The Hospital-Acquired Conditions (HAC) reduction program: using cranberry treatment to reduce catheter-associated urinary tract infections and avoid Medicare payment reduction penalties

Abstract

Objective: The Affordable Care Act (ACA) established the Hospital-Acquired Condition (HAC) Reduction Program. The Centers for Medicare and Medicaid Services (CMS) established a total HAC scoring methodology to rank hospitals based upon their HAC performance. Hospitals that rank in the lowest quartile based on their HAC score are subject to a 1% reduction in their total Medicare reimbursements. In FY 2017, 769 hospitals incurred payment reductions totaling $430 million. This study analyzes how improvements in the rate of catheter-associated urinary tract infections (CAUTI), based on the implementation of a cranberry-treatment regimen, impact hospitals’ HAC scores and likelihood of avoiding the Medicare-reimbursement penalty.

Methods: A simulation model is developed and implemented using public data from the CMS’ Hospital Compare website to determine how hospitals’ unilateral and simultaneous adoption of cranberry to improve CAUTI outcomes can affect HAC scores and the likelihood of a hospital incurring the Medicare payment reduction, given results on cranberry effectiveness in preventing CAUTI based on scientific trials. The simulation framework can be adapted to consider other initiatives to improve hospitals’ HAC scores.

Results: Nearly all simulated hospitals improved their overall HAC score by adopting cranberry as a CAUTI preventative, assuming mean effectiveness from scientific trials. Many hospitals with HAC scores in the lowest quartile of the HAC-score distribution and subject to Medicare reimbursement reductions can improve their scores sufficiently through adopting a cranberry-treatment regimen to avoid payment reduction.

Limitations: The study was unable to replicate exactly the data used by CMS to establish HAC scores for FY 2018. The study assumes that hospitals subject to the Medicare payment reduction were not using cranberry as a prophylactic treatment for their catheterized patients, but is unable to confirm that this is true in all cases. The study also assumes that hospitalized catheter patients would be able to consume cranberry in either juice or capsule form, but this may not be true in 100% of cases.

Conclusion: Most hospitals can improve their HAC scores and many can avoid Medicare reimbursement reductions if they are able to attain a percentage reduction in CAUTI comparable to that documented for cranberry-treatment regimes in the existing literature.

Keywords:

• Affordable Care Act
• CAUTI
• cranberry
• healthcare costs
• hospital-acquired condition
• Medicare payment reduction
• simulation

Introduction

Catheter-associated urinary tract infections (CAUTI) are a common and important healthcare-associated infection (HAI). The Centers for Disease Control and Prevention (CDC) report that urinary tract infections (UTIs) are the most common type of HAI, and ∼75% of UTIs are associated with the use of a urinary catheter¹. More than a quarter of all hospital inpatients may have a urinary catheter placed during their stay². CAUTI can result in increased length of hospital stay, patient discomfort, additional healthcare costs, and in some cases mortality.

Achieving progress in reducing CAUTI has been difficult. The CDC in its HAI Progress Report found no overall improvement in CAUTI between 2009–2014³. A meta-analysis of short-term catheterized patients concluded that the use of antibiotics at the time the catheter was removed reduces UTI incidence by ∼50%⁴. Thus, standard antibiotic treatments alone are unable to eliminate the risk of UTI after catheterization and present risks of complications including severe diarrhea. The unmet goal of substantially reducing CAUTI has led many to consider cranberries as a dietary supplement to prevent UTIs among catheterized patients. Compounds in cranberries prevent Escherichia coli, the most common cause of the infection, from adhering to cells in the urinary tract.

The Patient Protection and Affordable Care Act (ACA) of 2010 contains provisions to address hospital-acquired healthcare issues, including CAUTI. Section 3008 of the ACA established the Hospital-Acquired Condition (HAC) Reduction Program (HAC-RP). In rulemaking to implement the HAC-RP, the Centers for Medicare and Medicaid Services (CMS) established a scoring methodology to rank hospitals based upon their HAC performance. Beginning in FY 2015, the HAC-RP required the Secretary of Health and Human Services to reduce Medicare reimbursements to hospitals ranking in the lowest quartile based on their HAC performance. In FY 2017, 769 hospitals were assessed a 1% Medicare reimbursement reduction that totaled $430 million⁵.

The HAC-scoring methodology has been revised and updated frequently over the life of the HAC-RP. Under rules in place for FY 2018, CAUTI is one of five hospital-associated infections (HAIs) that jointly comprise Domain 2 of the HAC-scoring methodology and share an 85% weight in the total score⁶. The other components of Domain 2 are Central Line-Associated Bloodstream Infection (CLABSI), Surgical Site Infection (SSI)–colon and hysterectomy, Methicillin-resistant Staphylococcus aureus (MRSA) bacteremia, and Clostridium Difficile Infection (CDI). Domain 1, with 15% weight in the total HAC score, consists of a hospital’s Recalibrated Patient Safety Indicator (PSI) measure known as Recalibrated PSI 90 Composite, an aggregate of 10 patient-safety indicators⁷.

In this paper, we utilize the clinical research supporting reduction in the incidence of UTI and, in particular, CAUTI among patients treated with cranberry to simulate how a cranberry-treatment protocol for CAUTI patients can impact hospitals’ HAC scores. We are interested in particular in hospitals presently incurring a Medicare reimbursement penalty and whether they can, through introducing a cranberry-treatment protocol for their catheterized patients, improve their overall HAC scores sufficiently to avoid incurring the payment-reduction penalty, which averaged $559,000 per penalized hospital in FY 2017.

Although our analysis focuses on the CMS reimbursement penalty, we note that hospitals also incur additional costs for treating CAUTI cases. These costs are not reimbursed by Medicare or Medicaid under the FY 2009 Inpatient Prospective Payment System Final Rule, and reducing them represents an additional financial benefit to hospitals from reduced rates of CAUTI.

