Colleen Carey

Need for and Use of Family Leave Among Parents of Children With Special Health Care Needs

OBJECTIVE. Parents of children with special health care needs are especially vulnerable to work–family conflicts that family leave benefits might help resolve. We examined leave-taking among full-time–employed parents of children with special health care needs. METHODS. We identified all children with special health care needs in 2 large inpatient/outpatient systems in Chicago, Illinois, and Los Angeles, California, and randomly selected 800 per site. From November 2003 to January 2004, we conducted telephone interviews with 1105 (87% of eligible and successfully contacted) parents. Among the samples 574 full-time–employed parents, we examined whether leave benefits predicted missing any work for child illness, missing >4 weeks for child illness, and ability to miss work whenever their child needed them. RESULTS. Forty-eight percent of full-time–employed parents qualified for federal Family and Medical Leave Act benefits; 30% reported employer-provided leave benefits (not including sick leave/vacation). In the previous year, their children averaged 20 missed school/child care days, 12 doctor/emergency department visits, and 1.7 hospitalizations. Although 81% of parents missed work for child illness, 41% reported not always missing work when their child needed them, and 40% of leave-takers reported returning to work too soon. In multivariate regressions, parents who were eligible for Family and Medical Leave Act benefits and aware of their eligibility had 3.0 times greater odds of missing work for child illness than ineligible parents. Parents with >4 weeks of employer-provided leave benefits had 4.7 times greater odds of missing >4 weeks than parents without benefits. Parents with paid leave benefits had 2.8 times greater odds than other parents of missing work whenever their child needed them. CONCLUSIONS. Full-time–employed parents of children with special health care needs experience severe work–family conflicts. Although most have leave benefits, many report unmet need for leave. Access to Family and Medical Leave Act benefits and employer-provided leave may greatly affect leave-taking.

Patterns of Potential Opioid Misuse and Subsequent Adverse Outcomes in Medicare, 2008 to 2012

