HEDIS 1 2018 Ad-Hoc Public Comment: Table of Contents HEDIS Overview... 1 The HEDIS Measure Development Process... Synopsis... Submitting Comments... NCQA Review of Public Comments... Value Set Directory... Items for Public Comment... Questions... Proposed Changes to Existing Measures... 4 Hospitalization for Potentially Preventable Complications...8 Inpatient Hospital Utilization...17 Proposed Measures to Retire... 38 Annual Monitoring for Patients on Persistent Medications...40 Children and Adolescents Access to Primary Care Practitioners...63 Frequency of Ongoing Prenatal Care...71 1 HEDIS is a registered trademark of the National Committee for Quality Assurance (NCQA).
HEDIS 1 2018 Ad-Hoc Public Comment Overview HEDIS Overview HEDIS is a set of standardized performance measures designed to ensure that purchasers and consumers can reliably compare the performance of health plans. It also serves as a model for emerging systems of performance measurement in other areas of health care delivery. HEDIS is maintained by NCQA, a not-for-profit organization committed to evaluating and publicly reporting on the quality of physicians, HMOs, PPOs, ACOs and other organizations. The HEDIS measurement set consists of 95 measures across seven domains of care. Items available for Public Comment are being considered for the 2018 HEDIS Technical Specifications Update, which will be published in October 2017 and reported in June 2018, based on activity that occurred during 2017 and, where applicable, prior years. HEDIS Measure Development Process To develop measures, NCQA uses a consensus development process that involves a rigorous review of published guidelines and published scientific evidence and feedback from multi-stakeholder advisory panels. The NCQA Committee on Performance Measurement, a diverse panel of independent scientists and representatives from health plans, consumers, federal policymakers, purchasers and clinicians, oversees the evolution of the measurement set. Numerous Measurement Advisory Panels provide clinical and technical knowledge required to develop the measures. Additional HEDIS Expert Panels and the Technical Measurement Advisory Panel provide invaluable assistance by identifying methodological issues and providing feedback on new and existing measures. Synopsis In order to be responsive to the need for quick analysis and updates, NCQA seeks feedback on proposed changes to two existing measures and the retirement of three measures this July. Changes are being considered for HEDIS 2018. Reviewers are asked to submit their comments to NCQA in writing via the Public Comment Web site by 11:59 PM (ET) Thursday, July 27. Submitting Comments Submit all comments via NCQA s Public Comment Web site, using the following link: http://publiccomments.ncqa.org Note: NCQA does not accept comments via mail, e-mail or fax. How to Submit a Comment 1. Enter the following information: Your e-mail address. Your contact information. 2. Choose from the following options: 1 HEDIS is a registered trademark of the National Committee for Quality Assurance (NCQA). 2017 National Committee for Quality Assurance 1
Select HEDIS 2018 Ad-Hoc Public Comment. Select the measure on which you would like to comment. Select your support option (e.g., Support, Support with modifications, Do not support,). Note: If you choose Support with modifications, enter the suggested modification in the text box. If you choose Do not support, include your rationale in the text box. 3. Enter your comment into the text box. There is a 2,500 character limit. If you exceed the limit, your comment will be cut off at 2,500 characters. Note: We suggest that you develop your comment in Word, in order to check your character limit, and save a copy for reference. Use the cut and paste function to copy your comment into the text box. 4. If you are submitting more than one comment use the Submit and Return button. When you have submitted all comments use the Submit and Logout button to receive an e-mail notification with all submitted comments. NCQA must receive your comments by 11:59 PM (ET) Thursday, July 27. NCQA Review of Public Comments NCQA appreciates the time and effort required to submit comments, and reviews all feedback submitted within the Public Comment period. Due to the high volume of comments received, NCQA cannot respond to individual comments; however, NCQA advisory panels and the Committee on Performance Measurement will consider comments and advise NCQA staff. Value Set Directory Effective with HEDIS 2014, code tables are not included in measure specifications; they are included in a separate Excel workbook called the Value Set Directory. Measure specifications reference value sets that must be used for HEDIS reporting. A value set contains the complete set of codes used to identify the service or condition included in the measure. Codes for the measures listed below are in the HEDIS 2018 Ad-Hoc Public Comment Value Set Directory, which is included with the measure materials on the NCQA Public Comment Web page. Items for Public Comment Refer to the NCQA Public Comment page for detailed documentation on the items described below. Proposed Changes to Existing Measures NCQA proposes changes to the following measures: 1. Hospitalization for Potentially Preventable Complications. 2. Inpatient Hospital Utilization. Proposed Measures to Retire NCQA proposes to retire the following measures: 1. Annual Monitoring for Patients on Persistent Medications. 2017 National Committee for Quality Assurance 2
2. Children and Adolescents Access to Primary Care Practitioners. 3. Frequency of Ongoing Prenatal Care. Questions? Contact NCQA Customer Support at 888-275-7585, Monday Friday, 8:30 a.m. 5:00p.m. (EST). 2017 National Committee for Quality Assurance 3
Proposed Changes to Existing Measures for HEDIS 1 2018: Inpatient Hospital Utilization (IHU) and Hospitalization for Potentially Preventable Complications (HPC) NCQA seeks comments on the following proposed revisions to two risk-adjusted utilization measures: Inpatient Hospital Utilization (IHU) and Hospitalization for Potentially Preventable Complications (HPC) for inclusion in the HEDIS 2018 measurement set. 1. Inclusion of observation stays in addition to inpatient stays in the measure rates 2. Removal of individuals with high-frequency hospitalization from the measure rates Measure Description Inpatient Hospital Utilization (IHU): For members 18 years of age and older, the risk-adjusted ratio of observed to expected acute inpatient discharges during the measurement year reported by Surgery, Medicine and Total. Hospitalization of Potentially Preventable Complications (HPC): For members 67 years of age and older, the rate of discharge for ambulatory care sensitive conditions (ACSC) per 1,000 members and the risk-adjusted ratio of observed to expected discharges for ACSC by chronic and acute conditions. These measures are risk adjusted for age, sex and co-morbid conditions using the CMS Hierarchical Condition Categories (HCC) model. Inclusion of Observation Stays Background There is growing concern among stakeholders regarding the increasing use of observation stays and that inappropriate use may serve as a mechanism to avoid an inpatient admission. To address these concerns, NCQA performed a thorough literature review, stakeholder interviews and empirical testing. Observation stays are intended for short-term assessment and reassessment of patient acuity and symptom etiology to inform a decision regarding further treatment as an inpatient admission or hospital discharge. From a payment perspective, observation stays are considered an outpatient service, however frequently from the patient s perspective the observation stay may be identical to an inpatient stay with care provided on the same floor from the same care team. The use of observation stays can have significant financial implications for patients, providers and health plans when used as an alternative to inpatient admission 2. Our feedback, from literature and stakeholders confirmed that observation stay prevalence and duration is increasing, while inpatient utilization appears to be decreasing. Although there are many factors influencing observation stay use, our feedback emphasized that observation stay utilization is primarily driven by reimbursement factors, by available resources, and to lesser extent by quality measurement efforts. Methods Currently, no HEDIS measure tracks observation stays; however they may be a concern for measurement efforts intending to capture utilization and hospitalization events. We are conducting testing using a large 1 HEDIS is a registered trademark of the National Committee for Quality Assurance (NCQA). 2 Observation stays are covered by Medicare Part B, which results in increased cost-sharing for patients due to 20% coinsurance and non-coverage of self-administered medications (i.e. prescription and over-the-counter drugs you get in an outpatient setting). Observation stays also do not contribute towards the three-day hospitalization prerequisite for Medicare coverage of skilled nursing facility care. Health plans typically reimburse providers at a lower rate for observation stays compared to inpatient admissions; in addition, providers are also less vulnerable to claims denial for observation stays than inpatient admissions. 2017 National Committee for Quality Assurance 4
