Search

To evaluate the effect of a predictive algorithm-driven disease management (DM) outreach program compared with non-predictive algorithm-driven DM program participation on health care spending and utilization.We used propensity score matching forMedicare Advantage members with chronic heart failure (CHF) to evaluate the impact of predictive algorithm-driven DM outreach using claims data from 2013 to 2018 from a large commercial health insurer.The insurer ran a predictive algorithm to identify members with CHF with a high likelihood of hospitalization (LOH), and a DM outreach was initiated to those identified as being at high risk of hospitalization (high-LOH intervention group). The intervention group was matched to members with similar concurrent medical risk profiles, based on the DxCG/Verisk score, who received the same DM outreach through the insurer's standard process (low-LOH intervention group). This approach allowed an evaluation of the predictive algorithm in targeting individuals suitable for DM outreach.Regression models showed that high-LOH intervention members had a lower probability of hospitalization (0.032; P = .075) and emergency department (ED) visit (0.039; P = .043) in the year after the outreach compared with low-LOH intervention members, leading to lower total outpatient spending ($1517; P .001). Analyses for no-intervention members showed that predictive outreach members would have been expected to have higher inpatient and ED utilization and higher medical spending compared with the traditional care members.A prediction-driven DM outreach program among patients with CHF was effective in reducing medical spending in the year after the outreach compared with traditional DM outreach programs.

OBJECTIVE: To measure the extent to which the provision of mammograms was impacted by the COVID‐19 pandemic and surrounding guidelines. DATA SOURCES: De‐identified summary data derived from medical claims and eligibility files were provided by Independence Blue Cross for women receiving mammograms. STUDY DESIGN: We used a difference‐in‐differences approach to characterize the change in mammograms performed over time and a queueing formula to estimate the time to clear the queue of missed mammograms. DATA COLLECTION: We used data from the first 30 weeks of each year from 2018 to 2020. PRINCIPAL FINDINGS: Over the 20 weeks following March 11, 2020, the volume of screening mammograms and diagnostic mammograms fell by 58% and 38% of expected levels, on average. Lowest volumes were observed in week 15 (April 8 to 14), when screening and diagnostic mammograms fell by 99% and 74%, respectively. Volumes began to rebound in week 19 (May), with diagnostic mammograms reaching levels to similar to previous years’ and screening mammograms remaining 14% below expectations. We estimate it will take a minimum of 22 weeks to clear the queue of missed mammograms in our study sample. CONCLUSIONS: The provision of mammograms has been significantly disrupted due to the COVID‐19 pandemic.

OBJECTIVES To determine whether implementation of patient-centered oncology standards in 5 medical oncology practices improved patient experiences and quality and reduced emergency department (ED) and hospital use. STUDY DESIGN Retrospective, pre-post study design with a concurrent nonrandomized control group. METHODS We used insurance claims to calculate all-cause hospitalizations and ED visits and primary care and specialist office visits (n = 28,826 eligible patients during baseline and 30,843 during follow-up) and identify patients for a care experiences survey (n = 715 preintervention and 437 postintervention respondents). For utilization and patient experience outcomes, we compared pilot practices' performance with 18 comparison practices using difference-in-differences (DID) regression models accounting for practice-level clustering. We assessed pilot practice performance on 31 quality measures from the American Society of Clinical Oncology Quality Oncology Practice Initiative program. RESULTS There were no statistically significant differences in hospital, ED, or primary care visits between the pilot and comparison groups over time, but there was a significant increase in specialty visits for the pilot group (adjusted DID of 0.07; 95% CI, 0.01-0.13; P = .03). For care experiences, pilot practices improved more on shared decision-making (4.03 DID composite score; P = .013), whereas the comparison group improved more on access (-6.36 DID composite score; P

