Abstract
Importance
National patterns of low-value and high-value care delivered to patients without insurance or with Medicaid could inform public policy but have not been previously examined.
Objective
To measure rates of low-value care and high-value care received by patients without insurance or with Medicaid, compared with privately insured patients, and provided by safety-net physicians vs non–safety-net physicians.
Design, Setting, and Participants
This multiyear cross-sectional observational study included all patients ages 18 to 64 years from the National Ambulatory Medical Care Survey (2005-2013) and the National Hospital Ambulatory Medical Care Survey (2005-2011) eligible for any of the 21 previously defined low-value or high-value care measures. All measures were analyzed with multivariable logistic regression and adjusted for patient and physician characteristics.
Exposures
Comparison of patients by insurance status (uninsured/Medicaid vs privately insured) and safety-net physicians (seeing >25% uninsured/Medicaid patients) vs non–safety-net physicians (seeing 1%-10%).
Main Outcomes and Measures
Delivery of 9 low-value or 12 high-value care measures, based on previous research definitions, and composite measures for any high-value or low-value care delivery during an office visit.
Results
Overall, 193 062 office visits were eligible for at least 1 measure. Mean (95% CI) age for privately insured patients (n = 94 707) was 44.7 (44.5-44.9) years; patients on Medicaid (n = 45 123), 39.8 (39.3-40.3) years; and uninsured patients (n = 19 530), 41.9 (41.5-42.4) years. Overall, low-value and high-value care was delivered in 19.4% (95% CI, 18.5%-20.2%) and 33.4% (95% CI, 32.4%-34.3%) of eligible encounters, respectively. Rates of low-value and high-value care delivery were similar across insurance types for the majority of services examined. Among Medicaid patients, adjusted rates of use were no different for 6 of 9 low-value and 9 of 12 high-value services compared with privately insured beneficiaries, whereas among the uninsured, rates were no different for 7 of 9 low-value and 9 of 12 high-value services. Safety-net physicians provided similar care compared with non–safety-net physicians, with no difference for 8 out of 9 low-value and for all 12 high-value services.
Conclusions and Relevance
Overuse of low-value care is common among patients without insurance or with Medicaid. Rates of low-value and high-value care were similar among physicians serving vulnerable patients and other physicians. Overuse of low-value care is a potentially important focus for state Medicaid programs and safety-net institutions to pursue cost savings and improved quality of health care delivery.
This cross-sectional study measure rates of low-value and high-value care received by patients without insurance or with Medicaid compared with privately insured patients, and provided by safety-net vs non–safety-net physicians.
Key Points
Question
How often do patients with Medicaid or without insurance receive low-value care compared with privately insured patients; are there any differences related to the physicians treating these patient groups?
Findings
Analyses of nationally representative survey data from 2005 to 2013 show that low-value care was delivered in nearly 1 in 5 visits, with no overall difference between Medicaid or uninsured patients vs privately insured patients. Rates of low-value care were similar between safety-net physicians and non–safety-net physicians.
Meaning
Overuse of low-value care is just as common among patients with Medicaid or without insurance as among privately insured patients.
Introduction
Wasteful spending on low-value medical services is an important source of unnecessary health care costs. Recently, there has been increasing scrutiny of low-value care, in part due to the American Board of Internal Medicine (ABIM) Foundation’s Choosing Wisely campaign. This campaign, combined with other professional guidelines, has shaped the definition of low-value services, which are evolving as quality measures. While eliminating low-value care is a goal for all physicians, it may be particularly important for physicians in settings with limited financial resources.
Prior research has documented the use of several low-value services in Medicare nationally, as well as among privately and publicly insured patients in particular states. However, that we know of, no research to date has examined the national prevalence of low-value care delivery among vulnerable patients in the health care safety net, such as those without insurance or with Medicaid. It is unclear whether patterns of care for these services differ by patient insurance status or by the type of physician delivering care. One possibility is that overuse of low-value care for uninsured patients and those with Medicaid could be driven by the types of physicians treating them, since care for these patients is disproportionately concentrated among a small percentage of physicians. Alternatively, patient insurance status may be a more important predictor of low-value care. The extent to which low-value care is common in the treatment of uninsured and Medicaid patients has important implications for whether state Medicaid programs and safety-net physicians should focus quality improvement efforts on reducing rates of low-value care.
Another important policy question is the relative magnitude of the problems of overuse of low-value care vs underuse of high-value care. It is possible that patients without insurance or with Medicaid may not receive a significant amount of low-value care but still experience lower quality of care due to underuse of high-value care. Therefore, both low-value and high-value care delivery need to be included in an assessment of quality of care in Medicaid.
Our objective was to compare patterns of high-value and low-value care among uninsured or Medicaid patients vs privately insured patients, and between physicians serving high vs low proportions of these vulnerable patients using nationally representative survey data.
Methods
Study Population and Data
We used data from the National Ambulatory Medical Care Surveys (NAMCS) and the outpatient department National Hospital Ambulatory Medical Care Surveys (NHAMCS) administered by the National Center for Health Statistics at the Centers for Disease Control and Prevention. We used data from 2005 through the most recently available year (2011 in NHAMCS and 2013 in NAMCS). The NAMCS and NHAMCS are annual, multistage probability surveys that are nationally representative of ambulatory physician visits in the United States. The NAMCS surveys visits to nonfederally employed office–based physicians’ practices, and NHAMCS surveys visits to hospital–based outpatient departments (it also samples emergency departments, which were excluded from our analysis). We included visits from patients ages 18 to 64 years. Our analysis used publicly available data and was deemed nonhuman subjects research by the institutional review board at the Harvard T. H. Chan School of Public Health.
