Using MedPAR Data as a Measure of Urinary Tract Infection Rates: Implications for the Medicare Inpatient DRG Payment System

Abstract

A valuable metric of hospital performance is the rate of nosocomial infections, particularly urinary tract infections (UTIs). Current measurement techniques are expensive to administer and are not widely available. Determining a measurement index of nosocomial UTI incidence using Medicare Provider Analysis and Review (MedPAR) data to make recommendations may better align hospital payment with delivery of quality healthcare.

There is significant variation among hospitals’ calculated nosocomial UTI rates. In a sample hospital, the hospital received an estimated $675,000 in additional payments from Medicare due to payments for secondary nosocomial UTIs. The Comparative MedPAR Nosocomial UTI Index is a meaningful tool for determining nosocomial UTI rates as a measure of hospital quality. Additional improvements to the tool include incorporating risk factors based upon initial diagnosis, Major Diagnostic Category (MDC), and other diagnoses. Patients would benefit if the Centers for Medicare and Medicaid Services (CMS) discontinued the practice of paying hospitals for hospital-acquired infections, as this practice discourages adoption of infection-reducing initiatives.

Key Words: Nosocomial UTI, Quality Measurement, MedPAR Claims Analysis

Introduction

Healthcare regulators and payers are increasingly interested in having statistical tools to assess the relative performance of hospitals in treating illness and reducing preventable complications. Currently, the Medicare payment system provides pay for services rather than for results, with adjustments made annually to coincide with the underlying costs of services. Medicare’s Diagnosis-Related Group (DRG) payment system contains a disincentive for improvement because it provides additional payment for complications, even if those complications might have been prevented through quality enhancements. The Medicare Payment Advisory Commission (MedPAC) has promoted tying hospital payment to better quality.1 For the first time, the Centers for Medicare and Medicaid Services (CMS) proposed a system to pay hospitals a bonus of 0.4 percent higher than their standard payment based upon reporting quality criteria. This proposed move of paying a bonus based upon quality criteria would represent a departure from the current service-based system and would eventually provide an additional incentive to improve quality. This paper addresses utilizing existing MedPAR data to determine nosocomial urinary tract infection (UTI) rates at Medicare hospitals and assesses how this potentially preventable complicating condition may affect payment levels to these hospitals.

Background

An important aspect of paying for quality is measuring quality; however, most quality tools proposed for use with Medicare patients would be burdensome to hospitals. A valuable metric of hospital performance is the nosocomial infection rate, particularly the rate of UTIs. The urinary tract remains a significant site for hospital-acquired infections, with 66 percent to 86 percent of UTIs being associated with urinary catheterization.2 The prevention of UTIs represents a potentially rich opportunity to reduce the incidence of hospital-acquired infections.

Recent literature has indicated that it may be possible to use Medicare discharge data to obtain a meaningful measure of hospital quality performance. In their recent 11-state study, Needleman et al. discussed the applicability of Medicare discharge data as a substitute for overall hospital discharge data or ten adverse patient reactions, including UTIs. Needleman et al. discussed the correlation of Medicare data to non-Medicare data and demonstrated that Medicare data are a strong proxy for data from the entire population.3 While Needleman et al. established the validity of using Medicare inpatient data to measure hospital complications, they did not address the issue of which hospitals are doing well in the prevention of UTIs or the role of Medicare’s national payment system in providing incentives to reduce UTIs.

If a reliable metric could be developed, hospitals could take a number of steps to reduce their UTI rates among the Medicare patient population.4 Needleman et al. indicated through regression analysis that increased nurse-to-patient ratios resulted in fewer hospital-acquired UTIs.5 A number of studies have demonstrated that the use of silver-alloy coated urinary drainage catheters rather than standard urinary drainage catheters can lead to a reduction in UTIs.6, 7 Hospital training and safety practices have been linked to a reduction of nosocomial UTI rates.8 Hospitals could evaluate the potential benefit of different programs based upon cost and quality effectiveness and then choose accordingly.

Research Question and Hypothesis

Existing Medicare data may provide a resource for determining hospital-specific nosocomial UTI rates for Medicare hospitals and a foundation to assess Medicare payments for nosocomial UTI infections. Because Medicare’s current DRG payment system incorporates a disincentive for quality improvement related to complications, this paper examines whether nosocomial UTI rates vary significantly among Medicare hospitals and whether Medicare payments are shifted to higher-paying DRGs due solely to nosocomial UTIs.

