 To describe lipid management over time in a cohort of insured patients with diabetes and evaluate differences between African American and white patients.

 Automated claims data were used to identify a cohort of 11,411 patients with diabetes in 1997 to 1998. Patients were followed through 2007.

Rates of hypercholesterolemia testing, treatment, and goal attainment were measured annually. Treatment was determined by a claim for lipid-lowering agents, and goal attainment was defined as a low-density lipoprotein cholesterol (LDL-C) level <100 mg/dL.

 During the study period, LDL-C testing increased from 48% to 70% among African American patients and from 61% to 77% among white patients. Treatment with lipid-lowering drugs increased from 23% to 56% among African American patients and 33% to 61% among white patients. The proportion at goal increased from 35% to 76% and from 24% to 59% among white and African American patients, respectively. African American patients were less likely to be tested for LDL-C (odds ratio [OR] 0.79; 95% confidence interval [CI] 0.73-0.86), treated with lipidlowering agents (OR 0.72; 95% CI 0.65-0.80), have their medication dosage altered (OR 0.65; 95% CI 0.59-0.73), or attain LDL-C goal (OR 0.59; 95% CI 0.56-0.63) compared with white patients.

Although rates of LDL-C testing, treatment, and goal attainment improved over time, racial disparities in dyslipidemia management continued to exist. Further studies to determine the causes of differences in management by race are warranted.

		Using claims data, we determined that insured African Americans with diabetes were less likely than whites to be treated with lipid-lowering agents, have their medication altered, or reach low-density lipoprotein cholesterol (LDL-C) goal.

		Rates of LDL-C testing, treatment, and goal attainment significantly improved in both races over time.

		Nonadherence to lipid-lowering drugs was higher among African Americans than among whites.

Patients diagnosed with diabetes mellitus (DM) are at higher risk for cardiovascular disease (CVD) events and mortality than patients with no history of DM.

 In an effort to reduce this risk, national guidelines recommend strict hypercholesterolemia management, among other measures, in patients with DM. Racial disparities have been observed not only in the prevalence of DM and its complications but also in the management of hypercholesterolemia (lipid testing, treatment, and control/goal attainment).

 In 1 published study, investigators found that even among patients treated for hypercholesterolemia, African American patients were less likely to reach their low-density lipoprotein cholesterol (LDL-C) goal compared with white patients.

 Several reports have shown that even among patients with coronary heart disease (CHD), DM, or hypertension, African Americans are less likely to receive 3-hydroxy-3-methylglutaryl coenzyme A reductase inhibitors (ie, statin therapy) for dyslipidemia and/or achieve LDL-C control compared with white patients.

Disparities in access to healthcare and healthcare-seeking behavior may explain why lipid management impact is better among whites than among African Americans.

 These disparities have been attributed to difficulties in accessing healthcare among uninsured minorities, and lower socioeconomic status has been associated with an inferior quality of care received.

 Even among insured African Americans, quality of care, particularly lipid treatment and control, is inferior to that received by other racial groups.

 However, some findings suggest that patients of differing race and ethnic groups receive equal benefits when treated appropriately.

 Further complicating matters, previous studies have also shown racial differences in adherence to lipid-lowering medications among patients with diabetes which might contribute to ethnic and racial disparities.

 This paper builds on previous literature by including information on care processes, clinical outcomes, patient sociodemographic and clinical characteristics, office visit and prescription drug copayments, treatment intensification, and medication adherence in the same study. With its large sample size, high proportion of African Americans, and long observation period, this study strengthens and expands previous findings.

To more fully investigate the question of racial disparities in lipid control, we describe annual rates of testing, treatment, and LDL-C goal achievement over a 10-year period in a large cohort of insured patients with diabetes receiving care in an integrated healthcare delivery system. We also evaluate whether dyslipidemia management differed between African American and white patients after controlling for numerous patient clinical characteristics and sociodemographic factors. This includes controlling for economic barriers beyond the mere presence of health insurance with variables such as prescription drug and physician office visit copayments. Further, we explore whether racial differences in rates of LDL-C goal achievement could be explained by racial differences in treatment intensification of and adherence to lipid-lowering drugs.

All study patients received care through a large integrated health system serving southeastern Michigan. This health system includes a 900-member multispecialty salaried medical group that delivers care in Detroit and surrounding communities. We identified a retrospective cohort of patients from multiple sites who were managed by the medical group and had insurance coverage through an affiliated health maintenance organization. All patients had prescription coverage, with tiered copayments based on the covering entity’s formulary. We followed cohort members from baseline (January 1, 1997, to December 31, 1998) until the first of either death, health plan disenrollment, or the end of the study period (ie, December 31, 2007). The Institutional Review Board of the Henry Ford Health System approved the study as described.

We used the National Committee for Quality Assurance’s Health Plan Employer Data and Information Set (HEDIS) criteria to identify patients diagnosed with diabetes in the baseline period.29 The patient had to meet at least 1 of the following 3 diagnostic definitions for diabetes: (1) 

2 outpatient visits with a diagnosis of DM (

 250.xx); (2) a dispensing for insulin or an oral hypoglycemic medication (therapeutic class codes C4G, C4K, C4L, C4M, and C4N); or (3) a mean glycated hemoglobin (A1C) level 

126 mg/dL on 2 separate occasions with a mean A1C 

6.5%. Patients had to be 18 years or older at baseline and be continuously enrolled in the health plan, with pharmacy benefit coverage, for the 2-year baseline period.

Information on patient characteristics, including age, sex, marital status, and race, was available from electronic data sources maintained by the health system. At the time of this data collection, race was usually based on self-report, but could have been assigned by administrative staff at the time of the initial clinical encounter. Medical claims and encounter data were used to identify and construct the following: clinical characteristics and diagnosis variables, the Deyo adaptation of the Charlson Comorbidity Index,

 and measures reflective of medical care use (ie, frequency of outpatient visits and cardiology visits). Measures of laboratory test receipt and test results were obtained from an automated clinical laboratory system. Prescription drug claims data were used to compile prescription drug use and adherence measures. Medical group and health plan databases were linked using patients’ unique medical record numbers. Use of automated data to identify patients with diabetes has been previously validated.

