The diagnosis of schizophrenia among nursing home residents with ADRD: Does race matter?

Abstract

Objective:

To examine racial differences in the frequency of schizophrenia diagnosis codes used among nursing home (NH) residents with Alzheimer’s Disease and Related Dementias (ADRD), pre and post the implementation of public reporting of antipsychotic use in NHs).

Methods:

The 2011–2017 Minimum Data Set and Medicare Master Beneficiary Summary File were linked. We identified long-stay NH residents (i.e. those who had quarterly or annual assessments) with ADRD aged 55 years and older (N=7,734,348). Outcome variable was defined as the diagnosis of schizophrenia documented in the MDS assessments. Main variables of interest included individual race (Black versus White), the percent of Blacks in a NH and time trend. Multivariate regressions were estimated.

Results:

The frequency of schizophrenia diagnosis codes among NH residents with ADRD steadily increased over the study period, and Blacks experienced a greater increase than their White counterparts. For example, the overall likelihood of having schizophrenia diagnosis increased 1.9 percentage points (95% confidence interval [CI]:[0.019, 0.020], P<0.01) from 2011 to 2017 among Whites, while Blacks had an addition 1.3 percentage points increase (95%CI: [0.011, 0.015], P<0.01). The increase in the likelihood of having schizophrenia diagnosis code was higher in NHs with higher percent of Blacks: the increase from 2011 to 2017 was 2.6 percentage point (95%CI: [0.023, 0.029], P<0.01) higher in NHs with the highest percent of Blacks, compared to NHs with lowest percent of Blacks. Racial differences in the growth of schizophrenia diagnosis also existed within a NH after accounting for NH factors.

Conclusion:

Following the implementation of public reporting of antipsychotic use in NH, Black residents experienced a greater increase in the likelihood of having schizophrenia diagnosis than White NH residents. NHs with a higher proportion of Blacks had a greater increase in schizophrenia diagnosis, and Blacks experienced an increased likelihood of schizophrenia diagnosis than Whites within a NH. Further research is need to determine a causal relationship between the federal policy mandating public reporting and disparities in schizophrenia diagnostic coding.

INTRODUCTION

Antipsychotic medications are commonly used among nursing home (NH) residents with Alzheimer’s Disease and Related Dementias (ADRD) to address dementia related behavioral problems.^1,2 Due to concerns about the adverse events associated with antipsychotic use among this population,^3–6 the Centers for Medicare and Medicaid Services (CMS) started to publicly report antipsychotic use in NHs in July 2012aiming to reduce it.⁷ In this antipsychotic reduction initiative, CMS excludes residents with the code for schizophrenia diagnosis when reporting the prevalence of antipsychotic use.⁸ In other words, NHs are not penalized for using antipsychotics if a resident has the schizophrenia diagnosis code.

Although there has been a steady decrease in antipsychotic use in NHs,^9,10 it is unclear whether it indeed reflects an improvement in dementia care or is accompanied by other changes that may not necessarily result in better care. For example, a newly released New York Times report (09/11/2021), “Phony Diagnoses Hide High Rates of Drugging at Nursing Homes”, alerted to the increasing trend in unjustified diagnosing of schizophrenia in NHs in recent years.¹¹ Such concern was shared by other recent reports, which questioned whether the motivation to reduce the publicly reported antipsychotic use in NHs could be associated with increased coding of schizophrenia diagnosis.^12,13 Indeed, some studies of the impact of CMS’ public reporting on other quality measures have suggested that NHs may respond to public reporting by changing their practices or gaming the data, leading to unintended consequences for the residents.¹⁴ Thus, NHs may be incentivized to code schizophrenia in order to use antipsychotics, which for patients with ADRD, if not justified by clinical criteria, may lead to inappropriate dementia management and poor quality of care.

