Centers for Medicare and Medicaid Services’ readmission reports inaccurately describe an institution's decompensated heart failure admissions

Abstract

Hospitals typically use Center for Medicare and Medicaid Services’ (CMS) Hospital Readmission Reduction Program (HRRP) administrative reports as the standard of heart failure (HF) admission quantification. We aimed to evaluate the HF admission population identified by CMS HRRP definition of HF hospital admissions compared with a clinically based HF definition. We evaluated all hospital admissions at an academic medical center over 16 months in patients with Medicare fee‐for service benefits and age ≥65 years. We compared the CMS HRRP HF definition against an electronic HF identification algorithm. Admissions identified solely by the CMS HF definition were manually reviewed by HF providers. Admissions confirmed with having decompensated HF as the primary problem by manual review or by the HF ID algorithm were deemed “HF positive,” whereas those refuted were “HF negative.” Of the 1672 all‐cause admissions evaluated, 708 (42%) were HF positive. The CMS HF definition identified 440 admissions: sensitivity (54%), specificity (94%), positive predictive value (87%), negative predictive value (74%). The CMS HF definition missed 324 HF admissions because of inclusion/exclusion criteria (15%) and decompensated HF being a secondary diagnosis (85%). The CMS HF definition falsely identified 56 admissions as HF. The most common admission reasons in this cohort included elective pacemaker or defibrillator implantations (n = 13), noncardiac dyspnea (n = 9), left ventricular assist device complications (n = 8), and acute coronary syndrome (n = 6). The CMS HRRP HF report is a poor representation of an institution's HF admissions because of limitations in administrative coding and the HRRP HF report inclusion/exclusion criteria.

1. INTRODUCTION

Decompensated heart failure (HF) is the most common cause of hospital admission among the Centers for Medicare and Medicaid Services’ (CMS) population, comprising more than half of the national cost of HF care.1, 2, 3, 4 In an effort to improve the quality of care and reduce preventable HF admissions, CMS initiated the Hospital Readmission Reduction Program (HRRP) in October 2012, which imposes a percentage reduction in Medicare base reimbursements to hospitals underperforming based on their 3‐year risk‐standardized all‐cause 30‐day readmission rate following an index HF admission.5, 6, 7 Although an unintended consequence of the HRRP's financial incentives, hospitals often use the CMS HRRP report as the sole benchmark for defining and quantifying their HF admission population and making quality‐improvement decisions without understanding the methodology or the inclusion and exclusion criteria of the HRRP HF report.8, 9, 10 The CMS HF admission definition used in the HRRP requires a primary discharge International Classification of Diseases (ICD) HF diagnostic code in addition to satisfying several significant inclusion and exclusion criteria (Table 1).11

Table 1.

CMS Hospital Readmission Reduction Program's definition of an HF admission

Inclusion criteria
Principal discharge diagnosis of HF1
Enrolled in Medicare FFS or VA beneficiaries
Age ≥65 years
Discharged alive from a nonfederal acute‐care hospital or VA hospital
Not transferred to another acute‐care facility
Enrolled in Part A and Part B Medicare for the 12 months prior to admission date
Enrolled in Part A during the hospital admission
Exclusion criteria
Without ≥30 days’ enrollment in Medicare FFS after index hospital discharge2
Discharged AMA

Abbreviations: AMA, against medical advice; CMS, Centers for Medicare and Medicaid Services; FFS, fee‐for‐service; HF, heart failure; ICD‐9, International Classification of Diseases, Ninth Revision; VA, Veterans Affairs.

Adapted from Appendix D.2 HF of the Yale New Haven Health Services Corp., Center for Outcomes Research and Evaluation (CORE), Prepared for CMS. Condition‐Specific Measures Updates and Specifications Report: Hospital‐Level 30‐Day Risk‐Standardized Readmission Measures. Baltimore, MD: Centers for Medicare and Medicaid Services; March 2015.11

¹ICD‐9 list: 402.01, 402.11, 402.91, 404.01, 404.03, 404.11, 404.13, 404.91, 404.93, 428.xx.