To examine trends in lipid management, we created indicator variables for LDL-C testing, treatment with a lipidlowering agent (therapeutic class codes D7L, M4E, and M4F), and LDL-C goal attainment during baseline and in each year of follow-up. Patient LDL-C values were based on the average annual value for the 2-year baseline period and for each follow-up year. The clinical laboratory system was also used to derive variables reflective of baseline and annual mean A1C.

Comorbidity scores were calculated using the diagnostic data available during the baseline years to construct the Deyo adaptation of the Charlson Comorbidity Index.

2 outpatient diagnoses) and procedural data were used to construct indicators for baseline cardiovascular risk factors and diseases. These included hypertension, left ventricular hypertrophy (LVH), peripheral vascular disease (PVD), cerebrovascular diseases (including stroke or transient ischemic attack), CHD (including unstable angina and myocardial infarction), retinopathy, non-traumatic lower extremity amputation, end-stage renal disease (ESRD), and smoking status. Smoking status was determined by the presence of at least 1 diagnosis of tobacco use (

code 305.1) in the 1997-1998 baseline period.

Prescription claims data were used to compute the continuous measure of medication gaps (CMG), a measure of nonadherence in pharmacotherapy.

 CMG is the sum of treatment gaps in medication refills over the total number of days in the observation period. Since some subjects were taking more than 1 lipid-modifying medication, CMGs for all drugs within the lipid-lowering class were averaged to create a composite CMG. CMG was calculated for 4 classes of hyperlipidemic drugs: statins, fibrates, nicotinic acid, and bile acid sequestrants. Within each drug class, CMG indexes were computed for those patients who filled at least 1 prescription per year (N = 2553 for statins, N = 415 for fibrates, N = 11 for nicotinic acid, and N = 63 for bile acid sequestrants) in the period between the first prescription claim after January 1, 1997, and the last prescription available or December 31, 2007 (ie, if the last prescription extended past the last day of observation).

medication nonadherence was measured as 1-CMG (ie, reverse-coded) and multiplied by 100 to provide a scale of 0 to 100. Therefore, higher scores reflected better medication adherence. Dichotomous versions of the reverse-coded CMG were created to categorize patients as either “adherent” or “nonadherent.” Patients whose reversecoded CMG was less than 80% (ie, a gap in therapy greater than 20%) were classified as nonadherent whereas patients with a reverse-coded CMG greater than or equal to 80% were classified as adherent. Patients with no prescribed treatment were considered to have adherence equal to zero in the multivariable models. A cutoff of 80% has been used historically to differentiate adherent from nonadherent behavior

 and it is associated with the likelihood of achieving LDL-C goal.

Sociodemographic information included age, race, gender, and marital status. Medical claims and encounter data were used to determine clinic, copayment amounts, and number of outpatient and specialty care (cardiology) physician visits during the 2-year baseline period. Race was categorized as “white,” “African American,” or “other.” The “other” population was included in the analyses and is presented in the tables. However, interpretation of these data is difficult due to the heterogeneity of the population and its small sample size (n = 423; 3.7%); therefore, results specific to this subgroup are not described. Residential street address was used to estimate median household income and level of education (as represented by proportion of males and females with high school or lower level of education) using geographical information system technology which assigned values based on the census block-group of residence.

Rates of testing, treatment, and LDL-C goal attainment stratified by race were estimated and compared by chi-square tests for each follow-up year (1999-2007). Values were declared missing if the patient was not continuously enrolled for the entire year of interest. Multivariable logistic regression models (generalized estimating equation, or GEE)

 that account for repeated events (ie, testing, treatment, dose adjustment, and goal achievement during the subsequent follow-up years), while controlling for baseline factors, were used to estimate racial effects on testing, treatment, and goal attainment variables.

We used a stepped approach to modeling the association between race and lipid management among patients with diabetes. Several important covariates could confound and/or mediate an association between race and lipid management. A stepped approach to multivariate modeling allowed us to evaluate the effects of adding important covariates to the model. The first model was adjusted by year of follow-up. In addition to year of follow-up, the second model included baseline sociodemographic characteristics (gender, marital status, estimated household income, level of education, prescription drug and office visit copayments); lipid measures (LDL-C); baseline cardiovascular risk factors, such as hypertension, LVH, retinopathy, smoking status, PVD, CVD, CHD, amputation, ESRD; A1C; Charlson Comorbidity Index; number of outpatient visits; and number of cardiology visits. To determine the role of medication nonadherence on racial differences in LDLC goal attainment, the third model included a flag for treatment with a lipid-lowering drug along with a dichotomous medication adherence variable. These time-varying variables reflected the annual averages for the previous calendar year (or the last available year if the previous calendar year was not available). Last, to control for African American patients clustering at particular care sites, clinic was added as an additional covariate to the second and third models.

A total of 11,411 patients with diabetes were identified at 28 sites during the baseline period. The mean and median years of observation were 6.2 and 7.0, respectively, for both the overall study population and for each racial category. African American patients comprised 43.0% of the sample. Among African American patients, the mean age at baseline was 56.9 years (SD ± 12.9 years) and 55.5% were female. Among the white patients, the mean age at baseline was 58.9 years (SD ± 13.0 years) and 47.3% were female. The majority of cohort members had insurance coverage via an employer and were married; however, a higher proportion of the white patients were married relative to the African American patients. The mean number of outpatient visits during the 2-year baseline period was 12.0 for African Americans and 12.7 for whites; the mean number of cardiology visits during the baseline period was 0.6 for African Americans and 0.8 for whites. Racial distributions across clinics varied considerably. Of the 28 sites, 11 (39%) had >25% African American patients, 6 (21%) had >40% African American patients, and 3 (11%) had >85% African American patients. Hypertension was highly prevalent in the total patient population (77.2%) and African American patients had a prevalence rate at least 10% higher than the other groups. Significant differences in rates were also found among other baseline clinical covariates, including PVD, CVD, CHD, LVH, ESRD, and smoking status. The proportion of African American patients with LDL-C testing in the 2-year baseline period was 56.7%. At baseline, 18.3% of African American patients had an LDL-C level <100 mg/dL and 23% were treated with lipid-lowering therapy. Among white patients, the proportion with LDL-C testing in the 2-year baseline period was 69.9%. At baseline, 25.6% of white patients had an LDL-C level <100 mg/dL and 33% were treated with lipid-lowering therapy. Nonadherence to lipid-lowering agents was higher among African American patients (41% of patients categorized as nonadherent) compared with white patients (26% of patients categorized as nonadherent); these differences were statistically significant (