The concern over this diagnosis, and the related quality of care, may be even greater among Black NH residents with ADRD than Whites. Studies have suggested that Blacks are more likely to be diagnosed with schizophrenia,^15,16 and that the misdiagnosis of schizophrenia for other mental illness are more likely among Blacks.^7,17–19 Among those with ADRD, Blacks may have different care needs than Whites – for example, they tend to have a higher level of cognitive impairment and are more likely to develop behavioral issues.^20–24,25 Moreover, Blacks have been found to receive lower quality of care compared to their White counterparts in NHs. For example, Blacks are more likely to reside in NHs with inadequate staffing and lower quality, which may not have sufficient resources to provide appropriate care to address behavioral issues among residents with ADRD.^26–28 Thus, the public reporting of antipsychotics in NHs may have different impacts on Black compared to White residents with ADRD.

Despite the recent concerns and findings, the extent of the increase in the diagnosis of schizophrenia and its variation between for Black and White residents with ADRD are not clear. Therefore, the main objectives of this study are to: a) examine whether the increase in the diagnosis of schizophrenia among NH residents with ADRD varies as a function of race; and b) identify potential sources of such differences. Specifically, we examined whether the increase in the diagnosis of schizophrenia differed between NHs based on the racial composition of their residents and/or occurs within NHs.

METHODS

Data.

We linked the Minimum Data Set (MDS3.0) with the Master Beneficiary Summary File (MBSF) for calendar years 2011–2017. The MDS is an assessment tool conducted for all residents in Medicare- and/or Medicaid- certified NHs. The MDS assessments contain information on mental health disorder diagnosis (e.g. schizophrenia and ADRD) for NH residents. The MBSF contains basic information on individuals’ demographics, including race, gender and age. The MBSF chronic condition file contains information on a set of chronic conditions (including ADRD).

Cohort.

The study cohort consisted of Black and White NH residents with ADRD who were 55 years or older and who were long-stay residents. We did not include residents with other race or ethnicity because they accounted for a small proportion of the sample (4.5%). The identification of ADRD was based on both the MDS (diagnosis checkboxes and International Classification of Diseases version 9 [ICD-9] or ICD-10) and the MBSF chronic condition segment. We defined NH long-stay residents as those who had MDS annual, quarterly, or significant correction to a prior quarterly assessment. In general, residents with these types of assessments have stayed in NHs for at least 90 days. For residents with more than one qualified assessments in a given year, one was randomly selected each year as the “target” assessment. In total, the analytical sample included 7,734,348 person-years over the 2011–2017 time period (2,965,096 unique individuals).

Variables.

The outcome variable was defined as whether a resident with ADRD had the schizophrenia diagnosis code, based on the “target” MDS assessment. We used the MDS to determine the diagnosis of schizophrenia because this is the approach CMS uses to exclude residents when constructing the antipsychotic use measure. The key independent variables of interest included individuals’ race (i.e. Black or White), racial composition of a NH, and year indicators. Racial composition of a NH was calculated as the proportion of Black residents among all NH residents in a given year, based on the MDS. We then categorized NHs into quartiles based on the distribution of NH racial composition, with homes in the first quartile serving the fewest Black residents, and NHs in the 4th quartile serving the highest percentage. A set of year indicators were generated to capture the secular trend in the prevalence of schizophrenia among residents with ADRD. We combined year 2011 and the first two quarters of year 2012 and used it as the “baseline” period in our analysis because the CMS started to publicly report antipsychotic use on July 1^st, 2012. We then combined the last two quarters of year 2012 with 2013, and generated a set of year indicator for the rest of the years (i.e. 20140–2017).

We included individual age and gender as covariates in the regression analyses. We did not include additional individual characteristics (e.g. behavioral symptom or cognitive functional status) in the regression analyses because they were not likely to be clinical risk factors for the onset of schizophrenia.

Statistical Analysis.

