Currently Viewing:
The American Journal of Managed Care August 2018
Currently Reading
Impact of a Medical Home Model on Costs and Utilization Among Comorbid HIV-Positive Medicaid Patients
Paul Crits-Christoph, PhD; Robert Gallop, PhD; Elizabeth Noll, PhD; Aileen Rothbard, ScD; Caroline K. Diehl, BS; Mary Beth Connolly Gibbons, PhD; Robert Gross, MD, MSCE; and Karin V. Rhodes, MD, MS
Precision Medicine and Sharing Medical Data in Real Time: Opportunities and Barriers
Y. Tony Yang, ScD, and Brian Chen, PhD, JD
Levers to Reduce Use of Unnecessary Services: Creating Needed Headroom to Enhance Spending on Evidence-Based Care
Michael Budros, MPH, MPP, and A. Mark Fendrick, MD
From the Editorial Board: Michael E. Chernew, PhD
Michael E. Chernew, PhD
Optimizing Number and Timing of Appointment Reminders: A Randomized Trial
John F. Steiner, MD, MPH; Michael R. Shainline, MS, MBA; Jennifer Z. Dahlgren, MS; Alan Kroll, MSPT, MBA; and Stan Xu, PhD
Impact of After-Hours Telemedicine on Hospitalizations in a Skilled Nursing Facility
David Chess, MD; John J. Whitman, MBA; Diane Croll, DNP; and Richard Stefanacci, DO
Baseline and Postfusion Opioid Burden for Patients With Low Back Pain
Kevin L. Ong, PhD; Kirsten E. Stoner, PhD; B. Min Yun, PhD; Edmund Lau, MS; and Avram A. Edidin, PhD
Patient and Physician Predictors of Hyperlipidemia Screening and Statin Prescription
Sneha Kannan, MD; David A. Asch, MD, MBA; Gregory W. Kurtzman, BA; Steve Honeywell Jr, BS; Susan C. Day, MD, MPH; and Mitesh S. Patel, MD, MBA, MS
Evaluating HCV Screening, Linkage to Care, and Treatment Across Insurers
Karen Mulligan, PhD; Jeffrey Sullivan, MS; Lara Yoon, MPH; Jacki Chou, MPP, MPL; and Karen Van Nuys, PhD
Reducing Coprescriptions of Benzodiazepines and Opioids in a Veteran Population
Ramona Shayegani, PharmD; Mary Jo Pugh, PhD; William Kazanis, MS; and G. Lucy Wilkening, PharmD
Medicare Advantage Enrollees’ Use of Nursing Homes: Trends and Nursing Home Characteristics
Hye-Young Jung, PhD; Qijuan Li, PhD; Momotazur Rahman, PhD; and Vincent Mor, PhD

Impact of a Medical Home Model on Costs and Utilization Among Comorbid HIV-Positive Medicaid Patients

Paul Crits-Christoph, PhD; Robert Gallop, PhD; Elizabeth Noll, PhD; Aileen Rothbard, ScD; Caroline K. Diehl, BS; Mary Beth Connolly Gibbons, PhD; Robert Gross, MD, MSCE; and Karin V. Rhodes, MD, MS
Among HIV-positive Medicaid patients with comorbid medical and psychiatric disorders, there was increased outpatient service utilization, yet relative cost savings, for patients who were treated in patient-centered medical homes.
METHODS

We examined pre-post healthcare utilization and costs for HIV-positive Medicaid patients with at least 1 of 4 chronic medical conditions, plus at least 1 psychiatric and/or substance use disorder, comparing changes in healthcare utilization and costs among patients treated in CCI and non-CCI practices. This study was approved by the University of Pennsylvania Institutional Review Board with a waiver of informed consent and complied with the ethical standards of the Office for Human Research Protections.

Data

We initially obtained a list of 147 CCI practices from the Pennsylvania Department of Health. Identifying information for 5 of the 152 practices in the CCI was not in the list provided. Of the remaining 147, we excluded 10 that did not serve Medicaid patients, reducing the sample to 137 CCI practices. Using a Medicaid claims dataset obtained from the Pennsylvania Department of Human Services, we identified HIV-positive patients treated in the CCI practices and those treated in non-CCI practices during the time period the CCI was active. We identified HIV-positive patients with an International Classification of Diseases, Ninth Revision code of 042 (attached to at least 2 claims) who also had primary or secondary diagnostic codes for at least 1 of 4 chronic medical conditions (in any type of claim) and at least 1 claim (outpatient or inpatient nonlaboratory) for a psychiatric and/or substance use disorder. The 4 comorbid chronic medical conditions (diabetes, chronic obstructive pulmonary disease [COPD], asthma, and congestive heart failure [CHF]) were selected because they were identified as sources of disproportionate health and financial burdens by the Pennsylvania Health Care Cost Containment Council.31 Comorbid behavioral health conditions included psychiatric (major depressive disorder, schizophrenia/schizoaffective disorder, bipolar disorder, posttraumatic stress disorder, and anxiety disorders) and substance use (opioid, cocaine, and alcohol) disorders. The psychiatric disorders chosen were the most prevalent and are associated with high healthcare costs within a Medicaid population.32 The substance use disorders chosen were the 3 most common nationally (other than marijuana) for patients presenting at substance use treatment facilities during the target years.33