subset of Medicare claims covering 2,395,212 beneficiaries and a large subset of Commercial claims covering 5,259,228 eligible members. The Medicare and Commercial databases included claims from October 2015 through September 2016. The initial testing focuses on descriptive analysis of each of the proposed revisions (observation stays and high-frequency hospital utilizers). We examined the impact of each revision on the unadjusted measure rates for both Inpatient Hospital Utilization and Hospitalization for Potentially Preventable Complications. These results are presented across the entire sample of Medicare beneficiaries and Commercial members. The next phase of testing will focus on developing risk adjustment models for each of the measures to account for underlying differences in the population characteristics served by each plan. Initial Measure Testing NCQA identified observation stays using Uniform Billing revenue codes 0760, 0762 and 0769 on outpatient claims and restricted analysis to observation stays that do not result in inpatient admission (observation stays that convert to inpatient admissions are already included in the risk-adjusted utilization measures). Analyses of Medicare Advantage and Commercial samples indicate there are 57.8 observation stays per 1,000 Medicare beneficiaries and 7.1 observation stays per 1,000 commercial members. Stays are predominantly short (1 day or less) and the most common diagnosis is disease of the circulatory system. We also examined the impact of adding observation stays to both the Inpatient Hospital Utilization and Hospitalization for Potentially Preventable Complications measures. Inclusion of observation stays in the numerator of the Medicare Inpatient Hospital Utilization measure increased the total rate from 130.0 events per 1,000 beneficiaries to 185.4 events per 1,000 beneficiaries, a 43 percent increase. In the Commercial population, inclusion of observation stays in the measure increased the total rate from 22.9 events per 1,000 members to 29.8 events per 1,000 members, a 30 percent increase. The possible inclusion of observation stays in the Hospitalization for Potentially Preventable Complications measure as a separate rate indicates that among Medicare beneficiaries there is 1 observation stay for every 3.6 inpatient stays for acute ambulatory care sensitive conditions and 1 observation stay for every 3.6 inpatient stays for chronic ambulatory care sensitive conditions. Inclusion of observation stays in the total rate increased both the chronic condition rate and the acute condition rate by 27 percent. Our advisory panels voiced strong support for inclusion of observation stays in the measures and we considered three options for making this potential change: 1) to report observation stays as a separate rate 2) to include observation stays with inpatient stays for and overall rate of hospitalization 3) to implement both options 1 and 2. Overall, the feedback highlighted the value in measuring overall riskadjusted rate of hospitalization reflecting both inpatient admission and observation stay discharges (Option 2). Although, we received some feedback that we should present observation stays separately from inpatient stays, we also heard concerns that multiple rates could be confusing to consumers. Also, our research to date suggests that the use of observation stays is driven heavily by system and hospital factors and less by individual characteristics so it would be difficult to develop an accurate risk adjustment model for a standalone observation rate. Therefore, we are recommending inclusion of observation stays with inpatient admissions to reflect and overall rate of hospitalization. High Frequency Hospitalization Background In addition to the review of observation stays, NCQA is re-examining the approach to risk-adjustment for these measures with specific attention to high-frequency hospital utilizers. Research conducted by the Agency for Healthcare Research and Quality and others describe a subpopulation of individuals across all payers who experience markedly high frequency of hospitalization, without evidence of clinical or demographic characteristics differentiating them from other patients (Jiang et al. 2015). In our first-year 2017 National Committee for Quality Assurance 5
analysis of the Hospitalization for Potentially Preventable Complications measure we identified several plans with outlier performance where performance was significantly higher (or worse) than would reasonably be expected. One hypothesis is that these plans were small plans with a handful of high frequency hospital utilizers, which both inflated the observed rate and affected the ability of the risk adjustment model to adequately predict an expected rate of hospitalization. Initial Testing Results Hospitalization for Potentially Preventable Complications To better understand the prevalence of high-frequency hospital utilizers, we explored the rate of repeated hospitalizations for both all cause (IHU) and ambulatory care sensitive conditions (HPC). Using the testing database, we conducted an outlier analyses to assess the distribution of hospitalizations across the sample. We first examined the frequency of inpatient and observation stay hospitalizations for acute and chronic conditions specified in the Hospitalization for Potentially Preventable Complications measure. Testing results indicate very few patients experience a hospitalization event for an acute (1.1 percent) or a chronic (1.5 percent) condition identified in the measure during a measurement year. Among those hospitalized, 93 percent of these patients only experience one event for an acute condition and 84 percent only experience one event for a chronic condition. However, in testing we found a subgroup of individuals with high and outlying utilization; among patients with an acute hospitalization event, less than one percent (0.83 percent) experience three or more acute hospitalizations and among patients with a chronic hospitalization event, less than five percent (4.3 percent) experience three or more chronic hospitalizations. Inpatient Hospital Utilization We also assessed frequency of inpatient and observation stay hospitalization for medical and surgical events as specified in the Inpatient Hospital Utilization measure. Among Medicare beneficiaries, 9.8 percent experience at least one medical hospitalization and 4.4 percent experience a surgical hospitalization during a measurement year. There is less hospital utilization among Commercial members, with 1.7 percent experiencing a medical hospitalization and 0.7 percent experiencing a surgical hospitalization. We looked at the distribution of discharge counts among Medicare beneficiaries and Commercial members who did experience a medical or surgical observation stay or inpatient admission within the measurement year. Results show that the majority of Medicare beneficiaries with at least one discharge are only hospitalized once (78 percent-medical; 90 percent-surgical). Also, the majority of commercial members with at least one discharge are only hospitalized once (77 percent medical; 93 percent-surgical). We observed a subset of individuals with high and outlying hospital utilization. Less than 7 percent of beneficiaries with at least one medical hospitalization event experience three or more medical hospitalization events. Similarly, less than two percent of beneficiaries with at least one surgical hospitalization event experience three or more surgical hospitalization events. While individuals with high utilization represent a very small proportion of patients hospitalized as specified for both measures, including them in the denominator will distort the risk model. Therefore, we are proposing to remove hospitalizations experienced by individuals with high frequency of hospitalization, classify those individuals as outliers and report an outlier count. NCQA seeks feedback on the following proposed revisions: Inpatient Hospital Utilization (IHU) Revise the definition of acute hospital discharge to include discharges for observation stays that did not result in inpatient admissions Revise the measure title to Acute Hospital Utilization to indicate the measure is not focused solely on inpatient admissions Remove individuals who experience more than 3 hospitalizations in the measurement year and classify this population as outliers 2017 National Committee for Quality Assurance 6