BACKGROUND: Teledermatology may increase access to care but has not been widely implemented due, in part, to lack of insurance coverage and reimbursement. We assessed the impact of implementing a consultative store-and-forward teledermatology model on access to care, medical cost, and utilization. MATERIALS AND METHODS: Prospective implementation of teledermatology occurred at five University of Pennsylvania Health System primary care practices from June 27, 2016, to May 25, 2017. Primary outcomes included time to case completion, proportion of patients completing in-person dermatology visits, and total outpatient costs. Medical and pharmacy claims data were used for utilization and cost subanalysis. RESULTS: The study included 167 patients and 1,962 controls with a 6-month follow-up. Median time to definitive dermatologist response was 0.19 days (interquartile range [IQR]: 0.03–2.92) for intervention and 83.60 days (IQR: 19.74–159.73) for controls. In medical claims subanalysis, no significant differences in mean outpatient costs ($3,366 vs. $2,232, p = 0.1356) or total medical costs ($3,535 vs. $2,654, p = 0.2899) were detected. CONCLUSIONS: Implementation of teledermatology improved access to care, and within this small sample, remained comparable in terms of cost and utilization. Thus, these data suggest teledermatology may improve access without increasing utilization or cost.

This paper studies a commercial insurer-driven intervention to improve resource allocation. The insurer developed a claims-based algorithm to derive a member-level healthcare utilization risk score. Members with the highest scores were contacted by a care management team tasked with closing gaps in care. The number of members outreached was dictated by resource availability and not by severity, creating a set of arbitrary cutoff points, separating treated and untreated members with very similar predicted risk scores. Using a regression discontinuity approach, we find evidence that predictive analytics-driven interventions directed at high-risk individuals reduced emergency room and specialist visits, yet not hospitalizations.

In response to the Covid-19 pandemic, many localities instituted non-essential business closure orders, keeping individuals categorized as essential workers at the frontlines while sending their non-essential counterparts home. We examine the extent to which being designated as an essential or non-essential worker impacts one’s risk of being Covid-positive following the non-essential business closure order in Pennsylvania. We also assess the intrahousehold transmission risk experienced by their cohabiting family members and roommates. Using a difference-in-differences framework, we estimate that workers designated as essential have a 55% higher likelihood of being positive for Covid-19 than those classified as non-essential; in other words, non-essential workers experience a protective effect. While members of the health care and social assistance subsector contribute significantly to this overall effect, it is not completely driven by them. We also find evidence of intrahousehold transmission that differs in intensity by essential status. Dependents cohabiting with an essential worker have a 17% higher likelihood of being Covid-positive compared to those cohabiting with a non-essential worker. Roommates cohabiting with an essential worker experience a 38% increase in likelihood of being Covid-positive. Analysis of households with a Covid-positive member suggests that intrahousehold transmission is an important mechanism driving these effects.

Introduction Cancer patients' experience of care is an important component of quality that has not previously been used for comparing performance. We administered a new patient experience survey to cancer patients receiving outpatient chemotherapy treatment. We examined its measures for sensitivity to adjustment for case-mix and response tendency (level of general optimism/pessimism) and reliability for making performance comparisons between practices. Methods We surveyed 2304 cancer patients who received chemotherapy at 23 medical oncology practices in Southeastern Pennsylvania, receiving 715 responses (response rate 31%; 14 practices had 10 or more responses). We aggregated patient responses to calculate practice-level scores on 5 predefined composites: Affective Communication, Shared Decision-Making, Patient Self-Management, Exchanging Information, and Access. We then ranked the practices on each composite before and after adjustment for standard case-mix variables and supplemental adjustment for response tendency (measured via the Life Orientation Test-Revised). We calculated the reliability of practice scores on each composite using hierarchical linear models and calculated minimum sample sizes necessary to achieve reliabilities exceeding 0.7. Results After adjusting responses for case-mix and converting to a 0-100 scale, composite scores ranged from 77 for the Patient Self-Management composite to 92 for the Access composite. Adjustment for response tendency had an impact on practice rankings only for the Shared Decision-Making composite. The number of responses necessary to create reliable practice-level measurements ranged from 17 (Access composite) to 96 (Affective Communication composite). Conclusions Patient experiences at oncology practices can be measured reliably using reasonable sample sizes. Standard case-mix adjustment is adequate for making comparisons on most composites.