Methods

The Medicare program is a national health insurance program intended for people aged 65 or older, some people under age 65 with disabilities, and people with end-stage renal disease. The national Medicare Provider Analysis and Review (MedPAR) file contains records of Medicare-covered inpatient discharges in the United States. Each patient record contains up to nine diagnosis codes, up to six procedure codes, claim costs and charges, the DRG, the length of stay, and many other admission-specific parameters. Each claim also contains the hospital’s provider number, making it possible to correlate diagnoses with providers. MedPAR tracks discharges rather than patients, making a distinct record available for each patient encounter. For example, if a patient is admitted and discharged in January and is again admitted and discharged in September, MedPAR will produce two records for this patient. This permits analysis based on hospital encounters rather than on patients. To determine a measure of the potential number of preventable UTIs, we examined the principal and secondary diagnosis fields in the MedPAR data for specific UTI diagnosis codes.

Our analysis encompassed all patients discharged during the fiscal year ending September 30, 2002, as reported in the MedPAR data set. As a first step, we identified all claims in which the patient had a UTI diagnosis in any of the nine diagnosis fields, using ICD-9-CM diagnosis codes. The codes that we tracked to UTI are Urinary tract infection (599.0) and Infection and inflammatory reaction due to indwelling urinary catheter (996.64).

Determining MedPAR-Derived Hospital-Specific UTI Rates

Our formula to calculate the UTI rate was developed in a number of steps. We began with the following basic formula:

Number of UTIs ÷Number of Discharges

As a simple measure of UTI rate, we could use the above calculation. While this calculation does provide the percentage of patients who had a diagnosis of UTI (for fiscal year 2002, this figure is 10.14 percent), we are most interested in using MedPAR data to provide a valuable metric of nosocomial UTI rates per hospital.

Although the ICD-9-CM diagnosis codes do not lend themselves to distinguishing between nosocomial infections and infections unrelated to the quality of care at the hospital, MedPAR does distinguish between diagnoses by creating nine separate diagnosis fields. These fields are numbered sequentially, with the first diagnosis also known as the principal diagnosis. The principal diagnosis is defined as the diagnosis that most immediately caused the patient to be hospitalized. For example, if a diabetic patient breaks a hip and has hip surgery, the fracture would be the cause of the hospitalization, and the diabetes, which may complicate treatment, would be a secondary diagnosis. Other analyses have adjusted the number of UTIs by eliminating claims in which the patient was admitted with a principal diagnosis of UTI.9

Therefore, as a second step, we eliminated from our analysis all claims in which the principal diagnosis was UTI. If the patient was admitted with a UTI, the hospital could not have prevented the UTI, making it obvious that the UTI was not nosocomial. Thus, these claims were eliminated from both the numerator and the denominator of the formula. We refined the analytical data set by eliminating claims that included either of the ICD-9-CM UTI diagnosis codes as the principal diagnosis. Our revised formula became  

([Number of Total UTIs]—[Number of Principal-Diagnosis UTIs]) ÷
([Number of Discharges]—[Number of Principal-Diagnosis UTIs])

For 2002, this resulted in a national figure of 8.69 percent.

Additionally, we reviewed patterns of secondary UTIs. We determined that a substantial number of these infections occurred in patients with diagnoses in Major Diagnostic Category (MDC) 18, Infectious and Parasitic Diseases. Of 311,891 patients in the 2002 data set assigned to DRGs within MDC 18, 101,428 (32.5 percent) had a secondary UTI. We concluded that hospitals with a high number of admissions within MDC 18 would have a bias toward less favorable UTI statistics and therefore MDC 18 UTI rates should be examined separately. There were no significant differences in secondary UTI rates among the other MDCs. Thus, to fairly compare hospital performance, we did not include claims associated with MDC 18 in establishing the comparative hospital-specific UTI metric. 

The principal diagnoses associated with the 101,428 secondary UTIs in MDC 18 that were separated are shown in Table 1. Table 1 includes the ICD-9 diagnosis code, its description, and the number of discharges having UTI as a secondary diagnosis, as reported in the 2002 MedPAR data. 

The comparative hospital-specific nosocomial UTI rate calculation then became

([Number of Discharges with Secondary But Not Principal UTI]—[Number of Secondary UTIs in MDC 18]) ÷
([Number of Total Discharges]—[Number of Principal UTIs]—[Number of Discharges in MDC 18])   

This calculation led to a national average nosocomial UTI rate of 8.10 percent for the fiscal year ending September 30, 2002.