Results from the simulation analysis show that cranberry treatment at mean effectiveness based on the scientific evidence improves the overall HAC score for nearly all hospitals. Many hospitals with HAC scores in the lowest quartile can improve their scores sufficiently through adopting cranberry to avoid penalty. A cranberry treatment regimen may also be part of a multifaceted approach for hospitals to improve HAC scores and avoid the Medicare reimbursement penalty.

Clinical research on cranberry and UTI

A number of clinical trials have been conducted to test the efficacy of cranberry on UTI prevention, with most focusing on women with a history of UTIs. Table 1 contains a summary of results from these trials. Most trials that measured the incidence of UTI among treatment and control groups have found a cranberry treatment regimen to be effective in reducing the incidence of UTIs relative to a control group in a relatively consistent range of 40–67%^8–14. A minority of studies^15–19, however, have found no significant impact of cranberry treatment. For the most part these studies have focused on narrow population segments—women aged 65+ in nursing homes¹⁵, children with neurogenic bladder^16,17, or elderly female hip-fracture patients, provided treatment for a limited period of time, and focused not on incidence of UTI per se but, rather, presence and frequency of bacteriuria in treatment and control groups during and well after treatment.

Table 1. Summary of literature of efficacy of cranberry in treating UTI-related conditions.

Reference no.	Authors	Year	Patient population	Patient count	Intervention	Primary outcome	Outcome
Studies finding a significant impact for cranberry
[8]	Avorn et al.	1994	Asymptomatic elderly women	153	10.14 oz (300 mL) cranberry juice cocktail daily for 6 months	Presence of bacteriuria with pyuria in urine	42% reduction in the presence of bacteriuria with pyuria in urine
[9]	Walker et al.	1997	Women with recurrent UTI	19	Cranberry capsule (with 400 mg cranberry solids)	Episodes of UTI	60% reduction in UTI episodes
[10]	Kontiokari et al.	2001	Women with urinary tract infection caused by Escherichia coli	150	1.7 oz (50 ml) of cranberry-lingonberry juice concentrate daily for 6 months	First recurrence of symptomatic urinary tract infection	56% reduction in the rate of the first occurrence of UTI during 1 year
[11]	Stothers	2002	Sexually active women with a history of UTI (21–72 years)	150	25.5 oz cranberry juice daily or cranberry tablets	Symptomatic UTI within 1 year	38% reduction in UTI
[12]	Barnoiu et al.	2015	Patients with double J catheter	62	120 mg cranberry extract added to routine prophylactic therapy	Episodes of UTI	66.7% reduction
[13]	Foxman et al.	2015	Gynecological surgery patients involving urinary catheterization	160	Cranberry juice capsules (equivalent of 16 oz of juice) for 6 weeks	Clinically diagnosed and treated UTI (CAUTI) with or without positive urine culture within 6 weeks	50% reduction in UTI after catheterization
[14]	Maki et al.	2016	Women with a history of a recent UTI (20–70 years)	373	8.12 oz (240-mL) of cranberry beverage daily	UTIs in 24 weeks	41% reduction in UTI
Studies finding no significant impact for cranberry
[15]	Juthani-Mehta et al.	2016	Women nursing home residents (65+ years)	185	Two cranberry capsules containing 72 mg total of pro-anthocyanidin (equivalent to 20 oz juice)	Presence of bacteriuria plus pyuria assessed every 2 months for 1 year and symptomatic UTIs	No significant difference
[16]	Schlanger et al.	1999	Children with neurogenic bladder receiving intermittent catheterization	15	2 oz cranberry concentrate (300 mL cranberry juice cocktail equivalent)	Frequency of bacteriuria	No significant difference
[17]	Foda et al.	1995	Children with neurogenic bladder receiving intermittent catheterization	40	15 mL/kg/day of cranberry cocktail for 6 months	Number of negative culture months	No significant difference
[18]	Gunnarsson et al.	2017	Women with hip fracture (60+ years)	227	Cranberry capsules (550 mg) daily for 5 days post-operation	Urine cultures obtained at 5 and 14 days post-operation	No significant difference
[19]	Letouzey et al.	2017	Women undergoing pelvic surgery (18+ years)	272	36 mg cranberry once daily for 10 days	Post-operative bacteriuria (positive urine culture) with 15 and 40 days	No significant difference

Of most direct relevance to the present study are trials that focused on cranberry treatment for catheterized patients. To locate the universe of such studies we searched PubMed for articles published between 1990 and 2017 that contained the terms “cranberry”, “urinary tract infection”, and all possible forms and spellings pertaining to use of a catheter. We further limited the search to studies in humans and publications in English. This search yielded eight studies, one of which focused on rats and was, thus, excluded. Of the remaining seven publications, three (Barnoiu et al.¹², Foxman et al.¹³, and Gunnarsson et al.¹⁸ provided estimates of the effectiveness of cranberry in reducing CAUTI in clinical trials; two provided reviews of possible interventions for preventing CAUTI, including cranberries; one discussed the impact of cranberries on the swarming behavior of the bacteria, but did not provide estimates of efficacy in preventing CAUTI²⁰; and another²¹ provided a short and favorable commentary on the Foxman et al.¹³ trial.

Foxman et al.¹³ studied the effects of cranberry on UTIs among 160 women undergoing elective gynecological surgery with urinary catheterization. Participants ranged from 23–88 years of age. For the cranberry group, the mean age was 56, with a standard deviation of ±12.5. For the placebo group, the mean age was 56, with a standard deviation of ±14.3. The treatment group received four capsules of cranberry, equivalent to 16 oz of cranberry juice, for a 6-week period. The rate of UTI was significantly lower by 50% (15/80) in the treatment group compared to the placebo group (30/80).