²This information was not available and therefore was not included as either an inclusion or exclusion criterion in our analysis.

Identification of HF hospitalizations is complex because the presenting phenotype varies and no single confirmatory test exists. CMS HRRP relies upon administrative diagnostic claims coding for HF admission quantification. ICD codes are insufficient in reliably identifying patient cohorts, with approximately one‐third of hospitalizations identified as HF by ICD‐9 codes judged as false positives upon medical‐record review.12, 13 Recent studies simply evaluating ICD‐9 and ICD‐10 codes’ ability to identify decompensated HF admissions have not studied the performance in the CMS beneficiary population alone or applied the inclusion and exclusion criteria in the CMS HRRP definition of a HF admission.14, 15, 16, 17

In this study, we compare the HF admission population of the CMS HRRP definition for acute decompensated HF hospital admissions to a clinically based decompensated HF definition in a CMS beneficiary population. Through this comparison, we aimed to quantify and describe the impact that HRRP HF report definitions and methodology have on omitting actual HF admissions and incorrectly identifying non‐HF admission as HF admissions. With this knowledge, hospitals can better interpret their HRRP report, make better quality‐improvement resource‐allocation decisions, and impact their HRRP performance by identifying practices of inappropriately labeling a non‐HF admission as a HF admission in the data submitted to CMS.

2. METHODS

We retrospectively identified all hospital admissions at Vanderbilt University Medical Center from August 15, 2013, to January 31, 2015, that met the following inclusion criteria: age ≥65 years and enrollment in CMS fee‐for‐service benefits. We then compared the CMS definition of an HF admission to a clinical HF definition in the 1672 all‐cause hospitalizations identified. For the CMS HF definition, we employed the inclusion and exclusion criteria used by CMS HRRP for the definition of a HF index admission and all HF readmissions within 30 days (Table 1) to measure all HF admissions included in the HRRP report, with one exception2, 11, 18: We were unable to ascertain the exclusion criterion “Lack of at least 30 days enrollment in Medicare Fee‐for‐Service after index hospital discharge.” We extracted patient‐level data from Vanderbilt University Medical Center's institutional enterprise data warehouse on patients’ primary diagnosis ICD‐9‐CM codes of admission, age at admission, payer, discharge disposition, and subsequent hospitalizations within 30 days. We applied CMS algorithms to identify index admission and its subsequent readmission within 30 days, ensuring that a 30‐day rehospitalization was not counted as both an index hospitalization and a readmission within 30 days. We followed CMS HRRP HF readmission measure methodology to set up the HF cohort to mimic what CMS would use as the HF index admissions as the denominator for the calculation of risk‐standardized readmission rate.11

For the standard of comparison, we utilized a clinically based decompensated HF identification algorithm that has been previously validated in our institution's population to identify a decompensated HF admission with 95% specificity.19 The HF identification algorithm utilized requires an admission to meet 3 of 4 criteria: intravenous loop diuretic administration, b‐type natriuretic peptide (BNP) > 400 pg/mL, history of ICD‐9 code for HF, or admitting clinical diagnosis of decompensated HF selected on an electronic hospital admission forms within the electronic medical record (EMR). As the HF identification algorithm had already been validated against the gold standard of provider review and demonstrated 95% specificity, admissions classified by the HF identification algorithm did not undergo further provider review. To address potential false negatives from the HF identification algorithm's limitations in sensitivity, hospital admissions identified by the CMS HRRP HF definition but not identified by the HF identification algorithm were reviewed by 3 HF providers. The HF provider panel met as a group and manually reviewed the EMR using Framingham HF symptom list and the Atherosclerosis Risk in Communities (ARIC) criteria for acute decompensated HF to reach a consensus decision.20, 21 The EMRs were adjudicated without knowledge of the administrative coding. If HF was deemed not to be the primary admission reason, the group reached a consensus on the actual primary admission reason. Admissions confirmed as HF admissions by clinician EMR review or by the HF identification algorithm were “HF positive,” whereas those refuted by the HF ID algorithm or manual review were “HF negative.” The primary outcome was the HF admission identification accuracy of the CMS HF admission definition compared with the clinically defined HF positive definition. Secondary outcomes included quantification and description of admissions that were false negatives and false positives by the CMS HF definition.

