Discrepancies Among Hospitals and Regions in the Provision of Low-Value Care

.


Background
5][6] In an effort to stem the financial burden imposed by such services, policy makers and experts have passed a number of initiatives, such as "Do not Do" (National Institute for Health and Care Excellence, NICE) 7 and "Choosing Wisely" (American Board of Internal Medicine, ABIM). 8Researchers have also created country-specific lists of examples of low-value care.
Most previous research on low-value care utilization were conducted in the United States, 2,4,9- 11 Canada, [12][13][14] Australia, [15][16][17][18] and European countries. 19,206][27][28][29][30] Researchers identified a number of measures that are associated with the utilization of low-value care, including patient characteristics (e.g., age, sex, ethnicity, socioeconomic status, and comorbidities) 21,24,31 and physician characteristics (e.g., seniority, specialty, and patient panel size). 13,23,32Note that factors related to the utilization of low-value care in Asia have yet to be elucidated and country-specific lists of low-value interventions have not been developed for most Asian countries.We posited that the research performed in other countries could be used to lay the groundwork for future work in this area.
In 1995, Taiwan implemented a single-payer mandatory National Health Insurance (NHI) program, now encompasses over 99% of 23 million residents and 93% of the hospitals and clinics. 33The NHI program offers comprehensive healthcare service, including outpatient visit, hospitalization, examinations, prescriptions, rehabilitation, and home care, with 30% of the contracted facilities being public.The NHI program is known for its high accessibility and affordability.However, despite its success, there are concerns about the ineffective gatekeeping of specialist services and the general quality of care. 34In the current study, we aimed to assess the situation of low-value care in terms of utilization, cost, and trend over a five-year period.We also evaluated characteristics associated with the increased risk of lowvalue care at the hospital and regional levels.

Data Source and Study Design
We adopted a non-interventional, retrospective cohort design to measure the prevalence of INTERNATIONAL JOURNAL OF HEALTH POLICY AND MANAGEMENT (IJHPM) ONLINE ISSN: 2322-5939 JOURNAL HOMEPAGE: HTTPS://WWW.IJHPM.COM grouped by quartiles into four subcategories.All regional level variables of interest were split at the median to form high and low groups.Trend analyses on the utilization rate of low-value care services, the number of affected beneficiaries, the number of episodes, and the corresponding costs were performed using the general linear model (GLM).The generalized estimating equation (GEE) model was used to determine whether variations observed at the hospital and regional level were associated with the utilization of low-value care services (in terms of the number of episodes per 10,000 beneficiaries).Affected hospitals were classified into high-and low-cost groups based on the 75 th percentile of corresponding costs.We also analyzed the relationship between characteristics of interest and high-cost group using GEE.
Additional analyses which excluded two sex-specific low-value care services (e.g., prostate specific antigen test for men aged over 75 years old and repeated X-ray bone densitometry in short intervals) 42 were performed to examine the substantive associated factors.All analyses were performed using SAS, 9.4 version (SAS, Gray, North Carolina) with the level of statistical significance set at p<.05 based on two-tailed tests.

Ethical Issues/Statement
This study has been approved by the Joint Institutional Review Board (IRB no: 17-S-017-1).

Characteristics of Beneficiaries, Hospitals, and Region
Between 1 January, 2013 to 31 December, 2017, 914,191 beneficiaries (about 1.03% of all beneficiaries) received at least one of the selected low-value care services, for a total of 1,218,146 beneficiary-year observations.We identified 493 hospitals that were providing lowvalue care for a total of 2,054 hospital-year observations.Table 1 presents the baseline characteristics at the beneficiary, hospital, and regional levels.The mean age of affected beneficiaries was 68.97 years (SD, 15.63), most of whom were male (65.83%).The majority of identified hospitals were local facilities (75.26%), and 52.54% of them were private.The mean proportion of male patients treated in the hospitals was 42.81% and the average experience of physicians was 15.16 years (SD, 4.54).Only 2.78% of the hospitals were located in mountainous areas, offshore islands, or districts with insufficient medical resources.

Extent and Trend of Low-Value Care
In measuring the utilization of seven low-value care services during the study period, we identified 1,970,496 distinctive episodes, with a corresponding cost of US$30.41 million.
Figure 1 shows the utilization rate per 10,000 beneficiaries and the associated costs of the seven services..59]for Q1-Q2, respectively).The utilization was also positively correlated with the age of the patients, the proportion of male patients, and the presence of comorbidities.In terms of regional factors, low-value care utilization was inversely proportional to the proportion of residents who completed senior secondary education (episodes per 10 000 beneficiaries [95%CI], -5.65 [-10.99 to -0.32]).
Other characteristics were not significantly related to the utilization of low-value care, including average combined comorbidity score, the ratio of specialists to primary care physicians, the proportion of low-income households and remoteness of location.Figure 3 presents the result of sensitivity analysis on the association between characteristics of interest and the utilization of non-sex-specific low-value care services.We found that the correlation remained significantly positive between use and the proportion of male patients within hospitals (episodes per 10,000 beneficiaries [95%CI], 0.54 [0.09 to 1.00]).Abbreviation: LVC, low-value care; PSA, prostate specific antigen.