Barnoiu et al.¹² studied 62 patients with a double J catheter, with 31 treatment participants receiving 120 mg of cranberry extract in addition to routine prophylactic therapy. For the cranberry group the mean age was 50.8, with a standard deviation of ±14.7, and for the placebo group the mean age was 48.9, with a standard deviation of ±12.8. The percentage of positive UTI cultures was significantly lower, 12.9%, in the treatment group compared to the group that did not receive cranberry, 38.7% (p = 0.04)—an improvement by two-thirds in treatment relative to control.

Gunnarsson et al.¹⁸ studied women over age 60 with hip fractures. Although 227 patients entered the study, several factors caused the authors’ usable sample size to be reduced substantially below the initial goal of 200 across treatment and placebo groups. Among 65 participants—29 in placebo, 36 in treatment—who took their capsules consistently and whose trial was not contaminated by use of antibiotics other than the prophylaxis, incidence of bacteriuria at 5 days post-op was ∼38% among control and 22% among treatment—a 42% decrease in incidence (p = 0.166). Incidence at either day 5 or day 14 post-op was 44% among control and 29% among treatment—a 34% decrease in incidence (p = 0.236). These results are at the lower end of the 40–67% efficacy rate noted previously, but are not statistically significant, largely due to the shrinking of the authors’ sample population.

Based on the larger usable study population in Foxman et al.¹³ compared to Barnoiu et al.¹² or Gunnarsson et al.¹⁸ and their mid-range result regarding cranberry efficacy in preventing CAUTI relative to the efficacy found in Barnoiu et al.¹² and Gunnarsson et al.¹⁸, the baseline analysis assumes the 50% efficacy found by Foxman et al.¹³. In sensitivity analysis, we consider both a higher and lower level of effectiveness for cranberry, based on the Foxman et al.¹³ 95% confidence bounds.

The HAC-RP

As noted, CMS computes a hospital’s total HAC score as the weighted sum of its Domain 1 (15%) and Domain 2 (85%) scores. The Domain 1 score is based on the points assigned for a hospital’s PSI 90 Composite Index Value. The Domain 2 score is computed for FY 2018 as the average of the hospital’s scores for CAUTI, CLABSI, SSI, MRSA, and CDI.

The CDC calculates each hospital’s standardized infection ratios (SIRs) for each Domain 2 measure for which the hospital provides data. The SIRs are the ratio of observed-to-predicted numbers of HAIs. The CDC calculates the number of predicted HAIs for each patient-care location within a hospital, based on its National Healthcare Safety Network (NHSN) database. HAI for each Domain 2 measure are reported to NHSM across all applicable Intensive Care Unit locations for participating hospitals. A hospital’s total predicted number of HAIs for FY 2018 was the sum of the predicted number of HAIs across medical wards, surgical wards, medical-surgical wards, and ICU locations within the hospital. CLABSI, CAUTI, SSI, MRSA, and CDI HAIs are calculated using chart-abstracted surveillance data reported to NHSN for infections occurring January 1, 2015–December 31, 2016²².

A hospital often lacks one or more of the component scores for Domain 2, and a few hospitals lack a Domain 1 score. This occurs if a hospital lacks sufficient data to compute a score, or it has received an exception from CMS for a particular HAI. Hospitals that do not submit data for an HAI and have not received an exception are assigned the maximum score for that measure. Importantly, when component scores are missing, the scores that are available acquire the weights for the missing scores. Thus, for hospitals that have a Domain 1 score, the percentage weight in the total HAC score assigned to any one of the five Domain 2 HAI is 85/N, where N∈{1,2,3,4,5} is the number of available Domain 2 scores for the hospital. For example, suppose a hospital has a Domain 1 score and scores for CAUTI, CLABSI, and CDI in Domain 2, but no scores for MRSA and SSI, then the percentage weight assigned to the CAUTI, CLABSI, and CDI scores in the HAC score formula is 85/3 = 28.33.

If a hospital lacks a Domain 1 score, then the probability weight for each Domain 2 score is 100/N. If a hospital lacks data for all five Domain 2 HAIs, but has a Domain 1 score, then the Domain 1 score has a 100% weight in the total HAC score. Hospitals that don’t have either a Domain 1 or Domain 2 score do not receive a HAC score.

CMS is fundamentally changing the HAC scoring methodology beginning in FY 2018. Prior to that time, hospitals received an integer score ranging from 1–10 for each HAI based upon the hospital’s decile rank for that HAI among all hospitals with a score, with higher scores indicating poorer performance. This scoring system caused the HAC-RP to have several idiosyncratic features. Many hospitals, for example, shared the same total HAC score, and many shared the HAC score that defined the 75^th percentile cut-off (e.g. 134 hospitals had the 6.75 HAC score that included the 75^th percentile in FY 2016), meaning that, in practice, fewer than 25% of hospitals received the reimbursement penalty, since it applied only to hospitals with scores strictly above the 75^th percentile. It also meant that improved relative performance for any of the HAIs would not improve a hospital’s overall HAC score, unless the improvement caused the hospital to advance at least one decile in the rankings.

For FY 2018, CMS adopted a Winsorized z score methodology to replace the decile-scoring methodology. For Domain 2, Winsorization begins by constructing the distribution across hospitals of SIRs for each HAI in Domain 2. Winsorization seeks to avoid the influence of outliers. Thus, scores that lie in the lower 5% tail are assigned the value of the distribution at the 5^th percentile, and scores in the upper 5% tail are assigned the value of the distribution at the 95^th percentile. SIR scores within the 5–95 percentiles are unaffected by Winsorization.

In the next step these adjusted SIRs are normalized by subtracting the mean Winsorized SIR and dividing that difference by the standard deviation of the Winsorized scores. Thus, for example, if hospital i’s Winsorized CAUTI SIR is X_i, and the mean Winsorized score is where N represents all eligible hospitals with a measure score for CAUTI, then hospital i’s Winsorized z score is where s_X denotes the standard deviation of the Winsorized CAUTI SIR scores computed in the usual manner. A hospital’s overall Domain 2 score is the average of its z_i across all applicable HAI for which the hospital has a score. Hospitals’ PSI 90 scores are Winsorized and normalized in the same manner to construct the Domain 1 score.

After determining each hospital’s HAC score in this manner, CMS constructs a distribution of these scores across all hospitals to establish the score, H^*, associated with the 75^th percentile cut-off; hospitals with HAC scores greater than H^* are subject to the payment reduction. In FY 2018, H^*= 0.3687. The simulation model described in the next section seeks to replicate this ex ante distribution of HAC scores to the extent possible, based upon available data and then to examine how individual scores and the distribution of scores are impacted under alternative cranberry-adoption scenarios.

Methods and data

The simulation analysis focused on hospitals with CAUTI scores and with total HAC scores in the highest 25^th percentile and, thus, subject to the Medicare payment reduction. The baseline simulation involved constructing synthetic hospitals by drawing PSI 90 (Domain 1) and CLABSI, CAUTI, SSI, MRSA, and CDI (Domain 2) scores from distributions created from the data for hospitals in each of five percentage point ranges or ventiles (75–80, 80–85, 85–90, 90–95, 95–100) for the lowest performing quartile of hospitals for the FY 2018 (October 2017–September 2018) program.

Take, for example, the 75–80 percentile ventile. For each of these hospitals we collected data on the CAUTI score and the PSI 90, CLABSI, SSI, MRSA, and CDI scores (if available), and used these scores to generate distributions of the six scores that are characteristic of hospitals in the 75–80 percentile ventile. The same procedure was used to create distributions of scores from hospitals in the other four ventiles comprising the lowest quartile.

We then created 1,000 synthetic hospitals for each of the five ventiles by randomly drawing with replacement a value for each of the six performance measures comprising the HAC score. Each synthetic hospital’s HAC score was then computed based on the CMS formula for FY 2018, as described in the previous section.

Data used in the simulation are from the CMS Hospital Compare website. We were unable to replicate exactly the data used by CMS to construct the FY 2018 scores and hospital rankings because CMS has not yet released all of the data for the FY 2018 program. Recalibrated PSI 90 scores for the actual FY 2018 program are based on data from July 2014–September 2015, whereas presently PSI 90 data are only available for the FY 2017 program—the 2-year period July 2013–June 2015. The PSI measures also differ somewhat between the FY 2017 and 2018 programs. PSI 90 composite claims for FY 2018 are based on PSI 03, 06, 08, and 09–15, whereas for FY 2017 the PSI 90 composite included PSI 03–08 and 12–15.

Domain 2 measures for the actual FY 2018 program are based on data for the 2-year period from January 2015–December 2016, whereas the Domain 2 data presently available from CMS and used in the simulation are for the 1-year period October 2015–September 2016. Given that the most recent data available are associated with the FY 2017 program, the CLABSI and CAUTI measures were not expanded to include ward data, as mandated for FY 2018.

We do not know with certainty if any of the hospitals with baseline HAC scores in the highest 25^th percentile used cranberry to treat catheterized patients in FY 2018. Communications with a leading cranberry processor and marketer revealed that, although many hospitals offer cranberry juice as a drink choice for patients and in the cafeteria, very few use it as a CAUTI-prevention strategy. Our assumption, thus, was that these low-performing hospitals were not using cranberry as a CAUTI preventative, and their SIRs for CAUTI represent pre-treatment outcomes.

In the baseline simulation, cranberry was used in each synthetic hospital to treat catheterized patients, assuming 100% of catheterized patients could be treated with an effectiveness rate of 50% reduction in cases of CAUTI based on Foxman et al.¹³. A hospital that reduces its number of CAUTI cases through cranberry treatment reduces the numerator in its CAUTI SIR, while the denominator is unchanged. The CAUTI Winsorized z-score, Domain 2 score, and overall HAC score were then re-computed for each hospital, and its improvement in the HAC-score distribution, if any, was noted, including whether the hospital was able to move from the lowest-performing quartile and thereby avoid the payment reduction.

The extent to which an improved CAUTI z-score reduces the hospital’s Domain 2 and total HAC scores depends upon the weight assigned to the CAUTI score for that hospital. For hospitals with a Domain 1 score, the percentage weight for FY 2018 was 85/5 = 17% for hospitals that have all five Domain 2 scores. However, hospitals with missing Domain 2 scores are not uncommon, making it important to also simulate the potential impact of a cranberry treatment regimen for hospitals that have a CAUTI score but lack one or more of the other Domain 2 scores. Because the number of possible combinations of available and missing Domain 2 scores is very large, we focused, in addition to the case where a hospital had all five Domain 2 scores, on the following three cases for missing scores that were most common in the data:

• Hospitals with only CAUTI, CLABSI, SSI, and CDI Domain 2 scores (221 hospitals);

• Hospitals with only CAUTI, CLABSI, and CDI Domain 2 scores (151 hospitals); and

• Hospitals with only CAUTI and CDI Domain 2 scores (267 hospitals).

Thus, we also generated 1,000 synthetic hospitals from the original distributions for each of these three scenarios. The case of a hospital having only a CAUTI Domain 2 score was not simulated because it applied to only 19 hospitals in the data. Similarly, the case of a missing Domain 1 score (and, hence, 100% weight for Domain 2) was not simulated because it applied to comparatively few hospitals (223) and the percentage weights assigned to CAUTI would have been closely comparable to the Domain 2 scenarios that were simulated.

We also conducted sensitivity analysis for different rates of cranberry effectiveness in preventing CAUTI for the unilateral-adoption scenario. For this purpose, we used the approximate upper and lower bounds of the asymmetric 95% confidence interval from the Foxman et al.¹³ study, 15% reduction in CAUTI for treatment patients and 70% reduction. All simulations for the unilateral-adoption case were repeated for these percentage reductions in CAUTI.

The baseline simulation considered a hospital’s adoption of cranberry as a CAUTI preventative in isolation, i.e. holding other hospitals’ HAC scores constant. We also conducted simulations wherein 10, 20, and 40% of hospitals in the lowest-performing quartile simultaneously adopted a cranberry-treatment regimen. Studying simultaneous adoption enables the simulation to reflect that hospitals in effect compete in CMS’s HAC-scoring tournament, with the goal to not be among those in the lowest-performing quartile; under FY 2018 rules, the program is a pure zero-sum game in this regard.

Simultaneous adoption of a cranberry-treatment regimen by multiple hospitals introduces two new features to the simulation analysis. First, multiple adoptions will cause the mean Winsorized CAUTI SIR, X, to be reduced. This means that an adopting hospital’s CAUTI z-score will improve less than in the unilateral adoption scenario because improvement in CAUTI SIR, in the form of lower X_i, will be counterbalanced by lower X in the computation for z_i. Second, the total HAC score, H^*, defining the 75^th percentile cut-off, will also be reduced, making it less likely, ceteris paribus, that a hospital’s improved HAC score from adopting a cranberry-treatment regimen will enable it to escape from the lowest-performing quartile.

For the three simultaneous-adoption simulations, we used random draws of actual hospitals from each ventile in the lowest-performing quartile to preserve possible correlation among the Domain 1 and Domain 2 scores. In each simultaneous-adoption scenario we simulated an equal percentage of adoptions (i.e. 10, 20, and 40%) for each of the five ventiles comprising the lowest-performing quartile. The weight afforded the CAUTI score was set based on the presence or absence of CLABSI, SSI, MRSA, and CDI scores for each of the real-world hospitals chosen from the random draw. To avoid dependence of results on any particular set of draws, each of the three simultaneous-adoption scenarios was repeated 100 times.

In summary, for the base simulation with unilateral adoption of cranberry by a hospital for CAUTI treatment, 60 simulations were run, each with 1,000 draws of synthetic hospitals: five HAC performance ventiles × four Domain 2 score scenarios × three cranberry-effectiveness scenarios—i.e. the Foxman et al.¹³ mean and upper and lower 95% confidence bounds. In addition, three simultaneous-adoption scenarios were considered: 10%, 20%, and 40%, each equally distributed within the lowest-performing quartile, with 100 draws of actual hospitals in each case.

Results

Unilateral adoption

Table 2 provides summary statistics, by ventile, for each of the HAIs that go into the total HAC calculations. Figure 1 shows the total HAC baseline distribution, with vertical lines indicating the delineations for the ventiles used in the simulation analysis. The left-most line indicates H^*, the 75^th percentile cut-off. Although the HAC-score distribution is smooth under the Winsorized z methodology, scores are “binned” for purposes of constructing the figure, causing some lumpiness in its appearance.

Figure 1. Total HAC distribution with ventile cut-offs.

Table 2. Summary statistics of Winsorized z-scores by ventile.

	Ventiles of total HAC score
	75–80	80–85	85–90	90–95	95–100
# of Hospitals with CAUTI score	158	154	147	133	77
Domain 1 (PSI 90)
Mean	0.37	0.28	0.35	0.43	0.45
SD	0.71	0.74	0.70	0.70	0.74
Min	−1.32	–1.32	–1.32	–1.32	–1.32
Max	1.61	1.67	1.67	1.67	1.61
CLABSI
Mean	0.36	0.44	0.75	0.94	1.59
SD	0.82	0.78	0.91	0.97	0.95
Min	–0.90	–0.85	–0.78	–0.98	–0.74
Max	2.44	2.44	2.44	2.44	2.44
CAUTI
Mean	0.41	0.59	0.73	0.81	1.53
SD	0.78	0.77	0.94	0.89	0.86
Min	–1.07	–0.93	–0.81	–0.94	–0.60
Max	2.28	2.28	2.28	2.28	2.28
SSI
Mean	0.50	0.69	0.67	1.02	1.24
SD	0.79	0.90	0.87	0.97	1.00
Min	–0.91	–0.86	–0.86	–0.78	–0.64
Max	2.35	2.35	2.35	2.35	2.35
MRSA
Mean	0.44	0.41	0.68	1.04	1.75
SD	0.87	0.84	0.99	1.03	1.07
Min	–0.87	–0.96	–0.89	–0.75	–0.70
Max	2.63	2.63	2.63	2.63	2.63
CDI
Mean	0.29	0.44	0.48	0.68	1.30
SD	0.64	0.71	0.71	0.75	0.76
Min	–1.09	–1.10	–1.26	–1.41	–1.42
Max	1.86	1.86	1.86	1.86	1.86

Table 3 contains results for the base unilateral-adoption simulation. All of the 1,000 synthetic hospitals in each simulation are subject ex ante to the Medicare payment reduction, i.e. have H_i > H^*= 0.345. An improved CAUTI SIR through adoption of a cranberry-treatment regimen offers the strong possibility but not a guarantee that the hospital will improve its Winsorized z-score for CAUTI and in turn its Domain 2 and overall HAC scores. The only exceptions would be for (i) hospitals with zero CAUTI cases in the SIR numerator ex ante, and (ii) hospitals with a CAUTI SIR in the 0–5 or 95–100 percentile ranges both prior to and after implementing a cranberry-treatment regimen. Under the Winsorization process, the hospital’s SIR would remain at either the 5^th or 95^th percentile cut-off.

Table 3. Number of hospitals avoiding payment reductions in unilateral-adoption simulation.

Initial percentile in total HAC score distribution	75–80	80–85	85–90	90–95	95–100
Hospitals with all scores	942	568	116	0	0
Hospitals with CAUTI, CLABSI, SSI, & CDI	965	741	205	29	0
Hospitals with CAUTI, CLABSI, & CDI	987	921	442	126	0
Hospitals with CAUTI & CDI	1,000	966	828	535	78

Results show that nearly all hospitals in the 75–80 ventile avoid the payment reduction through adoption of the cranberry treatment at mean Foxman et al.¹³ effectiveness, regardless of the weight afforded the CAUTI score in Domain 2. Only 58 among the 1,000 synthetic hospitals with all Domain 2 scores remained subject to the reduced payment, and even fewer retained the penalty if they were missing one or more of the Domain 2 scores, thereby affording greater weight to the improved CAUTI performance.

Most of the hospitals in the 80–85 ventile were also able to avoid the Medicare payment reduction by adopting a cranberry-treatment regimen: 568/1,000 hospitals with all Domain 2 scores avoided the payment reduction, 741 of hospitals lacking a MRSA score avoided the penalty, and 921 avoided it if missing both MRSA and SSI scores.

In the 85–90 ventile, only 116 synthetic hospitals with all Domain 2 scores avoided the payment reduction through adopting cranberry. Substantially more avoided it if they were missing one or more scores. Results for the 90–95 and 95–100 ventiles show that very few hospitals are able to avoid payment reduction solely by adopting a cranberry-treatment regimen, even if they are missing one or more Domain 2 scores.

Adoption of cranberry may be part of a multi-faceted strategy for hospitals to improve their HAC scores and avoid the Medicare payment reduction. Table 4 provides insights into how a cranberry-treatment regimen would contribute to such a strategy. It provides the average HAC score improvement, standard deviation, and maximum and minimum improvement for each ventile and each of the four Domain 2 possibilities under consideration.

Table 4. Change in total HAC score from using cranberries.

Initial percentile in total HAC score distribution	75–80	80–85	85–90	90–95	95–100
Hospitals with all scores
Average change	–0.146	–0.157	–0.161	–0.183	–0.218
Maximum change	–0.337	–0.340	–0.337	–0.339	–0.338
Minimum change	–0.011	0.000	0.000	–0.052	0.000
Standard deviation	0.063	0.069	0.081	0.077	0.079
Hospitals with CAUTI, CLABSI, SSI, & CDI
Average	–0.185	–0.195	–0.197	–0.239	–0.285
Maximum change	–0.014	–0.425	–0.421	–0.424	–0.423
Minimum change	–0.422	0.000	0.000	–0.065	0.000
Standard deviation	0.079	0.081	0.095	0.097	0.102
Hospitals with CAUTI, CLABSI, & CDI
Average	–0.251	–0.269	–0.266	–0.320	–0.382
Maximum change	–0.562	–0.567	–0.562	–0.566	–0.564
Minimum change	–0.018	0.000	0.000	–0.087	0.000
Standard deviation	0.090	0.094	0.121	0.128	0.133
Hospitals with CAUTI & CDI
Average	–0.355	–0.389	–0.403	–0.501	–0.610
Maximum change	–0.761	–0.837	–0.843	–0.849	–0.846
Minimum change	–0.081	0.000	0.000	–0.131	0.000
Standard deviation	0.105	0.121	0.142	0.166	0.198

Table 4 reveals occurrence of the possibility noted earlier that cranberry treatment with 50% improvement in CAUTI may not improve the overall total HAC score. To no surprise, average improvements in total HAC scores from adopting cranberry are larger for hospitals lacking one or more Domain 2 scores, but are, nonetheless, substantial, even for hospitals with all scores, where they range from −0.146 for the 75–80 percentile to −0.218 for the 95–100 percentile range. Absolute improvement in CAUTI scores tends to be greatest for hospitals in the lowest-performing ventiles because, on average, these hospitals have the highest CAUTI Winsorized z-scores and, thus, obtain the greatest benefit from the simulated 50% reduction. Also notable from Table 4 is that the range and standard deviation of HAC-score improvements is quite substantial, even within individual cells of the table. The variance is due to ex ante differences in CAUTI z-scores among synthetic hospitals within the cell and emphasizes the intuitive point that hospitals with the most to gain from adopting a cranberry-treatment regimen for their catheterized patients are those with the highest incidence of UTI within that patient population.

Appendix Tables A1–A4 replicate the information contained in Tables 3 and 4 for the approximate Foxman et al.¹³ 95% confidence bounds on effectiveness of cranberry in reducing CAUTI. Appendix Tables A1 and A2 give results for the upper bound of effectiveness, 70% improvement, and Appendix Tables A3 and A4 do the same for the lower bound, 15% improvement. As expected, if hospitals are able to obtain a 70% improvement (which is similar to the efficacy found in Barnoiu et al.¹²), then even more are able to avoid the Medicare payment reduction under unilateral adoption, including nearly all in the 75–80 and 80–85 ventiles, regardless of the number of Domain 2 scores, and substantial numbers in the 85–90 and 90–95 ventiles, if missing one or more Domain 2 scores. Even with 70% efficacy, most hospitals in the 95–100 ventile cannot escape the Medicare penalty solely by implementing cranberry treatment for their catheterized patients. The opposite is true if hospitals attain only a 15% improvement. Avoidance of the Medicare payment reduction is attained mainly for hospitals in the 75–80 ventile, and average improvements in HAC score are lower for all cases considered.

Simultaneous adoption

Table 5 summarizes simulation results when 10, 20, and 40% of hospitals in each ventile within the lowest-performing quartile simultaneously adopt a cranberry-treatment regimen for their catheterized patients and obtain a 50% reduction in CAUTI. Note that each ventile does not contain the same number of hospitals, due to limiting the simulation to hospitals that have a CAUTI score. For example, in the 10% adoption scenario, 67 hospitals in the lowest quartile were studied, with each trial under each adoption scenario replicated 100 times.