Study Variables
Insurance Coverage and Safety Net Physician Categorization
The 2 main exposures of interest were a patient’s insurance at the time of an outpatient visit and the proportion of vulnerable patients (defined here as uninsured or having Medicaid) seen by a physician. The survey collects information on source of payment for each visit, generally filled out by physician office staff. Since some patients have multiple types of coverage, we assigned a primary insurance type to each visit based on the following hierarchy: Medicaid, Medicare, private insurance, other insurance, self-pay or charity and/or no charge (which we combined into “uninsured”), or unknown. Patients with both Medicaid and Medicare coverage were coded as having Medicaid coverage. We chose this approach because dual-eligible patients are an important, high-risk segment of the Medicaid population. Our main comparisons of interest were rates of low-value and high-value care delivery to Medicaid or uninsured patients vs those with private insurance coverage.
Then, we defined physicians as “safety-net” and “non–safety-net” physicians based on the survey-weighted proportion of visits to each physician by patients with Medicaid or no insurance; this operationalizes the Institute of Medicine’s definition of safety-net physicians as those who care for a substantial share of patients without insurance or with Medicaid. We categorized physicians into 4 roughly equal-sized categories by the proportion of patients without insurance or with Medicaid: 0%, 1% to 10%, 11% to 25%, and greater than 25%. Our study analyzed the 1% to 10% group vs the greater than 25% group as the primary comparison of interest, because comparing physicians with 0% precludes any assessment of insurance type (a problem also referred to as “complete lack of overlap”). For simplicity, we refer from here to the 1% to 10% group as “non–safety-net” and the greater 25% as “safety-net” physicians (results from the 0% group are available in the eAppendix in the Supplement).
Low-Value and High-Value Care Measures
We used 21 previously defined low-value care or guideline-recommended high-value care measures (detailed definitions are available in the eAppendix in the Supplement). In general, visits were eligible for a category when a patient had a reason for visit, physician diagnosis, or chronic illness comorbidity coded relevant for a given quality measure. This means that some visits for patients with chronic illness were considered eligible, even if the visit was not for that illness. For low-value care, we identified 9 measures defined in published guidelines and prior literature within the NAMCS and/or NHAMCS surveys in 4 clinical scenarios: antibiotics for upper respiratory infections (URI); advanced imaging for sinusitis; nonindicated screening tests in general medical examinations (GME); and inappropriate imaging or narcotics in acute low back pain and uncomplicated headache (Table 1) (eAppendix in the Supplement). For high-value care measures, we selected 12 measures endorsed by quality measurement guidelines or the US Preventive Services Task Force and previously analyzed using NAMCS and/or NHAMCS data (Table 1) (eAppendix in the Supplement). These included tobacco cessation and weight loss counseling; appropriate medical treatment for coronary artery disease (CAD), cerebrovascular disease (CVD), congestive heart failure (CHF), depression, and osteoporosis; anticoagulation for atrial fibrillation, and statin use for diabetes.
Table 1. Low-Value and High-Value Quality Measures by Insurance Typea.
| Quality Measure | Office Visits, No. | % (95% CI) | ||
|---|---|---|---|---|
| Private Insurance | Medicaid | Uninsured | ||
| Low-value care measures | ||||
| Overall low-value compositeb | 55 554 | 19.1 (18.1-20.2) | 20.6 (19.0-22.2) | 20.8 (18.4-23.2) |
| Antibiotics for never URIs | 4698 | 55.1 (51.9-58.2) | 49.4 (41.8-57.0) | 72.7 (66.5-79.0) |
| CT scan for sinusitis | 1774 | 1.3 (0.59-2.0) | 0.7 (0.0-2.0) | 1.0 (0.0-2.5) |
| Screening EKG in GME | 18 169 | 6.3 (4.8-7.8) | 5.4 (1.7-9.1) | 3.9 (0.46-7.3) |
| Screening CBC in GME | 22 272 | 26.8 (24.3-29.3) | 26.7 (23.0-30.4) | 20.8 (16.3-25.3) |
| Screening UA in GME | 21 952 | 20.1 (18.1-22.1) | 20.2 (16.2-24.3) | 17.6 (12.7-22.6) |
| Narcotics for back/neck pain | 20 283 | 28.4 (26.3-30.5) | 37.5 (33.5-41.5) | 43.9 (36.5-51.3) |
| CT/MRI for back/neck pain | 20 283 | 9.7 (8.6-10.8) | 7.6 (6.2-9.1) | 5.0 (3.4-6.6) |
| Narcotics for headache | 6912 | 10.5 (8.7-12.3) | 20.2 (15.2-25.2) | 15.6 (9.4-21.8) |
| CT/MRI for headache | 6912 | 6.3 (4.7-8.0) | 7.2 (4.3-10.1) | 4.1 (1.3-6.8) |
| High-value care measures | ||||
| Overall high-value compositeb | 164 209 | 34.1 (33.1-35.1) | 31.5 (29.7-33.4) | 40.5 (37.8-43.2) |
| Tobacco cessation counseling | 51 789 | 19.1 (17.7-20.4) | 21.3 (18.8-23.7) | 21.3 (17.9-24.7) |
| Weight reduction counseling | 68 448 | 28.4 (26.9-29.9) | 28.0 (25.2-30.9) | 38.5 (30.9-46.0) |
| Anticoagulant use in Afib | 1169 | 70.3 (64.1-76.5) | 63.4 (51.0-75.9) | 78.4 (61.2-95.7) |
| Aspirin use in CAD | 7230 | 45.7 (42.7-48.7) | 37.8 (32.6-42.9) | 39.0 (29.6-48.4) |
| β-Blocker use in CAD | 6257 | 43.8 (41.0-46.7) | 33.0 (28.0-38.1) | 43.1 (33.3-53.0) |
| Statin use in CAD | 7168 | 46.4 (43.7-49.2) | 36.4 (31.3-41.4) | 41.7 (30.8-52.5) |
