Table 1 Summary of Included Studies
From: Identifying drivers of health care value: a scoping review of the literature
First Author Year Country | Value Definition | Population | Independent Variable(s) and Covariates | Outcome | Major Findings (Adjusted, if presented) |
|---|---|---|---|---|---|
Insurance or Health System Characteristics | |||||
Barnetta, [29] 2017 US | Defined based on a set of low value services | Patient visits recorded in NAMCS and NHAMCS between 2005 and 13 and 2005–11, respectively (n = 193,062 eligible office visits) | Independent Variables: patient’s insurance at the time of the outpatient visit and the proportion of vulnerable patients seen by a given physician Covariates: year, age, sex, race/ethnicity, comorbidity, region, rural, practice setting | 9 low value and 12 high value services identified in published guidelines and prior literature and composite measures of high and low value service utilization | Similar low and high value service use in Medicaid, uninsured, and privately insured. One exception was Medicaid and uninsured patients (compared to private) were more likely to get inappropriate narcotic prescriptions. In composite measures, Medicaid patients had lower receipt of high value services compared to privately insured (aOR = 0.88, 0.83–0.94). Uninsured had higher receipt of high value services compared to privately insured (aOR = 1.44, 1.28–1.61). No significant difference in high/low value service use between safety-net and non-safety-net physicians. Sensitivity analyses did not change results. |
Braithwaite, [35] 2010 US | “Defined by the ratio of incremental benefits to incremental costs” | US Population | Independent Variables: VBID reducing cost sharing for high value services with or without increasing cost sharing for lower value services (no cost-offset, cost-offset without uninsured subsidy, cost-neutral with uninsured subsidy) Covariates: age, insurance | Life years gained, spending changes | 60% of US health expenditures spent on low value services (ICER>$300,000/life year), 20% on intermediate value services (ICER $100–$300,000/life year), 20% on high value services (ICER $100,000/ life year). Applying VBID with no cost offset led to 0.24 life years gained and $22 billion increased spending. Cost offset without uninsured subsidy led to 0.25 life years gained and unchanged to reduced spending. Cost offset and uninsured subsidy led to 0.44 life years gained and no change in spending. Sensitivity analyses did not change results. |
Charlesworth, [17] 2016 US | Defined as a set of services that “provide little clinical benefit and may even cause harm” | 2012–13 Oregon Medicaid and commercial claims for 18–64 in Oregon All Payer All Claims Database | Independent Variable: Medicaid vs commercial insurance Covariates: age, sex, rural, comorbidities, primary care service area | 13 low value services from Schwartz et al. and 3 CMS Quality Net measures | No consistent association was found between insurance type and low value care. Medicaid patients were more likely to receive low value care for services in the ED (pinteraction < 0.001), Medicaid patients were more likely to receive low value care if they lived in an area with a higher rate of commercial low value care for 7/11 services. |
Colla, [19] 2018 US | Low value services are those whose “avoidance would increase the quality and value of care provided” | Fee for service Medicare beneficiaries enrolled in parts A and B from 2009 to 11 and commercially insured with claims in the Health Care Cost Institute claims base (n = 933.4 M) | Independent Variable: payer type Covariates: Hospital referral region, Medicare spending, physician group concentration, ratio of specialists to PCPs, mortality, proportion in poor or fair health, race, ethnicity, Medicare effective use score, rural, income, quality | 7 low value services and a composite measure of them | Across payer types, use of low value care varied little for almost all measures. Those with commercial insurance had higher use of 3 measures of low value care. Those with Medicare had higher use of a different 3 measures of low value care. Correlation between Medicare and commercial low value care ranged from 0.540–0.905. Some regional health system characteristics (e.g. specialist to PCP ratio, Medicare spending, and physician group concentration) and patient characteristics (e.g. proportion in poor or fair health, race, poverty) were associated with low value care in both cohorts, but most of the associations were weak. |