 <.0001). These baseline sociodemographic, healthcare access, and clinical characteristics are shown in 

Over the follow-up period (1999-2007), 2 trends in lipid management were identified. First, rates of LDL-C testing, treatment, and goal attainment significantly improved in both races over time (

<.001). Second, the racial disparities in LDL-C testing, treatment, and goal attainment remained highly significant over time (

Among African American patients, LDL-C testing rates increased from 48% in 1999 to 70% in 2007. The overall use of lipid-lowering agents increased from 23% to 39%. Consistent with increased testing and treatment, there was significant improvement in the proportion of patients at LDL-C goal. The proportion of African American patients with average LDL-C levels <100 mg/dL rose by 35% (from 24% to 59%) between 1999 and 2007.

Articles

CLINICAL

African Americans with diabetes are less likely than whites to be treated with lipid-lowering agents, have their medication altered, or reach LDL-C goal.

Racial Disparities in Lipid Control in Patients With Diabetes

To assess the performance of a health information technology platform that enables automated measurement of asthma care quality using comprehensive electronic medical record (EMR) data, including providers’ free-text notes.

 Retrospective data study of outpatient asthma care in Kaiser Permanente Northwest (KPNW), a midsized health maintenance

organization (HMO), and OCHIN, Inc, a group of Federally Qualified Health Centers.

We created 22 automated quality measures addressing guideline-recommended outpatient asthma care. We included EMRs of asthma patients aged 

12 years during a 3-year observation window and narrowed this group to those with persistent asthma (13,918 KPNW; 1825 OCHIN). We validated our automated quality measures using chart review for 818 randomly selected patients, stratified by age and sex for each health system. In both health systems, we compared the performance of these measures against chart review.

Most measures performed well in the KPNW system, where accuracy averaged 88% (95% confidence interval [CI] 82%-93%). Mean sensitivity was 77% (95% CI 62%-92%) and mean specificity was 84% (95% CI 75%-93%). The automated analysis was less accurate at OCHIN, where mean accuracy was 80% (95% CI 72%-89%) with mean sensitivity and specificity 52% (95% CI 35%-69%) and 82% (95% CI 69%-95%) respectively.

To achieve comprehensive quality measurement in many clinical domains, the capacity to analyze text clinical notes is required. The automated measures performed well in the HMO, where practice is more standardized. The measures need to be refined for health systems with more diversity in clinical practice, patient populations, and setting.

	Quality improvement requires comprehensive quality measurement, which in turn requires robust automation to be broadly applicable and reliable.

		We designed a health information technology platform and quality assessment method that overcome informatics challenges created by text-based guidelines, nonstandard electronic clinical data elements, and text clinical documentation.

		We implemented our method in 2 healthcare systems to assess outpatient asthma care.

		The automated measures generally performed well in the health maintenance organization setting, where clinical practice is more standardized. Refinements are needed for health systems with more diversity in clinical practice, patient population, and setting.

To guide quality improvement, we must comprehensively measure the quality of healthcare. Currently, this process is hindered by expensive, time-consuming, and sometimes inconsistent manual chart review.

 Electronic medical records (EMRs), which have become more prevalent as a result of the American Recovery and Reinvestment Act of 2009, promise to make routine and comprehensive quality measurement a reality.

 However, informatics challenges have hindered progress: care guidelines may not be specified to allow for automated measurement; the needed data are not standardized and are subject to variations in EMRs and clinical practice; and much of the data required are in the free-text notes that care providers use to document clinical encounters, which may not be accessible.

Our research team, funded by the Agency for Healthcare Research and Quality (AHRQ), designed and implemented an automated method to comprehensively assess outpatient asthma care. In doing so, we aimed to develop a platform that could automate care measurement for any condition.

We conducted a retrospective data study of outpatient asthma care in 2 distinct healthcare systems: Kaiser Permanente Northwest (KPNW) and the Federally Qualified Health Centers (FQHCs) associated with OCHIN, Inc. We obtained institutional review board approval and executed Data Use Agreements between research organizations for this study.

OCHIN serves the data management needs of FQHCs and other community health centers that care for indigent, uninsured, and underinsured populations. OCHIN has licensed an integrated Practice Management and EMR data system from Epic Systems. We approached 8 FQHCs associated with OCHIN (caring for 173,640 patients at 44 locations through 2010), and all agreed to participate in this study.

KPNW is a nonprofit, groupmodel health maintenance organization (HMO) that provides comprehensive, prepaid healthcare to members. KPNW serves about 475,000 members in Oregon and Washington. All patient contacts are recorded in a single, comprehensive EMR—the HealthConnect system.

We included the electronic records of patients 12 years or older at the start of 2001 (KPNW) or 2006 (OCHIN) who had at least 1 diagnosis code for asthma (35,775 in KPNW and 6880 in OCHIN). We then narrowed this population to those defined as having persistent asthma (

) to reach the target population of interest (13,918 KPNW; 1825 OCHIN).

To automate the assessment of asthma care quality, recommended care steps from credentialed guidelines needed to be converted into quantifiable measures. The development of each measure began with a concise proposition about the recommended care for specific patients, such as “patients seen for asthma exacerbation should have a chest exam.” We used an 8-stage iterative process to ensure that our quality measures would be comprehensive and current. This process included 4

vetting steps with local and national experts. We identified 25 measures from comprehensive, rigorous quality measure sets, primarily derived from RAND’s Quality Assessment system.

 We then revised the measures to reflect updated guidelines

 and restricted our attention to asthma care in the outpatient setting, resulting in a set of 22 measures that we labeled the Asthma Care Quality (ACQ) measure set (

Each measure was specified as a ratio based on its applicability to individual patients (denominator) and evidence that the recommended care had been delivered (numerator). Performance on each measure can then be reported as the percentage of patients who received recommended care from among those for whom that care was indicated. For example, the national RAND study of McGlynn and colleagues demonstrated that across 30 clinical conditions, Americans received about 55% of recommended care.1 Ratios for each measure can be produced at the patient, provider, clinic, and healthsystem levels.