| β-Blocker use in CHF | 3343 | 44.8 (39.9-49.8) | 34.7 (26.7-42.6) | 53.0 (40.0-65.9) |
| ACE/ARB use in CHF | 4252 | 28.6 (23.9-33.4) | 30.8 (24.9-36.7) | 40.0 (28.7-51.3) |
| Antiplatelet use in CVD | 4341 | 27.3 (23.6-30.9) | 24.2 (18.9-29.4) | 28.7 (14.5-42.8) |
| Statin use in DM | 37 589 | 28.6 (27.1-30.2) | 22.7 (20.8-24.7) | 25.3 (21.8-28.9) |
| Treatment for depression | 49 014 | 59.8 (58.1-61.4) | 54.6 (51.6-57.7) | 71.1 (67.9-74.3) |
| Treatment for osteoporosis | 4877 | 23.9 (21.2-26.6) | 17.9 (12.5-23.3) | 21.2 (10.9-31.4) |
Abbreviations: ACE, angiotensin-converting enzyme inhibitor; AF, atrial fibrillation; ARB, angiotensin receptor blocker; CAD, coronary artery disease; CBC, complete blood count; CHF, congestive heart failure; CT, computed tomography; CVD, cerebrovascular disease; DM, diabetes mellitus; ECG, electrocardiogram; GME, general medical examination; MRI, magnetic resonance imaging; UA, urinalysis; URI, upper respiratory tract infection.
All estimates (except unweighted) reflect survey-weighted proportions accounting for the complex sample design of National Ambulatory Medical Care Survey.
Low-value and high-value care composites calculated at the visit level as the proportion of low-value or high-value services delivered at a visit. To account for visits qualifying for more services than others, survey weights were multiplied by the number of eligible low-value or high-value measures.
Statistical Analysis
For each outcome, we calculated an unadjusted survey-weighted rate stratified by patient insurance coverage or physician safety net status. The denominator for each measure was the number of visits that patients might be eligible to receive a service, accounting for the inclusion and exclusion criteria in eTable 1 in the Supplement. Then, for each numerator, we identified visits in which the low-value or high-value services was delivered and calculated each measure’s rate of delivery (0%-100%).
We also generated 2 composite measures to quantify the delivery of any low-value or high-value service. To estimate these measures, we calculated for each visit the proportion of high-value or low-value services delivered as the number of services delivered (numerator) divided by the number of services that visit was eligible for (denominator). Because visits varied in the number of services eligible for the composite, for these outcomes we multiplied visit weights by the number of eligible low-value or high-value services (eAppendix in the Supplement). Therefore, the rate for each composite measure represents the rate at which low-value or high-value care was given among all opportunities for a measure.
To examine the independent association of health insurance or safety-net physicians with low-value or high-value care, we estimated logistic regression models for each of the 21 low-value and high-value care measures, including only the visits eligible for that measure. For the 2 composite measures, the outcome was the proportion of eligible low-value or high-value services delivered (using fractional logistic regression), weighted by the product of the survey weight and the number of eligible visits. The primary variable of interest in each model was either an indicator for Medicaid or uninsured patients (vs patients with private coverage), or an indicator for being seen by a safety-net physician.
All models are adjusted for year fixed effects, patient age, sex, race/ethnicity, number of chronic illnesses, census region, rural office location, and practice setting (solo practice, group practice, or outpatient hospital department). Models are simultaneously adjusted for patient insurance status and safety net physician status.
In all analyses, we used robust design–based variance estimators to account for clustering within geographic areas or physicians and NAMCS and/or NHAMCS survey weights to account for the survey design and survey nonresponse. Analyses were performed with R version 3.3.0 (R Foundation). We report 95% CIs for each estimate, reflecting a 2-sided significant threshold of P < .05.
Sensitivity Analyses
We tested for changes over time in rates of the composite low-value and high-value measures from 2005 to 2011, the years for which NAMCS and NHAMCS data were both available (eAppendix in the Supplement). We also performed additional sensitivity analyses to examine the robustness of our findings to different definitions of insurance coverage (eg, excluding dual-eligible Medicaid-Medicare patients from Medicaid group), excluding the narcotic low-value care measures, and limiting our analysis to care delivered solely by primary care physicians (eAppendix in the Supplement).
Results
Our sample included 193 062 sampled office visits from 2005 to 2013 that were eligible for at least 1 low-value or high-value service. Compared with those with private insurance, Medicaid and uninsured patients were significantly younger, less likely to be white, and more likely to be seen in outpatient hospital departments (Table 2). Stratifying visits by physician safety-net status showed similar patterns of demographic features (eTable 2 in the Supplement). Even though safety-net physicians accounted for only 26% of the total visits in our study sample, they delivered 73% of all visits for Medicaid or uninsured patients.