Coronini-Cronberg, [33] 2015 England | Low value procedures are those considered to be “ineffective, overused, or inappropriate” | 2002–12 admissions and outpatient appointments at NHS hospitals in England recorded in HES dataset (n = 5,248,808 episodes) | Independent Variables: Before/after efficiency savings program, commissioning organization Covariates: year, sex, age, socioeconomic status | 3 services considered ineffective and 3 services considered effective only in certain situations | Annual rate (per-1000) for each procedure varied from 0.99–3.87. Changes following implementation of the efficiency savings program varied with rates of utilization declining for 3 low value services or increasing or remaining unchanged for others. Rates of utilization of two benchmark procedures during this year were unchanged. Changes in utilization of low value services also varied considerably across commissioning organizations. |
Koehlmoos, [20] 2019 US | Low value: overuse or inappropriate care, including procedures and treatments that are clinically inappropriate, excessively intensive, or too frequent | All TRICARE Prime and Prime Plus beneficiaries 18+ in 2014. This includes 20% active military, dependents, and retired military personnel (n = 2.9 M) | Independent variable: direct vs purchased care Covariates: age, sex, race | 19 low value service indicators identified by Segal et al. | Six procedures were used more frequently in direct care settings, 11 more frequently in purchased care settings, and two showed no significant difference. Note: no adjustment for patient characteristics |
Reid, [31] 2017 US | Low value “services are medical tests and procedures that provide unclear or no clinical benefit, but still expose patients to risk and expense” | 25% random sample of patients 18–64 in a United health plan from 2011 to 13 not in HMOs or exclusive provider organizations (n = 11,074 CDHP, 365,016 traditional) | Independent Variable: CDHP vs traditional plan Covariates: age, sex, race, household income, region, comorbidities | 26 measures of outpatient low value services and their associated costs. Also looked at costs for frequently co-occurring services or outpatient costs for complex services | Switching to a CDHP was associated with a $231.60 reduction (95% CI: -$341.65, −$121.53) in yearly outpatient spending. There was no significant association between switching to a CDHP and low value spending. There was a small association between switching to a CDHP and reduced low value outpatient spending on imaging (−$1.76, −$3.39, −$0.14), but no difference in low value imaging spending relative to overall imaging spending. Results remained true after sensitivity analyses. |
Schwartz, [32] 2015 US | Low value services are those that “provide minimal clinical benefit on average” | Random sample of 2009–12 Medicare beneficiaries enrolled in Parts A and B (nACO = 693,218 person-yrs, ntraditional = 17,453,423) | Independent Variable: ACO vs traditional Medicare plan Covariates: age, sex, race, ethnicity, disability, comorbidities, hierarchical condition category risk score, geographic area | 31 claims-based measures of low value care derived from evidence-based lists of low value services | In year 1 of Pioneer ACO contracts, there was a differential reduction in low value service use in the ACO group (− 0.8 services/100 beneficiaries, − 1.2 to − 0.4). This corresponded to a 4.5% (− 7.5% to − 1.5%) differential reduction in spending on low value services. Similar reductions were seen for high- and low-priced services and services that were more or less sensitive to patient preferences. ACOs with higher baseline levels of low value service use exhibited a higher differential reduction of services. |
Hospital Characteristics | |||||
Badgery-Parker, [34] 2019 Australia | Low-value: “tests and interventions for which the benefit is not expected to outweigh the harm and/or costs” | All patients in public hospitals in New South Wales hospitals who receive a procedure considered to be low value (n = 225 hospitals) | Independent Variables: Hospital Local Health District Statistical Local Area Covariates: age, comorbidities, private insurance | Receipt of any of 9 low value inpatient procedures selected from Choosing Wisely, RACP Evolve, and NICE | Hospital is a more significant contributor to low value care than local health district or statistical local area. For knee arthroscopy, the hospital MOR was 4.3 (95% CI: 3.3–5.7), meaning the odds of a patient receiving the procedure was, on average, 4.3 times higher if the patient went to a hospital in the same local health district with more low value utilization. Hospital MORs for other services were 1.7–3.5. Younger age and higher comorbidity score were associated with reduced odds of low value care. |