Table 5. 10%, 20%, and 40% simultaneous adoption.

	10% Adoption		20% Adoption		40% Adoption
	Mean	%	Mean	%	Mean	%
Number of hospitals adopting	67		134		267
Change in total HAC score	–0.228	–34.20%	–0.218	–33.00%	–0.211	–31.9%
Change in 75th percentile of total HAC	–0.009	–2.60%	–0.017	–4.94%	–0.027	–7.7%
Adopting hospitals avoiding penalty	27	40%	49	37%	89	33%
Adopting hospitals with HAC1< HACi < HAC0	1	1%	4	3%	14	5%

Whereas 46.5% of the 20,000 synthetic hospitals in the unilateral-adoption simulation summarized in Table 3 avoided the Medicare-payment reduction, on average across 100 trials 27/67 = 40.3% of actual hospitals avoided it under 10% simultaneous adoption within the lowest-performing quartile. The percentage avoiding the payment reduction drops to 37% and 33% as the simultaneous adopters increase to 20% and 40%, respectively. Simultaneous adoption causes the critical HAC value, H^*, to decrease—only by 2.6% in the 10% adoption case, increasing to 4.9% and 7.7% in the 20% and 40% adoption scenarios. The bottom row of Table 5 shows hospitals that would have avoided the penalty at the ex ante (i.e. pre-cranberry adoption) critical value, H⁰_*, but who do not avoid it, given the ex post-critical value H₁^*<H₀^*. Only about one hospital, on average, falls into this category for 10% adoption, but the number increases to four and 14 for the 20% and 40% adoption scenarios.

The other factor at work is that simultaneous adoption improves the mean Winsorized z-score for CAUTI, reducing the amount of normalized improvement in CAUTI z_i for any cranberry-adopting hospital and, in turn, the magnitude of improvement in its overall HAC score. This effect is reflected in the average absolute and percentage change in HAC scores across the three simulations, which declines as the percentage of adopters increases from 10 to 20 to 40. Nonetheless, fully one third of adopting hospitals on average avoid the Medicare-payment reduction penalty, even in the presence of 40% of hospitals in the lowest-performing quartile, adopting cranberry treatment for their catheterized patients. Given the FY 2018 structure of the program, all hospitals moving from the lowest quartile would be replaced by hospitals acquiring the penalty, namely those with HAC scores, H_j, such that H₁^*<H_j<H₀^*

Discussion

Hospitals that avoid a Medicare payment reduction through adopting cranberry obtain an immediate financial benefit in that their revenues from serving Medicare patients are not reduced by 1%. The monetary value of this benefit hinges on the magnitude of a hospital’s Medicare billings. For FY 2017, across the 769 hospitals incurring a penalty, the total reimbursement reduction was $430 million, or $559,000 per hospital, on average. In addition, hospitals that are able to reduce the incidence of CAUTI avoid the costs of treating the infection, costs that are not reimbursed by Medicare and Medicaid under the “no pay rule”, implemented under the HAC Initiative²³.

Communications with a leading cranberry processing and marketing firm indicate that, at 2017 prices, hospitals could supply an 8 oz serving of cranberry juice to patients at a cost of ∼ $0.40. Thus, for example, the 16 oz treatment provided by Foxman et al.¹³ would have an ingredient cost of less than $1.00 per day. Labor and other variable costs would largely be unaffected by introduction of cranberry as a CAUTI preventative, because it would be provided as part of meals or when other medications were being dispensed. Given the very nominal costs of providing cranberry to catheterized patients, it seems certain that any hospital that is able to avoid Medicare payment reduction through adopting a cranberry treatment regimen would reap financial benefits well in excess of the costs of providing the cranberry treatment. Even if the HAC-score improvement, by itself, is insufficient to avoid the payment reduction, hospitals can benefit from adopting cranberry as part of a multi-faceted strategy to improve their HAC scores and avoid the payment reduction.

A parallel rationale applies to hospitals that were not in the lowest quartile and, thus, not subject to the payment reduction for FY 2018. Although these hospitals were not part of the simulation analysis, the Foxman et al.¹³ results and our work for the lowest quartile of hospitals suggest that many of them could improve their CAUTI and HAC scores through adopting a cranberry treatment regimen, thereby providing a level of protection against adverse developments in any of the other scores that could drop them into the lowest quartile and subject them to the payment reduction.

Some limitations of the study should be noted. They include the inability to replicate exactly the data used by CMS to establish HAC scores for FY 2018 and an inability to confirm the assumption that all hospitals subject to the Medicare payment reduction were not using cranberry as a prophylactic treatment for their catheterized patients. We also assume that the data reported to the NHSN and available from CMS regarding CAUTI and the other HAI are accurate, but we cannot rule out that some hospitals could act strategically in what they report. The study also assumes that hospitalized catheter patients would be able to consume cranberry in either juice or capsule form, but this need not be true in 100% of cases. Further, the simulations studied adoption only among hospitals in the lowest 25^th percentile and were subject to the reimbursement penalty. Adoption of cranberry as a CAUTI treatment among the upper 75^th percentile would reduce the ability of hospitals in the lowest quartile to avoid the penalty through their own adoption.

Finally, the study relied upon the Foxman et al.¹³ 50% estimate of efficacy for cranberry treatment in reducing the incidence of CAUTI. Although this estimate is in line with much of the prior literature, it is subject to error, and some of the recent literature, particularly studies that have measured presence and frequency of bacteriuria instead of incidence of UTI in treatment and control groups, have found no significant effect for cranberry treatment.