We investigated providers’ clinical practices related to each measure in the ACQ measure set and how care was documented and stored in the EMR. Each measure’s numerator requires a “measure interval,” which is the time window during which the care events must take place. The measure interval is oriented around an index date that is a property of denominator inclusion. For example, for the measure “patients seen for asthma exacerbation should have a chest exam,” the index date is the exacerbation encounter and the measure interval includes only that encounter. On the other hand, for the measure “patients with persistent asthma should have a flu vaccination annually,” the index date is the event that qualifies the patient as having persistent asthma and the measure interval is operationalized to include encounters 6 months before through 12 months after the index date.

For each of the 22 quality measures, we first defined an observation period (in our case, 3 years of clinical events) and divided it into a period for denominator qualification (the selection period) followed by an evaluation period, during which, in most cases, the prescribed care events were identified. For this study, we used a 2-year selection period based on a modified version of the Healthcare Effectiveness Data and Information Set asthma criteria to identify patients with persistent asthma (used in all of our measures) or those presenting with an asthma exacerbation (used in 36% of the measures in our set). We identified patients as having persistent asthma if they met minimum criteria for asthma-related utilization or if this diagnosis could be specifically determined from the provider’s clinical notes (Table 1). Asthma exacerbation criteria were based on hospital discharge diagnosis or an outpatient visit associated with a glucocorticoid order/dispensing and a text note about exacerbation.

We designed a quality measurement system as a “pipeline” of transformation and markup steps taken on encounter-level EMR data. The goal was to capture all of the clinical events required to assess care quality (

Data begin traveling through the pipeline when they are extracted from each EMR system’s data warehouse. These data extracts—produced by a component called the EMR Adapter&mdash;contain data aggregated into records at the encounter (visit) level for all patients. In our study, these records included the coded diagnoses, problems, and medical history updates; medications ordered, dispensed, and noted as current or discontinued; immunizations, allergies, and health maintenance topics addressed; and procedures ordered, progress notes, and patient instructions.

The data are then exported from the EMR data warehouse (typically, a relational database) into file-based eXtensible Markup Language (XML) documents according to a local specification. The first transformation step involves converting locally defined XML formats into a common, standard XML format conforming to the HL7 Clinical Document Architecture (CDA) specification.

METHODS

The authors discuss the design and evaluation of a health information technology platform that enables comprehensive, automated assessment of care quality in electronic medical records.

Automating Care Quality Measurement With Health Information Technology

 To examine knowledge of and financial barriers to early adoption of human papillomavirus (HPV) vaccination, specifically the role played by insurance, income, and affordability (measured by forgoing or delaying needed medical care due to cost/no insurance).

We used the 2007 California Health Interview Survey. Females aged 18 to 26 years (n = 1840) and parents with daughters aged 8 to 17 years (n = 5765) were analyzed separately.

Logistic regression models were used with the following dependent variables: (1) heard of the HPV vaccine, (2) received 1 dose only, (3) completed the series, (4) have not previously heard of HPV vaccine but interested in receiving it, and (5) interested and willing to pay $360 for it.

 Individuals enrolled in private health maintenance organizations (HMOs) were more likely to have heard of the vaccine compared with the uninsured and those enrolled in public HMOs. Young adults enrolled in private HMOs were also more likely to have initiated HPV vaccination or completed the series compared with uninsured

young adults or those insured in non-HMO plans. Higher income parents were more willing to pay the cost of the vaccine. Forgoing needed care due

to costs led to lower odds of initiating HPV vaccination among parents and completing the series among young adults.

: Strategies to increase HPV vaccination rates should consider insurance or cost barriers for adults and those with high medical care expenditures. Disparities in receipt of the HPV vaccine are likely to continue without targeted outreach to more vulnerable populations.

Insurance coverage, income, and other indicators of affordability affect human papillomavirus (HPV) vaccination rates.

		To ensure effective prevention of HPV, it is important to provide targeted education to subgroups, particularly young adults aged 22 to 26 years, parents 56 years and older with young daughters aged 8 to 17 years, Latinos and Asian Americans, and members of public HMOs.

		Strategies to increase HPV vaccination rates should consider insurance or cost barriers for adults and those with high medical care expenditures.

Infection with human papillomavirus (HPV) is a serious public health concern with significant morbidity and financial consequences. The most prevalent strains of HPV, subtypes 16 and 18, are leading causes of cervical and anal cancers nationally.

 Cervical cancers caused by persistent HPV infections have led to an estimated 8.1 deaths per 100,000 women nationally and 8.3, 6.4, and 11 deaths per 100,000 for whites, African Americans, and Latinos, respectively, in California.

The prevalence of HPV is highest (45%) among young women aged 20 to 24 years,

and the likelihood of HPV infection is about 75% among adults aged 15 to 49 years.

 The costs of treating HPV and its related disease are estimated at or above $4 billion annually.

Two HPV vaccines, Gardasil and Cervarix, became available in 2006 and 2009, respectively.

 Gardasil guards against subtypes 6, 11, 16, and 18; Cervarix guards against subtypes 16 and 18 only. Both are nearly 100% effective against precancerous lesions, with Cervarix more effective against persistent infections (nearly 100%) than Gardasil (86-89%).

 The vaccines are recommended for young females aged 12 to 26 years and may have some benefits for older women.

 Financial constraints may be a significant barrier to HPV vaccination. The vaccine, at an estimated cost of $360 for the 3-dose vaccine plus the associated administration and visit fees (estimated at $92), is too expensive without insurance coverage for most low-income families.

Free vaccine for children may be obtained from various sources including the federally funded Vaccine for Children (VFC) program, which covers about 40% of poor children

; and the Immunization Grant Program (Section 317), which is subject to budget shortfalls.

 Vaccine manufacturers may cover uninsured adults and Medicaid may cover vaccine costs for adults.

The scope of these forms of coverage is unknown and the underinsured may still have limited access.