Opioid misuse has reached epidemic proportions in the United States, leading to a substantial morbidity and mortality burden (1, 2). In 2015, nearly 5% of U.S. adults misused opioids in the previous year (3). The rise of this epidemic seems to have contributed to a historic increase in deaths among middle-aged white Americans (4) after decades of decline, and recent research suggests that it may be associated with low labor market participation in the wake of the Great Recession (5). According to the National Survey on Drug Use and Health, most opioid misuse cases originated with a legitimate prescription from a provider, with the drugs either prescribed directly to the patient or obtained through diversion (6). The scale of misuse of legitimately prescribed opioids has led to an intense focus on helping health care providers identify patients at high risk for misuse. The most prominent examples are state databases known as prescription drug monitoring programs (PDMPs) (7), which collect data from the states pharmacies (and, in some cases, from pharmacies in other states) on all prescription opioid fills. Although PDMPs originally were designed to assist law enforcement in investigations, 49 states (excluding Missouri) and the District of Columbia have implemented a PDMP that allows access to providers, and a growing number of states are mandating that providers review PDMP data before prescribing opioids (8, 9). In principle, PDMPs allow providers to infer a patients risk for opioid misuse by examining details of the patients prescription history. However, despite the wealth of opioid prescribing data contained in PDMPs, clear guidance is lacking on what specific use patterns are correlated with adverse opioid-related patient outcomes and should therefore prompt clinical attention (1012). Previous studies on patterns of potential opioid misuse generally focused on single measures and examined only a narrow range of outcomes, potentially underestimating the depth of information present in a complex prescription history (1316). Several metrics have been proposed to assess potential opioid misuse, including acquisition from more than 1 prescriber or pharmacy, receipt of opioids across state borders, and overlapping days of supply. However, these patterns are not mutually exclusive, and relying on any one measure to define misuse may exclude a substantial proportion of potential misuse. Systematic empirical evidence is lacking on how the many dimensions of potential prescription opioid misuse relate to subsequent opioid-related adverse events. Using data on Medicare beneficiaries, we examined the relationship between several patterns of potential opioid misuse and adverse opioid-related outcomes, such as opioid overdose and death. We used longitudinal data to study how 6 measures of potential opioid misuse each relate to adverse opioid-related outcomes during the following year, with the goal of identifying clinically relevant patterns of high-risk prescription opioid use. Methods Study Population and Data Sources We used prescription drug and medical claims for a 5% random sample of Medicare beneficiaries between 2008 and 2012. We included beneficiaries of all ages because the subset younger than 65 years, who are largely disabled, has a high prevalence of opioid use (17). We divided claims into half-years (6-month periods), and our analysis sample included patients who filled at least 1 opioid prescription in any half-year and were not enrolled in Medicare Advantage for any month during the current or next calendar year. We excluded persons who had cancer diagnosed in the year of the opioid fill, because many patients with advanced cancer use opioids appropriately for pain control. After sample restrictions, we noted opioid fills and adverse opioid outcomes in 1933232 beneficiary half-year observations between 2008 and 2012. This number represents 627391 distinct persons who received opioids for an average of 3 half-years. The study was approved by the Cornell University Institutional Review Board for Human Participants. Defining Opioid Prescriptions We defined an opioid fill as an observation in the Part D (prescription drug insurance) event file of any drug containing an opioid analgesic ingredient, as classified by the U.S. Pharmacopeia (18). We extracted the number of days for which opioids were supplied and calculated the morphine-equivalent dosage (MED) of total medication dispensed by using standard conversion tables (19). Measures of Possible Opioid Misuse We defined 6 measures of possible opioid misuse on the basis of 3 separate dimensions of potential misuse: high prescription quantity, fragmented prescribing (sometimes referred to as doctor shopping), and use of out-of-state providers. These measures are not diagnostic of misuse but serve as potential markers for high-risk opioid use that may deserve scrutiny by providers. Within these misuse measures, we examined both dichotomous indicators for particularly high-risk patterns and several categorical definitions of these measures to determine whether there was a doseresponse association. We defined 2 measures to assess high prescription quantity. Recognizing that a patient may appropriately receive a prescription for a long-acting opioid plus a short-acting one for breakthrough pain, we first classified each claim as long- or short-acting and assessed quantity by using the maximum amount supplied of either type. To illustrate, our first quantity measure was whether a patient obtained more than a 210-day (about a 7-month) supply of either long- or short-acting opioids during the half-year. A person who obtained an 8-month supply of tramadol would fulfill this measure, but a person who obtained 4 months of fentanyl transdermal patches and 4 months of hydrocodone would not. Second, we defined a claims overlap measure according to whether a patient filled a prescription for the same opioid ingredient with the same duration of action (long- or short-acting) more than 1 week before the number of days supplied on the previous prescription were exhausted. The claim must have been for the same ingredient to exclude the possibility that a patient may have had poorly controlled pain requiring different types of opioids. For example, a patient who filled prescriptions for a fentanyl patch and hydrocodone tablets on the same day would not be coded as having overlapping claims. For sensitivity analyses, we also explored measures that did not adjust for combinations of long- and short-acting opioids (Supplement). Supplement. Supplemental Appendix To measure fragmented prescribing, we counted the number of prescribers and the number of pharmacies from which a patient obtained opioids during the half-year. Prescribers were identified by their National Provider Identifier, and pharmacies in the Part D event file were identified by using an encrypted tag specific to individual pharmacy locations. Our main analysis focused on a binary cutoff to define potential misuse: whether a patient obtained opioids from 5 or more prescribers or 5 or more pharmacies in each half-year (20, 21). We considered 2 additional measures that involved prescriptions crossing state lines, because some persons may have sought to evade new state regulations by obtaining opioids from prescribers or pharmacies in another state. Because provider addresses in publicly available files might be inaccurate, we inferred a state for each prescriber on the basis of the modal state of residence each year for beneficiaries who received nonopioid prescriptions from that prescriber; we used a similar approach for pharmacies. We counted the number of opioid prescriptions a patient obtained from an out-of-state prescriber or filled at an out-of-state pharmacy. Outcomes Our primary outcome was any occurrence of opioid overdose, defined as the presence of any opioid overdoserelated diagnosis code (also labeled as opioid poisoning in diagnosis descriptions [Supplement]) (14) in any position on a claim from any care setting (that is, an inpatient, an outpatient, or a carrier claim), in the year after a half-year with an opioid fill. These diagnoses included intentional and unintentional overdose or poisoning diagnoses attributed to prescription opioids, as well as suicide attempts or self-inflicted injuries involving opioids. We observed no evidence for redaction of opioid overdose claims from our data set, compared with previous work with nonredacted claims (Supplement Figure 1) (14). We measured overdoses in the full year (that is, 2 half-years) after the half-year of prescription opioid use to mimic the information available to a PDMP user observing data from the past half-year and interested in outcomes during the coming year. As a secondary outcome, we examined all-cause mortality during the year after a half-year with a Part D opioid claim. For patients with an opioid overdose incident in the subsequent 12 months, we defined an alternative secondary outcome, 30-day postoverdose mortality, for those who died of any cause during the 30 days after an opioid overdose claim. In the absence of cause-of-death data, this measure serves as a marker for deaths that may be directly related to opioids. In the Supplement, we demonstrate similar results from measuring mortality within 15 or 90 days after an overdose event. Patient Covariates We collected information on beneficiaries regarding age, sex, race, and state of residence; whether they had dual eligibility for Medicaid and Medicare coverage; and whether they were entitled to Medicare because of age or disability. Using the medical claims, we created binary variables for each Charlson comorbidity score category based on claims in the current year (22). We also assessed a patients average daily MED as a key measure potentially correlated with both opioid misuse and adverse outcomes (23, 24). However, because many appropriate clinical factors may influence a per

The Impact of Insurance Expansions on the Already Insured: The Affordable Care Act and Medicare

Some states have not adopted the Affordable Care Act (ACA) Medicaid expansions due to concerns that the expansions may impair access to care and utilization for those who are already insured. We investigate such negative spillovers using a large panel of Medicare beneficiaries. Across many subgroups and outcomes, we find no evidence that the expansions reduced utilization among Medicare beneficiaries, and can rule out all but very small changes in utilization or spending. These results indicate that the expansions in Medicaid did not impair access to care or utilization for the Medicare population.

Drug Firms' Payments and Physicians' Prescribing Behavior in Medicare Part D

In a pervasive but controversial practice, drug firms frequently make monetary or in-kind payments to medical providers. Critics are concerned that drug firms are distorting prescribing behavior away from the best interests of patients, while defenders of the practice claim that payments arise from the need to educate providers about changing drug technologies. Using two different identification strategies, we investigate the effect of payments from drug firms on individual-level prescribing behavior in Medicare Part D. We find that individuals whose providers receive payments from a drug firm tend to increase expenditure on the firms products. Our method accounts for the selection of physicians into payments (which may result if, e.g., pharmaceutical firms target payments to physicians who see a large number of patients) and our finding holds even when we look over time within individuals who change providers. However, using hand-collected efficacy data on four major therapeutic classes, we find that those receiving payments also prescribe higher-quality drugs on average. In addition, we examine four case studies of major drugs going off patent. Providers receiving payments from the firms experiencing the patent expiry transition their patients just as quickly to generics as prescribers who do not receive such payments. These results suggest that, absent other interventions to facilitate education, policies such as the Physician Payments Sunshine Act may reduce the efficacy of drugs prescribed.