Weeks, [27] 2016 US | Value equals quality divided by expenditures | All hospitals with available Hospital Compare and Medicare spending data for AMI, CABG, colectomy, and hip replacement in 2012 | Independent Variables: Hospital Size Hospital Census Number of Beds Accreditation Hospital Profit Status | Value as defined by a composite measure of patient experience, quality of care processes, outcomes, and safety divided by expenditures for the episode of care | Hospitals in highest value quintiles had higher census (p < 0.003), more beds (p < 0.019), and more operations (p < 0.001 for 2/4 conditions) compared to those in the lowest value quintile. Accreditation was not associated with value. CABG did not show the same associations with hospital census, beds, operations, and accreditation. Only hip replacement was associated with hospital profit status (p < 0.006). Half of hospitals in highest or lowest value quintiles also in highest or lowest quality quintile. |
Physician/Practice Characteristics | |||||
Barreto, [15] 2019 US | Low value: “unnecessary services” or “‘overuse beyond evidence based levels and unnecessary choice of higher cost services”’ | Stratified random sample of PCPs matched to 2011 Medicare Part B beneficiaries (n = 6873 PCPs, 1,078,840 patients) | Independent Variables: PCP sex, years in practice, specialty, credential, international graduate status, patient panel size, location Covariates: age, sex,race, ethnicity, comorbidities | Cost of primary care services considered low value based on Choosing Wisely recommendations | Lower annual per patient low value care Medicare spending was associated with allopathic training (β = −$1.65), smaller Medicare patient panel (β = −$3.98 for panels < 50 vs panels > 300), family medicine practice (β = −$1.03 vs internal medicine), practicing in the Midwest (β = −$2.80 vs Northeast), practicing in a rural area (β = −$1.75), and being a recent graduate. |
Bouck, [16] 2018 Canada | Low value care “represents little to no patient benefit, or comparatively greater risk of harm” | Identified test-specific cohorts across Ontario outpatient visits for 4 low value tests. Excluded physicians with < 50 in 2012–14. (n = 2394 physicians) | Independent Variables: physician sex, international graduate status, years since graduation, billing group care model and size Covariates: patient sociodemographic variables | 4 screening tests identified as low value based on Choosing Wisely Canada recommendations | Physicians who were male (OR = 1.29, 1.01–1.64)), further out from medical school graduation (OR = 1.03, 1.02–1.04), domestic graduates (OR = 1.56, 1.19–2.04), or in an enhanced fee for service payment model (OR = 2.04, 1.42–2.94 vs capitated) had increased odds of being frequent users of low value tests. 18.4% of physicians ordered 39.2% of low value tests. Billing group differences of greater relevance than patient or physician factors. |
Mafi, [21] 2016 US | Low value care is care that often has a “greater probability of harm than benefit” | Patients with primary care visits recorded in NAMCS or NHAMCS from 1997 to 2011 for 3 conditions. (n = 25,529 physician and 3420 APC visits) | Independent Variable: APC vs MD Covariates: Age, sex, race/ethnicity, comorbidities, symptom acuity, insurance status, rural, region, year, PCP vs not PCP, hospital vs community practice, practice setting type | Low value services identified for each of these conditions based on Choosing Wisely recommendations | APCs were not significantly more likely to order antibiotics (p = 0.96 office-based, p = 0.92 hospital-based), CT/MRI (p = 0.086 office-based, p = 0.48 hospital-based) or radiography (p = 0.65 office-based, p = 0.73 hospital-based), or refer to specialist physicians compared to physicians (p = − 0.64 office-based, p = 0.091 hospital-based). When stratified by PCP vs non-PCP, hospital-based PCP APCs ordered more antibiotics and made more referrals than hospital-based PCP physicians. |