Although this analysis has focused specifically on provision of cranberry as a CAUTI preventative, the results regarding impacts of improved CAUTI SIRs on the CAUTI Winsorized z, Domain 2 score, and total HAC score apply to any other prophylactic measures introduced to reduce the incidence of CAUTI, such as use of antibiotics at the time of catheter removal or the adoption of a comprehensive set of strategies to reduce CAUTI, such as discussed and studied by Saint et al.²⁴. We also suggest that the simulation framework developed here can be adapted readily to study strategies to improve SIRs for other Domain 2 HAIs.

Conclusion

Hospitals ranking in the lowest quartile based on the CMS HAC score are subject to a 1% Medicare payment reduction. This study has developed and implemented a simulation framework to analyze the impacts of hospitals adopting a cranberry-treatment regimen for their catheterized patients to reduce the incidence of CAUTI. Results show that many hospitals currently subject to the reimbursement reduction can avoid it if they are able to attain a percentage reduction in CAUTI comparable to that achieved in the Foxman et al.¹³ study.

Key considerations in determining whether a hospital can avoid the Medicare-payment reduction by implementing a cranberry-treatment regimen for its catheterized patients are (i) the hospital’s relative ranking within the lowest-performing quartile and (ii) the number of Domain 2 measures for which it has a score, thereby determining the percentage weight afforded to the CAUTI score and any improvements in it. Even hospitals that remain in the lowest quartile despite adopting cranberry as a CAUTI preventative are highly likely to improve their CAUTI SIR and overall HAC scores, meaning that a cranberry-treatment regimen to reduce CAUTI can be part of a hospital’s overall strategy to improve its HAC score and avoid the Medicare payment reduction.

Transparency

Declaration of funding

Major funding for this study was provided by Ocean Spray Cranberries, Inc.

Declaration of financial/other relationships

RJS has received fees from Ocean Spray Cranberries as an outside economic consultant. None of the authors has financial interest in Ocean Spray Cranberries nor any other company in the cranberry industry, nor any other conflicts of interest. JME peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Appendix

Table A1. Number of hospitals avoiding payment reductions (upper CI; 70%).

Initial percentile in total HAC score distribution	75–80	80–85	85–90	90–95	95–100
Hospitals with all scores	969	810	315	43	0
Hospitals with CAUTI, CLABSI, SSI, & CDI	983	891	475	156	0
Hospitals with CAUTI, CLABSI, & CDI	995	962	691	391	24
Hospitals with CAUTI & CDI	1,000	971	938	792	382

Table A2. Change in total HAC score from using cranberries (upper CI; 70%).

Initial percentile in total HAC score distribution	75−80	80−85	85−90	90−95	95−100
Hospitals with all scores
Average change	−0.204	−0.219	−0.225	−0.254	−0.317
Maximum change	−0.439	−0.441	−0.439	−0.441	−0.440
Minimum change	−0.015	0.000	0.000	−0.073	−0.072
Standard deviation	0.088	0.094	0.108	0.103	0.096
Hospitals with CAUTI, CLABSI, SSI, & CDI
Average	−0.259	−0.272	−0.277	−0.333	−0.414
Maximum change	−0.549	−0.551	−0.549	−0.551	−0.550
Minimum change	−0.019	0.000	0.000	−0.092	−0.090
Standard deviation	0.109	0.110	0.125	0.130	0.118
Hospitals with CAUTI, CLABSI, & CDI
Average	−0.352	−0.376	−0.372	−0.444	−0.557
Maximum change	−0.732	−0.735	−0.732	−0.734	−0.733
Minimum change	−0.026	0.000	0.000	−0.122	−0.120
Standard deviation	0.126	0.130	0.163	0.171	0.152
Hospitals with CAUTI & CDI
Average	−0.497	−0.544	−0.563	−0.697	−0.900
Maximum change	−1.049	−1.095	−1.098	−1.102	−1.100
Minimum change	−0.113	0.000	0.000	−0.183	−0.360
Standard deviation	0.147	0.168	0.192	0.222	0.198

Table A3. Number of hospitals avoiding payment reductions (lower CI; 15%).

Initial percentile in total HAC score distribution	75–80	80–85	85–90	90–95	95–100
Hospitals with all scores	455	23	0	0	0
Hospitals with CAUTI, CLABSI, SSI, & CDI	620	34	0	0	0
Hospitals with CAUTI, CLABSI, & CDI	770	61	8	0	0
Hospitals with CAUTI & CDI	948	271	28	2	0

Table A4. Change in total HAC score from using cranberries (lower CI; 15%).

Initial percentile in total HAC score distribution	75–80	80–85	85–90	90–95	95–100
Hospitals with all scores
Average change	−0.043	−0.048	−0.049	−0.056	−0.051
Maximum change	−0.159	−0.163	−0.158	−0.162	−0.160
Minimum change	0.000	0.000	0.000	0.000	0.000
Standard deviation	0.023	0.029	0.033	0.035	0.050
Hospitals with CAUTI, CLABSI, SSI, & CDI
Average	−0.055	−0.060	−0.061	−0.074	−0.069
Maximum change	−0.199	−0.204	−0.198	−0.203	−0.201
Minimum change	0.000	0.000	0.000	0.000	0.000
Standard deviation	0.027	0.033	0.039	0.047	0.068
Hospitals with CAUTI, CLABSI, & CDI
Average	−0.075	−0.081	−0.082	−0.100	−0.090
Maximum change	−0.265	−0.272	−0.264	−0.270	−0.267
Minimum change	0.000	0.000	0.000	0.000	0.000
Standard deviation	0.031	0.037	0.049	0.062	0.091
Hospitals with CAUTI & CDI
Average	−0.107	−0.117	−0.123	−0.154	−0.138
Maximum change	−0.257	−0.387	−0.396	−0.406	−0.401
Minimum change	−0.024	0.000	0.000	0.000	0.000
Standard deviation	0.032	0.040	0.056	0.083	0.153