Accreditation of providers helps resolve the pervasive information asymmetries in health care markets. However, meeting accreditation standards typically involves flexibility in implementation, leading to heterogeneity in performance. For example, the patient-centered medical home (PCMH) is a leading model for recognizing high-performing primary care practices. Flexibility in PCMH implementation allows for varying degrees of emphasis on processes designed to enhance medication adherence. To assess the impact of the PCMH on adherence, we combine 6 years of detailed patient claims data with a novel dataset containing detailed practice-level PCMH attributes. We study the effects of the number and configuration of adherence-relevant capabilities, using variation in the timing of PCMH adoption to estimate its impact. While PCMH adoption improved overall medication adherence, when combining claims data with the unique recognition data detailing what PCMH capabilities were adopted, we find that these gains are concentrated among patients in practices that adopted more adherence-relevant capabilities. Despite mixed evidence in the literature concerning costs and utilization, our results indicate that PCMH recognition improves medication adherence.

Text analysis can reveal patterns of association among medical terms and medical codes.

The Patient-Centered Medical Home (PCMH) is a widely-implemented model for improving primary care, emphasizing care coordination, information technology, and process improvements. However, its treatment as an undifferentiated intervention in policy evaluation obscures meaningful variation in implementation. This heterogeneity leads to contracting inefficiencies between insurers and practices and may account for mixed evidence on its success. Using a novel dataset we group practices into meaningful implementation clusters and then link these clusters with detailed patient claims data. We find implementation choice affects performance, suggesting that generally-unobserved features of primary care reorganization influence patient outcomes. Reporting these features may be valuable to insurers and their members.

There is growing interest in using predictive analytics to drive interventions that reduce avoidable healthcare utilization. This study evaluates the impact of such an intervention utilizing claims from 2013 to 2017 for high-risk Medicare Advantage patients with congestive heart failure. A predictive algorithm using clinical and nonclinical information produced a risk score ranking for health plan members in 10 separate waves between July 2013 and May 2015. Each wave was followed by an outreach intervention. The varying capacity for outreach across waves created a set of arbitrary intervention treatment cutoff points, separating treated and untreated members with very similar predicted risk scores. We estimate a difference-in-differences model to identify the effects of the intervention program among patients with a high score on care utilization. We find that enrollment in the intervention decreased the probability and number of hospitalizations (by 43% and 50%, respectively) and emergency room visits (10% and 14%, respectively), reduced the time until a primary care visit (8.2 days), and reduced total medical cost by $716 per month in the first 6 months following outreach.

Purpose We assessed the associations between the 21-gene recurrence score assay (RS) receipt, subsequent chemotherapy use, and medical expenditures among patients with early-stage breast cancer. Patients and Methods Data from the Pennsylvania Cancer Registry were used to assemble a retrospective cohort of women with early-stage breast cancer from 2007 to 2010 who underwent initial surgical treatment. These data were merged with administrative claims from the 12-month periods before and after diagnosis to identify comorbidities, treatments, and expenditures (n = 7,287). Propensity score–weighted regression models were estimated to identify the effects of RS receipt on chemotherapy use and medical spending in the year after diagnosis. Results The associations between RS receipt and outcomes varied markedly by patient age. RS use was associated with lower chemotherapy use among women younger than 55 (19.2% lower; 95% CI, 10.6 to 27.9). RS use was associated with higher chemotherapy use among women 75 to 84 years old (5.7% higher; 95% CI, 0.4 to 11.0). RS receipt was associated with lower adjusted 1-year medical spending among women younger than 55 ($15,333 lower; 95% CI, $2,841 to $27,824) and with higher spending among women who were 75 to 84 years old ($3,489 higher; 95% CI, $857 to $6,122). Conclusion RS receipt was associated with reduced use of adjuvant chemotherapy and lower health care spending among women with breast cancer who were younger than 55. Conversely, among women 75 and older, RS testing was associated with a modest increase in chemotherapy use and slightly higher spending. From a population perspective, the impact of RS testing on breast cancer treatment and health care costs is much greater in younger women.