Among privately insured patients, the vaccine may not be covered or covered fully.

 In early 2008, some preferred provider organizations (PPOs) in California did not cover the vaccine.

 Lack of coverage or knowledge of coverage for the vaccine is a frequently reported reason for not vaccinating against HPV.

 Differential rates of compliance with the federal Advisory Committee on Immunization Practices by PPOs versus health maintenance organizations (HMOs) shortly after availability of the HPV vaccine, as well as higher cost-sharing requirements by PPOs, may have led to additional barriers to vaccination according to the type of private insurance coverage. Overall, high costs and inadequate coverage of the HPV vaccine are likely to lead to uneven vaccination rates among lowincome and other disadvantaged groups. Understanding the impact of financial barriers on adherence to HPV vaccination is needed to improve HPV vaccination rates and to prevent HPV infection and its disease sequelae among all eligible-age females.

We used the 2007 California Health Interview Survey (CHIS) for this study. The survey is a representative randomdial telephone survey of more than 53,600 households, including about 51,000 adults, and is conducted in English, Spanish, Mandarin, Cantonese, Korean, and Vietnamese languages to capture the great majority of non—English speaking populations. It has an overall response rate of 21.2% and has been found in multiple studies to be representative of the California population.

 The sample included all females aged 18 to 26 years (n = 1840) as well as girls aged 8 to 17 years (n = 5765) who are recommended to receive HPV vaccination. Parents of teenagers answered the CHIS vaccination questions to increase the reliability of responses and because the vaccination decision is most likely to be made by parents of minors.

Conceptual Framework, Research Questions, and Hypotheses

We used Andersen and Davidson’s conceptual framework for healthcare use.

 Based on this conceptual framework, HPV vaccination or intent to vaccinate was determined by enabling factors (insurance, income, and affordability of care), predisposing factors (eg, age, race), and need (eg, health status). Our primary interest was to assess the impact of enabling factors. Specifically, we wanted to determine how insurance, income, and affordability of care affect HPV vaccination rates and patients’ willingness to pay.

We hypothesized that rates of HPV vaccination or intent to vaccinate and willingness to pay for the vaccination are lower for individuals without insurance coverage or with PPO or fee-for-service (FFS) coverage, for some low-income populations (as measured by federal poverty level [FPL]), and for individuals who reported that care has been unaffordable (as measured by having delayed or forgone needed healthcare due to cost/no insurance in the past year). We assumed that some eligible individuals fell through the gaps despite availability of VFC and other programs. We also examined the role of enabling factors in knowledge of the HPV vaccine, hypothesizing that vaccine knowledge was lower among lower income and uninsured groups who were not targeted by vaccine manufacturers’ advertising campaigns early on.

Five different dependent variables are included in this study. Adults and parents (

) were asked (1) if they had heard of the HPV vaccine prior to the survey. Those who had heard of the vaccine were asked (2) if they/their daughters had received at least 1 dose of the HPV vaccine, and if so (3) whether they/their daughters had received all 3 doses. Respondents who had not heard of the vaccine were informed of the existence and intent of the vaccine and were then asked (4) if they were interested in receiving the vaccine (adults) or giving it to their daughter if a doctor recommended it. Those who were interested in receiving the vaccine were asked (5) if they were willing to pay $360 for the 3-dose vaccine.

The independent variables of interest were insurance, poverty level, and affordability, as indicated by forgone or delayed care due to cost/no insurance. Insurance coverage was defined as private HMO, private PPO, public HMO, public FFS, or uninsured. The private HMO category includes variations in HMO types including point-of-service plans and exclusive provider organizations. The private PPO category includes PPOs and FFS options; the latter is an increasingly uncommon form of insurance in California. Public HMOs may also include some exclusive provider organizations but are primarily HMOs. Public non-HMO coverage in California is an FFS form of coverage. The distinctions in form or coverage were included to account for differences in benefit levels, cost-sharing requirements, restrictions imposed on use of various services, provider networks, and provider compensation and incentives. HMOs are more likely to cover all vaccinations and have lower cost sharing but may impose authorization requirements on services, while PPOs are more likely to have higher cost sharing but fewer restrictions on use. Public FFS coverage does not include cost sharing, but provider participation in the program and subsequent access to care may be more limited.

Poverty level was indicated by FPL—less than 100%, 100% to 299%, and 300% or higher&mdash;and was used to assess the individual’s ability to pay for the associated out-of-pocket costs of HPV vaccine if insured or the full cost of the vaccine if uninsured and unable to receive free vaccination. Affordability was indicated by reports of having forgone or delayed receipt of needed medical care due to cost/no insurance in the past year and was intended to account for financial barriers to access not

captured by other variables in the models. Forgone or delayed care due to other reasons was also included in the models to control for nonfinancial reasons such as lack of time that lead individuals and parents not to obtain the HPV vaccination. All models were also controlled for predisposing factors and other potentially confounding factors. Predisposing control variables included age, race/ethnicity, education, and family status. Age was categorized as 18 to 21 and 22 to 26 years for the young adult females and 18 to 35, 36 to 55, and 56 years and older for the parents. Race and ethnicity were defined as white, Latino, Asian American, African American, and other groups. Education was dichotomized into any college education versus less. Marital and family status was defined as married, single with children, and single without children in the adult sample, and single with children versus married with children in the parent sample. We included the receipt of the flu vaccine in the past year as a measure of a health-seeking behavior. Need was controlled for with perceived health status noted as excellent, very good, or good versus fair or poor.

Logistic regression models were used to assess the role of financial determinants of HPV vaccination, controlling for other potential modifiers. Insurance categories with very small sample sizes were combined in the models predicting completion of the vaccine series to improve reliability of findings. Health status and education were excluded from the model predicting completion of the vaccine series for adults due to small sample size and collinearity. All analyses were weighted and adjusted for the complex survey design of CHIS. The pseudo 

 was calculated based on the Nagelkerke method and indicates the approximate variance in the dependent variable explained by the independent variables.