We first compared the proportion of residents with schizophrenia diagnosis codes for Blacks and Whites with ADRD, as well as across NHs with different racial compositions. We then examined NH characteristics across the four groups of NHs with different racial compositions. To explore the relationship between race and the likelihood of having schizophrenia diagnosis codes, we estimated a set of linear probability models with robust standard errors among NH residents with ADRD. We used the linear probability model because it was computationally efficient, given the study’s large sample size, and allowed for the direct interpretation of the β coefficients (e.g. difference in the probability of having schizophrenia diagnosis between Blacks and Whites), especially the interaction terms.²⁹ More specifically, we first estimated a “base” model with individual race (Black versus White), age, gender, year indicators, and the interactions between year indicators and race. The effect of race from this model captured the overall racial differences in the likelihood of having schizophrenia diagnosis codes at the baseline year (i.e. 2011 and the first two quarters of 2012), and the interactions between year and race reflected changes in racial differences in the likelihood of having schizophrenia diagnosis codes over time. The overall racial differences may arise from differences within a NH or variations across NHs with different proportions of Black residents.

We then estimated a model with NH fixed-effects and added an additional interaction between NH racial composition (%Blacks) and year indicators. The estimates of individual race from the fix-effects model captured differences in the likelihood of having schizophrenia diagnosis between Blacks and Whites within a NH, accounting for time-invariant facility level factors. The interaction terms between NH racial composition and year indicators represented the changes in the likelihood of having schizophrenia diagnosis over time at the NH level. Although the fixed-effects model provided consistent estimators (e.g. within-NH racial differences), and accounted for time-invariant NH effects without assumptions about the distribution of NH effects or the correlation between NH effects and other independent variables, the model was not able to capture the overall relationship between NH racial composition and the likelihood of schizophrenia if there was no or very small changes in NH characteristics over time.

Thus, we also estimated a model with NH random-effects to explore variations in the likelihood of having schizophrenia diagnosis codes across NHs with different racial compositions. Other NH facility characteristics, such as NH quality of care, were not included because they could be related with NH racial composition, and potentially “dilute” the overall relationship between NH racial composition and the diagnosis of schizophrenia among residents with ADRD. The limitation of the random-effects model was that it required the assumption of independence between unobserved facility characteristics and other independent variables. Otherwise the estimators would be biased.

For ease of the interpretation, we also calculated the average adjusted probabilities of having schizophrenia diagnosis for Blacks and Whites with ADRD over time, with and without accounting for NH level factors. The overall adjusted probabilities of schizophrenia for Blacks and Whites were calculated on the base model. The average adjusted probabilities of schizophrenia for each subgroups of NHs were calculated based on the random-effects model.

Lastly, to check the robustness of the findings, we performed two sets of sensitivity analyses. First, for those residents who were in NHs for more than one year, we randomly selected one of the years and repeated the analyses to avoid the concern of within-individual correlations. Second, we repeated the analyses by restricting the study cohort to those who were 65 years and older as they accounted for the majority of NH population. The statistical analyses were perform by STATA 16. This study was approved by the institutional review board (IRB) at the University of Rochester.

RESULTS

Descriptive analyses

Among the long-stay NH residents with ADRD, about 14% were Black and 86% were White. The overall frequency of schizophrenia diagnosis codes in this cohort was 10.62% for Blacks and 5.75% for Whites. In Table 1 we compared the proportion of residents with schizophrenia diagnosis codes by individual race and by the NH proportion of Blacks.

Table-1:

Proportion of residents with schizophrenia diagnosis, individual and NH characteristics, by individual race and NH racial compsotion

	Individual level		NH subgroups (quartile of %Blacks in a NH)[a]
	White (N=6,636,423)	Black (N=1,097,925)	%Black - quartile 1 (N=26,709)	%Black - quartile 2 (N=26,739)	%Black - quartile 3 (N=26,723)	%Black - quartile 4 (N=26,709)
Age	83.02(9.01)	78.22(11.05)
Male	29.15%	38.60%
Proportion of residents with schizophrenia diagnosis
2011 + 2012Q1,2	4.98%	9.16%	2.59%	3.24%	5.39%	9.59%
2012 Q2,3+2013	5.20%	9.81%	2.69%	3.36%	5.63%	10.21%
2014	5.55%	10.42%	2.75%	3.79%	5.99%	10.82%
2015	5.90%	10.92%	2.85%	4.13%	6.51%	11.26%
2016	6.56%	11.82%	3.21%	4.56%	7.44%	12.21%
2017	7.20%	12.78%	3.34%	5.09%	8.34%	13.19%
Selected NH characteristics
Star rating: 4 or 5 starts			54.27%	48.50%	40.57%	32.38%
For-profit			47.13%	64.37%	73.24%	76.79%
CNA hours per resident per day			2.60 (0.64)	2.48 (0.62)	2.43 (0.59)	2.37 (0.58)
RN hours per resident per day			0.63 (0.31)	0.55 (0.25)	0.49 (0.23)	0.44 (0.22)
%Medicaid			54.29%	55.58%	61.70%	70.68%

