Currently Viewing:
The American Journal of Managed Care September 2018
Food Insecurity, Healthcare Utilization, and High Cost: A Longitudinal Cohort Study
Seth A. Berkowitz, MD, MPH; Hilary K. Seligman, MD, MAS; James B. Meigs, MD, MPH; and Sanjay Basu, MD, PhD
Language Barriers and LDL-C/SBP Control Among Latinos With Diabetes
Alicia Fernandez, MD; E. Margaret Warton, MPH; Dean Schillinger, MD; Howard H. Moffet, MPH; Jenna Kruger, MPH; Nancy Adler, PhD; and Andrew J. Karter, PhD
Hepatitis C Care Cascade Among Persons Born 1945-1965: 3 Medical Centers
Joanne E. Brady, PhD; Claudia Vellozzi, MD, MPH; Susan Hariri, PhD; Danielle L. Kruger, BA; David R. Nerenz, PhD; Kimberly Ann Brown, MD; Alex D. Federman, MD, MPH; Katherine Krauskopf, MD, MPH; Natalie Kil, MPH; Omar I. Massoud, MD; Jenni M. Wise, RN, MSN; Toni Ann Seay, MPH, MA; Bryce D. Smith, PhD; Anthony K. Yartel, MPH; and David B. Rein, PhD
“Precision Health” for High-Need, High-Cost Patients
Dhruv Khullar, MD, MPP, and Rainu Kaushal, MD, MPH
From the Editorial Board: A. Mark Fendrick, MD
A. Mark Fendrick, MD
Health Literacy, Preventive Health Screening, and Medication Adherence Behaviors of Older African Americans at a PCMH
Anil N.F. Aranha, PhD, and Pragnesh J. Patel, MD
Early Experiences With the Acute Community Care Program in Eastern Massachusetts
Lisa I. Iezzoni, MD, MSc; Amy J. Wint, MSc; W. Scott Cluett III; Toyin Ajayi, MD, MPhil; Matthew Goudreau, BS; Bonnie B. Blanchfield, CPA, SM, ScD; Joseph Palmisano, MA, MPH; and Yorghos Tripodis, PhD
Economic Evaluation of Patient-Centered Care Among Long-Term Cancer Survivors
JaeJin An, BPharm, PhD, and Adrian Lau, PharmD
Currently Reading
Fragmented Ambulatory Care and Subsequent Healthcare Utilization Among Medicare Beneficiaries
Lisa M. Kern, MD, MPH; Joanna K. Seirup, MPH; Mangala Rajan, MBA; Rachel Jawahar, PhD, MPH; and Susan S. Stuard, MBA
Adjusting Medicare Advantage Star Ratings for Socioeconomic Status and Disability
Melony E. Sorbero, PhD, MS, MPH; Susan M. Paddock, PhD; Cheryl L. Damberg, PhD; Ann Haas, MS, MPH; Mallika Kommareddi, MPH; Anagha Tolpadi, MS; Megan Mathews, MA; and Marc N. Elliott, PhD

Fragmented Ambulatory Care and Subsequent Healthcare Utilization Among Medicare Beneficiaries

Lisa M. Kern, MD, MPH; Joanna K. Seirup, MPH; Mangala Rajan, MBA; Rachel Jawahar, PhD, MPH; and Susan S. Stuard, MBA
Among Medicare beneficiaries, the relationship between fragmented ambulatory care and subsequent emergency department visits and hospital admissions varies with the number of chronic conditions.
Check out our website’s new table/figure pop-up feature! Click on the name of a table or figure in the text to see it in your browser.
Statistical Analysis