Table 2. Office Visit Characteristics by Patient Insurance Type, 2005-2012a.
| Patient Characteristic | Weighted % (95% CI) | ||
|---|---|---|---|
| Private | Medicaid | Uninsured | |
| Unweighted, No. | 94 707 | 45 123 | 19 530 |
| Weighted, No. (millions) | 1888 | 414 | 216 |
| Age, mean, y | 44.7 (44.5-44.9) | 39.8 (39.3-40.3) | 41.9 (41.5-42.4) |
| Female | 61.7 (60.9-62.6) | 70.1 (68.6-71.6) | 60.0 (57.8-62.1) |
| Race/ethnicity | |||
| Hispanic | 8.9 (7.9-9.9) | 18.1 (15.4-20.8) | 14.9 (12.4-17.3) |
| Non-Hispanic black | 9.9 (8.7-11.2) | 21.7 (18.6-24.9) | 14.9 (11.4-18.3) |
| Non-Hispanic white | 77.1 (75.8-78.4) | 56.3 (52.5-60.0) | 66.7 (63.4-69.9) |
| Other | 4.0 (3.4-4.6) | 3.9 (3.1-4.7) | 3.6 (2.8-4.4) |
| Chronic illnesses, No.b | |||
| 0 | 30.6 (29.4-31.7) | 25.7 (24.0-27.3) | 31.7 (29.3-34.1) |
| 1 | 30.5 (29.9-31.1) | 28.8 (27.8-29.9) | 37.6 (35.3-39.8) |
| 2-3 | 30.1 (29.2-30.9) | 32.0 (30.8-33.3) | 24.3 (22.5-26.1) |
| ≥4 | 7.2 (6.7-7.7) | 11.9 (10.9-12.9) | 4.9 (4.0-5.7) |
| Missing | 1.7 (1.5-1.9) | 1.6 (1.3-1.9) | 1.6 (1.1-2.0) |
| US Region | |||
| Northeast | 23.1 (19.9-26.3) | 23.5 (20.5-26.6) | 17.0 (14.1-19.8) |
| Midwest | 20.0 (17.8-22.2) | 23.7 (20.3-27.1) | 18.1 (14.7-21.6) |
| South | 38.0 (34.3-41.6) | 34.4 (30.6-38.1) | 47.5 (42.7-52.4) |
| West | 19.0 (16.4-21.6) | 18.4 (15.4-21.4) | 17.4 (14.4-20.4) |
| Rural vs urban MSA | 10.9 (6.4-15.4) | 18.7 (10.4-26.9) | 12.3 (6.5-18.0) |
| Practice setting | |||
| Group practice | 63.8 (61.8-65.8) | 48.1 (45.0-51.1) | 40.3 (36.6-44.0) |
| Outpatient hospital | 5.3 (4.4-6.2) | 18.9 (16.6-21.2) | 15.2 (10.9-19.5) |
| Solo practice | 30.9 (29.1-32.7) | 33.0 (29.9-36.1) | 44.5 (40.3-48.6) |
Abbreviation: MSA, metropolitan statistical area.
All estimates (except unweighted) reflect survey-weighted proportions accounting for the complex sample design of National Ambulatory Medical Care Survey. Results for visits with Medicare or other insurance not shown (n = 33 702).
Chronic illnesses include 14 indicators for arthritis, asthma, any cancer, cerebrovascular disease, chronic obstructive pulmonary disease, chronic kidney disease, congestive heart failure, depression, diabetes mellitus, hyperlipidemia, hypertension, ischemic heart disease, obesity, and osteoporosis.
Overall, low-value and high-value care was delivered in 19.4% (95% CI 18.5%-20.2%) and 33.4% (95% CI, 32.4%-34.3%) of eligible opportunities for all patients in our sample, respectively.
Out of 9 low-value services, 7 were not significantly different, and 2 inappropriate prescribing measures (narcotics for back/neck pain and headache) were higher for Medicaid than for privately insured patients (Table 1). Likewise, out of 12 high-value services, 9 were not significantly different for Medicaid than for privately insured patients (Table 1). Combining the measures in a composite outcome, Medicaid patients had similar rates of low-value care compared with the privately insured (20.6% vs 19.1% unadjusted rates, respectively; adjusted odds ratio [aOR], 1.11; 95% CI, 0.99-1.25) (Table 1; Figure 1A). In contrast, overall receipt of pooled high-value services was lower among Medicaid patients than among the privately insured (34.1% vs 31.5% unadjusted rates, respectively; aOR, 0.88; 95% CI, 0.83-0.94) (Table 1; Figure 1A).
Figure 1. Adjusted Odds Ratios for Low-Value and High-Value Care by Insurance Type.
A, Adjusted odds ratios (ORs) comparing rates of low-value or high-value services delivered to Medicaid vs privately insured patients; B, uninsured patients vs privately insured patients, including composite measures for the delivery of any low-value or high-value care service. Adjusted ORs and 95% CIs are estimated from survey-weighted logistic regression models clustering for the complex sample design of National Ambulatory Medical Care Survey. All estimates are adjusted for the characteristics listed in Table 1. For the purposes of plotting, the upper CI for 2 ORs (CT scan for sinusitis and anticoagulant use in AFib) are truncated. AFib indicates atrial fibrillation; ACE, angiotensin-converting enzyme inhibitor; ARB, angiotensin receptor blocker; CAD, coronary artery disease; CBC, complete blood cell count; CHF, congestive heart failure; CT, computed tomography; CVD, cerebrovascular disease; ECG, electrocardiogram; GME, general medical examination; MRI, magnetic resonance imaging; URI, upper respiratory tract infection.
Among patients without insurance, out of 9 low-value care measures, 6 were not significantly different compared with privately insured patients, while 2 inappropriate prescribing measures were more likely (antibiotics for URI and narcotics for back/neck pain), and 1 was less likely (computed tomography/magnetic resonance imaging for back/neck pain) (Table 1). Similarly, there was no difference between the uninsured and privately insured for 9 of 12 high-value care measures. Pooling across all measures, there was no overall difference in the low-value care composite among uninsured vs privately insured patients (20.8 vs 19.1%; aOR, 1.08; 95% CI, 0.91-1.27) (Table 1; Figure 1B), while the uninsured were more likely to receive any high-value care services compared with the privately insured (aOR, 1.44; 95% CI, 1.28-1.61) (Figure 1B).