^[a]The numbers of NHs represent NH-year

Between 2011 and 2017, the proportion of residents with schizophrenia diagnosis codes increased from 4.98 % to 7.21% (i.e. 2.23 percentage points) among Whites, and from 9.16% to 12.78% (i.e. 3.62 percentage points) among Blacks. NHs with a higher proportion of Blacks appeared to have a higher proportion of residents with schizophrenia diagnosis and greater increase over time: from 2.59% to 3.34% in NHs with the lowest percent of Blacks and from 9.59% to 13.19% in NHs with the highest percent of Blacks. In addition, NHs serving a high proportion of Blacks were less likely to have 4 or 5 star quality ratings, had lower CNA and RN staffing, had higher percent of Medicaid residents, and were more likely to be for-profit facilities, compared to those serving a low proportion of Blacks.

Overall racial differences in the diagnosis of schizophrenia

The base model in Table-2 and Figure 1.a captured the overall racial differences in the likelihood of having schizophrenia diagnosis and the variation over time. As illustrated in Figure 1.a, the likelihood of schizophrenia among residents with ADRD increased over time, and this increase was greater among Blacks. For example, as presented in the base model, the overall likelihood of having schizophrenia diagnosis was 1.4 percentage points higher for Blacks than Whites (95% CI: [0.013, 0.016], P<0.01, based on Heteroskedasticity-robust t statistics) prior to the public reporting (2011 and the first two quarters of 2012). The likelihood of having schizophrenia diagnosis increased 1.9 percentage points from 2011 to 2017 among Whites (95% CI: [0.019, 0.020], P<0.01), while Blacks experienced an additional 1.3 percentage point increase, compared with Whites, over this period (i.e. captured by the interaction term between Black and year 2017, 95% CI:[0.011, 0.015], P<0.01).

Table-2:

Results from regression on the likelihood of having schizophrenia diagnosis, race, NH racial composition and time