Mafi, [14] 2017 US | Low value care is “care that provides minimal average benefit in specific clinical scenarios” | Patients with primary care visits recorded in NAMCS or NHAMCS from 1997 to 2013 for URTI, back pain, or headache. (n = 31,162) | Independent Variables: hospital/community-based practice, hospital or physician owned, Covariates: age, sex, race/ethnicity, usual PCP, comorbidities, insurance status, rural, region, year, symptom acuity | Low value services identified for each of these conditions based on Choosing Wisely recommendations | Visits to hospital-based practices had higher use of CT or MRI (OR = 1.44, 1.13–1.85), radiographs (1.41, 1.16–1.71), and specialty referrals (2.74, 2.23–3.36) compared to community based. Antibiotic use was similar in both locations. Non-PCP visits associated with higher use of imaging and specialty referral, primarily in hospital practices. No significant differences in hospital owned vs physician owned practice in terms of antibiotic and imaging use, but hospital owned had more specialty referrals. |
Oronce, [23] 2019 US | High Value: “guideline-concordant care that improves health, avoids harm, eliminates waste;” Low value: overuse of “services that provide marginal or unknown benefit” | All visits from patients 18+ to PCPs with eligible ICD9 codes recorded in NAMCS in 2010–12. (n = 29,155) | Independent Variables: Community health center vs private practice Covariates: age, sex, race/ethnicity, insurance, comorbidities, rural, region, year | 12 measures of high value care and 7 measures of low value care consistent with previously published guidelines and studies | Community health centers were more likely to prescribe beta blockers for CHF visits (OR = 2.56, 1.18–5.56), statins for diabetes (OR = 1.35, 1.02–1.79), and treatment for osteoporosis (OR = 1.77, 1.05–3.00) compared to private practices. They were also more likely to avoid screening urinalysis (OR = 1.87, 1.11–3.14), complete blood count (OR = 1.72, 1.18–2.53), and EKGs (OR = 11.03, 2.67–45.52), and they were less likely to prescribe antibiotics for upper respiratory infections (OR = 0.59, 0.40–0.88). Overall, community health centers performed better or comparably to private practices on high and low value care measures explored in this study. |
Schwartz, [30] 2018 US | Low value services “provide minimal average clinical benefit in specific clinical scenarios” | 20% sample of 2007–11 Medicare FFS patients in parts A and B matched to provider organizations (n = 4,039,733 patients and 3137 provider organizations) | Independent variable: provider organization defined by TIN Covariates: age, sex, race/ethnicity, disability, ESRD, comorbidities, area-level socioeconomic status | Rate of utilization of 31 low value services based on previous studies and guidelines as well as a composite measure of low value service utilization | Between organization standard deviation in use of low value services were 10 (95% CI: 9.4–40.5). Variation in services more sensitive to patient preference (90th:10th percentile = 1.61, 1.58–1.64) was much less than variation in services categorized as less sensitive (90th:10th percentile = 2.84, 2.6–3.13). Within organizations, service use was positively correlated between almost all pairs of categories of low value services. |
Schwartz, [28] 2019 US | Low value services are those that “produce minimal average clinical benefit in specific clinical scenarios” | 20% sample of Medicare FFS beneficiaries enrolled in parts A and B from 2008 to 13 matched to PCPs (n = 3,159,834 beneficiaries and 41,773 PCPs) | Independent variables: physician age, sex, training, academic engagement, payment from pharmaceutical or device manufacturer, patient panel size Covariates: Patient age, sex, race/ethnicity, disability, dual eligibility, ESRD, area-level SES | Rate of use of 17 low-value services (based on previous studies and guidelines) reported as number of services per 100 patients per year | Mean rate of low value service provision was 33.1 services per 100 beneficiaries per year. Smaller variation in rates of low value service utilization observed than predicted by differences in physician characteristics. Within organizations, ratio of low value service use between providers at the 90th and 10th percentiles of low value use was 1.67 per 100 beneficiaries per year. Variation in low value service utilization across organizations was much greater than would be expected due to chance. The only physician characteristic predicting substantially higher low value service utilization was patient panel size. |
Physician and Patient Characteristics | |||||
