Maureen O'Keeffe-Rosetti

Using risk-adjustment models to identify high-cost risks.

Background. We examine the ability of various publicly available risk models to identify high-cost individuals and enrollee groups using multi-HMO administrative data. Methods. Five risk-adjustment models (the Global Risk-Adjustment Model [GRAM], Diagnostic Cost Groups [DCGs], Adjusted Clinical Groups [ACGs], RxRisk, and Prior-expense) were estimated on a multi-HMO administrative data set of 1.5 million individual-level observations for 1995–1996. Models produced distributions of individual-level annual expense forecasts for comparison to actual values. Prespecified “high-cost” thresholds were set within each distribution. The area under the receiver operating characteristic curve (AUC) for “high-cost” prevalences of 1% and 0.5% was calculated, as was the proportion of “high-cost” dollars correctly identified. Results are based on a separate 106,000-observation validation dataset. Main Results. For “high-cost” prevalence targets of 1% and 0.5%, ACGs, DCGs, GRAM, and Prior-expense are very comparable in overall discrimination (AUCs, 0.83–0.86). Given a 0.5% prevalence target and a 0.5% prediction threshold, DCGs, GRAM, and Prior-expense captured

Excess risk of diabetes in persons with hypertension.

963,000 (approximately 3%) more “high-cost” sample dollars than other models. DCGs captured the most “high-cost” dollars among enrollees with asthma, diabetes, and depression; predictive performance among demographic groups (Medicaid members, members over 64, and children under 13) varied across models. Conclusions. Risk models can efficiently identify enrollees who are likely to generate future high costs and who could benefit from case management. The dollar value of improved prediction performance of the most accurate risk models should be meaningful to decision-makers and encourage their broader use for identifying high costs.

Diffusion of Intraperitoneal Chemotherapy in Women with Advanced Ovarian Cancer in Community Settings 2003–2008: The Effect of the NCI Clinical Recommendation

PROBLEM Persons with hypertension appear to be at increased risk of diabetes, an important predictor of cardiovascular disease. Whether, and to what extent, this risk may vary across subgroups defined on the basis of important clinical characteristics has not been well characterized. METHODS Study population included members of Kaiser Permanente Northwest Region, a large health maintenance organization, aged > or = 35 years and free of diabetes in 1998. Persons in the study population were stratified based on whether or not they had hypertension, and onset of diabetes was ascertained over a 6-year period beginning January 1999. Excess risk of diabetes was characterized in terms of risk differences between persons with and without hypertension, and was estimated on an overall basis and for subgroups defined on the basis of age, sex, and body mass index (BMI). RESULTS Study population totaled 104,368; 44% had hypertension. Relative risk (RR) of developing diabetes was 2.7 (95% CI: 2.6-2.8) for those with vs. without hypertension [21.0 (95% CI: 20.7-21.4) vs. 7.8 (95% CI: 7.6-8.0) per 1000 person-years, respectively]. Adjusted for age, sex, and BMI, RR of diabetes was 1.8 (95% CI: 1.7-1.9). With one exception (men, aged > or = 75 years), risk of diabetes was higher across all age and BMI strata for both men and women with vs. without hypertension; differences in risk were greatest among those with high BMI (> or = 35 kg/m(2)). Across BMI strata, RR of developing diabetes was generally higher at younger ages. CONCLUSION All persons with hypertension, irrespective of age, sex, and BMI, are at elevated risk of developing diabetes. Men and women with hypertension who are overweight or obese are at substantially elevated risk of diabetes, regardless of age, and should be monitored especially closely for the development of this disease.

When does an episode of care for cancer begin

Purpose: A 2006 National Cancer Institute clinical announcement recommended the use of combined intravenous (IV) and intraperitoneal (IP) chemotherapy over IV chemotherapy alone for women with International Federation of Gynecology and Obstetrics (FIGO) stage 3 optimally debulked ovarian cancer due to significant survival benefit demonstrated in multiple randomized clinical trials. We examined uptake of IP chemotherapy in community practice before and after this recommendation. Methods: We identified 288 women with FIGO stage 2 or greater incident ovarian cancer diagnosed from 2003 to 2008 at three integrated delivery systems in the US. Administrative health plan data were used to determine patient characteristics and receipt of IV and IP chemotherapy within 12 months of diagnosis. We compared characteristics of women receiving IV chemotherapy alone vs. IP chemotherapy (with or without IV chemotherapy) and assessed temporal trends in IP chemotherapy use. Results: Overall 12.5% (n = 36) of women received IP chemotherapy during the study period. IP chemotherapy use was non-existent between 2003 and 2005. Use of IP chemotherapy occurred among 26.9% of women diagnosed in 2006 and plateaued at 20.4% of women diagnosed in 2008. IP recipients were younger (mean age 55.9 vs. 63.5 years, p = < 0.001) and more likely to have stage 3 ovarian cancer (77.8 vs. 50.4% p = 0.039) compared to their IV-only chemotherapy counterparts. Conclusion: Use of IP chemotherapy for newly diagnosed advanced stage ovarian cancer patients was uncommon in this community setting. Future research should identify potential patient, physician, and system barriers and facilitators to using IP chemotherapy in this setting.

Pooling multisite administrative data for economic analysis

Background:Little is known about the medical care resources devoted to diagnosing and treating cancer-related symptoms before a definitive cancer diagnosis. Previous research using SEER-Medicare data to measure incremental costs and utilization associated with cancer started with the date of diagnosis. We hypothesized that health care use increases before diagnosis of a new primary cancer. Methods:We used a longitudinal case-control design to estimate incremental medical care utilization rates. Cases were 121,293 persons enrolled between January 2000 and December 2008 with ≥1 primary cancers. We selected 522,839 controls randomly from among all health plan members who had no tumor registry evidence of cancer before January 2009, and we frequency matched controls to cancer cases on a 5:1 ratio by age group, sex, and having health plan eligibility in the year of diagnosis of the index cancer case. Utilization data were extracted for all cases and controls for the period 2000 to 2008 from standardized distributed data warehouses. To determine when and the extent to which patterns of medical care use change preceding a cancer diagnosis, we compute hospitalization rates, hospital days, emergency department visits, same-day surgical procedures, ambulatory medical office visits, imaging procedures, laboratory tests, and ambulatory prescription dispensings per 1000 persons per month within integrated delivery systems. Results:One- to 3-fold increases in monthly utilization rates were observed during the 3 to 5 months before a cancer diagnosis, compared with matched noncancer control groups. This pattern was consistent for both aged and nonaged cancer patients. Aged cancer patients had higher utilization rates than nonaged cancer patients throughout the year before a cancer diagnosis. Conclusions:The prediagnosis phase is a resource-intensive component of cancer care episodes and should be included in cost of cancer estimates. More research is needed to determine whether reliable prognostic markers can be identified as the start of a cancer episode before a pathology-based diagnosis.

Risk-Factor Clustering and Cardiovascular Disease Risk in Hypertensive Patients*

Economic evaluations, such as cost-offset analyses, are receiving increased attention by US health insurers and payers. Administrative data collected by health maintenance organizations (HMOs) are considered an efficient source of utilization and cost measures for multisite economic analyses that increase external validity. However, pooling administrative data are problematic because HMO data sources reflect differences in systems of care, costing and coding. This paper describes issues arising from the pooling of HMO administrative cost data for use in multisite economic evaluations. We describe the attributes of administrative data that are relevant to costing and discuss their implications. We then briefly describe our experience with pooling these data, discuss lessons learned and offer suggestions for researchers working with such data.