Hospitalization for Potentially Preventable Complications (HPC) Revise the definition of acute hospital discharge to include discharges for observation stays that did not result in inpatient admissions Remove individuals who experience more than 3 hospitalizations for each of the Acute and Chronic defined diagnoses in the measurement year and classify these populations as outliers Supporting documents include draft measure specifications and associated measure rationale workup. NCQA acknowledges the contributions of the Utilization and Geriatric Measurement Advisory Panels. 2017 National Committee for Quality Assurance 7
Hospitalization for Potentially Preventable Complications (HPC) Developed by the Agency for Healthcare Research and Quality (AHRQ) and adapted by NCQA with permission. SUMMARY OF CHANGES TO HEDIS 2018 Added observation stay discharges. Added step to remove discharges for members with three or more chronic ambulatory care sensitive condition inpatient and observation stay discharges during the measurement year from the chronic ACSC measure calculation and report these members as chronic ACSC outliers. Added step to remove discharges for members with three or more acute ambulatory care sensitive condition inpatient and observation stay discharges during the measurement year from the acute ACSC measure calculation and report these members as acute ACSC outliers. Added Count of Chronic ACSC Outliers and Count of Acute ACSC Outliers to Table HPC-A-3. Clarified the definition of direct transfer : when the discharge date from the first inpatient setting precedes the admission date to a second inpatient setting by one calendar day or less. Clarified to round to 10 decimal places using the.5 rule during the intermediate calculations of Expected events. Added steps 5 and 6 to the calculation of the PPD risk weights to calculate covariance and total variance for each category. Removed the Risk Adjustment Weighting Process diagram. Added Total Variance as a data element to Table HPC-B-2/3, Table HPC-C-3 and Table HPC-D-3. Description For members 67 years of age and older, the rate of discharges for ambulatory care sensitive conditions (ACSC) per 1,000 members and the risk-adjusted ratio of observed to expected discharges for ACSC by chronic and acute conditions. Definitions ACSC Ambulatory care sensitive condition. An acute or chronic health condition that can be managed or treated in an outpatient setting. The ambulatory care conditions included in this measure are: Chronic ACSC Diabetes short-term complications. Diabetes long-term complications. Uncontrolled diabetes. Lower-extremity amputation among patients with diabetes. COPD. Asthma. Hypertension. Heart failure. Acute ACSC Bacterial pneumonia. Urinary tract infection. Cellulitis. Pressure ulcer. 2017 National Committee for Quality Assurance 8
Chronic ACSC Outlier Acute ACSC Outlier Classification period PPD PUCD Members with three or more inpatient and observation stay chronic ambulatory care sensitive conditions during the measurement year. Members with three or more inpatient and observation stay acute ambulatory care sensitive conditions during the measurement year. The year prior to the measurement year. Predicted probability of discharge. The predicted probability of a member having any discharge in the measurement year. Predicted unconditional count of discharge. The predicted unconditional count of discharges for members during the measurement year Eligible Population Note: Members in hospice are excluded from the eligible population. Refer to General Guideline 20: Members in Hospice. Product lines Ages Continuous enrollment Allowable gap Anchor date Benefit Event/diagnosis Required exclusions Medicare. 67 years and older as of December 31 of the measurement year. The measurement year and the year prior to the measurement year. No more than one gap in enrollment of up to 45 days during each year of continuous enrollment. December 31 of the measurement year. Medical. None. Members who are enrolled in an Institutional Special Needs Plan (I-SNP) any time during the measurement year. Calculation of Observed Events Report each ACSC category separately and as a combined total. The total is the sum of the acute and chronic ACSC categories. Chronic ACSC Step 1 Step 2 Follow the steps below to identify the number of chronic ACSC acute inpatient and observation stay discharges. Identify all acute inpatient and observation stay discharges during the measurement year. To identify acute inpatient and observation stay discharges: 1. Identify all acute and nonacute inpatient stays (Inpatient Stay Value Set) and observation stays (Observation Stay Value Set). 2. Exclude nonacute inpatient stays (Nonacute Inpatient Stay Value Set). 3. Identify the discharge date for the stay. If an observation stay results in an acute inpatient stay, include only the acute inpatient stay discharge. Acute-to-acute direct transfers: Keep the original discharge and drop the direct transfer s discharge. 2017 National Committee for Quality Assurance 9
A direct transfer is when the discharge date from the first inpatient setting precedes the admission date to a second inpatient setting by one calendar day or less. For example: An inpatient discharge on June 1, followed by an admission to another inpatient setting on June 1, is a direct transfer. An inpatient discharge on June 1, followed by an admission to an inpatient setting on June 2, is a direct transfer. An inpatient discharge on June 1, followed by an admission to another inpatient setting on June 3, is not a direct transfer; these are two distinct inpatient stays. Use the following method to identify acute-to-acute direct transfers: 1. Identify all acute and nonacute inpatient stays (Inpatient Stay Value Set). 2. Exclude nonacute inpatient stays (Nonacute Inpatient Stay Value Set). 3. Identify the admission and discharge dates for the stay. Step 3 For the remaining acute inpatient and observation stay discharges, identify discharges with any of the following: Primary diagnosis for diabetes short-term complications (ketoacidosis, hyperosmolarity or coma; Diabetes Short Term Complications Value Set). Primary diagnosis for diabetes with long-term complications (renal, eye, neurological, circulatory or unspecified complications; Diabetes Long Term Complications Value Set). Primary diagnosis for uncontrolled diabetes (Uncontrolled Diabetes Value Set). A procedure code for lower extremity amputation (Lower Extremity Amputation Procedures Value Set) with any diagnosis for diabetes (Diabetes Diagnosis Value Set). Exclude any discharge with a diagnosis for traumatic amputation of the lower extremity (Traumatic Amputation of Lower Extremity Value Set). Exclude any discharge with a diagnosis for toe amputation procedure (Toe Amputation Value Set). Primary diagnosis of COPD (COPD Diagnosis Value Set). Exclude any discharge with a diagnosis for cystic fibrosis or anomaly of the respiratory system (Cystic Fibrosis and Respiratory System Anomalies Value Set). Primary diagnosis for asthma (Asthma Diagnosis Value Set). Exclude any discharge with a diagnosis for cystic fibrosis or anomaly of the respiratory system (Cystic Fibrosis and Respiratory System Anomalies Value Set). Primary diagnosis for acute bronchitis (Acute Bronchitis Diagnosis Value Set) with diagnosis for COPD (COPD Diagnosis Value Set). Exclude any discharge with a diagnosis for cystic fibrosis or anomaly of the respiratory system (Cystic Fibrosis and Respiratory System Anomalies Value Set). Primary diagnosis for heart failure (Heart Failure Diagnosis Value Set), excluding Exclude any discharges with a cardiac procedure (Cardiac Procedure Value Set). Primary diagnosis for hypertension (Hypertension Value Set). Exclude any discharge with a cardiac procedure (Cardiac Procedure Value Set). Exclude any discharge with a diagnosis of Stage I-IV kidney disease (Stage I-IV Kidney Disease Value Set) with a dialysis procedure (Dialysis Value Set). Step 4 Remove discharges for members with 3 or more chronic ACSC discharges in the measurement year and report these members as chronic ACSC outliers. 2017 National Committee for Quality Assurance 10
