and access to medical care: does it matter? J Health Soc Behav 1995;36:1–10. [19] Brennenstuhl S, Fuller-Thomson E, Popova S. Prevalence and factors associated with colorectal cancer screening in Canadian women. J Womens Health (Larchmt) 2010;19:775–84. [20] Rahman SM, Dignan MB, Shelton BJ. A theory-based model for predicting adherence to guidelines for screening mammography among women age 40 and older. Int J Cancer Prev 2005;2:169–79.
References
1 .9 .8 .7 .6 .5
Probability of blood pressure measurement
[1] Saha S, Chant D, McGrath J. A systematic review of mortality in schizophrenia: is the differential mortality gap worsening over time? Arch Gen Psychiatry 2007;64:1123–31. [2] Brown S, Kim M, Mitchell C, Inskip H. Twenty-five year mortality of a community cohort with schizophrenia. Br J Psychiatry 2010;196:116–21. [3] Crump C, Winkleby MA, Sundquist K, Sundquist J. Comorbidities and mortality in persons with schizophrenia: a Swedish national cohort study. Am J Psychiatry 2013;170:324–33. [4] Hoang U, Stewart R, Goldacre MJ. Mortality after hospital discharge for people with schizophrenia or bipolar disorder: retrospective study of linked English hospital episode statistics, 1999–2006. BMJ 2011;343:d5422. [5] Talaslahti T, Alanen HM, Hakko H, Isohanni M, Hakkinen U, Leinonen E. Mortality and causes of death in older patients with schizophrenia. Int J Geriatr Psychiatry 2012;27:1131–7. [6] Bushe CJ, Taylor M, Haukka J. Mortality in schizophrenia: a measurable clinical endpoint. J Psychopharmacol 2010;24:17–25. [7] Essock SM, Covell NH, Leckman-Westin E, Lieberman JA, Sederer LI, Kealey E, et al. Identifying clinically questionable psychotropic prescribing practices for Medicaid recipients in New York State. Psychiatr Serv 2009;60:1595–602. [8] Morrato EH, Druss B, Hartung DM, Valuck RJ, Allen R, Campagna E, et al. Metabolic testing rates in 3 state Medicaid programs after FDA warnings and ADA/APA recommendations for second-generation antipsychotic drugs. Arch Gen Psychiatry 2010;67:17–24. [9] Druss BG, Rosenheck RA, Desai MM, Perlin JB. Quality of preventive medical care for patients with mental disorders. Med Care 2002;40:129–36.
0
1
2
3
4
5
Years since 2005 Non-SMI
SMI
Fig. 2. Change in probability of blood pressure measurement over time, comparing visits with and without and SMI diagnosis, after adjustment for covariates.
Fig. 4. Change in probability of weight measurement by number of medications, comparing visits with and without and SMI diagnosis, after adjustment for covariates.
66
S.P. Iyer, A.S. Young / General Hospital Psychiatry 37 (2015) 60–66
Fig. 5. Change in probability of hypertension control over the age spectrum, comparing visits with and without and SMI diagnosis, after adjustment for covariates.
[21] Mangurian C, Giwa F, Shumway M, Fuentes-Afflick E, Perez-Stable EJ, Dilley JW, et al. Primary care providers' views on metabolic monitoring of outpatients taking antipsychotic medication. Psychiatr Serv 2013;64:597–9. [22] Parameswaran SG, Chang C, Swenson AK, Shumway M, Olfson M, Mangurian CV. Roles in and barriers to metabolic screening for people taking antipsychotic medications: a survey of psychiatrists. Schizophr Res 2013;143:395–6. [23] McDonell MG, Kaufman EA, Srebnik DS, Ciechanowski PS, Ries RK. Barriers to metabolic care for adults with serious mental illness: provider perspectives. Int J Psychiatry Med 2011;41:379–87. [24] Ronsley R, Raghuram K, Davidson J, Panagiotopoulos C. Barriers and facilitators to implementation of a metabolic monitoring protocol in hospital and community settings for second-generation antipsychotic-treated youth. J Can Acad Child Adolesc Psychiatry 2011;20:134–41.