2.1. Statistical analysis

Statistical analysis was performed using the χ² test for categorical variables and the Student t test or Mann–Whitney U test for continuous variables, as appropriate.

3. RESULTS

Of the 1672 all‐cause hospital admissions evaluated, 764 (45%) were identified as an HF admission by either the CMS HF definition or the clinically based decompensated HF identification algorithm. After manual EMR review, 708 of these 764 admissions were validated as HF positive, indicating the primary admission reason was decompensated HF (Figure 1). Thus, the CMS HF definition identified 440 admissions as HF admissions, of which 384 were validated as HF positive and 56 were refuted as HF negative. The CMS HF definition missed an additional 324 HF positive admissions (46% of total HF positive admissions), performing with a sensitivity of 54%, specificity of 94%, positive predictive value of 87%, and negative predictive value of 74% (see Supporting Information, Table 1, in the online version of this article).

Figure 1.

A total of 764 hospitalizations in which decompensated HF was the primary admission problem were identified by the CMS HF definition (n = 440) and the clinical HF definition (n = 553) based upon the HF identification algorithm. Of those identified by the CMS HF definition (n = 440), 229 were validated as HF positive by the clinical HF definition, 155 were validated as HF positive by provider review, and 56 were deemed HF negative upon provider review. Therefore, 708 HF positive admissions were verified to result from decompensated HF as the primary problem. Excluding HF admissions identified by both definitions (n = 229), 324 HF positive admissions were missed by the CMS HF definition. Abbreviations: CMS, Center for Medicare and Medicaid Services; HF, heart failure.

Compared with the 324 HF positive admissions the CMS HF definition missed, the 440 admissions identified as HF admissions by the CMS HF definition were more likely to have an intravenous loop diuretic order on the first day of admission (91% vs 74%; P = 0.001) and more likely to be on cardiology inpatient services (74% vs 60%; P = 0.001; Table 2). Notably, the groups did not differ in the known history of HF, percentage of admissions without a BNP test, mean BNP, or percentage of HF with reduced ejection fraction. The CMS HF definition is dependent upon ICD coding for HF as the primary discharge diagnosis. The most common primary discharge diagnoses in the 324 missed HF positive admissions were cardiac arrhythmias, aortic valve disorders, and acute coronary syndromes (ACS; Table 3). The majority of missed HF positive admissions (85%) had an ICD‐9 code for decompensated HF listed as a secondary diagnosis. Further illustrating the nuances of the CMS HRRP HF report, 48 (15%) of the missed HF positive admissions did have a primary discharge diagnosis of HF, but they were excluded based on other CMS HRRP criteria. Of these 48, 21 were excluded because the HF admission was within 30 days of a previous HF admission. The CMS HRRP HF report does not include these admissions because they are considered linked to the previous index admission. The remaining exclusion reasons included Medicare HMO but not fee‐for‐service benefits (n = 13), in‐hospital mortality (n = 10), hospital‐to‐hospital transfer prior to discharge (n = 3), and enrolled in Medicare < 1 year (n = 1).

Table 2.