Acute ACSC Step 1 Follow the steps below to identify the number of acute ACSC acute inpatient and observation stay discharges. Identify all acute inpatient and observation stay discharges during the measurement year. To identify acute inpatient and observation stay discharges: 1. Identify all acute and nonacute inpatient stays (Inpatient Stay Value Set) and observation stays (Observation Stay Value Set). 2. Exclude nonacute inpatient stays (Nonacute Inpatient Stay Value Set). 3. Identify the discharge date for the stay. If an observation stay results in an acute inpatient stay, include only the acute inpatient stay discharge. Step 2 Acute-to-acute direct transfers: Keep the original discharge and drop the direct transfer discharge. A direct transfer is when the discharge date from the first inpatient setting precedes the admission date to a second inpatient setting by one calendar day or less. For example: An inpatient discharge on June 1, followed by an admission to another inpatient setting on June 1, is a direct transfer. An inpatient discharge on June 1, followed by an admission to an inpatient setting on June 2, is a direct transfer. An inpatient discharge on June 1, followed by an admission to another inpatient setting on June 3, is not a direct transfer; these are two distinct inpatient stays. Use the following method to identify acute-to-acute direct transfers: 1. Identify all acute and nonacute inpatient stays (Inpatient Stay Value Set). 2. Exclude nonacute inpatient stays (Nonacute Inpatient Stay Value Set). 3. Identify the admission and discharge dates for the stay. Step 3 For the remaining acute inpatient and observation stay discharges, identify discharges with the any of the following: Primary diagnosis of bacterial pneumonia (Bacterial Pneumonia Value Set). Exclude any discharge with a diagnosis of sickle cell anemia, HB-S disease (Sickle Cell Anemia and HB-S Disease Value Set). Exclude any discharge with a procedure or diagnosis for immunocompromised state (Immunocompromised State Value Set). Primary diagnosis of urinary tract infection (Urinary Tract Infection Value Set). Exclude any discharge with a diagnosis of kidney/urinary tract disorder (Kidney and Urinary Tract Disorder Value Set). Exclude any discharge with a procedure or diagnosis for immunocompromised state (Immunocompromised State Value Set). Primary diagnosis of cellulitis (Cellulitis Value Set). Primary diagnosis of pressure ulcer (Pressure Ulcer Value Set). Step 4 Total ACSC Remove discharges for members with 3 or more acute ACSC discharges in the measurement year and report these members as acute ACSC outliers. Count of inpatient stays with a discharge date during the measurement year for a chronic or acute ACSC. Sum the events from the Chronic ACSC and Acute ACSC categories to obtain a total ACSC. 2017 National Committee for Quality Assurance 11
Risk Adjustment Determination For each member in the eligible population, except all outliers, use the steps in the Utilization Risk Adjustment Determination section in the Guidelines for Risk Adjusted Utilization Measures to identify risk adjustment categories based on presence of comorbidities. Risk Adjustment Weighting and Calculation of Expected Events Calculation of risk-adjusted outcomes (counts of discharges) uses predetermined risk weights generated by two separate regression models. Weights from each model are combined to predict how many discharges each member may have during the measurement year given their age, gender and the presence or absence of a comorbid condition. Refer to the Risk Adjustment Weight Process diagram for an overview of the process. For each member in the eligible population, except all outliers, assign Predicted Probability of Discharge (PPD) risk weights. Calculate the PPD for each ACSC category: Chronic ACSC, Acute ACSC, Total ACSC. Step 1 Step 2 Step 3 Step 4 Step 5 For each member with a comorbidity HCC Category, link the PPD weights. For Chronic ACSC: Use Table HPCCh-PPD-ComorbidHCC. For Acute ACSC: Use Table HPCA-PPD-ComorbidHCC. For Total ACSC: Use Table HPCT-PPD-ComorbidHCC. Link the age and gender weights for each member. For Chronic ACSC: Use Table HPCCh-PPD. For Acute ACSC: Use Table HPCA-PPD. For Total ACSC: Use Table HPCT-PPD. Identify the base risk weight. For Chronic ACSC: Use Table HPCCh-PPD. For Acute ACSC: Use Table HPCA-PPD. For Total ACSC: Use Table HPCT-PPD. Sum all PPD weights associated with the member (HCC, age, gender and base weight) for each category (Chronic ACSC, Acute ACSC, Total ACSC). Calculate the predicted probability of having at least one discharge in the measurement year, based on the sum of the weights for each member, for each category (Chronic ACSC, Acute ACSC, Total ACSC) using the formula below. PPD = e ( PPDWeightsForEachMember) 1+e ( PPD WeightsForEachMember) Note: The risk adjustment tables will be released on November 1, 2017, and posted to www.ncqa.org. For each member in the eligible population, except all outliers, assign Predicted Unconditional Count of Discharge (PUCD) risk weights. Calculate the PUCD for each ACSC category: Chronic ACSC, Acute ACSC, Total ACSC. Step 1 For each member with a comorbidity HCC Category, link the weights. If a member does not have any comorbidities to which weights can be linked, assign a weight of 1. For Chronic ACSC: Use Table HPCCh-PUCD-ComorbidHCC. For Acute ACSC: Use Table HPCA-PUCD-ComorbidHCC. For Total ACSC: Use Table HPCT-PUCD-ComorbidHCC. 2017 National Committee for Quality Assurance 12
Step 2 Step 3 Link the age and gender weights for each member. For Chronic ACSC: Use Table HPCCh-PUCD. For Acute ACSC: Use Table HPCA-PUCD. For Total ACSC: Use Table HPCT-PUCD. Identify the base risk weight. For Chronic ACSC: Use Table HPCCh-PUCD. For Acute ACSC: Use Table HPCA-PUCD. For Total ACSC: Use Table HPCT-PUCD. Step 4 Expected count of hospitalization Calculate the predicted unconditional count of discharges in the measurement year, by multiplying all PUCD weights (HCC, age, gender and base weight) associated with the member for each ACSC category (Chronic ACSC, Acute ACSC, Total ACSC). Use the following formula: PUCD = Base Weight * Age/gender Weight * HCC Weight Note: Multiply by each HCC associated with the member. For example, assume a member with HCC-2, HCC-10, HCC-47. The formula would be: PUCD = Base Weight * Age/gender Weight * HCC-2 * HCC-10 * HCC-47 Round intermediate calculations to 10 decimal places using the.5 rule. Report the final member-level expected count of discharges for each category using the formula below. Round to four decimal places using the.5 rule and enter these values into the reporting table. Expected Count of ACSC Discharges = PPD x PUCD Step 5 Use the formula below to calculate to calculate the covariance of the predicted outcomes for each category (i.e., gender, age group and type of ACSC). CCCCCC = nn ii=1 PPPPPP mmmmmmmm(pppppp) PPPPPPPP mmmmmmmm(pppppppp) nn 1 Step 6 Use the formula below to calculate the variance for each category. nn TTTTTTTTTT VVVVVVVVVVVVVVVV = (PPPPPP PPPPPPPP) 2 1 + (1 PPPPPP) 2 2 CCCCCC + PPPPPP PPPPPPPP ii=1 Reporting: Members in the Eligible Population and Outliers The number of members in the eligible population, including all outliers, for each age and gender group and the overall total. Enter these values into the reporting table (Table HPC-A-3). Reporting: Observed Events The number of observed discharges within each age and gender group and the overall total for each ACSC category and Total ACSC. Reporting: Observed Discharges per 1,000 Members The number of observed discharges divided by the number of members in the eligible population, multiplied by 1,000 within each age and gender group and the overall total for each ACSC category and Total ACSC. 2017 National Committee for Quality Assurance 13
Reporting: Expected Events The number of expected discharges within each age and gender group and the overall total for each ACSC category and Total ACSC. Reporting: Expected Discharges per 1,000 Members The number of expected discharges divided by the number of members in the eligible population, multiplied by 1,000 within each age and gender group and the overall total for each ACSC category and Total ACSC. Reporting: Total Variance The variance (from Risk Adjustment Weighting and Calculation of Expected Events, PUCD, step 6) within each age and gender group and the overall total for each category (i.e., Chronic ACSC, Acute ACSC and Total ACSC). Note Organizations may not use risk assessment protocols to supplement diagnoses for calculating risk adjustment scores for this measure. The HPC measurement model was developed and tested using only claims-based diagnoses; diagnoses from additional data sources would affect the validity of the models as they are currently implemented in the specifications. Table HPC-A-3: Members in the Eligible Population Chronic ACSC Chronic ACSC Acute ACSC Age Sex Members Outliers Outlier Rate Outliers Male 67-74 Female Total Male 75-84 Female Total Male 85+ Female Total Male Total Female Total Acute ACSC Outlier Rate 2017 National Committee for Quality Assurance 14
Table HPC-B-3: Age 67-74 75-84 85+ Total Table HPC-C-3: Age 67-74 75-84 85+ Total Hospitalization for Potentially Preventable Complication Rates by Age and Risk Adjustment: Chronic ACSC Sex Observed Chronic ACSC Discharges Observed Chronic ACSC Discharges/ 1,000 Members Expected Chronic ACSC Discharges Expected Chronic ACSC Discharges/ 1,000 Members Total Variance (O/E) O/E Ratio (Observed Discharges/Expected Discharges) Male Female Total Male Female Total Male Female Total Male Female Total: Hospitalization for Potentially Preventable Complication Rates by Age, Gender and Risk Adjustment: Acute ACSC Observed Acute ACSC Discharges/ 1,000 Members O/E Ratio (Observed Discharges/ Expected Discharges) Sex Observed Acute ACSC Discharges Expected Acute ACSC Discharges Expected Acute ACSC Discharges/ 1,000 Members Total Variance (O/E) Male Female Total Male Female Total Male Female Total Male Female Total 2017 National Committee for Quality Assurance 15