The characteristics of the adult and parent samples are presented in 

. Most adult females aged 18 to 26 years (70%) and parents of girls aged 8 to 17 years (63%) had heard of the HPV vaccine prior to the survey. Among those who had heard of the vaccine, 16% of adults and 21% of girls aged 8 to 17 years had received at least 1 dose of the vaccine. Among those who had received at least 1 dose of the vaccine, 36% of adults and 43% of young girls had completed the 3-dose series. The respondents who had not heard of the vaccine were informed of the existence and purpose of the vaccine. Many adults (61%) responded that they would be interested in receiving it and 57% of parents reported that they would be interested if the doctor recommended it. Among those who said they would be interested or would follow the doctor’s recommendation, 41% of adults and 79% of parents said that they would be willing to pay $360 for the 3-dose vaccine. Nearly all adults (99%) and parents (98%) who were unwilling to pay for the vaccine reported willingness to receive it for free in a follow-up question to this effect (data not shown in table).

Predictors of Having Heard of the HPV Vaccine.

 The  odds of having heard of the vaccine were lower for those who were uninsured (odds ratio [OR] 0.5) and those with public HMO coverage (0.5), but higher for those with private PPO compared with private HMO coverage (

). Adults with family incomes under 100% of FPL had lower odds than those with incomes at or above 300% of FPL. The odds of forgoing or delaying needed medical care were not significant. The odds of having heard of the vaccine were also lower for Latinos and Asian Americans, and higher for those with some college education and single adults without children.

Predictors of Having Received at Least 1 Dose of HPV Vaccine.

 Among adults who had heard of the vaccine, the odds of having received at least 1 dose of the vaccine were lower for uninsured individuals (0.3) compared with those who had private HMO coverage. The odds were higher for younger adults and for those who received the flu shot. Poverty level and having forgone or delayed needed care were not significant. Predictors of Having Received All 3 Doses of HPV Vaccine. Among adults who had received at least 1 dose of the HPV vaccine, the odds of having completed the vaccine series were not significantly affected by insurance coverage or poverty, or by having forgone or delayed needed medical care due to cost/no insurance. However, the odds were lower for adults who did not get or delayed needed medical care in the past 12 months for other reasons (0.04) compared with those who did not forgo or delay care. The odds were lower for younger adults and Latinos.

Predictors of Interest in Receiving the HPV Vaccine. 

Among adults who had not previously heard of the vaccine, insurance did not play a significant role in the odds of being interested in the vaccine, but the odds were lower for those earning between 100% and 299% of FPL (0.7) compared with those earning 300% or more of FPL. The odds were higher for those who had forgone or delayed care due to cost/no insurance (1.6). The odds were higher for those who received the flu vaccine.

POLICY

Compared with insured adults, uninsured adults in California had less knowledge of the human papillomavirus vaccine and were less likely to be vaccinated.

Role of Insurance, Income, and Affordability in Human Papillomavirus Vaccination

 To evaluate the real-world use of venous thromboembolism (VTE) prophylaxis among medical inpatients and the impact of VTE prophylaxis on outcomes and cost.

 Retrospective analysis of patientlevel administrative claims data for medical inpatients at risk of VTE and linked outpatient data.

Data were analyzed from patients admitted to the hospital from 2005 to 2007 (calendar years) with a primary diagnosis of chronic heart failure, thromboembolic stroke, severe lung disease, acute infection, or cancer (index hospitalization), according to whether they received VTE prophylaxis or not. The number of VTE events, time to VTE event, length of hospital stay, and number of major or minor bleeding events were analyzed from the index date until the end of follow-up (180 days postdischarge) or death.

: Overall, 7127 of 13,293 patients (53.6%) received VTE prophylaxis. Prophylaxis significantly reduced the incidence of VTE compared with no prophylaxis (0.06% vs 3.44%, respectively;

 <.00001) and increased the median time to VTE (182 vs 27 days, respectively). Prophylaxis also significantly reduced the incidence of VTE in the 180 days postdischarge. Readmission rates were similar between groups. Major bleeding occurred in 1.57% of patients receiving low molecular weight heparin + warfarin versus <.6% receiving any other form of prophylaxis. The development of VTE or major or minor bleeding events significantly increased total medical costs versus no VTE events (

 This real-world analysis showed that thromboprophylaxis was underutilized in medical patients, even though the clinical and economic impact of VTE was significant.

A retrospective analysis of data from medical inpatients at risk of venous thromboembolism (VTE) (n = 13,293) examined the impact of VTE prophylaxis during admission on clinical outcomes and costs:

		During the index hospitalization, 53.6% received VTE prophylaxis.

		VTE prophylaxis significantly reduced the incidence of VTE up to 180 days postdischarge and prolonged median time to VTE development.

		Significantly higher total medical costs were incurred in patients developing VTE or major or minor bleeding.

		Greater emphasis should be placed on ensuring that at-risk medical patients receive VTE prophylaxis to reduce VTE events and associated costs.

Venous thromboembolism (VTE) is a significant medical problem, with an estimated 200,000 to 600,000 Americans developing VTE each year.

 It is estimated that more than three-fourths of hospitalized patients in the United States have at least 1 risk factor for VTE and 48% have 2 or more risk factors.

 VTE risk is lower in medical patients than in surgical patients, but still substantial (10%-20%).

 In addition, there is increasing evidence that medical patients are less likely than surgical patients to receive thromboprophylaxis, even when it is indicated or recommended.

 In the US cohort from the multinational Epidemiologic International Day for the Evaluation of Patients at Risk for Venous Thromboembolism in the Acute Hospital Care Setting (ENDORSE) study, 48% of medical patients at risk of VTE received the recommended prophylaxis, compared with 71% of at-risk surgical patients.

 Postoperative VTE has been clearly shown to increase length of hospital stay, medical costs, and mortality,

 but far less is known about the impact VTE has on medical inpatients.

The aim of this study was to analyze the effect of pharmacologic VTE prophylaxis among medical inpatients on the incidence and timing of VTE, readmission due to VTE, bleeding events, and cost of care in the 30, 90, and 180 days in the postdischarge period after the initial (index) admission.

