CMS indicates Centers for Medicare & Medicaid Services. The “predicted trend” was the projected rate of diagnoses in the absence of the CMS suppression procedures, based on a continuation of the baseline trend. The shaded area indicates the 95% confidence band for the modeled trend.
Customize your JAMA Network experience by selecting one or more topics from the list below.
Rough K, Bateman BT, Patorno E, et al. Suppression of Substance Abuse Claims in Medicaid Data and Rates of Diagnoses for Non–Substance Abuse Conditions. JAMA. 2016;315(11):1164–1166. doi:10.1001/jama.2015.18417
In a change from longstanding practice, the Centers for Medicare & Medicaid Services (CMS) recently began suppressing substance abuse–related claims in the Medicare and Medicaid Research Identifiable Files.1,2 This change was enacted to comply with a 1987 federal regulation barring third party payers from releasing information from federally funded substance abuse treatment programs without patient consent.3
CMS removes any claim containing a diagnostic or procedure code related to substance abuse, meaning that the entire encounter captured by the claim is deleted (including all diagnosis and procedure codes).1,4 Therefore, important diagnoses linked to substance abuse might also be suppressed.
We examined the association between implementation of the CMS suppression policy and rates of diagnoses for non–substance abuse conditions in Medicaid data.
We received Medicaid data for 2000-2006 prior to implementation of the suppression policy (ie, containing substance abuse codes) and data for 2007-2010 after the policy was enacted, allowing comparison of data without vs with claim suppression. Use of this deidentified database was approved by Partners’ institutional review board and the need for informed consent was waived.
Based on all diagnosis fields for International Classification of Diseases, Ninth Revision, Clinical Modification codes, we calculated annual inpatient and outpatient rates (per 100 000 patients utilizing inpatient and outpatient services, respectively) of diagnoses for 6 conditions that commonly co-occur with substance abuse (hepatitis C, human immunodeficiency virus, cirrhosis, tobacco use, depression, and anxiety) and 4 conditions unrelated to substance abuse (type II diabetes, stroke, hypertension, and kidney disease).
Segmented linear regression models allowing for first-order autocorrelation were used to test for changes in the rate of each condition in the years after suppression was implemented. Models included a term for the baseline trend (2000-2006) and terms to estimate changes in level and trend after implementation of suppression procedures (2007-2010). For each parameter, 95% confidence intervals were calculated and a 2-sided Wald test was conducted. P values less than .05 were considered statistically significant.
Analyses were performed in SAS (SAS Institute), version 9.4.
Over the 11-year study, there were 63 million inpatient and 13.6 billion outpatient claims. For inpatient diagnoses, regression models showed a statistically significant negative level change (ie, immediate reduction in the first year affected by suppression) for conditions commonly co-occurring with substance abuse (Table). Relative to rates observed in 2006, there was an immediate reduction in the 2007 inpatient diagnosis rates (per 100 000 patients) of 56.7% for hepatitis C (−1233 [95% CI, −1588 to −908]; P < .001) (Figure), 51.3% for tobacco use (−5015 [95% CI, −6073 to −3957]; P < .001), 48.9% for cirrhosis (−675 [95% CI, −864 to −486]; P < .001), 38.4% for depression (−2712 [95% CI, −4377 to −1047]; P = .02), 26.6% for anxiety (−795 [95% CI, −1220 to −371]; P = .01), and 24.0% for HIV (−498 [95% CI, −665 to −330]; P < .001).
Reductions in outpatient diagnosis rates were less pronounced. Although all conditions that commonly co-occur with substance abuse had a negative level change, this decrease was only statistically significant for anxiety, with a 6.3% reduction (−2512 [95% CI, −4811 to −213]; P = .02).
Conditions unrelated to substance abuse appeared generally unassociated with the CMS suppression practices. However, implementation of the policy coincided with sudden and substantial decreases in the rates of inpatient diagnoses for conditions commonly co-occurring with substance abuse, and anxiety showed significant reductions in outpatient diagnosis rates.
Underestimation of diagnoses has the potential to bias health services research studies and epidemiological analyses for which affected conditions are outcomes or confounders. In studies of health care utilization, the number of missing claims may vary in a nonrandom fashion between groups defined by demographics, disease, or locality. Comparisons between groups may lead to spurious conclusions—a hospital that regularly admits substance abusers will have artificially low rates of readmission, giving a false appearance of better performance.
A potential limitation is that the observations are susceptible to influence from secular trends, including changes in Medicaid eligibility, coding practices, or the true underlying prevalence of the medical conditions. However, the marked, immediate decline in inpatient rates of comorbidities related to substance abuse following the beginning of the suppression period suggests that these findings were likely the result of the CMS suppression policies.
Corresponding Author: Kathryn Rough, ScM, Division of Pharmacoepidemiology and Pharmacoeconomics, Brigham and Women’s Hospital, 1620 Tremont St, Ste 3030, Boston, MA 02120 (firstname.lastname@example.org).
Author Contributions: Ms Rough had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Rough, Bateman, Patorno, Desai, Hernandez-Diaz, Huybrechts.
Acquisition, analysis, or interpretation of data: Rough, Bateman, Patorno, Park, Hernandez-Diaz, Huybrechts.
Drafting of the manuscript: Rough.
Critical revision of the manuscript for important intellectual content: Rough, Bateman, Patorno, Desai, Park, Hernandez-Diaz, Huybrechts.
Statistical analysis: Rough, Desai, Park.
Obtained funding: Hernandez-Diaz.
Administrative, technical, or material support: Rough, Huybrechts.
Study supervision: Patorno, Hernandez-Diaz, Huybrechts.
Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Ms Rough reports receiving a 3-month student internship from Bayer. Dr Hernandez-Diaz reports receiving personal fees from AstraZeneca and University of California, Berkeley, and grant funding from Pfizer. No other disclosures were reported.
Funding/Support: This article was supported by training grants from the Pharmacoepidemiology Program at the Harvard T. H. Chan School of Public Health (Ms Rough and Ms Park), grant T32 AI007433 from the National Institute of Allergy and Infectious Diseases (Ms Rough), career development grant K01MH099141 (Dr Huybrechts) and grant R01MH100216 (Dr Hernandez-Diaz) from the National Institute of Mental Health, and career development grant K08HD075831 from the Eunice Kennedy Shriver National Institute of Child Health & Human Development (Dr Bateman).
Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Additional Contributions: We thank Helen Mogun, MS (Brigham and Women’s Hospital), for preparing the analytic datasets for this study and Cora Allen-Coleman, BA (University of Wisconsin-Madison), for her assistance in creating the Figure. These contributors did not receive compensation apart from their salary.
Create a personal account or sign in to: