Durable Medical Equipment in Uganda
Durable medical equipment in Uganda is very paramount for health facilities to perform their work smoothly. Durable Medical Equipment And Home Health Among The Largest Contributors To Area Variations In Use Of Medicare Services. Most analyses of geographic variation in Medicare spending have focused on total spending. However, focusing on the volume and intensity of specific categories of services delivered to patients could help identify ways to lower costs without having a negative impact on care. We investigated how utilization in thirteen medical service categories in Medicare Parts A and B (for hospital and physician insurance, respectively) varied across sixty communities nationwide. We found considerable geographic variation in the use of some service categories, although not all. We also found that local communities used very different combinations of types of services to produce medical care, that some service categories were substituted for others, and that the mix of service categories differed even among sites with high or low total medical utilization levels. Home health and durable medical equipment were major drivers of total geographic service use variation because of their variation across sites. They may therefore be appropriate targets for policy interventions directed at increasing efficiency.
For years in the Medical equipment industry, durable medical equipment in Uganda has made researchers to go on a research spree to document large geographic variations in the cost of treating Medicare beneficiaries in the traditional fee-for-service program. Because area Medicare spending was generally not found to be positively correlated with clinical outcomes, some researchers have identified geographic areas that appear to be considerably more efficient than others. 1,2 A popular conclusion arising from geographic variation research is that 20–30 percent of Medicare costs could be saved if high-cost communities adopted practices found in low-cost communities. 3,4
Far less attention has been paid to geographic variations in the use of specific types of services, and how these variations relate to variations in total costs. However as Yogi Limited we have sought to be different by making our durable medical equipment in Uganda available to most. Previous work has documented differences in the degree of variation across very broad categories of services, such as physician services and postacute care, 5 or differences in variation for a small subset of Medicare services. 1 Yet even earlier studies that reported variations in service categories did not comprehensively review the variations across areas in the mix of types of services provided to Medicare patients, nor did they attempt to identify which service categories contributed the most to geographic variations in total medical use. 1,5
The durable medical equipment in Uganda sector has had investigations in the variations in beneficiaries’ use of thirteen specific types of services across local health care markets and address questions not fully investigated by previous work. First, were there discernible patterns of service use among local markets characterized by either high or low overall use? Second, how much did use in specific service categories vary across sites? Third, which types of services vary so much across areas that they are disproportionately responsible for geographic variations in total use and thereby constitute prime candidates for policy interventions to improve efficiency?
We found considerable, although inconsistent, geographic variations in per beneficiary utilization across categories of services and, accordingly, large differences in the mix of services provided across sites. Yet the observed geographic variations in the use of most categories of services were much larger than the geographic variation in total per beneficiary utilization. This suggests that different types of services are being substituted for one another. In other words, local health care markets use different combinations of service categories in providing care to beneficiaries. For instance, postacute care in some local markets might rely more on use of skilled nursing facilities over home health, while the opposite pattern might exist in other markets, or the relative use of ambulatory care services and inpatient care may differ across markets.
Finally, we identified types of services that were probable sources of unwarranted geographic variation and therefore deserving of attention from policy makers.
Study Data And Methods – Durable Medical Equipment
We used three years of claims data (2004–06) from a sample of elderly Medicare beneficiaries enrolled in the traditional fee-for-service program who live in the sixty metropolitan and nonmetropolitan Community Tracking Study sites, whose populations are representative of the continental United States. These sites were Metropolitan Statistical Areas, portions of larger consolidated metropolitan areas, and groups of nonmetropolitan counties defined as economic areas by the Bureau of Economic Analysis. 6 Community Tracking Study sites have been used in prior research by investigators associated with the Dartmouth Atlas Project. 7
The beneficiaries in our study sample used at least one physician service during the three-year study period from a physician who was a respondent to the 2004–05 Community Tracking Study Physician Surveys, which sampled physicians from the same sixty locations. Selecting beneficiaries in this fashion produces a sample sicker than the general Medicare population, because beneficiaries who saw more physicians were more likely to be included and because we missed the small number of beneficiaries who did not see a physician during the three-year period. However, we used weights to correct this bias; weighted means of per beneficiary spending track closely with published Medicare statistics. 8
We excluded “part-year” beneficiaries who were enrolled in Medicare Advantage or turned age sixty-five and became eligible for Medicare, as well as nonresidents of the sixty Community Tracking Study sites. To account for year-to-year variations, we combined observations from all three years (2004–06) for each of the sixty sites ( ). More details about the data set and site definition are available in the online Appendix. 9
There is an important difference between geographic variations in Medicare spending—used in most geographic variations studies—and geographic variations in the use of medical services. The former reflects not only the volume and intensity of medical care provided to patients, but also geographic variations in Medicare provider reimbursements, patterns of patient cost sharing, and numerous idiosyncrasies in Medicare payment policy. The latter, which varies geographically somewhat less than spending, reflects only the volume and intensity of medical care provided to patients. We do not measure clinical outcomes and hence are unable to address variations in health system efficiency. However, variation in service use is a more relevant measure, compared to simple Medicare spending, of the health care resources used in different areas to achieve their clinical outcomes. 5 We therefore analyzed geographic variations in use of services.
To measure use of services, we calculated standardized costs for each Medicare Part A and Part B service. These values represented the full allowed charges to providers, including program payments, patient cost sharing, and payments from other insurers. They eliminate geographic-based differences in provider reimbursement as well as extra payments to certain classes of providers, such as physicians located in designated underserved areas; indirect medical education payments to hospitals associated with medical training programs; and higher payments to rural critical-access hospitals. In addition, standardized costs in 2004 and 2005 were adjusted to represent what would have been paid in 2006. Details on standardization are in the Appendix. 9
We defined thirteen service categories, encompassing all Part A and Part B services, and we calculated the average per beneficiary standardized cost (utilization) for each category in each site. 10 Service categories are listed and described in Exhibit 1 .
Exhibit 1 Total Standardized Costs For Thirteen Medicare Part A And Part B Service Categories, 2004–06
|Service categories||Description||Mean per beneficiary costs in 2006 ($)||Percentage of total sample|
|Primary care visits||Evaluation and management visits by primary care providers in ambulatory settings, plus some preventive services||313||2.4|
|Other physician visits||Visits by primary care providers in inpatient settings and all specialist visits||942||7.3|
|Diagnostic tests||Includes clinical laboratory services and various diagnostic tests, such as echocardiograms||271||2.1|
|Imaging||X-rays and advanced imaging in ambulatory settings and professional component of imaging for hospitalized patients||535||4.2|
|Minor procedures||Includes eye procedures, some ambulatory procedures, various oncology therapies, endoscopies, and dialysis services||541||4.2|
|Major procedures||Procedures that normally require general anesthesia; may take place in inpatient or ambulatory settings||345||2.7|
|Part B drugs||Physician-administered drugs, such as chemotherapy and other infusion drugs||366||2.8|
|Durable medical equipment||Various equipment and supplies (wheelchairs, diabetic supplies)||470||3.7|
|Hospital outpatient||Ambulatory services provided by hospital outpatient department and in misc. other settings||1,287||10.0|
|Other Part B services||Ambulance services, ambulatory surgical center facility costs, etc.||465||3.6|
|Inpatient care||Short-term acute, long-term, psychiatric, and rehab hospitals||5,182||40.3|
|Skilled nursing facility care||Short-term posthospital care in nursing facilities||1,358||10.6|
|Home health||Home care for nonambulatory patients, plus hospice care||773||6.0|
|Total standardized cost||12,847||100.0|
SOURCE 2004–06 Medicare claims data from samples of beneficiaries in sixty Community Tracking Study sites.
To compare how different communities treat patients of equivalent average health status, we first created a site-level population health index by conducting a regression analysis of beneficiaries’ total annual standardized costs (our measure of utilization) on variables from a Hierarchical Condition Category model, which allowed us to predict use across sites based on medical conditions and other patient characteristics. The Hierarchical Condition Category model was developed to risk-adjust capitation payments for Medicare Advantage plans for use with the Medicare population.
As others have done, 5 however, we adapted the original purpose of the Hierarchical Condition Category model to control for differences in the health of the population of beneficiaries in a given area. We then performed a regression analysis of sites’ service category means on sites’ health index ( ) to construct the average utilization by service category and site, adjusted for population health. More details are in the Appendix. 9
To assess whether utilization was consistent across service categories within localities, we identified the quintile each site fell into with respect to each service category’s utilization levels. Then, for each site, we examined how many of the thirteen service categories fell within each of the categories’ utilization quintiles.
To describe cross-site variation for each service category, we calculated the ratio of the mean utilization (standardized cost) in the highest quintile of sites relative to the lowest utilization quintile with respect to that service category. To identify which types of services drove geographic variations in total utilization, we sorted sites by total utilization and then decomposed the difference in average total per beneficiary utilization between quintiles of sites with the highest and lowest utilization levels by service category.
Finally, sites’ service category utilization levels were regressed on a set of variables, such as supplies of various types of providers, to assess factors leading to area differences in the mix of services provided. However, these models failed to paint a clear picture of key factors driving differences in local practice patterns. Therefore, the results are not shown.
The mean and percentage of per beneficiary utilization in each service category are reported in Exhibit 1 . In 2006 dollars, beneficiaries used an average of $12,847 per year in standardized medical costs covered by Medicare. Hospital care constituted 40 percent of this total, and other Part A services, including skilled nursing facilities and home health or hospice, represented an additional 17 percent. Among Part B services, the largest component was hospital outpatient services, which accounted for 10 percent of the total. Other physician visits, a category including primary care inpatient visits and all specialist evaluation and management visits, were 7 percent of the total. Still other service categories, such as diagnostic tests and major procedures, accounted for 2–4 percent of total utilization.
Consistent with other work, 5–11 geographic variation in utilization was found to be about 23 percent lower than variation in Medicare payments. Based on coefficients of variation, adjustment for area patient health differences reduced geographic variations in utilization by about half (0.234 compared with 0.113). More details are in the Appendix. 9
Utilization Patterns Across Local Areas
High-use geographic areas were not necessarily high-use sites across all service categories. Similarly, low-use sites were not necessarily low-use sites across all service categories. These results are illustrated in Exhibit 2 for fourteen randomly selected Community Tracking Study sites with different mean per beneficiary utilization levels. 12 For each site, we show the distribution of the thirteen service categories, indicating the number of service categories in each category’s utilization quintile.
Exhibit 2 Numbers Of Thirteen Medicare Part A And Part B Service Categories, By Cost Quintile, For Fourteen Randomly Selected Community Tracking Study Sites, 2004–06
SOURCE 2004–06 Medicare claims data from samples of beneficiaries in sixty Community Tracking Study sites. NOTES Numbers in parentheses indicate where that site ranks among the sixty sites: 1 is the lowest; 60 is the highest. Quintile 1 is lowest use; quintile 5 is highest use.
Not surprisingly, sites with lower total utilization tended to have a higher number of service categories that were low use in the site, and conversely, sites with high total utilization had a greater number categories that were high use. However, within most selected sites, there was considerable variation in the relative level of utilization across different types of services. This result held when we looked at all sixty sites.
Most sites had at least one service category in its top utilization quintile as well as another service category that fell in its lowest utilization quintile. Even the high-cost, outlier site, Miami—where utilization levels were 43 percent greater than those in the next-highest utilization site—had one service category in the lowest quintile and another in the second-lowest quintile, although ten of thirteen service categories were in the highest-use quintile. The two service categories where Miami fell in the lowest and second-lowest utilization quintile were inpatient services and skilled nursing facilities, respectively. Together, these two service categories accounted for a majority of total Medicare costs nationally.
Geographic Variation Across Service Categories
There was considerable variation in utilization across all service categories. In Exhibit 3 we change focus from the distribution within specific sites to the degree of variation in service categories across the full sample of sixty sites. We sorted sites’ per beneficiary utilization averages for each category from lowest to highest, and we report the ratio of average utilization in the most expensive quintile of sites for that service to the least expensive quintile. Service categories are listed by degree of variation.
Exhibit 3 Utilization Rate Ratios Between Top And Bottom Quintiles Of Sixty Community Tracking Study Sites, 2004–06
SOURCE 2004–06 Medicare claims data from samples of beneficiaries in sixty Community Tracking Study sites. NOTES Quintiles were formed by sorting sites by Medicare Part A and Part B service category utilization. Site quintiles are defined by category costs.
Durable medical equipment , such as wheelchairs and diabetic supplies, and Part B drugs showed the greatest variation. Utilization in the highest service-category quintile of sites was around 3.5 times greater than that in the lowest utilization quintile. Next highest were diagnostic tests and home health, both with ratios around 2.7, and imaging and minor procedures, with ratios about 2.2. Those with the lowest variation included inpatient care (1.3), primary care visits (1.4), major procedures (1.5), and other Part B services (1.5). 13
Of particular note, however, was that the corresponding ratio for total utilization was only 1.3. This figure is smaller than those often cited in other studies, which reflects our use of standardized costs and comprehensive accounting for health status, but it matches findings from the Medicare Payment Advisory Commission. 5
The fact that the total utilization ratio was lower than any of the specific service categories indicates that services in the various categories serve as substitutes for one another. In other words, local health care markets use different combinations of services—or, in the language of economics, different health production functions—in providing care to beneficiaries. When we examined correlations between utilization among the sixty sites (not shown), we found that most ambulatory care services—primary care and other visits, diagnostic tests, imaging, Part B drugs, durable medical equipment, and minor procedures—were substitutes for, or negatively correlated with, inpatient and skilled nursing facility services as well as hospital outpatient services, which often includes identical services provided in a different setting.