Patient and hospital admission characteristics

Characteristic	HF Admissions Identified by CMS, n = 440	HF Admissions Missed by CMS, n = 324	P Value
Age y	77 ± 8.2	77 ± 7.6	0.4817
Male sex	249 (57)	203 (63)	0.0920
Caucasian	378 (86)	269 (83)	0.2739
LVEF unknown	45 (10)	41 (13)
LVEF ≤40%	175 (44)	122 (43)	0.757
BNP not ordered in admission	48 (11)	39 (12)	0.6276
BNP, pg/mL	1259 ± 1229	1316 ± 1565	0.6099
History of HF ICD‐9 code in past year	191 (43)	156 (48)	0.1935
IV loop diuretic order within 24 h of admission	400 (91)	239 (74)	0.0001
Admission SBP, mm Hg	131 ± 25.4	126 ± 25.2	0.0211
eGFR, mL/min/m2	42 ± 16	41 ± 16	0.5424
Discharge service
Cardiology	327 (74)	196 (60)
Noncardiology	113 (26)	128 (40)	0.0001
Medical comorbidities
CAD	289 (66)	238 (73)	0.0217
DM	232 (53)	166 (51)	0.6832
CKD	143 (33)	101 (31)	0.6974
COPD	119 (27)	99 (31)	0.2883
AF	238 (54)	169 (52)	0.5971
HTN	358 (81)	274 (85)	0.2470
Length of hospital stay, d	5.8 ± 6.0	8.4 ± 8.7	0.0001
Discharged to home	339 (77)	202 (62)
Discharged to care facility1	101 (23)	122 (38)	0.0001
30‐day all‐cause readmissions rate to index hospital, %2	16.1	20.5	0.1474

Abbreviations: AF, atrial fibrillation; BNP, B‐type natriuretic peptide; CAD, coronary artery disease; CKD, chronic kidney disease; CMS, Center for Medicare and Medicaid Services; COPD, chronic obstructive pulmonary disease; DM, diabetes mellitus; eGFR, estimated glomerular filtration rate (by MDRD equation); HF, heart failure; HTN, hypertension; ICD‐9, International Classification of Diseases, Ninth Revision; IV, intravenous; LVEF, left ventricular ejection fraction; MDRD, Modification of Diet in Renal Disease; SBP, systolic blood pressure; SD, standard deviation.

Data are presented as n (%) or mean ± SD.

¹Care facilities include skilled nursing facility, inpatient rehabilitation, long‐term acute‐care facility, and hospital transfer.

²Readmission rate algorithm employed CMS Hospital Readmission Reduction Program's inclusion and exclusion criteria.11

Table 3.

Most frequent primary discharge diagnoses of HF admissions not identified by CMS HF definition

CMS Clinical Classification	ICD‐9 Diagnostic Code	Primary Discharge Diagnosis	N (% of 324 Admissions)
Cardiac dysrhythmias	427.xx		41 (13)
	427.31	AF	27
	427.1	Paroxysmal ventricular tachycardia	9
	427.89	Other specified cardiac dysrhythmias	5
Heart valve disorders	424.1	Aortic valve disorders	20 (6)
Acute MI	410.71	Acute MI, subendocardial infarction, init.	18 (6)
Acute and unspecified renal failure	584.9	Acute kidney failure, unspecified	14 (4)
Coronary atherosclerosis	414.01	Coronary atherosclerosis of native coronary vessel	13 (4)
Respiratory failure	518.xx		13 (4)
	518.81	Acute respiratory failure	7
	518.84	Acute and chronic respiratory failure	6
Septicemia	038.9	Unspecified septicemia	7 (2)

Abbreviations: AF, atrial fibrillation; CMS, Center for Medicare and Medicaid Services; HF, heart failure; ICD‐9, International Classification of Diseases, Ninth Revision; init., initial episode of care; MI, myocardial infarction.

The CMS HF definition incorrectly labeled 56 HF negative admissions as HF admissions. Manual EMR review revealed the most common admission reasons in this cohort to be elective pacemaker or defibrillator implantations (n = 13), noncardiac dyspnea (n = 9), left ventricular assist device (LVAD) complications (n = 8), and ACS (n = 6; Table 4). Thus, 13% of the HF admissions identified by the CMS HF definition were not secondary to HF and represent a major opportunity for reducing an institution's 30‐day all cause readmission rate in the CMS HRRP HF report.

Table 4.