HIV-positive patients were considered part of the CCI group if they had at least 1 claim from a CCI practice. Patients with HIV who never had a claim from a CCI practice were placed in the comparison group. For each CCI patient, the first claim filed after the date the patient’s practice joined the CCI was identified as the “index episode.” For patients in the comparison (non-CCI) group, the index episode was defined as the first claim filed after commencement of the CCI program. In the original sample, there were 404 CCI patients and 4039 non-CCI patients diagnosed as HIV positive. To be included in the study, a patient needed to have at least 6 months of Medicaid eligibility during the year prior and the year following the index episode. This restriction resulted in a sample of 302 CCI HIV-positive patients and 2577 non-CCI HIV-positive patients. The date of the index episode was used to mark the first exposure of each patient to the “intervention” of the CCI, allowing for a pre-post intervention comparison, as has been done in other studies.34,35

Patients eligible for both Medicaid and Medicare were included, as many patients with chronic mental health and substance abuse conditions have such dual eligibility. For the target sample, we identified the presence of both chronic medical conditions and behavioral health diagnoses from claims at any point during the 2005-2010 time period, whether before or after the index episode date. This was done because the medical and psychiatric disorders targeted in this study typically are chronic conditions that manifest first as subthreshold symptoms and, if managed well, can be prevented from evolving into full diagnoses.

Costs were calculated using standardized prices for Medicaid claims. Outpatient costs were standardized using the Medicaid outpatient fee schedule, regardless of fee-for-service or capitation. Notably, capitated Medicaid managed care plans also submit claims for provided services. Pharmacy costs were based on fee-for-service Medicaid pharmacy costs specified for each Hierarchical Ingredient Code. To standardize inpatient costs, we computed 2008 Pennsylvania Medicaid average costs by diagnosis-related groups using fee-for-service data. We separately calculated pharmacy costs for antiretroviral drugs and other medications.

Outcomes

We compared pre-post changes in healthcare utilization and costs for HIV-positive patients treated at CCI practices with changes in utilization and costs for non-CCI HIV-positive patients in the same year. Changes in utilization and costs were measured from the 1-year pre–index episode period to the 1-year post–index episode period, with costs and counts calculated per month eligible for Medicaid.

Statistical Analysis

The primary outcome measure was the difference between CCI and non-CCI patients in total healthcare cost changes per month eligible between the 1-year preindex period and the 1-year postindex period. Secondary outcomes included between-group differences in pre-post changes within specific ED, inpatient, pharmacy, and outpatient cost and utilization variables. Negative values for differences favored the CCI group over the non-CCI group and vice versa for positive values.

A propensity score was derived to address the potential lack of comparability of patients in CCI and non-CCI practices at the time of treatment initiation using variables listed in Table 1. Other variables included in the propensity score (not listed in Table 1) were disability status or Supplemental Security Income benefits, months of Medicaid eligibility, dual Medicaid and Medicare eligibility, year of index episode, and region within Pennsylvania. The propensity score was used as an inverse weighting factor in all analyses.36,37 Covariates included in all models were pre–index year total costs for each patient, pre–index year total utilization for each patient, and year of preindex episode.

Primary analysis of the cost variables consisted of a weighted least squares model of difference-in-differences (DID) scores, including the covariates described above and the propensity score as a weighting factor. Analyses of healthcare utilization counts proceeded using either generalized Poisson (GP) models, zero-inflated GP models, negative binomial regression models, or zero-inflated negative binomial models with SAS PROC-GENMOD, PROC-GLIMMIX, and PROC-NLMIXED.38,39 The choice between zero-inflated and noninflated models depended on the magnitude of the zero counts, assessed using Vuong’s test.40 The choice between GP and negative binomial was a function of whether there was underdispersion or overdispersion of the data, respectively. For the zero versus nonzero component of the zero-inflated models, differences between the CCI and non-CCI groups were described using odds ratios (ORs) with 95% CIs. For the GP, negative binomial regression, and count portion of the zero-inflated models, the exponential of the regression coefficient for the comparison between CCI and non-CCI cost differences was interpreted as the percent increase or decrease in the expected count for CCI compared with non-CCI.41

As an assessment of whether the propensity procedure successfully balanced the groups, we implemented the goodness-of-fit diagnostic steps described by Austin.42 First, we derived and visually examined quintile side-by-side boxplots of the propensity score for the CCI and non-CCI groups. Finally, we derived the weighted conditional standardized difference for each of the baseline predictors in Table 1 and compared each with the unconditional standardized difference.43


 
Copyright AJMC 2006-2019 Clinical Care Targeted Communications Group, LLC. All Rights Reserved.
x
Welcome the the new and improved AJMC.com, the premier managed market network. Tell us about yourself so that we can serve you better.
Sign Up