Pinar Karaca-Mandic

Interaction terms in nonlinear models

OBJECTIVES To explain the use of interaction terms in nonlinear models. STUDY DESIGN We discuss the motivation for including interaction terms in multivariate analyses. We then explain how the straightforward interpretation of interaction terms in linear models changes in nonlinear models, using graphs and equations. We extend the basic results from logit and probit to difference-in-differences models, models with higher powers of explanatory variables, other nonlinear models (including log transformation and ordered models), and panel data models. EMPIRICAL APPLICATION: We show how to calculate and interpret interaction effects using a publicly available Stata data set with a binary outcome. Stata 11 has added several features which make those calculations easier. LIMDEP code also is provided. CONCLUSIONS It is important to understand why interaction terms are included in nonlinear models in order to be clear about their substantive interpretation.

Cost Sharing, Family Health Care Burden, and the Use of Specialty Drugs for Rheumatoid Arthritis

OBJECTIVES To examine the impact of benefit generosity and household health care financial burden on the demand for specialty drugs in the treatment of rheumatoid arthritis (RA). DATA SOURCES/STUDY SETTING Enrollment, claims, and benefit design information for 35 large private employers during 2000-2005. STUDY DESIGN We estimated multivariate models of the effects of benefit generosity and household financial burden on initiation and continuation of biologic therapies. DATA EXTRACTION METHODS We defined initiation of biologic therapy as first-time use of etanercept, adalimumab, or infliximab, and we constructed an index of plan generosity based on coverage of biologic therapies in each plan. We estimated the households burden by summing up the annual out-of-pocket (OOP) expenses of other family members. PRINCIPAL FINDINGS Benefit generosity affected both the likelihood of initiating a biologic and continuing drug therapy, although the effects were stronger for initiation. Initiation of a biologic was lower in households where other family members incurred high OOP expenses. CONCLUSIONS The use of biologic therapy for RA is sensitive to benefit generosity and household financial burden. The increasing use of coinsurance rates for specialty drugs (as under Medicare Part D) raises concern about adverse health consequences.

Behavioral impact of graduated driver licensing on teenage driving risk and exposure

Graduated driver licensing (GDL) is a critical policy tool for potentially improving teenage driving while reducing teen accident exposure. While previous studies demonstrated that GDL reduces teenage involvement in fatal crashes, much remains unanswered. We explore the mechanisms through which GDL influences accident rates as well as its long term effectiveness on teen driving. In particular, we investigate: (1) whether GDL policies improve teenage driving behavior, or simply reduce teenage prevalence on the roads; (2) whether GDL exposed teens become better drivers in later years. We employ a unique data source, the State Data System, which contains all police reported accidents (fatal and non-fatal) during 1990-2005 for 12 states. We estimate a structural model that separately identifies GDLs effect on relative teenage prevalence and relative teenage riskiness. Identification of the model is driven by the relative numbers of crashes between two teenagers, two adults, or a teenager and an adult. We find that the GDL policies reduce the number of 15-17-year-old accidents by limiting the amount of teenage driving rather than by improving teenage driving. This prevalence reduction primarily occurs at night and stricter GDL policies, especially those with night-time driving restrictions, are the most effective. Finally, we find that teen driving quality does not improve ex post GDL exposure.

Health and Health Risks Among Sexual Minority Women: An Examination of 3 Subgroups

We used 2001-2010 National Health and Nutrition Examination Survey data to examine insurance status, source of routine care, cigarette and alcohol use, and self-rated health among lesbian, bisexual, and heterosexual women who have sex with women, compared with heterosexual women who do not have sex with women. We found higher risks of being uninsured among lesbian and bisexual women, worse self-rated health among bisexual women, higher alcohol use among bisexual and heterosexual women who have sex with women, and higher smoking across all subgroups.

Lymphedema Prevalence and Treatment Benefits in Cancer: Impact of a Therapeutic Intervention on Health Outcomes and Costs

Background Lymphedema is a common complication of cancer therapeutics; its prevalence, treatment outcomes, and costs have been poorly defined. The objective of this study was to examine lymphedema prevalence among cancer survivors and to characterize changes in clinical outcomes and costs associated with a defined therapeutic intervention (use of a pneumatic compression devices [PCD]) in a representative, privately insured population. Methods and Findings Retrospective analysis of de-identified health claims data from a large national insurer for calendar years 2007 through 2013. Patients were required to have 12 months of continuous insurance coverage prior to PCD receipt (baseline), as well as a 12-month follow-up period. Analyses were performed for individuals with cancer-related lymphedema (n = 1,065). Lymphedema prevalence was calculated: number of patients with a lymphedema claim in a calendar year divided by total number of enrollees. The impact of PCD use was evaluated by comparing rates of a pre-specified set of health outcomes and costs for the 12 months before and after, respectively, PCD receipt. Lymphedema prevalence among cancer survivors increased from 0.95% in 2007 to 1.24% in 2013. PCD use was associated with decreases in rates of hospitalizations (45% to 32%, p<0.0001), outpatient hospital visits (95% to 90%, p<0.0001), cellulitis diagnoses (28% to 22%, p = 0.003), and physical therapy use (50% to 41%, p<0.0001). The average baseline health care costs were high (

Predicting New Diagnoses of HIV Infection Using Internet Search Engine Data

53,422) but decreased in the year after PCD acquisition (−

How do health insurance loading fees vary by group size?: implications for Healthcare reform.

11,833, p<0.0001). Conclusions Lymphedema is a prevalent medical condition that is often a defining attribute of cancer survivorship. The problem is associated with high health care costs; Treatment (in this instance, use of PCD) is associated with significant decreases in adverse clinical outcomes and costs.