Table HPC-D-3: Hospitalization for Potentially Preventable Complication Rates by Age, Gender and Risk Adjustment: Total ACSC Age 67-74 75-84 85+ Total Sex Observed Total ACSC Discharges Observed Total ACSC Discharges/ 1,000 Members Expected Total ACSC Discharges Expected Total ACSC Discharges/ 1,000 Members Total Variance (O/E) O/E Ratio (Observed Discharges/ Expected Discharges) Male Female Total Male Female Total Male Female Total Male Female Total 2017 National Committee for Quality Assurance 16
Acute Hospital Utilization (AHU) SUMMARY OF CHANGES TO HEDIS 2018 Renamed the measure. Added observation stay discharges. Added step to remove discharges for members with three or more inpatient and observation stay discharges during the measurement year and report these members as outliers. Added Count of Outliers to Table AHU-A-2/3. Added a note to clarify that Total Inpatient will not equal Surgery and Medicine sum if reporting using MS-DRGs. Clarified to round to ten decimal places using the.5 rule during the intermediate calculations of Expected events. Added steps 5 and 6 to the calculation of the PUCD risk weights to calculate covariance and total variance for each category. Removed the Risk Adjustment Weighting Process diagram. Added Total Variance as a data elements to Table AHU-B-2/3, Table AHU-C-2/3 and Table AHU-D-2/3. Description For members 18 years of age and older, the risk-adjusted ratio of observed to expected acute inpatient and observation stay discharges during the measurement year reported by Surgery, Medicine and Total. Definitions Outlier Classification period PPD PUCD Members with three or more inpatient and observation stay discharges during the measurement year. The year prior to the measurement year. Predicted probability of discharge. The predicted probability of a member having any discharge in the measurement year. Predicted unconditional count of discharge. The predicted unconditional count of discharges for members during the measurement year. Eligible Population Note: Members in hospice are excluded from the eligible population. Refer to General Guideline 20: Members in Hospice. Product lines Ages Continuous enrollment Allowable gap Anchor date Benefit Commercial, Medicare (report each product line separately). 18 and older as of December 31 of the measurement year. The measurement year and the year prior to the measurement year. No more than one gap in enrollment of up to 45 days during each year of continuous enrollment. December 31 of the measurement year. Medical. 2017 National Committee for Quality Assurance 17
Event/diagnosis None. Calculation of Observed Events Use the following steps to identify and categorize acute inpatient and observation stay discharges. Step 1 Step 2 Step 3 Identify all acute inpatient and observation discharges during the measurement year. To identify acute inpatient and observation discharges: 1. Identify all acute and nonacute inpatient stays (Inpatient Stay Value Set) and observation stays (Observation Stay Value Set). 2. Exclude nonacute inpatient stays (Nonacute Inpatient Stay Value Set). 3. Identify the discharge date for the stay. If an observation stay results in an acute inpatient stay, include only the acute inpatient stay discharge. Exclude inpatient and observation stay discharges with: A principal diagnosis of mental health or chemical dependency (Mental and Behavioral Disorders Value Set). A principal diagnosis of live-born infant (Deliveries Infant Record Value Set). A maternity-related principal diagnosis (Maternity Diagnosis Value Set). A maternity-related stay (Maternity Value Set; Maternity MS-DRG Value Set). A mental health, chemical dependency or rehabilitation related stay (IPU Exclusions MS- DRG Value Set). Newborn care (Newborns/Neonates MS-DRG Value Set). Inpatient and observation stays with a discharge for death. Remove discharges for members with three or more discharges in the measurement year and report these members as Outliers in Table AHU-A. Step 4 Calculate the total inpatient using all discharges identified after completing steps 1 3. Step 5 Identify surgery and medicine using MS-DRGs. For organizations that use DRGs, categorize each discharge as surgery or medicine. Surgery (Surgery MS-DRG Value Set). Medicine (Medicine MS-DRG Value Set). Note: If reporting using MS-DRGs, Total Inpatient will not equal the sum of Surgery and Medicine because DRGs for Principal Diagnosis Invalid as Discharge Diagnosis and Ungroupable are included in Total Inpatient, but are not included in surgery or medicine. For observation stays and if the organization does not use MS-DRGs for inpatient stays, follow steps 6 7 to categorize discharges. Step 6 Step 7 Calculate surgery. Identify the surgery discharges (Surgery Value Set) from the total inpatient discharges (step 4). Calculate medicine. Categorize as medicine the discharges remaining after removing surgery (step 6) from the total inpatient discharges (step 4). 2017 National Committee for Quality Assurance 18
Risk Adjustment Determination For each member in the eligible population, use the steps in the Utilization Risk Adjustment Determination section in the Guidelines for Risk Adjusted Utilization Measures to identify risk adjustment categories based on presence of comorbidities. Risk Adjustment Weighting and Calculation of Expected Events Calculation of risk-adjusted outcomes (counts of discharges) uses predetermined risk weights generated by two separate regression models. Weights from each model are combined to predict how many discharges each member may have during the measurement year, given age, gender and presence or absence of a comorbid condition. Refer to the Risk Adjustment Weight Process diagram for an overview of the process. For each member in the eligible population, except for outliers, assign Predicted Probability of Discharge (PPD) risk weights. Calculate the PPD for each service utilization category: Surgery, Medicine, Total. Step 1 For each member with a comorbidity HCC category, link the PPD weights. For the Medicare product line, use the following tables: For Surgery: Use Table AHUS-MA-PPD-ComorbidHCC. For Medicine: Use Table AHUM-MA-PPD-ComorbidHCC. For Total: Use Table AHUT-MA-PPD-ComorbidHCC. For the commercial product line, use the following tables: For Surgery: Use Table AHUS-Comm-PPD-ComorbidHCC. For Medicine: Use Table AHUM-Comm-PPD-ComorbidHCC. For Total: Use Table AHUT-Comm-PPD-ComorbidHCC. Step 2 Link the age-gender PPD weights for each member. For the Medicare product line, use the following tables: For Surgery: Use Table AHUS-MA-PPD. For Medicine: Use Table AHUM-MA-PPD. For Total: Use Table AHUT-MA-PPD. For the commercial product line, use the following tables: For Surgery: Use Table AHUS-Comm-PPD. For Medicine: Use Table AHUM-Comm-PPD. For Total: Use Table AHUT-Comm-PPD. Step 3 Identify the base PPD risk weight for each member. For the Medicare product line, use the following tables: For Surgery: Use Table AHUS-MA-PPD. For Medicine: Use Table AHUM-MA-PPD. For Total: Use Table AHUT-MA-PPD. For the commercial product line, use the following tables: For Surgery: Use Table AHUS-Comm-PPD. For Medicine: Use Table AHUM-Comm-PPD. For Total: Use Table AHUT-Comm-PPD. Step 4 Sum all PPD weights (HCC, age, gender, base weight) associated with the member for each category (Medicine, Surgery, Total). 2017 National Committee for Quality Assurance 19