	Base model	Fixed-effects model	Random-effects model
Racial differences at the baseline (2011 and the first two quarters of 2012)
Black (White as reference)	0.014*** (0.013 – 0.016)	−0.008*** (−0.011 – −0.006)	−0.008*** (−0.010 – −0.006)
Time trend (reference: 2011 and the first two quarters of 2012)
Year: 2012Q3Q4+ 2013	0.001*** (0.000 – 0.001)	0.001* (−0.000 – 0.001)	0.001* (−0.000 – 0.001)
Year: 2014	0.004*** (0.004 – 0.005)	0.002*** (0.001 – 0.003)	0.002*** (0.001 – 0.003)
Year: 2015	0.007*** (0.007 – 0.008)	0.004*** (0.003 – 0.005)	0.004*** (0.003 – 0.005)
Year: 2016	0.013*** (0.013 – 0.014)	0.007*** (0.006 – 0.008)	0.007*** (0.006 – 0.008)
Year: 2017	0.019*** (0.019 – 0.020)	0.008*** (0.006 – 0.009)	0.008*** (0.007 – 0.009)
Changes in racial differences within a NH over time
Black×(2012Q3Q4+ 2013)	0.003*** (0.002 – 0.005)	0.003*** (0.001 – 0.005)	0.003*** (0.001 – 0.005)
Black×2014	0.006*** (0.004 – 0.008)	0.004*** (0.002 – 0.006)	0.004*** (0.002 – 0.006)
Black×2015	0.007*** (0.005 – 0.009)	0.005*** (0.003 – 0.007)	0.005*** (0.003 – 0.007)
Black×2016	0.009*** (0.007 – 0.011)	0.005*** (0.002 – 0.007)	0.005*** (0.002 – 0.007)
Black×2017	0.013*** (0.011 – 0.015)	0.005*** (0.003 – 0.008)	0.005*** (0.003 – 0.008)
Differences across NH by race composition (reference: NH in Quartile 1:0.04%)
%Black in NHs (Quantile 2: 1.55%)		−0.001[a] (−0.002 – 0.001)	0.001** (0.000 – 0.003)
%Black in NHs (Quantile 3:7.19%)		−0.001 (−0.003 – 0.001)	0.007*** (0.005 – 0.009)
%Black in NHs (Quantile 4: 34.91%)		−0.002 (−0.006 – 0.001)	0.015*** (0.012 – 0.018)
Changes across NHs over time
%Black in NHs (Quartile 2)×(2012Q3Q4+ 2013)		−0.000 (−0.001 – 0.001)	−0.000 (−0.001 – 0.001)
%Black in NHs (Quartile 2)×(2014)		0.002*** (0.001 – 0.003)	0.002*** (0.001 – 0.003)
%Black in NHs (Quartile 2)×(2015)		0.003*** (0.001 – 0.004)	0.003*** (0.001 – 0.004)
%Black in NHs (Quartile 2)×(2016)		0.004*** (0.002 – 0.006)	0.004*** (0.002 – 0.006)
%Black in NHs (Quartile 2)×(2017)		0.008*** (0.007 – 0.010)	0.008*** (0.006 – 0.010)
%Black in NHs (Quartile3)×(2012Q3Q4+ 2013)		0.001 (−0.000 – 0.002)	0.001 (−0.000 – 0.002)
%Black in NHs (Quartile 3)×(2014)		0.003*** (0.002 – 0.004)	0.003*** (0.001 – 0.004)
%Black in NHs (Quartile 3)×(2015)		0.005*** (0.004 – 0.007)	0.005*** (0.004 – 0.007)
%Black in NHs (Quartile 3)×(2016)		0.010*** (0.008 – 0.012)	0.010*** (0.008 – 0.012)
%Black in NHs (Quartile 3)×(2017)		0.018*** (0.016 – 0.021)	0.018*** (0.015 – 0.020)
%Black in NHs (Quartile4)×(2012Q3Q4+ 2013)		0.003*** (0.001 – 0.004)	0.003*** (0.001 – 0.004)
%Black in NHs (Quartile 4)×(2014)		0.007*** (0.005 – 0.009)	0.006*** (0.005 – 0.008)
%Black in NHs (Quartile 4)×(2015)		0.010*** (0.008 – 0.012)	0.009*** (0.007 – 0.012)
%Black in NHs (Quartile 4)×(2016)		0.017*** (0.014 – 0.020)	0.017*** (0.014 – 0.019)
%Black in NHs (Quartile 4)×(2017)		0.027*** (0.024 – 0.030)	0.026*** (0.023 – 0.029)
Age	−0.006*** (−0.006 – −0.006)	−0.004*** (−0.004 – −0.004)	−0.004*** (−0.004 – −0.004)
Male	−0.009*** (−0.009 – −0.008)	−0.019*** (−0.020 – −0.018)	−0.018*** (−0.019 – −0.018)
Intercept	0.557*** (0.555 – 0.559)	0.417*** (0.411 – 0.422)	0.417*** (0.410 – 0.423)
Observations	7,734,348	7,734,348	7,734,348
Number of NHs		16,888	16,888

95% Confidence Interval in parentheses, based on the robust standard errors.

^***p<0.01,

^**p<0.05,

^*p<0.1.

Heteroskedasticity-robust t statistics were used to determine the P value.

^[a]This set of coefficients captured the effect of within-NH changes in racial composition on outcome. However, as the within-NH changes was very small, this set of coefficients were not significant.

Figure 1.