BACKGROUND: Targeted therapy for patients with lung and colon cancer based on tumor molecular profiles is an important cancer treatment strategy, but the impact of gene mutation tests on cancer treatment and outcomes in large populations is not clear. In this study, we assessed the accuracy of an algorithm to identify tumor mutation testing in administrative claims data during a period before test-specific Current Procedural Terminology (CPT) codes were available. METHODS: We used Pennsylvania Cancer Registry data to select patients with lung or colon cancer diagnosed between 2007 and 2011 who were treated at the University of Pennsylvania Health System, and we obtained their administrative claims. A combination of CPT laboratory codes (“stacking codes”) was used to identify potential tumor mutation testing in the claims data. Patients’ electronic medical records were then searched to determine whether tumor mutation testing actually had been performed. The sensitivity, specificity, positive and negative predictive values (PPV and NPV) were calculated. RESULTS: An algorithm using stacking codes had moderate sensitivity (86% for lung cancer and 81% for colon cancer) and high specificity (98% for lung cancer and 96% for colon cancer). Sensitivity and specificity did not vary significantly during 2007–2011. In patients with lung cancer, PPV was 98% and NPV was 92%. In patients with colon cancer, PPV was 96% and NPV was 83%. CONCLUSIONS: An algorithm using stacking codes can identify tumor mutation testing in administrative claims data among patients with lung and colon cancer with a high degree of accuracy.

Purpose: To (1) compare the effects of two worksite-based walking interventions on employee participation rates; (2) compare average daily step counts between conditions, and; (3) examine the effects of increases in average daily step counts on biometric and psychologic outcomes. Design: We conducted a cluster-randomized trial in which six employer groups were randomly selected and randomly assigned to condition. Setting: Four manufacturing worksites and two office-based worksite served as the setting. Subjects: A total of 474 employees from six employer groups were included. Intervention: A standard walking program was compared to an enhanced program that included incentives, feedback, competitive challenges, and monthly wellness workshops. Measures: Walking was measured by self-reported daily step counts. Survey measures and biometric screenings were administered at baseline and 3, 6, and 9 months after baseline. Analysis: Analysis used linear mixed models with repeated measures. Results: During 9 months, participants in the enhanced condition averaged 726 more steps per day compared with those in the standard condition (p < .001). A 1000-step increase in average daily steps was associated with significant weight loss for both men (−3.8 lbs.) and women (−2.1 lbs.), and reductions in body mass index (−0.41 men, −0.31 women). Higher step counts were also associated with improvements in mood, having more energy, and higher ratings of overall health. Conclusions: An enhanced walking program significantly increases participation rates and daily step counts, which were associated with weight loss and reductions in body mass index.

We present a new approach to population health, in which data-driven predictive models are learned for outcomes such as type 2 diabetes. Our approach enables risk assessment from readily available electronic claims data on large populations, without additional screening cost. Proposed model uncovers early and late-stage risk factors. Using administrative claims, pharmacy records, healthcare utilization, and laboratory results of 4.1 million individuals between 2005 and 2009, an initial set of 42,000 variables were derived that together describe the full health status and history of every individual. Machine learning was then used to methodically enhance predictive variable set and fit models predicting onset of type 2 diabetes in 2009-2011, 2010-2012, and 2011-2013. We compared the enhanced model with a parsimonious model consisting of known diabetes risk factors in a real-world environment, where missing values are common and prevalent. Furthermore, we analyzed novel and known risk factors emerging from the model at different age groups at different stages before the onset. Parsimonious model using 21 classic diabetes risk factors resulted in area under ROC curve (AUC) of 0.75 for diabetes prediction within a 2-year window following the baseline. The enhanced model increased the AUC to 0.80, with about 900 variables selected as predictive (p 0.0001 for differences between AUCs). Similar improvements were observed for models predicting diabetes onset 1-3 years and 2-4 years after baseline. The enhanced model improved positive predictive value by at least 50% and identified novel surrogate risk factors for type 2 diabetes, such as chronic liver disease (odds ratio [OR] 3.71), high alanine aminotransferase (OR 2.26), esophageal reflux (OR 1.85), and history of acute bronchitis (OR 1.45). Liver risk factors emerge later in the process of diabetes development compared with obesity-related factors such as hypertension and high hemoglobin A1c. In conclusion, population-level risk prediction for type 2 diabetes using readily available administrative data is feasible and has better prediction performance than classical diabetes risk prediction algorithms on very large populations with missing data. The new model enables intervention allocation at national scale quickly and accurately and recovers potentially novel risk factors at different stages before the disease onset.