There were no significant differences in the delivery of any low-value services between safety-net and non–safety-net physicians (19.9% vs 19.6% in unadjusted estimates; aOR, 0.98; 95% CI, 0.85-1.14) (Table 3; Figure 2) (eTable 3 in the Supplement). There was also no difference between the physician groups for 8 of the 9 individual low-value services. Similarly, there was no difference in high-value composite measure for safety-net vs non–safety-net physicians (34.1% vs 34.4%; aOR, 0.97; 95% CI, 0.87-1.08) (Table 3; Figure 2).
Table 3. Low-Value and High-Value Quality Measures by Physician Safety-Net Proportiona.
| Quality Measure | No. | % (95% CI) | |
|---|---|---|---|
| Non–Safety-Net Physicians (1%-10%) | Safety-Net Physicians (>25%) | ||
| Low-value care measures | |||
| Overall low-value compositeb | 55 554 | 19.6 (17.8-21.3) | 19.9 (18.3-21.5) |
| Antibiotics for “never” URIs | 4698 | 57.1 (51.5-62.7) | 55.6 (49.7-61.5) |
| CT scan for sinusitis | 1774 | 1.4 (0.2-2.5) | 0.2 (0.0-0.4) |
| Screening EKG in GME | 18 169 | 7.4 (4.6-10.1) | 4.7 (1.6-7.8) |
| Screening CBC in GME | 22 272 | 26.2 (22.4-29.9) | 24.2 (20.1-28.4) |
| Screening UA in GME | 21 952 | 19.1 (15.4-22.9) | 19.7 (15.9-23.5) |
| Narcotics for back/neck pain | 20 283 | 30.7 (27.3-34.1) | 37.9 (34.2-41.7) |
| CT/MRI for back/neck pain | 20 283 | 11.5 (9.0-14.0) | 7.1 (5.8-8.4) |
| Narcotics for headache | 6912 | 11.8 (8.2-15.4) | 14.6 (11.9-17.3) |
| CT/MRI for headache | 6912 | 6.4 (4.2-8.5) | 5.9 (3.9-7.9) |
| High-value care measures | |||
| Overall high-value compositeb | 164 209 | 34.1 (32.5-35.8) | 34.4 (32.4-36.4) |
| Tobacco cessation counseling | 51 789 | 19.6 (17.2-22.0) | 20.9 (18.5-23.3) |
| Weight reduction counseling | 68 448 | 28.4 (25.5-31.3) | 30.1 (26.8-33.3) |
| Anticoagulant use in Afib | 1169 | 71.2 (59.9-82.5) | 68.1 (57.6-78.6) |
| Aspirin use in CAD | 7230 | 42.5 (37.0-47.9) | 39.7 (35.3-44.0) |
| β-Blocker use in CAD | 6257 | 44.8 (39.1-50.6) | 39.1 (34.7-43.5) |
| Statin use in CAD | 7168 | 46 (41.0-51.1) | 39.4 (35.5-43.2) |
| β-Blocker use in CHF | 3343 | 43.3 (34.2-52.5) | 42.2 (36.0-48.4) |
| ACE/ARB use in CHF | 4252 | 27.8 (20.5-35.1) | 34.8 (28.4-41.1) |
| Antiplatelet use in CVD | 4341 | 30.2 (23.5-36.8) | 25.8 (20.8-30.7) |
| Statin use in diabetes | 37 589 | 28.1 (25.2-30.9) | 24.4 (22.3-26.5) |
| Treatment for depression | 49 014 | 60.6 (57.8-63.3) | 61.2 (58.5-63.9) |
| Treatment for osteoporosis | 4877 | 25.0 (20.1-29.8) | 18.7 (14.7-22.7) |
Abbreviations: ACE, angiotensin-converting enzyme inhibitor; AF, atrial fibrillation; ARB, angiotensin receptor blocker; CAD, coronary artery disease; CBC, complete blood cell count; CHF, congestive heart failure; CT, computed tomography; CVD, cerebrovascular disease; ECG, electrocardiogram; GME, general medical examination; MRI, magnetic resonance imaging; UA, urinalysis; URI, upper respiratory tract infection.
All estimates (except unweighted) reflect survey-weighted proportions accounting for the complex sample design of NAMCS. Physicians categorized according to the proportion of safety-net patients seen, defined as those with Medicaid insurance or without health insurance coverage (self-pay or charity care).
Low-value and high-value care composites calculated at the visit level as the proportion of low-value or high-value services delivered at a visit. To account for visits qualifying for more services than others, survey weights were multiplied by the number of low-value or high-value services eligible. This weight was used for the logistic regression adjusting for all patient factors in Figure 2 to estimate the adjusted odds ratios.
Figure 2. Adjusted Odds Ratios for Low-Value and High-Value Care for Safety-Net vs Non–Safety-Net Physicians.