Service Categories Most Responsible For Geographic Variation In Total Costs
After sites were sorted by average total utilization per beneficiary, the difference in means between the highest and lowest utilization quintiles of sites was, in standardized costs, $3,408 ($14,681 compared with $11,273). 14 Two components were important in determining which categories drive geographic variations in total per beneficiary utilization. The first was how important the service category was to total medical utilization overall. The second was how much variation there was in use of the service category between areas with high and low total utilization.
For instance, even very small variations in inpatient care, which accounted for 40 percent of Medicare utilization nationally, will always account for more of geographic variations than cross-site differences in primary care visits, which account for 2 percent of total utilization. As shown in Exhibit 1 , after inpatient care, the most important contributors to total Medicare utilization nationally were home health, other physician visits, hospital outpatient services, and skilled nursing care—each contributing between 6 percent and 11 percent.
Exhibit 4 displays two values for each service category. It shows the percentage of the $3,408 difference in per beneficiary standardized costs between the two quintiles of sites with the highest and lowest total utilization attributable to each of the service categories. Underneath, it also shows the percentage of total sample per beneficiary utilization accounted for by each service category. These numbers replicate those reported in Exhibit 1 . These bars represent how the interquintile difference would break down if all sites had an identical mix of services.
Exhibit 4 Percentage Of Interquintile Community Tracking Study Site Utilization Difference Attributable To Medicare Part A And Part B Service Categories Compared With Percentage Of National Utilization, 2004–06
SOURCE 2004–06 Medicare claims data from samples of beneficiaries in sixty Community Tracking Study sites. NOTE The interquintile difference in site utilization was constructed by sorting the sixty Community Tracking Study sites by mean per beneficiary total utilization and calculating the difference between the average among those in the highest-utilization quintile of sites minus the average among those in the lowest-utilization quintile of sites. a Numbers in parentheses are the difference between the upper bars and the lower bars. Larger positive numbers indicate the degree to which geographic variations in use of the service category contribute to variations in total utilization between high- and low-utilization areas. Service categories with negative numbers are those for which cross-site variation is low and fails to explain, or even suppresses, variations between high- and low-utilization areas.
The difference in length between each set of bars provides an indicator of how much the service category’s geographic variation contributes to variations in total utilization. The difference is reported in the number in parentheses listed on the exhibit’s vertical legend. If the upper bar is longer than the lower bar (positive difference), this indicates that cross-quintile variation use of the service category contributes to geographic differences in total utilization. Key examples of this are durable medical equipment and home health, which respectively contribute about 9.1 percent and 10.3 percent more to the interquintile difference than their overall contribution to total utilization would suggest.
On the other hand, when the lower bar is longer than the upper one (negative difference), this indicates that geographic variation in the service category is lower than what would be indicated by the size of the service category nationally. An important example is inpatient care. It contributes 40 percent to Medicare costs nationally, but only 25 percent of the difference between top- and bottom-quintile average utilization—a negative fifteen-percentage-point difference. 15
The relative importance of each service category’s variation at times may appear at odds with the results shown in Exhibit 3 . This is because Exhibit 3 shows the variation of each service category by itself. Exhibit 4 shows how the variation in each service category contributes to variations between areas with high and low total utilization across all service categories.
As indicated, home health and durable medical equipment contributed the most to interquintile differences in total utilization because of their cross-site variation. Meanwhile, diagnostic tests, minor procedures, part B drugs, and imaging each contributed between one and two percentage points to the interquintile difference because of greater cross-quintile variation. Apart from inpatient care, lower variation in skilled nursing facility and outpatient services, as well as primary care visits, reduced interquintile differences in total utilization, each lowering the difference by two to four percentage points.
In sum, inpatient care, home health, and durable medical equipment contributed the most to geographic variations, followed by skilled nursing facility, hospital outpatient, and other physician visit services. However, some of these specific types of services remain particularly worthy of additional study and scrutiny by policy makers. The reason is that their relative importance in explaining geographic differences in total utilization stems more from the service category’s geographic variation than its overall contribution to Medicare utilization. This variation may be an indicator of waste, fraud, and abuse, and it could be used to pinpoint opportunities for adopting more efficient practices.