Independent variable. Our base-case analysis measured fragmentation with the Bice-Boxerman Index (BBI) (eAppendix A [eAppendices available at]),15,16 a previously validated measure.3,4,6,8,16 We reversed raw BBI scores, so that higher scores would reflect more fragmentation. Patterns of care that reflect high dispersion (many providers) and low density (a relatively low proportion of ambulatory visits by each provider) yield worse (higher) scores. Because the distribution of BBI scores is inherently skewed, we divided scores into quintiles, an approach we successfully used previously,3 to maximize clarity of interpretation. We conducted sensitivity analyses with 2 other fragmentation indices, the Herfindahl-Hirschman Index and the Usual Provider Continuity Index (eAppendix A).16

Dependent variables. We identified ED visits and hospital admissions in the claims, using definitions from NCQA.14 An “ED visit” resulted in discharge to home or elsewhere. If an ED visit resulted in hospital admission, it was considered part of that admission and counted only as an admission.

Potential confounders. We used ICD-9 codes to calculate the number of chronic conditions for each beneficiary (0, 1-2, 3-4, or ≥5)3 of 26 unique chronic conditions defined by CMS (eAppendix B).17 We also calculated a severity of illness index.18,19 We considered beneficiary age and gender as potential confounders. In addition, we considered the number of ambulatory visits as a potential confounder, because the number of ambulatory visits was weakly correlated with fragmentation score (Spearman correlation coefficient, 0.22; P <.0001).

Descriptive statistics. We characterized beneficiaries in terms of age, gender, and number of chronic conditions. We compared the characteristics of the study sample with those who were excluded using pairwise Wilcoxon rank-sum tests for continuous variables and χ2 tests for categorical variables.

For the study sample, we calculated descriptive statistics regarding the number of ambulatory visits, number of unique providers, and proportion of ambulatory visits with the most frequently seen provider in the baseline year (overall and stratified by fragmentation quintile). We also determined the proportions of beneficiaries who had 1 or more ED visits and, separately, 1 or more hospital admissions during follow-up.

Statistical models. Because fragmentation can change over time and because the hypothesized consequences of fragmented care may unfold relatively quickly, we used Cox models and treated fragmentation as time-dependent. That is, we first calculated fragmentation in the first 12 months (calendar year 2010) and determined whether the beneficiary had an ED visit in month 13 (January 2011) or not. We then moved this window of observation by 1 month, recalculating the model using fragmentation in months 2 to 13 as a potential predictor of an ED visit in month 14, and so on. If the number of ambulatory visits in any 12-month window fell below 4 (making it difficult to calculate fragmentation scores), we carried forward the last fragmentation score that was based on 4 or more ambulatory visits. We used the same approach in separate models predicting hospital admissions.

For each model, observation continued until an outcome or censoring occurred. For the ED visit models, censoring occurred (1) if a beneficiary was admitted to the hospital (with the reasoning that the beneficiary was not at risk of an ED visit) or (2) at the end of the beneficiary’s continuous enrollment. For the hospital admission models, censoring occurred only at the end of the beneficiary’s continuous enrollment; any ED visit had no effect on the hospital models, because the beneficiary was still at risk of a hospital admission.

We adjusted for beneficiary gender and for the following time-varying covariates: age, Charlson-Deyo score, and number of ambulatory visits. By adjusting for number of ambulatory visits, we sought to fully tease apart fragmentation (the diffuseness of care) from the volume of ambulatory visits, as we have done previously.3 We calculated an interaction term for fragmentation category*chronic condition count and then, because this term was statistically significant for at least 1 model, we stratified our analyses by chronic condition count.

To test the appropriateness of the statistical assumptions of our models, we generated weighted plots of Schoenfeld residuals by time and calculated zph tests for nonproportional hazards (which is recommended in situations with time-dependent predictors and/or covariates).20-22

We considered P values <.05 to be statistically significant. We used SAS version 9.4 (SAS Institute; Cary, North Carolina).

Copyright AJMC 2006-2020 Clinical Care Targeted Communications Group, LLC. All Rights Reserved.
Welcome the the new and improved, the premier managed market network. Tell us about yourself so that we can serve you better.
Sign Up