Racial differences in the probabilities of having schizophrenia diagnosis over time

Figure 1.a was generated based on base model; and Figure 1.b was generated based on random-effects model. Quartile-1 includes NHs with the loweset propo

Racial differences in the likelihood of having schizophrenia diagnosis within a NH

The results from the fixed-effects model in Table-2 represent racial differences in the likelihood of having schizophrenia diagnosis within a NH. Unlike the findings from the base model, Blacks were 0.8 percentage point (95% CI:[−0.011, −0.006], P<0.01) less likely to have the diagnosis of schizophrenia than their White counterparts within the same NH prior to the public reporting. This suggests that the overall higher prevalence of schizophrenia diagnosis detected in the base model was mainly due to the variations across NHs.

At the same time, we observed an increasing trend in the likelihood of schizophrenia diagnosis over time within a NH, and this trend was greater among Blacks with ADRD. For example, from 2011 to 2017, the likelihood of schizophrenia increased 0.8 percentage points for Whites with ADRD, and Blacks had an additional 0.5 percentage points (95% CI; [0.003, 0.008], P<0.01) increase than their White counterparts within the same facility. However, the magnitude of the increase in the likelihood of schizophrenia diagnosis, and the related racial differences, appeared to be smaller than those estimated from the base model, further pointing to the role that NHs play in schizophrenia diagnosis coding.

The likelihood of schizophrenia across NH with different racial composition.

The random-effects model in Table-2 captured the relationship between NH-level proportion of Blacks and the likelihood of schizophrenia diagnosis. As shown in Table-2, residents in NHs with a higher percent of Blacks generally had higher likelihood of schizophrenia diagnosis. For example, prior to the public reporting, residents in NHs with the highest percent of Blacks had 1.5 percentage points (95% CI: [0.012–0.012], P<0.01) higher likelihood of having schizophrenia diagnosis. Furthermore, the model also suggested that the extent of growth in the likelihood of schizophrenia was greater in NHs with a higher percent of Blacks. For example, from 2011 to 2017 the growth in the likelihood of schizophrenia diagnosis was 0.8 percentage point (95% CI:[0.007,0.009], P<0.01) in NHs with the lowest percent of Blacks, and NHs with the highest percent of Blacks experienced an additional 2.6 percentage points increase, as compared to those with the NHs with the lowest percent of Blacks (95% CI [0.023–0.029], P<0.01). These results, together with the findings from the base and the fixed-effects models, suggested the important role of NHs in the likelihood of having schizophrenia diagnosis.

Lastly, for the ease of presentation, we graphed the probabilities of having schizophrenia diagnosis for these 4 groups of NHs over the study period in Figure 1b, based on the results from the random-effects model. The main findings from the sensitivity analyses (Appendix 1&2) were consistent with our findings from the main analyses.

DISCUSSION:

In this study, we examined racial differences in the likelihood of schizophrenia diagnosis codes among NH residents with ADRD, following the implementation of CMS’ public reporting of antipsychotics. We found that the proportion of residents with schizophrenia diagnosis codes steadily increased over the study period, and that Blacks experienced a greater increase in the likelihood of having schizophrenia codes than their White counterparts. Residents with ADRD who lived in facilities with a higher proportion of Blacks were more likely to have the diagnosis code of schizophrenia. Additionally, among residents of the same NH, increase in schizophrenia diagnosis was greater over time for Blacks than for their White counterparts.

The observed increase in schizophrenia diagnosis among NH residents with ADRD is not supported by existing epidemiological evidence.³⁰ Furthermore, our findings suggest the extent of increase in schizophrenia diagnosis codes vary with the racial composition of a NH. Studies have suggested that older adults with ADRD are likely to develop behavioral and psychological symptoms of dementia (BPSD).^20–24,25 Although non-pharmacological approaches are generally recommended to treat behavioral problems among older adults with ADRD,^31–33 these approaches can be labor-intensive and potentially costly.^34–38 For example, NHs may need to modify their environment, invest in having professionals with training in mental health, or train staff to better communicate with residents with ADRD and to identify and manage the underlying causes of BPSD.^{33,36,38–43} More staff time may also be needed in order to increase physical exercise or outdoor time for residents with ADRD and to prevent the occurrence or exacerbation of BPSD.^33,44,45 NHs serving a higher proportion of Blacks are more likely to be short of resources and staffing,²⁶ and thus are less able to provide appropriate non-pharmacological dementia care. Indeed, we found that NHs with a high percent of Blacks had lower CNA and RN staffing. Therefore, these facilities may be more likely to turn to antipsychotics for the management of residents’ behavioral problems, and “code” residents with schizophrenia to justify the use of antipsychotics, especially following the implementation of public reporting on antipsychotic use. In addition, schizophrenia diagnosis involves careful observations of individuals’ behavioral symptoms. Staff in resource poor NHs may be less likely to have adequate training in mental health to adequately evaluate residents’ symptoms /behaviors, or to communicate effectively with psychiatrists to assist in the diagnosis of schizophrenia, all of which may compromise the accuracy of correctly diagnosing this condition and subsequently negatively impact the quality of care.

