City Research Online

Abstract

Low value health care in low-income populations: A comparison of three US state Medicaid programs

Low value care, a global problem that is particularly serious in the US, is health care that is inconsistent with evidence-based clinical guidelines, exposes patients to needless risk, can cause adverse outcomes, and unnecessarily raises the cost of health care. Using administrative claims and eligibility (ACE) datasets from Medicaid programs in three US states (the primary provider of health care to low-income populations in the US), our research found variation in each of the ten low value care measures in utilization rates and expenditures by state by Medicaid category of aid (COA), age, gender, and two geographical proxies, ethnicity and income. We also found that the five largest (defined in terms of total dollars) measures are sensitive to variations in the measurement definitions and so propose a standardized template to achieve more consistent measurement definitions.
Using four regression models, our research found weak evidence that COA, age, gender, ethnicity, and income are patient characteristics associated with the five largest low value care measures. The four types of models included 1) an ordinary least squares model with a transformed dependent variable (ln(y+1)), and a series of generalized linear models (GLM) using 2) a Tweedie distribution, 3) a compound Poisson model with part 1, a Poisson distribution and part 2, a gamma distribution, and 4) a Frequency-Severity model with part 1, a binomial distribution and part 2, a gamma distribution). Our five largest low value care measures, ‘Prescribing opioids for acute or chronic low back pain before other options (Opioids LBP)’, ‘Performing imaging for uncomplicated acute back pain in the first six weeks (LBP Imag 6 Wks)’, ‘Prescribing antibiotics for acute upper respiratory and ear infections (AntiB URI)’, ‘Performing CT scans in the evaluation of pediatric abdominal pain (Abdom Pn CT)’, and ‘Performing imaging for uncomplicated headache (H-Ache Imag’), were derived from the Choosing Wisely definitions in the US.
The results of our descriptive analyses highlight the need for Medicaid low value care analyses to incorporate actuarial principles and adjust for differences among the states in the study populations’ COA, age, gender, and geography (e.g., community income and ethnicity). Our sensitivity analyses confirm the need for consistent definitions, documentation, and calculations for low value care measures to ensure low value care research results are comparable and we propose a template to help achieve this consistency. Our regression modeling found that our limited set of demographic variables is unable to explain most of the variation in low value care expenditures. The results, using only a limited set of measures and three state Medicaid programs, are not necessarily representative of all fifty US state Medicaid programs, and further illustrate the need for additional, more granular state-specific Medicaid low value care research.

**Keywords**: health expenditures, quality measurement, value-based purchasing, low value care, Medicaid, measurement variation, patient risk characteristics

Publication Type:	Thesis (Doctoral)
Subjects:	H Social Sciences > HA Statistics H Social Sciences > HJ Public Finance R Medicine > RA Public aspects of medicine
Departments:	Bayes Business School > Actuarial Science & Insurance Bayes Business School > Bayes Business School Doctoral Theses Doctoral Theses

Export

Downloads