This was a retrospective analysis of patient-level data from the MarketScan Hospital Drug Database (HDD) and linked outpatient files from the MarketScan Commercial and Medicare Supplemental Database from Thomson Reuters for calendar years 2005 to 2007. This is a proprietary database containing the largest collection of US employer-based patient data. Inpatient, outpatient, pharmacy, and enrollment data from MarketScan, together with linked HDD data, were used in this study. Data from 172 hospitals were identifiable for linkage in both hospital and claims databases and covered a geographically diverse area and both public and private health plans. These databases capture clinical and prescription data for the full continuum of care, including physician office visits; hospital stays; retail, mail order, and specialty pharmacies; and carve-out care.

Hospital data comprised submitted claims linked to detailed service-level hospital bills for each admission. Outpatient claims data were matched to each admission using the following patient level identifiers: date of admission, date of discharge, age, gender, and principal diagnosis. Admissions not uniquely identified by this key set of variables were excluded.

The study population consisted of all patients admitted to hospital with a primary diagnosis of chronic heart failure (CHF), thromboembolic stroke, severe lung disease, acute infection, or cancer, based on the 

International Classification of Diseases, Ninth Revision

) during calendar years 2005 to 2007. This was defined as the index hospitalization. The population was grouped into 2 cohorts: those who developed VTE during index hospitalization and those who did not. VTE during hospitalization was identified by the presence of

 codes for deep vein thrombosis (DVT; 451.1x-451.81, 451.83-

451.9x, 452.xx, 453.2-453.9x) or pulmonary embolism ([PE] 415.1x) on hospital records. In order to ensure that patients were not erroneously designated as having VTE, true VTE was defined as any VTE event for which anticoagulant therapy was prescribed within 15 days of diagnosis.6 In accordance with current privacy rule guidelines, no patient’s identity or medical records were disclosed, except in compliance with applicable law.

Patient age, gender, and comorbidity data were collected at the index hospitalization. Comorbidity information included the Charlson Comorbidity Index,

 and a primary or secondary diagnosis (based on

 codes) of congestive heart failure, peripheral arterial disease, acute coronary syndromes, hyperthyroidism, obesity, diabetes, hypertension, ischemic or hemorrhagic stroke, noncentral nervous system systemic embolism, transient ischemic attack, catheter ablation, dyspepsia, or preperiod VTE (see 

). The preperiod began 180 days before the index event (date of hospital admission).

Outcomes were analyzed according to whether or not patients received any anticoagulant prophylaxis prior to VTE or true VTE diagnosis. In addition, the following prophylactic treatment groups were identified: low molecular weight heparin (LMWH) only; warfarin only; unfractionated heparin (UFH) only; fondaparinux; LMWH + warfarin; UFH + warfarin. The drug use observation period was defined as time from admission for the index hospitalization until 30 days after index hospital discharge.

All outcomes were measured from the index admission date up to 180 days postdischarge or death, whichever occurred first. These outcomes were: the number of VTE and true VTE events; time to the VTE event; length of hospital stay (calculated at patient level); the number of major or minor bleeding events using

); the number of patients readmitted for true VTE within 30, 90, and 180 days after index hospital discharge, based on subsequent hospital admission diagnosis of true VTE, or primary or secondary diagnosis of true VTE within 1 to 2 days of subsequent hospital admission; the number of patients readmitted with major bleeding within 30, 90, and 180 days after index hospital discharge, based on subsequent hospital admission diagnosis (primary or secondary) of major bleeding, or primary or secondary diagnosis of major bleeding 1 week before or 1 week after subsequent hospital admission; discharge status (ie, home, short-term hospital stay,

transfer to other facility, miscellaneous); and all-cause healthcare costs paid by the patient and health plan. These costs were computed for medical services (inpatient stay, emergency department visits, and ambulatory care, including physician visits and other outpatient services), pharmacy dispensing, and total combined costs. Prices were adjusted using the annual medical care component of the consumer price index to reflect inflation between the year of the claim and 2007.

All study variables, including baseline and outcome measures, were analyzed descriptively as numbers and percentages for dichotomous and polychotomous variables, and as means, medians, standard deviations (SDs), and percentiles for continuous variables. Bivariate comparisons of baseline variables and outcomes were made using the appropriate test (

2 test), depending on the distribution of the variable. Standardized differences were calculated. The time to VTE event was calculated using Kaplan-Meier curves and analyzed using Cox regressions. A multivariable risk adjustment model was undertaken to estimate outcome measures, incorporating age, gender, region, baseline Elixhauser Comorbidity Index, CHF, thromboembolic stroke, severe lung disease, acute infection, and cancer as confounding factors. Total costs were estimated using log transformation and general linear models, depending on the distribution and presence of heteroskedasticity.

The study cohort comprised 13,293 medically ill hospitalized patients (

). The most common reasons for hospitalization were severe lung disease (n = 4473; 33.65%) and cancer (n = 3050; 22.94%). The mean (SD) age of the cohort was 66.5 (17.0) years; 48% of patients were male. Patients with CHF or stroke were older than those in other subgroups (Table 1). The cancer subgroup had the highest proportion of patients with an Elixhauser Comorbidity Index 

2 (16%). Overall, 399 patients (3.0%) had experienced a VTE in the 6 months preceding the index hospitalization.

During the index hospitalization, 7127 patients (53.6%) received thromboembolic prophylaxis and 6166 (46.4%) did not. The most commonly prescribed agents were UFH (n = 3531) or LMWH (n = 3390), followed by warfarin (n = 1629); few patients (n = 42) received fondaparinux. The duration of treatment was longest for warfarin, with a mean of 21.49 days compared with 12.98 days for fondaparinux, 6.30 days for LMWH, and 5.24 days for UFH.

Of the 7127 patients who received any anticoagulant prophylaxis, 4 (0.06%) developed a true VTE, compared with 212 (3.44%) of the 6166 patients who did not receive any anticoagulant prophylaxis (

). Patients who did not fall into any of the 6 anticoagulant categories described in columns 3 through 8 were included in column 2, “Any Anticoagulant.” The patient number for this category will therefore be greater than the total for the 6 treatment categories specified.