Colla, [18] 2014 US | Low value defined by “services whose avoidance improve efficiency through higher quality, reduced risks, and lower costs” | Fee for service Medicare beneficiaries enrolled in Parts A and B from 2006 to 11 | Variables: Hospital referral region, per-beneficiary Medicare spending, physician group concentration, ratio of specialists to PCPs, quality, mortality, percent in poor or fair health, race, ethnicity, effective use score, rural, income | 11 services representing 37 Choosing Wisely recommendations that can be measured using Medicare claims data | Higher Medicare spending was associated with low value care utilization (coeff = 0.099, 0.053–0.145) as was higher ratio of specialists to PCPs (coeff = 0.343, 0.060–0.626), higher proportion of black (coeff = 0.018, 0.011–0.025) and Hispanic (coeff = 0.015, 0.001–0.023), and higher proportion of residents in poor/fair health (coeff = 0.026, 0.001–0.052). Higher poverty rate (coeff = − 0.025, − 0.044, − 0.006) and higher physician group concentration (coeff = − 0.008, − 0.148, − 0.002) was associated with lower low value care utilization. Quality was not associated with low value service use. |
McAlister, [22] 2017 Canada | Low value care is defined as “health care practices that provide minimal or no benefit to recipients” | All patients 18+ in Alberta that presented to a health care provider between 2012 and 15 (n = 3,423,135) | Independent Variables: age, sex, comorbidities, region, physician contacts in past 12 months, median household income, specialist to PCP ratio | 10 low value services identified from Choosing Wisely lists that have been evaluated before and can be identified using claims data | Higher socioeconomic status (OR = 1.14–1.46 for each of 4 low value services), increased frequency of specialist contact (OR = 1.003–1.006 for each of 4 low value services), and higher ratio of specialists to PCPs (OR = 1.22–7.79 for each of 4 low value services) were associated with increased likelihood of receiving low value services. Provision of low value services varied significantly across eight regions, but not consistently in one direction. |
Patient Characteristics | |||||
Pendrith, [24] 2017 Canada | Low value care is care with a lack of benefit that can lead to higher health care costs and inconvenience or harm to patients | Ontario patients meeting eligibility criteria as recorded in the OHIP claims database between 2008 and 2013 (n = 9,918,145) | Variables: age, sex, rural, neighborhood income quintile, practice, Local Health Integration Network | 3 Choosing Wisely recommendations relevant to primary care | Substantial variation in all services, with repeat DEXA scan being the most common (21% of those receiving an index scan) and low value cervical cancer screening being the only one to decrease significantly in the study period. Predictors of repeat DEXA scan include older (p < 0.001), female (p < 0.001), urban (p < 0.001), high income area (p < 0.001), and high risk index scan (p < 0.001). Note: results not adjusted |
Reid, [25] 2016 US | Defined based on a set of low value services | 25% random sample of 2011–13 Optum Clinformatics Data Mart claims for United Healthcare patients 18–64 across the US (n = 1,468,689) | Variables: age, sex, race, ethnicity, income, area, Consumer-Directed Health Plan | 28 low value care measures previously published. Assessed number of patients receiving each service and associated costs | Low value spending per $10,000 in total spending was less among patients who were older (35–49 yrs.: − 20.42 [− 22.03, − 18.81]; 50–64 years: − 11.30 [− 12.94, − 9.66] compared to 18–34), male (− 18.19 [− 19.34, − 17.03]), black (− 3.81 [− 5.68, − 1.95]), Asian (− 4.40 [− 7.23, − 1.57]), low-income (− 8.10 [− 10.55, − 5.66]), or enrolled in a Consumer Directed Health Plan (− 5.86 [− 7.51, − 4.21]). West South Central region had highest effect of low value spending per $10,000 (14.26 [12.17, 16.35]). |
Schpero, [26] 2017 US | Low value services are those which are “unnecessary and economically inefficient, many are potentially harmful” | 2006–11 Medicare beneficiaries at risk of receiving a specified low value service (n varied by test) | Independent Variables: Race, ethnicity Covariates: age, income, sex, disability status, Medicaid enrollment, risk score, health care utilization, year, hospital referral region | 11 low value services identified by Choosing Wisely | Black and Hispanic beneficiaries had qualitatively higher receipt of low value services compared to white beneficiaries before adjusting. After adjustments, receipt of low value care was significantly higher among blacks for 5/11 services and for 6/11 services for Hispanics. |