Step 5 Calculate the predicted probability of having at least one discharge in the measurement year based on the sum of the weights for each member, for each category (Surgery, Medicine, Total), using the formula below. PPD = e ( PPD WeightsForEachMember) 1+e ( PPD WeightsForEachMember) Note: The risk adjustment tables will be released on November 1, 2017, and posted to www.ncqa.org. For each member in the eligible population, except for outliers, assign Predicted Unconditional Count of Discharge (PUCD) risk weights. Step 1 For each member with a comorbidity HCC Category, link the PUCD weights. If a member does not have any comorbidities to which a weight could be linked, assign a weight of 1. For Medicare product line, use the following tables: For Surgery: Use Table AHUS-MA-PUCD-ComorbidHCC. For Medicine: Use Table AHUM-MA-PUCD-ComorbidHCC. For Total: Use Table AHUT-MA-PUCD-ComorbidHCC. For the commercial product line, use the following tables: For Surgery: Use Table AHUS-Comm-PUCD-ComorbidHCC. For Medicine: Use Table AHUM-Comm-PUCD-ComorbidHCC. For Total: Use Table AHUT-Comm-PUCD-ComorbidHCC. Step 2 Link the PUCD age-gender weights for each member. For Medicare product line, use the following tables: For Surgery: Use Table AHUS-MA-PUCD. For Medicine: Use Table AHUM-MA-PUCD. For Total: Use Table AHUT-MA-PUCD. For the commercial product line, use the following tables: For Surgery: Use Table AHUS-Comm-PUCD. For Medicine: Use Table AHUM-Comm-PUCD. For Total: Use Table AHUT-Comm-PUCD. Step 3 Identify the base PUCD risk weight. For Medicare product line use the following tables: For Surgery: Use Table AHUS-MA-PUCD. For Medicine: Use Table AHUM-MA-PUCD. For Total: Use Table AHUT-MA-PUCD. For the commercial product line, use the following tables: For Surgery: Use Table AHUS-Comm-PUCD. For Medicine: Use Table AHUM-Comm-PUCD. For Total: Use Table AHUT-Comm-PUCD. Step 4 Calculate the predicted unconditional count of discharges in the measurement year, by multiplying all PUCD weights (HCC, age, gender and base weight) associated with the member for each category (Surgery, Medicine, Total). Use the following formula: PUCD = Base Weight * Age/gender Weight * HCC Weight Note: Multiply by each HCC associated with the member. For example, assume a member with HCC-2, HCC-10, HCC-47. The formula would be: PUCD = Base Weight * Age/gender Weight * HCC-2 * HCC-10 * HCC-47 2017 National Committee for Quality Assurance 20
Round intermediate calculations to 10 decimal places using the.5 rule. Expected count of hospitalization Step 5 Step 6 Report the final member-level expected count of discharges for each category using the formula below. Round to four decimal places using the.5 rule and enter these values into the reporting table. Expected Count of Discharges = PPD x PUCD Use the formula below to calculate the covariance of the predicted outcomes for each category (i.e., gender, age group and type of hospital stay). CCCCCC = nn ii=1(pppppp mmmmmmmm(pppppp)) (PPPPPPPP mmmmmmmm(pppppppp)) nn 1 Use the formula below to calculate the variance for each category. nn TTTTTTTTTT VVVVVVVVVVVVVVVV = (PPPPPP PPPPPPPP) 2 1 + (1 PPPPPP) 2 2 CCCCCC + PPPPPP PPPPPPPP ii=1 Reporting: Members in the Eligible Population and Outliers The number of members in the eligible population, including outliers, for each age and gender group and the overall total. Enter these values into the reporting table (Table AHU-A-2/3). Reporting: Observed Events The number of observed discharges within each age and gender group and the overall total for each category (Surgery, Medicine, Total). Reporting: Observed Discharges per 1,000 Members The number of observed discharges divided by the number of members in the eligible population, multiplied by 1,000 within each age and gender group and the overall total for each category (Surgery, Medicine, Total). Reporting: Expected Events The number of expected inpatient discharges within each age and gender group and the overall total for each category (Surgery, Medicine, Total). Reporting: Expected Discharges per 1,000 Members The number of expected inpatient discharges divided by the number of members in the eligible population, multiplied by 1,000 within each age and gender group and the overall total for each category (Surgery, Medicine, Total). Reporting: Total Variance The variance (from Risk Adjustment Weighting and Calculation of Expected Events, PUCD, step 6) within each age and gender group and the overall total for each category (i.e., Surgery, Medicine and Total Inpatient). 2017 National Committee for Quality Assurance 21
Table AHU-A-2/3: Members in the Eligible Population Age Sex Members Outliers Outlier Rate Male 18-44 Female Total Male 45-54 Female Total Male 55-64 Female Total Male 65-74 Female Total Male 75-84 Female Total Male 85+ Female Total Male Total Female Total 2017 National Committee for Quality Assurance 22
Table AHU-B-2/3: Acute Inpatient and Observation Stay Discharges by Age and Risk Adjustment: Surgery Age 18-44 45-54 55-64 65-74 75-84 85+ Total Sex Observed Inpatient Discharges Observed Inpatient Discharges/ 1,000 Members Expected Inpatient Discharges Expected Inpatient Discharges/ 1,000 Members Total Variance (O/E) O/E Ratio (Observed Discharges/ Expected Discharges) Male Female Total Male Female Total Male Female Total Male Female Total Male Female Total Male Female Total Male Female Total 2017 National Committee for Quality Assurance 23
Table AHU-C-2/3: Acute Inpatient and Observation Stay Discharges by Age and Risk Adjustment: Medicine Age 18-44 45-54 55-64 65-74 75-84 85+ Total Sex Observed Inpatient Discharges Observed Inpatient Discharges/ 1,000 Members Expected Inpatient Discharges Expected Inpatient Discharges/ 1,000 Members Total Variance (O/E) O/E Ratio (Observed Discharges/ Expected Discharges) Male Female Total Male Female Total Male Female Total Male Female Total Male Female Total Male Female Total Male Female Total 2017 National Committee for Quality Assurance 24
Table AHU-D-2/3: Acute Inpatient and Observation Stay Discharges by Age and Risk Adjustment: Total Inpatient Acute Hospitalization Age Sex Observed Inpatient Discharges Observed Inpatient Discharges/ 1,000 Members Expected Inpatient Discharges Expected Inpatient Discharges/ 1,000 Members Total Variance (O/E) O/E Ratio (Observed Discharges/ Expected Discharges) Male 18-44 Female Total Male 45-54 Female Total Male 55-64 Female Total Male 65-74 Female Total Male 75-84 Female Total Male 85+ Female Total Male Total Female Total 2017 National Committee for Quality Assurance 25
Inclusion of Observation Stays in Risk Adjusted Utilization Measures Topic Workup Executive Summary This workup summarizes evidence on the rising prevalence of observation stays, factors influencing observation stay utilization and implications of observational care for patients, providers and health plans. Findings were obtained through a literature review of studies focused primarily on the Medicare population and through interviews with external stakeholders. The prevalence of observation stays is rising among Medicare beneficiaries, while the prevalence of inpatient admissions is decreasing. Clinically, observation stay patients are often indistinguishable from admitted patients and consume comparable health care resources. Observational care is typically delivered on inpatient floors without separate established protocols and for a wide variety of conditions. At the same time, there is significant variation across hospitals in the use of observation stays and their duration, with more than 1 in 10 observation stays lasting longer than 3 midnights. The similarity between observation status and inpatient stay suggests that observation classification may not be based primarily on clinical differentiation, but could also be influenced by concern that an inpatient stay will not be reimbursed by CMS or other payers. In recent years, there has been increased scrutiny on payment for short inpatient stays, due in part to Medicare audit scrutiny and the Two-Midnight Rule, which specifies that inpatient admissions should be used when the admission is expected to last at least 2 midnights (Overman et al. 2014). While value-based purchasing policies around readmissions (e.g., the CMS Hospital Readmission Reduction Program and its 6 condition-specific readmission measures) may result in some shifts from inpatient to observation status, the relative contribution is uncertain. NCQA is conducting analyses to understand the impact of including observation stays in our measures of risk-adjusted utilization: Inpatient Hospital Utilization (under consideration for HEDIS 2018 with proposed name change to Acute Hospital Utilization). Hospitalization for Potentially Preventable Complications (under consideration for HEDIS 2018). Plan All-Cause Readmissions (under consideration for HEDIS 2019). Characteristics of Observations Stays CMS defines observation stays as a well-defined set of specific, clinically appropriate services, which include ongoing short-term treatment, assessment, and reassessment before a decision can be made regarding whether patients will require further treatment as hospital inpatients or if they are able to be discharged from the hospital. 78% of Medicare observation stays begin through presentation and treatment in the Emergency Department (OIG 2013). There are no regulatory guidelines on where observation services are delivered. Care can either be protocol-driven, guided by predetermined care pathways founded on scientific evidence, or discretionary, relying on independent clinician orders. Table 1 describes four types of observation care as defined by Ross et al. (2013). The exact prevalence of these types of care is not known. 56% of dedicated observation units (types 2 and 3) are managed by ED physicians, although a growing number of hospitalists now oversee observational care (Wiler et al. 2011). One physician, with the support of a nurse practitioner or physician assistant, can often support the care of 10 15 observation stay patients (Napolitano et al. 2014). Type 4 observational care outside a dedicated observation unit and without established protocols is the most common (Ross et al. 2013). Only 33% of hospitals have a dedicated observation unit, with patients otherwise using inpatient or ED beds (Wiler et al. 2011). 2017 National Committee for Quality Assurance 26