Results

Of the 13,021,642 discharges in the 2002 MedPAR database, 12,502,700 were used in the nosocomial UTI rate calculations for each hospital. Approximately 200,000 were not used because the patient was admitted with a principal diagnosis of UTI, and approximately 310,000 were not used because the patient had a principal diagnosis in MDC 18, Infectious and Parasitic Diseases. Overall, there were 1,012,041 qualifying UTIs out of the 12,502,700 discharges, or a national rate of 8.10 percent.

Hospital Performance

The analysis showed that there is significant variation among hospitals’ nosocomial UTI rates. We obtained data from every hospital in the MedPAR database having at least one UTI reported in any of the diagnosis fields. Using this measure, which we call the Comparative MedPAR Nosocomial UTI Index, and examining just 1,000 of the largest hospitals in the United States (as measured by the total number of discharges reported in the data set), the secondary UTI rates for the 10 best performers ranged from 3.10 percent to 4.09 percent (Table 2), and the rates for the 10 worst performers ranged from 13.42 percent to 15.49 percent (Table 3).

Implications for Medicare Reimbursement

Medicare pays hospitals using a payment algorithm whereby similar diagnoses and procedures are lumped into DRGs. The predominant belief among healthcare professionals is that hospitals are paid a given amount under this system regardless of the quality of care that is provided. However, the DRG payment system has a quality disincentive contained within the payment formula. The payment system frequently provides hospitals with additional payment if the patient acquires a UTI. This occurs when the principal diagnosis could be assigned to one of two DRGs, where one DRG is for a condition with complication or comorbidity (CC) and the other is for the same condition without CC.

As an example, transurethral resection of prostate (TURP; ICD-9-CM procedure code 60.29) is performed on two patients without comorbid conditions, and the patients subsequently receive temporary urinary drainage catheters. The hypothetical first patient receives catheter A and is treated at a hospital that has implemented an infection-reducing clinical practice program. The patient does not develop a UTI. The hospital is paid for DRG 337, Transurethral prostatectomy without CC. The hypothetical second patient receives the less-expensive catheter B in a hospital that does not have an infection-reducing program, and the patient develops a UTI. The hospital codes the UTI as a secondary diagnosis and is paid for DRG 336 (Transurethral prostatectomy with CC) because of the CC. In this case, Medicare pays an extra 44 percent for the potentially preventable secondary UTI diagnosis.

Hospitals can employ a combination of technologies, expenditures, and staff training to reduce UTI rates. In Saint’s meta-analysis, he concluded that improved catheter technologies, such as silver-alloy urinary catheters that cost approximately $6 more than standard catheters, may be worth the additional cost.10 While we cannot say with certainty that the patient who had a UTI would not have had a UTI if a different catheter had been used, randomized controlled studies do show a marked reduction in the incidence of UTIs in patients who receive anti-infective (silver-alloy) Foley catheters.11, 12 Hospitals with higher nurse-to-staff ratios have a lower incidence of UTIs.13 Likewise, training programs have been shown to reduce the rate of nosocomial UTIs.14, 15, 16

Net Impact of DRG Shifts

To illustrate the net impact of DRG shifts on hospital reimbursement, we previously examined New York Hospital in detail.17 Out of 9,987 total discharges having DRGs for which with-CC and without-CC options exist, we identified 784 claims at this hospital with a secondary UTI diagnosis (2001 MedPAR data). For each claim, we calculated the net financial impact of the complication, assuming the UTI complication was responsible for moving payment from the lower-paying (without-CC) DRG to the higher-paying (with-CC) DRG. For example, there were 17 claims in DRG 1 with a secondary UTI. Had these UTIs not been present, New York Hospital would have received approximately $14,439 less for each of these 17 claims, assuming no other CCs existed. The total difference in payment is over $245,000. Continuing this exercise for each DRG pair and then adding up the payment differences for all DRGs resulted in UTI-related complication payments of $4.5 million for this hospital.

Because a patient may have more than one complication or comorbid condition, we further analyzed how many of these claims would shift from a lower-paying DRG to a higher-paying DRG even if a UTI was not on the claim. As we had to examine each claim individually to make this assessment, we examined claims for three randomly selected DRGs. An analysis of New York Hospital’s specific claims for three DRGs is detailed in Table 4. This table lists the DRG, the number of discharges with a secondary UTI, the number of discharges with a CC that is due to the secondary UTI, the percentage of discharges with a CC due to the secondary UTI, and the additional payment as a result of the CC due to the secondary UTI.