A, Adjusted odds ratios (ORs) comparing rates of low-value or high-value services delivered by physicians serving a high proportion of safety-net patients (defined as >25% of visits with Medicaid or uninsured patients) vs those serving a low proportion of safety-net patients (0%-10% of visits with Medicaid or uninsured patients), including composite measures for the delivery of any low-value or high-value care service. Adjusted ORs and 95% CIs are estimated from survey-weighted logistic regression models clustering for the complex sample design of National Ambulatory Medical Care Survey. All estimates are adjusted for the characteristics listed in Table 1. For the purposes of plotting, the upper CI for 1 OR is truncated (CT scan for sinusitis). AFib indicates atrial fibrillation; ACE, angiotensin-converting enzyme inhibitor; ARB, angiotensin receptor blocker; CAD, coronary artery disease; CBC, complete blood cell count; CHF, congestive heart failure; CT, computed tomography; CVD, cerebrovascular disease; ECG, electrocardiogram; GME, general medical examination; MRI, magnetic resonance imaging; URI, upper respiratory tract infection.
Sensitivity Analyses
We found no significant changes over time in the composite low-value or high-value measures for any subgroup, except for a modest decrease in high-value care composite delivery in the privately insured group (OR, 0.99 per year; P = .01) (eFigure 1 and 2 in the Supplement). Our main findings were essentially unchanged under alternative insurance coverage definitions (eTable 4 in the Supplement). Restricting our analysis to primary care visits, rates of low value care delivery were minimally changed for non–safety-net and safety-net physicians, whereas high-value care was less frequent among safety-net physicians (eTable 4 in the Supplement). Finally, if we excluded the narcotic prescribing measures from our composite outcome, overall rates of low-value care were lower among Medicaid or uninsured patients vs privately insured patients (eTable 4 in the Supplement).
Discussion
In our analysis of nationally representative sample of ambulatory visits from 2005 to 2013, we found that the overall delivery of low-value care was as common among patients without insurance or with Medicaid as among the privately insured. This was also true for physicians serving a high proportion of Medicaid and uninsured patients compared with those serving a low proportion. While one reasonable hypothesis might be that patients without insurance or with Medicaid would receive less low-value care due to financial and social barriers to care, we do not find evidence for this. Instead, we observe similar levels of low-value care across different patient and physician types, suggesting that national concerns about overuse of low-value care in Medicare and private insurance can be extended to lower-income patients with Medicaid and even uninsured patients. Our findings are consistent with a previous single-state study of Medicaid and suggest that overuse is an important target for state policymakers seeking to manage their Medicaid budgets.
More broadly, these results suggest that physicians do not systematically discriminate in treatment patterns across insurance types for low-value care, which is consistent with the literature on managed care and other payment reforms. Ingrained patterns of overuse among physicians appear to transcend the insurance status of the individual patient, even among the uninsured. One other key insight from this study is that the differences in care that we do observe by insurance status do not appear to be driven by the different types of physicians seen by lower-income patients, given that we find similar patterns of low-value care delivered by safety net and non–safety-net physicians. This is an important finding, because many policy proposals aimed at Medicaid presuppose that differential access to high quality physicians is a key challenge for beneficiaries. Our study does not provide support for that argument.
While overall rates of low-value care were generally comparable across insurance types and physician types, there were important differences for specific measures. Most notably, rates of inappropriate narcotic prescribing were significantly higher in the Medicaid and uninsured populations (though the difference in narcotic prescribing for headaches in the uninsured was not statistically significant). These differences drove a significant proportion of overall low-value care. Because we lack context on the patients receiving narcotics outside of a single visit, we are unable to identify the drivers of this difference. However, this is a worthwhile area for future research.
We also observed differing rates of high-value care for Medicaid and uninsured patients in our pooled composite measures, though the majority of individual measures were not significantly different compared with privately insured patients. For Medicaid patients, overall rates of high-value care were lower than for the privately insured. This could be explained by our lack of data on important social confounders, such as education, income, or social support, which may be responsible for much of the observed differences. There have been few quasiexperimental comparisons of private and public insurance, though one such recent study indicated similar outcomes for low-income populations with Medicaid vs private coverage.
In contrast, we actually find higher pooled rates of recommended care among the uninsured compared with private insurance, though the majority of high-value measures were not significantly different. This may be surprising given the well-known difficulty uninsured patients have accessing care. However, for uninsured patients to be present in this analysis, they must first be accessing outpatient care, implying that our uninsured sample disproportionately includes patients who are motivated and able to receive care. This also points to the risk of residual confounding in our study design and our inability to make causal statements about the quality difference attributable to an individual having one insurance type over another. Any argument that our findings point to a quality decrement in Medicaid would also imply that being uninsured is preferable to having private insurance, which is highly unlikely.
Aside from the contrast between patient and physician groups, overall rates of low-value care were substantial, while rates of high-value care were generally below 50% of visits. This is broadly consistent with other studies examining national trends in the delivery of low-value or high-value care. Combining our results with other prior literature, adherence to standard high-value care appears to have improved substantially in the 1990s but has been largely flat since then.
Limitations
Our study has important limitations. First, this is a cross-sectional, observational analysis and cannot demonstrate causal relationships between insurance coverage or physician type and the delivery of low-value or high-value care. Second, we rely on NAMCS and NHAMCS, which are high-quality and validated national surveys, but nonetheless subject to both sampling and measurement error. We addressed these issues by pooling multiple years of data to increase statistical power and assessing multiple quality measures to avoid reliance on any single measure. Another concern is that for most measures we are unable to assess a comprehensive set of inclusion or exclusion criteria, such as drug intolerance and/or allergies, because we can only use data available within a single encounter. However, we have no reason to believe that this measurement error should be systematically different between different patient or physician subgroups. Other potential sources of bias are that we rely on insurance coverage as reported by physician offices in NAMCS and NHAMCS and cannot observe whether patients are in managed care, which may be particularly relevant for states with large Medicaid managed care programs. In addition, our statistical tests do not account for the false-positive rate associated with testing across multiple outcomes and subgroups; therefore, the significant differences noted between subgroups for individual quality measures should be regarded as exploratory. Last, 8 of the 12 high-value measures we used involved prescribing for cardiovascular disease, which may limit the generalizability of our composite high-value care measure.