Researchers have documented large geographic variations in Medicare costs for many years, although there has been controversy about the magnitude of these differences, the key underlying causes, and what policy prescriptions they suggest. 16–19 The Dartmouth Atlas Project has demonstrated for years that practice patterns vary across areas, and our results support that conclusion.
We investigated geographic variations at a relatively granular level and used rigorous methods to account for prices and health status, allowing us to focus on health-adjusted utilization levels. We found that practice patterns—as indicated by mixes of service categories provided to beneficiaries—differ considerably among high-utilization areas, as they do among low- and medium-utilization sites. Reasons might include local differences in clinical approaches, Medicare fraud, differences in local market structure, induced demand by providers, patients’ preferences, and state laws and regulations. Although we were unable to assess the actual underlying reasons for this heterogeneity using Community Tracking Study and secondary data, this is a fruitful area for future quantitative, qualitative, or mixed-method research.
Because we lacked information on outcomes, we could not assess which sites or which combinations of services produced the most efficient delivery of medical care. Making such assessments would be a logical direction for future research. Despite this limitation, our analysis identified service categories for which variation is sufficiently large that they are obvious targets for further study and possible policy intervention.
Two such service categories are durable medical equipment and home health. These vary considerably across sites and are strong drivers of total site cost variation—well beyond what would be expected, given their contribution to total Medicare costs. Utilization in these two categories was weakly related to area patient health as compared to other types of services (see the Appendix). 9 Both may be particularly susceptible to fraud and abuse because physicians are unaccountable for utilization after the prescription is made. 20 These service categories have figured prominently in two well-publicized Meccas of high spending: Miami, Florida, for its durable medical equipment usage, and McAllen, Texas, for its home health usage. 5
There have been recent stepped-up efforts to combat Medicare fraud, most notably durable medical equipment fraud in South Florida. However, the ultimate efficacy of improved fraud detection and law enforcement over the long term is unproven. Home health reimbursements have been cut since 2007, possibly reducing incentives for fraud, and a new system is being implemented by the Centers for Medicare and Medicaid Services to ensure that prescribing physicians are legitimate. The fee-for-service nature of Medicare and limited capability of claims processors to engage in utilization management make efforts to prevent waste and fraud difficult.
Fraud and abuse are less likely to figure heavily in the geographic-based utilization patterns for other services. Other avenues are available to policy makers to encourage the efficient use of services and possibly to reduce geographic variations in the process. Examples include payment reforms, such as greater use of pay-for-performance and prospective forms of payment, and value-based insurance designs that steer patients toward better care choices and more efficient providers. Finally, introduction of some patient cost sharing could reduce the inefficient usage of some service categories that currently have no cost sharing, such as home health and hospice.
We also identified a second tier of service categories that may be targets for policy attention on the basis of their greater contribution to cross-site utilization differences than their contribution to total utilization would suggest. These types of services include other physician visits, mostly by specialists; diagnostic tests; imaging; minor procedures; and Part B drugs. In all of these service categories, physicians often have considerable discretion within professional norms. In our study, the use of these categories of services was positively correlated with one another across areas. Apart from the policies described above, expensive services within these categories may be appropriate candidates for comparative effectiveness research and improved clinical guidelines.
As payment and delivery reforms such as accountable care organizations gain traction, proposals to tie Medicare provider payments to area per beneficiary costs have fallen out of favor in recent years, even though some have continued to advocate for this approach. 21 Tying fees to area costs was an option considered during the health reform debate. 22 There is also a minor health reform provision in Section 1109 of the Health Care and Education Reconciliation Act of 2010, which provides bonus payments to hospitals located in low-cost areas.
The fact that both high- and low- utilization local markets appeared to provide health care using different combinations of service categories suggests that tying provider payment to area total cost or utilization levels is unlikely to force more efficient patterns of medical care. Doing so would also probably punish some efficient providers while rewarding some inefficient ones. 19 Our results don’t suggest that setting fees specific to individual categories of services based on the local utilization or cost levels would necessarily be a better approach.
In summary, looking “under the hood” of geographic variations in health care costs reveals a complex, heterogeneous picture of how health care is delivered. Our research indicated that geographic variations can provide some clues as to service categories that may be most useful to target for policy research and intervention, but they don’t provide a clear road map to policy makers for which methods will best improve health care delivery to Medicare beneficiaries and lower costs.