Overall, major bleeding developed in 37 (0.52%) patients receiving VTE prophylaxis and 42 (0.68%) receiving no prophylaxis (

 = .226); the corresponding rates of minor bleeding were 373 (5.23%) and 364 (5.90%), respectively (

 =.0924). With the exception of patients receiving LMWH + warfarin, who had a major bleeding rate of 1.57%, major bleeding occurred in <.6% of patients receiving any of the other anticoagulant therapies. The rate of major bleeding was significantly higher in patients receiving LMWH + warfarin versus LMWH alone (1.57% vs 0.21%; 

 = .0002; Table 2). Minor bleeding occurred in 4.4% to 7.4% of patients across all groups (including those receiving no VTE thromboprophylaxis), and there were no significant differences between groups (Table 2). Results were similar in the risk-adjusted analysis, except

that the minor bleeding rate was significantly higher with UFH + warfarin versus LMWH alone (

VTE prophylaxis is underutilized in medical patients in US hospitals, but the occurrence of VTE has a major clinical and economic impact.

Thromboembolism Prophylaxis in Medical Inpatients: Effect on Outcomes and Costs

		There is a link between high levels of social capital and good health.

		Care delivery models have historically ignored opportunities to enhance social capital between providers and patients and among patients themselves.

		 As resources to manage care diminish, medical practices and payers should increasingly consider how medical practices and hospitals can leverage social capital as a means of improving the quality and efficiency of healthcare.

In 2001, the political scientist Robert Putnam described the causes and effects of declining social connectedness in his book

 Among Putnam’s conclusions: Americans are both happier and healthier when they have strong relationships with one another. Not surprisingly, the literature offers several pillars of support for Putnam’s thesis that social connectedness is associated with improved clinical outcomes.

As delivery reform unfolds and payers and delivery systems begin to experiment more broadly with new organizational forms such as accountable care organizations and patientcentered medical homes, “social capital”—the term Putnam and other social scientists use to quantify levels of social connectedness&mdash; will be a powerful concept to incorporate into their design and evaluation. The concept of social capital has been defined variably in the literature. For the purposes of its application to healthcare delivery, “social capital” refers to the collective value of bonds formed between and among individuals within social networks. These networks create a setting of trust and support in which people learn to reciprocate and perform actions that are mutually beneficial to others.

Putnam defines 2 types of social capital: bridging and bonding capital. Bridging capital is the social connectedness that results when members of unlike groups engage with one another. Bridging across occupational lines, racial and ethnic groups, and socioeconomic classes often produces an exchange of information, ideas, norms, and values that cannot be achieved if individuals restrict social contact to others like themselves.

Bonding capital, conversely, is the social connectedness that uniquely follows when individuals from within a particular group relate closely to one another. The experience and background shared by group members draws them close and promotes rewarding interaction. Examples of bonding capital might be the feelings experienced by individuals who share a common racial identity, profession, or interest.

Through the lens of this distinction, the models of care delivery that pervade medical practice are oriented more around maximizing bridging capital between patients and providers—physicians, nurses, and allied professionals&mdash; than toward generating bonding social capital among patients and their families.

Bonding social capital is an overlooked tool to improve the quality and reduce the cost of care for patients with chronic disease. The use of bonding social capital reflects the belief that levels of interpersonal trust and reciprocity norms can enhance cooperation around common medical conditions; patients who participate within a network of other patients with the same illness may be more likely to follow care recommendations and encourage others to follow if they trust the people within their network and receive support gratification from the members for their efforts.

With significant payment and medical education reform simultaneously under way, an important opportunity exists to enhance the importance of both types of social capital in care delivery and physician training models.

Physicians and the organizations in which they work may want to actively consider how they connect willing patients with one another—as supporters, teachers, and advocates for one another. Patient-centered medical homes may include patients and family members as integral members of care teams. The medical practice could become both a center for individualized care and a source for disease group management.

Diabetes management, for example, has been shown to benefit from efforts to promote social capital development among patients. Patients who participate in group visit programs over extended periods have been shown to have better glycemic control and compliance with evidence-based guidelines.

 Patients are more engaged and satisfied with their care and more likely to achieve better outcomes for themselves when involved in helping others achieve their goals. Group visit programs have been successfully implemented in many clinical settings, but have historically struggled to gain traction because third party payers have failed to reimburse them.

Enhancing social capital between patients and family members may also improve care. By focusing on the practical issues of supporting healthier lifestyles, family members and close social contacts have been shown to substantially increase the rates of healthy behaviors.

Physician practices can systematically enhance social capital by more regularly and systematically enrolling family members in the oversight of treatment and in promoting emotional support from family members. Such approaches cost little, and may encourage meaningful behavior change and improved health outcomes.

Online communication tools may be an important facilitator in promoting social capital.

 Online communities focused on specific conditions have been shown to offer rich, supportive environments where patients and their caregivers share personal stories, provide emotional support, and offer advice about up-to-date clinical care. In new care delivery models that integrate online communications into their work flow, clinicians may engage patients with common conditions within their practice to help manage one another’s conditions. Moreover, they may use linkages to other practices and clinicians to seek help for patients with more rare conditions. In these ways, the clinical practice might expand its walls to be available to help patients find support at any time of the day.

The ImproveCareNow network of pediatric gastroenterologists and patients demonstrates how online collaboration can build social capital by encouraging social connectedness to improve health outcomes. Sponsored by the American Society for Pediatric Gastroenterology, Hepatology, and Nutrition and the American Board of Pediatrics, ImproveCareNow has reported improvements in care processes such as appropriate medication dosing, as well as in clinical outcomes

As delivery reform unfolds and leads to new models of care delivery, social capital will be a powerful concept to incorporate into their design and evaluation.

June 2012

Racial Disparities in Lipid Control in Patients With Diabetes

Automating Care Quality Measurement With Health Information Technology

Role of Insurance, Income, and Affordability in Human Papillomavirus Vaccination

Thromboembolism Prophylaxis in Medical Inpatients: Effect on Outcomes and Costs

Bowling Alone, Healing Together: The Role of Social Capital in Delivery Reform

Quality Care Opportunities: Refining Physician Performance Measurement in Ambulatory Care

Unintended Consequences of a Quality Measure for Acute Bronchitis

Impact of Positive Airway Pressure Among Obstructive Sleep Apnea Patients

Barriers to Transition of Obese Patients From Hospital to Community