Despite the relationship between NH racial composition and the increase in the diagnosis of schizophrenia, the racial differences in the increase of schizophrenia diagnoses among NH residents with ADRD cannot be fully explained by this alone because such racial differences remained within NHs, even after accounting NH factors. It is possible that Blacks with ADRD are at higher risk of having BPSD than their White counterparts,^20–24,25 and studies found that Blacks are more likely to be misdiagnosed with schizophrenia for other behavioral problems.^7,17–19 However, these reasons alone do not seem to explain the within-NH racial differences in growth of schizophrenia unless systematic changes in clinicians’ practice patterns can be shown to have occurred in recent years.

There are some limitations of this study. First, we were not able to explicitly determine the reasons for the growth in schizophrenia. Although we suspect such increase is driven by CMS’s public reporting of NH antipsychotic us, it may also be related to some other unobserved concurrent changes. However, the systematic variations in the magnitude of growth in schizophrenia diagnosis across NHs suggest such increase is likely to be related to system-level factors, such as NH practices and policies. Second, our data only allowed us to observe a relatively short time period prior the public reporting, and thus we were not able to examine the trends in the schizophrenia diagnosis prior to the public reporting of NH antipsychotic use. Lastly, it is possible that the increased prevalence of schizophrenia, especially among Blacks, may be due to NHs’ documentation or identification of schizophrenia. It is interesting that we observed that Black residents were less likely to have the diagnosis of schizophrenia than Whites within a NH, prior to the public reporting, after accounting for NH effects. Although we were not able to determine the underlying reason, it is possible that schizophrenia among Blacks was under-documented prior to the public reporting. However, this would not explain the relationship between NH racial composition and the increase in the diagnosis of schizophrenia over time, as it is not very likely that NHs with a higher proportion of Blacks experienced higher levels of improvement in documenting mental illnesses post public reporting implementation.

CONCLUSIONS

In conclusion, we found an increase in the frequency of schizophrenia diagnosis codes following the implementation of public reporting of antipsychotic use in NHs, especially for Blacks with ADRD. Future research is need to examine the root cause for this increase and to re-examine policies that may incentivize inaccurate diagnosis and exacerbate racial disparities in NH care. The increase in the diagnosis of schizophrenia, if motivated by the CMS’ antipsychotic reduction policy, may result in residents being exposed to treatment with significant risks of adverse effects and poor health outcomes that further exacerbate the long-standing racial disparities in care quality among NH residents with ADRD.

Highlights:

1. What is the primary question addressed by this study?

   To examine racial differences in the change in the frequency of schizophrenia diagnosis among nursing home residents with ADRD following the public reporting of nursing home antipsychotic use, and sources of such racial differences.

2. What is the main finding of this study?

   The increase in the frequency of schizophrenia diagnosis during the study period was greater for Black residents with ADRD as compared with whites. Such racial differences were related to NHs’ racial composition, but racial differences in the growth of schizophrenia diagnosis persisted within a NH after accounting for NH factors.

3. What is the meaning of the finding?

   Efforts are needed to reexamine policies that may incentivize inaccurate diagnosis and exacerbate racial differences in nursing home care.

Data Statement:

This study has been presented in 2021 Academyhealth virtual meeting (June 2021).