Characteristics Associated with Costs of Low-Value Care
Compared to low-cost facilities, high-cost ones were more likely to have a large number of outpatient visits and patient stays of longer duration (P<.0001) (Table S2  [1.02 to 1.11]) were more likely to be in the high-cost group.Hospitals with a higher proportion of male patients were less likely to be in the high-cost group (aOR [95%CI], 0.97 [0.95 to 1.00]).Regions with higher combined comorbidity scores were more likely to be in the low-cost group (aOR [95%CI], 0.69 [0.52 to 0.92]), meaning that areas with poor or fair health tend to have lower costs associated with low-value care.Abbreviation: aOR, adjusted odds ratio.

Discussion
Low-value care is a critical issue in terms of patient safety and fiscal policy. 18Most previous studies on the prevalence and utilization patterns of low-value care were conducted in western countries.In the current study, we sought to extend their work to the situation at the hospital beneficiaries during the five-year follow-up period, resulting in annual losses of US$6.08 million.The two most common low-value services were PSA tests for men aged over 75 years old and screening for carotid artery disease in asymptomatic adults.This should not be surprising, given the broad base of clinicians ordering these examinations. 43The findings corroborate their inclusion in Choosing Wisely lists and Do Not Do recommendations as targets for interventions. 44,45Note that PSA tests and preoperative chest radiography are low-cost (<US$50) yet commonly-used examinations.These results are consistent with prior research which determined that low-cost high-volume services contribute significantly to health care spending. 6,46e observed increases in the utilization rate of low-value care did not match previous observations indicating no change or a decrease in use. 11,30,47This can perhaps be attributed to the fact that Taiwan's NHI provides easy access to health care with many beneficiaries engaging in doctor-shopping and undergoing overlapping examinations or treatments. 48,49cording to the published statistics, the average number of visits per capita for ambulatory care was 13.2 in Taiwan in 2019, which was significantly higher than in Canada (6.6), Australia (7.3), and Germany (9.8). 50Earlier work has demonstrated that the Choosing Wisely Campaign and payment reforms would help reduce low-value care; 4,51 nevertheless, little awareness has been raised among health care providers and policy makers across Asia.
Our findings at the hospital level are consistent with previous studies.Low-value care utilization appears to be less of a problem in local hospitals, private hospitals, and the hospitals with fewer outpatient visits.Mafi et al. 22 formerly reported that community-based practices were less likely to promulgate low-value care.We identified only a small number of regional factors that were predictive of low-value care utilization or the associated costs.
Badgery-Parker et al. 15 also reported that efforts to curb low-value care should be at the hospital level rather than the regional level.Note that the factors most strongly correlated with low-value care utilization were hospital service volumes and particularly ambulatory visits, indicating that larger institutions are more prone to unnecessary costs.These findings support preceding studies. 23,52Researchers have previously reported correlations between the utilization of low-value care and male patients, old age, and multiple comorbidities. 21,23,24,47 the current study, we found that hospitals with older patient populations and greater comorbidity burden were more likely to provide low-value care; moreover, the utilization were slightly higher in hospitals with a large proportion of male patients.It was very likely that sex-specific measures (e.g., PSA tests and X-ray bone densitometry) could bias our results; therefore, we conducted sensitivity analyses to clarify these relationships.Overall, we determined that the correlation between sex and low-value care remained significant.
We believe that our study will contribute valuable insights into low-value care within the Asian context.This study was subject to several limitations.First, the administrative claims data in this study lacked information related to clinical testing, which would have been valuable in defining low-value services more precisely.Note also that coding errors in large-scale databases may be inevitable.Nonetheless, we sought to minimize misclassification bias by applying procedural billing codes and adopting specific definitions available to facilitate the identification of low-value care.Second, this study focused on a single country that provides unrestricted access to medical services under a universal coverage National Health Insurance (NHI) program.As a result, our findings may not extrapolate to other healthcare systems, such as self-pay systems.The seven low-value services in this study are common among international recommendations and are easily defined in administrative data.Thus, our findings can be considered preliminary results relevant to the shaping of policies.Third, potential confounders at the physician level (e.g., specialty and patient panel size) were not addressed in this study; however, we considered the seniority of physicians at the hospital

Figure 1 .
Figure 1.Utilization and Associated Cost of Selected Low-Value Care Services, 2013-2017.Counts of episodes refers to unique incidences of service provision, associated costs only include the fee for specific examinations within each episode of low-value care services.Abbreviations: PSA, prostate specific antigen.

