INTRODUCTION
Increasing awareness of healthcare disparities has prompted re-examination of the National Quality Forum recommendation that measures of healthcare quality not be adjusted for patients’ sociodemographic characteristics. Adjustment might appear to endorse poorer quality care for those traditionally under-served. Fiscella and colleagues (1) point out, however, that failure to adjust for sociodemographic differences might unfairly penalize health systems serving disadvantaged groups. Jha and Zaslavsky (2) argue that quality measures should be adjusted for patient characteristics when differences between health systems are confounded by differences between patients they serve. In those cases, stratified reporting of quality measures would both reveal health disparities and permit fairer comparisons of quality across health systems or facilities.
Given that rates of mental health treatment differ markedly by race and ethnicity (3), we examined how stratifying by race/ethnicity would affect a specific mental healthcare quality measure: the proportion of outpatients starting antidepressant treatment who receive adequate or potentially effective acute-phase treatment (4).
METHODS
Participating health systems (Group Health Cooperative, HealthPartners, Kaiser Permanente Colorado, Kaiser Permanente Hawaii, and Kaiser Permanente Southern California) serve over six million patients, with each system's membership representing the racial/ethnic distribution of its service area (Table 1). Using methods previously described (5), health system records identified adult outpatients beginning a new episode of antidepressant treatment for a depressive disorder between 1/1/2010 and 12/31/2012. Self-reported race/ethnicity was identified from electronic medical records. Pharmacy dispensing records were used to identify patients receiving more than 90 days supply of any antidepressant medication over 180 days, beginning with the index prescription, similar to the NCQA/HEDIS measure of adequate acute-phase treatment (4). The sample was limited to patients continuously enrolled from 270 days before to 180 days after the index prescription. Health system institutional review boards granted waivers of consent for use of de-identified data.
Table 1.
Racial / ethnic composition of health system populations
| Site A | Site B | Site C | Site D | Site E | |
|---|---|---|---|---|---|
| Non-Hispanic White | 62.2% | 43.8% | 36.0% | 70.6% | 53.6% |
| African American | 3.5% | 9.6% | 1.2% | 3.4% | 5.9% |
| Asian | 3.9% | 5.1% | 25.5% | 1.5% | 1.7% |
| Hawaiian/Pac. Islander | 0.8% | 0.6% | 23.3% | 0.2% | 0.1% |
| Native American | 1.5% | 0.3% | 1.5% | 0.8% | 0.9% |
| Hispanic | 4.5% | 37.2% | 6.8% | 13.7% | 1.0% |
| Other or Unknown | 23.7% | 3.5% | 5.6% | 9.7% | 36.9% |
RESULTS
The overall proportion of patients meeting this threshold for effective acute-phase treatment varied from 58.2% to 69.9% across the five health systems (Table 2). This proportion varied markedly across racial/ethnic groups, but showed a similar pattern across health systems (highest in Non-Hispanic Whites, lower in Asians and Hispanics, lowest in African Americans). Rates for Native Hawaiians and Pacific Islanders were the most variable across sites. When rates for each health system were standardized to replicate the race/ethnicity distribution of the entire sample (6), overall rates of adequate treatment varied from 60.5% to 65.6%.
Table 2.
Proportion of adult outpatients starting antidepressant treatment for depression who receive at least 90 days of medication over the following 180 days (with 95% confidence limits).