The patient-centered medical home (PCMH) is a model for restructuring primary care with a focus on improved access to care and clinical excellence. However, to date, the evidence on its effect on healthcare utilization and expenditures has been quite mixed. One possible reason for this may lie in the flexibility with which a practice can meet the thresholds for PCMH recognition, adopting practice reforms in vastly different domains. Hence, practices with the same recognition level may in fact demonstrate divergent capabilities, and subsequently have different leverage over the achievement of the PCMH goals. We study this idea by using previously unavailable data that spans all PCMH recognition domains. The richness of our data allows us to group practices into clusters based on their choice of attributes during the recognition process, and then examine the performance of individual clusters in reducing healthcare utilization and expenditures. We find that treating the PCMH model as an undifferentiated intervention obscures meaningful variation in implementation across practices. In particular, clusters emphasizing practice improvements like use of decision support, enhanced access to care and population management tools have demonstrated some success in impacting utilization and expenditures patterns after PCMH recognition.

This study explores the utility of using the Integrative Model of Behavioral Prediction as a framework for predicting cancer patients' intentions to seek information about their cancer from sources other than a physician, and to examine the relation between patients' baseline intentions to seek information and their actual seeking behavior at follow-up. Within 1 year of their diagnosis with colon, breast, or prostate cancer, 1,641 patients responded to a mailed questionnaire assessing intentions to seek cancer-related information from a source other than their doctor, as well as their attitudes, perceived normative pressure, and perceived behavioral control with respect to this behavior. In addition, the survey assessed their cancer-related information seeking. One year later, 1,049 of these patients responded to a follow-up survey assessing cancer-related information seeking during the previous year. Attitudes, perceived normative pressure, and perceived behavioral control were predictive of information-seeking intentions, although attitudes emerged as the primary predictor. Intentions to seek information, perceived normative pressure regarding information seeking, baseline information-seeking behavior, and being diagnosed with stage 4 cancer were predictive of actual information-seeking behavior at follow-up. Practical implications are discussed.

The amount of cancer-related information available in the media and other sources continues to increase each year. We wondered how people make use of such content in making specific health decisions. We studied both the information they actively seek ("seeking") and that which they encounter in a less purposive way ("scanning") through a nationally representative survey of adults aged 40-70 years (n = 2,489) focused on information use around three prevention behaviors (dieting, fruit and vegetable consumption, and exercising) and three screening test behaviors (prostate-specific antigen, colonoscopy, mammogram). Overall, respondents reported a great deal of scanning and somewhat less seeking (on average 62% versus 28% for each behavior), and they used a range of sources including mass media, interpersonal conversations, and the Internet, alongside physicians. Seeking was predicted by female gender, age of 55-64 vs. 40-44, higher education, Black race and Hispanic ethnicity, and being married. Scanning was predicted by older age, female gender, and education. Respondents were fairly consistent in their place on a typology of scanning and seeking across behaviors. Seeking was associated with all six behaviors, and scanning was associated with three of six behaviors.

Research using the Implicit Association Test (IAT) has consistently shown that White participants demonstrate an implicit preference for White, race–related stimuli over similar Black stimuli. Scholars in many domains have also documented that people generally have more positive associations with the color white and more negative associations with the color black. The present research, consisting of three studies, examined the potential contribution of general implicit evaluative associations with the colors white and black to implicit race preferences as measured by the IAT. Across three studies, evaluative associations with the colors white and black were significantly related to evaluative racial associations. Nevertheless, implicit preferences for Whites relative to Blacks remained significant after controlling for the effect of implicit color preferences. Results support the position that racial IAT responses substantially reflect racial evaluative associations. Theoretical and methodologica...