Hospital Prescribing of Opioids to Medicare Beneficiaries

TO THE EDITOR—Internet search engine data have been demonstrated to be a powerful tool to predict outbreaks of infectious diseases such as influenza and food-borne illnesses [1–3]. This method has not been applied, however, to predict incidence of chronic infections such as human immunodeficiency virus (HIV). HIV represents an important public health problem, yet there is a significant lag between initial reporting of diagnosis to public health authorities and report of regional incidence statistics by the Centers for Disease Control and Prevention (CDC). Providing more up-to-date information on incidence could help raise public awareness and aid prevention efforts. We examined the association between annual state-level incidence of HIV diagnosis and the volume of Internet searches for the term “HIV” from 2007 to 2010 using Google Trends [4], a publicly available search-volume tool from Google. State estimates of HIV incidence were obtained from the CDC, based on mandatory confidential reporting of new HIV diagnoses. State search-volume data from Google Trends were available only in relative, not absolute, terms. For instance, the state with the highest frequency of searches for the term “HIV” in a given year (relative to all other search terms) was reported as 100, with other states measured relative to it. In 2007, for example, Maryland had the highest frequency of searches for “HIV” relative to all other Internet search terms, while California had 72% of this frequency; Google Trends reported the rate of search for “HIV” as 100 in Maryland and 72 in California. To maintain comparability across years, we normalized search rates for “HIV” to be relative to Maryland’s rate in 2007. Results were the same regardless of which state–year pair was chosen for normalization. We estimated the bivariate association between Internet searches for “HIV” and annual HIV incidence at the state level from 2007 to 2010. To assess how closely searches for “HIV” predicted HIV incidence, we also estimated the bivariate association for 2007–2008 only and used estimates from this model to predict state HIV incidence for 2009–2010 based on state Internet searches for “HIV” during this period. We compared predicted to actual rates in 2009–2010. States varied considerably in Internet searches for “HIV”; in 2010, for example, the frequency of searches for “HIV” in Utah was only 40% of Maryland’s rate,

Impact of New Drugs and Biologics on Colorectal Cancer Treatment and Costs

The health insurance loading fee represents the portion of the premium above the expected amount of medical care expenditures paid by the insurance company. The size of the loading fees and how they vary by employer group size have important implications for health policy given the recent passage of the Patient Protection and Affordable Care Act. Despite their policy relevance, there is surprisingly little empirical evidence on the magnitude and the determinants of health insurance loading fees. This paper provides estimates of the loading fees by firm size using data from the confidential Medical Expenditure Panel Survey Household Component–Insurance Component Linked File. Overall, we find an inverse relationship between employer group size and loading fees. Firms of up to 100 employees face similar loading fees of approximately 34%. Loads decline with firm size and are estimated to be on average 15% for firms with more than 100 employees, but less than 10,000 employees, and 4% for firms with more than 10,000 workers.

The Cutaneous, Net Clinical, and Health Economic Benefits of Advanced Pneumatic Compression Devices in Patients With Lymphedema

IMPORTANCE Use of opioids during and shortly after an acute hospitalization is warranted in some clinical settings. However, given the potential of opioids for short-term adverse events and long-term physiologic tolerance, it is important to understand the frequency of opioid prescribing at hospital discharge, hospital variation, and patient and hospital factors associated with opioid prescribing, which is currently unknown in the United States. OBJECTIVE To estimate the frequency of opioid prescribing at hospital discharge among Medicare beneficiaries without an opioid prescription claim 60 days prior to hospitalization; to document hospital variation in prescribing; and to analyze patient and hospital factors associated with prescribing, including hospital average performance on pain-related Hospital Consumer Assessment of Healthcare Providers and Systems (HCAHPS) measures. DESIGN, SETTING, AND PARTICIPANTS Analysis of pharmacy claims of a 20% random sample of Medicare beneficiaries hospitalized in 2011 without an opioid prescription claim in the 60 days before hospitalization. MAIN OUTCOMES AND MEASURES Our main outcome was a new opioid claim within 7 days of hospital discharge. We estimated a multivariable linear probability model of patient factors associated with new opioid use and described hospital variation in adjusted rates of new opioid use. In multivariable linear regression analysis, we also analyzed hospital factors associated with average adjusted new opioid use at the hospital level, including the percentage of each hospitals patients who reported that their pain during hospitalization was always well controlled in the 2011 HCAHPS surveys. RESULTS Among 623 957 hospitalizations, 92 882 (14.9%) were associated with a new opioid claim. Among those hospitalizations with an associated opioid claim within 7 days of hospital discharge, 32 731 (42.5%) of 77 092 were associated with an opioid claim after 90 days postdischarge. Across 2512 hospitals, the average adjusted rate of new opioid use within 7 days of hospitalization was 15.1% (interquartile range, 12.3%-17.4%; interdecile range, 10.5%-20.0%). A hospitals adjusted rate of new opioid use was modestly positively associated with the percentage of its inpatients reporting that their pain was always well managed (increase from 25th to the 75th percentile in the HCAHPS measure was associated with an absolute increase in new opioid use of 0.89 percentage points or a relative increase of 6.0%; P < .001). CONCLUSIONS AND RELEVANCE New opioid use after hospitalization is common among Medicare beneficiaries, with substantial variation across hospitals and a large proportion of patients using a prescription opioid 90 days after hospitalization. The degree to which observed hospital variation in short- and longer-term opioid use reflects variation in inappropriate prescribing at hospital discharge is unknown.