Hospital admissions incorrectly identified by CMS HF definition

Admission Indication by Clinician EMR Analysis	n (%)
Elective pacemaker/defibrillator implantation	13 (23)
Noncardiac dyspnea	9 (16)
LVAD complication	8 (14)
ACS	6 (11)
Heart or renal transplant complication	3 (5)
Valvular surgery	3 (5)
Hypertensive urgency	3 (5)
Oncology	2 (4)
Pulmonary hypertension evaluation	2 (4)
Other1	7 (13)
Total	56 (100)

Abbreviations: ACS, acute coronary syndrome; AF, atrial fibrillation; CMS, Center for Medicare and Medicaid Services; EMR, electronic medical record; HF, heart failure; ICD, implantable cardioverter‐defibrillator; LVAD, left ventricular assist device.

¹Other causes include syncope, hyperkalemia, hemodialysis complication, AF, noncardiac distributive shock, diarrhea, and inappropriate ICD defibrillation.

4. DISCUSSION

Our study is the first to our knowledge to compare the CMS HRRP HF definition to both a validated HF definition based on clinical variables and the gold standard of provider EMR review. We found that the CMS HRRP HF definition misses 46% of our institution's decompensated HF admissions, and 13% of the HF admissions identified by CMS HRRP HF definition were non‐HF admissions. Therefore, hospitals should use the CMS HRRP report only for its intended purpose, and clinical definitions should be utilized to accurately measure the HF admission burden and quality care intervention planning. Although these findings are directly applicable to hospitals within the United States, our findings should generate questions and study hypotheses for any country using population reports for quality of care or disease population identification.

Knowledge of the CMS HRRP report limitations can affect an institution's performance in the CMS HRRP in 2 main scenarios:

1. A patient, within the 30‐day window from a recent index hospitalization for HF, is electively admitted for a defibrillator implantation. When this admission is incorrectly coded as HF, which was the primary cause of incorrect coding in our study (n = 13), this admission now counts as a 30‐day readmission in the HRRP report. An institution with 800 annual index HF hospitalizations and an annual 30‐day all‐cause readmission rate of 21% (n = 168) could lower its 30‐day all‐cause readmission rate to 19% by simply correcting the 13 elective defibrillator admissions incorrectly coded as HF admissions. Of note, CMS allows hospitals to submit supplemental data to refute ICD‐coded admissions. Thus, by either correcting coding practices or by retroactively submitting supplemental data to CMS refuting the incorrect ICD code, hospitals can potentially reduce the number of 30‐day unplanned readmissions comprising their readmission percentage.

2: Institutions using the CMS HRRP report to quantify HF admissions will underestimate the true burden of HF admissions. The CMS HRRP HF definition missed 324 (46%) of the HF admissions in our study, and 86% of these had HF as a secondary diagnosis. Although the institution does benefit in the short term from this diagnostic flaw, it can blind an institution to the enormity of the disease burden. Thus, small fluctuations in the percentage of admissions reaching the full potential of being coded as HF (eg, HF moving from a secondary diagnosis to a primary diagnosis) can result in a large increase in the number of HF index or 30‐day readmissions. If financial and resource‐allocation decisions are based upon the CMS HRRP quantification, clinicians could be underpowered to address the magnitude of the HF population, leading to continued failure in reducing 30‐day all‐cause readmission rates. Given the multitude of potential coding scenarios with varying immediate and long‐term economic consequences, we were unable to estimate the financial impact of our findings at the hospital level.

Recent comparisons of ICD administrative coding and clinical diagnoses of acute decompensated HF have yielded mixed results, likely due to the differences in the populations studied and the ICD coding definitions employed. The ARIC study examined the agreement of ICD‐9 codes and a clinician panel in defining a decompensated HF hospital admission. Using ICD‐9 HF codes in the primary diagnosis position, the authors reported a similarly low sensitivity of 21.3% and a specificity of 98.3%, with a κ coefficient of 0.80 (95% confidence interval: 0.77‐0.83).14 Likewise, ICD‐10 codes in the emergency room were recently compared with clinician adjudication for acute decompensated HF. Using those deemed as a high probability of HF (n = 809) by clinicians as the standard, primary problem ICD‐10 codes in the emergency room identified acute decompensated HF with a sensitivity of 76% and a specificity of 50%.15 Our study differs from these in that we applied the many other inclusion and exclusion criteria employed by the CMS HRRP HF definition to the CMS Medicare 65‐or‐older beneficiary population, not ICD codes alone. In summary reviews of studies attempting to validate ICD codes for HF admission identification, ICD codes have performed unreliably.13, 22, 23 Notable reasons for this failure may include coder experience and institutional coder “culture.”6, 22, 24 Institutions should evaluate their coding culture to identify potential coding patterns in need of improvement, such as those listed in Table 4.