Figure 2 .
Figure 2. Correlation Between Utilization of Low-Value Care Services and Characteristics at the Hospital and Regional Levels.More utilization (on the right side) signifies that factors are associated with increased use of low-value care services, while less utilization (on the left side) indicates factors are correlated with reduced use of low-value service.*Estimates are the differences of LVC episodes per 10,000 beneficiaries from the reference group for each comparison group.Abbreviation: LVC, low-value care.

Figure 3 .
Figure 3. Correlation Between Utilization of Low-Value Care Services and Characteristics at the Hospital and Regional Levels, excluding sex-specific measures.Sex-specific LVC measures include PSA test and X-ray bone densitometry.Estimates are the differences of episodes per

Figure 4
Figure 4 displays the relationship between characteristics of interest and associated costs of low-value care.Costs were shown to increase yearly, with a corresponding increase in the adjusted odds ratio (aOR) from 1.36 (95% CI, 0.81 to 1.43) in 2015 to 1.73 (95% CI, 1.16 to 2.59) in 2017.A significantly positive correlation was observed between the volume of outpatient department visits and the cost associated with low-value care.Hospitals with a larger volume of outpatient visits (aOR [95%CI], 2.10 [1.26 to 3.49] for Q2-Q3, 2.88 [1.45 to 5.75] for ≥Q3) and those treated a higher proportion of older patients (aOR [95%CI], 1.06

Figure 4 .
Figure 4. Correlation Between High Cost on Low-Value Care Services and Characteristics at the Hospital and Regional Levels.Higher odds of high cost (displayed on the right side of the figure) suggest that the factor is associated with an increased probability of higher low-value care service cost, whereas lower odds of higher cost (shown on the left side) indicate a decreased probability of incurring higher costs for low-value care services.

Table 1 .
Baseline Characteristics of Beneficiaries, Hospitals and Regions of the Selected Low- INTERNATIONAL JOURNAL OF HEALTH POLICY AND MANAGEMENT (IJHPM) ONLINE ISSN: 2322-5939 JOURNAL HOMEPAGE: HTTPS://WWW.IJHPM.COM 9 INTERNATIONAL JOURNAL OF HEALTH POLICY AND MANAGEMENT (IJHPM) ONLINE ISSN: 2322-5939 JOURNAL HOMEPAGE: HTTPS://WWW.IJHPM.COM 10 The most common low-value intervention was the PSA test for men aged over 75 years old, which increased from 59.63 per 10,000 beneficiaries in 2013 (US$1.77 million) to 68.46 per 10,000 beneficiaries in 2017 (US$2.01million).The second most common intervention was screening for carotid artery disease in asymptomatic adults, which increased during the study period from 30.80 to 42.45 per 10,000 beneficiaries.The measure accounted for 36.14% of the total low-value care services, increasing during the study period from US$1.86 to US$2.57 million (P for trend<.001).Table2demonstrates the trend on the utilization and costs of low-value care.Other low-value care services, such as preoperative chest radiography, preoperative echocardiography, and preoperative stress test, were also shown to increase in the prevalence and corresponding costs.Only the utilization of X-ray bone densitometry decreased in the prevalence and costs.As for the composite measure, the utilization rate increased from 150.70 to 186.23 episodes per 10,000 beneficiaries (i.e., a 23.57% relative increase; P for trend=.001)with an increase in cost from US$5.40 to US$6.90 million (i.e., a relative change of 27.78%; P for trend=.001).

Table 2 .
Utilization and Associated Cost of Selected and Composite Low-Value Care Services, * The composite measure was created by summing the total utilization and associated cost of selected low-value care services.Abbreviations: PSA, prostate specific antigen; LVC, low-value care.Characteristics Associated with Low-Value Care UtilizationFigure2illustrates the association between characteristics of interest and the utilization of low-value care services.In general, low-value care appeared to increase over time; however, 40e results of this study demonstrate the extent of overuse, which also support the idea that the measurement of such services from several initiatives are applicable to a administrative database under a national health insurance program.Moreover, the comparison within and/or between hospitals provides preliminary information by which to formulate strategies to reduce costs.Our observations on utilization being associated with the volume of outpatient visits and the presence of multiple comorbidities indicates that future research should explore the causes of low-value care and potential remedies.In addition, despite the abundance of low-value lists, there is still a limited understanding of the extent of low-value care globally due to a lack of measurement, especially on Asia context.While historical measures of geographical variation in service utilization have provided insights into healthcare utilization patterns, they often do not account for the appropriateness of care.40Thisstudyanalyzed the nationwide patient-level data to evaluate the appropriateness of healthcare services based on patient characteristics and indications.Our results would enhance the understanding of low-value care in an Asian setting.Researchers have highlighted utilization patterns and potential contributors to hospital-level low-value care.The annual rate of low-value care was 166.19 episodes per 10,000