| Total Number | Site A | Site B | Site C | Site D | Site E | |
|---|---|---|---|---|---|---|
| Total | 183,395 | 67.8% (67.5-68.1) | 63.4% (63.2-63.5) | 58.2% (57.5-58.9) | 63.9% (63.5-64.2) | 69.9% (69.5-70.2) |
| Non-Hispanic White | 91,602 | 70.1% (69.8-70.5) | 71.1% (70.9-71.3) | 65.1 (63.9-66.3) | 67.6% (67.2-68.0) | 72.8% (72.4-73.2) |
| African American | 13,697 | 49.6% (47.9-51.4) | 54.5% (54.0-55.0) | 54.4% (47.8-61.0) | 46.5% (44.6-48.5) | 40.2% (38.8-41.6) |
| Asian | 8596 | 54.5% (52.9-56.1) | 61.1% (55.8-58.7) | 57.3% (55.8-58.7) | 55.3% (52.4-58.3) | 58.6% (56.1-61.2) |
| Hawaiian/Pac. Islander | 1996 | 57.6% (54.0-61.2) | 63.2% (61.3-65.0) | 48.6% (47.1-50.1) | 60.0% (51.7-68.3) | 33.3% (19.7-46.9) |
| Native American | 1159 | 67.9% (65.5-70.3) | 67.3% (55.6-67.2) | 61.4% (55.6-67.2) | 63.3% (59.5-67.1) | 63.6% (60.0-67.1) |
| Hispanic | 46,592 | 62.8% (61.4-64.3) | 56.6% (56.4-56.9) | 53.0% 50.2-55.8) | 49.8% (48.8-50.8) | 53.7% (50.3-57.1) |
| Other or Unknown | 19,753 | 67.6% (67.0-68.2) | 66.4% (65.6-67.1) | 64.0% (61.0-67.0) | 64.4% (63.3-65.6) | 71.5% (71.0-72.0) |
| Standardized Total* | 183,395 | 65.6% | 65.1% | 60.5% | 60.5% | 64.2% |
total rate for each site directly standardized to replicate race/ethnicity distribution of entire sample
DISCUSSION
Most of the observed variation across health systems in overall rates of effective acute-phase antidepressant treatment reflected differences in the racial/ethnic distribution of patient populations. Standardizing across health systems reduced the range in performance from approximately 12% to approximately 5% and significantly altered the overall ranking of these health systems. Rankings also varied widely across racial/ethnic groups, with every health system ranking first or second in at least one group.
We cannot determine whether lower rates of adequate treatment in some racial or ethnic groups reflect disparities in clinical practice (which should be reduced) or differences in patients’ informed treatment preferences (which should be respected). In either case, use of unadjusted overall rates would bias comparisons between health systems. Unadjusted comparisons could create perverse incentives, punishing a health system for identifying and treating depression in traditionally under-served groups. Stratified rates also reveal important opportunities for improving care in traditionally under-served groups.
In these five health systems, most of the variation in overall rates of adequate acute-phase antidepressant treatment was due to confounding by racial/ethnic differences in the patients they serve. Consistent with recommendations of Fiscella (1) and Jha (2), comparison of depression care across health systems – and incentives to improve health system performance – should be based on stratified performance measures.
ACKNOWLEDGMENTS
The Mental Health Research Network is supported by NIMH Cooperative Agreement U19MH092201. The National Institute of Mental Health had no role in the design and conduct of this study; collection, management, and analysis and interpretation of the data; preparation, review, or approval of the manuscript, or the decision to submit the manuscript for publication. Drs. Stewart and Simon conducted data analyses. Dr. Simon had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Drs. Simon and Penfold have received salary support from research grants by Bristol Myers Squibb and Otsuka Pharmaceuticals to Group Health Research Institute. We do not believe that any of these financial interests are relevant to the question addressed by this letter.
The Mental Health Research Network is supported by NIMH Cooperative Agreement U19MH092201
REFERENCES
- 1.Fiscella K, Burstin HR, Nerenz DR. Quality measures and sociodemographic risk factors: to adjust or not to adjust. JAMA. 2014;312(24):2615–6. doi: 10.1001/jama.2014.15372. [DOI] [PubMed] [Google Scholar]
- 2.Jha AK, Zaslavsky AM. Quality reporting that addresses disparities in health care. JAMA. 2014;312(3):225–6. doi: 10.1001/jama.2014.7204. [DOI] [PubMed] [Google Scholar]
- 3.Substance Abuse and Mental Health Services Administration . Racial/Ethnic Differences in Mental Health Service Use Among Adults. Substance Abuse and Mental Health Services Administration; Rockville, MD: 2015. [Google Scholar]
- 4.National Committee for Quality Assurance . HEDIS 2015 Volume 1: Narrative. National Committee for Quality Assurance; Washington, DC: 2014. [Google Scholar]
- 5.Simon GE, Rossom RC, Beck A, Waitzfelder BE, Coleman KJ, Stewart C, et al. Antidepressants are not over-prescribed for mild depression. J Clin Psychiatry. doi: 10.4088/JCP.14m09162. in press. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Kalton G. Standardization: A technique to control for extraneous variables. Applied Statistics. 1968;17:118–36. [Google Scholar]