Conclusions
We found that rates of low-value care were similar among patients with Medicaid, no insurance, or private insurance coverage, and between physicians serving high and low proportions of uninsured and Medicaid patients. Our results appear to be driven more by patient factors than the quality of care delivered by safety-net physicians, whose rates of low-value care were similar to other physicians. These results show that in addition to improving underuse of high-value care, overuse of low-value care is a potentially important focus for state Medicaid programs and safety-net institutions to pursue cost-savings and improved quality.
eAppendix. Developing and Modeling Low-Value and High-Value Composite Measures
eTable 1. Low-value and high-value care measure definitions
eTable 2. Visit Characteristics by Proportion of Safety Net Patients Seen by Physicians, 2005-2012
eTable 3. Low and High-value Quality Measures by Physician Safety Net Proportion, including 0% Safety Net Providers
eTable 4. Other sensitivity analyses
eFigure 1. Trends in Low- and High-Value Care Delivery Composite Measures by Patient Insurance Status, 2005-2011
eFigure 2. Trends in Low- and High-Value Care Delivery Composite Measures by Proportion, 2005-2011
eReferences.
References
- 1.ABIM Foundation. Choosing Wisely: Promoting conversations between providers and patients. http://www.choosingwisely.org/. Accessed June 6, 2016.
- 2.Schwartz AL, Landon BE, Elshaug AG, Chernew ME, McWilliams JM. Measuring low-value care in Medicare. JAMA Intern Med. 2014;174(7):1067-1076. doi: 10.1001/jamainternmed.2014.1541 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Charlesworth CJ, Meath TH, Schwartz AL, McConnell KJ. Comparison of low-value care in Medicaid vs commercially insured populations. JAMA Intern Med. 2016;176(7):998-1004. doi: 10.1001/jamainternmed.2016.2086 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Kale MS, Bishop TF, Federman AD, Keyhani S. Trends in the overuse of ambulatory health care services in the United States. JAMA Intern Med. 2013;173(2):142-148. doi: 10.1001/2013.jamainternmed.1022 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Rosenberg A, Agiro A, Gottlieb M, et al. Early trends among seven recommendations from the choosing wisely campaign. JAMA Intern Med. 2015;175(12):1913-1920. doi: 10.1001/jamainternmed.2015.5441 [DOI] [PubMed] [Google Scholar]
- 6.Reid RO, Rabideau B, Sood N. Low-value health care services in a commercially insured population. JAMA Intern Med. 2016;176(10):1567-1571. doi: 10.1001/jamainternmed.2016.5031 [DOI] [PubMed] [Google Scholar]
- 7.Cunningham P, May J. Medicaid patients increasingly concentrated among physicians. Track Rep. 2006;(16):1-5. [PubMed] [Google Scholar]
- 8.Landon BE, Schneider EC, Normand SL, Scholle SH, Pawlson LG, Epstein AM. Quality of care in Medicaid managed care and commercial health plans. JAMA. 2007;298(14):1674-1681. doi: 10.1001/jama.298.14.1674 [DOI] [PubMed] [Google Scholar]
- 9.Centers for Disease Control and Prevention. NAMCS/NHAMCS—Questionnaires, Datasets, and Related Documentation. http://www.cdc.gov/nchs/ahcd/ahcd_questionnaires.htm. Accessed July 14, 2016.
- 10.Kaiser Family Foundation Dual Eligibles: Medicaid’s Role for Low-Income Medicare Beneficiaries. http://kff.org/medicaid/fact-sheet/dual-eligibles-medicaids-role-for-low-income-2/. Accessed July 18, 2016.
- 11.America’s Health Care Safety Net Intact but Endangered. Washington, DC: National Academies Press; 2000. http://www.nap.edu/catalog/9612. Accessed September 26, 2016. [PubMed]
- 12.Gelman A, Hill J. Data Analysis Using Regression and Multilevel/Hierarchical Models. New York, NY: Cambridge University Press; 2007. [Google Scholar]
- 13.Barnett ML, Linder JA. Antibiotic prescribing to adults with sore throat in the United States, 1997-2010. JAMA Intern Med. 2014;174(1):138-140. doi: 10.1001/jamainternmed.2013.11673 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Barnett ML, Linder JA. Antibiotic prescribing for adults with acute bronchitis in the United States, 1996-2010. JAMA. 2014;311(19):2020-2022. doi: 10.1001/jama.2013.286141 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Mehrotra A, Gidengil CA, Setodji CM, Burns RM, Linder JA. Antibiotic prescribing for respiratory infections at retail clinics, physician practices, and emergency departments. Am J Manag Care. 2015;21(4):294-302. [PubMed] [Google Scholar]
- 16.Mehrotra A, Zaslavsky AM, Ayanian JZ. Preventive health examinations and preventive gynecological examinations in the United States. Arch Intern Med. 2007;167(17):1876-1883. doi: 10.1001/archinte.167.17.1876 [DOI] [PubMed] [Google Scholar]
- 17.Linder JA, Ma J, Bates DW, Middleton B, Stafford RS. Electronic health record use and the quality of ambulatory care in the United States. Arch Intern Med. 2007;167(13):1400-1405. doi: 10.1001/archinte.167.13.1400 [DOI] [PubMed] [Google Scholar]
- 18.Mafi JN, McCarthy EP, Davis RB, Landon BE. Worsening trends in the management and treatment of back pain. JAMA Intern Med. 2013;173(17):1573-1581. doi: 10.1001/jamainternmed.2013.8992 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Mafi JN, Edwards ST, Pedersen NP, Davis RB, McCarthy EP, Landon BE. Trends in the ambulatory management of headache: analysis of NAMCS and NHAMCS data 1999-2010. J Gen Intern Med. 2015;30(5):548-555. doi: 10.1007/s11606-014-3107-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Edwards ST, Mafi JN, Landon BE. Trends and quality of care in outpatient visits to generalist and specialist physicians delivering primary care in the United States, 1997-2010. J Gen Intern Med. 2014;29(6):947-955. doi: 10.1007/s11606-014-2808-y [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Ma J, Stafford RS. Quality of US outpatient care: temporal changes and racial/ethnic disparities. Arch Intern Med. 2005;165(12):1354-1361. doi: 10.1001/archinte.165.12.1354 [DOI] [PubMed] [Google Scholar]
- 22.Papke LE, Wooldridge JM Econometric Methods for Fractional Response Variables with an Application to 401(k) Plan Participation Rates. National Bureau of Economic Research. 1993. http://www.nber.org/papers/t0147. Accessed May 13, 2016.