Table 1. Hospital Settings in Which Observation Services Are Provided Type 1 Type 2 Type 3 Type 4 Setting Description Characteristics Source: Ross et al. 2013. Protocol driven, dedicated observation unit Discretionary care, dedicated observation unit Protocol driven, bed in any location Discretionary care, bed in any location Highest level of evidence for favorable outcomes. Care typically directed by ED Care directed by a variety of specialists. Unit typically based in ED Often called a virtual observation unit Most common practice. Unstructured care, poor alignment of resources with patients needs. The prevalence of observation stays is rising among Medicare beneficiaries, while the prevalence of inpatient admissions is decreasing (Feng et al. 2012). CMS reported that observation stays doubled from 2006 2014 (950,000 nearly 1.9 million). A retrospective review of records from one academic medical center from July 2010 December 2011 found that over 25% of patients on the general medical service were classified as observation stays; 40% of these were patients 65 years of age or older (Sheehy et al. 2013). The average duration of observation stays is also increasing. In 2014, the average length of observation stays was 28 hours, a 4-hour increase from the average length of stay in 2006 (25.6 hours) (MedPAC 2016). Although Medicare states observation services lasting more than 48 hours should be for rare and exceptional cases, both the Medicare Payment Advisory Committee (MedPAC) and Office of Inspector General (OIG) found 11% of all observation stays last 48 hours or longer (Table 2) (CMS Benefit Policy, Chapter 6). Table 2. Number and Percentage of Observation Stays by Length of Stay, 2012 Length of Stay Number (%) of Observation Stays 0 nights (1 calendar day) 126,264 (8) 1 midnight 833,583 (55) 2 midnights 385,830 (26) At least 3 midnights 166,198 (11) Source: OIG analysis of CMS data, 2013 Observation stays are used by providers for patients with different conditions (Hockenberry et al. 2014). Data from 2012 indicated that the most common reasons for observation services were similar to the most common reasons for short term-inpatient admissions (fewer than two midnights) (OIG 2013). These results suggest observation stays and short-term inpatient admissions may be similar in nature. 2017 National Committee for Quality Assurance 27
Table 3. Most Common Reasons for Observation Stays and Short Inpatient Stays, 2012 Common Reasons 1 Number (%) of Observation Stays 2 Number (%) of Short Inpatient Stays 3 Chest pain 340,484 (22.5) 49,716 (4.3) Digestive disorders 93,091 (6.2) 37,649 (3.3) Fainting 81,349 (5.4) 32,656 (2.8) Signs and symptoms 47,439 (3.1) Nutritional disorders 39,227 (2.6) 24,624 (2.1) Dizziness 34,455 (2.3) Irregular heartbeat 31,390 (2.1) 38,961 (3.4) Circulatory disorders 31,163 (2.1) 29,515 (2.6) Respiratory signs and symptoms 24,715 (1.6) Medical back problems 23,946 (1.6) Coronary stent insertion 45,658 (4.0) Loss of blood flow to the brain 25,355 (2.2) Red blood cell disorders 20,977 (1.8) Irregular heartbeat (medium severity) 20,064 (1.7) 1 The common reasons described in the table include the top 10 reasons for observation stays and the top 10 reasons for short inpatient stays. 2 There were 1,511,875 observation stays in the analysis. 3 There were 1,146,925 short inpatient stays in the analysis. Source: OIG analysis of CMS data, 2013. Patients in observation stays and patients in inpatient stays have similar outcomes. A recent systematic review found no statistically significant difference in mortality and readmission rates for various medical conditions including chest pain, atrial fibrillation, asthma and pyelonephritis when comparing inpatient and observation stay patients. Observation stay patients did have significantly higher satisfaction and some international studies show lower total hospital costs compared with admitted inpatients (Conley et al. 2016). Policy Influences on Utilization of Observation Stays Observation services were initially developed to monitor patients with low acuity chest pain through established care pathways, but have expanded to provide care for patients with a variety of conditions, diagnoses and severity of illness (Mosely et al. 2013). Expansion is attributed to policy and financial pressures (Barrett et al. 2015). The CMS enactment of the Medicare Modernization Act instituted the Recovery Audit Program to recoup payment of inpatient care that should have been administered on an outpatient basis. The program initially started as a pilot demonstration in six states between 2005 and 2008, and was rolled out nationally by 2010. The administrative and financial pressures associated with denial of inpatient claims experienced by providers encouraged increased use of observation stays to avoid targeting of Recovery Audit Contractors (RAC). After implementation of the Medicare Recovery Audit Program, MedPAC found the number of observation hours for Medicare beneficiaries increased by 70%, from 23 million hours in 2006 to 39 million hours in 2010 (MedPAC 2012). Additionally, among observation stays exceeding 72 hours, the average monthly hours doubled from 1,025 in 2007 to 2,258 in 2009 (Feng et al. 2012). 2017 National Committee for Quality Assurance 28
RACs frequently targeted short inpatient admissions lasting less than two midnights, often finding no medical necessity for admission and arguing that care could have been delivered under outpatient or observation status. Of the overpayments recovered by RACs in the first year of the program, half were for short inpatient admissions (Ross et al. 2013). In 2013 CMS issued the Two-Midnight Rule, which states that the Inpatient Prospective Payment System (IPPS) rate will only be paid for patients expected to require more than two midnights of inpatient care. Any inpatient admission lasting less than two midnights requires detailed documentation of medical necessity to receive reimbursement as an inpatient stay. (CMS 2015). Ambiguity and scrutiny of inpatient versus observation status encouraged reliance on level-of-care criteria tools, such as Milliman or McKesson s InterQual, to guide patient classification. Medicare follows the InterQual guidelines when determining medical necessity for reimbursement of claims at the IPPS rate (Platts et al. 2014). However, determination of observation stay status versus inpatient admission can be unclear using InterQual criteria. A study evaluating observation stays and inpatient admissions at an academic medical center in 2011 found wide variation in diagnoses and length of stay for observation stays, suggesting that the use of the InterQual criteria at the medical center did not clearly define observation stays. The authors concluded that flexibility in classification guidelines and conversion of initial patient status after admission, but prior to discharge, suggests an absence of clear guidance regarding appropriate observation stay utilization (Sheehy et al. 2013). The use of observation stays was also influenced by the passing of the Patient Protection and Affordable Care Act (ACA). The legislation pushed the agenda of providing high-quality and lower-cost care, encouraging establishment of value-based purchasing programs (VBP) and the Hospital Readmissions Reduction Program (HRRP). Both VBP and HRRP target 30-day readmissions, penalizing providers with higher rates of patients returning to the hospital. Using observation classification for patients may allow providers to avoid index admission events and 30-day readmission events (Barrett et al. 2015). Value-based purchasing arrangements and technical assistance programs targeting readmissions may not be sole drivers of observation stay increases, as providers were already reducing readmission rates prior to passage of the of ACA. However, there was a more dramatic decline after the legislation was in place (Zuckerman et al. 2016). Rates for conditions covered and not covered by HRRP had a statistically significant decline at hospitals eligible for HRRP penalty and hospitals not eligible for HRRP penalty. Among hospitals eligible for HRRP penalty, readmission rates for conditions covered under HRRP declined at a significantly faster rate than those not covered, while hospitals not eligible for HRRP penalty had similar rates of decline for conditions covered and not covered (Desai et al. 2016). A recent study evaluating readmissions among Medicare fee-for-service (FFS) patients with acute myocardial infarction, heart