In this sample, the number of claims affected (i.e., claims for which the DRG shift was attributed solely to the secondary UTI) was reduced to 31 percent of the original number, and payment was reduced to 16 percent of the original potential dollar impact. Even if only 15 percent of the original claims shifted due to this change, a payment shift of $675,000 is a huge disincentive to investing in UTI reduction programs. By way of comparison, the cost to the hospital of converting to a more-expensive urinary drainage catheter would be about $50,000 for this group of 9,987 patients. If the results obtained in the anti-infective (silver-alloy) Foley catheter clinical studies followed the sample modeled results for this hospital, New York Hospital would stand to lose approximately $300,000 in reimbursement in addition to having to pay for the more-expensive urinary drainage catheters.

Discussion

The existing payment disincentive scenario works against the spirit of the DRG system. Hospitals’ payments for identical patients receiving identical procedures should be the same. Hospitals could then base decisions on technology acquisition and quality improvement programs solely on the analysis of costs to be averted rather than factoring in lost revenue. Providing hospitals with incremental payment for nosocomial infections, especially when the infection might be preventable, is not good policy. This is particularly true for UTI, an often-preventable complication. Hospitals should have every financial incentive to improve patient care by making every possible effort to reduce the incidence of hospital-acquired infections.

CMS announced that it intends to provide a bonus for superior outcomes to hospitals based upon hospital performance measures. While it is advisable to measure nosocomial infection rates to influence hospitals’ behavior by tying a portion of their payment directly to adverse reactions and quality, it is more urgent to remove the direct payments for adverse events under the Medicare DRG system. In the case of New York Hospital, total payments increased by an estimated $675,000 (for the 2001 fiscal year) because of higher payments for DRGs with complications or comorbidity. Such a financial system discourages the consideration and implementation of infection-reducing strategies. The UTI disincentive could be removed from the current system by excluding ICD-9-CM diagnosis codes 599.0 and 996.64 as complications in CMS’s CC list. CMS should reconsider the advisability of a CC list in the inpatient payment system at all.

The measurement index described in this report is easily duplicated on a real-time basis. It provides a simple measurement for comparing hospital performance and evaluating outcomes for patients in the Medicare system.

The measurement does depend critically upon hospitals using similar criteria to code for UTIs. It is possible that the UTI diagnosis codes may be omitted from claims due to field limitation constraints or for nefarious reasons, such as to affect quality measurements. Even when coding is consistent, our analysis, like other analyses using national data sets, is limited in that we did not and could not distinguish between secondary diagnoses that were present upon admission and those that appeared after admission.18, 19, 20, 21 Lawthers et al. indicated that in 13 percent of surgical cases, the condition represented by a complicating condition (not just UTI) in their Complications Screening Program (CSP) was judged to be present upon admission and therefore not due to the patient’s hospitalization.22 The calculated UTI rate still may overstate the true number of nosocomial UTIs. The process does typically remove readmissions for recurrence of a prior UTI, as UTI is most frequently the principal diagnosis for such readmissions. Therefore, these patients are not included.

The inclusion of the date of onset of a complicating condition would be helpful in differentiating secondary diagnoses present upon admission from diagnoses acquired during the hospital stay. Our efforts to include procedures such as catheter placement in our formula were not successful, as hospitals do not typically include this procedure on the bill and it is not one of the six procedures reported on the claim. Ultimately, using data on hospitals’ acquisition of specific drugs and number of urinary drainage catheters purchased may be helpful in developing more accurate estimates.

Conclusion

This paper discusses a new tool for measuring nosocomial UTI rates as a possible indicator of hospital quality. Hospitals showed marked differences in UTI rates. Since it is important to know whether hospitals use equivalent approaches to coding, additional studies comparing actual UTI rates to those derived from MedPAR data warrant further study. Additional improvements in the MedPAR-derived tool could be achieved by incorporating risk factors based upon initial diagnosis, MDC, other diagnoses, age, and source of admission. Analyzing an age cohort (e.g., all patients over 85 years of age) may also provide additional insight, particularly as the population continues to live longer. The UTI measurement system described here provides a meaningful comparative measure of hospitals’ performance in preventing infections. Comparison between hospitals of UTI rates in patients with MDC 18 diagnoses may provide additional insight into the potential to reduce nosocomial UTI rates for these patients as well. Finally, patients would benefit if CMS discontinued the practice of paying hospitals for hospital-acquired urinary tract infections through the presence of CC codes in the DRG payment system. Such a system discourages adoption of infection-reducing initiatives in America’s hospitals. 