- 23.R Core Team R: A Language and Environment for Statistical Computing. Vienna, Austria: R Foundation for Statistical Computing. 2014. http://www.R-project.org/. Accessed March 25, 2015.
- 24.Lumley T. Analysis of complex survey samples. J Stat Softw. 2004;9(1):1-19. [Google Scholar]
- 25.Glied S, Zivin JG. How do doctors behave when some (but not all) of their patients are in managed care? J Health Econ. 2002;21(2):337-353. [DOI] [PubMed] [Google Scholar]
- 26.McWilliams JM, Landon BE, Chernew ME. Changes in health care spending and quality for Medicare beneficiaries associated with a commercial ACO contract. JAMA. 2013;310(8):829-836. doi: 10.1001/jama.2013.276302 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Baicker K, Robbins JA. Medicare payments and system-level health-care use: the spillover effects of Medicare managed care. Am J Health Econ. 2015;1(4):399-431. doi: 10.1162/AJHE_a_00024 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Decker SL. In 2011 nearly one-third of physicians said they would not accept new Medicaid patients, but rising fees may help. Health Aff (Millwood). 2012;31(8):1673-1679. doi: 10.1377/hlthaff.2012.0294 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Ryan P. A Better Way. http://abetterway.speaker.gov/?page=health-care. Accessed July 14, 2016.
- 30.Sommers BD, Blendon RJ, Orav EJ, Epstein AM. Changes in utilization and health among low-income adults after Medicaid expansion or expanded private insurance. JAMA Intern Med. 2016;176(10):1501-1509. doi: 10.1001/jamainternmed.2016.4419 [DOI] [PubMed] [Google Scholar]
- 31.Frakt A, Carroll AE, Pollack HA, Reinhardt U. Our flawed but beneficial Medicaid program. N Engl J Med. 2011;364(16):e31. doi: 10.1056/NEJMp1103168 [DOI] [PubMed] [Google Scholar]
- 32.Levine DM, Linder JA, Landon BE. The quality of outpatient care delivered to adults in the United States, 2002 to 2013. JAMA Intern Med. 2016;176(12):1778-1790. doi: 10.1001/jamainternmed.2016.6217 [DOI] [PubMed] [Google Scholar]
- 33.Gilchrist VJ, Stange KC, Flocke SA, McCord G, Bourguet CC. A comparison of the National Ambulatory Medical Care Survey (NAMCS) measurement approach with direct observation of outpatient visits. Med Care. 2004;42(3):276-280. [DOI] [PubMed] [Google Scholar]
- 34.Sommers BD, Gunja MZ, Finegold K, Musco T. Changes in self-reported insurance coverage, access to care, and health under the Affordable Care Act. JAMA. 2015;314(4):366-374. doi: 10.1001/jama.2015.8421 [DOI] [PubMed] [Google Scholar]
- 35.U.S. House of Representatives Committee On the Budget. Fiscal Year 2012. Budget. http://budget.house.gov/fy2012budget/. Accessed July 14, 2016.
- 36.Farley R. Is Medicaid Bad for Your Health? http://www.factcheck.org/2015/07/is-medicaid-bad-for-your-health/. Published July 10, 2015. Accessed September 12, 2016.
- 37.Health Resources & Service Administration. How is rural defined? http://www.hrsa.gov/healthit/toolbox/RuralHealthITtoolbox/Introduction/defined.html. Accessed July 19, 2016.
- 38.The White House: President Barack Obama. Presidential Memorandum—Addressing Prescription Drug Abuse and Heroin Use. https://obamawhitehouse.archives.gov/the-press-office/2015/10/21/presidential-memorandum-addressing-prescription-drug-abuse-and-heroin Published October 21, 2015. Accessed July 31, 2016.
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
eAppendix. Developing and Modeling Low-Value and High-Value Composite Measures
eTable 1. Low-value and high-value care measure definitions
eTable 2. Visit Characteristics by Proportion of Safety Net Patients Seen by Physicians, 2005-2012
eTable 3. Low and High-value Quality Measures by Physician Safety Net Proportion, including 0% Safety Net Providers
eTable 4. Other sensitivity analyses
eFigure 1. Trends in Low- and High-Value Care Delivery Composite Measures by Patient Insurance Status, 2005-2011
eFigure 2. Trends in Low- and High-Value Care Delivery Composite Measures by Proportion, 2005-2011
eReferences.