We compared the CMS HRRP HF definition, which is dependent on decompensated HF as the primary diagnosis, to a clinical HF definition that could identify an admission with an acute problem of decompensated HF, but not necessarily the “primary problem.” Thus, an argument can be made that the more inclusive clinical definition is biased to identify more HF admissions. We view this as the objective of the study and not a significant methodological limitation for 2 reasons. First, our primary objective was to quantify the gap between the CMS HRRP report's and the actual number of decompensated HF admissions, illustrating these different methods of HF identification have different purposes. The CMS HRRP report, because of its reliance on administrative coding and strict inclusion/exclusion criteria, should only be used for its intended purpose. However, if the institution's goal is to identify all patients with decompensated HF to better quantify the disease burden and quality‐of‐care issues, the CMS HRRP report should not be utilized. Second, the assignment of a primary problem is often complicated in HF, where it is unclear if HF exacerbated another comorbidity (such as atrial fibrillation) or if this occurred in reverse order.25, 26, 27 Therefore, the issue of the initial insult is irrelevant from an administrative viewpoint if both medical problems require acute medical care.

4.1. Study limitations

Several limitations of our study warrant discussion. First, we performed our analysis at a single tertiary‐care, academic medical center. Our findings may not be indicative of other institutions with different HF populations and coding practices. Yet, our findings should prompt other institutions to explore their accuracy of ICD codes as a potential method of HRRP performance improvement. Second, we used an HF identification algorithm in addition to clinician EMR review as the standard of comparison. Though this clinically based algorithm has been previously validated in our institution to have a specificity of 95% for decompensated HF admissions, it has not been validated in other populations and is dependent upon providers to generate the data points required for identification.19 To overcome the algorithm's limitations in sensitivity from admissions with incomplete identification data, we performed manual EMR review of admissions identified by the CMS definition but not the HF identification algorithm. However, it is possible we did not identify some HF admissions missed by both the CMS and HF identification algorithm. Third, we were unable to ascertain one HRRP exclusion “Lack of at least 30 days enrollment in Medicare Fee‐for‐Service after index hospital discharge.” Therefore, it is possible that our findings slightly overestimate the agreement between the HRRP definition and actual HF admissions because we did not exclude admissions for this reason. Additionally, LVAD complications were a recurrent cause of false positives by the CMS HF definition. This may not be relevant to institutions who do not care for this population. Furthermore, the CMS HRRP report will now exclude patients with a procedure code for LVAD implantation or heart transplantation starting in the 2016 HRRP measures.28 This change further underscores the importance of understanding the inclusion and exclusion criteria in the CMS HRRP definition of HF admissions, as fluctuations in the readmission rates can be due to definition changes and not the quality of care.

5. CONCLUSION

The CMS HRRP HF report should not be used as an institution's standard for HF admission measurement because of limitations in administrative coding, reliance upon a primary discharge diagnosis, and the inclusion/exclusion criteria of the HRRP HF report. Knowledge of these limitations by clinicians and administration is critical for accurate assessment of the HF admission burden and allocation of resources to impact the quality of care delivered, and can be an opportunity to lower a hospital's 30‐day readmission rate in the CMS HRRP.

Conflicts of interest

The authors declare no potential conflicts of interest.

Supporting information

Table S1. Sensitivity and Specificity of CMS HF Definition

Cox ZL, Lai P, Lewis CM and Lenihan DJ. Centers for Medicare and Medicaid Services’ readmission reports inaccurately describe an institution's decompensated heart failure admissions. Clin Cardiol. 2017;40:620–625. 10.1002/clc.22711