failure and pneumonia found that 2.2% of AMI patients, 1.6% of heart failure patients and 1.2% of pneumonia patients had an observation stay within 30 days after initial hospital discharge (Venkatesh et al. 2016). A study looking at a sample of both commercial payer claims and Medicare supplemental databases from 2002 2011 found that 17.6% of observation stays had an inpatient admission within 30 days prior. Additionally, 14.5% of patients experienced a hospitalization within 30 days of an observation stay episode (Overman et al. 2014). This suggests for health plans with readmission rates above 14.5% that adding observation stays to a measure of readmissions could increase the rate of readmissions (i.e., more numerator events), but could also decrease the rate of readmissions (i.e., more observation stays as denominator events without a corresponding numerator event). Financial Impact of Observation Stays on Beneficiaries and Medicare Rising prevalence of observation services has financial implications for both patients, providers, and health plans. In Medicare FFS observation stays are considered an outpatient service and paid through Medicare Part B (CMS Benefit Policy, Chapter 6). Thus, patients may experience increased cost-sharing for an observation stay, compared with an inpatient stay, due to 20% coinsurance (for all Medicare Part B covered services), no coverage of self-administered medications and ineligibility to qualify for the Medicare skilled nursing services benefit (Baugh et al. 2013). While most Medicare beneficiaries have 2017 National Committee for Quality Assurance 29
supplemental coverage such as employer sponsored plans, Medicaid or Medigap to assist with coinsurance costs (KFF 2016), there is no cap on beneficiary cost-sharing under Medicare Part B, which creates increased financial vulnerability for patients (CMS 2014). In contrast, the 2017 cost-sharing for beneficiaries associated with inpatient stays is a $1,316 deductible without additional coinsurance costs. Despite concerns about increased patient costs, analysis of 2012 CMS claims found observation stays for most beneficiaries were less expensive than short inpatient admissions. The average per stay beneficiary cost-sharing for observation care ($401) was lower than the average cost-sharing for short-term inpatient care ($725) (Table 4) (OIG 2013). The same analysis found that only 6% of observation stays in 2012, impacting 83,000 beneficiaries, resulted in patients paying out-of-pocket amounts exceeding their inpatient deductible (OIG 2013). The results were similar when comparing the cost to Medicare for shortterm inpatient stays and observation stays. Table 4. Differences Between Average Payments* for Short Inpatient Stays (1 night or less) and Observation Stays, by Most Common Reasons for Treatment, 2012 Top Reasons for Observation or Short Inpatient Stays** Difference in Average Medicare Payments (Short-Term Inpatient Cost Observation Stay Cost)*** Difference in Average Beneficiary Payments (Short-Term Inpatient Cost Observation Stay Cost) Red blood cell disorders $2,801 $373 Irregular heartbeat (medium severity) $2,444 $457 Circulatory disorders $2,312 -$167 Coronary stent insertion $2,267 -$817 Medical back problems $2,085 $404 Digestive disorders $2,047 $425 Nutritional disorders $1,977 $474 Fainting $1,890 $417 Signs and symptoms $1,854 $359 Respiratory signs and symptoms $1,792 $396 Loss of blood flow to the brain $1,677 $415 Dizziness $1,320 $466 Irregular heartbeat $943 $572 Chest pain $870 $419 Calculated average $1,877 $300 * Average payments for observation stays are estimates because each reason is estimated based on information from the Part B hospital claim. **This list includes the top 10 reasons both for observation and short inpatient stays. Source: OIG analysis of CMS data, 2013. ***A positive number indicates higher inpatient costs than observation stay costs. The financial implications of observation stays are not limited to services received during acute-care episodes, but extend to post-acute care episodes. Coverage of skilled nursing facility (SNF) services by Medicare requires a 3-day inpatient admission and hours of observational care do not contribute toward this prerequisite. In 2012, there were 600,000 beneficiaries with hospital stays lasting 3 midnights or more, through an observation stay, a long outpatient stay or a 2-midnight inpatient stay preceded by a 1- midnight outpatient stay. Although these beneficiaries received 3 midnights of care in a hospital, they were not eligible for the Medicare skilled nursing facility benefit (OIG 2013). Skilled nursing care without Medicare coverage costs beneficiaries an average of $10,503 (OIG 2013). Patients with Medicare 2017 National Committee for Quality Assurance 30
Advantage and commercial payers may not be subject to the inpatient admission requirement for coverage of SNF services. Medicare Advantage plans are required to cover, at a minimum, Medicare Parts A and B benefits; however, they may cover additional benefits such as SNF services, without the requisite prior 3 midnights of inpatient care. This varies by Medicare Advantage plans. Medicare beneficiaries under observation often receive the same care services, interact with similar providers and share the same physical environment as admitted inpatients. Until they are informed of their observation stay status, it is difficult for patients to recognize they are not admitted and do not qualify for Medicare Part A coverage. Attention to this issue resulted in the enactment of the Notice of Observation Treatment and Implication for Care Eligibility Act (NOTICE Act). Implemented in 2017, it requires providers to notify Medicare FFS and Medicare Advantage beneficiaries who need more than 24 hours of observational care of their outpatient classification status within 36 hours, through oral and written administration of a Medicare Outpatient Observation Notice (MOON). No published data describing the impact of the NOTICE Act on the prevalence and duration of observation stays has been reported. Variation in Utilization of Observation Stays Utilization, administrative claim coding and reimbursement for observation stays varies among providers and regions. Medicare FFS established more standardized guidance than other payers, but there is still variability in how providers use observation stays in practice. There is wide variation in how hospitals categorize and code stays as observation stays, long outpatient stays (Part B claims lasting longer than 1 midnight not coded as observation stays) and inpatient stays. Analysis of coding patterns among providers of observation stays and long outpatient stays lasting over 1 midnight found wide variability among 3,330 hospitals in reporting patterns (Figure 1). Some hospitals coded almost all outpatient stays and very few observation stays; some coded almost all observation stays and few outpatient stays. Similar variation was found in coding of observation stays and long outpatient stays, compared with short inpatient admissions, among 3,330 hospitals (Figure 2) (OIG 2013). Figure 1. Variation in the Use of Observation and Long Outpatient Stays Among Hospitals, 2012 Source: OIG analysis of CMS data, 2013. 2017 National Committee for Quality Assurance 31
Figure 2. Variation in the Use of Short Inpatient Stays and the Use of Observation and Long Outpatient Stays Among Hospitals, 2012. Source: OIG analysis of CMS data, 2013. Other studies evaluating the association between providers and the prevalence of observation stays found that 18% of providers delivered no observational care in 2009 Medicare FFS claims (Wright et al. 2014). Table 5 describes hospital and patient characteristics associated with the likelihood of providing any observational care such as critical access hospital status, provider bed count and average age of patients served. Critical access hospitals were 97% less likely to use observation services than noncritical access hospitals. Hospitals with bed counts below 100 were around 50% less likely to use observation services than hospitals with bed counts above 100. There is an 18% increase in the likelihood of using observation services as the mean age of patients served increases by one year (Wright et al. 2014). Table 5. Hospital and Patient Characteristics Associated with the Likelihood of Providing Any Observation Care Factors Associated With Providers Using Any Observation Stays Bed Count Above 200 Outpatient Surgery Unit Micropolitan Area 1 * Average Age* Factors Associated With Providers Using No Observation Stays Critical Access Hospital* Bed Count Below 100* Government, Nonfederal Ownership* For-Profit Ownership 1 Micropolitan statistical areas include populations between 10,000 and 50,000 *Statistically significant at 5%. Source: Wright et al., 2014. Stakeholder Interviews 2017 National Committee for Quality Assurance 32