Jerry Stringham is president of Medical Technology Partners, Inc. in Rockville, MD. Nancy Young is a director at Medical Technology Partners, Inc. in Rockville, MD.

Notes

1. Medicare Payment Advisory Commission. “Report to the Congress: Medicare Payment Policy.” March 2003. Available at the MedPAC Web site at www.medpac.gov/publications/congressional_reports/Mar04_Entire_report.pdf. Accessed February 15, 2004.
2.  Wong, E. S. “Guidelines in Preventing Urinary Tract Infections.” Date unknown. Available at the CDC Web site at www.cdc.gov/ncidod/hip/Guide/uritract.htm. Accessed February 15, 2004.
3. Needleman, J. P., S. Buerhaus, M. Mattke, et al. “Measuring Hospital Quality: Can Medicare Data Substitute for All Payer Data?” Health Services Research 38, no. 6, pt. 1 (2003): 1487–1508.
4. Wong, E. S. “Guidelines in Preventing Urinary Tract Infections.”  
5. Needleman, J. P., S. Buerhaus, M. Mattke, et al. “Measuring Hospital Quality: Can Medicare Data Substitute for All Payer Data?”
6.  Saint, S., J. G. Elmore, S. D. Sullivan, et al. “The Efficacy of Silver Alloy-Coated Urinary Catheters in Preventing Urinary Tract Infection: A Meta-Analysis.” American Journal of Medicine 105 (1998): 236-41.
7.  Karchmer, T. B., E. T. Giannetta, C. A. Muto, et al. “A Randomized Crossover Study of Silver-Coated Urinary Catheters in Hospitalized Patients.” Archives of Internal Medicine 160 (2000): 3294–98.
8. Wong, E. S. “Guidelines in Preventing Urinary Tract Infections.”  
9. Needleman, J. P., S. Buerhaus, M. Mattke, et al. “Nurse-Staffing Levels and Quality of Care in Hospitals.” New England Journal of Medicine 346, no. 22 (2002): 1415-22.
10. Saint, S., J. G. Elmore, S. D. Sullivan, et al. “The Efficacy of Silver Alloy-Coated Urinary Catheters in Preventing Urinary Tract Infection: A Meta-Analysis.”  
11. Ibid.
12. Karchmer, T. B., E. T. Giannetta, C. A. Muto, et al. “A Randomized Crossover Study of Silver-Coated Urinary Catheters in Hospitalized Patients.”
13. Needleman, J. P., S. Buerhaus, M. Mattke, et al. “Nurse-Staffing Levels and Quality of Care in Hospitals.”
14. Wong, E. S. “Guidelines in Preventing Urinary Tract Infections.”  
15. Burke, J. P., and D. K. Riley. “Nosocomial Urinary Tract Infections.” In C. G. Mayhall (Editor), Hospital Epidemiology and Infection Control. Philadelphia: Lippincott Williams & Wilkins, 1996.
16. Vance, J. “Diagnosing & Managing Urinary Tract Infections: Myths, Mysteries, & Realities.” October 2002. Available at the AMDA Web site at www.amda.com/caring/october2002/utis.htm. Accessed February 15, 2004.
17. Stringham, J. M. “UTI-CC DRG Reassignment Report.” Personal correspondence, Centers for Medicare and Medicaid Services, October 2003.
18. Lawthers, A. G., E. P. McCarthy, R. B. Davis, et al. “Identification of In-Hospital Complications from Claims Data.” Medical Care 38, no. 8 (2000): 785–95.
19. Geraci, J. M. “In-Hospital Complication Occurrence as a Screen for Quality-of-Care Problems: What’s Next?” Medical Care 38, no. 8 (2000): 777–80.
20. Geraci, J. M., C. M. Ashton, D. H. Kuykendall, et al. “International Classification of Disease, 9th Revised, Clinical Modification Codes in Discharge Abstracts Are Poor Measures of Complication Occurrence in Medical Inpatients.” Medical Care 35, no. 6 (1997): 589–602.
21. Weingart, S. N., L. I. Iezzoni, R. B. Davis, et al. “Use of Administrative Data to Find Substandard Care: Validation of the Complications Screening Program.” Medical Care 38, no. 8 (2000): 796-806.
22. Lawthers, A. G., E. P. McCarthy, R. B. Davis, et al. “Identification of In-Hospital Complications from Claims Data.” 

Article citation: Perspectives in Health Information Management 2; 12, Fall 2005

Printer friendly version of this article

Posted in:

Leave a Reply